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 <linux/irqbypass.h>
28 #include <asm/signal.h>
30 #include <linux/kvm.h>
31 #include <linux/kvm_para.h>
33 #include <linux/kvm_types.h>
35 #include <asm/kvm_host.h>
38 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
39 * in kvm, other bits are visible for userspace which are defined in
40 * include/linux/kvm_h.
42 #define KVM_MEMSLOT_INVALID (1UL << 16)
43 #define KVM_MEMSLOT_INCOHERENT (1UL << 17)
45 /* Two fragments for cross MMIO pages. */
46 #define KVM_MAX_MMIO_FRAGMENTS 2
48 #ifndef KVM_ADDRESS_SPACE_NUM
49 #define KVM_ADDRESS_SPACE_NUM 1
53 * For the normal pfn, the highest 12 bits should be zero,
54 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
55 * mask bit 63 to indicate the noslot pfn.
57 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
58 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
59 #define KVM_PFN_NOSLOT (0x1ULL << 63)
61 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
62 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
63 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
66 * error pfns indicate that the gfn is in slot but faild to
67 * translate it to pfn on host.
69 static inline bool is_error_pfn(pfn_t pfn)
71 return !!(pfn & KVM_PFN_ERR_MASK);
75 * error_noslot pfns indicate that the gfn can not be
76 * translated to pfn - it is not in slot or failed to
77 * translate it to pfn.
79 static inline bool is_error_noslot_pfn(pfn_t pfn)
81 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
84 /* noslot pfn indicates that the gfn is not in slot. */
85 static inline bool is_noslot_pfn(pfn_t pfn)
87 return pfn == KVM_PFN_NOSLOT;
91 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
92 * provide own defines and kvm_is_error_hva
94 #ifndef KVM_HVA_ERR_BAD
96 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
97 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
99 static inline bool kvm_is_error_hva(unsigned long addr)
101 return addr >= PAGE_OFFSET;
106 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
108 static inline bool is_error_page(struct page *page)
114 * vcpu->requests bit members
116 #define KVM_REQ_TLB_FLUSH 0
117 #define KVM_REQ_MIGRATE_TIMER 1
118 #define KVM_REQ_REPORT_TPR_ACCESS 2
119 #define KVM_REQ_MMU_RELOAD 3
120 #define KVM_REQ_TRIPLE_FAULT 4
121 #define KVM_REQ_PENDING_TIMER 5
122 #define KVM_REQ_UNHALT 6
123 #define KVM_REQ_MMU_SYNC 7
124 #define KVM_REQ_CLOCK_UPDATE 8
125 #define KVM_REQ_KICK 9
126 #define KVM_REQ_DEACTIVATE_FPU 10
127 #define KVM_REQ_EVENT 11
128 #define KVM_REQ_APF_HALT 12
129 #define KVM_REQ_STEAL_UPDATE 13
130 #define KVM_REQ_NMI 14
131 #define KVM_REQ_PMU 15
132 #define KVM_REQ_PMI 16
133 #define KVM_REQ_WATCHDOG 17
134 #define KVM_REQ_MASTERCLOCK_UPDATE 18
135 #define KVM_REQ_MCLOCK_INPROGRESS 19
136 #define KVM_REQ_EPR_EXIT 20
137 #define KVM_REQ_SCAN_IOAPIC 21
138 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 22
139 #define KVM_REQ_ENABLE_IBS 23
140 #define KVM_REQ_DISABLE_IBS 24
141 #define KVM_REQ_APIC_PAGE_RELOAD 25
142 #define KVM_REQ_SMI 26
143 #define KVM_REQ_HV_CRASH 27
144 #define KVM_REQ_IOAPIC_EOI_EXIT 28
145 #define KVM_REQ_HV_RESET 29
147 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
148 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
150 extern struct kmem_cache *kvm_vcpu_cache;
152 extern spinlock_t kvm_lock;
153 extern struct list_head vm_list;
155 struct kvm_io_range {
158 struct kvm_io_device *dev;
161 #define NR_IOBUS_DEVS 1000
166 struct kvm_io_range range[];
172 KVM_VIRTIO_CCW_NOTIFY_BUS,
177 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
178 int len, const void *val);
179 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
180 gpa_t addr, int len, const void *val, long cookie);
181 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
183 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
184 int len, struct kvm_io_device *dev);
185 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
186 struct kvm_io_device *dev);
188 #ifdef CONFIG_KVM_ASYNC_PF
189 struct kvm_async_pf {
190 struct work_struct work;
191 struct list_head link;
192 struct list_head queue;
193 struct kvm_vcpu *vcpu;
194 struct mm_struct *mm;
197 struct kvm_arch_async_pf arch;
201 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
202 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
203 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
204 struct kvm_arch_async_pf *arch);
205 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
212 READING_SHADOW_PAGE_TABLES,
216 * Sometimes a large or cross-page mmio needs to be broken up into separate
217 * exits for userspace servicing.
219 struct kvm_mmio_fragment {
227 #ifdef CONFIG_PREEMPT_NOTIFIERS
228 struct preempt_notifier preempt_notifier;
234 unsigned long requests;
235 unsigned long guest_debug;
241 int guest_fpu_loaded, guest_xcr0_loaded;
242 unsigned char fpu_counter;
243 wait_queue_head_t wq;
247 struct kvm_vcpu_stat stat;
248 unsigned int halt_poll_ns;
250 #ifdef CONFIG_HAS_IOMEM
252 int mmio_read_completed;
254 int mmio_cur_fragment;
255 int mmio_nr_fragments;
256 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
259 #ifdef CONFIG_KVM_ASYNC_PF
262 struct list_head queue;
263 struct list_head done;
268 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
270 * Cpu relax intercept or pause loop exit optimization
271 * in_spin_loop: set when a vcpu does a pause loop exit
272 * or cpu relax intercepted.
273 * dy_eligible: indicates whether vcpu is eligible for directed yield.
281 struct kvm_vcpu_arch arch;
284 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
286 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
290 * Some of the bitops functions do not support too long bitmaps.
291 * This number must be determined not to exceed such limits.
293 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
295 struct kvm_memory_slot {
297 unsigned long npages;
298 unsigned long *dirty_bitmap;
299 struct kvm_arch_memory_slot arch;
300 unsigned long userspace_addr;
305 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
307 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
310 struct kvm_s390_adapter_int {
318 struct kvm_kernel_irq_routing_entry {
321 int (*set)(struct kvm_kernel_irq_routing_entry *e,
322 struct kvm *kvm, int irq_source_id, int level,
330 struct kvm_s390_adapter_int adapter;
332 struct hlist_node link;
335 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
336 struct kvm_irq_routing_table {
337 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
340 * Array indexed by gsi. Each entry contains list of irq chips
341 * the gsi is connected to.
343 struct hlist_head map[0];
347 #ifndef KVM_PRIVATE_MEM_SLOTS
348 #define KVM_PRIVATE_MEM_SLOTS 0
351 #ifndef KVM_MEM_SLOTS_NUM
352 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
355 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
356 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
364 * memslots are not sorted by id anymore, please use id_to_memslot()
365 * to get the memslot by its id.
367 struct kvm_memslots {
369 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
370 /* The mapping table from slot id to the index in memslots[]. */
371 short id_to_index[KVM_MEM_SLOTS_NUM];
378 struct mutex slots_lock;
379 struct mm_struct *mm; /* userspace tied to this vm */
380 struct kvm_memslots *memslots[KVM_ADDRESS_SPACE_NUM];
381 struct srcu_struct srcu;
382 struct srcu_struct irq_srcu;
383 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
384 atomic_t online_vcpus;
385 int last_boosted_vcpu;
386 struct list_head vm_list;
388 struct kvm_io_bus *buses[KVM_NR_BUSES];
389 #ifdef CONFIG_HAVE_KVM_EVENTFD
392 struct list_head items;
393 struct list_head resampler_list;
394 struct mutex resampler_lock;
396 struct list_head ioeventfds;
398 struct kvm_vm_stat stat;
399 struct kvm_arch arch;
400 atomic_t users_count;
401 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
402 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
403 spinlock_t ring_lock;
404 struct list_head coalesced_zones;
407 struct mutex irq_lock;
408 #ifdef CONFIG_HAVE_KVM_IRQCHIP
410 * Update side is protected by irq_lock.
412 struct kvm_irq_routing_table __rcu *irq_routing;
414 #ifdef CONFIG_HAVE_KVM_IRQFD
415 struct hlist_head irq_ack_notifier_list;
418 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
419 struct mmu_notifier mmu_notifier;
420 unsigned long mmu_notifier_seq;
421 long mmu_notifier_count;
424 struct list_head devices;
427 #define kvm_err(fmt, ...) \
428 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
429 #define kvm_info(fmt, ...) \
430 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
431 #define kvm_debug(fmt, ...) \
432 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
433 #define kvm_pr_unimpl(fmt, ...) \
434 pr_err_ratelimited("kvm [%i]: " fmt, \
435 task_tgid_nr(current), ## __VA_ARGS__)
437 /* The guest did something we don't support. */
438 #define vcpu_unimpl(vcpu, fmt, ...) \
439 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
441 #define vcpu_debug(vcpu, fmt, ...) \
442 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
444 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
446 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
447 * the caller has read kvm->online_vcpus before (as is the case
448 * for kvm_for_each_vcpu, for example).
451 return kvm->vcpus[i];
454 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
456 idx < atomic_read(&kvm->online_vcpus) && \
457 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
460 #define kvm_for_each_memslot(memslot, slots) \
461 for (memslot = &slots->memslots[0]; \
462 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
465 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
466 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
468 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
469 void vcpu_put(struct kvm_vcpu *vcpu);
471 #ifdef __KVM_HAVE_IOAPIC
472 void kvm_vcpu_request_scan_ioapic(struct kvm *kvm);
473 void kvm_arch_irq_routing_update(struct kvm *kvm);
475 static inline void kvm_vcpu_request_scan_ioapic(struct kvm *kvm)
478 static inline void kvm_arch_irq_routing_update(struct kvm *kvm)
483 #ifdef CONFIG_HAVE_KVM_IRQFD
484 int kvm_irqfd_init(void);
485 void kvm_irqfd_exit(void);
487 static inline int kvm_irqfd_init(void)
492 static inline void kvm_irqfd_exit(void)
496 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
497 struct module *module);
500 void kvm_get_kvm(struct kvm *kvm);
501 void kvm_put_kvm(struct kvm *kvm);
503 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
505 return rcu_dereference_check(kvm->memslots[as_id],
506 srcu_read_lock_held(&kvm->srcu)
507 || lockdep_is_held(&kvm->slots_lock));
510 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
512 return __kvm_memslots(kvm, 0);
515 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
517 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
519 return __kvm_memslots(vcpu->kvm, as_id);
522 static inline struct kvm_memory_slot *
523 id_to_memslot(struct kvm_memslots *slots, int id)
525 int index = slots->id_to_index[id];
526 struct kvm_memory_slot *slot;
528 slot = &slots->memslots[index];
530 WARN_ON(slot->id != id);
535 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
536 * - create a new memory slot
537 * - delete an existing memory slot
538 * - modify an existing memory slot
539 * -- move it in the guest physical memory space
540 * -- just change its flags
542 * Since flags can be changed by some of these operations, the following
543 * differentiation is the best we can do for __kvm_set_memory_region():
552 int kvm_set_memory_region(struct kvm *kvm,
553 const struct kvm_userspace_memory_region *mem);
554 int __kvm_set_memory_region(struct kvm *kvm,
555 const struct kvm_userspace_memory_region *mem);
556 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
557 struct kvm_memory_slot *dont);
558 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
559 unsigned long npages);
560 void kvm_arch_memslots_updated(struct kvm *kvm, struct kvm_memslots *slots);
561 int kvm_arch_prepare_memory_region(struct kvm *kvm,
562 struct kvm_memory_slot *memslot,
563 const struct kvm_userspace_memory_region *mem,
564 enum kvm_mr_change change);
565 void kvm_arch_commit_memory_region(struct kvm *kvm,
566 const struct kvm_userspace_memory_region *mem,
567 const struct kvm_memory_slot *old,
568 const struct kvm_memory_slot *new,
569 enum kvm_mr_change change);
570 bool kvm_largepages_enabled(void);
571 void kvm_disable_largepages(void);
572 /* flush all memory translations */
573 void kvm_arch_flush_shadow_all(struct kvm *kvm);
574 /* flush memory translations pointing to 'slot' */
575 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
576 struct kvm_memory_slot *slot);
578 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
579 struct page **pages, int nr_pages);
581 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
582 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
583 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
584 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
585 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
587 void kvm_release_page_clean(struct page *page);
588 void kvm_release_page_dirty(struct page *page);
589 void kvm_set_page_accessed(struct page *page);
591 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
592 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
593 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
595 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
596 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
597 pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn, bool atomic,
598 bool *async, bool write_fault, bool *writable);
600 void kvm_release_pfn_clean(pfn_t pfn);
601 void kvm_set_pfn_dirty(pfn_t pfn);
602 void kvm_set_pfn_accessed(pfn_t pfn);
603 void kvm_get_pfn(pfn_t pfn);
605 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
607 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
609 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
610 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
611 void *data, unsigned long len);
612 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
613 int offset, int len);
614 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
616 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
617 void *data, unsigned long len);
618 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
619 gpa_t gpa, unsigned long len);
620 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
621 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
622 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
623 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
624 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
625 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
627 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
628 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
629 pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
630 pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
631 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
632 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
633 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
634 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
636 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
638 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
640 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
641 int offset, int len);
642 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
644 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
646 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
647 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
648 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
649 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
650 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
651 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
653 void kvm_flush_remote_tlbs(struct kvm *kvm);
654 void kvm_reload_remote_mmus(struct kvm *kvm);
655 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
656 void kvm_make_scan_ioapic_request(struct kvm *kvm);
657 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
659 long kvm_arch_dev_ioctl(struct file *filp,
660 unsigned int ioctl, unsigned long arg);
661 long kvm_arch_vcpu_ioctl(struct file *filp,
662 unsigned int ioctl, unsigned long arg);
663 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
665 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
667 int kvm_get_dirty_log(struct kvm *kvm,
668 struct kvm_dirty_log *log, int *is_dirty);
670 int kvm_get_dirty_log_protect(struct kvm *kvm,
671 struct kvm_dirty_log *log, bool *is_dirty);
673 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
674 struct kvm_memory_slot *slot,
678 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
679 struct kvm_dirty_log *log);
681 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
683 long kvm_arch_vm_ioctl(struct file *filp,
684 unsigned int ioctl, unsigned long arg);
686 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
687 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
689 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
690 struct kvm_translation *tr);
692 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
693 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
694 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
695 struct kvm_sregs *sregs);
696 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
697 struct kvm_sregs *sregs);
698 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
699 struct kvm_mp_state *mp_state);
700 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
701 struct kvm_mp_state *mp_state);
702 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
703 struct kvm_guest_debug *dbg);
704 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
706 int kvm_arch_init(void *opaque);
707 void kvm_arch_exit(void);
709 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
710 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
712 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
714 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
715 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
716 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
717 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
718 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
719 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
720 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
722 int kvm_arch_hardware_enable(void);
723 void kvm_arch_hardware_disable(void);
724 int kvm_arch_hardware_setup(void);
725 void kvm_arch_hardware_unsetup(void);
726 void kvm_arch_check_processor_compat(void *rtn);
727 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
728 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
730 void *kvm_kvzalloc(unsigned long size);
732 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
733 static inline struct kvm *kvm_arch_alloc_vm(void)
735 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
738 static inline void kvm_arch_free_vm(struct kvm *kvm)
744 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
745 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
746 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
747 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
749 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
753 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
757 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
762 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
763 void kvm_arch_start_assignment(struct kvm *kvm);
764 void kvm_arch_end_assignment(struct kvm *kvm);
765 bool kvm_arch_has_assigned_device(struct kvm *kvm);
767 static inline void kvm_arch_start_assignment(struct kvm *kvm)
771 static inline void kvm_arch_end_assignment(struct kvm *kvm)
775 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
781 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
783 #ifdef __KVM_HAVE_ARCH_WQP
784 return vcpu->arch.wqp;
790 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
792 * returns true if the virtual interrupt controller is initialized and
793 * ready to accept virtual IRQ. On some architectures the virtual interrupt
794 * controller is dynamically instantiated and this is not always true.
796 bool kvm_arch_intc_initialized(struct kvm *kvm);
798 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
804 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
805 void kvm_arch_destroy_vm(struct kvm *kvm);
806 void kvm_arch_sync_events(struct kvm *kvm);
808 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
809 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
811 bool kvm_is_reserved_pfn(pfn_t pfn);
813 struct kvm_irq_ack_notifier {
814 struct hlist_node link;
816 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
819 int kvm_irq_map_gsi(struct kvm *kvm,
820 struct kvm_kernel_irq_routing_entry *entries, int gsi);
821 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
823 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
825 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
826 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
827 int irq_source_id, int level, bool line_status);
828 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
829 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
830 void kvm_register_irq_ack_notifier(struct kvm *kvm,
831 struct kvm_irq_ack_notifier *kian);
832 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
833 struct kvm_irq_ack_notifier *kian);
834 int kvm_request_irq_source_id(struct kvm *kvm);
835 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
837 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
838 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
839 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
841 static inline int kvm_iommu_map_pages(struct kvm *kvm,
842 struct kvm_memory_slot *slot)
847 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
848 struct kvm_memory_slot *slot)
853 /* must be called with irqs disabled */
854 static inline void __kvm_guest_enter(void)
857 /* KVM does not hold any references to rcu protected data when it
858 * switches CPU into a guest mode. In fact switching to a guest mode
859 * is very similar to exiting to userspace from rcu point of view. In
860 * addition CPU may stay in a guest mode for quite a long time (up to
861 * one time slice). Lets treat guest mode as quiescent state, just like
862 * we do with user-mode execution.
864 if (!context_tracking_cpu_is_enabled())
865 rcu_virt_note_context_switch(smp_processor_id());
868 /* must be called with irqs disabled */
869 static inline void __kvm_guest_exit(void)
874 static inline void kvm_guest_enter(void)
878 local_irq_save(flags);
880 local_irq_restore(flags);
883 static inline void kvm_guest_exit(void)
887 local_irq_save(flags);
889 local_irq_restore(flags);
893 * search_memslots() and __gfn_to_memslot() are here because they are
894 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
895 * gfn_to_memslot() itself isn't here as an inline because that would
896 * bloat other code too much.
898 static inline struct kvm_memory_slot *
899 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
901 int start = 0, end = slots->used_slots;
902 int slot = atomic_read(&slots->lru_slot);
903 struct kvm_memory_slot *memslots = slots->memslots;
905 if (gfn >= memslots[slot].base_gfn &&
906 gfn < memslots[slot].base_gfn + memslots[slot].npages)
907 return &memslots[slot];
909 while (start < end) {
910 slot = start + (end - start) / 2;
912 if (gfn >= memslots[slot].base_gfn)
918 if (gfn >= memslots[start].base_gfn &&
919 gfn < memslots[start].base_gfn + memslots[start].npages) {
920 atomic_set(&slots->lru_slot, start);
921 return &memslots[start];
927 static inline struct kvm_memory_slot *
928 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
930 return search_memslots(slots, gfn);
933 static inline unsigned long
934 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
936 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
939 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
941 return gfn_to_memslot(kvm, gfn)->id;
945 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
947 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
949 return slot->base_gfn + gfn_offset;
952 static inline gpa_t gfn_to_gpa(gfn_t gfn)
954 return (gpa_t)gfn << PAGE_SHIFT;
957 static inline gfn_t gpa_to_gfn(gpa_t gpa)
959 return (gfn_t)(gpa >> PAGE_SHIFT);
962 static inline hpa_t pfn_to_hpa(pfn_t pfn)
964 return (hpa_t)pfn << PAGE_SHIFT;
967 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
969 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
971 return kvm_is_error_hva(hva);
974 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
976 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
984 struct kvm_stats_debugfs_item {
987 enum kvm_stat_kind kind;
988 struct dentry *dentry;
990 extern struct kvm_stats_debugfs_item debugfs_entries[];
991 extern struct dentry *kvm_debugfs_dir;
993 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
994 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
996 if (unlikely(kvm->mmu_notifier_count))
999 * Ensure the read of mmu_notifier_count happens before the read
1000 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1001 * mmu_notifier_invalidate_range_end to make sure that the caller
1002 * either sees the old (non-zero) value of mmu_notifier_count or
1003 * the new (incremented) value of mmu_notifier_seq.
1004 * PowerPC Book3s HV KVM calls this under a per-page lock
1005 * rather than under kvm->mmu_lock, for scalability, so
1006 * can't rely on kvm->mmu_lock to keep things ordered.
1009 if (kvm->mmu_notifier_seq != mmu_seq)
1015 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1018 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1020 #define KVM_MAX_IRQ_ROUTES 1024
1023 int kvm_setup_default_irq_routing(struct kvm *kvm);
1024 int kvm_setup_empty_irq_routing(struct kvm *kvm);
1025 int kvm_set_irq_routing(struct kvm *kvm,
1026 const struct kvm_irq_routing_entry *entries,
1029 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
1030 const struct kvm_irq_routing_entry *ue);
1031 void kvm_free_irq_routing(struct kvm *kvm);
1035 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1039 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1041 #ifdef CONFIG_HAVE_KVM_EVENTFD
1043 void kvm_eventfd_init(struct kvm *kvm);
1044 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1046 #ifdef CONFIG_HAVE_KVM_IRQFD
1047 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1048 void kvm_irqfd_release(struct kvm *kvm);
1049 void kvm_irq_routing_update(struct kvm *);
1051 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1056 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1061 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1063 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1068 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1070 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1071 static inline void kvm_irq_routing_update(struct kvm *kvm)
1076 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1081 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1083 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
1084 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
1086 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
1089 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1091 set_bit(req, &vcpu->requests);
1094 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1096 if (test_bit(req, &vcpu->requests)) {
1097 clear_bit(req, &vcpu->requests);
1104 extern bool kvm_rebooting;
1107 struct kvm_device_ops *ops;
1110 struct list_head vm_node;
1113 /* create, destroy, and name are mandatory */
1114 struct kvm_device_ops {
1116 int (*create)(struct kvm_device *dev, u32 type);
1119 * Destroy is responsible for freeing dev.
1121 * Destroy may be called before or after destructors are called
1122 * on emulated I/O regions, depending on whether a reference is
1123 * held by a vcpu or other kvm component that gets destroyed
1124 * after the emulated I/O.
1126 void (*destroy)(struct kvm_device *dev);
1128 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1129 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1130 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1131 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1135 void kvm_device_get(struct kvm_device *dev);
1136 void kvm_device_put(struct kvm_device *dev);
1137 struct kvm_device *kvm_device_from_filp(struct file *filp);
1138 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1139 void kvm_unregister_device_ops(u32 type);
1141 extern struct kvm_device_ops kvm_mpic_ops;
1142 extern struct kvm_device_ops kvm_xics_ops;
1143 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1144 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1146 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1148 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1150 vcpu->spin_loop.in_spin_loop = val;
1152 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1154 vcpu->spin_loop.dy_eligible = val;
1157 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1159 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1163 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1166 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1168 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1169 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1170 struct irq_bypass_producer *);
1171 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1172 struct irq_bypass_producer *);
1173 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1174 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1175 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1176 uint32_t guest_irq, bool set);
1177 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */