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 #define KVM_MMIO_SIZE 8
41 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
42 * in kvm, other bits are visible for userspace which are defined in
43 * include/linux/kvm_h.
45 #define KVM_MEMSLOT_INVALID (1UL << 16)
47 /* Two fragments for cross MMIO pages. */
48 #define KVM_MAX_MMIO_FRAGMENTS 2
51 * For the normal pfn, the highest 12 bits should be zero,
52 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
53 * mask bit 63 to indicate the noslot pfn.
55 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
56 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
57 #define KVM_PFN_NOSLOT (0x1ULL << 63)
59 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
60 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
61 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
64 * error pfns indicate that the gfn is in slot but faild to
65 * translate it to pfn on host.
67 static inline bool is_error_pfn(pfn_t pfn)
69 return !!(pfn & KVM_PFN_ERR_MASK);
73 * error_noslot pfns indicate that the gfn can not be
74 * translated to pfn - it is not in slot or failed to
75 * translate it to pfn.
77 static inline bool is_error_noslot_pfn(pfn_t pfn)
79 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
82 /* noslot pfn indicates that the gfn is not in slot. */
83 static inline bool is_noslot_pfn(pfn_t pfn)
85 return pfn == KVM_PFN_NOSLOT;
89 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
90 * provide own defines and kvm_is_error_hva
92 #ifndef KVM_HVA_ERR_BAD
94 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
95 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
97 static inline bool kvm_is_error_hva(unsigned long addr)
99 return addr >= PAGE_OFFSET;
104 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
106 static inline bool is_error_page(struct page *page)
112 * vcpu->requests bit members
114 #define KVM_REQ_TLB_FLUSH 0
115 #define KVM_REQ_MIGRATE_TIMER 1
116 #define KVM_REQ_REPORT_TPR_ACCESS 2
117 #define KVM_REQ_MMU_RELOAD 3
118 #define KVM_REQ_TRIPLE_FAULT 4
119 #define KVM_REQ_PENDING_TIMER 5
120 #define KVM_REQ_UNHALT 6
121 #define KVM_REQ_MMU_SYNC 7
122 #define KVM_REQ_CLOCK_UPDATE 8
123 #define KVM_REQ_KICK 9
124 #define KVM_REQ_DEACTIVATE_FPU 10
125 #define KVM_REQ_EVENT 11
126 #define KVM_REQ_APF_HALT 12
127 #define KVM_REQ_STEAL_UPDATE 13
128 #define KVM_REQ_NMI 14
129 #define KVM_REQ_PMU 15
130 #define KVM_REQ_PMI 16
131 #define KVM_REQ_WATCHDOG 17
132 #define KVM_REQ_MASTERCLOCK_UPDATE 18
133 #define KVM_REQ_MCLOCK_INPROGRESS 19
134 #define KVM_REQ_EPR_EXIT 20
135 #define KVM_REQ_SCAN_IOAPIC 21
136 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 22
138 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
139 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
143 extern struct kmem_cache *kvm_vcpu_cache;
145 extern spinlock_t kvm_lock;
146 extern struct list_head vm_list;
148 struct kvm_io_range {
151 struct kvm_io_device *dev;
154 #define NR_IOBUS_DEVS 1000
159 struct kvm_io_range range[];
165 KVM_VIRTIO_CCW_NOTIFY_BUS,
169 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
170 int len, const void *val);
171 int kvm_io_bus_write_cookie(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
172 int len, const void *val, long cookie);
173 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
175 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
176 int len, struct kvm_io_device *dev);
177 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
178 struct kvm_io_device *dev);
180 #ifdef CONFIG_KVM_ASYNC_PF
181 struct kvm_async_pf {
182 struct work_struct work;
183 struct list_head link;
184 struct list_head queue;
185 struct kvm_vcpu *vcpu;
186 struct mm_struct *mm;
189 struct kvm_arch_async_pf arch;
193 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
194 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
195 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
196 struct kvm_arch_async_pf *arch);
197 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
204 READING_SHADOW_PAGE_TABLES,
208 * Sometimes a large or cross-page mmio needs to be broken up into separate
209 * exits for userspace servicing.
211 struct kvm_mmio_fragment {
219 #ifdef CONFIG_PREEMPT_NOTIFIERS
220 struct preempt_notifier preempt_notifier;
226 unsigned long requests;
227 unsigned long guest_debug;
233 int guest_fpu_loaded, guest_xcr0_loaded;
234 wait_queue_head_t wq;
238 struct kvm_vcpu_stat stat;
240 #ifdef CONFIG_HAS_IOMEM
242 int mmio_read_completed;
244 int mmio_cur_fragment;
245 int mmio_nr_fragments;
246 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
249 #ifdef CONFIG_KVM_ASYNC_PF
252 struct list_head queue;
253 struct list_head done;
258 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
260 * Cpu relax intercept or pause loop exit optimization
261 * in_spin_loop: set when a vcpu does a pause loop exit
262 * or cpu relax intercepted.
263 * dy_eligible: indicates whether vcpu is eligible for directed yield.
271 struct kvm_vcpu_arch arch;
274 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
276 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
280 * Some of the bitops functions do not support too long bitmaps.
281 * This number must be determined not to exceed such limits.
283 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
285 struct kvm_memory_slot {
287 unsigned long npages;
288 unsigned long *dirty_bitmap;
289 struct kvm_arch_memory_slot arch;
290 unsigned long userspace_addr;
295 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
297 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
300 struct kvm_kernel_irq_routing_entry {
303 int (*set)(struct kvm_kernel_irq_routing_entry *e,
304 struct kvm *kvm, int irq_source_id, int level,
313 struct hlist_node link;
316 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
318 struct kvm_irq_routing_table {
319 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
320 struct kvm_kernel_irq_routing_entry *rt_entries;
323 * Array indexed by gsi. Each entry contains list of irq chips
324 * the gsi is connected to.
326 struct hlist_head map[0];
331 struct kvm_irq_routing_table {};
335 #ifndef KVM_PRIVATE_MEM_SLOTS
336 #define KVM_PRIVATE_MEM_SLOTS 0
339 #ifndef KVM_MEM_SLOTS_NUM
340 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
345 * memslots are not sorted by id anymore, please use id_to_memslot()
346 * to get the memslot by its id.
348 struct kvm_memslots {
350 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
351 /* The mapping table from slot id to the index in memslots[]. */
352 short id_to_index[KVM_MEM_SLOTS_NUM];
357 struct mutex slots_lock;
358 struct mm_struct *mm; /* userspace tied to this vm */
359 struct kvm_memslots *memslots;
360 struct srcu_struct srcu;
361 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
364 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
365 atomic_t online_vcpus;
366 int last_boosted_vcpu;
367 struct list_head vm_list;
369 struct kvm_io_bus *buses[KVM_NR_BUSES];
370 #ifdef CONFIG_HAVE_KVM_EVENTFD
373 struct list_head items;
374 struct list_head resampler_list;
375 struct mutex resampler_lock;
377 struct list_head ioeventfds;
379 struct kvm_vm_stat stat;
380 struct kvm_arch arch;
381 atomic_t users_count;
382 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
383 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
384 spinlock_t ring_lock;
385 struct list_head coalesced_zones;
388 struct mutex irq_lock;
389 #ifdef CONFIG_HAVE_KVM_IRQCHIP
391 * Update side is protected by irq_lock and,
392 * if configured, irqfds.lock.
394 struct kvm_irq_routing_table __rcu *irq_routing;
395 struct hlist_head mask_notifier_list;
396 struct hlist_head irq_ack_notifier_list;
399 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
400 struct mmu_notifier mmu_notifier;
401 unsigned long mmu_notifier_seq;
402 long mmu_notifier_count;
405 struct list_head devices;
408 #define kvm_err(fmt, ...) \
409 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
410 #define kvm_info(fmt, ...) \
411 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
412 #define kvm_debug(fmt, ...) \
413 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
414 #define kvm_pr_unimpl(fmt, ...) \
415 pr_err_ratelimited("kvm [%i]: " fmt, \
416 task_tgid_nr(current), ## __VA_ARGS__)
418 /* The guest did something we don't support. */
419 #define vcpu_unimpl(vcpu, fmt, ...) \
420 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
422 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
425 return kvm->vcpus[i];
428 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
430 idx < atomic_read(&kvm->online_vcpus) && \
431 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
434 #define kvm_for_each_memslot(memslot, slots) \
435 for (memslot = &slots->memslots[0]; \
436 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
439 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
440 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
442 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
443 void vcpu_put(struct kvm_vcpu *vcpu);
445 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
446 int kvm_irqfd_init(void);
447 void kvm_irqfd_exit(void);
449 static inline int kvm_irqfd_init(void)
454 static inline void kvm_irqfd_exit(void)
458 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
459 struct module *module);
462 void kvm_get_kvm(struct kvm *kvm);
463 void kvm_put_kvm(struct kvm *kvm);
465 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
467 return rcu_dereference_check(kvm->memslots,
468 srcu_read_lock_held(&kvm->srcu)
469 || lockdep_is_held(&kvm->slots_lock));
472 static inline struct kvm_memory_slot *
473 id_to_memslot(struct kvm_memslots *slots, int id)
475 int index = slots->id_to_index[id];
476 struct kvm_memory_slot *slot;
478 slot = &slots->memslots[index];
480 WARN_ON(slot->id != id);
485 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
486 * - create a new memory slot
487 * - delete an existing memory slot
488 * - modify an existing memory slot
489 * -- move it in the guest physical memory space
490 * -- just change its flags
492 * Since flags can be changed by some of these operations, the following
493 * differentiation is the best we can do for __kvm_set_memory_region():
502 int kvm_set_memory_region(struct kvm *kvm,
503 struct kvm_userspace_memory_region *mem);
504 int __kvm_set_memory_region(struct kvm *kvm,
505 struct kvm_userspace_memory_region *mem);
506 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
507 struct kvm_memory_slot *dont);
508 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
509 unsigned long npages);
510 void kvm_arch_memslots_updated(struct kvm *kvm);
511 int kvm_arch_prepare_memory_region(struct kvm *kvm,
512 struct kvm_memory_slot *memslot,
513 struct kvm_userspace_memory_region *mem,
514 enum kvm_mr_change change);
515 void kvm_arch_commit_memory_region(struct kvm *kvm,
516 struct kvm_userspace_memory_region *mem,
517 const struct kvm_memory_slot *old,
518 enum kvm_mr_change change);
519 bool kvm_largepages_enabled(void);
520 void kvm_disable_largepages(void);
521 /* flush all memory translations */
522 void kvm_arch_flush_shadow_all(struct kvm *kvm);
523 /* flush memory translations pointing to 'slot' */
524 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
525 struct kvm_memory_slot *slot);
527 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
530 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
531 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
532 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
533 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
534 void kvm_release_page_clean(struct page *page);
535 void kvm_release_page_dirty(struct page *page);
536 void kvm_set_page_accessed(struct page *page);
538 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
539 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
540 bool write_fault, bool *writable);
541 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
542 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
544 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
545 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
547 void kvm_release_pfn_clean(pfn_t pfn);
548 void kvm_set_pfn_dirty(pfn_t pfn);
549 void kvm_set_pfn_accessed(pfn_t pfn);
550 void kvm_get_pfn(pfn_t pfn);
552 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
554 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
556 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
557 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
558 void *data, unsigned long len);
559 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
560 int offset, int len);
561 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
563 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
564 void *data, unsigned long len);
565 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
566 gpa_t gpa, unsigned long len);
567 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
568 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
569 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
570 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
571 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
572 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
574 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
575 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
576 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
577 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
578 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
579 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
581 void kvm_flush_remote_tlbs(struct kvm *kvm);
582 void kvm_reload_remote_mmus(struct kvm *kvm);
583 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
584 void kvm_make_scan_ioapic_request(struct kvm *kvm);
586 long kvm_arch_dev_ioctl(struct file *filp,
587 unsigned int ioctl, unsigned long arg);
588 long kvm_arch_vcpu_ioctl(struct file *filp,
589 unsigned int ioctl, unsigned long arg);
590 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
592 int kvm_dev_ioctl_check_extension(long ext);
594 int kvm_get_dirty_log(struct kvm *kvm,
595 struct kvm_dirty_log *log, int *is_dirty);
596 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
597 struct kvm_dirty_log *log);
599 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
601 long kvm_arch_vm_ioctl(struct file *filp,
602 unsigned int ioctl, unsigned long arg);
604 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
605 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
607 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
608 struct kvm_translation *tr);
610 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
611 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
612 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
613 struct kvm_sregs *sregs);
614 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
615 struct kvm_sregs *sregs);
616 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
617 struct kvm_mp_state *mp_state);
618 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
619 struct kvm_mp_state *mp_state);
620 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
621 struct kvm_guest_debug *dbg);
622 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
624 int kvm_arch_init(void *opaque);
625 void kvm_arch_exit(void);
627 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
628 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
630 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
631 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
632 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
633 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
634 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
635 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
636 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
638 int kvm_arch_hardware_enable(void *garbage);
639 void kvm_arch_hardware_disable(void *garbage);
640 int kvm_arch_hardware_setup(void);
641 void kvm_arch_hardware_unsetup(void);
642 void kvm_arch_check_processor_compat(void *rtn);
643 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
644 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
646 void *kvm_kvzalloc(unsigned long size);
647 void kvm_kvfree(const void *addr);
649 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
650 static inline struct kvm *kvm_arch_alloc_vm(void)
652 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
655 static inline void kvm_arch_free_vm(struct kvm *kvm)
661 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
662 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
663 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
664 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
666 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
670 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
674 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
680 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
682 #ifdef __KVM_HAVE_ARCH_WQP
683 return vcpu->arch.wqp;
689 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
690 void kvm_arch_destroy_vm(struct kvm *kvm);
691 void kvm_arch_sync_events(struct kvm *kvm);
693 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
694 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
696 bool kvm_is_mmio_pfn(pfn_t pfn);
698 struct kvm_irq_ack_notifier {
699 struct hlist_node link;
701 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
704 struct kvm_assigned_dev_kernel {
705 struct kvm_irq_ack_notifier ack_notifier;
706 struct list_head list;
711 unsigned int entries_nr;
713 bool host_irq_disabled;
715 struct msix_entry *host_msix_entries;
717 struct msix_entry *guest_msix_entries;
718 unsigned long irq_requested_type;
723 spinlock_t intx_lock;
724 spinlock_t intx_mask_lock;
726 struct pci_saved_state *pci_saved_state;
729 struct kvm_irq_mask_notifier {
730 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
732 struct hlist_node link;
735 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
736 struct kvm_irq_mask_notifier *kimn);
737 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
738 struct kvm_irq_mask_notifier *kimn);
739 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
742 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
744 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
745 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
746 int irq_source_id, int level, bool line_status);
747 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
748 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
749 void kvm_register_irq_ack_notifier(struct kvm *kvm,
750 struct kvm_irq_ack_notifier *kian);
751 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
752 struct kvm_irq_ack_notifier *kian);
753 int kvm_request_irq_source_id(struct kvm *kvm);
754 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
756 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
757 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
758 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
759 int kvm_iommu_map_guest(struct kvm *kvm);
760 int kvm_iommu_unmap_guest(struct kvm *kvm);
761 int kvm_assign_device(struct kvm *kvm,
762 struct kvm_assigned_dev_kernel *assigned_dev);
763 int kvm_deassign_device(struct kvm *kvm,
764 struct kvm_assigned_dev_kernel *assigned_dev);
766 static inline int kvm_iommu_map_pages(struct kvm *kvm,
767 struct kvm_memory_slot *slot)
772 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
773 struct kvm_memory_slot *slot)
777 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
783 static inline void kvm_guest_enter(void)
787 BUG_ON(preemptible());
789 local_irq_save(flags);
791 local_irq_restore(flags);
793 /* KVM does not hold any references to rcu protected data when it
794 * switches CPU into a guest mode. In fact switching to a guest mode
795 * is very similar to exiting to userspace from rcu point of view. In
796 * addition CPU may stay in a guest mode for quite a long time (up to
797 * one time slice). Lets treat guest mode as quiescent state, just like
798 * we do with user-mode execution.
800 rcu_virt_note_context_switch(smp_processor_id());
803 static inline void kvm_guest_exit(void)
807 local_irq_save(flags);
809 local_irq_restore(flags);
813 * search_memslots() and __gfn_to_memslot() are here because they are
814 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
815 * gfn_to_memslot() itself isn't here as an inline because that would
816 * bloat other code too much.
818 static inline struct kvm_memory_slot *
819 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
821 struct kvm_memory_slot *memslot;
823 kvm_for_each_memslot(memslot, slots)
824 if (gfn >= memslot->base_gfn &&
825 gfn < memslot->base_gfn + memslot->npages)
831 static inline struct kvm_memory_slot *
832 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
834 return search_memslots(slots, gfn);
837 static inline unsigned long
838 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
840 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
843 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
845 return gfn_to_memslot(kvm, gfn)->id;
849 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
851 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
853 return slot->base_gfn + gfn_offset;
856 static inline gpa_t gfn_to_gpa(gfn_t gfn)
858 return (gpa_t)gfn << PAGE_SHIFT;
861 static inline gfn_t gpa_to_gfn(gpa_t gpa)
863 return (gfn_t)(gpa >> PAGE_SHIFT);
866 static inline hpa_t pfn_to_hpa(pfn_t pfn)
868 return (hpa_t)pfn << PAGE_SHIFT;
871 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
873 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
881 struct kvm_stats_debugfs_item {
884 enum kvm_stat_kind kind;
885 struct dentry *dentry;
887 extern struct kvm_stats_debugfs_item debugfs_entries[];
888 extern struct dentry *kvm_debugfs_dir;
890 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
891 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
893 if (unlikely(kvm->mmu_notifier_count))
896 * Ensure the read of mmu_notifier_count happens before the read
897 * of mmu_notifier_seq. This interacts with the smp_wmb() in
898 * mmu_notifier_invalidate_range_end to make sure that the caller
899 * either sees the old (non-zero) value of mmu_notifier_count or
900 * the new (incremented) value of mmu_notifier_seq.
901 * PowerPC Book3s HV KVM calls this under a per-page lock
902 * rather than under kvm->mmu_lock, for scalability, so
903 * can't rely on kvm->mmu_lock to keep things ordered.
906 if (kvm->mmu_notifier_seq != mmu_seq)
912 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
914 #define KVM_MAX_IRQ_ROUTES 1024
916 int kvm_setup_default_irq_routing(struct kvm *kvm);
917 int kvm_set_irq_routing(struct kvm *kvm,
918 const struct kvm_irq_routing_entry *entries,
921 int kvm_set_routing_entry(struct kvm_irq_routing_table *rt,
922 struct kvm_kernel_irq_routing_entry *e,
923 const struct kvm_irq_routing_entry *ue);
924 void kvm_free_irq_routing(struct kvm *kvm);
926 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
930 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
934 #ifdef CONFIG_HAVE_KVM_EVENTFD
936 void kvm_eventfd_init(struct kvm *kvm);
937 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
939 #ifdef CONFIG_HAVE_KVM_IRQCHIP
940 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
941 void kvm_irqfd_release(struct kvm *kvm);
942 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
944 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
949 static inline void kvm_irqfd_release(struct kvm *kvm) {}
954 static inline void kvm_eventfd_init(struct kvm *kvm) {}
956 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
961 static inline void kvm_irqfd_release(struct kvm *kvm) {}
963 #ifdef CONFIG_HAVE_KVM_IRQCHIP
964 static inline void kvm_irq_routing_update(struct kvm *kvm,
965 struct kvm_irq_routing_table *irq_rt)
967 rcu_assign_pointer(kvm->irq_routing, irq_rt);
971 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
976 #endif /* CONFIG_HAVE_KVM_EVENTFD */
978 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
979 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
981 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
984 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
988 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
992 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
994 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
997 void kvm_free_all_assigned_devices(struct kvm *kvm);
1001 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
1007 static inline void kvm_free_all_assigned_devices(struct kvm *kvm) {}
1011 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1013 set_bit(req, &vcpu->requests);
1016 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1018 if (test_bit(req, &vcpu->requests)) {
1019 clear_bit(req, &vcpu->requests);
1026 extern bool kvm_rebooting;
1028 struct kvm_device_ops;
1031 struct kvm_device_ops *ops;
1034 struct list_head vm_node;
1037 /* create, destroy, and name are mandatory */
1038 struct kvm_device_ops {
1040 int (*create)(struct kvm_device *dev, u32 type);
1043 * Destroy is responsible for freeing dev.
1045 * Destroy may be called before or after destructors are called
1046 * on emulated I/O regions, depending on whether a reference is
1047 * held by a vcpu or other kvm component that gets destroyed
1048 * after the emulated I/O.
1050 void (*destroy)(struct kvm_device *dev);
1052 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1053 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1054 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1055 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1059 void kvm_device_get(struct kvm_device *dev);
1060 void kvm_device_put(struct kvm_device *dev);
1061 struct kvm_device *kvm_device_from_filp(struct file *filp);
1063 extern struct kvm_device_ops kvm_mpic_ops;
1064 extern struct kvm_device_ops kvm_xics_ops;
1065 extern struct kvm_device_ops kvm_vfio_ops;
1066 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1068 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1070 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1072 vcpu->spin_loop.in_spin_loop = val;
1074 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1076 vcpu->spin_loop.dy_eligible = val;
1079 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1081 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1085 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1088 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */