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 <asm/signal.h>
32 #include <linux/kvm.h>
33 #include <linux/kvm_para.h>
35 #include <linux/kvm_types.h>
37 #include <asm/kvm_host.h>
39 #ifndef KVM_MAX_VCPU_ID
40 #define KVM_MAX_VCPU_ID KVM_MAX_VCPUS
44 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
45 * in kvm, other bits are visible for userspace which are defined in
46 * include/linux/kvm_h.
48 #define KVM_MEMSLOT_INVALID (1UL << 16)
50 /* Two fragments for cross MMIO pages. */
51 #define KVM_MAX_MMIO_FRAGMENTS 2
53 #ifndef KVM_ADDRESS_SPACE_NUM
54 #define KVM_ADDRESS_SPACE_NUM 1
58 * For the normal pfn, the highest 12 bits should be zero,
59 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
60 * mask bit 63 to indicate the noslot pfn.
62 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
63 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
64 #define KVM_PFN_NOSLOT (0x1ULL << 63)
66 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
67 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
68 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
71 * error pfns indicate that the gfn is in slot but faild to
72 * translate it to pfn on host.
74 static inline bool is_error_pfn(kvm_pfn_t pfn)
76 return !!(pfn & KVM_PFN_ERR_MASK);
80 * error_noslot pfns indicate that the gfn can not be
81 * translated to pfn - it is not in slot or failed to
82 * translate it to pfn.
84 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn)
86 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
89 /* noslot pfn indicates that the gfn is not in slot. */
90 static inline bool is_noslot_pfn(kvm_pfn_t pfn)
92 return pfn == KVM_PFN_NOSLOT;
96 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
97 * provide own defines and kvm_is_error_hva
99 #ifndef KVM_HVA_ERR_BAD
101 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
102 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
104 static inline bool kvm_is_error_hva(unsigned long addr)
106 return addr >= PAGE_OFFSET;
111 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
113 static inline bool is_error_page(struct page *page)
118 #define KVM_REQUEST_MASK GENMASK(7,0)
119 #define KVM_REQUEST_NO_WAKEUP BIT(8)
120 #define KVM_REQUEST_WAIT BIT(9)
122 * Architecture-independent vcpu->requests bit members
123 * Bits 4-7 are reserved for more arch-independent bits.
125 #define KVM_REQ_TLB_FLUSH (0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
126 #define KVM_REQ_MMU_RELOAD (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
127 #define KVM_REQ_PENDING_TIMER 2
128 #define KVM_REQ_UNHALT 3
129 #define KVM_REQUEST_ARCH_BASE 8
131 #define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
132 BUILD_BUG_ON((unsigned)(nr) >= 32 - KVM_REQUEST_ARCH_BASE); \
133 (unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) | (flags)); \
135 #define KVM_ARCH_REQ(nr) KVM_ARCH_REQ_FLAGS(nr, 0)
137 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
138 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
140 extern struct kmem_cache *kvm_vcpu_cache;
142 extern spinlock_t kvm_lock;
143 extern struct list_head vm_list;
145 struct kvm_io_range {
148 struct kvm_io_device *dev;
151 #define NR_IOBUS_DEVS 1000
156 struct kvm_io_range range[];
162 KVM_VIRTIO_CCW_NOTIFY_BUS,
167 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
168 int len, const void *val);
169 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
170 gpa_t addr, int len, const void *val, long cookie);
171 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
173 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
174 int len, struct kvm_io_device *dev);
175 void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
176 struct kvm_io_device *dev);
177 struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx,
180 #ifdef CONFIG_KVM_ASYNC_PF
181 struct kvm_async_pf {
182 struct work_struct work;
183 struct list_head link;
184 struct list_head queue;
185 struct kvm_vcpu *vcpu;
186 struct mm_struct *mm;
189 struct kvm_arch_async_pf arch;
193 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
194 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
195 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
196 struct kvm_arch_async_pf *arch);
197 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
204 READING_SHADOW_PAGE_TABLES,
208 * Sometimes a large or cross-page mmio needs to be broken up into separate
209 * exits for userspace servicing.
211 struct kvm_mmio_fragment {
219 #ifdef CONFIG_PREEMPT_NOTIFIERS
220 struct preempt_notifier preempt_notifier;
226 unsigned long requests;
227 unsigned long guest_debug;
230 struct list_head blocked_vcpu_list;
235 int guest_xcr0_loaded;
236 struct swait_queue_head wq;
237 struct pid __rcu *pid;
240 struct kvm_vcpu_stat stat;
241 unsigned int halt_poll_ns;
244 #ifdef CONFIG_HAS_IOMEM
246 int mmio_read_completed;
248 int mmio_cur_fragment;
249 int mmio_nr_fragments;
250 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
253 #ifdef CONFIG_KVM_ASYNC_PF
256 struct list_head queue;
257 struct list_head done;
262 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
264 * Cpu relax intercept or pause loop exit optimization
265 * in_spin_loop: set when a vcpu does a pause loop exit
266 * or cpu relax intercepted.
267 * dy_eligible: indicates whether vcpu is eligible for directed yield.
275 struct kvm_vcpu_arch arch;
276 struct dentry *debugfs_dentry;
279 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
282 * The memory barrier ensures a previous write to vcpu->requests cannot
283 * be reordered with the read of vcpu->mode. It pairs with the general
284 * memory barrier following the write of vcpu->mode in VCPU RUN.
286 smp_mb__before_atomic();
287 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
291 * Some of the bitops functions do not support too long bitmaps.
292 * This number must be determined not to exceed such limits.
294 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
296 struct kvm_memory_slot {
298 unsigned long npages;
299 unsigned long *dirty_bitmap;
300 struct kvm_arch_memory_slot arch;
301 unsigned long userspace_addr;
306 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
308 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
311 struct kvm_s390_adapter_int {
324 struct kvm_kernel_irq_routing_entry {
327 int (*set)(struct kvm_kernel_irq_routing_entry *e,
328 struct kvm *kvm, int irq_source_id, int level,
342 struct kvm_s390_adapter_int adapter;
343 struct kvm_hv_sint hv_sint;
345 struct hlist_node link;
348 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
349 struct kvm_irq_routing_table {
350 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
353 * Array indexed by gsi. Each entry contains list of irq chips
354 * the gsi is connected to.
356 struct hlist_head map[0];
360 #ifndef KVM_PRIVATE_MEM_SLOTS
361 #define KVM_PRIVATE_MEM_SLOTS 0
364 #ifndef KVM_MEM_SLOTS_NUM
365 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
368 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
369 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
377 * memslots are not sorted by id anymore, please use id_to_memslot()
378 * to get the memslot by its id.
380 struct kvm_memslots {
382 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
383 /* The mapping table from slot id to the index in memslots[]. */
384 short id_to_index[KVM_MEM_SLOTS_NUM];
391 struct mutex slots_lock;
392 struct mm_struct *mm; /* userspace tied to this vm */
393 struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
394 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
397 * created_vcpus is protected by kvm->lock, and is incremented
398 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
399 * incremented after storing the kvm_vcpu pointer in vcpus,
400 * and is accessed atomically.
402 atomic_t online_vcpus;
404 int last_boosted_vcpu;
405 struct list_head vm_list;
407 struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
408 #ifdef CONFIG_HAVE_KVM_EVENTFD
411 struct list_head items;
412 struct list_head resampler_list;
413 struct mutex resampler_lock;
415 struct list_head ioeventfds;
417 struct kvm_vm_stat stat;
418 struct kvm_arch arch;
419 refcount_t users_count;
420 #ifdef CONFIG_KVM_MMIO
421 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
422 spinlock_t ring_lock;
423 struct list_head coalesced_zones;
426 struct mutex irq_lock;
427 #ifdef CONFIG_HAVE_KVM_IRQCHIP
429 * Update side is protected by irq_lock.
431 struct kvm_irq_routing_table __rcu *irq_routing;
433 #ifdef CONFIG_HAVE_KVM_IRQFD
434 struct hlist_head irq_ack_notifier_list;
437 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
438 struct mmu_notifier mmu_notifier;
439 unsigned long mmu_notifier_seq;
440 long mmu_notifier_count;
443 struct list_head devices;
444 struct dentry *debugfs_dentry;
445 struct kvm_stat_data **debugfs_stat_data;
446 struct srcu_struct srcu;
447 struct srcu_struct irq_srcu;
451 #define kvm_err(fmt, ...) \
452 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
453 #define kvm_info(fmt, ...) \
454 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
455 #define kvm_debug(fmt, ...) \
456 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
457 #define kvm_debug_ratelimited(fmt, ...) \
458 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
460 #define kvm_pr_unimpl(fmt, ...) \
461 pr_err_ratelimited("kvm [%i]: " fmt, \
462 task_tgid_nr(current), ## __VA_ARGS__)
464 /* The guest did something we don't support. */
465 #define vcpu_unimpl(vcpu, fmt, ...) \
466 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
467 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
469 #define vcpu_debug(vcpu, fmt, ...) \
470 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
471 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
472 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
474 #define vcpu_err(vcpu, fmt, ...) \
475 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
477 static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx)
479 return srcu_dereference_check(kvm->buses[idx], &kvm->srcu,
480 lockdep_is_held(&kvm->slots_lock) ||
481 !refcount_read(&kvm->users_count));
484 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
486 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
487 * the caller has read kvm->online_vcpus before (as is the case
488 * for kvm_for_each_vcpu, for example).
491 return kvm->vcpus[i];
494 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
496 idx < atomic_read(&kvm->online_vcpus) && \
497 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
500 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
502 struct kvm_vcpu *vcpu = NULL;
507 if (id < KVM_MAX_VCPUS)
508 vcpu = kvm_get_vcpu(kvm, id);
509 if (vcpu && vcpu->vcpu_id == id)
511 kvm_for_each_vcpu(i, vcpu, kvm)
512 if (vcpu->vcpu_id == id)
517 static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu)
519 struct kvm_vcpu *tmp;
522 kvm_for_each_vcpu(idx, tmp, vcpu->kvm)
528 #define kvm_for_each_memslot(memslot, slots) \
529 for (memslot = &slots->memslots[0]; \
530 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
533 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
534 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
536 void vcpu_load(struct kvm_vcpu *vcpu);
537 void vcpu_put(struct kvm_vcpu *vcpu);
539 #ifdef __KVM_HAVE_IOAPIC
540 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
541 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
543 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
546 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
551 #ifdef CONFIG_HAVE_KVM_IRQFD
552 int kvm_irqfd_init(void);
553 void kvm_irqfd_exit(void);
555 static inline int kvm_irqfd_init(void)
560 static inline void kvm_irqfd_exit(void)
564 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
565 struct module *module);
568 void kvm_get_kvm(struct kvm *kvm);
569 void kvm_put_kvm(struct kvm *kvm);
571 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
573 return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu,
574 lockdep_is_held(&kvm->slots_lock) ||
575 !refcount_read(&kvm->users_count));
578 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
580 return __kvm_memslots(kvm, 0);
583 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
585 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
587 return __kvm_memslots(vcpu->kvm, as_id);
590 static inline struct kvm_memory_slot *
591 id_to_memslot(struct kvm_memslots *slots, int id)
593 int index = slots->id_to_index[id];
594 struct kvm_memory_slot *slot;
596 slot = &slots->memslots[index];
598 WARN_ON(slot->id != id);
603 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
604 * - create a new memory slot
605 * - delete an existing memory slot
606 * - modify an existing memory slot
607 * -- move it in the guest physical memory space
608 * -- just change its flags
610 * Since flags can be changed by some of these operations, the following
611 * differentiation is the best we can do for __kvm_set_memory_region():
620 int kvm_set_memory_region(struct kvm *kvm,
621 const struct kvm_userspace_memory_region *mem);
622 int __kvm_set_memory_region(struct kvm *kvm,
623 const struct kvm_userspace_memory_region *mem);
624 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
625 struct kvm_memory_slot *dont);
626 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
627 unsigned long npages);
628 void kvm_arch_memslots_updated(struct kvm *kvm, struct kvm_memslots *slots);
629 int kvm_arch_prepare_memory_region(struct kvm *kvm,
630 struct kvm_memory_slot *memslot,
631 const struct kvm_userspace_memory_region *mem,
632 enum kvm_mr_change change);
633 void kvm_arch_commit_memory_region(struct kvm *kvm,
634 const struct kvm_userspace_memory_region *mem,
635 const struct kvm_memory_slot *old,
636 const struct kvm_memory_slot *new,
637 enum kvm_mr_change change);
638 bool kvm_largepages_enabled(void);
639 void kvm_disable_largepages(void);
640 /* flush all memory translations */
641 void kvm_arch_flush_shadow_all(struct kvm *kvm);
642 /* flush memory translations pointing to 'slot' */
643 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
644 struct kvm_memory_slot *slot);
646 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
647 struct page **pages, int nr_pages);
649 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
650 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
651 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
652 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
653 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
655 void kvm_release_page_clean(struct page *page);
656 void kvm_release_page_dirty(struct page *page);
657 void kvm_set_page_accessed(struct page *page);
659 kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
660 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
661 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
663 kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
664 kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
665 kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
666 bool atomic, bool *async, bool write_fault,
669 void kvm_release_pfn_clean(kvm_pfn_t pfn);
670 void kvm_release_pfn_dirty(kvm_pfn_t pfn);
671 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
672 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
673 void kvm_get_pfn(kvm_pfn_t pfn);
675 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
677 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
679 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
680 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
681 void *data, unsigned long len);
682 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
683 int offset, int len);
684 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
686 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
687 void *data, unsigned long len);
688 int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
689 void *data, int offset, unsigned long len);
690 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
691 gpa_t gpa, unsigned long len);
692 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
693 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
694 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
695 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
696 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
697 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
699 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
700 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
701 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
702 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
703 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
704 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
705 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
706 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
708 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
710 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
712 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
713 int offset, int len);
714 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
716 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
718 void kvm_sigset_activate(struct kvm_vcpu *vcpu);
719 void kvm_sigset_deactivate(struct kvm_vcpu *vcpu);
721 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
722 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
723 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
724 bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
725 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
726 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
727 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool usermode_vcpu_not_eligible);
728 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
729 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
731 void kvm_flush_remote_tlbs(struct kvm *kvm);
732 void kvm_reload_remote_mmus(struct kvm *kvm);
733 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
735 long kvm_arch_dev_ioctl(struct file *filp,
736 unsigned int ioctl, unsigned long arg);
737 long kvm_arch_vcpu_ioctl(struct file *filp,
738 unsigned int ioctl, unsigned long arg);
739 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
741 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
743 int kvm_get_dirty_log(struct kvm *kvm,
744 struct kvm_dirty_log *log, int *is_dirty);
746 int kvm_get_dirty_log_protect(struct kvm *kvm,
747 struct kvm_dirty_log *log, bool *is_dirty);
749 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
750 struct kvm_memory_slot *slot,
754 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
755 struct kvm_dirty_log *log);
757 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
759 long kvm_arch_vm_ioctl(struct file *filp,
760 unsigned int ioctl, unsigned long arg);
762 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
763 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
765 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
766 struct kvm_translation *tr);
768 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
769 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
770 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
771 struct kvm_sregs *sregs);
772 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
773 struct kvm_sregs *sregs);
774 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
775 struct kvm_mp_state *mp_state);
776 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
777 struct kvm_mp_state *mp_state);
778 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
779 struct kvm_guest_debug *dbg);
780 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
782 int kvm_arch_init(void *opaque);
783 void kvm_arch_exit(void);
785 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
786 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
788 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
790 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
791 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
792 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
793 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
794 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
795 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
796 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
798 bool kvm_arch_has_vcpu_debugfs(void);
799 int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu);
801 int kvm_arch_hardware_enable(void);
802 void kvm_arch_hardware_disable(void);
803 int kvm_arch_hardware_setup(void);
804 void kvm_arch_hardware_unsetup(void);
805 void kvm_arch_check_processor_compat(void *rtn);
806 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
807 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
808 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
810 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
811 static inline struct kvm *kvm_arch_alloc_vm(void)
813 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
816 static inline void kvm_arch_free_vm(struct kvm *kvm)
822 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
823 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
824 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
825 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
827 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
831 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
835 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
840 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
841 void kvm_arch_start_assignment(struct kvm *kvm);
842 void kvm_arch_end_assignment(struct kvm *kvm);
843 bool kvm_arch_has_assigned_device(struct kvm *kvm);
845 static inline void kvm_arch_start_assignment(struct kvm *kvm)
849 static inline void kvm_arch_end_assignment(struct kvm *kvm)
853 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
859 static inline struct swait_queue_head *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
861 #ifdef __KVM_HAVE_ARCH_WQP
862 return vcpu->arch.wqp;
868 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
870 * returns true if the virtual interrupt controller is initialized and
871 * ready to accept virtual IRQ. On some architectures the virtual interrupt
872 * controller is dynamically instantiated and this is not always true.
874 bool kvm_arch_intc_initialized(struct kvm *kvm);
876 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
882 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
883 void kvm_arch_destroy_vm(struct kvm *kvm);
884 void kvm_arch_sync_events(struct kvm *kvm);
886 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
887 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
889 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
891 struct kvm_irq_ack_notifier {
892 struct hlist_node link;
894 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
897 int kvm_irq_map_gsi(struct kvm *kvm,
898 struct kvm_kernel_irq_routing_entry *entries, int gsi);
899 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
901 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
903 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
904 int irq_source_id, int level, bool line_status);
905 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
906 struct kvm *kvm, int irq_source_id,
907 int level, bool line_status);
908 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
909 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
910 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
911 void kvm_register_irq_ack_notifier(struct kvm *kvm,
912 struct kvm_irq_ack_notifier *kian);
913 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
914 struct kvm_irq_ack_notifier *kian);
915 int kvm_request_irq_source_id(struct kvm *kvm);
916 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
919 * search_memslots() and __gfn_to_memslot() are here because they are
920 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
921 * gfn_to_memslot() itself isn't here as an inline because that would
922 * bloat other code too much.
924 static inline struct kvm_memory_slot *
925 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
927 int start = 0, end = slots->used_slots;
928 int slot = atomic_read(&slots->lru_slot);
929 struct kvm_memory_slot *memslots = slots->memslots;
931 if (gfn >= memslots[slot].base_gfn &&
932 gfn < memslots[slot].base_gfn + memslots[slot].npages)
933 return &memslots[slot];
935 while (start < end) {
936 slot = start + (end - start) / 2;
938 if (gfn >= memslots[slot].base_gfn)
944 if (gfn >= memslots[start].base_gfn &&
945 gfn < memslots[start].base_gfn + memslots[start].npages) {
946 atomic_set(&slots->lru_slot, start);
947 return &memslots[start];
953 static inline struct kvm_memory_slot *
954 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
956 return search_memslots(slots, gfn);
959 static inline unsigned long
960 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
962 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
965 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
967 return gfn_to_memslot(kvm, gfn)->id;
971 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
973 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
975 return slot->base_gfn + gfn_offset;
978 static inline gpa_t gfn_to_gpa(gfn_t gfn)
980 return (gpa_t)gfn << PAGE_SHIFT;
983 static inline gfn_t gpa_to_gfn(gpa_t gpa)
985 return (gfn_t)(gpa >> PAGE_SHIFT);
988 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
990 return (hpa_t)pfn << PAGE_SHIFT;
993 static inline struct page *kvm_vcpu_gpa_to_page(struct kvm_vcpu *vcpu,
996 return kvm_vcpu_gfn_to_page(vcpu, gpa_to_gfn(gpa));
999 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
1001 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
1003 return kvm_is_error_hva(hva);
1006 enum kvm_stat_kind {
1011 struct kvm_stat_data {
1016 struct kvm_stats_debugfs_item {
1019 enum kvm_stat_kind kind;
1021 extern struct kvm_stats_debugfs_item debugfs_entries[];
1022 extern struct dentry *kvm_debugfs_dir;
1024 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1025 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1027 if (unlikely(kvm->mmu_notifier_count))
1030 * Ensure the read of mmu_notifier_count happens before the read
1031 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1032 * mmu_notifier_invalidate_range_end to make sure that the caller
1033 * either sees the old (non-zero) value of mmu_notifier_count or
1034 * the new (incremented) value of mmu_notifier_seq.
1035 * PowerPC Book3s HV KVM calls this under a per-page lock
1036 * rather than under kvm->mmu_lock, for scalability, so
1037 * can't rely on kvm->mmu_lock to keep things ordered.
