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_REQ_VM_BUGGED (4 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
131 #define KVM_REQUEST_ARCH_BASE 8
133 #define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
134 BUILD_BUG_ON((unsigned)(nr) >= 32 - KVM_REQUEST_ARCH_BASE); \
135 (unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) | (flags)); \
137 #define KVM_ARCH_REQ(nr) KVM_ARCH_REQ_FLAGS(nr, 0)
139 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
140 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
142 extern struct kmem_cache *kvm_vcpu_cache;
144 extern struct mutex kvm_lock;
145 extern struct list_head vm_list;
147 struct kvm_io_range {
150 struct kvm_io_device *dev;
153 #define NR_IOBUS_DEVS 1000
158 struct kvm_io_range range[];
164 KVM_VIRTIO_CCW_NOTIFY_BUS,
169 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
170 int len, const void *val);
171 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
172 gpa_t addr, int len, const void *val, long cookie);
173 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
175 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
176 int len, struct kvm_io_device *dev);
177 void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
178 struct kvm_io_device *dev);
179 struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx,
182 #ifdef CONFIG_KVM_ASYNC_PF
183 struct kvm_async_pf {
184 struct work_struct work;
185 struct list_head link;
186 struct list_head queue;
187 struct kvm_vcpu *vcpu;
188 struct mm_struct *mm;
191 struct kvm_arch_async_pf arch;
195 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
196 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
197 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
198 struct kvm_arch_async_pf *arch);
199 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
206 READING_SHADOW_PAGE_TABLES,
210 * Sometimes a large or cross-page mmio needs to be broken up into separate
211 * exits for userspace servicing.
213 struct kvm_mmio_fragment {
221 #ifdef CONFIG_PREEMPT_NOTIFIERS
222 struct preempt_notifier preempt_notifier;
228 unsigned long requests;
229 unsigned long guest_debug;
232 struct list_head blocked_vcpu_list;
237 int guest_xcr0_loaded;
238 struct swait_queue_head wq;
239 struct pid __rcu *pid;
242 struct kvm_vcpu_stat stat;
243 unsigned int halt_poll_ns;
246 #ifdef CONFIG_HAS_IOMEM
248 int mmio_read_completed;
250 int mmio_cur_fragment;
251 int mmio_nr_fragments;
252 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
255 #ifdef CONFIG_KVM_ASYNC_PF
258 struct list_head queue;
259 struct list_head done;
264 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
266 * Cpu relax intercept or pause loop exit optimization
267 * in_spin_loop: set when a vcpu does a pause loop exit
268 * or cpu relax intercepted.
269 * dy_eligible: indicates whether vcpu is eligible for directed yield.
277 struct kvm_vcpu_arch arch;
278 struct dentry *debugfs_dentry;
281 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
284 * The memory barrier ensures a previous write to vcpu->requests cannot
285 * be reordered with the read of vcpu->mode. It pairs with the general
286 * memory barrier following the write of vcpu->mode in VCPU RUN.
288 smp_mb__before_atomic();
289 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
293 * Some of the bitops functions do not support too long bitmaps.
294 * This number must be determined not to exceed such limits.
296 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
298 struct kvm_memory_slot {
300 unsigned long npages;
301 unsigned long *dirty_bitmap;
302 struct kvm_arch_memory_slot arch;
303 unsigned long userspace_addr;
308 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
310 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
313 struct kvm_s390_adapter_int {
326 struct kvm_kernel_irq_routing_entry {
329 int (*set)(struct kvm_kernel_irq_routing_entry *e,
330 struct kvm *kvm, int irq_source_id, int level,
344 struct kvm_s390_adapter_int adapter;
345 struct kvm_hv_sint hv_sint;
347 struct hlist_node link;
350 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
351 struct kvm_irq_routing_table {
352 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
355 * Array indexed by gsi. Each entry contains list of irq chips
356 * the gsi is connected to.
