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)
47 /* Two fragments for cross MMIO pages. */
48 #define KVM_MAX_MMIO_FRAGMENTS 2
51 * For the normal pfn, the highest 12 bits should be zero,
52 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
53 * mask bit 63 to indicate the noslot pfn.
55 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
56 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
57 #define KVM_PFN_NOSLOT (0x1ULL << 63)
59 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
60 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
61 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
64 * error pfns indicate that the gfn is in slot but faild to
65 * translate it to pfn on host.
67 static inline bool is_error_pfn(pfn_t pfn)
69 return !!(pfn & KVM_PFN_ERR_MASK);
73 * error_noslot pfns indicate that the gfn can not be
74 * translated to pfn - it is not in slot or failed to
75 * translate it to pfn.
77 static inline bool is_error_noslot_pfn(pfn_t pfn)
79 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
82 /* noslot pfn indicates that the gfn is not in slot. */
83 static inline bool is_noslot_pfn(pfn_t pfn)
85 return pfn == KVM_PFN_NOSLOT;
88 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
89 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
91 static inline bool kvm_is_error_hva(unsigned long addr)
93 return addr >= PAGE_OFFSET;
96 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
98 static inline bool is_error_page(struct page *page)
104 * vcpu->requests bit members
106 #define KVM_REQ_TLB_FLUSH 0
107 #define KVM_REQ_MIGRATE_TIMER 1
108 #define KVM_REQ_REPORT_TPR_ACCESS 2
109 #define KVM_REQ_MMU_RELOAD 3
110 #define KVM_REQ_TRIPLE_FAULT 4
111 #define KVM_REQ_PENDING_TIMER 5
112 #define KVM_REQ_UNHALT 6
113 #define KVM_REQ_MMU_SYNC 7
114 #define KVM_REQ_CLOCK_UPDATE 8
115 #define KVM_REQ_KICK 9
116 #define KVM_REQ_DEACTIVATE_FPU 10
117 #define KVM_REQ_EVENT 11
118 #define KVM_REQ_APF_HALT 12
119 #define KVM_REQ_STEAL_UPDATE 13
120 #define KVM_REQ_NMI 14
121 #define KVM_REQ_PMU 15
122 #define KVM_REQ_PMI 16
123 #define KVM_REQ_WATCHDOG 17
124 #define KVM_REQ_MASTERCLOCK_UPDATE 18
125 #define KVM_REQ_MCLOCK_INPROGRESS 19
126 #define KVM_REQ_EPR_EXIT 20
127 #define KVM_REQ_SCAN_IOAPIC 21
129 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
130 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
132 extern struct kmem_cache *kvm_vcpu_cache;
134 extern spinlock_t kvm_lock;
135 extern struct list_head vm_list;
137 struct kvm_io_range {
140 struct kvm_io_device *dev;
143 #define NR_IOBUS_DEVS 1000
147 struct kvm_io_range range[];
153 KVM_VIRTIO_CCW_NOTIFY_BUS,
157 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
158 int len, const void *val);
159 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
161 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
162 int len, struct kvm_io_device *dev);
163 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
164 struct kvm_io_device *dev);
166 #ifdef CONFIG_KVM_ASYNC_PF
167 struct kvm_async_pf {
168 struct work_struct work;
169 struct list_head link;
170 struct list_head queue;
171 struct kvm_vcpu *vcpu;
172 struct mm_struct *mm;
175 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, unsigned long hva,
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,
299 struct hlist_node link;
302 #ifndef KVM_PRIVATE_MEM_SLOTS
303 #define KVM_PRIVATE_MEM_SLOTS 0
306 #ifndef KVM_MEM_SLOTS_NUM
307 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
312 * memslots are not sorted by id anymore, please use id_to_memslot()
313 * to get the memslot by its id.
315 struct kvm_memslots {
317 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
318 /* The mapping table from slot id to the index in memslots[]. */
319 short id_to_index[KVM_MEM_SLOTS_NUM];
324 struct mutex slots_lock;
325 struct mm_struct *mm; /* userspace tied to this vm */
326 struct kvm_memslots *memslots;
327 struct srcu_struct srcu;
328 struct srcu_struct irq_srcu;
329 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
332 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
333 atomic_t online_vcpus;
334 int last_boosted_vcpu;
335 struct list_head vm_list;
337 struct kvm_io_bus *buses[KVM_NR_BUSES];
338 #ifdef CONFIG_HAVE_KVM_EVENTFD
341 struct list_head items;
342 struct list_head resampler_list;
343 struct mutex resampler_lock;
345 struct list_head ioeventfds;
347 struct kvm_vm_stat stat;
348 struct kvm_arch arch;
349 atomic_t users_count;
350 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
351 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
352 spinlock_t ring_lock;
353 struct list_head coalesced_zones;
356 struct mutex irq_lock;
357 #ifdef CONFIG_HAVE_KVM_IRQCHIP
359 * Update side is protected by irq_lock.
361 struct kvm_irq_routing_table __rcu *irq_routing;
362 struct hlist_head mask_notifier_list;
364 #ifdef CONFIG_HAVE_KVM_IRQFD
365 struct hlist_head irq_ack_notifier_list;
368 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
369 struct mmu_notifier mmu_notifier;
370 unsigned long mmu_notifier_seq;
371 long mmu_notifier_count;
374 struct list_head devices;
377 #define kvm_err(fmt, ...) \
378 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
379 #define kvm_info(fmt, ...) \
380 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
381 #define kvm_debug(fmt, ...) \
382 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
383 #define kvm_pr_unimpl(fmt, ...) \
384 pr_err_ratelimited("kvm [%i]: " fmt, \
385 task_tgid_nr(current), ## __VA_ARGS__)
387 /* The guest did something we don't support. */
388 #define vcpu_unimpl(vcpu, fmt, ...) \
389 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
391 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
394 return kvm->vcpus[i];
397 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
399 idx < atomic_read(&kvm->online_vcpus) && \
400 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
403 #define kvm_for_each_memslot(memslot, slots) \
404 for (memslot = &slots->memslots[0]; \
405 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
408 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
409 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
411 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
412 void vcpu_put(struct kvm_vcpu *vcpu);
414 #ifdef CONFIG_HAVE_KVM_IRQFD
415 int kvm_irqfd_init(void);
416 void kvm_irqfd_exit(void);
418 static inline int kvm_irqfd_init(void)
423 static inline void kvm_irqfd_exit(void)
427 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
428 struct module *module);
431 void kvm_get_kvm(struct kvm *kvm);
432 void kvm_put_kvm(struct kvm *kvm);
434 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
436 return rcu_dereference_check(kvm->memslots,
437 srcu_read_lock_held(&kvm->srcu)
438 || lockdep_is_held(&kvm->slots_lock));
441 static inline struct kvm_memory_slot *
442 id_to_memslot(struct kvm_memslots *slots, int id)
444 int index = slots->id_to_index[id];
445 struct kvm_memory_slot *slot;
447 slot = &slots->memslots[index];
449 WARN_ON(slot->id != id);
454 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
455 * - create a new memory slot
456 * - delete an existing memory slot
457 * - modify an existing memory slot
458 * -- move it in the guest physical memory space
459 * -- just change its flags
461 * Since flags can be changed by some of these operations, the following
462 * differentiation is the best we can do for __kvm_set_memory_region():
471 int kvm_set_memory_region(struct kvm *kvm,
472 struct kvm_userspace_memory_region *mem);
473 int __kvm_set_memory_region(struct kvm *kvm,
474 struct kvm_userspace_memory_region *mem);
475 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
476 struct kvm_memory_slot *dont);
477 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
478 unsigned long npages);
479 void kvm_arch_memslots_updated(struct kvm *kvm);
480 int kvm_arch_prepare_memory_region(struct kvm *kvm,
481 struct kvm_memory_slot *memslot,
482 struct kvm_userspace_memory_region *mem,
483 enum kvm_mr_change change);
484 void kvm_arch_commit_memory_region(struct kvm *kvm,
485 struct kvm_userspace_memory_region *mem,
486 const struct kvm_memory_slot *old,
487 enum kvm_mr_change change);
488 bool kvm_largepages_enabled(void);
489 void kvm_disable_largepages(void);
490 /* flush all memory translations */
491 void kvm_arch_flush_shadow_all(struct kvm *kvm);
492 /* flush memory translations pointing to 'slot' */
493 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
494 struct kvm_memory_slot *slot);
496 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
499 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
500 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
501 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
502 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
503 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
505 void kvm_release_page_clean(struct page *page);
506 void kvm_release_page_dirty(struct page *page);
507 void kvm_set_page_accessed(struct page *page);
509 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
510 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
511 bool write_fault, bool *writable);
512 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
513 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
515 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
516 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
518 void kvm_release_pfn_clean(pfn_t pfn);
519 void kvm_set_pfn_dirty(pfn_t pfn);
520 void kvm_set_pfn_accessed(pfn_t pfn);
521 void kvm_get_pfn(pfn_t pfn);
523 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
525 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
527 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
528 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
529 void *data, unsigned long len);
530 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
531 int offset, int len);
532 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
534 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
535 void *data, unsigned long len);
536 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
537 gpa_t gpa, unsigned long len);
538 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
539 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
540 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
541 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
542 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
543 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
545 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
546 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
547 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
548 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
549 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
550 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
552 void kvm_flush_remote_tlbs(struct kvm *kvm);
553 void kvm_reload_remote_mmus(struct kvm *kvm);
554 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
555 void kvm_make_scan_ioapic_request(struct kvm *kvm);
557 long kvm_arch_dev_ioctl(struct file *filp,
558 unsigned int ioctl, unsigned long arg);
559 long kvm_arch_vcpu_ioctl(struct file *filp,
560 unsigned int ioctl, unsigned long arg);
561 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
563 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
565 int kvm_get_dirty_log(struct kvm *kvm,
566 struct kvm_dirty_log *log, int *is_dirty);
567 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
568 struct kvm_dirty_log *log);
570 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
572 long kvm_arch_vm_ioctl(struct file *filp,
573 unsigned int ioctl, unsigned long arg);
575 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
576 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
578 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
579 struct kvm_translation *tr);
581 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
582 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
583 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
584 struct kvm_sregs *sregs);
585 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
586 struct kvm_sregs *sregs);
587 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
588 struct kvm_mp_state *mp_state);
589 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
590 struct kvm_mp_state *mp_state);
591 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
592 struct kvm_guest_debug *dbg);
593 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
595 int kvm_arch_init(void *opaque);
596 void kvm_arch_exit(void);
598 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
599 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
601 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
603 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
604 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
605 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
606 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
607 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
608 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
609 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
611 int kvm_arch_hardware_enable(void);
612 void kvm_arch_hardware_disable(void);
613 int kvm_arch_hardware_setup(void);
614 void kvm_arch_hardware_unsetup(void);
615 void kvm_arch_check_processor_compat(void *rtn);
616 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
617 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
619 void *kvm_kvzalloc(unsigned long size);
620 void kvm_kvfree(const void *addr);
622 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
623 static inline struct kvm *kvm_arch_alloc_vm(void)
625 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
628 static inline void kvm_arch_free_vm(struct kvm *kvm)
634 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
636 #ifdef __KVM_HAVE_ARCH_WQP
637 return vcpu->arch.wqp;
643 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
644 void kvm_arch_destroy_vm(struct kvm *kvm);
645 void kvm_arch_sync_events(struct kvm *kvm);
647 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
648 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
650 bool kvm_is_mmio_pfn(pfn_t pfn);
652 struct kvm_irq_ack_notifier {
653 struct hlist_node link;
655 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
658 struct kvm_assigned_dev_kernel {
659 struct kvm_irq_ack_notifier ack_notifier;
660 struct list_head list;
665 unsigned int entries_nr;
667 bool host_irq_disabled;
669 struct msix_entry *host_msix_entries;
671 struct msix_entry *guest_msix_entries;
672 unsigned long irq_requested_type;
677 spinlock_t intx_lock;
678 spinlock_t intx_mask_lock;
680 struct pci_saved_state *pci_saved_state;
683 struct kvm_irq_mask_notifier {
684 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
686 struct hlist_node link;
689 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
690 struct kvm_irq_mask_notifier *kimn);
691 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
692 struct kvm_irq_mask_notifier *kimn);
693 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
696 int kvm_irq_map_gsi(struct kvm *kvm,
697 struct kvm_kernel_irq_routing_entry *entries, int gsi);
698 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
700 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
702 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
703 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
704 int irq_source_id, int level, bool line_status);
705 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
706 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
707 void kvm_register_irq_ack_notifier(struct kvm *kvm,
708 struct kvm_irq_ack_notifier *kian);
709 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
710 struct kvm_irq_ack_notifier *kian);
711 int kvm_request_irq_source_id(struct kvm *kvm);
712 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
714 /* For vcpu->arch.iommu_flags */
715 #define KVM_IOMMU_CACHE_COHERENCY 0x1
717 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
718 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
719 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
720 int kvm_iommu_map_guest(struct kvm *kvm);
721 int kvm_iommu_unmap_guest(struct kvm *kvm);
722 int kvm_assign_device(struct kvm *kvm,
723 struct kvm_assigned_dev_kernel *assigned_dev);
724 int kvm_deassign_device(struct kvm *kvm,
725 struct kvm_assigned_dev_kernel *assigned_dev);
727 static inline int kvm_iommu_map_pages(struct kvm *kvm,
728 struct kvm_memory_slot *slot)
733 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
734 struct kvm_memory_slot *slot)
738 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
744 static inline void kvm_guest_enter(void)
748 BUG_ON(preemptible());
750 local_irq_save(flags);
752 local_irq_restore(flags);
754 /* KVM does not hold any references to rcu protected data when it
755 * switches CPU into a guest mode. In fact switching to a guest mode
756 * is very similar to exiting to userspace from rcu point of view. In
757 * addition CPU may stay in a guest mode for quite a long time (up to
758 * one time slice). Lets treat guest mode as quiescent state, just like
759 * we do with user-mode execution.
761 rcu_virt_note_context_switch(smp_processor_id());
764 static inline void kvm_guest_exit(void)
768 local_irq_save(flags);
770 local_irq_restore(flags);
774 * search_memslots() and __gfn_to_memslot() are here because they are
775 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
776 * gfn_to_memslot() itself isn't here as an inline because that would
777 * bloat other code too much.
779 static inline struct kvm_memory_slot *
780 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
782 struct kvm_memory_slot *memslot;
784 kvm_for_each_memslot(memslot, slots)
785 if (gfn >= memslot->base_gfn &&
786 gfn < memslot->base_gfn + memslot->npages)
792 static inline struct kvm_memory_slot *
793 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
795 return search_memslots(slots, gfn);
798 static inline unsigned long
799 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
801 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
804 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
806 return gfn_to_memslot(kvm, gfn)->id;
810 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
812 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
814 return slot->base_gfn + gfn_offset;
817 static inline gpa_t gfn_to_gpa(gfn_t gfn)
819 return (gpa_t)gfn << PAGE_SHIFT;
822 static inline gfn_t gpa_to_gfn(gpa_t gpa)
824 return (gfn_t)(gpa >> PAGE_SHIFT);
827 static inline hpa_t pfn_to_hpa(pfn_t pfn)
829 return (hpa_t)pfn << PAGE_SHIFT;
832 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
834 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
836 return kvm_is_error_hva(hva);
839 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
841 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
849 struct kvm_stats_debugfs_item {
852 enum kvm_stat_kind kind;
853 struct dentry *dentry;
855 extern struct kvm_stats_debugfs_item debugfs_entries[];
856 extern struct dentry *kvm_debugfs_dir;
858 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
859 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
861 if (unlikely(kvm->mmu_notifier_count))
864 * Ensure the read of mmu_notifier_count happens before the read
865 * of mmu_notifier_seq. This interacts with the smp_wmb() in
866 * mmu_notifier_invalidate_range_end to make sure that the caller
867 * either sees the old (non-zero) value of mmu_notifier_count or
868 * the new (incremented) value of mmu_notifier_seq.
869 * PowerPC Book3s HV KVM calls this under a per-page lock
870 * rather than under kvm->mmu_lock, for scalability, so
871 * can't rely on kvm->mmu_lock to keep things ordered.
874 if (kvm->mmu_notifier_seq != mmu_seq)
880 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
882 #define KVM_MAX_IRQ_ROUTES 1024
884 int kvm_setup_default_irq_routing(struct kvm *kvm);
885 int kvm_set_irq_routing(struct kvm *kvm,
886 const struct kvm_irq_routing_entry *entries,
889 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
890 const struct kvm_irq_routing_entry *ue);
891 void kvm_free_irq_routing(struct kvm *kvm);
895 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
899 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
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_IRQFD
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 *);
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)
936 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
941 #endif /* CONFIG_HAVE_KVM_EVENTFD */
943 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
944 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
946 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
949 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
953 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
957 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
959 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
962 void kvm_free_all_assigned_devices(struct kvm *kvm);
966 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
972 static inline void kvm_free_all_assigned_devices(struct kvm *kvm) {}
976 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
978 set_bit(req, &vcpu->requests);
981 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
983 if (test_bit(req, &vcpu->requests)) {
984 clear_bit(req, &vcpu->requests);
991 extern bool kvm_rebooting;
994 struct kvm_device_ops *ops;
997 struct list_head vm_node;
1000 /* create, destroy, and name are mandatory */
1001 struct kvm_device_ops {
1003 int (*create)(struct kvm_device *dev, u32 type);
1006 * Destroy is responsible for freeing dev.
1008 * Destroy may be called before or after destructors are called
1009 * on emulated I/O regions, depending on whether a reference is
1010 * held by a vcpu or other kvm component that gets destroyed
1011 * after the emulated I/O.
1013 void (*destroy)(struct kvm_device *dev);
1015 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1016 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1017 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1018 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1022 void kvm_device_get(struct kvm_device *dev);
1023 void kvm_device_put(struct kvm_device *dev);
1024 struct kvm_device *kvm_device_from_filp(struct file *filp);
1026 extern struct kvm_device_ops kvm_mpic_ops;
1027 extern struct kvm_device_ops kvm_xics_ops;
1028 extern struct kvm_device_ops kvm_vfio_ops;
1029 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1031 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1033 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1035 vcpu->spin_loop.in_spin_loop = val;
1037 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1039 vcpu->spin_loop.dy_eligible = val;
1042 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1044 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1048 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1051 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */