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 <asm/signal.h>
28 #include <linux/kvm.h>
29 #include <linux/kvm_para.h>
31 #include <linux/kvm_types.h>
33 #include <asm/kvm_host.h>
36 #define KVM_MMIO_SIZE 8
40 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
41 * in kvm, other bits are visible for userspace which are defined in
42 * include/linux/kvm_h.
44 #define KVM_MEMSLOT_INVALID (1UL << 16)
46 /* Two fragments for cross MMIO pages. */
47 #define KVM_MAX_MMIO_FRAGMENTS 2
50 * For the normal pfn, the highest 12 bits should be zero,
51 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
52 * mask bit 63 to indicate the noslot pfn.
54 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
55 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
56 #define KVM_PFN_NOSLOT (0x1ULL << 63)
58 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
59 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
60 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
63 * error pfns indicate that the gfn is in slot but faild to
64 * translate it to pfn on host.
66 static inline bool is_error_pfn(pfn_t pfn)
68 return !!(pfn & KVM_PFN_ERR_MASK);
72 * error_noslot pfns indicate that the gfn can not be
73 * translated to pfn - it is not in slot or failed to
74 * translate it to pfn.
76 static inline bool is_error_noslot_pfn(pfn_t pfn)
78 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
81 /* noslot pfn indicates that the gfn is not in slot. */
82 static inline bool is_noslot_pfn(pfn_t pfn)
84 return pfn == KVM_PFN_NOSLOT;
87 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
88 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
90 static inline bool kvm_is_error_hva(unsigned long addr)
92 return addr >= PAGE_OFFSET;
95 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
97 static inline bool is_error_page(struct page *page)
103 * vcpu->requests bit members
105 #define KVM_REQ_TLB_FLUSH 0
106 #define KVM_REQ_MIGRATE_TIMER 1
107 #define KVM_REQ_REPORT_TPR_ACCESS 2
108 #define KVM_REQ_MMU_RELOAD 3
109 #define KVM_REQ_TRIPLE_FAULT 4
110 #define KVM_REQ_PENDING_TIMER 5
111 #define KVM_REQ_UNHALT 6
112 #define KVM_REQ_MMU_SYNC 7
113 #define KVM_REQ_CLOCK_UPDATE 8
114 #define KVM_REQ_KICK 9
115 #define KVM_REQ_DEACTIVATE_FPU 10
116 #define KVM_REQ_EVENT 11
117 #define KVM_REQ_APF_HALT 12
118 #define KVM_REQ_STEAL_UPDATE 13
119 #define KVM_REQ_NMI 14
120 #define KVM_REQ_IMMEDIATE_EXIT 15
121 #define KVM_REQ_PMU 16
122 #define KVM_REQ_PMI 17
123 #define KVM_REQ_WATCHDOG 18
124 #define KVM_REQ_MASTERCLOCK_UPDATE 19
125 #define KVM_REQ_MCLOCK_INPROGRESS 20
126 #define KVM_REQ_EPR_EXIT 21
127 #define KVM_REQ_EOIBITMAP 22
129 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
130 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
134 extern struct kmem_cache *kvm_vcpu_cache;
136 struct kvm_io_range {
139 struct kvm_io_device *dev;
142 #define NR_IOBUS_DEVS 1000
146 struct kvm_io_range range[];
152 KVM_VIRTIO_CCW_NOTIFY_BUS,
156 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
157 int len, const void *val);
158 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
160 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
161 int len, struct kvm_io_device *dev);
162 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
163 struct kvm_io_device *dev);
165 #ifdef CONFIG_KVM_ASYNC_PF
166 struct kvm_async_pf {
167 struct work_struct work;
168 struct list_head link;
169 struct list_head queue;
170 struct kvm_vcpu *vcpu;
171 struct mm_struct *mm;
174 struct kvm_arch_async_pf arch;
179 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
180 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
181 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
182 struct kvm_arch_async_pf *arch);
183 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
190 READING_SHADOW_PAGE_TABLES,
194 * Sometimes a large or cross-page mmio needs to be broken up into separate
195 * exits for userspace servicing.
197 struct kvm_mmio_fragment {
205 #ifdef CONFIG_PREEMPT_NOTIFIERS
206 struct preempt_notifier preempt_notifier;
212 unsigned long requests;
213 unsigned long guest_debug;
219 int guest_fpu_loaded, guest_xcr0_loaded;
220 wait_queue_head_t wq;
224 struct kvm_vcpu_stat stat;
226 #ifdef CONFIG_HAS_IOMEM
228 int mmio_read_completed;
230 int mmio_cur_fragment;
231 int mmio_nr_fragments;
232 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
235 #ifdef CONFIG_KVM_ASYNC_PF
238 struct list_head queue;
239 struct list_head done;
244 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
246 * Cpu relax intercept or pause loop exit optimization
247 * in_spin_loop: set when a vcpu does a pause loop exit
248 * or cpu relax intercepted.
249 * dy_eligible: indicates whether vcpu is eligible for directed yield.
257 struct kvm_vcpu_arch arch;
260 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
262 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
266 * Some of the bitops functions do not support too long bitmaps.
267 * This number must be determined not to exceed such limits.
269 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
271 struct kvm_memory_slot {
273 unsigned long npages;
274 unsigned long *dirty_bitmap;
275 struct kvm_arch_memory_slot arch;
276 unsigned long userspace_addr;
281 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
283 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
286 struct kvm_kernel_irq_routing_entry {
289 int (*set)(struct kvm_kernel_irq_routing_entry *e,
290 struct kvm *kvm, int irq_source_id, int level);
298 struct hlist_node link;
301 #ifdef __KVM_HAVE_IOAPIC
303 struct kvm_irq_routing_table {
304 int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
305 struct kvm_kernel_irq_routing_entry *rt_entries;
308 * Array indexed by gsi. Each entry contains list of irq chips
309 * the gsi is connected to.
311 struct hlist_head map[0];
316 struct kvm_irq_routing_table {};
320 #ifndef KVM_PRIVATE_MEM_SLOTS
321 #define KVM_PRIVATE_MEM_SLOTS 0
324 #ifndef KVM_MEM_SLOTS_NUM
325 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
330 * memslots are not sorted by id anymore, please use id_to_memslot()
331 * to get the memslot by its id.
333 struct kvm_memslots {
335 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
336 /* The mapping table from slot id to the index in memslots[]. */
337 short id_to_index[KVM_MEM_SLOTS_NUM];
342 struct mutex slots_lock;
343 struct mm_struct *mm; /* userspace tied to this vm */
344 struct kvm_memslots *memslots;
345 struct srcu_struct srcu;
346 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
349 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
350 atomic_t online_vcpus;
351 int last_boosted_vcpu;
352 struct list_head vm_list;
354 struct kvm_io_bus *buses[KVM_NR_BUSES];
355 #ifdef CONFIG_HAVE_KVM_EVENTFD
358 struct list_head items;
359 struct list_head resampler_list;
360 struct mutex resampler_lock;
362 struct list_head ioeventfds;
364 struct kvm_vm_stat stat;
365 struct kvm_arch arch;
366 atomic_t users_count;
367 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
368 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
369 spinlock_t ring_lock;
370 struct list_head coalesced_zones;
373 struct mutex irq_lock;
374 #ifdef CONFIG_HAVE_KVM_IRQCHIP
376 * Update side is protected by irq_lock and,
377 * if configured, irqfds.lock.
379 struct kvm_irq_routing_table __rcu *irq_routing;
380 struct hlist_head mask_notifier_list;
381 struct hlist_head irq_ack_notifier_list;
384 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
385 struct mmu_notifier mmu_notifier;
386 unsigned long mmu_notifier_seq;
387 long mmu_notifier_count;
392 #define kvm_err(fmt, ...) \
393 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
394 #define kvm_info(fmt, ...) \
395 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
396 #define kvm_debug(fmt, ...) \
397 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
398 #define kvm_pr_unimpl(fmt, ...) \
399 pr_err_ratelimited("kvm [%i]: " fmt, \
400 task_tgid_nr(current), ## __VA_ARGS__)
402 /* The guest did something we don't support. */
403 #define vcpu_unimpl(vcpu, fmt, ...) \
404 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
406 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
409 return kvm->vcpus[i];
412 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
414 idx < atomic_read(&kvm->online_vcpus) && \
415 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
418 #define kvm_for_each_memslot(memslot, slots) \
419 for (memslot = &slots->memslots[0]; \
420 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
423 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
424 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
426 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
427 void vcpu_put(struct kvm_vcpu *vcpu);
429 #ifdef __KVM_HAVE_IOAPIC
430 int kvm_irqfd_init(void);
431 void kvm_irqfd_exit(void);
433 static inline int kvm_irqfd_init(void)
438 static inline void kvm_irqfd_exit(void)
442 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
443 struct module *module);
446 void kvm_get_kvm(struct kvm *kvm);
447 void kvm_put_kvm(struct kvm *kvm);
448 void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new,
449 u64 last_generation);
451 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
453 return rcu_dereference_check(kvm->memslots,
454 srcu_read_lock_held(&kvm->srcu)
455 || lockdep_is_held(&kvm->slots_lock));
458 static inline struct kvm_memory_slot *
459 id_to_memslot(struct kvm_memslots *slots, int id)
461 int index = slots->id_to_index[id];
462 struct kvm_memory_slot *slot;
464 slot = &slots->memslots[index];
466 WARN_ON(slot->id != id);
471 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
472 * - create a new memory slot
473 * - delete an existing memory slot
474 * - modify an existing memory slot
475 * -- move it in the guest physical memory space
476 * -- just change its flags
478 * Since flags can be changed by some of these operations, the following
479 * differentiation is the best we can do for __kvm_set_memory_region():
488 int kvm_set_memory_region(struct kvm *kvm,
489 struct kvm_userspace_memory_region *mem);
490 int __kvm_set_memory_region(struct kvm *kvm,
491 struct kvm_userspace_memory_region *mem);
492 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
493 struct kvm_memory_slot *dont);
494 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages);
495 int kvm_arch_prepare_memory_region(struct kvm *kvm,
496 struct kvm_memory_slot *memslot,
497 struct kvm_userspace_memory_region *mem,
498 enum kvm_mr_change change);
499 void kvm_arch_commit_memory_region(struct kvm *kvm,
500 struct kvm_userspace_memory_region *mem,
501 const struct kvm_memory_slot *old,
502 enum kvm_mr_change change);
503 bool kvm_largepages_enabled(void);
504 void kvm_disable_largepages(void);
505 /* flush all memory translations */
506 void kvm_arch_flush_shadow_all(struct kvm *kvm);
507 /* flush memory translations pointing to 'slot' */
508 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
509 struct kvm_memory_slot *slot);
511 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
514 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
515 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
516 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
517 void kvm_release_page_clean(struct page *page);
518 void kvm_release_page_dirty(struct page *page);
519 void kvm_set_page_dirty(struct page *page);
520 void kvm_set_page_accessed(struct page *page);
522 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
523 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
524 bool write_fault, bool *writable);
525 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
526 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
528 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
529 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
531 void kvm_release_pfn_dirty(pfn_t pfn);
532 void kvm_release_pfn_clean(pfn_t pfn);
533 void kvm_set_pfn_dirty(pfn_t pfn);
534 void kvm_set_pfn_accessed(pfn_t pfn);
535 void kvm_get_pfn(pfn_t pfn);
537 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
539 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
541 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
542 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
543 void *data, unsigned long len);
544 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
545 int offset, int len);
546 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
548 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
549 void *data, unsigned long len);
550 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
552 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
553 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
554 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
555 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
556 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
557 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
558 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
561 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
562 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
563 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
564 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
565 void kvm_resched(struct kvm_vcpu *vcpu);
566 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
567 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
569 void kvm_flush_remote_tlbs(struct kvm *kvm);
570 void kvm_reload_remote_mmus(struct kvm *kvm);
571 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
572 void kvm_make_update_eoibitmap_request(struct kvm *kvm);
574 long kvm_arch_dev_ioctl(struct file *filp,
575 unsigned int ioctl, unsigned long arg);
576 long kvm_arch_vcpu_ioctl(struct file *filp,
577 unsigned int ioctl, unsigned long arg);
578 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
580 int kvm_dev_ioctl_check_extension(long ext);
582 int kvm_get_dirty_log(struct kvm *kvm,
583 struct kvm_dirty_log *log, int *is_dirty);
584 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
585 struct kvm_dirty_log *log);
587 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
588 struct kvm_userspace_memory_region *mem);
589 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level);
590 long kvm_arch_vm_ioctl(struct file *filp,
591 unsigned int ioctl, unsigned long arg);
593 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
594 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
596 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
597 struct kvm_translation *tr);
599 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
600 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
601 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
602 struct kvm_sregs *sregs);
603 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
604 struct kvm_sregs *sregs);
605 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
606 struct kvm_mp_state *mp_state);
607 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
608 struct kvm_mp_state *mp_state);
609 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
610 struct kvm_guest_debug *dbg);
611 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
613 int kvm_arch_init(void *opaque);
614 void kvm_arch_exit(void);
616 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
617 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
619 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
620 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
621 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
622 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
623 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
624 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
625 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
627 int kvm_arch_hardware_enable(void *garbage);
628 void kvm_arch_hardware_disable(void *garbage);
629 int kvm_arch_hardware_setup(void);
630 void kvm_arch_hardware_unsetup(void);
631 void kvm_arch_check_processor_compat(void *rtn);
632 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
633 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
635 void kvm_free_physmem(struct kvm *kvm);
637 void *kvm_kvzalloc(unsigned long size);
638 void kvm_kvfree(const void *addr);
640 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
641 static inline struct kvm *kvm_arch_alloc_vm(void)
643 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
646 static inline void kvm_arch_free_vm(struct kvm *kvm)
652 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
654 #ifdef __KVM_HAVE_ARCH_WQP
655 return vcpu->arch.wqp;
661 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
662 void kvm_arch_destroy_vm(struct kvm *kvm);
663 void kvm_free_all_assigned_devices(struct kvm *kvm);
664 void kvm_arch_sync_events(struct kvm *kvm);
666 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
667 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
669 bool kvm_is_mmio_pfn(pfn_t pfn);
671 struct kvm_irq_ack_notifier {
672 struct hlist_node link;
674 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
677 struct kvm_assigned_dev_kernel {
678 struct kvm_irq_ack_notifier ack_notifier;
679 struct list_head list;
684 unsigned int entries_nr;
686 bool host_irq_disabled;
688 struct msix_entry *host_msix_entries;
690 struct msix_entry *guest_msix_entries;
691 unsigned long irq_requested_type;
696 spinlock_t intx_lock;
697 spinlock_t intx_mask_lock;
699 struct pci_saved_state *pci_saved_state;
702 struct kvm_irq_mask_notifier {
703 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
705 struct hlist_node link;
708 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
709 struct kvm_irq_mask_notifier *kimn);
710 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
711 struct kvm_irq_mask_notifier *kimn);
712 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
715 #ifdef __KVM_HAVE_IOAPIC
716 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
717 union kvm_ioapic_redirect_entry *entry,
718 unsigned long *deliver_bitmask);
720 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
721 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
722 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
723 int irq_source_id, int level);
724 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
725 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
726 void kvm_register_irq_ack_notifier(struct kvm *kvm,
727 struct kvm_irq_ack_notifier *kian);
728 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
729 struct kvm_irq_ack_notifier *kian);
730 int kvm_request_irq_source_id(struct kvm *kvm);
731 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
733 /* For vcpu->arch.iommu_flags */
734 #define KVM_IOMMU_CACHE_COHERENCY 0x1
736 #ifdef CONFIG_IOMMU_API
737 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
738 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
739 int kvm_iommu_map_guest(struct kvm *kvm);
740 int kvm_iommu_unmap_guest(struct kvm *kvm);
741 int kvm_assign_device(struct kvm *kvm,
742 struct kvm_assigned_dev_kernel *assigned_dev);
743 int kvm_deassign_device(struct kvm *kvm,
744 struct kvm_assigned_dev_kernel *assigned_dev);
745 #else /* CONFIG_IOMMU_API */
746 static inline int kvm_iommu_map_pages(struct kvm *kvm,
747 struct kvm_memory_slot *slot)
752 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
753 struct kvm_memory_slot *slot)
757 static inline int kvm_iommu_map_guest(struct kvm *kvm)
762 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
767 static inline int kvm_assign_device(struct kvm *kvm,
768 struct kvm_assigned_dev_kernel *assigned_dev)
773 static inline int kvm_deassign_device(struct kvm *kvm,
774 struct kvm_assigned_dev_kernel *assigned_dev)
778 #endif /* CONFIG_IOMMU_API */
780 static inline void __guest_enter(void)
783 * This is running in ioctl context so we can avoid
784 * the call to vtime_account() with its unnecessary idle check.
786 vtime_account_system(current);
787 current->flags |= PF_VCPU;
790 static inline void __guest_exit(void)
793 * This is running in ioctl context so we can avoid
794 * the call to vtime_account() with its unnecessary idle check.
796 vtime_account_system(current);
797 current->flags &= ~PF_VCPU;
800 #ifdef CONFIG_CONTEXT_TRACKING
801 extern void guest_enter(void);
802 extern void guest_exit(void);
804 #else /* !CONFIG_CONTEXT_TRACKING */
805 static inline void guest_enter(void)
810 static inline void guest_exit(void)
814 #endif /* !CONFIG_CONTEXT_TRACKING */
816 static inline void kvm_guest_enter(void)
820 BUG_ON(preemptible());
822 local_irq_save(flags);
824 local_irq_restore(flags);
826 /* KVM does not hold any references to rcu protected data when it
827 * switches CPU into a guest mode. In fact switching to a guest mode
828 * is very similar to exiting to userspase from rcu point of view. In
829 * addition CPU may stay in a guest mode for quite a long time (up to
830 * one time slice). Lets treat guest mode as quiescent state, just like
831 * we do with user-mode execution.
833 rcu_virt_note_context_switch(smp_processor_id());
836 static inline void kvm_guest_exit(void)
840 local_irq_save(flags);
842 local_irq_restore(flags);
846 * search_memslots() and __gfn_to_memslot() are here because they are
847 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
848 * gfn_to_memslot() itself isn't here as an inline because that would
849 * bloat other code too much.
851 static inline struct kvm_memory_slot *
852 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
854 struct kvm_memory_slot *memslot;
856 kvm_for_each_memslot(memslot, slots)
857 if (gfn >= memslot->base_gfn &&
858 gfn < memslot->base_gfn + memslot->npages)
864 static inline struct kvm_memory_slot *
865 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
867 return search_memslots(slots, gfn);
870 static inline unsigned long
871 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
873 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
876 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
878 return gfn_to_memslot(kvm, gfn)->id;
881 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
883 /* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
884 return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
885 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
889 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
891 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
893 return slot->base_gfn + gfn_offset;
896 static inline gpa_t gfn_to_gpa(gfn_t gfn)
898 return (gpa_t)gfn << PAGE_SHIFT;
901 static inline gfn_t gpa_to_gfn(gpa_t gpa)
903 return (gfn_t)(gpa >> PAGE_SHIFT);
906 static inline hpa_t pfn_to_hpa(pfn_t pfn)
908 return (hpa_t)pfn << PAGE_SHIFT;
911 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
913 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
921 struct kvm_stats_debugfs_item {
924 enum kvm_stat_kind kind;
925 struct dentry *dentry;
927 extern struct kvm_stats_debugfs_item debugfs_entries[];
928 extern struct dentry *kvm_debugfs_dir;
930 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
931 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
933 if (unlikely(kvm->mmu_notifier_count))
936 * Ensure the read of mmu_notifier_count happens before the read
937 * of mmu_notifier_seq. This interacts with the smp_wmb() in
938 * mmu_notifier_invalidate_range_end to make sure that the caller
939 * either sees the old (non-zero) value of mmu_notifier_count or
940 * the new (incremented) value of mmu_notifier_seq.
941 * PowerPC Book3s HV KVM calls this under a per-page lock
942 * rather than under kvm->mmu_lock, for scalability, so
943 * can't rely on kvm->mmu_lock to keep things ordered.
946 if (kvm->mmu_notifier_seq != mmu_seq)
952 #ifdef KVM_CAP_IRQ_ROUTING
954 #define KVM_MAX_IRQ_ROUTES 1024
956 int kvm_setup_default_irq_routing(struct kvm *kvm);
957 int kvm_set_irq_routing(struct kvm *kvm,
958 const struct kvm_irq_routing_entry *entries,
961 void kvm_free_irq_routing(struct kvm *kvm);
963 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
967 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
971 #ifdef CONFIG_HAVE_KVM_EVENTFD
973 void kvm_eventfd_init(struct kvm *kvm);
974 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
976 #ifdef CONFIG_HAVE_KVM_IRQCHIP
977 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
978 void kvm_irqfd_release(struct kvm *kvm);
979 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
981 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
986 static inline void kvm_irqfd_release(struct kvm *kvm) {}
991 static inline void kvm_eventfd_init(struct kvm *kvm) {}
993 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
998 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1000 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1001 static inline void kvm_irq_routing_update(struct kvm *kvm,
1002 struct kvm_irq_routing_table *irq_rt)
1004 rcu_assign_pointer(kvm->irq_routing, irq_rt);
1008 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1013 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1015 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
1016 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
1018 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
1021 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
1025 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
1029 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
1031 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
1036 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
1044 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1046 set_bit(req, &vcpu->requests);
1049 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1051 if (test_bit(req, &vcpu->requests)) {
1052 clear_bit(req, &vcpu->requests);
1059 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1061 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1063 vcpu->spin_loop.in_spin_loop = val;
1065 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1067 vcpu->spin_loop.dy_eligible = val;
1070 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1072 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1076 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1080 static inline bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
1085 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */