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 these bits to indicate the error.
63 #define KVM_PFN_ERR_MASK (0xfffULL << 52)
65 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
66 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
67 #define KVM_PFN_ERR_BAD (KVM_PFN_ERR_MASK + 2)
68 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 3)
70 static inline bool is_error_pfn(pfn_t pfn)
72 return !!(pfn & KVM_PFN_ERR_MASK);
75 static inline bool is_noslot_pfn(pfn_t pfn)
77 return pfn == KVM_PFN_ERR_BAD;
80 static inline bool is_invalid_pfn(pfn_t pfn)
82 return !is_noslot_pfn(pfn) && is_error_pfn(pfn);
85 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
87 static inline bool is_error_page(struct page *page)
93 * vcpu->requests bit members
95 #define KVM_REQ_TLB_FLUSH 0
96 #define KVM_REQ_MIGRATE_TIMER 1
97 #define KVM_REQ_REPORT_TPR_ACCESS 2
98 #define KVM_REQ_MMU_RELOAD 3
99 #define KVM_REQ_TRIPLE_FAULT 4
100 #define KVM_REQ_PENDING_TIMER 5
101 #define KVM_REQ_UNHALT 6
102 #define KVM_REQ_MMU_SYNC 7
103 #define KVM_REQ_CLOCK_UPDATE 8
104 #define KVM_REQ_KICK 9
105 #define KVM_REQ_DEACTIVATE_FPU 10
106 #define KVM_REQ_EVENT 11
107 #define KVM_REQ_APF_HALT 12
108 #define KVM_REQ_STEAL_UPDATE 13
109 #define KVM_REQ_NMI 14
110 #define KVM_REQ_IMMEDIATE_EXIT 15
111 #define KVM_REQ_PMU 16
112 #define KVM_REQ_PMI 17
114 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
118 extern struct kmem_cache *kvm_vcpu_cache;
120 struct kvm_io_range {
123 struct kvm_io_device *dev;
126 #define NR_IOBUS_DEVS 1000
130 struct kvm_io_range range[];
139 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
140 int len, const void *val);
141 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
143 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
144 int len, struct kvm_io_device *dev);
145 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
146 struct kvm_io_device *dev);
148 #ifdef CONFIG_KVM_ASYNC_PF
149 struct kvm_async_pf {
150 struct work_struct work;
151 struct list_head link;
152 struct list_head queue;
153 struct kvm_vcpu *vcpu;
154 struct mm_struct *mm;
157 struct kvm_arch_async_pf arch;
162 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
163 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
164 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
165 struct kvm_arch_async_pf *arch);
166 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
173 READING_SHADOW_PAGE_TABLES,
177 * Sometimes a large or cross-page mmio needs to be broken up into separate
178 * exits for userspace servicing.
180 struct kvm_mmio_fragment {
188 #ifdef CONFIG_PREEMPT_NOTIFIERS
189 struct preempt_notifier preempt_notifier;
195 unsigned long requests;
196 unsigned long guest_debug;
202 int guest_fpu_loaded, guest_xcr0_loaded;
203 wait_queue_head_t wq;
207 struct kvm_vcpu_stat stat;
209 #ifdef CONFIG_HAS_IOMEM
211 int mmio_read_completed;
213 int mmio_cur_fragment;
214 int mmio_nr_fragments;
215 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
218 #ifdef CONFIG_KVM_ASYNC_PF
221 struct list_head queue;
222 struct list_head done;
227 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
229 * Cpu relax intercept or pause loop exit optimization
230 * in_spin_loop: set when a vcpu does a pause loop exit
231 * or cpu relax intercepted.
232 * dy_eligible: indicates whether vcpu is eligible for directed yield.
239 struct kvm_vcpu_arch arch;
242 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
244 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
248 * Some of the bitops functions do not support too long bitmaps.
249 * This number must be determined not to exceed such limits.
251 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
253 struct kvm_memory_slot {
255 unsigned long npages;
257 unsigned long *dirty_bitmap;
258 struct kvm_arch_memory_slot arch;
259 unsigned long userspace_addr;
264 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
266 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
269 struct kvm_kernel_irq_routing_entry {
272 int (*set)(struct kvm_kernel_irq_routing_entry *e,
273 struct kvm *kvm, int irq_source_id, int level);
281 struct hlist_node link;
284 #ifdef __KVM_HAVE_IOAPIC
286 struct kvm_irq_routing_table {
287 int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
288 struct kvm_kernel_irq_routing_entry *rt_entries;
291 * Array indexed by gsi. Each entry contains list of irq chips
292 * the gsi is connected to.
294 struct hlist_head map[0];
299 struct kvm_irq_routing_table {};
303 #ifndef KVM_MEM_SLOTS_NUM
304 #define KVM_MEM_SLOTS_NUM (KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
309 * memslots are not sorted by id anymore, please use id_to_memslot()
310 * to get the memslot by its id.
312 struct kvm_memslots {
314 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
315 /* The mapping table from slot id to the index in memslots[]. */
316 int id_to_index[KVM_MEM_SLOTS_NUM];
321 struct mutex slots_lock;
322 struct mm_struct *mm; /* userspace tied to this vm */
323 struct kvm_memslots *memslots;
324 struct srcu_struct srcu;
325 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
328 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
329 atomic_t online_vcpus;
330 int last_boosted_vcpu;
331 struct list_head vm_list;
333 struct kvm_io_bus *buses[KVM_NR_BUSES];
334 #ifdef CONFIG_HAVE_KVM_EVENTFD
337 struct list_head items;
339 struct list_head ioeventfds;
341 struct kvm_vm_stat stat;
342 struct kvm_arch arch;
343 atomic_t users_count;
344 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
345 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
346 spinlock_t ring_lock;
347 struct list_head coalesced_zones;
350 struct mutex irq_lock;
351 #ifdef CONFIG_HAVE_KVM_IRQCHIP
353 * Update side is protected by irq_lock and,
354 * if configured, irqfds.lock.
356 struct kvm_irq_routing_table __rcu *irq_routing;
357 struct hlist_head mask_notifier_list;
358 struct hlist_head irq_ack_notifier_list;
361 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
362 struct mmu_notifier mmu_notifier;
363 unsigned long mmu_notifier_seq;
364 long mmu_notifier_count;
369 #define kvm_err(fmt, ...) \
370 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
371 #define kvm_info(fmt, ...) \
372 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
373 #define kvm_debug(fmt, ...) \
374 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
375 #define kvm_pr_unimpl(fmt, ...) \
376 pr_err_ratelimited("kvm [%i]: " fmt, \
377 task_tgid_nr(current), ## __VA_ARGS__)
379 /* The guest did something we don't support. */
380 #define vcpu_unimpl(vcpu, fmt, ...) \
381 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
383 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
386 return kvm->vcpus[i];
389 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
391 idx < atomic_read(&kvm->online_vcpus) && \
392 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
395 #define kvm_for_each_memslot(memslot, slots) \
396 for (memslot = &slots->memslots[0]; \
397 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
400 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
401 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
403 void vcpu_load(struct kvm_vcpu *vcpu);
404 void vcpu_put(struct kvm_vcpu *vcpu);
406 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
407 struct module *module);
410 void kvm_get_kvm(struct kvm *kvm);
411 void kvm_put_kvm(struct kvm *kvm);
412 void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new);
414 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
416 return rcu_dereference_check(kvm->memslots,
417 srcu_read_lock_held(&kvm->srcu)
418 || lockdep_is_held(&kvm->slots_lock));
421 static inline struct kvm_memory_slot *
422 id_to_memslot(struct kvm_memslots *slots, int id)
424 int index = slots->id_to_index[id];
425 struct kvm_memory_slot *slot;
427 slot = &slots->memslots[index];
429 WARN_ON(slot->id != id);
433 int kvm_is_error_hva(unsigned long addr);
434 int kvm_set_memory_region(struct kvm *kvm,
435 struct kvm_userspace_memory_region *mem,
437 int __kvm_set_memory_region(struct kvm *kvm,
438 struct kvm_userspace_memory_region *mem,
440 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
441 struct kvm_memory_slot *dont);
442 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages);
443 int kvm_arch_prepare_memory_region(struct kvm *kvm,
444 struct kvm_memory_slot *memslot,
445 struct kvm_memory_slot old,
446 struct kvm_userspace_memory_region *mem,
448 void kvm_arch_commit_memory_region(struct kvm *kvm,
449 struct kvm_userspace_memory_region *mem,
450 struct kvm_memory_slot old,
452 bool kvm_largepages_enabled(void);
453 void kvm_disable_largepages(void);
454 void kvm_arch_flush_shadow(struct kvm *kvm);
456 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
459 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
460 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
461 void kvm_release_page_clean(struct page *page);
462 void kvm_release_page_dirty(struct page *page);
463 void kvm_set_page_dirty(struct page *page);
464 void kvm_set_page_accessed(struct page *page);
466 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
467 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
468 bool write_fault, bool *writable);
469 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
470 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
472 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
473 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
475 void kvm_release_pfn_dirty(pfn_t pfn);
476 void kvm_release_pfn_clean(pfn_t pfn);
477 void kvm_set_pfn_dirty(pfn_t pfn);
478 void kvm_set_pfn_accessed(pfn_t pfn);
479 void kvm_get_pfn(pfn_t pfn);
481 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
483 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
485 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
486 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
487 void *data, unsigned long len);
488 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
489 int offset, int len);
490 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
492 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
493 void *data, unsigned long len);
494 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
496 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
497 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
498 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
499 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
500 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
501 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
502 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
505 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
506 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
507 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
508 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
509 void kvm_resched(struct kvm_vcpu *vcpu);
510 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
511 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
513 void kvm_flush_remote_tlbs(struct kvm *kvm);
514 void kvm_reload_remote_mmus(struct kvm *kvm);
516 long kvm_arch_dev_ioctl(struct file *filp,
517 unsigned int ioctl, unsigned long arg);
518 long kvm_arch_vcpu_ioctl(struct file *filp,
519 unsigned int ioctl, unsigned long arg);
520 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
522 int kvm_dev_ioctl_check_extension(long ext);
524 int kvm_get_dirty_log(struct kvm *kvm,
525 struct kvm_dirty_log *log, int *is_dirty);
526 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
527 struct kvm_dirty_log *log);
529 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
531 kvm_userspace_memory_region *mem,
533 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level);
534 long kvm_arch_vm_ioctl(struct file *filp,
535 unsigned int ioctl, unsigned long arg);
537 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
538 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
540 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
541 struct kvm_translation *tr);
543 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
544 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
545 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
546 struct kvm_sregs *sregs);
547 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
548 struct kvm_sregs *sregs);
549 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
550 struct kvm_mp_state *mp_state);
551 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
552 struct kvm_mp_state *mp_state);
553 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
554 struct kvm_guest_debug *dbg);
555 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
557 int kvm_arch_init(void *opaque);
558 void kvm_arch_exit(void);
560 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
561 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
563 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
564 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
565 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
566 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
567 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
568 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
570 int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu);
571 int kvm_arch_hardware_enable(void *garbage);
572 void kvm_arch_hardware_disable(void *garbage);
573 int kvm_arch_hardware_setup(void);
574 void kvm_arch_hardware_unsetup(void);
575 void kvm_arch_check_processor_compat(void *rtn);
576 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
577 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
579 void kvm_free_physmem(struct kvm *kvm);
581 void *kvm_kvzalloc(unsigned long size);
582 void kvm_kvfree(const void *addr);
584 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
585 static inline struct kvm *kvm_arch_alloc_vm(void)
587 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
590 static inline void kvm_arch_free_vm(struct kvm *kvm)
596 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
598 #ifdef __KVM_HAVE_ARCH_WQP
599 return vcpu->arch.wqp;
605 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
606 void kvm_arch_destroy_vm(struct kvm *kvm);
607 void kvm_free_all_assigned_devices(struct kvm *kvm);
608 void kvm_arch_sync_events(struct kvm *kvm);
610 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
611 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
613 bool kvm_is_mmio_pfn(pfn_t pfn);
615 struct kvm_irq_ack_notifier {
616 struct hlist_node link;
618 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
621 struct kvm_assigned_dev_kernel {
622 struct kvm_irq_ack_notifier ack_notifier;
623 struct list_head list;
628 unsigned int entries_nr;
630 bool host_irq_disabled;
632 struct msix_entry *host_msix_entries;
634 struct msix_entry *guest_msix_entries;
635 unsigned long irq_requested_type;
640 spinlock_t intx_lock;
641 spinlock_t intx_mask_lock;
643 struct pci_saved_state *pci_saved_state;
646 struct kvm_irq_mask_notifier {
647 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
649 struct hlist_node link;
652 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
653 struct kvm_irq_mask_notifier *kimn);
654 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
655 struct kvm_irq_mask_notifier *kimn);
656 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
659 #ifdef __KVM_HAVE_IOAPIC
660 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
661 union kvm_ioapic_redirect_entry *entry,
662 unsigned long *deliver_bitmask);
664 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
665 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
666 int irq_source_id, int level);
667 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
668 void kvm_register_irq_ack_notifier(struct kvm *kvm,
669 struct kvm_irq_ack_notifier *kian);
670 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
671 struct kvm_irq_ack_notifier *kian);
672 int kvm_request_irq_source_id(struct kvm *kvm);
673 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
675 /* For vcpu->arch.iommu_flags */
676 #define KVM_IOMMU_CACHE_COHERENCY 0x1
678 #ifdef CONFIG_IOMMU_API
679 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
680 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
681 int kvm_iommu_map_guest(struct kvm *kvm);
682 int kvm_iommu_unmap_guest(struct kvm *kvm);
683 int kvm_assign_device(struct kvm *kvm,
684 struct kvm_assigned_dev_kernel *assigned_dev);
685 int kvm_deassign_device(struct kvm *kvm,
686 struct kvm_assigned_dev_kernel *assigned_dev);
687 #else /* CONFIG_IOMMU_API */
688 static inline int kvm_iommu_map_pages(struct kvm *kvm,
689 struct kvm_memory_slot *slot)
694 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
695 struct kvm_memory_slot *slot)
699 static inline int kvm_iommu_map_guest(struct kvm *kvm)
704 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
709 static inline int kvm_assign_device(struct kvm *kvm,
710 struct kvm_assigned_dev_kernel *assigned_dev)
715 static inline int kvm_deassign_device(struct kvm *kvm,
716 struct kvm_assigned_dev_kernel *assigned_dev)
720 #endif /* CONFIG_IOMMU_API */
722 static inline void kvm_guest_enter(void)
724 BUG_ON(preemptible());
725 account_system_vtime(current);
726 current->flags |= PF_VCPU;
727 /* KVM does not hold any references to rcu protected data when it
728 * switches CPU into a guest mode. In fact switching to a guest mode
729 * is very similar to exiting to userspase from rcu point of view. In
730 * addition CPU may stay in a guest mode for quite a long time (up to
731 * one time slice). Lets treat guest mode as quiescent state, just like
732 * we do with user-mode execution.
734 rcu_virt_note_context_switch(smp_processor_id());
737 static inline void kvm_guest_exit(void)
739 account_system_vtime(current);
740 current->flags &= ~PF_VCPU;
744 * search_memslots() and __gfn_to_memslot() are here because they are
745 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
746 * gfn_to_memslot() itself isn't here as an inline because that would
747 * bloat other code too much.
749 static inline struct kvm_memory_slot *
750 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
752 struct kvm_memory_slot *memslot;
754 kvm_for_each_memslot(memslot, slots)
755 if (gfn >= memslot->base_gfn &&
756 gfn < memslot->base_gfn + memslot->npages)
762 static inline struct kvm_memory_slot *
763 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
765 return search_memslots(slots, gfn);
768 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
770 return gfn_to_memslot(kvm, gfn)->id;
773 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
775 /* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
776 return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
777 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
781 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
783 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
785 return slot->base_gfn + gfn_offset;
788 static inline unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot,
791 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
794 static inline gpa_t gfn_to_gpa(gfn_t gfn)
796 return (gpa_t)gfn << PAGE_SHIFT;
799 static inline gfn_t gpa_to_gfn(gpa_t gpa)
801 return (gfn_t)(gpa >> PAGE_SHIFT);
804 static inline hpa_t pfn_to_hpa(pfn_t pfn)
806 return (hpa_t)pfn << PAGE_SHIFT;
809 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
811 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
819 struct kvm_stats_debugfs_item {
822 enum kvm_stat_kind kind;
823 struct dentry *dentry;
825 extern struct kvm_stats_debugfs_item debugfs_entries[];
826 extern struct dentry *kvm_debugfs_dir;
828 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
829 static inline int mmu_notifier_retry(struct kvm_vcpu *vcpu, unsigned long mmu_seq)
831 if (unlikely(vcpu->kvm->mmu_notifier_count))
834 * Ensure the read of mmu_notifier_count happens before the read
835 * of mmu_notifier_seq. This interacts with the smp_wmb() in
836 * mmu_notifier_invalidate_range_end to make sure that the caller
837 * either sees the old (non-zero) value of mmu_notifier_count or
838 * the new (incremented) value of mmu_notifier_seq.
839 * PowerPC Book3s HV KVM calls this under a per-page lock
840 * rather than under kvm->mmu_lock, for scalability, so
841 * can't rely on kvm->mmu_lock to keep things ordered.
844 if (vcpu->kvm->mmu_notifier_seq != mmu_seq)
850 #ifdef KVM_CAP_IRQ_ROUTING
852 #define KVM_MAX_IRQ_ROUTES 1024
854 int kvm_setup_default_irq_routing(struct kvm *kvm);
855 int kvm_set_irq_routing(struct kvm *kvm,
856 const struct kvm_irq_routing_entry *entries,
859 void kvm_free_irq_routing(struct kvm *kvm);
861 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
865 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
869 #ifdef CONFIG_HAVE_KVM_EVENTFD
871 void kvm_eventfd_init(struct kvm *kvm);
872 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
873 void kvm_irqfd_release(struct kvm *kvm);
874 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
875 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
879 static inline void kvm_eventfd_init(struct kvm *kvm) {}
881 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
886 static inline void kvm_irqfd_release(struct kvm *kvm) {}
888 #ifdef CONFIG_HAVE_KVM_IRQCHIP
889 static inline void kvm_irq_routing_update(struct kvm *kvm,
890 struct kvm_irq_routing_table *irq_rt)
892 rcu_assign_pointer(kvm->irq_routing, irq_rt);
896 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
901 #endif /* CONFIG_HAVE_KVM_EVENTFD */
903 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
904 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
906 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
909 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
913 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
917 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
919 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
924 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
932 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
934 set_bit(req, &vcpu->requests);
937 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
939 if (test_bit(req, &vcpu->requests)) {
940 clear_bit(req, &vcpu->requests);
947 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
949 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
951 vcpu->spin_loop.in_spin_loop = val;
953 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
955 vcpu->spin_loop.dy_eligible = val;
958 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
960 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
964 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
968 static inline bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
973 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */