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 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
38 * in kvm, other bits are visible for userspace which are defined in
39 * include/linux/kvm_h.
41 #define KVM_MEMSLOT_INVALID (1UL << 16)
42 #define KVM_MEMSLOT_INCOHERENT (1UL << 17)
44 /* Two fragments for cross MMIO pages. */
45 #define KVM_MAX_MMIO_FRAGMENTS 2
47 #ifndef KVM_ADDRESS_SPACE_NUM
48 #define KVM_ADDRESS_SPACE_NUM 1
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
141 #define KVM_REQ_SMI 26
142 #define KVM_REQ_HV_CRASH 27
143 #define KVM_REQ_IOAPIC_EOI_EXIT 28
145 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
146 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
148 extern struct kmem_cache *kvm_vcpu_cache;
150 extern spinlock_t kvm_lock;
151 extern struct list_head vm_list;
153 struct kvm_io_range {
156 struct kvm_io_device *dev;
159 #define NR_IOBUS_DEVS 1000
164 struct kvm_io_range range[];
170 KVM_VIRTIO_CCW_NOTIFY_BUS,
175 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
176 int len, const void *val);
177 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
178 gpa_t addr, int len, const void *val, long cookie);
179 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
181 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
182 int len, struct kvm_io_device *dev);
183 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
184 struct kvm_io_device *dev);
186 #ifdef CONFIG_KVM_ASYNC_PF
187 struct kvm_async_pf {
188 struct work_struct work;
189 struct list_head link;
190 struct list_head queue;
191 struct kvm_vcpu *vcpu;
192 struct mm_struct *mm;
195 struct kvm_arch_async_pf arch;
199 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
200 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
201 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
202 struct kvm_arch_async_pf *arch);
203 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
210 READING_SHADOW_PAGE_TABLES,
214 * Sometimes a large or cross-page mmio needs to be broken up into separate
215 * exits for userspace servicing.
217 struct kvm_mmio_fragment {
225 #ifdef CONFIG_PREEMPT_NOTIFIERS
226 struct preempt_notifier preempt_notifier;
232 unsigned long requests;
233 unsigned long guest_debug;
239 int guest_fpu_loaded, guest_xcr0_loaded;
240 unsigned char fpu_counter;
241 wait_queue_head_t wq;
245 struct kvm_vcpu_stat stat;
246 unsigned int halt_poll_ns;
248 #ifdef CONFIG_HAS_IOMEM
250 int mmio_read_completed;
252 int mmio_cur_fragment;
253 int mmio_nr_fragments;
254 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
257 #ifdef CONFIG_KVM_ASYNC_PF
260 struct list_head queue;
261 struct list_head done;
266 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
268 * Cpu relax intercept or pause loop exit optimization
269 * in_spin_loop: set when a vcpu does a pause loop exit
270 * or cpu relax intercepted.
271 * dy_eligible: indicates whether vcpu is eligible for directed yield.
279 struct kvm_vcpu_arch arch;
282 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
284 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
288 * Some of the bitops functions do not support too long bitmaps.
289 * This number must be determined not to exceed such limits.
291 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
293 struct kvm_memory_slot {
295 unsigned long npages;
296 unsigned long *dirty_bitmap;
297 struct kvm_arch_memory_slot arch;
298 unsigned long userspace_addr;
303 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
305 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
308 struct kvm_s390_adapter_int {
316 struct kvm_kernel_irq_routing_entry {
319 int (*set)(struct kvm_kernel_irq_routing_entry *e,
320 struct kvm *kvm, int irq_source_id, int level,
328 struct kvm_s390_adapter_int adapter;
330 struct hlist_node link;
333 #ifndef KVM_PRIVATE_MEM_SLOTS
334 #define KVM_PRIVATE_MEM_SLOTS 0
337 #ifndef KVM_MEM_SLOTS_NUM
338 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
341 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
342 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
350 * memslots are not sorted by id anymore, please use id_to_memslot()
351 * to get the memslot by its id.
353 struct kvm_memslots {
355 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
356 /* The mapping table from slot id to the index in memslots[]. */
357 short id_to_index[KVM_MEM_SLOTS_NUM];
364 struct mutex slots_lock;
365 struct mm_struct *mm; /* userspace tied to this vm */
366 struct kvm_memslots *memslots[KVM_ADDRESS_SPACE_NUM];
367 struct srcu_struct srcu;
368 struct srcu_struct irq_srcu;
369 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
370 atomic_t online_vcpus;
371 int last_boosted_vcpu;
372 struct list_head vm_list;
374 struct kvm_io_bus *buses[KVM_NR_BUSES];
375 #ifdef CONFIG_HAVE_KVM_EVENTFD
378 struct list_head items;
379 struct list_head resampler_list;
380 struct mutex resampler_lock;
382 struct list_head ioeventfds;
384 struct kvm_vm_stat stat;
385 struct kvm_arch arch;
386 atomic_t users_count;
387 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
388 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
389 spinlock_t ring_lock;
390 struct list_head coalesced_zones;
393 struct mutex irq_lock;
394 #ifdef CONFIG_HAVE_KVM_IRQCHIP
396 * Update side is protected by irq_lock.
398 struct kvm_irq_routing_table __rcu *irq_routing;
400 #ifdef CONFIG_HAVE_KVM_IRQFD
401 struct hlist_head irq_ack_notifier_list;
404 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
405 struct mmu_notifier mmu_notifier;
406 unsigned long mmu_notifier_seq;
407 long mmu_notifier_count;
410 struct list_head devices;
413 #define kvm_err(fmt, ...) \
414 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
415 #define kvm_info(fmt, ...) \
416 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
417 #define kvm_debug(fmt, ...) \
418 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
419 #define kvm_pr_unimpl(fmt, ...) \
420 pr_err_ratelimited("kvm [%i]: " fmt, \
421 task_tgid_nr(current), ## __VA_ARGS__)
423 /* The guest did something we don't support. */
424 #define vcpu_unimpl(vcpu, fmt, ...) \
425 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
427 #define vcpu_debug(vcpu, fmt, ...) \
428 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
430 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
432 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
433 * the caller has read kvm->online_vcpus before (as is the case
434 * for kvm_for_each_vcpu, for example).
437 return kvm->vcpus[i];
440 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
442 idx < atomic_read(&kvm->online_vcpus) && \
443 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
446 #define kvm_for_each_memslot(memslot, slots) \
447 for (memslot = &slots->memslots[0]; \
448 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
451 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
452 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
454 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
455 void vcpu_put(struct kvm_vcpu *vcpu);
457 #ifdef __KVM_HAVE_IOAPIC
458 void kvm_vcpu_request_scan_ioapic(struct kvm *kvm);
460 static inline void kvm_vcpu_request_scan_ioapic(struct kvm *kvm)
465 #ifdef CONFIG_HAVE_KVM_IRQFD
466 int kvm_irqfd_init(void);
467 void kvm_irqfd_exit(void);
469 static inline int kvm_irqfd_init(void)
474 static inline void kvm_irqfd_exit(void)
478 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
479 struct module *module);
482 void kvm_get_kvm(struct kvm *kvm);
483 void kvm_put_kvm(struct kvm *kvm);
485 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
487 return rcu_dereference_check(kvm->memslots[as_id],
488 srcu_read_lock_held(&kvm->srcu)
489 || lockdep_is_held(&kvm->slots_lock));
492 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
494 return __kvm_memslots(kvm, 0);
497 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
499 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
501 return __kvm_memslots(vcpu->kvm, as_id);
504 static inline struct kvm_memory_slot *
505 id_to_memslot(struct kvm_memslots *slots, int id)
507 int index = slots->id_to_index[id];
508 struct kvm_memory_slot *slot;
510 slot = &slots->memslots[index];
512 WARN_ON(slot->id != id);
517 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
518 * - create a new memory slot
519 * - delete an existing memory slot
520 * - modify an existing memory slot
521 * -- move it in the guest physical memory space
522 * -- just change its flags
524 * Since flags can be changed by some of these operations, the following
525 * differentiation is the best we can do for __kvm_set_memory_region():
534 int kvm_set_memory_region(struct kvm *kvm,
535 const struct kvm_userspace_memory_region *mem);
536 int __kvm_set_memory_region(struct kvm *kvm,
537 const struct kvm_userspace_memory_region *mem);
538 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
539 struct kvm_memory_slot *dont);
540 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
541 unsigned long npages);
542 void kvm_arch_memslots_updated(struct kvm *kvm, struct kvm_memslots *slots);
543 int kvm_arch_prepare_memory_region(struct kvm *kvm,
544 struct kvm_memory_slot *memslot,
545 const struct kvm_userspace_memory_region *mem,
546 enum kvm_mr_change change);
547 void kvm_arch_commit_memory_region(struct kvm *kvm,
548 const struct kvm_userspace_memory_region *mem,
549 const struct kvm_memory_slot *old,
550 const struct kvm_memory_slot *new,
551 enum kvm_mr_change change);
552 bool kvm_largepages_enabled(void);
553 void kvm_disable_largepages(void);
554 /* flush all memory translations */
555 void kvm_arch_flush_shadow_all(struct kvm *kvm);
556 /* flush memory translations pointing to 'slot' */
557 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
558 struct kvm_memory_slot *slot);
560 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
561 struct page **pages, int nr_pages);
563 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
564 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
565 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
566 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
567 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
569 void kvm_release_page_clean(struct page *page);
570 void kvm_release_page_dirty(struct page *page);
571 void kvm_set_page_accessed(struct page *page);
573 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
574 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
575 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
577 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
578 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
579 pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn, bool atomic,
580 bool *async, bool write_fault, bool *writable);
582 void kvm_release_pfn_clean(pfn_t pfn);
583 void kvm_set_pfn_dirty(pfn_t pfn);
584 void kvm_set_pfn_accessed(pfn_t pfn);
585 void kvm_get_pfn(pfn_t pfn);
587 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
589 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
591 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
592 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
593 void *data, unsigned long len);
594 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
595 int offset, int len);
596 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
598 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
599 void *data, unsigned long len);
600 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
601 gpa_t gpa, unsigned long len);
602 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
603 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
604 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
605 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
606 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
607 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
609 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
610 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
611 pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
612 pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
613 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
614 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
615 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
616 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
618 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
620 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
622 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
623 int offset, int len);
624 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
626 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
628 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
629 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
630 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
631 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
632 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
633 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
635 void kvm_flush_remote_tlbs(struct kvm *kvm);
636 void kvm_reload_remote_mmus(struct kvm *kvm);
637 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
638 void kvm_make_scan_ioapic_request(struct kvm *kvm);
639 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
641 long kvm_arch_dev_ioctl(struct file *filp,
642 unsigned int ioctl, unsigned long arg);
643 long kvm_arch_vcpu_ioctl(struct file *filp,
644 unsigned int ioctl, unsigned long arg);
645 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
647 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
649 int kvm_get_dirty_log(struct kvm *kvm,
650 struct kvm_dirty_log *log, int *is_dirty);
652 int kvm_get_dirty_log_protect(struct kvm *kvm,
653 struct kvm_dirty_log *log, bool *is_dirty);
655 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
656 struct kvm_memory_slot *slot,
660 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
661 struct kvm_dirty_log *log);
663 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
665 long kvm_arch_vm_ioctl(struct file *filp,
666 unsigned int ioctl, unsigned long arg);
668 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
669 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
671 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
672 struct kvm_translation *tr);
674 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
675 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
676 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
677 struct kvm_sregs *sregs);
678 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
679 struct kvm_sregs *sregs);
680 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
681 struct kvm_mp_state *mp_state);
682 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
683 struct kvm_mp_state *mp_state);
684 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
685 struct kvm_guest_debug *dbg);
686 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
688 int kvm_arch_init(void *opaque);
689 void kvm_arch_exit(void);
691 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
692 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
694 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
696 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
697 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
698 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
699 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
700 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
701 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
702 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
704 int kvm_arch_hardware_enable(void);
705 void kvm_arch_hardware_disable(void);
706 int kvm_arch_hardware_setup(void);
707 void kvm_arch_hardware_unsetup(void);
708 void kvm_arch_check_processor_compat(void *rtn);
709 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
710 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
712 void *kvm_kvzalloc(unsigned long size);
714 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
715 static inline struct kvm *kvm_arch_alloc_vm(void)
717 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
720 static inline void kvm_arch_free_vm(struct kvm *kvm)
726 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
727 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
728 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
729 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
731 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
735 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
739 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
744 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
745 void kvm_arch_start_assignment(struct kvm *kvm);
746 void kvm_arch_end_assignment(struct kvm *kvm);
747 bool kvm_arch_has_assigned_device(struct kvm *kvm);
749 static inline void kvm_arch_start_assignment(struct kvm *kvm)
753 static inline void kvm_arch_end_assignment(struct kvm *kvm)
757 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
763 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
765 #ifdef __KVM_HAVE_ARCH_WQP
766 return vcpu->arch.wqp;
772 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
774 * returns true if the virtual interrupt controller is initialized and
775 * ready to accept virtual IRQ. On some architectures the virtual interrupt
776 * controller is dynamically instantiated and this is not always true.
778 bool kvm_arch_intc_initialized(struct kvm *kvm);
780 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
786 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
787 void kvm_arch_destroy_vm(struct kvm *kvm);
788 void kvm_arch_sync_events(struct kvm *kvm);
790 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
791 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
793 bool kvm_is_reserved_pfn(pfn_t pfn);
795 struct kvm_irq_ack_notifier {
796 struct hlist_node link;
798 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
801 int kvm_irq_map_gsi(struct kvm *kvm,
802 struct kvm_kernel_irq_routing_entry *entries, int gsi);
803 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
805 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
807 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
808 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
809 int irq_source_id, int level, bool line_status);
810 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
811 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
812 void kvm_register_irq_ack_notifier(struct kvm *kvm,
813 struct kvm_irq_ack_notifier *kian);
814 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
815 struct kvm_irq_ack_notifier *kian);
816 int kvm_request_irq_source_id(struct kvm *kvm);
817 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
819 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
820 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
821 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
823 static inline int kvm_iommu_map_pages(struct kvm *kvm,
824 struct kvm_memory_slot *slot)
829 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
830 struct kvm_memory_slot *slot)
835 /* must be called with irqs disabled */
836 static inline void __kvm_guest_enter(void)
839 /* KVM does not hold any references to rcu protected data when it
840 * switches CPU into a guest mode. In fact switching to a guest mode
841 * is very similar to exiting to userspace from rcu point of view. In
842 * addition CPU may stay in a guest mode for quite a long time (up to
843 * one time slice). Lets treat guest mode as quiescent state, just like
844 * we do with user-mode execution.
846 if (!context_tracking_cpu_is_enabled())
847 rcu_virt_note_context_switch(smp_processor_id());
850 /* must be called with irqs disabled */
851 static inline void __kvm_guest_exit(void)
856 static inline void kvm_guest_enter(void)
860 local_irq_save(flags);
862 local_irq_restore(flags);
865 static inline void kvm_guest_exit(void)
869 local_irq_save(flags);
871 local_irq_restore(flags);
875 * search_memslots() and __gfn_to_memslot() are here because they are
876 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
877 * gfn_to_memslot() itself isn't here as an inline because that would
878 * bloat other code too much.
880 static inline struct kvm_memory_slot *
881 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
883 int start = 0, end = slots->used_slots;
884 int slot = atomic_read(&slots->lru_slot);
885 struct kvm_memory_slot *memslots = slots->memslots;
887 if (gfn >= memslots[slot].base_gfn &&
888 gfn < memslots[slot].base_gfn + memslots[slot].npages)
889 return &memslots[slot];
891 while (start < end) {
892 slot = start + (end - start) / 2;
894 if (gfn >= memslots[slot].base_gfn)
900 if (gfn >= memslots[start].base_gfn &&
901 gfn < memslots[start].base_gfn + memslots[start].npages) {
902 atomic_set(&slots->lru_slot, start);
903 return &memslots[start];
909 static inline struct kvm_memory_slot *
910 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
912 return search_memslots(slots, gfn);
915 static inline unsigned long
916 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
918 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
921 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
923 return gfn_to_memslot(kvm, gfn)->id;
927 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
929 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
931 return slot->base_gfn + gfn_offset;
934 static inline gpa_t gfn_to_gpa(gfn_t gfn)
936 return (gpa_t)gfn << PAGE_SHIFT;
939 static inline gfn_t gpa_to_gfn(gpa_t gpa)
941 return (gfn_t)(gpa >> PAGE_SHIFT);
944 static inline hpa_t pfn_to_hpa(pfn_t pfn)
946 return (hpa_t)pfn << PAGE_SHIFT;
949 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
951 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
953 return kvm_is_error_hva(hva);
956 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
958 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
966 struct kvm_stats_debugfs_item {
969 enum kvm_stat_kind kind;
970 struct dentry *dentry;
972 extern struct kvm_stats_debugfs_item debugfs_entries[];
973 extern struct dentry *kvm_debugfs_dir;
975 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
976 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
978 if (unlikely(kvm->mmu_notifier_count))
981 * Ensure the read of mmu_notifier_count happens before the read
982 * of mmu_notifier_seq. This interacts with the smp_wmb() in
983 * mmu_notifier_invalidate_range_end to make sure that the caller
984 * either sees the old (non-zero) value of mmu_notifier_count or
985 * the new (incremented) value of mmu_notifier_seq.
986 * PowerPC Book3s HV KVM calls this under a per-page lock
987 * rather than under kvm->mmu_lock, for scalability, so
988 * can't rely on kvm->mmu_lock to keep things ordered.
991 if (kvm->mmu_notifier_seq != mmu_seq)
997 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1000 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1002 #define KVM_MAX_IRQ_ROUTES 1024
1005 int kvm_setup_default_irq_routing(struct kvm *kvm);
1006 int kvm_setup_empty_irq_routing(struct kvm *kvm);
1007 int kvm_set_irq_routing(struct kvm *kvm,
1008 const struct kvm_irq_routing_entry *entries,
1011 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
1012 const struct kvm_irq_routing_entry *ue);
1013 void kvm_free_irq_routing(struct kvm *kvm);
1017 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1021 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1023 #ifdef CONFIG_HAVE_KVM_EVENTFD
1025 void kvm_eventfd_init(struct kvm *kvm);
1026 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1028 #ifdef CONFIG_HAVE_KVM_IRQFD
1029 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1030 void kvm_irqfd_release(struct kvm *kvm);
1031 void kvm_irq_routing_update(struct kvm *);
1033 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1038 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1043 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1045 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1050 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1052 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1053 static inline void kvm_irq_routing_update(struct kvm *kvm)
1058 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1063 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1065 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
1066 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
1068 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
1071 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1073 set_bit(req, &vcpu->requests);
1076 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1078 if (test_bit(req, &vcpu->requests)) {
1079 clear_bit(req, &vcpu->requests);
1086 extern bool kvm_rebooting;
1089 struct kvm_device_ops *ops;
1092 struct list_head vm_node;
1095 /* create, destroy, and name are mandatory */
1096 struct kvm_device_ops {
1098 int (*create)(struct kvm_device *dev, u32 type);
1101 * Destroy is responsible for freeing dev.
1103 * Destroy may be called before or after destructors are called
1104 * on emulated I/O regions, depending on whether a reference is
1105 * held by a vcpu or other kvm component that gets destroyed
1106 * after the emulated I/O.
1108 void (*destroy)(struct kvm_device *dev);
1110 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1111 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1112 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1113 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1117 void kvm_device_get(struct kvm_device *dev);
1118 void kvm_device_put(struct kvm_device *dev);
1119 struct kvm_device *kvm_device_from_filp(struct file *filp);
1120 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1121 void kvm_unregister_device_ops(u32 type);
1123 extern struct kvm_device_ops kvm_mpic_ops;
1124 extern struct kvm_device_ops kvm_xics_ops;
1125 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1126 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1128 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1130 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1132 vcpu->spin_loop.in_spin_loop = val;
1134 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1136 vcpu->spin_loop.dy_eligible = val;
1139 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1141 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1145 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1148 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */