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)
46 * If we support unaligned MMIO, at most one fragment will be split into two:
48 #ifdef KVM_UNALIGNED_MMIO
49 # define KVM_EXTRA_MMIO_FRAGMENTS 1
51 # define KVM_EXTRA_MMIO_FRAGMENTS 0
54 #define KVM_USER_MMIO_SIZE 8
56 #define KVM_MAX_MMIO_FRAGMENTS \
57 (KVM_MMIO_SIZE / KVM_USER_MMIO_SIZE + KVM_EXTRA_MMIO_FRAGMENTS)
60 * For the normal pfn, the highest 12 bits should be zero,
61 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
62 * mask bit 63 to indicate the noslot pfn.
64 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
65 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
66 #define KVM_PFN_NOSLOT (0x1ULL << 63)
68 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
69 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
70 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
73 * error pfns indicate that the gfn is in slot but faild to
74 * translate it to pfn on host.
76 static inline bool is_error_pfn(pfn_t pfn)
78 return !!(pfn & KVM_PFN_ERR_MASK);
82 * error_noslot pfns indicate that the gfn can not be
83 * translated to pfn - it is not in slot or failed to
84 * translate it to pfn.
86 static inline bool is_error_noslot_pfn(pfn_t pfn)
88 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
91 /* noslot pfn indicates that the gfn is not in slot. */
92 static inline bool is_noslot_pfn(pfn_t pfn)
94 return pfn == KVM_PFN_NOSLOT;
97 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
98 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
100 static inline bool kvm_is_error_hva(unsigned long addr)
102 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_IMMEDIATE_EXIT 15
131 #define KVM_REQ_PMU 16
132 #define KVM_REQ_PMI 17
133 #define KVM_REQ_WATCHDOG 18
134 #define KVM_REQ_MASTERCLOCK_UPDATE 19
135 #define KVM_REQ_MCLOCK_INPROGRESS 20
137 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
138 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
142 extern struct kmem_cache *kvm_vcpu_cache;
144 struct kvm_io_range {
147 struct kvm_io_device *dev;
150 #define NR_IOBUS_DEVS 1000
154 struct kvm_io_range range[];
163 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
164 int len, const void *val);
165 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
167 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
168 int len, struct kvm_io_device *dev);
169 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
170 struct kvm_io_device *dev);
172 #ifdef CONFIG_KVM_ASYNC_PF
173 struct kvm_async_pf {
174 struct work_struct work;
175 struct list_head link;
176 struct list_head queue;
177 struct kvm_vcpu *vcpu;
178 struct mm_struct *mm;
181 struct kvm_arch_async_pf arch;
186 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
187 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
188 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
189 struct kvm_arch_async_pf *arch);
190 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
197 READING_SHADOW_PAGE_TABLES,
201 * Sometimes a large or cross-page mmio needs to be broken up into separate
202 * exits for userspace servicing.
204 struct kvm_mmio_fragment {
212 #ifdef CONFIG_PREEMPT_NOTIFIERS
213 struct preempt_notifier preempt_notifier;
219 unsigned long requests;
220 unsigned long guest_debug;
226 int guest_fpu_loaded, guest_xcr0_loaded;
227 wait_queue_head_t wq;
231 struct kvm_vcpu_stat stat;
233 #ifdef CONFIG_HAS_IOMEM
235 int mmio_read_completed;
237 int mmio_cur_fragment;
238 int mmio_nr_fragments;
239 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
242 #ifdef CONFIG_KVM_ASYNC_PF
245 struct list_head queue;
246 struct list_head done;
251 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
253 * Cpu relax intercept or pause loop exit optimization
254 * in_spin_loop: set when a vcpu does a pause loop exit
255 * or cpu relax intercepted.
256 * dy_eligible: indicates whether vcpu is eligible for directed yield.
263 struct kvm_vcpu_arch arch;
266 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
268 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
272 * Some of the bitops functions do not support too long bitmaps.
273 * This number must be determined not to exceed such limits.
275 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
277 struct kvm_memory_slot {
279 unsigned long npages;
281 unsigned long *dirty_bitmap;
282 struct kvm_arch_memory_slot arch;
283 unsigned long userspace_addr;
288 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
290 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
293 struct kvm_kernel_irq_routing_entry {
296 int (*set)(struct kvm_kernel_irq_routing_entry *e,
297 struct kvm *kvm, int irq_source_id, int level);
305 struct hlist_node link;
308 #ifdef __KVM_HAVE_IOAPIC
310 struct kvm_irq_routing_table {
311 int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
312 struct kvm_kernel_irq_routing_entry *rt_entries;
315 * Array indexed by gsi. Each entry contains list of irq chips
316 * the gsi is connected to.
318 struct hlist_head map[0];
323 struct kvm_irq_routing_table {};
327 #ifndef KVM_MEM_SLOTS_NUM
328 #define KVM_MEM_SLOTS_NUM (KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
333 * memslots are not sorted by id anymore, please use id_to_memslot()
334 * to get the memslot by its id.
336 struct kvm_memslots {
338 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
339 /* The mapping table from slot id to the index in memslots[]. */
340 int id_to_index[KVM_MEM_SLOTS_NUM];
345 struct mutex slots_lock;
346 struct mm_struct *mm; /* userspace tied to this vm */
347 struct kvm_memslots *memslots;
348 struct srcu_struct srcu;
349 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
352 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
353 atomic_t online_vcpus;
354 int last_boosted_vcpu;
355 struct list_head vm_list;
357 struct kvm_io_bus *buses[KVM_NR_BUSES];
358 #ifdef CONFIG_HAVE_KVM_EVENTFD
361 struct list_head items;
362 struct list_head resampler_list;
363 struct mutex resampler_lock;
365 struct list_head ioeventfds;
367 struct kvm_vm_stat stat;
368 struct kvm_arch arch;
369 atomic_t users_count;
370 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
371 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
372 spinlock_t ring_lock;
373 struct list_head coalesced_zones;
376 struct mutex irq_lock;
377 #ifdef CONFIG_HAVE_KVM_IRQCHIP
379 * Update side is protected by irq_lock and,
380 * if configured, irqfds.lock.
382 struct kvm_irq_routing_table __rcu *irq_routing;
383 struct hlist_head mask_notifier_list;
384 struct hlist_head irq_ack_notifier_list;
387 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
388 struct mmu_notifier mmu_notifier;
389 unsigned long mmu_notifier_seq;
390 long mmu_notifier_count;
395 #define kvm_err(fmt, ...) \
396 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
397 #define kvm_info(fmt, ...) \
398 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
399 #define kvm_debug(fmt, ...) \
400 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
401 #define kvm_pr_unimpl(fmt, ...) \
402 pr_err_ratelimited("kvm [%i]: " fmt, \
403 task_tgid_nr(current), ## __VA_ARGS__)
405 /* The guest did something we don't support. */
406 #define vcpu_unimpl(vcpu, fmt, ...) \
407 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
409 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
412 return kvm->vcpus[i];
415 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
417 idx < atomic_read(&kvm->online_vcpus) && \
418 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
421 #define kvm_for_each_memslot(memslot, slots) \
422 for (memslot = &slots->memslots[0]; \
423 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
426 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
427 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
429 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
430 void vcpu_put(struct kvm_vcpu *vcpu);
432 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
433 struct module *module);
436 void kvm_get_kvm(struct kvm *kvm);
437 void kvm_put_kvm(struct kvm *kvm);
438 void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new);
440 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
442 return rcu_dereference_check(kvm->memslots,
443 srcu_read_lock_held(&kvm->srcu)
444 || lockdep_is_held(&kvm->slots_lock));
447 static inline struct kvm_memory_slot *
448 id_to_memslot(struct kvm_memslots *slots, int id)
450 int index = slots->id_to_index[id];
451 struct kvm_memory_slot *slot;
453 slot = &slots->memslots[index];
455 WARN_ON(slot->id != id);
459 int kvm_set_memory_region(struct kvm *kvm,
460 struct kvm_userspace_memory_region *mem,
462 int __kvm_set_memory_region(struct kvm *kvm,
463 struct kvm_userspace_memory_region *mem,
465 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
466 struct kvm_memory_slot *dont);
467 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages);
468 int kvm_arch_prepare_memory_region(struct kvm *kvm,
469 struct kvm_memory_slot *memslot,
470 struct kvm_memory_slot old,
471 struct kvm_userspace_memory_region *mem,
473 void kvm_arch_commit_memory_region(struct kvm *kvm,
474 struct kvm_userspace_memory_region *mem,
475 struct kvm_memory_slot old,
477 bool kvm_largepages_enabled(void);
478 void kvm_disable_largepages(void);
479 /* flush all memory translations */
480 void kvm_arch_flush_shadow_all(struct kvm *kvm);
481 /* flush memory translations pointing to 'slot' */
482 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
483 struct kvm_memory_slot *slot);
485 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
488 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
489 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
490 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
491 void kvm_release_page_clean(struct page *page);
492 void kvm_release_page_dirty(struct page *page);
493 void kvm_set_page_dirty(struct page *page);
494 void kvm_set_page_accessed(struct page *page);
496 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
497 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
498 bool write_fault, bool *writable);
499 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
500 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
502 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
503 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
505 void kvm_release_pfn_dirty(pfn_t pfn);
506 void kvm_release_pfn_clean(pfn_t pfn);
507 void kvm_set_pfn_dirty(pfn_t pfn);
508 void kvm_set_pfn_accessed(pfn_t pfn);
509 void kvm_get_pfn(pfn_t pfn);
511 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
513 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
515 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
516 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
517 void *data, unsigned long len);
518 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
519 int offset, int len);
520 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
522 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
523 void *data, unsigned long len);
524 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
526 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
527 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
528 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
529 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
530 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
531 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
532 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
535 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
536 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
537 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
538 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
539 void kvm_resched(struct kvm_vcpu *vcpu);
540 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
541 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
543 void kvm_flush_remote_tlbs(struct kvm *kvm);
544 void kvm_reload_remote_mmus(struct kvm *kvm);
545 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
547 long kvm_arch_dev_ioctl(struct file *filp,
548 unsigned int ioctl, unsigned long arg);
549 long kvm_arch_vcpu_ioctl(struct file *filp,
550 unsigned int ioctl, unsigned long arg);
551 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
553 int kvm_dev_ioctl_check_extension(long ext);
555 int kvm_get_dirty_log(struct kvm *kvm,
556 struct kvm_dirty_log *log, int *is_dirty);
557 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
558 struct kvm_dirty_log *log);
560 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
562 kvm_userspace_memory_region *mem,
564 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level);
565 long kvm_arch_vm_ioctl(struct file *filp,
566 unsigned int ioctl, unsigned long arg);
568 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
569 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
571 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
572 struct kvm_translation *tr);
574 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
575 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
576 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
577 struct kvm_sregs *sregs);
578 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
579 struct kvm_sregs *sregs);
580 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
581 struct kvm_mp_state *mp_state);
582 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
583 struct kvm_mp_state *mp_state);
584 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
585 struct kvm_guest_debug *dbg);
586 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
588 int kvm_arch_init(void *opaque);
589 void kvm_arch_exit(void);
591 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
592 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
594 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
595 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
596 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
597 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
598 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
599 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
600 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
602 int kvm_arch_hardware_enable(void *garbage);
603 void kvm_arch_hardware_disable(void *garbage);
604 int kvm_arch_hardware_setup(void);
605 void kvm_arch_hardware_unsetup(void);
606 void kvm_arch_check_processor_compat(void *rtn);
607 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
608 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
610 void kvm_free_physmem(struct kvm *kvm);
612 void *kvm_kvzalloc(unsigned long size);
613 void kvm_kvfree(const void *addr);
615 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
616 static inline struct kvm *kvm_arch_alloc_vm(void)
618 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
621 static inline void kvm_arch_free_vm(struct kvm *kvm)
627 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
629 #ifdef __KVM_HAVE_ARCH_WQP
630 return vcpu->arch.wqp;
636 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
637 void kvm_arch_destroy_vm(struct kvm *kvm);
638 void kvm_free_all_assigned_devices(struct kvm *kvm);
639 void kvm_arch_sync_events(struct kvm *kvm);
641 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
642 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
644 bool kvm_is_mmio_pfn(pfn_t pfn);
646 struct kvm_irq_ack_notifier {
647 struct hlist_node link;
649 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
652 struct kvm_assigned_dev_kernel {
653 struct kvm_irq_ack_notifier ack_notifier;
654 struct list_head list;
659 unsigned int entries_nr;
661 bool host_irq_disabled;
663 struct msix_entry *host_msix_entries;
665 struct msix_entry *guest_msix_entries;
666 unsigned long irq_requested_type;
671 spinlock_t intx_lock;
672 spinlock_t intx_mask_lock;
674 struct pci_saved_state *pci_saved_state;
677 struct kvm_irq_mask_notifier {
678 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
680 struct hlist_node link;
683 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
684 struct kvm_irq_mask_notifier *kimn);
685 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
686 struct kvm_irq_mask_notifier *kimn);
687 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
690 #ifdef __KVM_HAVE_IOAPIC
691 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
692 union kvm_ioapic_redirect_entry *entry,
693 unsigned long *deliver_bitmask);
695 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
696 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
697 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
698 int irq_source_id, int level);
699 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
700 void kvm_register_irq_ack_notifier(struct kvm *kvm,
701 struct kvm_irq_ack_notifier *kian);
702 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
703 struct kvm_irq_ack_notifier *kian);
704 int kvm_request_irq_source_id(struct kvm *kvm);
705 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
707 /* For vcpu->arch.iommu_flags */
708 #define KVM_IOMMU_CACHE_COHERENCY 0x1
710 #ifdef CONFIG_IOMMU_API
711 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
712 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
713 int kvm_iommu_map_guest(struct kvm *kvm);
714 int kvm_iommu_unmap_guest(struct kvm *kvm);
715 int kvm_assign_device(struct kvm *kvm,
716 struct kvm_assigned_dev_kernel *assigned_dev);
717 int kvm_deassign_device(struct kvm *kvm,
718 struct kvm_assigned_dev_kernel *assigned_dev);
719 #else /* CONFIG_IOMMU_API */
720 static inline int kvm_iommu_map_pages(struct kvm *kvm,
721 struct kvm_memory_slot *slot)
726 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
727 struct kvm_memory_slot *slot)
731 static inline int kvm_iommu_map_guest(struct kvm *kvm)
736 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
741 static inline int kvm_assign_device(struct kvm *kvm,
742 struct kvm_assigned_dev_kernel *assigned_dev)
747 static inline int kvm_deassign_device(struct kvm *kvm,
748 struct kvm_assigned_dev_kernel *assigned_dev)
752 #endif /* CONFIG_IOMMU_API */
754 static inline void kvm_guest_enter(void)
756 BUG_ON(preemptible());
757 vtime_account(current);
758 current->flags |= PF_VCPU;
759 /* KVM does not hold any references to rcu protected data when it
760 * switches CPU into a guest mode. In fact switching to a guest mode
761 * is very similar to exiting to userspase from rcu point of view. In
762 * addition CPU may stay in a guest mode for quite a long time (up to
763 * one time slice). Lets treat guest mode as quiescent state, just like
764 * we do with user-mode execution.
766 rcu_virt_note_context_switch(smp_processor_id());
769 static inline void kvm_guest_exit(void)
771 vtime_account(current);
772 current->flags &= ~PF_VCPU;
776 * search_memslots() and __gfn_to_memslot() are here because they are
777 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
778 * gfn_to_memslot() itself isn't here as an inline because that would
779 * bloat other code too much.
781 static inline struct kvm_memory_slot *
782 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
784 struct kvm_memory_slot *memslot;
786 kvm_for_each_memslot(memslot, slots)
787 if (gfn >= memslot->base_gfn &&
788 gfn < memslot->base_gfn + memslot->npages)
794 static inline struct kvm_memory_slot *
795 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
797 return search_memslots(slots, gfn);
800 static inline unsigned long
801 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
803 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
806 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
808 return gfn_to_memslot(kvm, gfn)->id;
811 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
813 /* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
814 return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
815 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
819 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
821 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
823 return slot->base_gfn + gfn_offset;
826 static inline gpa_t gfn_to_gpa(gfn_t gfn)
828 return (gpa_t)gfn << PAGE_SHIFT;
831 static inline gfn_t gpa_to_gfn(gpa_t gpa)
833 return (gfn_t)(gpa >> PAGE_SHIFT);
836 static inline hpa_t pfn_to_hpa(pfn_t pfn)
838 return (hpa_t)pfn << PAGE_SHIFT;
841 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
843 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
851 struct kvm_stats_debugfs_item {
854 enum kvm_stat_kind kind;
855 struct dentry *dentry;
857 extern struct kvm_stats_debugfs_item debugfs_entries[];
858 extern struct dentry *kvm_debugfs_dir;
860 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
861 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
863 if (unlikely(kvm->mmu_notifier_count))
866 * Ensure the read of mmu_notifier_count happens before the read
867 * of mmu_notifier_seq. This interacts with the smp_wmb() in
868 * mmu_notifier_invalidate_range_end to make sure that the caller
869 * either sees the old (non-zero) value of mmu_notifier_count or
870 * the new (incremented) value of mmu_notifier_seq.
871 * PowerPC Book3s HV KVM calls this under a per-page lock
872 * rather than under kvm->mmu_lock, for scalability, so
873 * can't rely on kvm->mmu_lock to keep things ordered.
876 if (kvm->mmu_notifier_seq != mmu_seq)
882 #ifdef KVM_CAP_IRQ_ROUTING
884 #define KVM_MAX_IRQ_ROUTES 1024
886 int kvm_setup_default_irq_routing(struct kvm *kvm);
887 int kvm_set_irq_routing(struct kvm *kvm,
888 const struct kvm_irq_routing_entry *entries,
891 void kvm_free_irq_routing(struct kvm *kvm);
893 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
897 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
901 #ifdef CONFIG_HAVE_KVM_EVENTFD
903 void kvm_eventfd_init(struct kvm *kvm);
904 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
906 #ifdef CONFIG_HAVE_KVM_IRQCHIP
907 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
908 void kvm_irqfd_release(struct kvm *kvm);
909 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
911 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
916 static inline void kvm_irqfd_release(struct kvm *kvm) {}
921 static inline void kvm_eventfd_init(struct kvm *kvm) {}
923 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
928 static inline void kvm_irqfd_release(struct kvm *kvm) {}
930 #ifdef CONFIG_HAVE_KVM_IRQCHIP
931 static inline void kvm_irq_routing_update(struct kvm *kvm,
932 struct kvm_irq_routing_table *irq_rt)
934 rcu_assign_pointer(kvm->irq_routing, irq_rt);
938 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
943 #endif /* CONFIG_HAVE_KVM_EVENTFD */
945 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
946 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
948 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
951 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
955 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
959 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
961 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
966 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
974 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
976 set_bit(req, &vcpu->requests);
979 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
981 if (test_bit(req, &vcpu->requests)) {
982 clear_bit(req, &vcpu->requests);
989 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
991 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
993 vcpu->spin_loop.in_spin_loop = val;
995 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
997 vcpu->spin_loop.dy_eligible = val;
1000 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1002 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1006 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1010 static inline bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
1015 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */