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 <asm/signal.h>
29 #include <linux/kvm.h>
30 #include <linux/kvm_para.h>
32 #include <linux/kvm_types.h>
34 #include <asm/kvm_host.h>
37 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
38 * in kvm, other bits are visible for userspace which are defined in
39 * include/linux/kvm_h.
41 #define KVM_MEMSLOT_INVALID (1UL << 16)
42 #define KVM_MEMSLOT_INCOHERENT (1UL << 17)
44 /* Two fragments for cross MMIO pages. */
45 #define KVM_MAX_MMIO_FRAGMENTS 2
47 #ifndef KVM_ADDRESS_SPACE_NUM
48 #define KVM_ADDRESS_SPACE_NUM 1
52 * For the normal pfn, the highest 12 bits should be zero,
53 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
54 * mask bit 63 to indicate the noslot pfn.
56 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
57 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
58 #define KVM_PFN_NOSLOT (0x1ULL << 63)
60 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
61 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
62 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
65 * error pfns indicate that the gfn is in slot but faild to
66 * translate it to pfn on host.
68 static inline bool is_error_pfn(pfn_t pfn)
70 return !!(pfn & KVM_PFN_ERR_MASK);
74 * error_noslot pfns indicate that the gfn can not be
75 * translated to pfn - it is not in slot or failed to
76 * translate it to pfn.
78 static inline bool is_error_noslot_pfn(pfn_t pfn)
80 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
83 /* noslot pfn indicates that the gfn is not in slot. */
84 static inline bool is_noslot_pfn(pfn_t pfn)
86 return pfn == KVM_PFN_NOSLOT;
90 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
91 * provide own defines and kvm_is_error_hva
93 #ifndef KVM_HVA_ERR_BAD
95 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
96 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
98 static inline bool kvm_is_error_hva(unsigned long addr)
100 return addr >= PAGE_OFFSET;
105 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
107 static inline bool is_error_page(struct page *page)
113 * vcpu->requests bit members
115 #define KVM_REQ_TLB_FLUSH 0
116 #define KVM_REQ_MIGRATE_TIMER 1
117 #define KVM_REQ_REPORT_TPR_ACCESS 2
118 #define KVM_REQ_MMU_RELOAD 3
119 #define KVM_REQ_TRIPLE_FAULT 4
120 #define KVM_REQ_PENDING_TIMER 5
121 #define KVM_REQ_UNHALT 6
122 #define KVM_REQ_MMU_SYNC 7
123 #define KVM_REQ_CLOCK_UPDATE 8
124 #define KVM_REQ_KICK 9
125 #define KVM_REQ_DEACTIVATE_FPU 10
126 #define KVM_REQ_EVENT 11
127 #define KVM_REQ_APF_HALT 12
128 #define KVM_REQ_STEAL_UPDATE 13
129 #define KVM_REQ_NMI 14
130 #define KVM_REQ_PMU 15
131 #define KVM_REQ_PMI 16
132 #define KVM_REQ_WATCHDOG 17
133 #define KVM_REQ_MASTERCLOCK_UPDATE 18
134 #define KVM_REQ_MCLOCK_INPROGRESS 19
135 #define KVM_REQ_EPR_EXIT 20
136 #define KVM_REQ_SCAN_IOAPIC 21
137 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 22
138 #define KVM_REQ_ENABLE_IBS 23
139 #define KVM_REQ_DISABLE_IBS 24
140 #define KVM_REQ_APIC_PAGE_RELOAD 25
141 #define KVM_REQ_SMI 26
142 #define KVM_REQ_HV_CRASH 27
143 #define KVM_REQ_IOAPIC_EOI_EXIT 28
144 #define KVM_REQ_HV_RESET 29
146 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
147 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
149 extern struct kmem_cache *kvm_vcpu_cache;
151 extern spinlock_t kvm_lock;
152 extern struct list_head vm_list;
154 struct kvm_io_range {
157 struct kvm_io_device *dev;
160 #define NR_IOBUS_DEVS 1000
165 struct kvm_io_range range[];
171 KVM_VIRTIO_CCW_NOTIFY_BUS,
176 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
177 int len, const void *val);
178 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
179 gpa_t addr, int len, const void *val, long cookie);
180 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
182 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
183 int len, struct kvm_io_device *dev);
184 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
185 struct kvm_io_device *dev);
187 #ifdef CONFIG_KVM_ASYNC_PF
188 struct kvm_async_pf {
189 struct work_struct work;
190 struct list_head link;
191 struct list_head queue;
192 struct kvm_vcpu *vcpu;
193 struct mm_struct *mm;
196 struct kvm_arch_async_pf arch;
200 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
201 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
202 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
203 struct kvm_arch_async_pf *arch);
204 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
211 READING_SHADOW_PAGE_TABLES,
215 * Sometimes a large or cross-page mmio needs to be broken up into separate
216 * exits for userspace servicing.
218 struct kvm_mmio_fragment {
226 #ifdef CONFIG_PREEMPT_NOTIFIERS
227 struct preempt_notifier preempt_notifier;
233 unsigned long requests;
234 unsigned long guest_debug;
240 int guest_fpu_loaded, guest_xcr0_loaded;
241 unsigned char fpu_counter;
242 wait_queue_head_t wq;
246 struct kvm_vcpu_stat stat;
247 unsigned int halt_poll_ns;
249 #ifdef CONFIG_HAS_IOMEM
251 int mmio_read_completed;
253 int mmio_cur_fragment;
254 int mmio_nr_fragments;
255 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
258 #ifdef CONFIG_KVM_ASYNC_PF
261 struct list_head queue;
262 struct list_head done;
267 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
269 * Cpu relax intercept or pause loop exit optimization
270 * in_spin_loop: set when a vcpu does a pause loop exit
271 * or cpu relax intercepted.
272 * dy_eligible: indicates whether vcpu is eligible for directed yield.
280 struct kvm_vcpu_arch arch;
283 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
285 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
289 * Some of the bitops functions do not support too long bitmaps.
290 * This number must be determined not to exceed such limits.
292 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
294 struct kvm_memory_slot {
296 unsigned long npages;
297 unsigned long *dirty_bitmap;
298 struct kvm_arch_memory_slot arch;
299 unsigned long userspace_addr;
304 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
306 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
309 struct kvm_s390_adapter_int {
317 struct kvm_kernel_irq_routing_entry {
320 int (*set)(struct kvm_kernel_irq_routing_entry *e,
321 struct kvm *kvm, int irq_source_id, int level,
329 struct kvm_s390_adapter_int adapter;
331 struct hlist_node link;
334 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
335 struct kvm_irq_routing_table {
336 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
339 * Array indexed by gsi. Each entry contains list of irq chips
340 * the gsi is connected to.
342 struct hlist_head map[0];
346 #ifndef KVM_PRIVATE_MEM_SLOTS
347 #define KVM_PRIVATE_MEM_SLOTS 0
350 #ifndef KVM_MEM_SLOTS_NUM
351 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
354 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
355 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
363 * memslots are not sorted by id anymore, please use id_to_memslot()
364 * to get the memslot by its id.
366 struct kvm_memslots {
368 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
369 /* The mapping table from slot id to the index in memslots[]. */
370 short id_to_index[KVM_MEM_SLOTS_NUM];
377 struct mutex slots_lock;
378 struct mm_struct *mm; /* userspace tied to this vm */
379 struct kvm_memslots *memslots[KVM_ADDRESS_SPACE_NUM];
380 struct srcu_struct srcu;
381 struct srcu_struct irq_srcu;
382 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
383 atomic_t online_vcpus;
384 int last_boosted_vcpu;
385 struct list_head vm_list;
387 struct kvm_io_bus *buses[KVM_NR_BUSES];
388 #ifdef CONFIG_HAVE_KVM_EVENTFD
391 struct list_head items;
392 struct list_head resampler_list;
393 struct mutex resampler_lock;
395 struct list_head ioeventfds;
397 struct kvm_vm_stat stat;
398 struct kvm_arch arch;
399 atomic_t users_count;
400 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
401 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
402 spinlock_t ring_lock;
403 struct list_head coalesced_zones;
406 struct mutex irq_lock;
407 #ifdef CONFIG_HAVE_KVM_IRQCHIP
409 * Update side is protected by irq_lock.
411 struct kvm_irq_routing_table __rcu *irq_routing;
413 #ifdef CONFIG_HAVE_KVM_IRQFD
414 struct hlist_head irq_ack_notifier_list;
417 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
418 struct mmu_notifier mmu_notifier;
419 unsigned long mmu_notifier_seq;
420 long mmu_notifier_count;
423 struct list_head devices;
426 #define kvm_err(fmt, ...) \
427 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
428 #define kvm_info(fmt, ...) \
429 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
430 #define kvm_debug(fmt, ...) \
431 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
432 #define kvm_pr_unimpl(fmt, ...) \
433 pr_err_ratelimited("kvm [%i]: " fmt, \
434 task_tgid_nr(current), ## __VA_ARGS__)
436 /* The guest did something we don't support. */
437 #define vcpu_unimpl(vcpu, fmt, ...) \
438 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
440 #define vcpu_debug(vcpu, fmt, ...) \
441 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
443 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
445 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
446 * the caller has read kvm->online_vcpus before (as is the case
447 * for kvm_for_each_vcpu, for example).
450 return kvm->vcpus[i];
453 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
455 idx < atomic_read(&kvm->online_vcpus) && \
456 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
459 #define kvm_for_each_memslot(memslot, slots) \
460 for (memslot = &slots->memslots[0]; \
461 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
464 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
465 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
467 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
468 void vcpu_put(struct kvm_vcpu *vcpu);
470 #ifdef __KVM_HAVE_IOAPIC
471 void kvm_vcpu_request_scan_ioapic(struct kvm *kvm);
472 void kvm_arch_irq_routing_update(struct kvm *kvm);
474 static inline void kvm_vcpu_request_scan_ioapic(struct kvm *kvm)
477 static inline void kvm_arch_irq_routing_update(struct kvm *kvm)
482 #ifdef CONFIG_HAVE_KVM_IRQFD
483 int kvm_irqfd_init(void);
484 void kvm_irqfd_exit(void);
486 static inline int kvm_irqfd_init(void)
491 static inline void kvm_irqfd_exit(void)
495 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
496 struct module *module);
499 void kvm_get_kvm(struct kvm *kvm);
500 void kvm_put_kvm(struct kvm *kvm);
502 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
504 return rcu_dereference_check(kvm->memslots[as_id],
505 srcu_read_lock_held(&kvm->srcu)
506 || lockdep_is_held(&kvm->slots_lock));
509 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
511 return __kvm_memslots(kvm, 0);
514 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
516 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
518 return __kvm_memslots(vcpu->kvm, as_id);
521 static inline struct kvm_memory_slot *
522 id_to_memslot(struct kvm_memslots *slots, int id)
524 int index = slots->id_to_index[id];
525 struct kvm_memory_slot *slot;
527 slot = &slots->memslots[index];
529 WARN_ON(slot->id != id);
534 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
535 * - create a new memory slot
536 * - delete an existing memory slot
537 * - modify an existing memory slot
538 * -- move it in the guest physical memory space
539 * -- just change its flags
541 * Since flags can be changed by some of these operations, the following
542 * differentiation is the best we can do for __kvm_set_memory_region():
551 int kvm_set_memory_region(struct kvm *kvm,
552 const struct kvm_userspace_memory_region *mem);
553 int __kvm_set_memory_region(struct kvm *kvm,
554 const struct kvm_userspace_memory_region *mem);
555 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
556 struct kvm_memory_slot *dont);
557 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
558 unsigned long npages);
559 void kvm_arch_memslots_updated(struct kvm *kvm, struct kvm_memslots *slots);
560 int kvm_arch_prepare_memory_region(struct kvm *kvm,
561 struct kvm_memory_slot *memslot,
562 const struct kvm_userspace_memory_region *mem,
563 enum kvm_mr_change change);
564 void kvm_arch_commit_memory_region(struct kvm *kvm,
565 const struct kvm_userspace_memory_region *mem,
566 const struct kvm_memory_slot *old,
567 const struct kvm_memory_slot *new,
568 enum kvm_mr_change change);
569 bool kvm_largepages_enabled(void);
570 void kvm_disable_largepages(void);
571 /* flush all memory translations */
572 void kvm_arch_flush_shadow_all(struct kvm *kvm);
573 /* flush memory translations pointing to 'slot' */
574 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
575 struct kvm_memory_slot *slot);
577 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
578 struct page **pages, int nr_pages);
580 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
581 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
582 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
583 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
584 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
586 void kvm_release_page_clean(struct page *page);
587 void kvm_release_page_dirty(struct page *page);
588 void kvm_set_page_accessed(struct page *page);
590 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
591 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
592 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
594 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
595 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
596 pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn, bool atomic,
597 bool *async, bool write_fault, bool *writable);
599 void kvm_release_pfn_clean(pfn_t pfn);
600 void kvm_set_pfn_dirty(pfn_t pfn);
601 void kvm_set_pfn_accessed(pfn_t pfn);
602 void kvm_get_pfn(pfn_t pfn);
604 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
606 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
608 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
609 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
610 void *data, unsigned long len);
611 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
612 int offset, int len);
613 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
615 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
616 void *data, unsigned long len);
617 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
618 gpa_t gpa, unsigned long len);
619 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
620 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
621 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
622 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
623 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
624 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
626 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
627 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
628 pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
629 pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
630 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
631 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
632 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
633 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
635 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
637 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
639 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
640 int offset, int len);
641 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
643 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
645 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
646 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
647 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
648 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
649 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
650 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
652 void kvm_flush_remote_tlbs(struct kvm *kvm);
653 void kvm_reload_remote_mmus(struct kvm *kvm);
654 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
655 void kvm_make_scan_ioapic_request(struct kvm *kvm);
656 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
658 long kvm_arch_dev_ioctl(struct file *filp,
659 unsigned int ioctl, unsigned long arg);
660 long kvm_arch_vcpu_ioctl(struct file *filp,
661 unsigned int ioctl, unsigned long arg);
662 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
664 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
666 int kvm_get_dirty_log(struct kvm *kvm,
667 struct kvm_dirty_log *log, int *is_dirty);
669 int kvm_get_dirty_log_protect(struct kvm *kvm,
670 struct kvm_dirty_log *log, bool *is_dirty);
672 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
673 struct kvm_memory_slot *slot,
677 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
678 struct kvm_dirty_log *log);
680 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
682 long kvm_arch_vm_ioctl(struct file *filp,
683 unsigned int ioctl, unsigned long arg);
685 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
686 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
688 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
689 struct kvm_translation *tr);
691 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
692 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
693 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
694 struct kvm_sregs *sregs);
695 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
696 struct kvm_sregs *sregs);
697 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
698 struct kvm_mp_state *mp_state);
699 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
700 struct kvm_mp_state *mp_state);
701 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
702 struct kvm_guest_debug *dbg);
703 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
705 int kvm_arch_init(void *opaque);
706 void kvm_arch_exit(void);
708 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
709 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
711 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
713 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
714 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
715 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
716 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
717 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
718 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
719 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
721 int kvm_arch_hardware_enable(void);
722 void kvm_arch_hardware_disable(void);
723 int kvm_arch_hardware_setup(void);
724 void kvm_arch_hardware_unsetup(void);
725 void kvm_arch_check_processor_compat(void *rtn);
726 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
727 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
729 void *kvm_kvzalloc(unsigned long size);
731 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
732 static inline struct kvm *kvm_arch_alloc_vm(void)
734 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
737 static inline void kvm_arch_free_vm(struct kvm *kvm)
743 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
744 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
745 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
746 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
748 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
752 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
756 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
761 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
762 void kvm_arch_start_assignment(struct kvm *kvm);
763 void kvm_arch_end_assignment(struct kvm *kvm);
764 bool kvm_arch_has_assigned_device(struct kvm *kvm);
766 static inline void kvm_arch_start_assignment(struct kvm *kvm)
770 static inline void kvm_arch_end_assignment(struct kvm *kvm)
774 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
780 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
782 #ifdef __KVM_HAVE_ARCH_WQP
783 return vcpu->arch.wqp;
789 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
791 * returns true if the virtual interrupt controller is initialized and
792 * ready to accept virtual IRQ. On some architectures the virtual interrupt
793 * controller is dynamically instantiated and this is not always true.
795 bool kvm_arch_intc_initialized(struct kvm *kvm);
797 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
803 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
804 void kvm_arch_destroy_vm(struct kvm *kvm);
805 void kvm_arch_sync_events(struct kvm *kvm);
807 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
808 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
810 bool kvm_is_reserved_pfn(pfn_t pfn);
812 struct kvm_irq_ack_notifier {
813 struct hlist_node link;
815 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
818 int kvm_irq_map_gsi(struct kvm *kvm,
819 struct kvm_kernel_irq_routing_entry *entries, int gsi);
820 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
822 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
824 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
825 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
826 int irq_source_id, int level, bool line_status);
827 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
828 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
829 void kvm_register_irq_ack_notifier(struct kvm *kvm,
830 struct kvm_irq_ack_notifier *kian);
831 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
832 struct kvm_irq_ack_notifier *kian);
833 int kvm_request_irq_source_id(struct kvm *kvm);
834 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
836 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
837 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
838 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
840 static inline int kvm_iommu_map_pages(struct kvm *kvm,
841 struct kvm_memory_slot *slot)
846 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
847 struct kvm_memory_slot *slot)
852 /* must be called with irqs disabled */
853 static inline void __kvm_guest_enter(void)
856 /* KVM does not hold any references to rcu protected data when it
857 * switches CPU into a guest mode. In fact switching to a guest mode
858 * is very similar to exiting to userspace from rcu point of view. In
859 * addition CPU may stay in a guest mode for quite a long time (up to
860 * one time slice). Lets treat guest mode as quiescent state, just like
861 * we do with user-mode execution.
863 if (!context_tracking_cpu_is_enabled())
864 rcu_virt_note_context_switch(smp_processor_id());
867 /* must be called with irqs disabled */
868 static inline void __kvm_guest_exit(void)
873 static inline void kvm_guest_enter(void)
877 local_irq_save(flags);
879 local_irq_restore(flags);
882 static inline void kvm_guest_exit(void)
886 local_irq_save(flags);
888 local_irq_restore(flags);
892 * search_memslots() and __gfn_to_memslot() are here because they are
893 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
894 * gfn_to_memslot() itself isn't here as an inline because that would
895 * bloat other code too much.
897 static inline struct kvm_memory_slot *
898 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
900 int start = 0, end = slots->used_slots;
901 int slot = atomic_read(&slots->lru_slot);
902 struct kvm_memory_slot *memslots = slots->memslots;
904 if (gfn >= memslots[slot].base_gfn &&
905 gfn < memslots[slot].base_gfn + memslots[slot].npages)
906 return &memslots[slot];
908 while (start < end) {
909 slot = start + (end - start) / 2;
911 if (gfn >= memslots[slot].base_gfn)
917 if (gfn >= memslots[start].base_gfn &&
918 gfn < memslots[start].base_gfn + memslots[start].npages) {
919 atomic_set(&slots->lru_slot, start);
920 return &memslots[start];
926 static inline struct kvm_memory_slot *
927 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
929 return search_memslots(slots, gfn);
932 static inline unsigned long
933 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
935 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
938 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
940 return gfn_to_memslot(kvm, gfn)->id;
944 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
946 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
948 return slot->base_gfn + gfn_offset;
951 static inline gpa_t gfn_to_gpa(gfn_t gfn)
953 return (gpa_t)gfn << PAGE_SHIFT;
956 static inline gfn_t gpa_to_gfn(gpa_t gpa)
958 return (gfn_t)(gpa >> PAGE_SHIFT);
961 static inline hpa_t pfn_to_hpa(pfn_t pfn)
963 return (hpa_t)pfn << PAGE_SHIFT;
966 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
968 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
970 return kvm_is_error_hva(hva);
973 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
975 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
983 struct kvm_stats_debugfs_item {
986 enum kvm_stat_kind kind;
987 struct dentry *dentry;
989 extern struct kvm_stats_debugfs_item debugfs_entries[];
990 extern struct dentry *kvm_debugfs_dir;
992 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
993 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
995 if (unlikely(kvm->mmu_notifier_count))
998 * Ensure the read of mmu_notifier_count happens before the read
999 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1000 * mmu_notifier_invalidate_range_end to make sure that the caller
1001 * either sees the old (non-zero) value of mmu_notifier_count or
1002 * the new (incremented) value of mmu_notifier_seq.
1003 * PowerPC Book3s HV KVM calls this under a per-page lock
1004 * rather than under kvm->mmu_lock, for scalability, so
1005 * can't rely on kvm->mmu_lock to keep things ordered.
1008 if (kvm->mmu_notifier_seq != mmu_seq)
1014 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1017 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1019 #define KVM_MAX_IRQ_ROUTES 1024
1022 int kvm_setup_default_irq_routing(struct kvm *kvm);
1023 int kvm_setup_empty_irq_routing(struct kvm *kvm);
1024 int kvm_set_irq_routing(struct kvm *kvm,
1025 const struct kvm_irq_routing_entry *entries,
1028 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
1029 const struct kvm_irq_routing_entry *ue);
1030 void kvm_free_irq_routing(struct kvm *kvm);
1034 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1038 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1040 #ifdef CONFIG_HAVE_KVM_EVENTFD
1042 void kvm_eventfd_init(struct kvm *kvm);
1043 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1045 #ifdef CONFIG_HAVE_KVM_IRQFD
1046 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1047 void kvm_irqfd_release(struct kvm *kvm);
1048 void kvm_irq_routing_update(struct kvm *);
1050 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1055 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1060 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1062 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1067 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1069 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1070 static inline void kvm_irq_routing_update(struct kvm *kvm)
1075 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1080 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1082 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
1083 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
1085 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
1088 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1090 set_bit(req, &vcpu->requests);
1093 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1095 if (test_bit(req, &vcpu->requests)) {
1096 clear_bit(req, &vcpu->requests);
1103 extern bool kvm_rebooting;
1106 struct kvm_device_ops *ops;
1109 struct list_head vm_node;
1112 /* create, destroy, and name are mandatory */
1113 struct kvm_device_ops {
1115 int (*create)(struct kvm_device *dev, u32 type);
1118 * Destroy is responsible for freeing dev.
1120 * Destroy may be called before or after destructors are called
1121 * on emulated I/O regions, depending on whether a reference is
1122 * held by a vcpu or other kvm component that gets destroyed
1123 * after the emulated I/O.
1125 void (*destroy)(struct kvm_device *dev);
1127 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1128 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1129 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1130 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1134 void kvm_device_get(struct kvm_device *dev);
1135 void kvm_device_put(struct kvm_device *dev);
1136 struct kvm_device *kvm_device_from_filp(struct file *filp);
1137 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1138 void kvm_unregister_device_ops(u32 type);
1140 extern struct kvm_device_ops kvm_mpic_ops;
1141 extern struct kvm_device_ops kvm_xics_ops;
1142 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1143 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1145 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1147 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1149 vcpu->spin_loop.in_spin_loop = val;
1151 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1153 vcpu->spin_loop.dy_eligible = val;
1156 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1158 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1162 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1165 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */