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/vmalloc.h>
23 #include <linux/rcupdate.h>
24 #include <linux/ratelimit.h>
25 #include <linux/err.h>
26 #include <linux/irqflags.h>
27 #include <linux/context_tracking.h>
28 #include <linux/irqbypass.h>
29 #include <linux/swait.h>
30 #include <linux/refcount.h>
31 #include <linux/nospec.h>
32 #include <asm/signal.h>
34 #include <linux/kvm.h>
35 #include <linux/kvm_para.h>
37 #include <linux/kvm_types.h>
39 #include <asm/kvm_host.h>
41 #ifndef KVM_MAX_VCPU_ID
42 #define KVM_MAX_VCPU_ID KVM_MAX_VCPUS
46 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
47 * in kvm, other bits are visible for userspace which are defined in
48 * include/linux/kvm_h.
50 #define KVM_MEMSLOT_INVALID (1UL << 16)
52 /* Two fragments for cross MMIO pages. */
53 #define KVM_MAX_MMIO_FRAGMENTS 2
55 #ifndef KVM_ADDRESS_SPACE_NUM
56 #define KVM_ADDRESS_SPACE_NUM 1
60 * For the normal pfn, the highest 12 bits should be zero,
61 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
62 * mask bit 63 to indicate the noslot pfn.
64 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
65 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
66 #define KVM_PFN_NOSLOT (0x1ULL << 63)
68 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
69 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
70 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
73 * error pfns indicate that the gfn is in slot but faild to
74 * translate it to pfn on host.
76 static inline bool is_error_pfn(kvm_pfn_t pfn)
78 return !!(pfn & KVM_PFN_ERR_MASK);
82 * error_noslot pfns indicate that the gfn can not be
83 * translated to pfn - it is not in slot or failed to
84 * translate it to pfn.
86 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn)
88 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
91 /* noslot pfn indicates that the gfn is not in slot. */
92 static inline bool is_noslot_pfn(kvm_pfn_t pfn)
94 return pfn == KVM_PFN_NOSLOT;
98 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
99 * provide own defines and kvm_is_error_hva
101 #ifndef KVM_HVA_ERR_BAD
103 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
104 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
106 static inline bool kvm_is_error_hva(unsigned long addr)
108 return addr >= PAGE_OFFSET;
113 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
115 static inline bool is_error_page(struct page *page)
120 #define KVM_REQUEST_MASK GENMASK(7,0)
121 #define KVM_REQUEST_NO_WAKEUP BIT(8)
122 #define KVM_REQUEST_WAIT BIT(9)
124 * Architecture-independent vcpu->requests bit members
125 * Bits 4-7 are reserved for more arch-independent bits.
127 #define KVM_REQ_TLB_FLUSH (0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
128 #define KVM_REQ_MMU_RELOAD (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
129 #define KVM_REQ_PENDING_TIMER 2
130 #define KVM_REQ_UNHALT 3
131 #define KVM_REQUEST_ARCH_BASE 8
133 #define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
134 BUILD_BUG_ON((unsigned)(nr) >= (FIELD_SIZEOF(struct kvm_vcpu, requests) * 8) - 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, gpa_t cr2_or_gpa,
198 unsigned long hva, 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;
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 static inline unsigned long *kvm_second_dirty_bitmap(struct kvm_memory_slot *memslot)
315 unsigned long len = kvm_dirty_bitmap_bytes(memslot);
317 return memslot->dirty_bitmap + len / sizeof(*memslot->dirty_bitmap);
320 struct kvm_s390_adapter_int {
333 struct kvm_kernel_irq_routing_entry {
336 int (*set)(struct kvm_kernel_irq_routing_entry *e,
337 struct kvm *kvm, int irq_source_id, int level,
351 struct kvm_s390_adapter_int adapter;
352 struct kvm_hv_sint hv_sint;
354 struct hlist_node link;
357 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
358 struct kvm_irq_routing_table {
359 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
362 * Array indexed by gsi. Each entry contains list of irq chips
363 * the gsi is connected to.
365 struct hlist_head map[0];
369 #ifndef KVM_PRIVATE_MEM_SLOTS
370 #define KVM_PRIVATE_MEM_SLOTS 0
373 #ifndef KVM_MEM_SLOTS_NUM
374 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
377 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
378 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
386 * memslots are not sorted by id anymore, please use id_to_memslot()
387 * to get the memslot by its id.
389 struct kvm_memslots {
391 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
392 /* The mapping table from slot id to the index in memslots[]. */
393 short id_to_index[KVM_MEM_SLOTS_NUM];
400 struct mutex slots_lock;
401 struct mm_struct *mm; /* userspace tied to this vm */
402 struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
403 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
406 * created_vcpus is protected by kvm->lock, and is incremented
407 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
408 * incremented after storing the kvm_vcpu pointer in vcpus,
409 * and is accessed atomically.
411 atomic_t online_vcpus;
413 int last_boosted_vcpu;
414 struct list_head vm_list;
416 struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
417 #ifdef CONFIG_HAVE_KVM_EVENTFD
420 struct list_head items;
421 struct list_head resampler_list;
422 struct mutex resampler_lock;
424 struct list_head ioeventfds;
426 struct kvm_vm_stat stat;
427 struct kvm_arch arch;
428 refcount_t users_count;
429 #ifdef CONFIG_KVM_MMIO
430 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
431 spinlock_t ring_lock;
432 struct list_head coalesced_zones;
435 struct mutex irq_lock;
436 #ifdef CONFIG_HAVE_KVM_IRQCHIP
438 * Update side is protected by irq_lock.
440 struct kvm_irq_routing_table __rcu *irq_routing;
442 #ifdef CONFIG_HAVE_KVM_IRQFD
443 struct hlist_head irq_ack_notifier_list;
446 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
447 struct mmu_notifier mmu_notifier;
448 unsigned long mmu_notifier_seq;
449 long mmu_notifier_count;
452 struct list_head devices;
453 struct dentry *debugfs_dentry;
454 struct kvm_stat_data **debugfs_stat_data;
455 struct srcu_struct srcu;
456 struct srcu_struct irq_srcu;
460 #define kvm_err(fmt, ...) \
461 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
462 #define kvm_info(fmt, ...) \
463 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
464 #define kvm_debug(fmt, ...) \
465 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
466 #define kvm_debug_ratelimited(fmt, ...) \
467 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
469 #define kvm_pr_unimpl(fmt, ...) \
470 pr_err_ratelimited("kvm [%i]: " fmt, \
471 task_tgid_nr(current), ## __VA_ARGS__)
473 /* The guest did something we don't support. */
474 #define vcpu_unimpl(vcpu, fmt, ...) \
475 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
476 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
478 #define vcpu_debug(vcpu, fmt, ...) \
479 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
480 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
481 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
483 #define vcpu_err(vcpu, fmt, ...) \
484 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
486 static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx)
488 return srcu_dereference_check(kvm->buses[idx], &kvm->srcu,
489 lockdep_is_held(&kvm->slots_lock) ||
490 !refcount_read(&kvm->users_count));
493 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
495 int num_vcpus = atomic_read(&kvm->online_vcpus);
496 i = array_index_nospec(i, num_vcpus);
498 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. */
500 return kvm->vcpus[i];
503 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
505 idx < atomic_read(&kvm->online_vcpus) && \
506 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
509 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
511 struct kvm_vcpu *vcpu = NULL;
516 if (id < KVM_MAX_VCPUS)
517 vcpu = kvm_get_vcpu(kvm, id);
518 if (vcpu && vcpu->vcpu_id == id)
520 kvm_for_each_vcpu(i, vcpu, kvm)
521 if (vcpu->vcpu_id == id)
526 static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu)
528 struct kvm_vcpu *tmp;
531 kvm_for_each_vcpu(idx, tmp, vcpu->kvm)
537 #define kvm_for_each_memslot(memslot, slots) \
538 for (memslot = &slots->memslots[0]; \
539 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
542 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
543 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
545 void vcpu_load(struct kvm_vcpu *vcpu);
546 void vcpu_put(struct kvm_vcpu *vcpu);
548 #ifdef __KVM_HAVE_IOAPIC
549 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
550 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
552 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
555 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
560 #ifdef CONFIG_HAVE_KVM_IRQFD
561 int kvm_irqfd_init(void);
562 void kvm_irqfd_exit(void);
564 static inline int kvm_irqfd_init(void)
569 static inline void kvm_irqfd_exit(void)
573 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
574 struct module *module);
577 void kvm_get_kvm(struct kvm *kvm);
578 void kvm_put_kvm(struct kvm *kvm);
580 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
582 as_id = array_index_nospec(as_id, KVM_ADDRESS_SPACE_NUM);
583 return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu,
584 lockdep_is_held(&kvm->slots_lock) ||
585 !refcount_read(&kvm->users_count));
588 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
590 return __kvm_memslots(kvm, 0);
593 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
595 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
597 return __kvm_memslots(vcpu->kvm, as_id);
600 static inline struct kvm_memory_slot *
601 id_to_memslot(struct kvm_memslots *slots, int id)
603 int index = slots->id_to_index[id];
604 struct kvm_memory_slot *slot;
606 slot = &slots->memslots[index];
608 WARN_ON(slot->id != id);
613 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
614 * - create a new memory slot
615 * - delete an existing memory slot
616 * - modify an existing memory slot
617 * -- move it in the guest physical memory space
618 * -- just change its flags
620 * Since flags can be changed by some of these operations, the following
621 * differentiation is the best we can do for __kvm_set_memory_region():
630 int kvm_set_memory_region(struct kvm *kvm,
631 const struct kvm_userspace_memory_region *mem);
632 int __kvm_set_memory_region(struct kvm *kvm,
633 const struct kvm_userspace_memory_region *mem);
634 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
635 struct kvm_memory_slot *dont);
636 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
637 unsigned long npages);
638 void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen);
639 int kvm_arch_prepare_memory_region(struct kvm *kvm,
640 struct kvm_memory_slot *memslot,
641 const struct kvm_userspace_memory_region *mem,
642 enum kvm_mr_change change);
643 void kvm_arch_commit_memory_region(struct kvm *kvm,
644 const struct kvm_userspace_memory_region *mem,
645 const struct kvm_memory_slot *old,
646 const struct kvm_memory_slot *new,
647 enum kvm_mr_change change);
648 bool kvm_largepages_enabled(void);
649 void kvm_disable_largepages(void);
650 /* flush all memory translations */
651 void kvm_arch_flush_shadow_all(struct kvm *kvm);
652 /* flush memory translations pointing to 'slot' */
653 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
654 struct kvm_memory_slot *slot);
656 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
657 struct page **pages, int nr_pages);
659 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
660 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
661 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
662 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
663 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
665 void kvm_release_page_clean(struct page *page);
666 void kvm_release_page_dirty(struct page *page);
667 void kvm_set_page_accessed(struct page *page);
669 kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
670 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
671 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
673 kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
674 kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
675 kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
676 bool atomic, bool *async, bool write_fault,
679 void kvm_release_pfn_clean(kvm_pfn_t pfn);
680 void kvm_release_pfn_dirty(kvm_pfn_t pfn);
681 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
682 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
683 void kvm_get_pfn(kvm_pfn_t pfn);
685 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
687 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
689 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
690 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
691 void *data, unsigned long len);
692 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
693 int offset, int len);
694 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
696 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
697 void *data, unsigned long len);
698 int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
699 void *data, unsigned int offset,
701 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
702 gpa_t gpa, unsigned long len);
703 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
704 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
705 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
706 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
707 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
708 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
710 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
711 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
712 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
713 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
714 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
715 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
716 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
717 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
719 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
721 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
723 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
724 int offset, int len);
725 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
727 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
729 void kvm_sigset_activate(struct kvm_vcpu *vcpu);
730 void kvm_sigset_deactivate(struct kvm_vcpu *vcpu);
732 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
733 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
734 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
735 bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
736 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
737 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
738 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool usermode_vcpu_not_eligible);
740 void kvm_flush_remote_tlbs(struct kvm *kvm);
741 void kvm_reload_remote_mmus(struct kvm *kvm);
743 bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
744 unsigned long *vcpu_bitmap, cpumask_var_t tmp);
745 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
747 long kvm_arch_dev_ioctl(struct file *filp,
748 unsigned int ioctl, unsigned long arg);
749 long kvm_arch_vcpu_ioctl(struct file *filp,
750 unsigned int ioctl, unsigned long arg);
751 vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
753 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
755 int kvm_get_dirty_log(struct kvm *kvm,
756 struct kvm_dirty_log *log, int *is_dirty);
758 int kvm_get_dirty_log_protect(struct kvm *kvm,
759 struct kvm_dirty_log *log, bool *is_dirty);
761 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
762 struct kvm_memory_slot *slot,
766 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
767 struct kvm_dirty_log *log);
769 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
771 long kvm_arch_vm_ioctl(struct file *filp,
772 unsigned int ioctl, unsigned long arg);
774 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
775 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
777 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
778 struct kvm_translation *tr);
780 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
781 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
782 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
783 struct kvm_sregs *sregs);
784 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
785 struct kvm_sregs *sregs);
786 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
787 struct kvm_mp_state *mp_state);
788 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
789 struct kvm_mp_state *mp_state);
790 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
791 struct kvm_guest_debug *dbg);
792 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
794 int kvm_arch_init(void *opaque);
795 void kvm_arch_exit(void);
797 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
798 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
800 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
802 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
803 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
804 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
805 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
806 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
807 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
808 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
810 bool kvm_arch_has_vcpu_debugfs(void);
811 int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu);
813 int kvm_arch_hardware_enable(void);
814 void kvm_arch_hardware_disable(void);
815 int kvm_arch_hardware_setup(void);
816 void kvm_arch_hardware_unsetup(void);
817 void kvm_arch_check_processor_compat(void *rtn);
818 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
819 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
820 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
821 bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu);
823 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
825 * All architectures that want to use vzalloc currently also
826 * need their own kvm_arch_alloc_vm implementation.
828 static inline struct kvm *kvm_arch_alloc_vm(void)
830 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
833 static inline void kvm_arch_free_vm(struct kvm *kvm)
839 #ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
840 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
846 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
847 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
848 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
849 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
851 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
855 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
859 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
864 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
865 void kvm_arch_start_assignment(struct kvm *kvm);
866 void kvm_arch_end_assignment(struct kvm *kvm);
867 bool kvm_arch_has_assigned_device(struct kvm *kvm);
869 static inline void kvm_arch_start_assignment(struct kvm *kvm)
873 static inline void kvm_arch_end_assignment(struct kvm *kvm)
877 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
883 static inline struct swait_queue_head *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
885 #ifdef __KVM_HAVE_ARCH_WQP
886 return vcpu->arch.wqp;
892 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
894 * returns true if the virtual interrupt controller is initialized and
895 * ready to accept virtual IRQ. On some architectures the virtual interrupt
896 * controller is dynamically instantiated and this is not always true.
898 bool kvm_arch_intc_initialized(struct kvm *kvm);
900 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
906 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
907 void kvm_arch_destroy_vm(struct kvm *kvm);
908 void kvm_arch_sync_events(struct kvm *kvm);
910 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
911 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
913 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
914 bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
916 struct kvm_irq_ack_notifier {
917 struct hlist_node link;
919 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
922 int kvm_irq_map_gsi(struct kvm *kvm,
923 struct kvm_kernel_irq_routing_entry *entries, int gsi);
924 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
926 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
928 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
929 int irq_source_id, int level, bool line_status);
930 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
931 struct kvm *kvm, int irq_source_id,
932 int level, bool line_status);
933 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
934 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
935 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
936 void kvm_register_irq_ack_notifier(struct kvm *kvm,
937 struct kvm_irq_ack_notifier *kian);
938 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
939 struct kvm_irq_ack_notifier *kian);
940 int kvm_request_irq_source_id(struct kvm *kvm);
941 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
944 * search_memslots() and __gfn_to_memslot() are here because they are
945 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
946 * gfn_to_memslot() itself isn't here as an inline because that would
947 * bloat other code too much.
949 static inline struct kvm_memory_slot *
950 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
952 int start = 0, end = slots->used_slots;
953 int slot = atomic_read(&slots->lru_slot);
954 struct kvm_memory_slot *memslots = slots->memslots;
956 if (gfn >= memslots[slot].base_gfn &&
957 gfn < memslots[slot].base_gfn + memslots[slot].npages)
958 return &memslots[slot];
960 while (start < end) {
961 slot = start + (end - start) / 2;
963 if (gfn >= memslots[slot].base_gfn)
969 if (gfn >= memslots[start].base_gfn &&
970 gfn < memslots[start].base_gfn + memslots[start].npages) {
971 atomic_set(&slots->lru_slot, start);
972 return &memslots[start];
978 static inline struct kvm_memory_slot *
979 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
981 return search_memslots(slots, gfn);
984 static inline unsigned long
985 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
987 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
990 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
992 return gfn_to_memslot(kvm, gfn)->id;
996 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
998 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
1000 return slot->base_gfn + gfn_offset;
1003 static inline gpa_t gfn_to_gpa(gfn_t gfn)
1005 return (gpa_t)gfn << PAGE_SHIFT;
1008 static inline gfn_t gpa_to_gfn(gpa_t gpa)
1010 return (gfn_t)(gpa >> PAGE_SHIFT);
1013 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
1015 return (hpa_t)pfn << PAGE_SHIFT;
1018 static inline struct page *kvm_vcpu_gpa_to_page(struct kvm_vcpu *vcpu,
1021 return kvm_vcpu_gfn_to_page(vcpu, gpa_to_gfn(gpa));
1024 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
1026 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
1028 return kvm_is_error_hva(hva);
1031 enum kvm_stat_kind {
1036 struct kvm_stat_data {
1042 struct kvm_stats_debugfs_item {
1045 enum kvm_stat_kind kind;
1048 extern struct kvm_stats_debugfs_item debugfs_entries[];
1049 extern struct dentry *kvm_debugfs_dir;
1051 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1052 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1054 if (unlikely(kvm->mmu_notifier_count))
1057 * Ensure the read of mmu_notifier_count happens before the read
1058 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1059 * mmu_notifier_invalidate_range_end to make sure that the caller
1060 * either sees the old (non-zero) value of mmu_notifier_count or
1061 * the new (incremented) value of mmu_notifier_seq.
1062 * PowerPC Book3s HV KVM calls this under a per-page lock
1063 * rather than under kvm->mmu_lock, for scalability, so
1064 * can't rely on kvm->mmu_lock to keep things ordered.
1067 if (kvm->mmu_notifier_seq != mmu_seq)
1073 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1075 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1077 bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
1078 int kvm_set_irq_routing(struct kvm *kvm,
1079 const struct kvm_irq_routing_entry *entries,
1082 int kvm_set_routing_entry(struct kvm *kvm,
1083 struct kvm_kernel_irq_routing_entry *e,
1084 const struct kvm_irq_routing_entry *ue);
1085 void kvm_free_irq_routing(struct kvm *kvm);
1089 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1093 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1095 #ifdef CONFIG_HAVE_KVM_EVENTFD
1097 void kvm_eventfd_init(struct kvm *kvm);
1098 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1100 #ifdef CONFIG_HAVE_KVM_IRQFD
1101 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1102 void kvm_irqfd_release(struct kvm *kvm);
1103 void kvm_irq_routing_update(struct kvm *);
1105 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1110 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1115 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1117 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1122 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1124 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1125 static inline void kvm_irq_routing_update(struct kvm *kvm)
1130 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1135 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1137 void kvm_arch_irq_routing_update(struct kvm *kvm);
1139 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1142 * Ensure the rest of the request is published to kvm_check_request's
1143 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1146 set_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1149 static inline bool kvm_request_pending(struct kvm_vcpu *vcpu)
1151 return READ_ONCE(vcpu->requests);
1154 static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)
1156 return test_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1159 static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu)
1161 clear_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1164 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1166 if (kvm_test_request(req, vcpu)) {
1167 kvm_clear_request(req, vcpu);
1170 * Ensure the rest of the request is visible to kvm_check_request's
1171 * caller. Paired with the smp_wmb in kvm_make_request.
1173 smp_mb__after_atomic();
1180 extern bool kvm_rebooting;
1182 extern unsigned int halt_poll_ns;
1183 extern unsigned int halt_poll_ns_grow;
1184 extern unsigned int halt_poll_ns_shrink;
1187 struct kvm_device_ops *ops;
1190 struct list_head vm_node;
1193 /* create, destroy, and name are mandatory */
1194 struct kvm_device_ops {
1198 * create is called holding kvm->lock and any operations not suitable
1199 * to do while holding the lock should be deferred to init (see
1202 int (*create)(struct kvm_device *dev, u32 type);
1205 * init is called after create if create is successful and is called
1206 * outside of holding kvm->lock.
1208 void (*init)(struct kvm_device *dev);
1211 * Destroy is responsible for freeing dev.
1213 * Destroy may be called before or after destructors are called
1214 * on emulated I/O regions, depending on whether a reference is
1215 * held by a vcpu or other kvm component that gets destroyed
1216 * after the emulated I/O.
1218 void (*destroy)(struct kvm_device *dev);
1220 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1221 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1222 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1223 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1227 void kvm_device_get(struct kvm_device *dev);
1228 void kvm_device_put(struct kvm_device *dev);
1229 struct kvm_device *kvm_device_from_filp(struct file *filp);
1230 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1231 void kvm_unregister_device_ops(u32 type);
1233 extern struct kvm_device_ops kvm_mpic_ops;
1234 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1235 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1237 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1239 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1241 vcpu->spin_loop.in_spin_loop = val;
1243 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1245 vcpu->spin_loop.dy_eligible = val;
1248 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1250 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1254 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1257 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1259 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1260 bool kvm_arch_has_irq_bypass(void);
1261 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1262 struct irq_bypass_producer *);
1263 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1264 struct irq_bypass_producer *);
1265 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1266 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1267 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1268 uint32_t guest_irq, bool set);
1269 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1271 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1272 /* If we wakeup during the poll time, was it a sucessful poll? */
1273 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1275 return vcpu->valid_wakeup;
1279 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1283 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
1285 #ifdef CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL
1286 long kvm_arch_vcpu_async_ioctl(struct file *filp,
1287 unsigned int ioctl, unsigned long arg);
1289 static inline long kvm_arch_vcpu_async_ioctl(struct file *filp,
1293 return -ENOIOCTLCMD;
1295 #endif /* CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL */
1297 int kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
1298 unsigned long start, unsigned long end, bool blockable);
1300 #ifdef CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE
1301 int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu);
1303 static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
1307 #endif /* CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE */
1309 typedef int (*kvm_vm_thread_fn_t)(struct kvm *kvm, uintptr_t data);
1311 int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn,
1312 uintptr_t data, const char *name,
1313 struct task_struct **thread_ptr);