358 struct hlist_head map[0];
362 #ifndef KVM_PRIVATE_MEM_SLOTS
363 #define KVM_PRIVATE_MEM_SLOTS 0
366 #ifndef KVM_MEM_SLOTS_NUM
367 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
370 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
371 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
379 * memslots are not sorted by id anymore, please use id_to_memslot()
380 * to get the memslot by its id.
382 struct kvm_memslots {
384 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
385 /* The mapping table from slot id to the index in memslots[]. */
386 short id_to_index[KVM_MEM_SLOTS_NUM];
393 struct mutex slots_lock;
394 struct mm_struct *mm; /* userspace tied to this vm */
395 struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
396 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
399 * created_vcpus is protected by kvm->lock, and is incremented
400 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
401 * incremented after storing the kvm_vcpu pointer in vcpus,
402 * and is accessed atomically.
404 atomic_t online_vcpus;
406 int last_boosted_vcpu;
407 struct list_head vm_list;
409 struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
410 #ifdef CONFIG_HAVE_KVM_EVENTFD
413 struct list_head items;
414 struct list_head resampler_list;
415 struct mutex resampler_lock;
417 struct list_head ioeventfds;
419 struct kvm_vm_stat stat;
420 struct kvm_arch arch;
421 refcount_t users_count;
422 #ifdef CONFIG_KVM_MMIO
423 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
424 spinlock_t ring_lock;
425 struct list_head coalesced_zones;
428 struct mutex irq_lock;
429 #ifdef CONFIG_HAVE_KVM_IRQCHIP
431 * Update side is protected by irq_lock.
433 struct kvm_irq_routing_table __rcu *irq_routing;
435 #ifdef CONFIG_HAVE_KVM_IRQFD
436 struct hlist_head irq_ack_notifier_list;
439 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
440 struct mmu_notifier mmu_notifier;
441 unsigned long mmu_notifier_seq;
442 long mmu_notifier_count;
445 struct list_head devices;
446 struct dentry *debugfs_dentry;
447 struct kvm_stat_data **debugfs_stat_data;
448 struct srcu_struct srcu;
449 struct srcu_struct irq_srcu;
454 #define kvm_err(fmt, ...) \
455 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
456 #define kvm_info(fmt, ...) \
457 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
458 #define kvm_debug(fmt, ...) \
459 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
460 #define kvm_debug_ratelimited(fmt, ...) \
461 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
463 #define kvm_pr_unimpl(fmt, ...) \
464 pr_err_ratelimited("kvm [%i]: " fmt, \
465 task_tgid_nr(current), ## __VA_ARGS__)
467 /* The guest did something we don't support. */
468 #define vcpu_unimpl(vcpu, fmt, ...) \
469 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
470 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
472 #define vcpu_debug(vcpu, fmt, ...) \
473 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
474 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
475 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
477 #define vcpu_err(vcpu, fmt, ...) \
478 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
480 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
481 static inline void kvm_vm_bugged(struct kvm *kvm)
483 kvm->vm_bugged = true;
484 kvm_make_all_cpus_request(kvm, KVM_REQ_VM_BUGGED);
487 #define KVM_BUG(cond, kvm, fmt...) \
489 int __ret = (cond); \
491 if (WARN_ONCE(__ret && !(kvm)->vm_bugged, fmt)) \
492 kvm_vm_bugged(kvm); \
496 #define KVM_BUG_ON(cond, kvm) \
498 int __ret = (cond); \
500 if (WARN_ON_ONCE(__ret && !(kvm)->vm_bugged)) \
501 kvm_vm_bugged(kvm); \
505 static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx)
507 return srcu_dereference_check(kvm->buses[idx], &kvm->srcu,
508 lockdep_is_held(&kvm->slots_lock) ||
509 !refcount_read(&kvm->users_count));
512 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
514 int num_vcpus = atomic_read(&kvm->online_vcpus);
515 i = array_index_nospec(i, num_vcpus);
517 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. */
519 return kvm->vcpus[i];
522 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
524 idx < atomic_read(&kvm->online_vcpus) && \
525 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
528 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
530 struct kvm_vcpu *vcpu = NULL;
535 if (id < KVM_MAX_VCPUS)
536 vcpu = kvm_get_vcpu(kvm, id);
537 if (vcpu && vcpu->vcpu_id == id)
539 kvm_for_each_vcpu(i, vcpu, kvm)
540 if (vcpu->vcpu_id == id)
545 static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu)
547 struct kvm_vcpu *tmp;
550 kvm_for_each_vcpu(idx, tmp, vcpu->kvm)
556 #define kvm_for_each_memslot(memslot, slots) \
557 for (memslot = &slots->memslots[0]; \
558 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
561 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
562 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
564 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
565 void vcpu_put(struct kvm_vcpu *vcpu);
567 #ifdef __KVM_HAVE_IOAPIC
568 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
569 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
571 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
574 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
579 #ifdef CONFIG_HAVE_KVM_IRQFD
580 int kvm_irqfd_init(void);
581 void kvm_irqfd_exit(void);
583 static inline int kvm_irqfd_init(void)
588 static inline void kvm_irqfd_exit(void)
592 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
593 struct module *module);
596 void kvm_get_kvm(struct kvm *kvm);
597 void kvm_put_kvm(struct kvm *kvm);
599 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
601 as_id = array_index_nospec(as_id, KVM_ADDRESS_SPACE_NUM);
602 return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu,
603 lockdep_is_held(&kvm->slots_lock) ||
604 !refcount_read(&kvm->users_count));
607 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
609 return __kvm_memslots(kvm, 0);
612 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
614 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
616 return __kvm_memslots(vcpu->kvm, as_id);
619 static inline struct kvm_memory_slot *
620 id_to_memslot(struct kvm_memslots *slots, int id)
622 int index = slots->id_to_index[id];
623 struct kvm_memory_slot *slot;
625 slot = &slots->memslots[index];
627 WARN_ON(slot->id != id);
632 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
633 * - create a new memory slot
634 * - delete an existing memory slot
635 * - modify an existing memory slot
636 * -- move it in the guest physical memory space
637 * -- just change its flags
639 * Since flags can be changed by some of these operations, the following
640 * differentiation is the best we can do for __kvm_set_memory_region():
649 int kvm_set_memory_region(struct kvm *kvm,
650 const struct kvm_userspace_memory_region *mem);
651 int __kvm_set_memory_region(struct kvm *kvm,
652 const struct kvm_userspace_memory_region *mem);
653 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
654 struct kvm_memory_slot *dont);
655 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
656 unsigned long npages);
657 void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen);
658 int kvm_arch_prepare_memory_region(struct kvm *kvm,
659 struct kvm_memory_slot *memslot,
660 const struct kvm_userspace_memory_region *mem,
661 enum kvm_mr_change change);
662 void kvm_arch_commit_memory_region(struct kvm *kvm,
663 const struct kvm_userspace_memory_region *mem,
664 const struct kvm_memory_slot *old,
665 const struct kvm_memory_slot *new,
666 enum kvm_mr_change change);
667 bool kvm_largepages_enabled(void);
668 void kvm_disable_largepages(void);
669 /* flush all memory translations */
670 void kvm_arch_flush_shadow_all(struct kvm *kvm);
671 /* flush memory translations pointing to 'slot' */
672 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
673 struct kvm_memory_slot *slot);
675 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
676 struct page **pages, int nr_pages);
678 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
679 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
680 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
681 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
682 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
684 void kvm_release_page_clean(struct page *page);
685 void kvm_release_page_dirty(struct page *page);
686 void kvm_set_page_accessed(struct page *page);
688 kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
689 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
690 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
692 kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
693 kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
694 kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
695 bool atomic, bool *async, bool write_fault,
698 void kvm_release_pfn_clean(kvm_pfn_t pfn);
699 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
700 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
701 void kvm_get_pfn(kvm_pfn_t pfn);
703 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
705 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
707 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
708 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
709 void *data, unsigned long len);
710 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
711 int offset, int len);
712 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
714 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
715 void *data, unsigned long len);
716 int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
717 void *data, unsigned int offset,
719 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
720 gpa_t gpa, unsigned long len);
721 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
722 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
723 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
724 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
725 unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn);
726 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
728 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
729 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
730 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
731 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
732 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
733 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
734 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
735 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
737 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
739 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
741 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
742 int offset, int len);
743 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
745 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
747 void kvm_sigset_activate(struct kvm_vcpu *vcpu);
748 void kvm_sigset_deactivate(struct kvm_vcpu *vcpu);
750 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
751 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
752 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
753 bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
754 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
755 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
756 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool usermode_vcpu_not_eligible);
757 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
758 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
760 void kvm_flush_remote_tlbs(struct kvm *kvm);
761 void kvm_reload_remote_mmus(struct kvm *kvm);
763 long kvm_arch_dev_ioctl(struct file *filp,
764 unsigned int ioctl, unsigned long arg);
765 long kvm_arch_vcpu_ioctl(struct file *filp,
766 unsigned int ioctl, unsigned long arg);
767 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
769 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
771 int kvm_get_dirty_log(struct kvm *kvm,
772 struct kvm_dirty_log *log, int *is_dirty);
774 int kvm_get_dirty_log_protect(struct kvm *kvm,
775 struct kvm_dirty_log *log, bool *is_dirty);
777 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
778 struct kvm_memory_slot *slot,
782 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
783 struct kvm_dirty_log *log);
785 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
787 long kvm_arch_vm_ioctl(struct file *filp,
788 unsigned int ioctl, unsigned long arg);
790 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
791 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
793 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
794 struct kvm_translation *tr);
796 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
797 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
798 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
799 struct kvm_sregs *sregs);
800 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
801 struct kvm_sregs *sregs);
802 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
803 struct kvm_mp_state *mp_state);
804 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
805 struct kvm_mp_state *mp_state);
806 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
807 struct kvm_guest_debug *dbg);
808 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
810 int kvm_arch_init(void *opaque);
811 void kvm_arch_exit(void);
813 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
814 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
816 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
818 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
819 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
820 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
821 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
822 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
823 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
824 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
826 bool kvm_arch_has_vcpu_debugfs(void);
827 int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu);
829 int kvm_arch_hardware_enable(void);
830 void kvm_arch_hardware_disable(void);
831 int kvm_arch_hardware_setup(void);
832 void kvm_arch_hardware_unsetup(void);
833 void kvm_arch_check_processor_compat(void *rtn);
834 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
835 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
836 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
837 bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu);
839 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
840 static inline struct kvm *kvm_arch_alloc_vm(void)
842 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
845 static inline void kvm_arch_free_vm(struct kvm *kvm)
851 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
852 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
853 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
854 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
856 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
860 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
864 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
869 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
870 void kvm_arch_start_assignment(struct kvm *kvm);
871 void kvm_arch_end_assignment(struct kvm *kvm);
872 bool kvm_arch_has_assigned_device(struct kvm *kvm);
874 static inline void kvm_arch_start_assignment(struct kvm *kvm)
878 static inline void kvm_arch_end_assignment(struct kvm *kvm)
882 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
888 static inline struct swait_queue_head *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
890 #ifdef __KVM_HAVE_ARCH_WQP
891 return vcpu->arch.wqp;
897 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
899 * returns true if the virtual interrupt controller is initialized and
900 * ready to accept virtual IRQ. On some architectures the virtual interrupt
901 * controller is dynamically instantiated and this is not always true.
903 bool kvm_arch_intc_initialized(struct kvm *kvm);
905 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
911 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
912 void kvm_arch_destroy_vm(struct kvm *kvm);
913 void kvm_arch_sync_events(struct kvm *kvm);
915 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
916 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
918 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
919 bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
921 struct kvm_irq_ack_notifier {
922 struct hlist_node link;
924 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
927 int kvm_irq_map_gsi(struct kvm *kvm,
928 struct kvm_kernel_irq_routing_entry *entries, int gsi);
929 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
931 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
933 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
934 int irq_source_id, int level, bool line_status);
935 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
936 struct kvm *kvm, int irq_source_id,
937 int level, bool line_status);
938 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
939 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
940 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
941 void kvm_register_irq_ack_notifier(struct kvm *kvm,
942 struct kvm_irq_ack_notifier *kian);
943 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
944 struct kvm_irq_ack_notifier *kian);
945 int kvm_request_irq_source_id(struct kvm *kvm);
946 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
949 * search_memslots() and __gfn_to_memslot() are here because they are
950 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
951 * gfn_to_memslot() itself isn't here as an inline because that would
952 * bloat other code too much.
954 static inline struct kvm_memory_slot *
955 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
957 int start = 0, end = slots->used_slots;
958 int slot = atomic_read(&slots->lru_slot);
959 struct kvm_memory_slot *memslots = slots->memslots;
961 if (gfn >= memslots[slot].base_gfn &&
962 gfn < memslots[slot].base_gfn + memslots[slot].npages)
963 return &memslots[slot];
965 while (start < end) {
966 slot = start + (end - start) / 2;
968 if (gfn >= memslots[slot].base_gfn)
974 if (start < slots->used_slots && gfn >= memslots[start].base_gfn &&
975 gfn < memslots[start].base_gfn + memslots[start].npages) {
976 atomic_set(&slots->lru_slot, start);
977 return &memslots[start];
983 static inline struct kvm_memory_slot *
984 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
986 return search_memslots(slots, gfn);
989 static inline unsigned long
990 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
993 * The index was checked originally in search_memslots. To avoid
994 * that a malicious guest builds a Spectre gadget out of e.g. page
995 * table walks, do not let the processor speculate loads outside
996 * the guest's registered memslots.
998 unsigned long offset = gfn - slot->base_gfn;
999 offset = array_index_nospec(offset, slot->npages);
1000 return slot->userspace_addr + offset * PAGE_SIZE;
1003 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
1005 return gfn_to_memslot(kvm, gfn)->id;
1009 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
1011 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
1013 return slot->base_gfn + gfn_offset;
1016 static inline gpa_t gfn_to_gpa(gfn_t gfn)
1018 return (gpa_t)gfn << PAGE_SHIFT;
1021 static inline gfn_t gpa_to_gfn(gpa_t gpa)
1023 return (gfn_t)(gpa >> PAGE_SHIFT);
1026 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
1028 return (hpa_t)pfn << PAGE_SHIFT;
1031 static inline struct page *kvm_vcpu_gpa_to_page(struct kvm_vcpu *vcpu,
1034 return kvm_vcpu_gfn_to_page(vcpu, gpa_to_gfn(gpa));
1037 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
1039 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
1041 return kvm_is_error_hva(hva);
1044 enum kvm_stat_kind {
1049 struct kvm_stat_data {
1055 struct kvm_stats_debugfs_item {
1058 enum kvm_stat_kind kind;
1061 extern struct kvm_stats_debugfs_item debugfs_entries[];
1062 extern struct dentry *kvm_debugfs_dir;
1064 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1065 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1067 if (unlikely(kvm->mmu_notifier_count))
1070 * Ensure the read of mmu_notifier_count happens before the read
1071 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1072 * mmu_notifier_invalidate_range_end to make sure that the caller
1073 * either sees the old (non-zero) value of mmu_notifier_count or
1074 * the new (incremented) value of mmu_notifier_seq.
1075 * PowerPC Book3s HV KVM calls this under a per-page lock
1076 * rather than under kvm->mmu_lock, for scalability, so
1077 * can't rely on kvm->mmu_lock to keep things ordered.
1080 if (kvm->mmu_notifier_seq != mmu_seq)
1086 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1088 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1090 bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
1091 int kvm_set_irq_routing(struct kvm *kvm,
1092 const struct kvm_irq_routing_entry *entries,
1095 int kvm_set_routing_entry(struct kvm *kvm,
1096 struct kvm_kernel_irq_routing_entry *e,
1097 const struct kvm_irq_routing_entry *ue);
1098 void kvm_free_irq_routing(struct kvm *kvm);
1102 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1106 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1108 #ifdef CONFIG_HAVE_KVM_EVENTFD
1110 void kvm_eventfd_init(struct kvm *kvm);
1111 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1113 #ifdef CONFIG_HAVE_KVM_IRQFD
1114 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1115 void kvm_irqfd_release(struct kvm *kvm);
1116 void kvm_irq_routing_update(struct kvm *);
1118 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1123 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1128 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1130 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1135 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1137 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1138 static inline void kvm_irq_routing_update(struct kvm *kvm)
1143 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1148 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1150 void kvm_arch_irq_routing_update(struct kvm *kvm);
1152 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1155 * Ensure the rest of the request is published to kvm_check_request's
1156 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1159 set_bit(req & KVM_REQUEST_MASK, &vcpu->requests);
1162 static inline bool kvm_request_pending(struct kvm_vcpu *vcpu)
1164 return READ_ONCE(vcpu->requests);
1167 static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)
1169 return test_bit(req & KVM_REQUEST_MASK, &vcpu->requests);
1172 static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu)
1174 clear_bit(req & KVM_REQUEST_MASK, &vcpu->requests);
1177 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1179 if (kvm_test_request(req, vcpu)) {
1180 kvm_clear_request(req, vcpu);
1183 * Ensure the rest of the request is visible to kvm_check_request's
1184 * caller. Paired with the smp_wmb in kvm_make_request.
1186 smp_mb__after_atomic();
1193 extern bool kvm_rebooting;
1195 extern unsigned int halt_poll_ns;
1196 extern unsigned int halt_poll_ns_grow;
1197 extern unsigned int halt_poll_ns_shrink;
1200 struct kvm_device_ops *ops;
1203 struct list_head vm_node;
1206 /* create, destroy, and name are mandatory */
1207 struct kvm_device_ops {
1211 * create is called holding kvm->lock and any operations not suitable
1212 * to do while holding the lock should be deferred to init (see
1215 int (*create)(struct kvm_device *dev, u32 type);
1218 * init is called after create if create is successful and is called
1219 * outside of holding kvm->lock.
1221 void (*init)(struct kvm_device *dev);
1224 * Destroy is responsible for freeing dev.
1226 * Destroy may be called before or after destructors are called
1227 * on emulated I/O regions, depending on whether a reference is
1228 * held by a vcpu or other kvm component that gets destroyed
1229 * after the emulated I/O.
1231 void (*destroy)(struct kvm_device *dev);
1233 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1234 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1235 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1236 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1240 void kvm_device_get(struct kvm_device *dev);
1241 void kvm_device_put(struct kvm_device *dev);
1242 struct kvm_device *kvm_device_from_filp(struct file *filp);
1243 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1244 void kvm_unregister_device_ops(u32 type);
1246 extern struct kvm_device_ops kvm_mpic_ops;
1247 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1248 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1250 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1252 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1254 vcpu->spin_loop.in_spin_loop = val;
1256 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1258 vcpu->spin_loop.dy_eligible = val;
1261 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1263 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1267 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1270 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1272 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1273 bool kvm_arch_has_irq_bypass(void);
1274 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1275 struct irq_bypass_producer *);
1276 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1277 struct irq_bypass_producer *);
1278 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1279 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1280 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1281 uint32_t guest_irq, bool set);
1282 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1284 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1285 /* If we wakeup during the poll time, was it a sucessful poll? */
1286 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1288 return vcpu->valid_wakeup;
1292 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1296 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
1298 typedef int (*kvm_vm_thread_fn_t)(struct kvm *kvm, uintptr_t data);
1300 int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn,
1301 uintptr_t data, const char *name,
1302 struct task_struct **thread_ptr);