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 <asm/signal.h>
31 #include <linux/kvm.h>
32 #include <linux/kvm_para.h>
34 #include <linux/kvm_types.h>
36 #include <asm/kvm_host.h>
38 #ifndef KVM_MAX_VCPU_ID
39 #define KVM_MAX_VCPU_ID KVM_MAX_VCPUS
43 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
44 * in kvm, other bits are visible for userspace which are defined in
45 * include/linux/kvm_h.
47 #define KVM_MEMSLOT_INVALID (1UL << 16)
49 /* Two fragments for cross MMIO pages. */
50 #define KVM_MAX_MMIO_FRAGMENTS 2
52 #ifndef KVM_ADDRESS_SPACE_NUM
53 #define KVM_ADDRESS_SPACE_NUM 1
57 * For the normal pfn, the highest 12 bits should be zero,
58 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
59 * mask bit 63 to indicate the noslot pfn.
61 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
62 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
63 #define KVM_PFN_NOSLOT (0x1ULL << 63)
65 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
66 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
67 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
70 * error pfns indicate that the gfn is in slot but faild to
71 * translate it to pfn on host.
73 static inline bool is_error_pfn(kvm_pfn_t pfn)
75 return !!(pfn & KVM_PFN_ERR_MASK);
79 * error_noslot pfns indicate that the gfn can not be
80 * translated to pfn - it is not in slot or failed to
81 * translate it to pfn.
83 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn)
85 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
88 /* noslot pfn indicates that the gfn is not in slot. */
89 static inline bool is_noslot_pfn(kvm_pfn_t pfn)
91 return pfn == KVM_PFN_NOSLOT;
95 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
96 * provide own defines and kvm_is_error_hva
98 #ifndef KVM_HVA_ERR_BAD
100 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
101 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
103 static inline bool kvm_is_error_hva(unsigned long addr)
105 return addr >= PAGE_OFFSET;
110 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
112 static inline bool is_error_page(struct page *page)
118 * Architecture-independent vcpu->requests bit members
119 * Bits 4-7 are reserved for more arch-independent bits.
121 #define KVM_REQ_TLB_FLUSH 0
122 #define KVM_REQ_MMU_RELOAD 1
123 #define KVM_REQ_PENDING_TIMER 2
124 #define KVM_REQ_UNHALT 3
126 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
127 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
129 extern struct kmem_cache *kvm_vcpu_cache;
131 extern spinlock_t kvm_lock;
132 extern struct list_head vm_list;
134 struct kvm_io_range {
137 struct kvm_io_device *dev;
140 #define NR_IOBUS_DEVS 1000
145 struct kvm_io_range range[];
151 KVM_VIRTIO_CCW_NOTIFY_BUS,
156 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
157 int len, const void *val);
158 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
159 gpa_t addr, int len, const void *val, long cookie);
160 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
162 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
163 int len, struct kvm_io_device *dev);
164 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
165 struct kvm_io_device *dev);
166 struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx,
169 #ifdef CONFIG_KVM_ASYNC_PF
170 struct kvm_async_pf {
171 struct work_struct work;
172 struct list_head link;
173 struct list_head queue;
174 struct kvm_vcpu *vcpu;
175 struct mm_struct *mm;
178 struct kvm_arch_async_pf arch;
182 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
183 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
184 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
185 struct kvm_arch_async_pf *arch);
186 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
193 READING_SHADOW_PAGE_TABLES,
197 * Sometimes a large or cross-page mmio needs to be broken up into separate
198 * exits for userspace servicing.
200 struct kvm_mmio_fragment {
208 #ifdef CONFIG_PREEMPT_NOTIFIERS
209 struct preempt_notifier preempt_notifier;
215 unsigned long requests;
216 unsigned long guest_debug;
219 struct list_head blocked_vcpu_list;
224 int guest_fpu_loaded, guest_xcr0_loaded;
225 struct swait_queue_head wq;
229 struct kvm_vcpu_stat stat;
230 unsigned int halt_poll_ns;
233 #ifdef CONFIG_HAS_IOMEM
235 int mmio_read_completed;
237 int mmio_cur_fragment;
238 int mmio_nr_fragments;
239 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
242 #ifdef CONFIG_KVM_ASYNC_PF
245 struct list_head queue;
246 struct list_head done;
251 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
253 * Cpu relax intercept or pause loop exit optimization
254 * in_spin_loop: set when a vcpu does a pause loop exit
255 * or cpu relax intercepted.
256 * dy_eligible: indicates whether vcpu is eligible for directed yield.
264 struct kvm_vcpu_arch arch;
265 struct dentry *debugfs_dentry;
268 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
270 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
274 * Some of the bitops functions do not support too long bitmaps.
275 * This number must be determined not to exceed such limits.
277 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
279 struct kvm_memory_slot {
281 unsigned long npages;
282 unsigned long *dirty_bitmap;
283 struct kvm_arch_memory_slot arch;
284 unsigned long userspace_addr;
289 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
291 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
294 struct kvm_s390_adapter_int {
307 struct kvm_kernel_irq_routing_entry {
310 int (*set)(struct kvm_kernel_irq_routing_entry *e,
311 struct kvm *kvm, int irq_source_id, int level,
325 struct kvm_s390_adapter_int adapter;
326 struct kvm_hv_sint hv_sint;
328 struct hlist_node link;
331 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
332 struct kvm_irq_routing_table {
333 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
336 * Array indexed by gsi. Each entry contains list of irq chips
337 * the gsi is connected to.
339 struct hlist_head map[0];
343 #ifndef KVM_PRIVATE_MEM_SLOTS
344 #define KVM_PRIVATE_MEM_SLOTS 0
347 #ifndef KVM_MEM_SLOTS_NUM
348 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
351 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
352 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
360 * memslots are not sorted by id anymore, please use id_to_memslot()
361 * to get the memslot by its id.
363 struct kvm_memslots {
365 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
366 /* The mapping table from slot id to the index in memslots[]. */
367 short id_to_index[KVM_MEM_SLOTS_NUM];
374 struct mutex slots_lock;
375 struct mm_struct *mm; /* userspace tied to this vm */
376 struct kvm_memslots *memslots[KVM_ADDRESS_SPACE_NUM];
377 struct srcu_struct srcu;
378 struct srcu_struct irq_srcu;
379 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
382 * created_vcpus is protected by kvm->lock, and is incremented
383 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
384 * incremented after storing the kvm_vcpu pointer in vcpus,
385 * and is accessed atomically.
387 atomic_t online_vcpus;
389 int last_boosted_vcpu;
390 struct list_head vm_list;
392 struct kvm_io_bus *buses[KVM_NR_BUSES];
393 #ifdef CONFIG_HAVE_KVM_EVENTFD
396 struct list_head items;
397 struct list_head resampler_list;
398 struct mutex resampler_lock;
400 struct list_head ioeventfds;
402 struct kvm_vm_stat stat;
403 struct kvm_arch arch;
404 atomic_t users_count;
405 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
406 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
407 spinlock_t ring_lock;
408 struct list_head coalesced_zones;
411 struct mutex irq_lock;
412 #ifdef CONFIG_HAVE_KVM_IRQCHIP
414 * Update side is protected by irq_lock.
416 struct kvm_irq_routing_table __rcu *irq_routing;
418 #ifdef CONFIG_HAVE_KVM_IRQFD
419 struct hlist_head irq_ack_notifier_list;
422 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
423 struct mmu_notifier mmu_notifier;
424 unsigned long mmu_notifier_seq;
425 long mmu_notifier_count;
428 struct list_head devices;
429 struct dentry *debugfs_dentry;
430 struct kvm_stat_data **debugfs_stat_data;
433 #define kvm_err(fmt, ...) \
434 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
435 #define kvm_info(fmt, ...) \
436 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
437 #define kvm_debug(fmt, ...) \
438 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
439 #define kvm_debug_ratelimited(fmt, ...) \
440 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
442 #define kvm_pr_unimpl(fmt, ...) \
443 pr_err_ratelimited("kvm [%i]: " fmt, \
444 task_tgid_nr(current), ## __VA_ARGS__)
446 /* The guest did something we don't support. */
447 #define vcpu_unimpl(vcpu, fmt, ...) \
448 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
449 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
451 #define vcpu_debug(vcpu, fmt, ...) \
452 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
453 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
454 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
456 #define vcpu_err(vcpu, fmt, ...) \
457 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
459 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
461 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
462 * the caller has read kvm->online_vcpus before (as is the case
463 * for kvm_for_each_vcpu, for example).
466 return kvm->vcpus[i];
469 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
471 idx < atomic_read(&kvm->online_vcpus) && \
472 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
475 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
477 struct kvm_vcpu *vcpu = NULL;
482 if (id < KVM_MAX_VCPUS)
483 vcpu = kvm_get_vcpu(kvm, id);
484 if (vcpu && vcpu->vcpu_id == id)
486 kvm_for_each_vcpu(i, vcpu, kvm)
487 if (vcpu->vcpu_id == id)
492 #define kvm_for_each_memslot(memslot, slots) \
493 for (memslot = &slots->memslots[0]; \
494 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
497 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
498 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
500 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
501 void vcpu_put(struct kvm_vcpu *vcpu);
503 #ifdef __KVM_HAVE_IOAPIC
504 void kvm_vcpu_request_scan_ioapic(struct kvm *kvm);
505 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
507 static inline void kvm_vcpu_request_scan_ioapic(struct kvm *kvm)
510 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
515 #ifdef CONFIG_HAVE_KVM_IRQFD
516 int kvm_irqfd_init(void);
517 void kvm_irqfd_exit(void);
519 static inline int kvm_irqfd_init(void)
524 static inline void kvm_irqfd_exit(void)
528 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
529 struct module *module);
532 void kvm_get_kvm(struct kvm *kvm);
533 void kvm_put_kvm(struct kvm *kvm);
535 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
537 return rcu_dereference_check(kvm->memslots[as_id],
538 srcu_read_lock_held(&kvm->srcu)
539 || lockdep_is_held(&kvm->slots_lock));
542 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
544 return __kvm_memslots(kvm, 0);
547 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
549 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
551 return __kvm_memslots(vcpu->kvm, as_id);
554 static inline struct kvm_memory_slot *
555 id_to_memslot(struct kvm_memslots *slots, int id)
557 int index = slots->id_to_index[id];
558 struct kvm_memory_slot *slot;
560 slot = &slots->memslots[index];
562 WARN_ON(slot->id != id);
567 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
568 * - create a new memory slot
569 * - delete an existing memory slot
570 * - modify an existing memory slot
571 * -- move it in the guest physical memory space
572 * -- just change its flags
574 * Since flags can be changed by some of these operations, the following
575 * differentiation is the best we can do for __kvm_set_memory_region():
584 int kvm_set_memory_region(struct kvm *kvm,
585 const struct kvm_userspace_memory_region *mem);
586 int __kvm_set_memory_region(struct kvm *kvm,
587 const struct kvm_userspace_memory_region *mem);
588 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
589 struct kvm_memory_slot *dont);
590 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
591 unsigned long npages);
592 void kvm_arch_memslots_updated(struct kvm *kvm, struct kvm_memslots *slots);
593 int kvm_arch_prepare_memory_region(struct kvm *kvm,
594 struct kvm_memory_slot *memslot,
595 const struct kvm_userspace_memory_region *mem,
596 enum kvm_mr_change change);
597 void kvm_arch_commit_memory_region(struct kvm *kvm,
598 const struct kvm_userspace_memory_region *mem,
599 const struct kvm_memory_slot *old,
600 const struct kvm_memory_slot *new,
601 enum kvm_mr_change change);
602 bool kvm_largepages_enabled(void);
603 void kvm_disable_largepages(void);
604 /* flush all memory translations */
605 void kvm_arch_flush_shadow_all(struct kvm *kvm);
606 /* flush memory translations pointing to 'slot' */
607 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
608 struct kvm_memory_slot *slot);
610 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
611 struct page **pages, int nr_pages);
613 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
614 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
615 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
616 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
617 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
619 void kvm_release_page_clean(struct page *page);
620 void kvm_release_page_dirty(struct page *page);
621 void kvm_set_page_accessed(struct page *page);
623 kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
624 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
625 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
627 kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
628 kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
629 kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
630 bool atomic, bool *async, bool write_fault,
633 void kvm_release_pfn_clean(kvm_pfn_t pfn);
634 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
635 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
636 void kvm_get_pfn(kvm_pfn_t pfn);
638 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
640 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
642 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
643 int kvm_vcpu_read_guest_cached(struct kvm_vcpu *vcpu, struct gfn_to_hva_cache *ghc,
644 void *data, unsigned long len);
645 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
646 int offset, int len);
647 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
649 int kvm_vcpu_write_guest_cached(struct kvm_vcpu *v, struct gfn_to_hva_cache *ghc,
650 void *data, unsigned long len);
651 int kvm_vcpu_write_guest_offset_cached(struct kvm_vcpu *v, struct gfn_to_hva_cache *ghc,
652 void *data, int offset, unsigned long len);
653 int kvm_vcpu_gfn_to_hva_cache_init(struct kvm_vcpu *v, struct gfn_to_hva_cache *ghc,
654 gpa_t gpa, unsigned long len);
655 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
656 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
657 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
658 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
659 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
660 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
662 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
663 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
664 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
665 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
666 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
667 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
668 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
669 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
671 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
673 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
675 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
676 int offset, int len);
677 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
679 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
681 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
682 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
683 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
684 void kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
685 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
686 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
687 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
688 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
689 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
691 void kvm_flush_remote_tlbs(struct kvm *kvm);
692 void kvm_reload_remote_mmus(struct kvm *kvm);
693 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
695 long kvm_arch_dev_ioctl(struct file *filp,
696 unsigned int ioctl, unsigned long arg);
697 long kvm_arch_vcpu_ioctl(struct file *filp,
698 unsigned int ioctl, unsigned long arg);
699 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
701 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
703 int kvm_get_dirty_log(struct kvm *kvm,
704 struct kvm_dirty_log *log, int *is_dirty);
706 int kvm_get_dirty_log_protect(struct kvm *kvm,
707 struct kvm_dirty_log *log, bool *is_dirty);
709 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
710 struct kvm_memory_slot *slot,
714 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
715 struct kvm_dirty_log *log);
717 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
719 long kvm_arch_vm_ioctl(struct file *filp,
720 unsigned int ioctl, unsigned long arg);
722 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
723 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
725 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
726 struct kvm_translation *tr);
728 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
729 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
730 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
731 struct kvm_sregs *sregs);
732 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
733 struct kvm_sregs *sregs);
734 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
735 struct kvm_mp_state *mp_state);
736 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
737 struct kvm_mp_state *mp_state);
738 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
739 struct kvm_guest_debug *dbg);
740 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
742 int kvm_arch_init(void *opaque);
743 void kvm_arch_exit(void);
745 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
746 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
748 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
750 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
751 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
752 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
753 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
754 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
755 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
756 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
758 bool kvm_arch_has_vcpu_debugfs(void);
759 int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu);
761 int kvm_arch_hardware_enable(void);
762 void kvm_arch_hardware_disable(void);
763 int kvm_arch_hardware_setup(void);
764 void kvm_arch_hardware_unsetup(void);
765 void kvm_arch_check_processor_compat(void *rtn);
766 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
767 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
769 void *kvm_kvzalloc(unsigned long size);
771 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
772 static inline struct kvm *kvm_arch_alloc_vm(void)
774 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
777 static inline void kvm_arch_free_vm(struct kvm *kvm)
783 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
784 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
785 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
786 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
788 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
792 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
796 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
801 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
802 void kvm_arch_start_assignment(struct kvm *kvm);
803 void kvm_arch_end_assignment(struct kvm *kvm);
804 bool kvm_arch_has_assigned_device(struct kvm *kvm);
806 static inline void kvm_arch_start_assignment(struct kvm *kvm)
810 static inline void kvm_arch_end_assignment(struct kvm *kvm)
814 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
820 static inline struct swait_queue_head *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
822 #ifdef __KVM_HAVE_ARCH_WQP
823 return vcpu->arch.wqp;
829 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
831 * returns true if the virtual interrupt controller is initialized and
832 * ready to accept virtual IRQ. On some architectures the virtual interrupt
833 * controller is dynamically instantiated and this is not always true.
835 bool kvm_arch_intc_initialized(struct kvm *kvm);
837 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
843 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
844 void kvm_arch_destroy_vm(struct kvm *kvm);
845 void kvm_arch_sync_events(struct kvm *kvm);
847 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
848 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
850 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
852 struct kvm_irq_ack_notifier {
853 struct hlist_node link;
855 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
858 int kvm_irq_map_gsi(struct kvm *kvm,
859 struct kvm_kernel_irq_routing_entry *entries, int gsi);
860 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
862 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
864 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
865 int irq_source_id, int level, bool line_status);
866 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
867 struct kvm *kvm, int irq_source_id,
868 int level, bool line_status);
869 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
870 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
871 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
872 void kvm_register_irq_ack_notifier(struct kvm *kvm,
873 struct kvm_irq_ack_notifier *kian);
874 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
875 struct kvm_irq_ack_notifier *kian);
876 int kvm_request_irq_source_id(struct kvm *kvm);
877 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
879 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
880 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
881 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
883 static inline int kvm_iommu_map_pages(struct kvm *kvm,
884 struct kvm_memory_slot *slot)
889 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
890 struct kvm_memory_slot *slot)
896 * search_memslots() and __gfn_to_memslot() are here because they are
897 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
898 * gfn_to_memslot() itself isn't here as an inline because that would
899 * bloat other code too much.
901 static inline struct kvm_memory_slot *
902 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
904 int start = 0, end = slots->used_slots;
905 int slot = atomic_read(&slots->lru_slot);
906 struct kvm_memory_slot *memslots = slots->memslots;
908 if (gfn >= memslots[slot].base_gfn &&
909 gfn < memslots[slot].base_gfn + memslots[slot].npages)
910 return &memslots[slot];
912 while (start < end) {
913 slot = start + (end - start) / 2;
915 if (gfn >= memslots[slot].base_gfn)
921 if (gfn >= memslots[start].base_gfn &&
922 gfn < memslots[start].base_gfn + memslots[start].npages) {
923 atomic_set(&slots->lru_slot, start);
924 return &memslots[start];
930 static inline struct kvm_memory_slot *
931 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
933 return search_memslots(slots, gfn);
936 static inline unsigned long
937 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
939 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
942 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
944 return gfn_to_memslot(kvm, gfn)->id;
948 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
950 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
952 return slot->base_gfn + gfn_offset;
955 static inline gpa_t gfn_to_gpa(gfn_t gfn)
957 return (gpa_t)gfn << PAGE_SHIFT;
960 static inline gfn_t gpa_to_gfn(gpa_t gpa)
962 return (gfn_t)(gpa >> PAGE_SHIFT);
965 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
967 return (hpa_t)pfn << PAGE_SHIFT;
970 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
972 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
974 return kvm_is_error_hva(hva);
982 struct kvm_stat_data {
987 struct kvm_stats_debugfs_item {
990 enum kvm_stat_kind kind;
992 extern struct kvm_stats_debugfs_item debugfs_entries[];
993 extern struct dentry *kvm_debugfs_dir;
995 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
996 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
998 if (unlikely(kvm->mmu_notifier_count))
1001 * Ensure the read of mmu_notifier_count happens before the read
1002 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1003 * mmu_notifier_invalidate_range_end to make sure that the caller
1004 * either sees the old (non-zero) value of mmu_notifier_count or
1005 * the new (incremented) value of mmu_notifier_seq.
1006 * PowerPC Book3s HV KVM calls this under a per-page lock
1007 * rather than under kvm->mmu_lock, for scalability, so
1008 * can't rely on kvm->mmu_lock to keep things ordered.
1011 if (kvm->mmu_notifier_seq != mmu_seq)
1017 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1020 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1021 #elif defined(CONFIG_ARM64)
1022 #define KVM_MAX_IRQ_ROUTES 4096
1024 #define KVM_MAX_IRQ_ROUTES 1024
1027 int kvm_set_irq_routing(struct kvm *kvm,
1028 const struct kvm_irq_routing_entry *entries,
1031 int kvm_set_routing_entry(struct kvm *kvm,
1032 struct kvm_kernel_irq_routing_entry *e,
1033 const struct kvm_irq_routing_entry *ue);
1034 void kvm_free_irq_routing(struct kvm *kvm);
1038 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1042 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1044 #ifdef CONFIG_HAVE_KVM_EVENTFD
1046 void kvm_eventfd_init(struct kvm *kvm);
1047 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1049 #ifdef CONFIG_HAVE_KVM_IRQFD
1050 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1051 void kvm_irqfd_release(struct kvm *kvm);
1052 void kvm_irq_routing_update(struct kvm *);
1054 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1059 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1064 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1066 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1071 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1073 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1074 static inline void kvm_irq_routing_update(struct kvm *kvm)
1078 void kvm_arch_irq_routing_update(struct kvm *kvm);
1080 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1085 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1087 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1090 * Ensure the rest of the request is published to kvm_check_request's
1091 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1094 set_bit(req, &vcpu->requests);
1097 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1099 if (test_bit(req, &vcpu->requests)) {
1100 clear_bit(req, &vcpu->requests);
1103 * Ensure the rest of the request is visible to kvm_check_request's
1104 * caller. Paired with the smp_wmb in kvm_make_request.
1106 smp_mb__after_atomic();
1113 extern bool kvm_rebooting;
1115 extern unsigned int halt_poll_ns;
1116 extern unsigned int halt_poll_ns_grow;
1117 extern unsigned int halt_poll_ns_shrink;
1120 struct kvm_device_ops *ops;
1123 struct list_head vm_node;
1126 /* create, destroy, and name are mandatory */
1127 struct kvm_device_ops {
1131 * create is called holding kvm->lock and any operations not suitable
1132 * to do while holding the lock should be deferred to init (see
1135 int (*create)(struct kvm_device *dev, u32 type);
1138 * init is called after create if create is successful and is called
1139 * outside of holding kvm->lock.
1141 void (*init)(struct kvm_device *dev);
1144 * Destroy is responsible for freeing dev.
1146 * Destroy may be called before or after destructors are called
1147 * on emulated I/O regions, depending on whether a reference is
1148 * held by a vcpu or other kvm component that gets destroyed
1149 * after the emulated I/O.
1151 void (*destroy)(struct kvm_device *dev);
1153 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1154 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1155 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1156 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1160 void kvm_device_get(struct kvm_device *dev);
1161 void kvm_device_put(struct kvm_device *dev);
1162 struct kvm_device *kvm_device_from_filp(struct file *filp);
1163 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1164 void kvm_unregister_device_ops(u32 type);
1166 extern struct kvm_device_ops kvm_mpic_ops;
1167 extern struct kvm_device_ops kvm_xics_ops;
1168 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1169 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1171 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1173 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1175 vcpu->spin_loop.in_spin_loop = val;
1177 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1179 vcpu->spin_loop.dy_eligible = val;
1182 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1184 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1188 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1191 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1193 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1194 bool kvm_arch_has_irq_bypass(void);
1195 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1196 struct irq_bypass_producer *);
1197 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1198 struct irq_bypass_producer *);
1199 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1200 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1201 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1202 uint32_t guest_irq, bool set);
1203 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1205 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1206 /* If we wakeup during the poll time, was it a sucessful poll? */
1207 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1209 return vcpu->valid_wakeup;
1213 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1217 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */