5 * This work is licensed under the terms of the GNU GPL, version 2. See
6 * the COPYING file in the top-level directory.
9 #include <linux/types.h>
10 #include <linux/hardirq.h>
11 #include <linux/list.h>
12 #include <linux/mutex.h>
13 #include <linux/spinlock.h>
14 #include <linux/signal.h>
15 #include <linux/sched.h>
16 #include <linux/bug.h>
18 #include <linux/mmu_notifier.h>
19 #include <linux/preempt.h>
20 #include <linux/msi.h>
21 #include <linux/slab.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/swait.h>
29 #include <linux/refcount.h>
30 #include <linux/nospec.h>
31 #include <asm/signal.h>
33 #include <linux/kvm.h>
34 #include <linux/kvm_para.h>
36 #include <linux/kvm_types.h>
38 #include <asm/kvm_host.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)
51 /* Two fragments for cross MMIO pages. */
52 #define KVM_MAX_MMIO_FRAGMENTS 2
54 #ifndef KVM_ADDRESS_SPACE_NUM
55 #define KVM_ADDRESS_SPACE_NUM 1
59 * For the normal pfn, the highest 12 bits should be zero,
60 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
61 * mask bit 63 to indicate the noslot pfn.
63 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
64 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
65 #define KVM_PFN_NOSLOT (0x1ULL << 63)
67 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
68 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
69 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
72 * error pfns indicate that the gfn is in slot but faild to
73 * translate it to pfn on host.
75 static inline bool is_error_pfn(kvm_pfn_t pfn)
77 return !!(pfn & KVM_PFN_ERR_MASK);
81 * error_noslot pfns indicate that the gfn can not be
82 * translated to pfn - it is not in slot or failed to
83 * translate it to pfn.
85 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn)
87 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
90 /* noslot pfn indicates that the gfn is not in slot. */
91 static inline bool is_noslot_pfn(kvm_pfn_t pfn)
93 return pfn == KVM_PFN_NOSLOT;
97 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
98 * provide own defines and kvm_is_error_hva
100 #ifndef KVM_HVA_ERR_BAD
102 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
103 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
105 static inline bool kvm_is_error_hva(unsigned long addr)
107 return addr >= PAGE_OFFSET;
112 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
114 static inline bool is_error_page(struct page *page)
119 #define KVM_REQUEST_MASK GENMASK(7,0)
120 #define KVM_REQUEST_NO_WAKEUP BIT(8)
121 #define KVM_REQUEST_WAIT BIT(9)
123 * Architecture-independent vcpu->requests bit members
124 * Bits 4-7 are reserved for more arch-independent bits.
126 #define KVM_REQ_TLB_FLUSH (0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
127 #define KVM_REQ_MMU_RELOAD (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
128 #define KVM_REQ_PENDING_TIMER 2
129 #define KVM_REQ_UNHALT 3
130 #define KVM_REQUEST_ARCH_BASE 8
132 #define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
133 BUILD_BUG_ON((unsigned)(nr) >= 32 - KVM_REQUEST_ARCH_BASE); \
134 (unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) | (flags)); \
136 #define KVM_ARCH_REQ(nr) KVM_ARCH_REQ_FLAGS(nr, 0)
138 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
139 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
141 extern struct kmem_cache *kvm_vcpu_cache;
143 extern struct mutex kvm_lock;
144 extern struct list_head vm_list;
146 struct kvm_io_range {
149 struct kvm_io_device *dev;
152 #define NR_IOBUS_DEVS 1000
157 struct kvm_io_range range[];
163 KVM_VIRTIO_CCW_NOTIFY_BUS,
168 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
169 int len, const void *val);
170 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
171 gpa_t addr, int len, const void *val, long cookie);
172 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
174 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
175 int len, struct kvm_io_device *dev);
176 void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
177 struct kvm_io_device *dev);
178 struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx,
181 #ifdef CONFIG_KVM_ASYNC_PF
182 struct kvm_async_pf {
183 struct work_struct work;
184 struct list_head link;
185 struct list_head queue;
186 struct kvm_vcpu *vcpu;
187 struct mm_struct *mm;
190 struct kvm_arch_async_pf arch;
194 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
195 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
196 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
197 struct kvm_arch_async_pf *arch);
198 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
205 READING_SHADOW_PAGE_TABLES,
209 * Sometimes a large or cross-page mmio needs to be broken up into separate
210 * exits for userspace servicing.
212 struct kvm_mmio_fragment {
220 #ifdef CONFIG_PREEMPT_NOTIFIERS
221 struct preempt_notifier preempt_notifier;
227 unsigned long requests;
228 unsigned long guest_debug;
231 struct list_head blocked_vcpu_list;
236 int guest_xcr0_loaded;
237 struct swait_queue_head wq;
238 struct pid __rcu *pid;
241 struct kvm_vcpu_stat stat;
242 unsigned int halt_poll_ns;
245 #ifdef CONFIG_HAS_IOMEM
247 int mmio_read_completed;
249 int mmio_cur_fragment;
250 int mmio_nr_fragments;
251 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
254 #ifdef CONFIG_KVM_ASYNC_PF
257 struct list_head queue;
258 struct list_head done;
263 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
265 * Cpu relax intercept or pause loop exit optimization
266 * in_spin_loop: set when a vcpu does a pause loop exit
267 * or cpu relax intercepted.
268 * dy_eligible: indicates whether vcpu is eligible for directed yield.
276 struct kvm_vcpu_arch arch;
277 struct dentry *debugfs_dentry;
280 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
283 * The memory barrier ensures a previous write to vcpu->requests cannot
284 * be reordered with the read of vcpu->mode. It pairs with the general
285 * memory barrier following the write of vcpu->mode in VCPU RUN.
287 smp_mb__before_atomic();
288 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
292 * Some of the bitops functions do not support too long bitmaps.
293 * This number must be determined not to exceed such limits.
295 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
297 struct kvm_memory_slot {
299 unsigned long npages;
300 unsigned long *dirty_bitmap;
301 struct kvm_arch_memory_slot arch;
302 unsigned long userspace_addr;
307 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
309 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
312 struct kvm_s390_adapter_int {
325 struct kvm_kernel_irq_routing_entry {
328 int (*set)(struct kvm_kernel_irq_routing_entry *e,
329 struct kvm *kvm, int irq_source_id, int level,
343 struct kvm_s390_adapter_int adapter;
344 struct kvm_hv_sint hv_sint;
346 struct hlist_node link;
349 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
350 struct kvm_irq_routing_table {
351 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
354 * Array indexed by gsi. Each entry contains list of irq chips
355 * the gsi is connected to.
357 struct hlist_head map[0];
361 #ifndef KVM_PRIVATE_MEM_SLOTS
362 #define KVM_PRIVATE_MEM_SLOTS 0
365 #ifndef KVM_MEM_SLOTS_NUM
366 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
369 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
370 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
378 * memslots are not sorted by id anymore, please use id_to_memslot()
379 * to get the memslot by its id.
381 struct kvm_memslots {
383 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
384 /* The mapping table from slot id to the index in memslots[]. */
385 short id_to_index[KVM_MEM_SLOTS_NUM];
392 struct mutex slots_lock;
393 struct mm_struct *mm; /* userspace tied to this vm */
394 struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
395 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
398 * created_vcpus is protected by kvm->lock, and is incremented
399 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
400 * incremented after storing the kvm_vcpu pointer in vcpus,
401 * and is accessed atomically.
403 atomic_t online_vcpus;
405 int last_boosted_vcpu;
406 struct list_head vm_list;
408 struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
409 #ifdef CONFIG_HAVE_KVM_EVENTFD
412 struct list_head items;
413 struct list_head resampler_list;
414 struct mutex resampler_lock;
416 struct list_head ioeventfds;
418 struct kvm_vm_stat stat;
419 struct kvm_arch arch;
420 refcount_t users_count;
421 #ifdef CONFIG_KVM_MMIO
422 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
423 spinlock_t ring_lock;
424 struct list_head coalesced_zones;
427 struct mutex irq_lock;
428 #ifdef CONFIG_HAVE_KVM_IRQCHIP
430 * Update side is protected by irq_lock.
432 struct kvm_irq_routing_table __rcu *irq_routing;
434 #ifdef CONFIG_HAVE_KVM_IRQFD
435 struct hlist_head irq_ack_notifier_list;
438 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
439 struct mmu_notifier mmu_notifier;
440 unsigned long mmu_notifier_seq;
441 long mmu_notifier_count;
444 struct list_head devices;
445 struct dentry *debugfs_dentry;
446 struct kvm_stat_data **debugfs_stat_data;
447 struct srcu_struct srcu;
448 struct srcu_struct irq_srcu;
452 #define kvm_err(fmt, ...) \
453 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
454 #define kvm_info(fmt, ...) \
455 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
456 #define kvm_debug(fmt, ...) \
457 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
458 #define kvm_debug_ratelimited(fmt, ...) \
459 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
461 #define kvm_pr_unimpl(fmt, ...) \
462 pr_err_ratelimited("kvm [%i]: " fmt, \
463 task_tgid_nr(current), ## __VA_ARGS__)
465 /* The guest did something we don't support. */
466 #define vcpu_unimpl(vcpu, fmt, ...) \
467 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
468 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
470 #define vcpu_debug(vcpu, fmt, ...) \
471 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
472 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
473 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
475 #define vcpu_err(vcpu, fmt, ...) \
476 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
478 static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx)
480 return srcu_dereference_check(kvm->buses[idx], &kvm->srcu,
481 lockdep_is_held(&kvm->slots_lock) ||
482 !refcount_read(&kvm->users_count));
485 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
487 int num_vcpus = atomic_read(&kvm->online_vcpus);
488 i = array_index_nospec(i, num_vcpus);
490 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. */
492 return kvm->vcpus[i];
495 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
497 idx < atomic_read(&kvm->online_vcpus) && \
498 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
501 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
503 struct kvm_vcpu *vcpu = NULL;
508 if (id < KVM_MAX_VCPUS)
509 vcpu = kvm_get_vcpu(kvm, id);
510 if (vcpu && vcpu->vcpu_id == id)
512 kvm_for_each_vcpu(i, vcpu, kvm)
513 if (vcpu->vcpu_id == id)
518 static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu)
520 struct kvm_vcpu *tmp;
523 kvm_for_each_vcpu(idx, tmp, vcpu->kvm)
529 #define kvm_for_each_memslot(memslot, slots) \
530 for (memslot = &slots->memslots[0]; \
531 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
534 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
535 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
537 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
538 void vcpu_put(struct kvm_vcpu *vcpu);
540 #ifdef __KVM_HAVE_IOAPIC
541 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
542 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
544 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
547 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
552 #ifdef CONFIG_HAVE_KVM_IRQFD
553 int kvm_irqfd_init(void);
554 void kvm_irqfd_exit(void);
556 static inline int kvm_irqfd_init(void)
561 static inline void kvm_irqfd_exit(void)
565 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
566 struct module *module);
569 void kvm_get_kvm(struct kvm *kvm);
570 void kvm_put_kvm(struct kvm *kvm);
572 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
574 as_id = array_index_nospec(as_id, KVM_ADDRESS_SPACE_NUM);
575 return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu,
576 lockdep_is_held(&kvm->slots_lock) ||
577 !refcount_read(&kvm->users_count));
580 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
582 return __kvm_memslots(kvm, 0);
585 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
587 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
589 return __kvm_memslots(vcpu->kvm, as_id);
592 static inline struct kvm_memory_slot *
593 id_to_memslot(struct kvm_memslots *slots, int id)
595 int index = slots->id_to_index[id];
596 struct kvm_memory_slot *slot;
598 slot = &slots->memslots[index];
600 WARN_ON(slot->id != id);
605 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
606 * - create a new memory slot
607 * - delete an existing memory slot
608 * - modify an existing memory slot
609 * -- move it in the guest physical memory space
610 * -- just change its flags
612 * Since flags can be changed by some of these operations, the following
613 * differentiation is the best we can do for __kvm_set_memory_region():
622 int kvm_set_memory_region(struct kvm *kvm,
623 const struct kvm_userspace_memory_region *mem);
624 int __kvm_set_memory_region(struct kvm *kvm,
625 const struct kvm_userspace_memory_region *mem);
626 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
627 struct kvm_memory_slot *dont);
628 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
629 unsigned long npages);
630 void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen);
631 int kvm_arch_prepare_memory_region(struct kvm *kvm,
632 struct kvm_memory_slot *memslot,
633 const struct kvm_userspace_memory_region *mem,
634 enum kvm_mr_change change);
635 void kvm_arch_commit_memory_region(struct kvm *kvm,
636 const struct kvm_userspace_memory_region *mem,
637 const struct kvm_memory_slot *old,
638 const struct kvm_memory_slot *new,
639 enum kvm_mr_change change);
640 bool kvm_largepages_enabled(void);
641 void kvm_disable_largepages(void);
642 /* flush all memory translations */
643 void kvm_arch_flush_shadow_all(struct kvm *kvm);
644 /* flush memory translations pointing to 'slot' */
645 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
646 struct kvm_memory_slot *slot);
648 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
649 struct page **pages, int nr_pages);
651 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
652 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
653 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
654 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
655 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
657 void kvm_release_page_clean(struct page *page);
658 void kvm_release_page_dirty(struct page *page);
659 void kvm_set_page_accessed(struct page *page);
661 kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
662 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
663 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
665 kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
666 kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
667 kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
668 bool atomic, bool *async, bool write_fault,
671 void kvm_release_pfn_clean(kvm_pfn_t pfn);
672 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
673 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
674 void kvm_get_pfn(kvm_pfn_t pfn);
676 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
678 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
680 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
681 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
682 void *data, unsigned long len);
683 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
684 int offset, int len);
685 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
687 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
688 void *data, unsigned long len);
689 int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
690 void *data, unsigned int offset,
692 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
693 gpa_t gpa, unsigned long len);
694 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
695 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
696 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
697 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
698 unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn);
699 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
701 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
702 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
703 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
704 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
705 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
706 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
707 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
708 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
710 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
712 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
714 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
715 int offset, int len);
716 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
718 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
720 void kvm_sigset_activate(struct kvm_vcpu *vcpu);
721 void kvm_sigset_deactivate(struct kvm_vcpu *vcpu);
723 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
724 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
725 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
726 bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
727 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
728 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
729 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool usermode_vcpu_not_eligible);
730 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
731 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
733 void kvm_flush_remote_tlbs(struct kvm *kvm);
734 void kvm_reload_remote_mmus(struct kvm *kvm);
735 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
737 long kvm_arch_dev_ioctl(struct file *filp,
738 unsigned int ioctl, unsigned long arg);
739 long kvm_arch_vcpu_ioctl(struct file *filp,
740 unsigned int ioctl, unsigned long arg);
741 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
743 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
745 int kvm_get_dirty_log(struct kvm *kvm,
746 struct kvm_dirty_log *log, int *is_dirty);
748 int kvm_get_dirty_log_protect(struct kvm *kvm,
749 struct kvm_dirty_log *log, bool *is_dirty);
751 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
752 struct kvm_memory_slot *slot,
756 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
757 struct kvm_dirty_log *log);
759 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
761 long kvm_arch_vm_ioctl(struct file *filp,
762 unsigned int ioctl, unsigned long arg);
764 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
765 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
767 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
768 struct kvm_translation *tr);
770 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
771 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
772 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
773 struct kvm_sregs *sregs);
774 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
775 struct kvm_sregs *sregs);
776 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
777 struct kvm_mp_state *mp_state);
778 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
779 struct kvm_mp_state *mp_state);
780 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
781 struct kvm_guest_debug *dbg);
782 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
784 int kvm_arch_init(void *opaque);
785 void kvm_arch_exit(void);
787 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
788 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
790 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
792 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
793 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
794 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
795 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
796 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
797 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
798 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
800 bool kvm_arch_has_vcpu_debugfs(void);
801 int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu);
803 int kvm_arch_hardware_enable(void);
804 void kvm_arch_hardware_disable(void);
805 int kvm_arch_hardware_setup(void);
806 void kvm_arch_hardware_unsetup(void);
807 void kvm_arch_check_processor_compat(void *rtn);
808 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
809 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
810 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
811 bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu);
813 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
814 static inline struct kvm *kvm_arch_alloc_vm(void)
816 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
819 static inline void kvm_arch_free_vm(struct kvm *kvm)
825 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
826 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
827 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
828 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
830 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
834 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
838 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
843 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
844 void kvm_arch_start_assignment(struct kvm *kvm);
845 void kvm_arch_end_assignment(struct kvm *kvm);
846 bool kvm_arch_has_assigned_device(struct kvm *kvm);
848 static inline void kvm_arch_start_assignment(struct kvm *kvm)
852 static inline void kvm_arch_end_assignment(struct kvm *kvm)
856 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
862 static inline struct swait_queue_head *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
864 #ifdef __KVM_HAVE_ARCH_WQP
865 return vcpu->arch.wqp;
871 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
873 * returns true if the virtual interrupt controller is initialized and
874 * ready to accept virtual IRQ. On some architectures the virtual interrupt
875 * controller is dynamically instantiated and this is not always true.
877 bool kvm_arch_intc_initialized(struct kvm *kvm);
879 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
885 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
886 void kvm_arch_destroy_vm(struct kvm *kvm);
887 void kvm_arch_sync_events(struct kvm *kvm);
889 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
890 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
892 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
893 bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
895 struct kvm_irq_ack_notifier {
896 struct hlist_node link;
898 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
901 int kvm_irq_map_gsi(struct kvm *kvm,
902 struct kvm_kernel_irq_routing_entry *entries, int gsi);
903 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
905 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
907 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
908 int irq_source_id, int level, bool line_status);
909 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
910 struct kvm *kvm, int irq_source_id,
911 int level, bool line_status);
912 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
913 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
914 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
915 void kvm_register_irq_ack_notifier(struct kvm *kvm,
916 struct kvm_irq_ack_notifier *kian);
917 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
918 struct kvm_irq_ack_notifier *kian);
919 int kvm_request_irq_source_id(struct kvm *kvm);
920 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
923 * search_memslots() and __gfn_to_memslot() are here because they are
924 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
925 * gfn_to_memslot() itself isn't here as an inline because that would
926 * bloat other code too much.
928 static inline struct kvm_memory_slot *
929 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
931 int start = 0, end = slots->used_slots;
932 int slot = atomic_read(&slots->lru_slot);
933 struct kvm_memory_slot *memslots = slots->memslots;
935 if (gfn >= memslots[slot].base_gfn &&
936 gfn < memslots[slot].base_gfn + memslots[slot].npages)
937 return &memslots[slot];
939 while (start < end) {
940 slot = start + (end - start) / 2;
942 if (gfn >= memslots[slot].base_gfn)
948 if (start < slots->used_slots && gfn >= memslots[start].base_gfn &&
949 gfn < memslots[start].base_gfn + memslots[start].npages) {
950 atomic_set(&slots->lru_slot, start);
951 return &memslots[start];
957 static inline struct kvm_memory_slot *
958 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
960 return search_memslots(slots, gfn);
963 static inline unsigned long
964 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
967 * The index was checked originally in search_memslots. To avoid
968 * that a malicious guest builds a Spectre gadget out of e.g. page
969 * table walks, do not let the processor speculate loads outside
970 * the guest's registered memslots.
972 unsigned long offset = gfn - slot->base_gfn;
973 offset = array_index_nospec(offset, slot->npages);
974 return slot->userspace_addr + offset * PAGE_SIZE;
977 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
979 return gfn_to_memslot(kvm, gfn)->id;
983 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
985 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
987 return slot->base_gfn + gfn_offset;
990 static inline gpa_t gfn_to_gpa(gfn_t gfn)
992 return (gpa_t)gfn << PAGE_SHIFT;
995 static inline gfn_t gpa_to_gfn(gpa_t gpa)
997 return (gfn_t)(gpa >> PAGE_SHIFT);
1000 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
1002 return (hpa_t)pfn << PAGE_SHIFT;
1005 static inline struct page *kvm_vcpu_gpa_to_page(struct kvm_vcpu *vcpu,
1008 return kvm_vcpu_gfn_to_page(vcpu, gpa_to_gfn(gpa));
1011 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
1013 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
1015 return kvm_is_error_hva(hva);
1018 enum kvm_stat_kind {
1023 struct kvm_stat_data {
1029 struct kvm_stats_debugfs_item {
1032 enum kvm_stat_kind kind;
1035 extern struct kvm_stats_debugfs_item debugfs_entries[];
1036 extern struct dentry *kvm_debugfs_dir;
1038 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1039 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1041 if (unlikely(kvm->mmu_notifier_count))
1044 * Ensure the read of mmu_notifier_count happens before the read
1045 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1046 * mmu_notifier_invalidate_range_end to make sure that the caller
1047 * either sees the old (non-zero) value of mmu_notifier_count or
1048 * the new (incremented) value of mmu_notifier_seq.
1049 * PowerPC Book3s HV KVM calls this under a per-page lock
1050 * rather than under kvm->mmu_lock, for scalability, so
1051 * can't rely on kvm->mmu_lock to keep things ordered.
1054 if (kvm->mmu_notifier_seq != mmu_seq)
1060 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1062 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1064 bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
1065 int kvm_set_irq_routing(struct kvm *kvm,
1066 const struct kvm_irq_routing_entry *entries,
1069 int kvm_set_routing_entry(struct kvm *kvm,
1070 struct kvm_kernel_irq_routing_entry *e,
1071 const struct kvm_irq_routing_entry *ue);
1072 void kvm_free_irq_routing(struct kvm *kvm);
1076 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1080 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1082 #ifdef CONFIG_HAVE_KVM_EVENTFD
1084 void kvm_eventfd_init(struct kvm *kvm);
1085 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1087 #ifdef CONFIG_HAVE_KVM_IRQFD
1088 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1089 void kvm_irqfd_release(struct kvm *kvm);
1090 void kvm_irq_routing_update(struct kvm *);
1092 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1097 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1102 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1104 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1109 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1111 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1112 static inline void kvm_irq_routing_update(struct kvm *kvm)
1117 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1122 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1124 void kvm_arch_irq_routing_update(struct kvm *kvm);
1126 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1129 * Ensure the rest of the request is published to kvm_check_request's
1130 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1133 set_bit(req & KVM_REQUEST_MASK, &vcpu->requests);
1136 static inline bool kvm_request_pending(struct kvm_vcpu *vcpu)
1138 return READ_ONCE(vcpu->requests);
1141 static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)
1143 return test_bit(req & KVM_REQUEST_MASK, &vcpu->requests);
1146 static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu)
1148 clear_bit(req & KVM_REQUEST_MASK, &vcpu->requests);
1151 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1153 if (kvm_test_request(req, vcpu)) {
1154 kvm_clear_request(req, vcpu);
1157 * Ensure the rest of the request is visible to kvm_check_request's
1158 * caller. Paired with the smp_wmb in kvm_make_request.
1160 smp_mb__after_atomic();
1167 extern bool kvm_rebooting;
1169 extern unsigned int halt_poll_ns;
1170 extern unsigned int halt_poll_ns_grow;
1171 extern unsigned int halt_poll_ns_shrink;
1174 struct kvm_device_ops *ops;
1177 struct list_head vm_node;
1180 /* create, destroy, and name are mandatory */
1181 struct kvm_device_ops {
1185 * create is called holding kvm->lock and any operations not suitable
1186 * to do while holding the lock should be deferred to init (see
1189 int (*create)(struct kvm_device *dev, u32 type);
1192 * init is called after create if create is successful and is called
1193 * outside of holding kvm->lock.
1195 void (*init)(struct kvm_device *dev);
1198 * Destroy is responsible for freeing dev.
1200 * Destroy may be called before or after destructors are called
1201 * on emulated I/O regions, depending on whether a reference is
1202 * held by a vcpu or other kvm component that gets destroyed
1203 * after the emulated I/O.
1205 void (*destroy)(struct kvm_device *dev);
1207 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1208 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1209 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1210 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1214 void kvm_device_get(struct kvm_device *dev);
1215 void kvm_device_put(struct kvm_device *dev);
1216 struct kvm_device *kvm_device_from_filp(struct file *filp);
1217 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1218 void kvm_unregister_device_ops(u32 type);
1220 extern struct kvm_device_ops kvm_mpic_ops;
1221 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1222 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1224 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1226 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1228 vcpu->spin_loop.in_spin_loop = val;
1230 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1232 vcpu->spin_loop.dy_eligible = val;
1235 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1237 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1241 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1244 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1246 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1247 bool kvm_arch_has_irq_bypass(void);
1248 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1249 struct irq_bypass_producer *);
1250 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1251 struct irq_bypass_producer *);
1252 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1253 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1254 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1255 uint32_t guest_irq, bool set);
1256 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1258 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1259 /* If we wakeup during the poll time, was it a sucessful poll? */
1260 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1262 return vcpu->valid_wakeup;
1266 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1270 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
1272 typedef int (*kvm_vm_thread_fn_t)(struct kvm *kvm, uintptr_t data);
1274 int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn,
1275 uintptr_t data, const char *name,
1276 struct task_struct **thread_ptr);