1040 if (kvm->mmu_notifier_seq != mmu_seq)
1046 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1048 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1050 bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
1051 int kvm_set_irq_routing(struct kvm *kvm,
1052 const struct kvm_irq_routing_entry *entries,
1055 int kvm_set_routing_entry(struct kvm *kvm,
1056 struct kvm_kernel_irq_routing_entry *e,
1057 const struct kvm_irq_routing_entry *ue);
1058 void kvm_free_irq_routing(struct kvm *kvm);
1062 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1066 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1068 #ifdef CONFIG_HAVE_KVM_EVENTFD
1070 void kvm_eventfd_init(struct kvm *kvm);
1071 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1073 #ifdef CONFIG_HAVE_KVM_IRQFD
1074 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1075 void kvm_irqfd_release(struct kvm *kvm);
1076 void kvm_irq_routing_update(struct kvm *);
1078 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1083 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1088 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1090 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1095 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1097 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1098 static inline void kvm_irq_routing_update(struct kvm *kvm)
1103 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1108 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1110 void kvm_arch_irq_routing_update(struct kvm *kvm);
1112 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1115 * Ensure the rest of the request is published to kvm_check_request's
1116 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1119 set_bit(req & KVM_REQUEST_MASK, &vcpu->requests);
1122 static inline bool kvm_request_pending(struct kvm_vcpu *vcpu)
1124 return READ_ONCE(vcpu->requests);
1127 static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)
1129 return test_bit(req & KVM_REQUEST_MASK, &vcpu->requests);
1132 static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu)
1134 clear_bit(req & KVM_REQUEST_MASK, &vcpu->requests);
1137 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1139 if (kvm_test_request(req, vcpu)) {
1140 kvm_clear_request(req, vcpu);
1143 * Ensure the rest of the request is visible to kvm_check_request's
1144 * caller. Paired with the smp_wmb in kvm_make_request.
1146 smp_mb__after_atomic();
1153 extern bool kvm_rebooting;
1155 extern unsigned int halt_poll_ns;
1156 extern unsigned int halt_poll_ns_grow;
1157 extern unsigned int halt_poll_ns_shrink;
1160 struct kvm_device_ops *ops;
1163 struct list_head vm_node;
1166 /* create, destroy, and name are mandatory */
1167 struct kvm_device_ops {
1171 * create is called holding kvm->lock and any operations not suitable
1172 * to do while holding the lock should be deferred to init (see
1175 int (*create)(struct kvm_device *dev, u32 type);
1178 * init is called after create if create is successful and is called
1179 * outside of holding kvm->lock.
1181 void (*init)(struct kvm_device *dev);
1184 * Destroy is responsible for freeing dev.
1186 * Destroy may be called before or after destructors are called
1187 * on emulated I/O regions, depending on whether a reference is
1188 * held by a vcpu or other kvm component that gets destroyed
1189 * after the emulated I/O.
1191 void (*destroy)(struct kvm_device *dev);
1193 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1194 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1195 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1196 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1200 void kvm_device_get(struct kvm_device *dev);
1201 void kvm_device_put(struct kvm_device *dev);
1202 struct kvm_device *kvm_device_from_filp(struct file *filp);
1203 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1204 void kvm_unregister_device_ops(u32 type);
1206 extern struct kvm_device_ops kvm_mpic_ops;
1207 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1208 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1210 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1212 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1214 vcpu->spin_loop.in_spin_loop = val;
1216 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1218 vcpu->spin_loop.dy_eligible = val;
1221 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1223 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1227 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1230 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1232 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1233 bool kvm_arch_has_irq_bypass(void);
1234 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1235 struct irq_bypass_producer *);
1236 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1237 struct irq_bypass_producer *);
1238 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1239 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1240 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1241 uint32_t guest_irq, bool set);
1242 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1244 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1245 /* If we wakeup during the poll time, was it a sucessful poll? */
1246 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1248 return vcpu->valid_wakeup;
1252 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1256 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
1258 #ifdef CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL
1259 long kvm_arch_vcpu_async_ioctl(struct file *filp,
1260 unsigned int ioctl, unsigned long arg);
1262 static inline long kvm_arch_vcpu_async_ioctl(struct file *filp,
1266 return -ENOIOCTLCMD;
1268 #endif /* CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL */
1270 void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
1271 unsigned long start, unsigned long end);