5 * This work is licensed under the terms of the GNU GPL, version 2. See
6 * the COPYING file in the top-level directory.
9 #include <linux/types.h>
10 #include <linux/hardirq.h>
11 #include <linux/list.h>
12 #include <linux/mutex.h>
13 #include <linux/spinlock.h>
14 #include <linux/signal.h>
15 #include <linux/sched.h>
16 #include <linux/bug.h>
18 #include <linux/mmu_notifier.h>
19 #include <linux/preempt.h>
20 #include <linux/msi.h>
21 #include <linux/slab.h>
22 #include <linux/rcupdate.h>
23 #include <linux/ratelimit.h>
24 #include <linux/err.h>
25 #include <linux/irqflags.h>
26 #include <linux/context_tracking.h>
27 #include <asm/signal.h>
29 #include <linux/kvm.h>
30 #include <linux/kvm_para.h>
32 #include <linux/kvm_types.h>
34 #include <asm/kvm_host.h>
37 #define KVM_MMIO_SIZE 8
41 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
42 * in kvm, other bits are visible for userspace which are defined in
43 * include/linux/kvm_h.
45 #define KVM_MEMSLOT_INVALID (1UL << 16)
46 #define KVM_MEMSLOT_INCOHERENT (1UL << 17)
48 /* Two fragments for cross MMIO pages. */
49 #define KVM_MAX_MMIO_FRAGMENTS 2
52 * For the normal pfn, the highest 12 bits should be zero,
53 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
54 * mask bit 63 to indicate the noslot pfn.
56 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
57 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
58 #define KVM_PFN_NOSLOT (0x1ULL << 63)
60 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
61 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
62 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
65 * error pfns indicate that the gfn is in slot but faild to
66 * translate it to pfn on host.
68 static inline bool is_error_pfn(pfn_t pfn)
70 return !!(pfn & KVM_PFN_ERR_MASK);
74 * error_noslot pfns indicate that the gfn can not be
75 * translated to pfn - it is not in slot or failed to
76 * translate it to pfn.
78 static inline bool is_error_noslot_pfn(pfn_t pfn)
80 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
83 /* noslot pfn indicates that the gfn is not in slot. */
84 static inline bool is_noslot_pfn(pfn_t pfn)
86 return pfn == KVM_PFN_NOSLOT;
90 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
91 * provide own defines and kvm_is_error_hva
93 #ifndef KVM_HVA_ERR_BAD
95 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
96 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
98 static inline bool kvm_is_error_hva(unsigned long addr)
100 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_PMU 15
131 #define KVM_REQ_PMI 16
132 #define KVM_REQ_WATCHDOG 17
133 #define KVM_REQ_MASTERCLOCK_UPDATE 18
134 #define KVM_REQ_MCLOCK_INPROGRESS 19
135 #define KVM_REQ_EPR_EXIT 20
136 #define KVM_REQ_SCAN_IOAPIC 21
137 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 22
138 #define KVM_REQ_ENABLE_IBS 23
139 #define KVM_REQ_DISABLE_IBS 24
140 #define KVM_REQ_APIC_PAGE_RELOAD 25
142 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
143 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
145 extern struct kmem_cache *kvm_vcpu_cache;
147 extern spinlock_t kvm_lock;
148 extern struct list_head vm_list;
150 struct kvm_io_range {
153 struct kvm_io_device *dev;
156 #define NR_IOBUS_DEVS 1000
161 struct kvm_io_range range[];
167 KVM_VIRTIO_CCW_NOTIFY_BUS,
172 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
173 int len, const void *val);
174 int kvm_io_bus_write_cookie(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
175 int len, const void *val, long cookie);
176 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
178 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
179 int len, struct kvm_io_device *dev);
180 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
181 struct kvm_io_device *dev);
183 #ifdef CONFIG_KVM_ASYNC_PF
184 struct kvm_async_pf {
185 struct work_struct work;
186 struct list_head link;
187 struct list_head queue;
188 struct kvm_vcpu *vcpu;
189 struct mm_struct *mm;
192 struct kvm_arch_async_pf arch;
196 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
197 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
198 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
199 struct kvm_arch_async_pf *arch);
200 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
207 READING_SHADOW_PAGE_TABLES,
211 * Sometimes a large or cross-page mmio needs to be broken up into separate
212 * exits for userspace servicing.
214 struct kvm_mmio_fragment {
222 #ifdef CONFIG_PREEMPT_NOTIFIERS
223 struct preempt_notifier preempt_notifier;
229 unsigned long requests;
230 unsigned long guest_debug;
236 int guest_fpu_loaded, guest_xcr0_loaded;
237 wait_queue_head_t wq;
241 struct kvm_vcpu_stat stat;
243 #ifdef CONFIG_HAS_IOMEM
245 int mmio_read_completed;
247 int mmio_cur_fragment;
248 int mmio_nr_fragments;
249 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
252 #ifdef CONFIG_KVM_ASYNC_PF
255 struct list_head queue;
256 struct list_head done;
261 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
263 * Cpu relax intercept or pause loop exit optimization
264 * in_spin_loop: set when a vcpu does a pause loop exit
265 * or cpu relax intercepted.
266 * dy_eligible: indicates whether vcpu is eligible for directed yield.
274 struct kvm_vcpu_arch arch;
277 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
279 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
283 * Some of the bitops functions do not support too long bitmaps.
284 * This number must be determined not to exceed such limits.
286 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
288 struct kvm_memory_slot {
290 unsigned long npages;
291 unsigned long *dirty_bitmap;
292 struct kvm_arch_memory_slot arch;
293 unsigned long userspace_addr;
298 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
300 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
303 struct kvm_s390_adapter_int {
311 struct kvm_kernel_irq_routing_entry {
314 int (*set)(struct kvm_kernel_irq_routing_entry *e,
315 struct kvm *kvm, int irq_source_id, int level,
323 struct kvm_s390_adapter_int adapter;
325 struct hlist_node link;
328 #ifndef KVM_PRIVATE_MEM_SLOTS
329 #define KVM_PRIVATE_MEM_SLOTS 0
332 #ifndef KVM_MEM_SLOTS_NUM
333 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
338 * memslots are not sorted by id anymore, please use id_to_memslot()
339 * to get the memslot by its id.
341 struct kvm_memslots {
343 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
344 /* The mapping table from slot id to the index in memslots[]. */
345 short id_to_index[KVM_MEM_SLOTS_NUM];
352 struct mutex slots_lock;
353 struct mm_struct *mm; /* userspace tied to this vm */
354 struct kvm_memslots *memslots;
355 struct srcu_struct srcu;
356 struct srcu_struct irq_srcu;
357 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
360 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
361 atomic_t online_vcpus;
362 int last_boosted_vcpu;
363 struct list_head vm_list;
365 struct kvm_io_bus *buses[KVM_NR_BUSES];
366 #ifdef CONFIG_HAVE_KVM_EVENTFD
369 struct list_head items;
370 struct list_head resampler_list;
371 struct mutex resampler_lock;
373 struct list_head ioeventfds;
375 struct kvm_vm_stat stat;
376 struct kvm_arch arch;
377 atomic_t users_count;
378 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
379 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
380 spinlock_t ring_lock;
381 struct list_head coalesced_zones;
384 struct mutex irq_lock;
385 #ifdef CONFIG_HAVE_KVM_IRQCHIP
387 * Update side is protected by irq_lock.
389 struct kvm_irq_routing_table __rcu *irq_routing;
391 #ifdef CONFIG_HAVE_KVM_IRQFD
392 struct hlist_head irq_ack_notifier_list;
395 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
396 struct mmu_notifier mmu_notifier;
397 unsigned long mmu_notifier_seq;
398 long mmu_notifier_count;
401 struct list_head devices;
404 #define kvm_err(fmt, ...) \
405 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
406 #define kvm_info(fmt, ...) \
407 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
408 #define kvm_debug(fmt, ...) \
409 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
410 #define kvm_pr_unimpl(fmt, ...) \
411 pr_err_ratelimited("kvm [%i]: " fmt, \
412 task_tgid_nr(current), ## __VA_ARGS__)
414 /* The guest did something we don't support. */
415 #define vcpu_unimpl(vcpu, fmt, ...) \
416 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
418 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
421 return kvm->vcpus[i];
424 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
426 idx < atomic_read(&kvm->online_vcpus) && \
427 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
430 #define kvm_for_each_memslot(memslot, slots) \
431 for (memslot = &slots->memslots[0]; \
432 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
435 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
436 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
438 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
439 void vcpu_put(struct kvm_vcpu *vcpu);
441 #ifdef __KVM_HAVE_IOAPIC
442 void kvm_vcpu_request_scan_ioapic(struct kvm *kvm);
444 static inline void kvm_vcpu_request_scan_ioapic(struct kvm *kvm)
449 #ifdef CONFIG_HAVE_KVM_IRQFD
450 int kvm_irqfd_init(void);
451 void kvm_irqfd_exit(void);
453 static inline int kvm_irqfd_init(void)
458 static inline void kvm_irqfd_exit(void)
462 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
463 struct module *module);
466 void kvm_get_kvm(struct kvm *kvm);
467 void kvm_put_kvm(struct kvm *kvm);
469 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
471 return rcu_dereference_check(kvm->memslots,
472 srcu_read_lock_held(&kvm->srcu)
473 || lockdep_is_held(&kvm->slots_lock));
476 static inline struct kvm_memory_slot *
477 id_to_memslot(struct kvm_memslots *slots, int id)
479 int index = slots->id_to_index[id];
480 struct kvm_memory_slot *slot;
482 slot = &slots->memslots[index];
484 WARN_ON(slot->id != id);
489 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
490 * - create a new memory slot
491 * - delete an existing memory slot
492 * - modify an existing memory slot
493 * -- move it in the guest physical memory space
494 * -- just change its flags
496 * Since flags can be changed by some of these operations, the following
497 * differentiation is the best we can do for __kvm_set_memory_region():
506 int kvm_set_memory_region(struct kvm *kvm,
507 struct kvm_userspace_memory_region *mem);
508 int __kvm_set_memory_region(struct kvm *kvm,
509 struct kvm_userspace_memory_region *mem);
510 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
511 struct kvm_memory_slot *dont);
512 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
513 unsigned long npages);
514 void kvm_arch_memslots_updated(struct kvm *kvm);
515 int kvm_arch_prepare_memory_region(struct kvm *kvm,
516 struct kvm_memory_slot *memslot,
517 struct kvm_userspace_memory_region *mem,
518 enum kvm_mr_change change);
519 void kvm_arch_commit_memory_region(struct kvm *kvm,
520 struct kvm_userspace_memory_region *mem,
521 const struct kvm_memory_slot *old,
522 enum kvm_mr_change change);
523 bool kvm_largepages_enabled(void);
524 void kvm_disable_largepages(void);
525 /* flush all memory translations */
526 void kvm_arch_flush_shadow_all(struct kvm *kvm);
527 /* flush memory translations pointing to 'slot' */
528 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
529 struct kvm_memory_slot *slot);
531 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
534 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
535 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
536 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
537 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
538 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
540 void kvm_release_page_clean(struct page *page);
541 void kvm_release_page_dirty(struct page *page);
542 void kvm_set_page_accessed(struct page *page);
544 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
545 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
546 bool write_fault, bool *writable);
547 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
548 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
550 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
551 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
553 void kvm_release_pfn_clean(pfn_t pfn);
554 void kvm_set_pfn_dirty(pfn_t pfn);
555 void kvm_set_pfn_accessed(pfn_t pfn);
556 void kvm_get_pfn(pfn_t pfn);
558 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
560 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
562 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
563 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
564 void *data, unsigned long len);
565 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
566 int offset, int len);
567 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
569 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
570 void *data, unsigned long len);
571 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
572 gpa_t gpa, unsigned long len);
573 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
574 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
575 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
576 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
577 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
578 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
580 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
581 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
582 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
583 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
584 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
585 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
587 void kvm_flush_remote_tlbs(struct kvm *kvm);
588 void kvm_reload_remote_mmus(struct kvm *kvm);
589 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
590 void kvm_make_scan_ioapic_request(struct kvm *kvm);
591 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
593 long kvm_arch_dev_ioctl(struct file *filp,
594 unsigned int ioctl, unsigned long arg);
595 long kvm_arch_vcpu_ioctl(struct file *filp,
596 unsigned int ioctl, unsigned long arg);
597 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
599 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
601 int kvm_get_dirty_log(struct kvm *kvm,
602 struct kvm_dirty_log *log, int *is_dirty);
603 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
604 struct kvm_dirty_log *log);
606 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
608 long kvm_arch_vm_ioctl(struct file *filp,
609 unsigned int ioctl, unsigned long arg);
611 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
612 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
614 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
615 struct kvm_translation *tr);
617 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
618 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
619 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
620 struct kvm_sregs *sregs);
621 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
622 struct kvm_sregs *sregs);
623 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
624 struct kvm_mp_state *mp_state);
625 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
626 struct kvm_mp_state *mp_state);
627 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
628 struct kvm_guest_debug *dbg);
629 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
631 int kvm_arch_init(void *opaque);
632 void kvm_arch_exit(void);
634 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
635 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
637 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
639 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
640 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
641 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
642 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
643 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
644 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
645 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
647 int kvm_arch_hardware_enable(void);
648 void kvm_arch_hardware_disable(void);
649 int kvm_arch_hardware_setup(void);
650 void kvm_arch_hardware_unsetup(void);
651 void kvm_arch_check_processor_compat(void *rtn);
652 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
653 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
655 void *kvm_kvzalloc(unsigned long size);
656 void kvm_kvfree(const void *addr);
658 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
659 static inline struct kvm *kvm_arch_alloc_vm(void)
661 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
664 static inline void kvm_arch_free_vm(struct kvm *kvm)
670 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
671 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
672 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
673 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
675 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
679 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
683 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
689 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
691 #ifdef __KVM_HAVE_ARCH_WQP
692 return vcpu->arch.wqp;
698 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
699 void kvm_arch_destroy_vm(struct kvm *kvm);
700 void kvm_arch_sync_events(struct kvm *kvm);
702 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
703 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
705 bool kvm_is_reserved_pfn(pfn_t pfn);
707 struct kvm_irq_ack_notifier {
708 struct hlist_node link;
710 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
713 int kvm_irq_map_gsi(struct kvm *kvm,
714 struct kvm_kernel_irq_routing_entry *entries, int gsi);
715 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
717 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
719 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
720 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
721 int irq_source_id, int level, bool line_status);
722 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
723 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
724 void kvm_register_irq_ack_notifier(struct kvm *kvm,
725 struct kvm_irq_ack_notifier *kian);
726 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
727 struct kvm_irq_ack_notifier *kian);
728 int kvm_request_irq_source_id(struct kvm *kvm);
729 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
731 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
732 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
733 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
735 static inline int kvm_iommu_map_pages(struct kvm *kvm,
736 struct kvm_memory_slot *slot)
741 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
742 struct kvm_memory_slot *slot)
747 static inline void kvm_guest_enter(void)
751 BUG_ON(preemptible());
753 local_irq_save(flags);
755 local_irq_restore(flags);
757 /* KVM does not hold any references to rcu protected data when it
758 * switches CPU into a guest mode. In fact switching to a guest mode
759 * is very similar to exiting to userspace from rcu point of view. In
760 * addition CPU may stay in a guest mode for quite a long time (up to
761 * one time slice). Lets treat guest mode as quiescent state, just like
762 * we do with user-mode execution.
764 rcu_virt_note_context_switch(smp_processor_id());
767 static inline void kvm_guest_exit(void)
771 local_irq_save(flags);
773 local_irq_restore(flags);
777 * search_memslots() and __gfn_to_memslot() are here because they are
778 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
779 * gfn_to_memslot() itself isn't here as an inline because that would
780 * bloat other code too much.
782 static inline struct kvm_memory_slot *
783 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
785 int start = 0, end = slots->used_slots;
786 int slot = atomic_read(&slots->lru_slot);
787 struct kvm_memory_slot *memslots = slots->memslots;
789 if (gfn >= memslots[slot].base_gfn &&
790 gfn < memslots[slot].base_gfn + memslots[slot].npages)
791 return &memslots[slot];
793 while (start < end) {
794 slot = start + (end - start) / 2;
796 if (gfn >= memslots[slot].base_gfn)
802 if (gfn >= memslots[start].base_gfn &&
803 gfn < memslots[start].base_gfn + memslots[start].npages) {
804 atomic_set(&slots->lru_slot, start);
805 return &memslots[start];
811 static inline struct kvm_memory_slot *
812 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
814 return search_memslots(slots, gfn);
817 static inline unsigned long
818 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
820 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
823 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
825 return gfn_to_memslot(kvm, gfn)->id;
829 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
831 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
833 return slot->base_gfn + gfn_offset;
836 static inline gpa_t gfn_to_gpa(gfn_t gfn)
838 return (gpa_t)gfn << PAGE_SHIFT;
841 static inline gfn_t gpa_to_gfn(gpa_t gpa)
843 return (gfn_t)(gpa >> PAGE_SHIFT);
846 static inline hpa_t pfn_to_hpa(pfn_t pfn)
848 return (hpa_t)pfn << PAGE_SHIFT;
851 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
853 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
855 return kvm_is_error_hva(hva);
858 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
860 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
868 struct kvm_stats_debugfs_item {
871 enum kvm_stat_kind kind;
872 struct dentry *dentry;
874 extern struct kvm_stats_debugfs_item debugfs_entries[];
875 extern struct dentry *kvm_debugfs_dir;
877 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
878 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
880 if (unlikely(kvm->mmu_notifier_count))
883 * Ensure the read of mmu_notifier_count happens before the read
884 * of mmu_notifier_seq. This interacts with the smp_wmb() in
885 * mmu_notifier_invalidate_range_end to make sure that the caller
886 * either sees the old (non-zero) value of mmu_notifier_count or
887 * the new (incremented) value of mmu_notifier_seq.
888 * PowerPC Book3s HV KVM calls this under a per-page lock
889 * rather than under kvm->mmu_lock, for scalability, so
890 * can't rely on kvm->mmu_lock to keep things ordered.
893 if (kvm->mmu_notifier_seq != mmu_seq)
899 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
902 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
904 #define KVM_MAX_IRQ_ROUTES 1024
907 int kvm_setup_default_irq_routing(struct kvm *kvm);
908 int kvm_set_irq_routing(struct kvm *kvm,
909 const struct kvm_irq_routing_entry *entries,
912 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
913 const struct kvm_irq_routing_entry *ue);
914 void kvm_free_irq_routing(struct kvm *kvm);
918 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
922 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
924 #ifdef CONFIG_HAVE_KVM_EVENTFD
926 void kvm_eventfd_init(struct kvm *kvm);
927 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
929 #ifdef CONFIG_HAVE_KVM_IRQFD
930 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
931 void kvm_irqfd_release(struct kvm *kvm);
932 void kvm_irq_routing_update(struct kvm *);
934 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
939 static inline void kvm_irqfd_release(struct kvm *kvm) {}
944 static inline void kvm_eventfd_init(struct kvm *kvm) {}
946 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
951 static inline void kvm_irqfd_release(struct kvm *kvm) {}
953 #ifdef CONFIG_HAVE_KVM_IRQCHIP
954 static inline void kvm_irq_routing_update(struct kvm *kvm)
959 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
964 #endif /* CONFIG_HAVE_KVM_EVENTFD */
966 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
967 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
969 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
972 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
976 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
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 extern bool kvm_rebooting;
998 struct kvm_device_ops *ops;
1001 struct list_head vm_node;
1004 /* create, destroy, and name are mandatory */
1005 struct kvm_device_ops {
1007 int (*create)(struct kvm_device *dev, u32 type);
1010 * Destroy is responsible for freeing dev.
1012 * Destroy may be called before or after destructors are called
1013 * on emulated I/O regions, depending on whether a reference is
1014 * held by a vcpu or other kvm component that gets destroyed
1015 * after the emulated I/O.
1017 void (*destroy)(struct kvm_device *dev);
1019 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1020 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1021 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1022 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1026 void kvm_device_get(struct kvm_device *dev);
1027 void kvm_device_put(struct kvm_device *dev);
1028 struct kvm_device *kvm_device_from_filp(struct file *filp);
1029 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1030 void kvm_unregister_device_ops(u32 type);
1032 extern struct kvm_device_ops kvm_mpic_ops;
1033 extern struct kvm_device_ops kvm_xics_ops;
1035 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1037 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1039 vcpu->spin_loop.in_spin_loop = val;
1041 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1043 vcpu->spin_loop.dy_eligible = val;
1046 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1048 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1052 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1055 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */