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 <asm/signal.h>
27 #include <linux/kvm.h>
28 #include <linux/kvm_para.h>
30 #include <linux/kvm_types.h>
32 #include <asm/kvm_host.h>
35 #define KVM_MMIO_SIZE 8
39 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
40 * in kvm, other bits are visible for userspace which are defined in
41 * include/linux/kvm_h.
43 #define KVM_MEMSLOT_INVALID (1UL << 16)
45 /* Two fragments for cross MMIO pages. */
46 #define KVM_MAX_MMIO_FRAGMENTS 2
49 * For the normal pfn, the highest 12 bits should be zero,
50 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
51 * mask bit 63 to indicate the noslot pfn.
53 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
54 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
55 #define KVM_PFN_NOSLOT (0x1ULL << 63)
57 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
58 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
59 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
62 * error pfns indicate that the gfn is in slot but faild to
63 * translate it to pfn on host.
65 static inline bool is_error_pfn(pfn_t pfn)
67 return !!(pfn & KVM_PFN_ERR_MASK);
71 * error_noslot pfns indicate that the gfn can not be
72 * translated to pfn - it is not in slot or failed to
73 * translate it to pfn.
75 static inline bool is_error_noslot_pfn(pfn_t pfn)
77 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
80 /* noslot pfn indicates that the gfn is not in slot. */
81 static inline bool is_noslot_pfn(pfn_t pfn)
83 return pfn == KVM_PFN_NOSLOT;
86 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
87 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
89 static inline bool kvm_is_error_hva(unsigned long addr)
91 return addr >= PAGE_OFFSET;
94 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
96 static inline bool is_error_page(struct page *page)
102 * vcpu->requests bit members
104 #define KVM_REQ_TLB_FLUSH 0
105 #define KVM_REQ_MIGRATE_TIMER 1
106 #define KVM_REQ_REPORT_TPR_ACCESS 2
107 #define KVM_REQ_MMU_RELOAD 3
108 #define KVM_REQ_TRIPLE_FAULT 4
109 #define KVM_REQ_PENDING_TIMER 5
110 #define KVM_REQ_UNHALT 6
111 #define KVM_REQ_MMU_SYNC 7
112 #define KVM_REQ_CLOCK_UPDATE 8
113 #define KVM_REQ_KICK 9
114 #define KVM_REQ_DEACTIVATE_FPU 10
115 #define KVM_REQ_EVENT 11
116 #define KVM_REQ_APF_HALT 12
117 #define KVM_REQ_STEAL_UPDATE 13
118 #define KVM_REQ_NMI 14
119 #define KVM_REQ_IMMEDIATE_EXIT 15
120 #define KVM_REQ_PMU 16
121 #define KVM_REQ_PMI 17
122 #define KVM_REQ_WATCHDOG 18
123 #define KVM_REQ_MASTERCLOCK_UPDATE 19
124 #define KVM_REQ_MCLOCK_INPROGRESS 20
125 #define KVM_REQ_EPR_EXIT 21
126 #define KVM_REQ_EOIBITMAP 22
128 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
129 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
133 extern struct kmem_cache *kvm_vcpu_cache;
135 struct kvm_io_range {
138 struct kvm_io_device *dev;
141 #define NR_IOBUS_DEVS 1000
145 struct kvm_io_range range[];
154 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
155 int len, const void *val);
156 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
158 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
159 int len, struct kvm_io_device *dev);
160 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
161 struct kvm_io_device *dev);
163 #ifdef CONFIG_KVM_ASYNC_PF
164 struct kvm_async_pf {
165 struct work_struct work;
166 struct list_head link;
167 struct list_head queue;
168 struct kvm_vcpu *vcpu;
169 struct mm_struct *mm;
172 struct kvm_arch_async_pf arch;
177 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
178 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
179 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
180 struct kvm_arch_async_pf *arch);
181 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
188 READING_SHADOW_PAGE_TABLES,
192 * Sometimes a large or cross-page mmio needs to be broken up into separate
193 * exits for userspace servicing.
195 struct kvm_mmio_fragment {
203 #ifdef CONFIG_PREEMPT_NOTIFIERS
204 struct preempt_notifier preempt_notifier;
210 unsigned long requests;
211 unsigned long guest_debug;
217 int guest_fpu_loaded, guest_xcr0_loaded;
218 wait_queue_head_t wq;
222 struct kvm_vcpu_stat stat;
224 #ifdef CONFIG_HAS_IOMEM
226 int mmio_read_completed;
228 int mmio_cur_fragment;
229 int mmio_nr_fragments;
230 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
233 #ifdef CONFIG_KVM_ASYNC_PF
236 struct list_head queue;
237 struct list_head done;
242 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
244 * Cpu relax intercept or pause loop exit optimization
245 * in_spin_loop: set when a vcpu does a pause loop exit
246 * or cpu relax intercepted.
247 * dy_eligible: indicates whether vcpu is eligible for directed yield.
254 struct kvm_vcpu_arch arch;
257 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
259 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
263 * Some of the bitops functions do not support too long bitmaps.
264 * This number must be determined not to exceed such limits.
266 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
268 struct kvm_memory_slot {
270 unsigned long npages;
271 unsigned long *dirty_bitmap;
272 struct kvm_arch_memory_slot arch;
273 unsigned long userspace_addr;
279 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
281 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
284 struct kvm_kernel_irq_routing_entry {
287 int (*set)(struct kvm_kernel_irq_routing_entry *e,
288 struct kvm *kvm, int irq_source_id, int level);
296 struct hlist_node link;
299 #ifdef __KVM_HAVE_IOAPIC
301 struct kvm_irq_routing_table {
302 int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
303 struct kvm_kernel_irq_routing_entry *rt_entries;
306 * Array indexed by gsi. Each entry contains list of irq chips
307 * the gsi is connected to.
309 struct hlist_head map[0];
314 struct kvm_irq_routing_table {};
318 #ifndef KVM_PRIVATE_MEM_SLOTS
319 #define KVM_PRIVATE_MEM_SLOTS 0
322 #ifndef KVM_MEM_SLOTS_NUM
323 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
328 * memslots are not sorted by id anymore, please use id_to_memslot()
329 * to get the memslot by its id.
331 struct kvm_memslots {
333 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
334 /* The mapping table from slot id to the index in memslots[]. */
335 short id_to_index[KVM_MEM_SLOTS_NUM];
340 struct mutex slots_lock;
341 struct mm_struct *mm; /* userspace tied to this vm */
342 struct kvm_memslots *memslots;
343 struct srcu_struct srcu;
344 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
347 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
348 atomic_t online_vcpus;
349 int last_boosted_vcpu;
350 struct list_head vm_list;
352 struct kvm_io_bus *buses[KVM_NR_BUSES];
353 #ifdef CONFIG_HAVE_KVM_EVENTFD
356 struct list_head items;
357 struct list_head resampler_list;
358 struct mutex resampler_lock;
360 struct list_head ioeventfds;
362 struct kvm_vm_stat stat;
363 struct kvm_arch arch;
364 atomic_t users_count;
365 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
366 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
367 spinlock_t ring_lock;
368 struct list_head coalesced_zones;
371 struct mutex irq_lock;
372 #ifdef CONFIG_HAVE_KVM_IRQCHIP
374 * Update side is protected by irq_lock and,
375 * if configured, irqfds.lock.
377 struct kvm_irq_routing_table __rcu *irq_routing;
378 struct hlist_head mask_notifier_list;
379 struct hlist_head irq_ack_notifier_list;
382 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
383 struct mmu_notifier mmu_notifier;
384 unsigned long mmu_notifier_seq;
385 long mmu_notifier_count;
390 #define kvm_err(fmt, ...) \
391 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
392 #define kvm_info(fmt, ...) \
393 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
394 #define kvm_debug(fmt, ...) \
395 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
396 #define kvm_pr_unimpl(fmt, ...) \
397 pr_err_ratelimited("kvm [%i]: " fmt, \
398 task_tgid_nr(current), ## __VA_ARGS__)
400 /* The guest did something we don't support. */
401 #define vcpu_unimpl(vcpu, fmt, ...) \
402 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
404 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
407 return kvm->vcpus[i];
410 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
412 idx < atomic_read(&kvm->online_vcpus) && \
413 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
416 #define kvm_for_each_memslot(memslot, slots) \
417 for (memslot = &slots->memslots[0]; \
418 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
421 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
422 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
424 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
425 void vcpu_put(struct kvm_vcpu *vcpu);
427 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
428 struct module *module);
431 void kvm_get_kvm(struct kvm *kvm);
432 void kvm_put_kvm(struct kvm *kvm);
433 void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new,
434 u64 last_generation);
436 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
438 return rcu_dereference_check(kvm->memslots,
439 srcu_read_lock_held(&kvm->srcu)
440 || lockdep_is_held(&kvm->slots_lock));
443 static inline struct kvm_memory_slot *
444 id_to_memslot(struct kvm_memslots *slots, int id)
446 int index = slots->id_to_index[id];
447 struct kvm_memory_slot *slot;
449 slot = &slots->memslots[index];
451 WARN_ON(slot->id != id);
455 int kvm_set_memory_region(struct kvm *kvm,
456 struct kvm_userspace_memory_region *mem,
458 int __kvm_set_memory_region(struct kvm *kvm,
459 struct kvm_userspace_memory_region *mem,
461 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
462 struct kvm_memory_slot *dont);
463 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages);
464 int kvm_arch_prepare_memory_region(struct kvm *kvm,
465 struct kvm_memory_slot *memslot,
466 struct kvm_memory_slot old,
467 struct kvm_userspace_memory_region *mem,
469 void kvm_arch_commit_memory_region(struct kvm *kvm,
470 struct kvm_userspace_memory_region *mem,
471 struct kvm_memory_slot old,
473 bool kvm_largepages_enabled(void);
474 void kvm_disable_largepages(void);
475 /* flush all memory translations */
476 void kvm_arch_flush_shadow_all(struct kvm *kvm);
477 /* flush memory translations pointing to 'slot' */
478 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
479 struct kvm_memory_slot *slot);
481 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
484 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
485 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
486 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
487 void kvm_release_page_clean(struct page *page);
488 void kvm_release_page_dirty(struct page *page);
489 void kvm_set_page_dirty(struct page *page);
490 void kvm_set_page_accessed(struct page *page);
492 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
493 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
494 bool write_fault, bool *writable);
495 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
496 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
498 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
499 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
501 void kvm_release_pfn_dirty(pfn_t pfn);
502 void kvm_release_pfn_clean(pfn_t pfn);
503 void kvm_set_pfn_dirty(pfn_t pfn);
504 void kvm_set_pfn_accessed(pfn_t pfn);
505 void kvm_get_pfn(pfn_t pfn);
507 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
509 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
511 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
512 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
513 void *data, unsigned long len);
514 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
515 int offset, int len);
516 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
518 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
519 void *data, unsigned long len);
520 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
522 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
523 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
524 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
525 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
526 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
527 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
528 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
531 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
532 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
533 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
534 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
535 void kvm_resched(struct kvm_vcpu *vcpu);
536 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
537 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
539 void kvm_flush_remote_tlbs(struct kvm *kvm);
540 void kvm_reload_remote_mmus(struct kvm *kvm);
541 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
542 void kvm_make_update_eoibitmap_request(struct kvm *kvm);
544 long kvm_arch_dev_ioctl(struct file *filp,
545 unsigned int ioctl, unsigned long arg);
546 long kvm_arch_vcpu_ioctl(struct file *filp,
547 unsigned int ioctl, unsigned long arg);
548 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
550 int kvm_dev_ioctl_check_extension(long ext);
552 int kvm_get_dirty_log(struct kvm *kvm,
553 struct kvm_dirty_log *log, int *is_dirty);
554 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
555 struct kvm_dirty_log *log);
557 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
559 kvm_userspace_memory_region *mem,
561 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level);
562 long kvm_arch_vm_ioctl(struct file *filp,
563 unsigned int ioctl, unsigned long arg);
565 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
566 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
568 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
569 struct kvm_translation *tr);
571 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
572 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
573 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
574 struct kvm_sregs *sregs);
575 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
576 struct kvm_sregs *sregs);
577 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
578 struct kvm_mp_state *mp_state);
579 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
580 struct kvm_mp_state *mp_state);
581 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
582 struct kvm_guest_debug *dbg);
583 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
585 int kvm_arch_init(void *opaque);
586 void kvm_arch_exit(void);
588 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
589 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
591 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
592 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
593 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
594 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
595 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
596 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
597 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
599 int kvm_arch_hardware_enable(void *garbage);
600 void kvm_arch_hardware_disable(void *garbage);
601 int kvm_arch_hardware_setup(void);
602 void kvm_arch_hardware_unsetup(void);
603 void kvm_arch_check_processor_compat(void *rtn);
604 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
605 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
607 void kvm_free_physmem(struct kvm *kvm);
609 void *kvm_kvzalloc(unsigned long size);
610 void kvm_kvfree(const void *addr);
612 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
613 static inline struct kvm *kvm_arch_alloc_vm(void)
615 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
618 static inline void kvm_arch_free_vm(struct kvm *kvm)
624 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
626 #ifdef __KVM_HAVE_ARCH_WQP
627 return vcpu->arch.wqp;
633 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
634 void kvm_arch_destroy_vm(struct kvm *kvm);
635 void kvm_free_all_assigned_devices(struct kvm *kvm);
636 void kvm_arch_sync_events(struct kvm *kvm);
638 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
639 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
641 bool kvm_is_mmio_pfn(pfn_t pfn);
643 struct kvm_irq_ack_notifier {
644 struct hlist_node link;
646 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
649 struct kvm_assigned_dev_kernel {
650 struct kvm_irq_ack_notifier ack_notifier;
651 struct list_head list;
656 unsigned int entries_nr;
658 bool host_irq_disabled;
660 struct msix_entry *host_msix_entries;
662 struct msix_entry *guest_msix_entries;
663 unsigned long irq_requested_type;
668 spinlock_t intx_lock;
669 spinlock_t intx_mask_lock;
671 struct pci_saved_state *pci_saved_state;
674 struct kvm_irq_mask_notifier {
675 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
677 struct hlist_node link;
680 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
681 struct kvm_irq_mask_notifier *kimn);
682 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
683 struct kvm_irq_mask_notifier *kimn);
684 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
687 #ifdef __KVM_HAVE_IOAPIC
688 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
689 union kvm_ioapic_redirect_entry *entry,
690 unsigned long *deliver_bitmask);
692 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
693 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
694 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
695 int irq_source_id, int level);
696 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
697 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
698 void kvm_register_irq_ack_notifier(struct kvm *kvm,
699 struct kvm_irq_ack_notifier *kian);
700 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
701 struct kvm_irq_ack_notifier *kian);
702 int kvm_request_irq_source_id(struct kvm *kvm);
703 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
705 /* For vcpu->arch.iommu_flags */
706 #define KVM_IOMMU_CACHE_COHERENCY 0x1
708 #ifdef CONFIG_IOMMU_API
709 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
710 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
711 int kvm_iommu_map_guest(struct kvm *kvm);
712 int kvm_iommu_unmap_guest(struct kvm *kvm);
713 int kvm_assign_device(struct kvm *kvm,
714 struct kvm_assigned_dev_kernel *assigned_dev);
715 int kvm_deassign_device(struct kvm *kvm,
716 struct kvm_assigned_dev_kernel *assigned_dev);
717 #else /* CONFIG_IOMMU_API */
718 static inline int kvm_iommu_map_pages(struct kvm *kvm,
719 struct kvm_memory_slot *slot)
724 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
725 struct kvm_memory_slot *slot)
729 static inline int kvm_iommu_map_guest(struct kvm *kvm)
734 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
739 static inline int kvm_assign_device(struct kvm *kvm,
740 struct kvm_assigned_dev_kernel *assigned_dev)
745 static inline int kvm_deassign_device(struct kvm *kvm,
746 struct kvm_assigned_dev_kernel *assigned_dev)
750 #endif /* CONFIG_IOMMU_API */
752 static inline void kvm_guest_enter(void)
754 BUG_ON(preemptible());
756 * This is running in ioctl context so we can avoid
757 * the call to vtime_account() with its unnecessary idle check.
759 vtime_account_system_irqsafe(current);
760 current->flags |= PF_VCPU;
761 /* KVM does not hold any references to rcu protected data when it
762 * switches CPU into a guest mode. In fact switching to a guest mode
763 * is very similar to exiting to userspase from rcu point of view. In
764 * addition CPU may stay in a guest mode for quite a long time (up to
765 * one time slice). Lets treat guest mode as quiescent state, just like
766 * we do with user-mode execution.
768 rcu_virt_note_context_switch(smp_processor_id());
771 static inline void kvm_guest_exit(void)
774 * This is running in ioctl context so we can avoid
775 * the call to vtime_account() with its unnecessary idle check.
777 vtime_account_system_irqsafe(current);
778 current->flags &= ~PF_VCPU;
782 * search_memslots() and __gfn_to_memslot() are here because they are
783 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
784 * gfn_to_memslot() itself isn't here as an inline because that would
785 * bloat other code too much.
787 static inline struct kvm_memory_slot *
788 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
790 struct kvm_memory_slot *memslot;
792 kvm_for_each_memslot(memslot, slots)
793 if (gfn >= memslot->base_gfn &&
794 gfn < memslot->base_gfn + memslot->npages)
800 static inline struct kvm_memory_slot *
801 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
803 return search_memslots(slots, gfn);
806 static inline unsigned long
807 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
809 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
812 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
814 return gfn_to_memslot(kvm, gfn)->id;
817 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
819 /* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
820 return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
821 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
825 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
827 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
829 return slot->base_gfn + gfn_offset;
832 static inline gpa_t gfn_to_gpa(gfn_t gfn)
834 return (gpa_t)gfn << PAGE_SHIFT;
837 static inline gfn_t gpa_to_gfn(gpa_t gpa)
839 return (gfn_t)(gpa >> PAGE_SHIFT);
842 static inline hpa_t pfn_to_hpa(pfn_t pfn)
844 return (hpa_t)pfn << PAGE_SHIFT;
847 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
849 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
857 struct kvm_stats_debugfs_item {
860 enum kvm_stat_kind kind;
861 struct dentry *dentry;
863 extern struct kvm_stats_debugfs_item debugfs_entries[];
864 extern struct dentry *kvm_debugfs_dir;
866 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
867 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
869 if (unlikely(kvm->mmu_notifier_count))
872 * Ensure the read of mmu_notifier_count happens before the read
873 * of mmu_notifier_seq. This interacts with the smp_wmb() in
874 * mmu_notifier_invalidate_range_end to make sure that the caller
875 * either sees the old (non-zero) value of mmu_notifier_count or
876 * the new (incremented) value of mmu_notifier_seq.
877 * PowerPC Book3s HV KVM calls this under a per-page lock
878 * rather than under kvm->mmu_lock, for scalability, so
879 * can't rely on kvm->mmu_lock to keep things ordered.
882 if (kvm->mmu_notifier_seq != mmu_seq)
888 #ifdef KVM_CAP_IRQ_ROUTING
890 #define KVM_MAX_IRQ_ROUTES 1024
892 int kvm_setup_default_irq_routing(struct kvm *kvm);
893 int kvm_set_irq_routing(struct kvm *kvm,
894 const struct kvm_irq_routing_entry *entries,
897 void kvm_free_irq_routing(struct kvm *kvm);
899 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
903 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
907 #ifdef CONFIG_HAVE_KVM_EVENTFD
909 void kvm_eventfd_init(struct kvm *kvm);
910 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
912 #ifdef CONFIG_HAVE_KVM_IRQCHIP
913 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
914 void kvm_irqfd_release(struct kvm *kvm);
915 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
917 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
922 static inline void kvm_irqfd_release(struct kvm *kvm) {}
927 static inline void kvm_eventfd_init(struct kvm *kvm) {}
929 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
934 static inline void kvm_irqfd_release(struct kvm *kvm) {}
936 #ifdef CONFIG_HAVE_KVM_IRQCHIP
937 static inline void kvm_irq_routing_update(struct kvm *kvm,
938 struct kvm_irq_routing_table *irq_rt)
940 rcu_assign_pointer(kvm->irq_routing, irq_rt);
944 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
949 #endif /* CONFIG_HAVE_KVM_EVENTFD */
951 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
952 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
954 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
957 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
961 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
965 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
967 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
972 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
980 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
982 set_bit(req, &vcpu->requests);
985 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
987 if (test_bit(req, &vcpu->requests)) {
988 clear_bit(req, &vcpu->requests);
995 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
997 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
999 vcpu->spin_loop.in_spin_loop = val;
1001 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1003 vcpu->spin_loop.dy_eligible = val;
1006 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1008 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1012 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1016 static inline bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
1021 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */