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)
48 #define KVM_MEMSLOT_INCOHERENT (1UL << 17)
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)
119 * Architecture-independent vcpu->requests bit members
120 * Bits 4-7 are reserved for more arch-independent bits.
122 #define KVM_REQ_TLB_FLUSH 0
123 #define KVM_REQ_MMU_RELOAD 1
124 #define KVM_REQ_PENDING_TIMER 2
125 #define KVM_REQ_UNHALT 3
127 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
128 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
130 extern struct kmem_cache *kvm_vcpu_cache;
132 extern spinlock_t kvm_lock;
133 extern struct list_head vm_list;
135 struct kvm_io_range {
138 struct kvm_io_device *dev;
141 #define NR_IOBUS_DEVS 1000
146 struct kvm_io_range range[];
152 KVM_VIRTIO_CCW_NOTIFY_BUS,
157 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
158 int len, const void *val);
159 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
160 gpa_t addr, int len, const void *val, long cookie);
161 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
163 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
164 int len, struct kvm_io_device *dev);
165 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
166 struct kvm_io_device *dev);
168 #ifdef CONFIG_KVM_ASYNC_PF
169 struct kvm_async_pf {
170 struct work_struct work;
171 struct list_head link;
172 struct list_head queue;
173 struct kvm_vcpu *vcpu;
174 struct mm_struct *mm;
177 struct kvm_arch_async_pf arch;
181 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
182 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
183 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
184 struct kvm_arch_async_pf *arch);
185 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
192 READING_SHADOW_PAGE_TABLES,
196 * Sometimes a large or cross-page mmio needs to be broken up into separate
197 * exits for userspace servicing.
199 struct kvm_mmio_fragment {
207 #ifdef CONFIG_PREEMPT_NOTIFIERS
208 struct preempt_notifier preempt_notifier;
214 unsigned long requests;
215 unsigned long guest_debug;
218 struct list_head blocked_vcpu_list;
224 int guest_fpu_loaded, guest_xcr0_loaded;
225 unsigned char fpu_counter;
226 struct swait_queue_head wq;
230 struct kvm_vcpu_stat stat;
231 unsigned int halt_poll_ns;
234 #ifdef CONFIG_HAS_IOMEM
236 int mmio_read_completed;
238 int mmio_cur_fragment;
239 int mmio_nr_fragments;
240 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
243 #ifdef CONFIG_KVM_ASYNC_PF
246 struct list_head queue;
247 struct list_head done;
252 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
254 * Cpu relax intercept or pause loop exit optimization
255 * in_spin_loop: set when a vcpu does a pause loop exit
256 * or cpu relax intercepted.
257 * dy_eligible: indicates whether vcpu is eligible for directed yield.
265 struct kvm_vcpu_arch arch;
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,
319 struct kvm_s390_adapter_int adapter;
320 struct kvm_hv_sint hv_sint;
322 struct hlist_node link;
325 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
326 struct kvm_irq_routing_table {
327 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
330 * Array indexed by gsi. Each entry contains list of irq chips
331 * the gsi is connected to.
333 struct hlist_head map[0];
337 #ifndef KVM_PRIVATE_MEM_SLOTS
338 #define KVM_PRIVATE_MEM_SLOTS 0
341 #ifndef KVM_MEM_SLOTS_NUM
342 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
345 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
346 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
354 * memslots are not sorted by id anymore, please use id_to_memslot()
355 * to get the memslot by its id.
357 struct kvm_memslots {
359 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
360 /* The mapping table from slot id to the index in memslots[]. */
361 short id_to_index[KVM_MEM_SLOTS_NUM];
368 struct mutex slots_lock;
369 struct mm_struct *mm; /* userspace tied to this vm */
370 struct kvm_memslots *memslots[KVM_ADDRESS_SPACE_NUM];
371 struct srcu_struct srcu;
372 struct srcu_struct irq_srcu;
373 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
376 * created_vcpus is protected by kvm->lock, and is incremented
377 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
378 * incremented after storing the kvm_vcpu pointer in vcpus,
379 * and is accessed atomically.
381 atomic_t online_vcpus;
383 int last_boosted_vcpu;
384 struct list_head vm_list;
386 struct kvm_io_bus *buses[KVM_NR_BUSES];
387 #ifdef CONFIG_HAVE_KVM_EVENTFD
390 struct list_head items;
391 struct list_head resampler_list;
392 struct mutex resampler_lock;
394 struct list_head ioeventfds;
396 struct kvm_vm_stat stat;
397 struct kvm_arch arch;
398 atomic_t users_count;
399 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
400 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
401 spinlock_t ring_lock;
402 struct list_head coalesced_zones;
405 struct mutex irq_lock;
406 #ifdef CONFIG_HAVE_KVM_IRQCHIP
408 * Update side is protected by irq_lock.
410 struct kvm_irq_routing_table __rcu *irq_routing;
412 #ifdef CONFIG_HAVE_KVM_IRQFD
413 struct hlist_head irq_ack_notifier_list;
416 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
417 struct mmu_notifier mmu_notifier;
418 unsigned long mmu_notifier_seq;
419 long mmu_notifier_count;
422 struct list_head devices;
423 struct dentry *debugfs_dentry;
424 struct kvm_stat_data **debugfs_stat_data;
427 #define kvm_err(fmt, ...) \
428 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
429 #define kvm_info(fmt, ...) \
430 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
431 #define kvm_debug(fmt, ...) \
432 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
433 #define kvm_pr_unimpl(fmt, ...) \
434 pr_err_ratelimited("kvm [%i]: " fmt, \
435 task_tgid_nr(current), ## __VA_ARGS__)
437 /* The guest did something we don't support. */
438 #define vcpu_unimpl(vcpu, fmt, ...) \
439 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
440 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
442 #define vcpu_debug(vcpu, fmt, ...) \
443 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
444 #define vcpu_err(vcpu, fmt, ...) \
445 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
447 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
449 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
450 * the caller has read kvm->online_vcpus before (as is the case
451 * for kvm_for_each_vcpu, for example).
454 return kvm->vcpus[i];
457 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
459 idx < atomic_read(&kvm->online_vcpus) && \
460 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
463 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
465 struct kvm_vcpu *vcpu = NULL;
470 if (id < KVM_MAX_VCPUS)
471 vcpu = kvm_get_vcpu(kvm, id);
472 if (vcpu && vcpu->vcpu_id == id)
474 kvm_for_each_vcpu(i, vcpu, kvm)
475 if (vcpu->vcpu_id == id)
480 #define kvm_for_each_memslot(memslot, slots) \
481 for (memslot = &slots->memslots[0]; \
482 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
485 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
486 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
488 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
489 void vcpu_put(struct kvm_vcpu *vcpu);
491 #ifdef __KVM_HAVE_IOAPIC
492 void kvm_vcpu_request_scan_ioapic(struct kvm *kvm);
493 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
495 static inline void kvm_vcpu_request_scan_ioapic(struct kvm *kvm)
498 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
503 #ifdef CONFIG_HAVE_KVM_IRQFD
504 int kvm_irqfd_init(void);
505 void kvm_irqfd_exit(void);
507 static inline int kvm_irqfd_init(void)
512 static inline void kvm_irqfd_exit(void)
516 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
517 struct module *module);
520 void kvm_get_kvm(struct kvm *kvm);
521 void kvm_put_kvm(struct kvm *kvm);
523 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
525 return rcu_dereference_check(kvm->memslots[as_id],
526 srcu_read_lock_held(&kvm->srcu)
527 || lockdep_is_held(&kvm->slots_lock));
530 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
532 return __kvm_memslots(kvm, 0);
535 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
537 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
539 return __kvm_memslots(vcpu->kvm, as_id);
542 static inline struct kvm_memory_slot *
543 id_to_memslot(struct kvm_memslots *slots, int id)
545 int index = slots->id_to_index[id];
546 struct kvm_memory_slot *slot;
548 slot = &slots->memslots[index];
550 WARN_ON(slot->id != id);
555 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
556 * - create a new memory slot
557 * - delete an existing memory slot
558 * - modify an existing memory slot
559 * -- move it in the guest physical memory space
560 * -- just change its flags
562 * Since flags can be changed by some of these operations, the following
563 * differentiation is the best we can do for __kvm_set_memory_region():
572 int kvm_set_memory_region(struct kvm *kvm,
573 const struct kvm_userspace_memory_region *mem);
574 int __kvm_set_memory_region(struct kvm *kvm,
575 const struct kvm_userspace_memory_region *mem);
576 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
577 struct kvm_memory_slot *dont);
578 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
579 unsigned long npages);
580 void kvm_arch_memslots_updated(struct kvm *kvm, struct kvm_memslots *slots);
581 int kvm_arch_prepare_memory_region(struct kvm *kvm,
582 struct kvm_memory_slot *memslot,
583 const struct kvm_userspace_memory_region *mem,
584 enum kvm_mr_change change);
585 void kvm_arch_commit_memory_region(struct kvm *kvm,
586 const struct kvm_userspace_memory_region *mem,
587 const struct kvm_memory_slot *old,
588 const struct kvm_memory_slot *new,
589 enum kvm_mr_change change);
590 bool kvm_largepages_enabled(void);
591 void kvm_disable_largepages(void);
592 /* flush all memory translations */
593 void kvm_arch_flush_shadow_all(struct kvm *kvm);
594 /* flush memory translations pointing to 'slot' */
595 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
596 struct kvm_memory_slot *slot);
598 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
599 struct page **pages, int nr_pages);
601 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
602 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
603 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
604 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
605 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
607 void kvm_release_page_clean(struct page *page);
608 void kvm_release_page_dirty(struct page *page);
609 void kvm_set_page_accessed(struct page *page);
611 kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
612 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
613 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
615 kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
616 kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
617 kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
618 bool atomic, bool *async, bool write_fault,
621 void kvm_release_pfn_clean(kvm_pfn_t pfn);
622 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
623 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
624 void kvm_get_pfn(kvm_pfn_t pfn);
626 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
628 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
630 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
631 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
632 void *data, unsigned long len);
633 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
634 int offset, int len);
635 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
637 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
638 void *data, unsigned long len);
639 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
640 gpa_t gpa, unsigned long len);
641 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
642 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
643 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
644 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
645 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
646 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
648 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
649 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
650 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
651 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
652 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
653 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
654 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
655 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
657 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
659 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
661 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
662 int offset, int len);
663 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
665 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
667 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
668 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
669 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
670 void kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
671 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
672 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
673 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
674 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
675 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
677 void kvm_flush_remote_tlbs(struct kvm *kvm);
678 void kvm_reload_remote_mmus(struct kvm *kvm);
679 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
681 long kvm_arch_dev_ioctl(struct file *filp,
682 unsigned int ioctl, unsigned long arg);
683 long kvm_arch_vcpu_ioctl(struct file *filp,
684 unsigned int ioctl, unsigned long arg);
685 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
687 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
689 int kvm_get_dirty_log(struct kvm *kvm,
690 struct kvm_dirty_log *log, int *is_dirty);
692 int kvm_get_dirty_log_protect(struct kvm *kvm,
693 struct kvm_dirty_log *log, bool *is_dirty);
695 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
696 struct kvm_memory_slot *slot,
700 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
701 struct kvm_dirty_log *log);
703 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
705 long kvm_arch_vm_ioctl(struct file *filp,
706 unsigned int ioctl, unsigned long arg);
708 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
709 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
711 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
712 struct kvm_translation *tr);
714 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
715 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
716 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
717 struct kvm_sregs *sregs);
718 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
719 struct kvm_sregs *sregs);
720 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
721 struct kvm_mp_state *mp_state);
722 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
723 struct kvm_mp_state *mp_state);
724 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
725 struct kvm_guest_debug *dbg);
726 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
728 int kvm_arch_init(void *opaque);
729 void kvm_arch_exit(void);
731 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
732 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
734 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
736 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
737 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
738 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
739 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
740 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
741 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
742 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
744 int kvm_arch_hardware_enable(void);
745 void kvm_arch_hardware_disable(void);
746 int kvm_arch_hardware_setup(void);
747 void kvm_arch_hardware_unsetup(void);
748 void kvm_arch_check_processor_compat(void *rtn);
749 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
750 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
752 void *kvm_kvzalloc(unsigned long size);
754 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
755 static inline struct kvm *kvm_arch_alloc_vm(void)
757 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
760 static inline void kvm_arch_free_vm(struct kvm *kvm)
766 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
767 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
768 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
769 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
771 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
775 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
779 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
784 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
785 void kvm_arch_start_assignment(struct kvm *kvm);
786 void kvm_arch_end_assignment(struct kvm *kvm);
787 bool kvm_arch_has_assigned_device(struct kvm *kvm);
789 static inline void kvm_arch_start_assignment(struct kvm *kvm)
793 static inline void kvm_arch_end_assignment(struct kvm *kvm)
797 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
803 static inline struct swait_queue_head *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
805 #ifdef __KVM_HAVE_ARCH_WQP
806 return vcpu->arch.wqp;
812 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
814 * returns true if the virtual interrupt controller is initialized and
815 * ready to accept virtual IRQ. On some architectures the virtual interrupt
816 * controller is dynamically instantiated and this is not always true.
818 bool kvm_arch_intc_initialized(struct kvm *kvm);
820 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
826 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
827 void kvm_arch_destroy_vm(struct kvm *kvm);
828 void kvm_arch_sync_events(struct kvm *kvm);
830 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
831 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
833 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
835 struct kvm_irq_ack_notifier {
836 struct hlist_node link;
838 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
841 int kvm_irq_map_gsi(struct kvm *kvm,
842 struct kvm_kernel_irq_routing_entry *entries, int gsi);
843 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
845 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
847 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
848 int irq_source_id, int level, bool line_status);
849 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
850 struct kvm *kvm, int irq_source_id,
851 int level, bool line_status);
852 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
853 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
854 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
855 void kvm_register_irq_ack_notifier(struct kvm *kvm,
856 struct kvm_irq_ack_notifier *kian);
857 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
858 struct kvm_irq_ack_notifier *kian);
859 int kvm_request_irq_source_id(struct kvm *kvm);
860 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
862 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
863 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
864 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
866 static inline int kvm_iommu_map_pages(struct kvm *kvm,
867 struct kvm_memory_slot *slot)
872 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
873 struct kvm_memory_slot *slot)
878 /* must be called with irqs disabled */
879 static inline void __kvm_guest_enter(void)
882 /* KVM does not hold any references to rcu protected data when it
883 * switches CPU into a guest mode. In fact switching to a guest mode
884 * is very similar to exiting to userspace from rcu point of view. In
885 * addition CPU may stay in a guest mode for quite a long time (up to
886 * one time slice). Lets treat guest mode as quiescent state, just like
887 * we do with user-mode execution.
889 if (!context_tracking_cpu_is_enabled())
890 rcu_virt_note_context_switch(smp_processor_id());
893 /* must be called with irqs disabled */
894 static inline void __kvm_guest_exit(void)
899 static inline void kvm_guest_enter(void)
903 local_irq_save(flags);
905 local_irq_restore(flags);
908 static inline void kvm_guest_exit(void)
912 local_irq_save(flags);
914 local_irq_restore(flags);
918 * search_memslots() and __gfn_to_memslot() are here because they are
919 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
920 * gfn_to_memslot() itself isn't here as an inline because that would
921 * bloat other code too much.
923 static inline struct kvm_memory_slot *
924 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
926 int start = 0, end = slots->used_slots;
927 int slot = atomic_read(&slots->lru_slot);
928 struct kvm_memory_slot *memslots = slots->memslots;
930 if (gfn >= memslots[slot].base_gfn &&
931 gfn < memslots[slot].base_gfn + memslots[slot].npages)
932 return &memslots[slot];
934 while (start < end) {
935 slot = start + (end - start) / 2;
937 if (gfn >= memslots[slot].base_gfn)
943 if (gfn >= memslots[start].base_gfn &&
944 gfn < memslots[start].base_gfn + memslots[start].npages) {
945 atomic_set(&slots->lru_slot, start);
946 return &memslots[start];
952 static inline struct kvm_memory_slot *
953 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
955 return search_memslots(slots, gfn);
958 static inline unsigned long
959 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
961 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
964 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
966 return gfn_to_memslot(kvm, gfn)->id;
970 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
972 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
974 return slot->base_gfn + gfn_offset;
977 static inline gpa_t gfn_to_gpa(gfn_t gfn)
979 return (gpa_t)gfn << PAGE_SHIFT;
982 static inline gfn_t gpa_to_gfn(gpa_t gpa)
984 return (gfn_t)(gpa >> PAGE_SHIFT);
987 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
989 return (hpa_t)pfn << PAGE_SHIFT;
992 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
994 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
996 return kvm_is_error_hva(hva);
1004 struct kvm_stat_data {
1009 struct kvm_stats_debugfs_item {
1012 enum kvm_stat_kind kind;
1014 extern struct kvm_stats_debugfs_item debugfs_entries[];
1015 extern struct dentry *kvm_debugfs_dir;
1017 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1018 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1020 if (unlikely(kvm->mmu_notifier_count))
1023 * Ensure the read of mmu_notifier_count happens before the read
1024 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1025 * mmu_notifier_invalidate_range_end to make sure that the caller
1026 * either sees the old (non-zero) value of mmu_notifier_count or
1027 * the new (incremented) value of mmu_notifier_seq.
1028 * PowerPC Book3s HV KVM calls this under a per-page lock
1029 * rather than under kvm->mmu_lock, for scalability, so
1030 * can't rely on kvm->mmu_lock to keep things ordered.
1033 if (kvm->mmu_notifier_seq != mmu_seq)
1039 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1042 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1044 #define KVM_MAX_IRQ_ROUTES 1024
1047 int kvm_setup_default_irq_routing(struct kvm *kvm);
1048 int kvm_setup_empty_irq_routing(struct kvm *kvm);
1049 int kvm_set_irq_routing(struct kvm *kvm,
1050 const struct kvm_irq_routing_entry *entries,
1053 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
1054 const struct kvm_irq_routing_entry *ue);
1055 void kvm_free_irq_routing(struct kvm *kvm);
1059 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1063 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1065 #ifdef CONFIG_HAVE_KVM_EVENTFD
1067 void kvm_eventfd_init(struct kvm *kvm);
1068 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1070 #ifdef CONFIG_HAVE_KVM_IRQFD
1071 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1072 void kvm_irqfd_release(struct kvm *kvm);
1073 void kvm_irq_routing_update(struct kvm *);
1075 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1080 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1085 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1087 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1092 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1094 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1095 static inline void kvm_irq_routing_update(struct kvm *kvm)
1099 void kvm_arch_irq_routing_update(struct kvm *kvm);
1101 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1106 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1108 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1111 * Ensure the rest of the request is published to kvm_check_request's
1112 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1115 set_bit(req, &vcpu->requests);
1118 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1120 if (test_bit(req, &vcpu->requests)) {
1121 clear_bit(req, &vcpu->requests);
1124 * Ensure the rest of the request is visible to kvm_check_request's
1125 * caller. Paired with the smp_wmb in kvm_make_request.
1127 smp_mb__after_atomic();
1134 extern bool kvm_rebooting;
1137 struct kvm_device_ops *ops;
1140 struct list_head vm_node;
1143 /* create, destroy, and name are mandatory */
1144 struct kvm_device_ops {
1146 int (*create)(struct kvm_device *dev, u32 type);
1149 * Destroy is responsible for freeing dev.
1151 * Destroy may be called before or after destructors are called
1152 * on emulated I/O regions, depending on whether a reference is
1153 * held by a vcpu or other kvm component that gets destroyed
1154 * after the emulated I/O.
1156 void (*destroy)(struct kvm_device *dev);
1158 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1159 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1160 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1161 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1165 void kvm_device_get(struct kvm_device *dev);
1166 void kvm_device_put(struct kvm_device *dev);
1167 struct kvm_device *kvm_device_from_filp(struct file *filp);
1168 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1169 void kvm_unregister_device_ops(u32 type);
1171 extern struct kvm_device_ops kvm_mpic_ops;
1172 extern struct kvm_device_ops kvm_xics_ops;
1173 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1174 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1176 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1178 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1180 vcpu->spin_loop.in_spin_loop = val;
1182 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1184 vcpu->spin_loop.dy_eligible = val;
1187 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1189 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1193 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1196 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1198 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1199 bool kvm_arch_has_irq_bypass(void);
1200 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1201 struct irq_bypass_producer *);
1202 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1203 struct irq_bypass_producer *);
1204 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1205 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1206 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1207 uint32_t guest_irq, bool set);
1208 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1210 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1211 /* If we wakeup during the poll time, was it a sucessful poll? */
1212 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1214 return vcpu->valid_wakeup;
1218 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1222 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */