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
134 extern struct kmem_cache *kvm_vcpu_cache;
136 extern spinlock_t kvm_lock;
137 extern struct list_head vm_list;
139 struct kvm_io_range {
142 struct kvm_io_device *dev;
145 #define NR_IOBUS_DEVS 1000
149 struct kvm_io_range range[];
155 KVM_VIRTIO_CCW_NOTIFY_BUS,
159 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
160 int len, const void *val);
161 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
163 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
164 int len, struct kvm_io_device *dev);
165 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
166 struct kvm_io_device *dev);
168 #ifdef CONFIG_KVM_ASYNC_PF
169 struct kvm_async_pf {
170 struct work_struct work;
171 struct list_head link;
172 struct list_head queue;
173 struct kvm_vcpu *vcpu;
174 struct mm_struct *mm;
177 struct kvm_arch_async_pf arch;
181 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
182 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
183 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
184 struct kvm_arch_async_pf *arch);
185 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
192 READING_SHADOW_PAGE_TABLES,
196 * Sometimes a large or cross-page mmio needs to be broken up into separate
197 * exits for userspace servicing.
199 struct kvm_mmio_fragment {
207 #ifdef CONFIG_PREEMPT_NOTIFIERS
208 struct preempt_notifier preempt_notifier;
214 unsigned long requests;
215 unsigned long guest_debug;
221 int guest_fpu_loaded, guest_xcr0_loaded;
222 wait_queue_head_t wq;
226 struct kvm_vcpu_stat stat;
228 #ifdef CONFIG_HAS_IOMEM
230 int mmio_read_completed;
232 int mmio_cur_fragment;
233 int mmio_nr_fragments;
234 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
237 #ifdef CONFIG_KVM_ASYNC_PF
240 struct list_head queue;
241 struct list_head done;
246 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
248 * Cpu relax intercept or pause loop exit optimization
249 * in_spin_loop: set when a vcpu does a pause loop exit
250 * or cpu relax intercepted.
251 * dy_eligible: indicates whether vcpu is eligible for directed yield.
259 struct kvm_vcpu_arch arch;
262 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
264 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
268 * Some of the bitops functions do not support too long bitmaps.
269 * This number must be determined not to exceed such limits.
271 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
273 struct kvm_memory_slot {
275 unsigned long npages;
276 unsigned long *dirty_bitmap;
277 struct kvm_arch_memory_slot arch;
278 unsigned long userspace_addr;
283 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
285 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
288 struct kvm_kernel_irq_routing_entry {
291 int (*set)(struct kvm_kernel_irq_routing_entry *e,
292 struct kvm *kvm, int irq_source_id, int level,
301 struct hlist_node link;
304 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
306 struct kvm_irq_routing_table {
307 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
308 struct kvm_kernel_irq_routing_entry *rt_entries;
311 * Array indexed by gsi. Each entry contains list of irq chips
312 * the gsi is connected to.
314 struct hlist_head map[0];
319 struct kvm_irq_routing_table {};
323 #ifndef KVM_PRIVATE_MEM_SLOTS
324 #define KVM_PRIVATE_MEM_SLOTS 0
327 #ifndef KVM_MEM_SLOTS_NUM
328 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
333 * memslots are not sorted by id anymore, please use id_to_memslot()
334 * to get the memslot by its id.
336 struct kvm_memslots {
338 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
339 /* The mapping table from slot id to the index in memslots[]. */
340 short id_to_index[KVM_MEM_SLOTS_NUM];
345 struct mutex slots_lock;
346 struct mm_struct *mm; /* userspace tied to this vm */
347 struct kvm_memslots *memslots;
348 struct srcu_struct srcu;
349 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
352 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
353 atomic_t online_vcpus;
354 int last_boosted_vcpu;
355 struct list_head vm_list;
357 struct kvm_io_bus *buses[KVM_NR_BUSES];
358 #ifdef CONFIG_HAVE_KVM_EVENTFD
361 struct list_head items;
362 struct list_head resampler_list;
363 struct mutex resampler_lock;
365 struct list_head ioeventfds;
367 struct kvm_vm_stat stat;
368 struct kvm_arch arch;
369 atomic_t users_count;
370 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
371 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
372 spinlock_t ring_lock;
373 struct list_head coalesced_zones;
376 struct mutex irq_lock;
377 #ifdef CONFIG_HAVE_KVM_IRQCHIP
379 * Update side is protected by irq_lock and,
380 * if configured, irqfds.lock.
382 struct kvm_irq_routing_table __rcu *irq_routing;
383 struct hlist_head mask_notifier_list;
384 struct hlist_head irq_ack_notifier_list;
387 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
388 struct mmu_notifier mmu_notifier;
389 unsigned long mmu_notifier_seq;
390 long mmu_notifier_count;
393 struct list_head devices;
396 #define kvm_err(fmt, ...) \
397 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
398 #define kvm_info(fmt, ...) \
399 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
400 #define kvm_debug(fmt, ...) \
401 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
402 #define kvm_pr_unimpl(fmt, ...) \
403 pr_err_ratelimited("kvm [%i]: " fmt, \
404 task_tgid_nr(current), ## __VA_ARGS__)
406 /* The guest did something we don't support. */
407 #define vcpu_unimpl(vcpu, fmt, ...) \
408 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
410 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
413 return kvm->vcpus[i];
416 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
418 idx < atomic_read(&kvm->online_vcpus) && \
419 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
422 #define kvm_for_each_memslot(memslot, slots) \
423 for (memslot = &slots->memslots[0]; \
424 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
427 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
428 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
430 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
431 void vcpu_put(struct kvm_vcpu *vcpu);
433 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
434 int kvm_irqfd_init(void);
435 void kvm_irqfd_exit(void);
437 static inline int kvm_irqfd_init(void)
442 static inline void kvm_irqfd_exit(void)
446 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
447 struct module *module);
450 void kvm_get_kvm(struct kvm *kvm);
451 void kvm_put_kvm(struct kvm *kvm);
453 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
455 return rcu_dereference_check(kvm->memslots,
456 srcu_read_lock_held(&kvm->srcu)
457 || lockdep_is_held(&kvm->slots_lock));
460 static inline struct kvm_memory_slot *
461 id_to_memslot(struct kvm_memslots *slots, int id)
463 int index = slots->id_to_index[id];
464 struct kvm_memory_slot *slot;
466 slot = &slots->memslots[index];
468 WARN_ON(slot->id != id);
473 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
474 * - create a new memory slot
475 * - delete an existing memory slot
476 * - modify an existing memory slot
477 * -- move it in the guest physical memory space
478 * -- just change its flags
480 * Since flags can be changed by some of these operations, the following
481 * differentiation is the best we can do for __kvm_set_memory_region():
490 int kvm_set_memory_region(struct kvm *kvm,
491 struct kvm_userspace_memory_region *mem);
492 int __kvm_set_memory_region(struct kvm *kvm,
493 struct kvm_userspace_memory_region *mem);
494 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
495 struct kvm_memory_slot *dont);
496 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
497 unsigned long npages);
498 void kvm_arch_memslots_updated(struct kvm *kvm);
499 int kvm_arch_prepare_memory_region(struct kvm *kvm,
500 struct kvm_memory_slot *memslot,
501 struct kvm_userspace_memory_region *mem,
502 enum kvm_mr_change change);
503 void kvm_arch_commit_memory_region(struct kvm *kvm,
504 struct kvm_userspace_memory_region *mem,
505 const struct kvm_memory_slot *old,
506 enum kvm_mr_change change);
507 bool kvm_largepages_enabled(void);
508 void kvm_disable_largepages(void);
509 /* flush all memory translations */
510 void kvm_arch_flush_shadow_all(struct kvm *kvm);
511 /* flush memory translations pointing to 'slot' */
512 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
513 struct kvm_memory_slot *slot);
515 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
518 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
519 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
520 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
521 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
522 void kvm_release_page_clean(struct page *page);
523 void kvm_release_page_dirty(struct page *page);
524 void kvm_set_page_accessed(struct page *page);
526 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
527 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
528 bool write_fault, bool *writable);
529 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
530 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
532 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
533 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
535 void kvm_release_pfn_clean(pfn_t pfn);
536 void kvm_set_pfn_dirty(pfn_t pfn);
537 void kvm_set_pfn_accessed(pfn_t pfn);
538 void kvm_get_pfn(pfn_t pfn);
540 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
542 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
544 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
545 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
546 void *data, unsigned long len);
547 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
548 int offset, int len);
549 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
551 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
552 void *data, unsigned long len);
553 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
554 gpa_t gpa, unsigned long len);
555 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
556 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
557 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
558 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
559 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
560 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
562 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
563 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
564 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
565 void kvm_vcpu_on_spin(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_scan_ioapic_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_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
589 long kvm_arch_vm_ioctl(struct file *filp,
590 unsigned int ioctl, unsigned long arg);
592 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
593 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
595 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
596 struct kvm_translation *tr);
598 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
599 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
600 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
601 struct kvm_sregs *sregs);
602 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
603 struct kvm_sregs *sregs);
604 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
605 struct kvm_mp_state *mp_state);
606 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
607 struct kvm_mp_state *mp_state);
608 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
609 struct kvm_guest_debug *dbg);
610 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
612 int kvm_arch_init(void *opaque);
613 void kvm_arch_exit(void);
615 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
616 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
618 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
619 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
620 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
621 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
622 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
623 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
624 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
626 int kvm_arch_hardware_enable(void *garbage);
627 void kvm_arch_hardware_disable(void *garbage);
628 int kvm_arch_hardware_setup(void);
629 void kvm_arch_hardware_unsetup(void);
630 void kvm_arch_check_processor_compat(void *rtn);
631 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
632 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
634 void *kvm_kvzalloc(unsigned long size);
635 void kvm_kvfree(const void *addr);
637 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
638 static inline struct kvm *kvm_arch_alloc_vm(void)
640 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
643 static inline void kvm_arch_free_vm(struct kvm *kvm)
649 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
651 #ifdef __KVM_HAVE_ARCH_WQP
652 return vcpu->arch.wqp;
658 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
659 void kvm_arch_destroy_vm(struct kvm *kvm);
660 void kvm_arch_sync_events(struct kvm *kvm);
662 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
663 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
665 bool kvm_is_mmio_pfn(pfn_t pfn);
667 struct kvm_irq_ack_notifier {
668 struct hlist_node link;
670 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
673 struct kvm_assigned_dev_kernel {
674 struct kvm_irq_ack_notifier ack_notifier;
675 struct list_head list;
680 unsigned int entries_nr;
682 bool host_irq_disabled;
684 struct msix_entry *host_msix_entries;
686 struct msix_entry *guest_msix_entries;
687 unsigned long irq_requested_type;
692 spinlock_t intx_lock;
693 spinlock_t intx_mask_lock;
695 struct pci_saved_state *pci_saved_state;
698 struct kvm_irq_mask_notifier {
699 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
701 struct hlist_node link;
704 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
705 struct kvm_irq_mask_notifier *kimn);
706 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
707 struct kvm_irq_mask_notifier *kimn);
708 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
711 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
713 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
714 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
715 int irq_source_id, int level, bool line_status);
716 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
717 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
718 void kvm_register_irq_ack_notifier(struct kvm *kvm,
719 struct kvm_irq_ack_notifier *kian);
720 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
721 struct kvm_irq_ack_notifier *kian);
722 int kvm_request_irq_source_id(struct kvm *kvm);
723 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
725 /* For vcpu->arch.iommu_flags */
726 #define KVM_IOMMU_CACHE_COHERENCY 0x1
728 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
729 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
730 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
731 int kvm_iommu_map_guest(struct kvm *kvm);
732 int kvm_iommu_unmap_guest(struct kvm *kvm);
733 int kvm_assign_device(struct kvm *kvm,
734 struct kvm_assigned_dev_kernel *assigned_dev);
735 int kvm_deassign_device(struct kvm *kvm,
736 struct kvm_assigned_dev_kernel *assigned_dev);
738 static inline int kvm_iommu_map_pages(struct kvm *kvm,
739 struct kvm_memory_slot *slot)
744 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
745 struct kvm_memory_slot *slot)
749 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
755 static inline void kvm_guest_enter(void)
759 BUG_ON(preemptible());
761 local_irq_save(flags);
763 local_irq_restore(flags);
765 /* KVM does not hold any references to rcu protected data when it
766 * switches CPU into a guest mode. In fact switching to a guest mode
767 * is very similar to exiting to userspace from rcu point of view. In
768 * addition CPU may stay in a guest mode for quite a long time (up to
769 * one time slice). Lets treat guest mode as quiescent state, just like
770 * we do with user-mode execution.
772 rcu_virt_note_context_switch(smp_processor_id());
775 static inline void kvm_guest_exit(void)
779 local_irq_save(flags);
781 local_irq_restore(flags);
785 * search_memslots() and __gfn_to_memslot() are here because they are
786 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
787 * gfn_to_memslot() itself isn't here as an inline because that would
788 * bloat other code too much.
790 static inline struct kvm_memory_slot *
791 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
793 struct kvm_memory_slot *memslot;
795 kvm_for_each_memslot(memslot, slots)
796 if (gfn >= memslot->base_gfn &&
797 gfn < memslot->base_gfn + memslot->npages)
803 static inline struct kvm_memory_slot *
804 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
806 return search_memslots(slots, gfn);
809 static inline unsigned long
810 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
812 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
815 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
817 return gfn_to_memslot(kvm, gfn)->id;
821 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
823 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
825 return slot->base_gfn + gfn_offset;
828 static inline gpa_t gfn_to_gpa(gfn_t gfn)
830 return (gpa_t)gfn << PAGE_SHIFT;
833 static inline gfn_t gpa_to_gfn(gpa_t gpa)
835 return (gfn_t)(gpa >> PAGE_SHIFT);
838 static inline hpa_t pfn_to_hpa(pfn_t pfn)
840 return (hpa_t)pfn << PAGE_SHIFT;
843 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
845 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
847 return kvm_is_error_hva(hva);
850 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
852 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
860 struct kvm_stats_debugfs_item {
863 enum kvm_stat_kind kind;
864 struct dentry *dentry;
866 extern struct kvm_stats_debugfs_item debugfs_entries[];
867 extern struct dentry *kvm_debugfs_dir;
869 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
870 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
872 if (unlikely(kvm->mmu_notifier_count))
875 * Ensure the read of mmu_notifier_count happens before the read
876 * of mmu_notifier_seq. This interacts with the smp_wmb() in
877 * mmu_notifier_invalidate_range_end to make sure that the caller
878 * either sees the old (non-zero) value of mmu_notifier_count or
879 * the new (incremented) value of mmu_notifier_seq.
880 * PowerPC Book3s HV KVM calls this under a per-page lock
881 * rather than under kvm->mmu_lock, for scalability, so
882 * can't rely on kvm->mmu_lock to keep things ordered.
885 if (kvm->mmu_notifier_seq != mmu_seq)
891 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
893 #define KVM_MAX_IRQ_ROUTES 1024
895 int kvm_setup_default_irq_routing(struct kvm *kvm);
896 int kvm_set_irq_routing(struct kvm *kvm,
897 const struct kvm_irq_routing_entry *entries,
900 int kvm_set_routing_entry(struct kvm_irq_routing_table *rt,
901 struct kvm_kernel_irq_routing_entry *e,
902 const struct kvm_irq_routing_entry *ue);
903 void kvm_free_irq_routing(struct kvm *kvm);
905 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
909 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
913 #ifdef CONFIG_HAVE_KVM_EVENTFD
915 void kvm_eventfd_init(struct kvm *kvm);
916 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
918 #ifdef CONFIG_HAVE_KVM_IRQCHIP
919 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
920 void kvm_irqfd_release(struct kvm *kvm);
921 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
923 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
928 static inline void kvm_irqfd_release(struct kvm *kvm) {}
933 static inline void kvm_eventfd_init(struct kvm *kvm) {}
935 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
940 static inline void kvm_irqfd_release(struct kvm *kvm) {}
942 #ifdef CONFIG_HAVE_KVM_IRQCHIP
943 static inline void kvm_irq_routing_update(struct kvm *kvm,
944 struct kvm_irq_routing_table *irq_rt)
946 rcu_assign_pointer(kvm->irq_routing, irq_rt);
950 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
955 #endif /* CONFIG_HAVE_KVM_EVENTFD */
957 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
958 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
960 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
963 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
967 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
971 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
973 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
976 void kvm_free_all_assigned_devices(struct kvm *kvm);
980 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
986 static inline void kvm_free_all_assigned_devices(struct kvm *kvm) {}
990 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
992 set_bit(req, &vcpu->requests);
995 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
997 if (test_bit(req, &vcpu->requests)) {
998 clear_bit(req, &vcpu->requests);
1005 extern bool kvm_rebooting;
1007 struct kvm_device_ops;
1010 struct kvm_device_ops *ops;
1013 struct list_head vm_node;
1016 /* create, destroy, and name are mandatory */
1017 struct kvm_device_ops {
1019 int (*create)(struct kvm_device *dev, u32 type);
1022 * Destroy is responsible for freeing dev.
1024 * Destroy may be called before or after destructors are called
1025 * on emulated I/O regions, depending on whether a reference is
1026 * held by a vcpu or other kvm component that gets destroyed
1027 * after the emulated I/O.
1029 void (*destroy)(struct kvm_device *dev);
1031 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1032 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1033 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1034 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1038 void kvm_device_get(struct kvm_device *dev);
1039 void kvm_device_put(struct kvm_device *dev);
1040 struct kvm_device *kvm_device_from_filp(struct file *filp);
1042 extern struct kvm_device_ops kvm_mpic_ops;
1043 extern struct kvm_device_ops kvm_xics_ops;
1044 extern struct kvm_device_ops kvm_vfio_ops;
1045 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1047 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1049 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1051 vcpu->spin_loop.in_spin_loop = val;
1053 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1055 vcpu->spin_loop.dy_eligible = val;
1058 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1060 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1064 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1067 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */