5 * This work is licensed under the terms of the GNU GPL, version 2. See
6 * the COPYING file in the top-level directory.
9 #include <linux/types.h>
10 #include <linux/hardirq.h>
11 #include <linux/list.h>
12 #include <linux/mutex.h>
13 #include <linux/spinlock.h>
14 #include <linux/signal.h>
15 #include <linux/sched.h>
16 #include <linux/bug.h>
18 #include <linux/mmu_notifier.h>
19 #include <linux/preempt.h>
20 #include <linux/msi.h>
21 #include <linux/slab.h>
22 #include <linux/rcupdate.h>
23 #include <linux/ratelimit.h>
24 #include <linux/err.h>
25 #include <linux/irqflags.h>
26 #include <linux/context_tracking.h>
27 #include <asm/signal.h>
29 #include <linux/kvm.h>
30 #include <linux/kvm_para.h>
32 #include <linux/kvm_types.h>
34 #include <asm/kvm_host.h>
37 #define KVM_MMIO_SIZE 8
41 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
42 * in kvm, other bits are visible for userspace which are defined in
43 * include/linux/kvm_h.
45 #define KVM_MEMSLOT_INVALID (1UL << 16)
46 #define KVM_MEMSLOT_INCOHERENT (1UL << 17)
48 /* Two fragments for cross MMIO pages. */
49 #define KVM_MAX_MMIO_FRAGMENTS 2
52 * For the normal pfn, the highest 12 bits should be zero,
53 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
54 * mask bit 63 to indicate the noslot pfn.
56 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
57 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
58 #define KVM_PFN_NOSLOT (0x1ULL << 63)
60 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
61 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
62 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
65 * error pfns indicate that the gfn is in slot but faild to
66 * translate it to pfn on host.
68 static inline bool is_error_pfn(pfn_t pfn)
70 return !!(pfn & KVM_PFN_ERR_MASK);
74 * error_noslot pfns indicate that the gfn can not be
75 * translated to pfn - it is not in slot or failed to
76 * translate it to pfn.
78 static inline bool is_error_noslot_pfn(pfn_t pfn)
80 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
83 /* noslot pfn indicates that the gfn is not in slot. */
84 static inline bool is_noslot_pfn(pfn_t pfn)
86 return pfn == KVM_PFN_NOSLOT;
90 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
91 * provide own defines and kvm_is_error_hva
93 #ifndef KVM_HVA_ERR_BAD
95 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
96 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
98 static inline bool kvm_is_error_hva(unsigned long addr)
100 return addr >= PAGE_OFFSET;
105 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
107 static inline bool is_error_page(struct page *page)
113 * vcpu->requests bit members
115 #define KVM_REQ_TLB_FLUSH 0
116 #define KVM_REQ_MIGRATE_TIMER 1
117 #define KVM_REQ_REPORT_TPR_ACCESS 2
118 #define KVM_REQ_MMU_RELOAD 3
119 #define KVM_REQ_TRIPLE_FAULT 4
120 #define KVM_REQ_PENDING_TIMER 5
121 #define KVM_REQ_UNHALT 6
122 #define KVM_REQ_MMU_SYNC 7
123 #define KVM_REQ_CLOCK_UPDATE 8
124 #define KVM_REQ_KICK 9
125 #define KVM_REQ_DEACTIVATE_FPU 10
126 #define KVM_REQ_EVENT 11
127 #define KVM_REQ_APF_HALT 12
128 #define KVM_REQ_STEAL_UPDATE 13
129 #define KVM_REQ_NMI 14
130 #define KVM_REQ_PMU 15
131 #define KVM_REQ_PMI 16
132 #define KVM_REQ_WATCHDOG 17
133 #define KVM_REQ_MASTERCLOCK_UPDATE 18
134 #define KVM_REQ_MCLOCK_INPROGRESS 19
135 #define KVM_REQ_EPR_EXIT 20
136 #define KVM_REQ_SCAN_IOAPIC 21
137 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 22
138 #define KVM_REQ_ENABLE_IBS 23
139 #define KVM_REQ_DISABLE_IBS 24
140 #define KVM_REQ_APIC_PAGE_RELOAD 25
142 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
143 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
145 extern struct kmem_cache *kvm_vcpu_cache;
147 extern spinlock_t kvm_lock;
148 extern struct list_head vm_list;
150 struct kvm_io_range {
153 struct kvm_io_device *dev;
156 #define NR_IOBUS_DEVS 1000
161 struct kvm_io_range range[];
167 KVM_VIRTIO_CCW_NOTIFY_BUS,
172 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
173 int len, const void *val);
174 int kvm_io_bus_write_cookie(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
175 int len, const void *val, long cookie);
176 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
178 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
179 int len, struct kvm_io_device *dev);
180 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
181 struct kvm_io_device *dev);
183 #ifdef CONFIG_KVM_ASYNC_PF
184 struct kvm_async_pf {
185 struct work_struct work;
186 struct list_head link;
187 struct list_head queue;
188 struct kvm_vcpu *vcpu;
189 struct mm_struct *mm;
192 struct kvm_arch_async_pf arch;
196 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
197 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
198 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
199 struct kvm_arch_async_pf *arch);
200 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
204 * Carry out a gup that requires IO. Allow the mm to relinquish the mmap
205 * semaphore if the filemap/swap has to wait on a page lock. pagep == NULL
206 * controls whether we retry the gup one more time to completion in that case.
207 * Typically this is called after a FAULT_FLAG_RETRY_NOWAIT in the main tdp
210 int kvm_get_user_page_io(struct task_struct *tsk, struct mm_struct *mm,
211 unsigned long addr, bool write_fault,
212 struct page **pagep);
218 READING_SHADOW_PAGE_TABLES,
222 * Sometimes a large or cross-page mmio needs to be broken up into separate
223 * exits for userspace servicing.
225 struct kvm_mmio_fragment {
233 #ifdef CONFIG_PREEMPT_NOTIFIERS
234 struct preempt_notifier preempt_notifier;
240 unsigned long requests;
241 unsigned long guest_debug;
247 int guest_fpu_loaded, guest_xcr0_loaded;
248 wait_queue_head_t wq;
252 struct kvm_vcpu_stat stat;
254 #ifdef CONFIG_HAS_IOMEM
256 int mmio_read_completed;
258 int mmio_cur_fragment;
259 int mmio_nr_fragments;
260 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
263 #ifdef CONFIG_KVM_ASYNC_PF
266 struct list_head queue;
267 struct list_head done;
272 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
274 * Cpu relax intercept or pause loop exit optimization
275 * in_spin_loop: set when a vcpu does a pause loop exit
276 * or cpu relax intercepted.
277 * dy_eligible: indicates whether vcpu is eligible for directed yield.
285 struct kvm_vcpu_arch arch;
288 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
290 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
294 * Some of the bitops functions do not support too long bitmaps.
295 * This number must be determined not to exceed such limits.
297 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
299 struct kvm_memory_slot {
301 unsigned long npages;
302 unsigned long *dirty_bitmap;
303 struct kvm_arch_memory_slot arch;
304 unsigned long userspace_addr;
309 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
311 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
314 struct kvm_s390_adapter_int {
322 struct kvm_kernel_irq_routing_entry {
325 int (*set)(struct kvm_kernel_irq_routing_entry *e,
326 struct kvm *kvm, int irq_source_id, int level,
334 struct kvm_s390_adapter_int adapter;
336 struct hlist_node link;
339 #ifndef KVM_PRIVATE_MEM_SLOTS
340 #define KVM_PRIVATE_MEM_SLOTS 0
343 #ifndef KVM_MEM_SLOTS_NUM
344 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
349 * memslots are not sorted by id anymore, please use id_to_memslot()
350 * to get the memslot by its id.
352 struct kvm_memslots {
354 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
355 /* The mapping table from slot id to the index in memslots[]. */
356 short id_to_index[KVM_MEM_SLOTS_NUM];
363 struct mutex slots_lock;
364 struct mm_struct *mm; /* userspace tied to this vm */
365 struct kvm_memslots *memslots;
366 struct srcu_struct srcu;
367 struct srcu_struct irq_srcu;
368 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
371 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
372 atomic_t online_vcpus;
373 int last_boosted_vcpu;
374 struct list_head vm_list;
376 struct kvm_io_bus *buses[KVM_NR_BUSES];
377 #ifdef CONFIG_HAVE_KVM_EVENTFD
380 struct list_head items;
381 struct list_head resampler_list;
382 struct mutex resampler_lock;
384 struct list_head ioeventfds;
386 struct kvm_vm_stat stat;
387 struct kvm_arch arch;
388 atomic_t users_count;
389 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
390 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
391 spinlock_t ring_lock;
392 struct list_head coalesced_zones;
395 struct mutex irq_lock;
396 #ifdef CONFIG_HAVE_KVM_IRQCHIP
398 * Update side is protected by irq_lock.
400 struct kvm_irq_routing_table __rcu *irq_routing;
402 #ifdef CONFIG_HAVE_KVM_IRQFD
403 struct hlist_head irq_ack_notifier_list;
406 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
407 struct mmu_notifier mmu_notifier;
408 unsigned long mmu_notifier_seq;
409 long mmu_notifier_count;
412 struct list_head devices;
415 #define kvm_err(fmt, ...) \
416 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
417 #define kvm_info(fmt, ...) \
418 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
419 #define kvm_debug(fmt, ...) \
420 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
421 #define kvm_pr_unimpl(fmt, ...) \
422 pr_err_ratelimited("kvm [%i]: " fmt, \
423 task_tgid_nr(current), ## __VA_ARGS__)
425 /* The guest did something we don't support. */
426 #define vcpu_unimpl(vcpu, fmt, ...) \
427 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
429 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
432 return kvm->vcpus[i];
435 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
437 idx < atomic_read(&kvm->online_vcpus) && \
438 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
441 #define kvm_for_each_memslot(memslot, slots) \
442 for (memslot = &slots->memslots[0]; \
443 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
446 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
447 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
449 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
450 void vcpu_put(struct kvm_vcpu *vcpu);
452 #ifdef __KVM_HAVE_IOAPIC
453 void kvm_vcpu_request_scan_ioapic(struct kvm *kvm);
455 static inline void kvm_vcpu_request_scan_ioapic(struct kvm *kvm)
460 #ifdef CONFIG_HAVE_KVM_IRQFD
461 int kvm_irqfd_init(void);
462 void kvm_irqfd_exit(void);
464 static inline int kvm_irqfd_init(void)
469 static inline void kvm_irqfd_exit(void)
473 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
474 struct module *module);
477 void kvm_get_kvm(struct kvm *kvm);
478 void kvm_put_kvm(struct kvm *kvm);
480 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
482 return rcu_dereference_check(kvm->memslots,
483 srcu_read_lock_held(&kvm->srcu)
484 || lockdep_is_held(&kvm->slots_lock));
487 static inline struct kvm_memory_slot *
488 id_to_memslot(struct kvm_memslots *slots, int id)
490 int index = slots->id_to_index[id];
491 struct kvm_memory_slot *slot;
493 slot = &slots->memslots[index];
495 WARN_ON(slot->id != id);
500 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
501 * - create a new memory slot
502 * - delete an existing memory slot
503 * - modify an existing memory slot
504 * -- move it in the guest physical memory space
505 * -- just change its flags
507 * Since flags can be changed by some of these operations, the following
508 * differentiation is the best we can do for __kvm_set_memory_region():
517 int kvm_set_memory_region(struct kvm *kvm,
518 struct kvm_userspace_memory_region *mem);
519 int __kvm_set_memory_region(struct kvm *kvm,
520 struct kvm_userspace_memory_region *mem);
521 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
522 struct kvm_memory_slot *dont);
523 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
524 unsigned long npages);
525 void kvm_arch_memslots_updated(struct kvm *kvm);
526 int kvm_arch_prepare_memory_region(struct kvm *kvm,
527 struct kvm_memory_slot *memslot,
528 struct kvm_userspace_memory_region *mem,
529 enum kvm_mr_change change);
530 void kvm_arch_commit_memory_region(struct kvm *kvm,
531 struct kvm_userspace_memory_region *mem,
532 const struct kvm_memory_slot *old,
533 enum kvm_mr_change change);
534 bool kvm_largepages_enabled(void);
535 void kvm_disable_largepages(void);
536 /* flush all memory translations */
537 void kvm_arch_flush_shadow_all(struct kvm *kvm);
538 /* flush memory translations pointing to 'slot' */
539 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
540 struct kvm_memory_slot *slot);
542 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
545 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
546 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
547 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
548 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
549 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
551 void kvm_release_page_clean(struct page *page);
552 void kvm_release_page_dirty(struct page *page);
553 void kvm_set_page_accessed(struct page *page);
555 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
556 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
557 bool write_fault, bool *writable);
558 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
559 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
561 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
562 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
564 void kvm_release_pfn_clean(pfn_t pfn);
565 void kvm_set_pfn_dirty(pfn_t pfn);
566 void kvm_set_pfn_accessed(pfn_t pfn);
567 void kvm_get_pfn(pfn_t pfn);
569 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
571 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
573 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
574 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
575 void *data, unsigned long len);
576 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
577 int offset, int len);
578 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
580 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
581 void *data, unsigned long len);
582 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
583 gpa_t gpa, unsigned long len);
584 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
585 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
586 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
587 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
588 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
589 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
591 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
592 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
593 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
594 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
595 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
596 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
598 void kvm_flush_remote_tlbs(struct kvm *kvm);
599 void kvm_reload_remote_mmus(struct kvm *kvm);
600 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
601 void kvm_make_scan_ioapic_request(struct kvm *kvm);
602 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
604 long kvm_arch_dev_ioctl(struct file *filp,
605 unsigned int ioctl, unsigned long arg);
606 long kvm_arch_vcpu_ioctl(struct file *filp,
607 unsigned int ioctl, unsigned long arg);
608 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
610 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
612 int kvm_get_dirty_log(struct kvm *kvm,
613 struct kvm_dirty_log *log, int *is_dirty);
614 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
615 struct kvm_dirty_log *log);
617 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
619 long kvm_arch_vm_ioctl(struct file *filp,
620 unsigned int ioctl, unsigned long arg);
622 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
623 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
625 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
626 struct kvm_translation *tr);
628 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
629 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
630 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
631 struct kvm_sregs *sregs);
632 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
633 struct kvm_sregs *sregs);
634 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
635 struct kvm_mp_state *mp_state);
636 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
637 struct kvm_mp_state *mp_state);
638 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
639 struct kvm_guest_debug *dbg);
640 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
642 int kvm_arch_init(void *opaque);
643 void kvm_arch_exit(void);
645 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
646 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
648 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
650 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
651 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
652 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
653 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
654 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
655 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
656 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
658 int kvm_arch_hardware_enable(void);
659 void kvm_arch_hardware_disable(void);
660 int kvm_arch_hardware_setup(void);
661 void kvm_arch_hardware_unsetup(void);
662 void kvm_arch_check_processor_compat(void *rtn);
663 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
664 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
666 void *kvm_kvzalloc(unsigned long size);
667 void kvm_kvfree(const void *addr);
669 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
670 static inline struct kvm *kvm_arch_alloc_vm(void)
672 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
675 static inline void kvm_arch_free_vm(struct kvm *kvm)
681 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
682 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
683 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
684 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
686 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
690 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
694 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
700 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
702 #ifdef __KVM_HAVE_ARCH_WQP
703 return vcpu->arch.wqp;
709 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
710 void kvm_arch_destroy_vm(struct kvm *kvm);
711 void kvm_arch_sync_events(struct kvm *kvm);
713 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
714 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
716 bool kvm_is_reserved_pfn(pfn_t pfn);
718 struct kvm_irq_ack_notifier {
719 struct hlist_node link;
721 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
724 int kvm_irq_map_gsi(struct kvm *kvm,
725 struct kvm_kernel_irq_routing_entry *entries, int gsi);
726 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
728 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
730 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
731 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
732 int irq_source_id, int level, bool line_status);
733 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
734 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
735 void kvm_register_irq_ack_notifier(struct kvm *kvm,
736 struct kvm_irq_ack_notifier *kian);
737 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
738 struct kvm_irq_ack_notifier *kian);
739 int kvm_request_irq_source_id(struct kvm *kvm);
740 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
742 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
743 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
744 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
746 static inline int kvm_iommu_map_pages(struct kvm *kvm,
747 struct kvm_memory_slot *slot)
752 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
753 struct kvm_memory_slot *slot)
758 static inline void kvm_guest_enter(void)
762 BUG_ON(preemptible());
764 local_irq_save(flags);
766 local_irq_restore(flags);
768 /* KVM does not hold any references to rcu protected data when it
769 * switches CPU into a guest mode. In fact switching to a guest mode
770 * is very similar to exiting to userspace from rcu point of view. In
771 * addition CPU may stay in a guest mode for quite a long time (up to
772 * one time slice). Lets treat guest mode as quiescent state, just like
773 * we do with user-mode execution.
775 rcu_virt_note_context_switch(smp_processor_id());
778 static inline void kvm_guest_exit(void)
782 local_irq_save(flags);
784 local_irq_restore(flags);
788 * search_memslots() and __gfn_to_memslot() are here because they are
789 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
790 * gfn_to_memslot() itself isn't here as an inline because that would
791 * bloat other code too much.
793 static inline struct kvm_memory_slot *
794 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
796 int start = 0, end = slots->used_slots;
797 int slot = atomic_read(&slots->lru_slot);
798 struct kvm_memory_slot *memslots = slots->memslots;
800 if (gfn >= memslots[slot].base_gfn &&
801 gfn < memslots[slot].base_gfn + memslots[slot].npages)
802 return &memslots[slot];
804 while (start < end) {
805 slot = start + (end - start) / 2;
807 if (gfn >= memslots[slot].base_gfn)
813 if (gfn >= memslots[start].base_gfn &&
814 gfn < memslots[start].base_gfn + memslots[start].npages) {
815 atomic_set(&slots->lru_slot, start);
816 return &memslots[start];
822 static inline struct kvm_memory_slot *
823 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
825 return search_memslots(slots, gfn);
828 static inline unsigned long
829 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
831 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
834 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
836 return gfn_to_memslot(kvm, gfn)->id;
840 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
842 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
844 return slot->base_gfn + gfn_offset;
847 static inline gpa_t gfn_to_gpa(gfn_t gfn)
849 return (gpa_t)gfn << PAGE_SHIFT;
852 static inline gfn_t gpa_to_gfn(gpa_t gpa)
854 return (gfn_t)(gpa >> PAGE_SHIFT);
857 static inline hpa_t pfn_to_hpa(pfn_t pfn)
859 return (hpa_t)pfn << PAGE_SHIFT;
862 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
864 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
866 return kvm_is_error_hva(hva);
869 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
871 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
879 struct kvm_stats_debugfs_item {
882 enum kvm_stat_kind kind;
883 struct dentry *dentry;
885 extern struct kvm_stats_debugfs_item debugfs_entries[];
886 extern struct dentry *kvm_debugfs_dir;
888 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
889 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
891 if (unlikely(kvm->mmu_notifier_count))
894 * Ensure the read of mmu_notifier_count happens before the read
895 * of mmu_notifier_seq. This interacts with the smp_wmb() in
896 * mmu_notifier_invalidate_range_end to make sure that the caller
897 * either sees the old (non-zero) value of mmu_notifier_count or
898 * the new (incremented) value of mmu_notifier_seq.
899 * PowerPC Book3s HV KVM calls this under a per-page lock
900 * rather than under kvm->mmu_lock, for scalability, so
901 * can't rely on kvm->mmu_lock to keep things ordered.
904 if (kvm->mmu_notifier_seq != mmu_seq)
910 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
913 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
915 #define KVM_MAX_IRQ_ROUTES 1024
918 int kvm_setup_default_irq_routing(struct kvm *kvm);
919 int kvm_set_irq_routing(struct kvm *kvm,
920 const struct kvm_irq_routing_entry *entries,
923 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
924 const struct kvm_irq_routing_entry *ue);
925 void kvm_free_irq_routing(struct kvm *kvm);
929 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
933 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
935 #ifdef CONFIG_HAVE_KVM_EVENTFD
937 void kvm_eventfd_init(struct kvm *kvm);
938 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
940 #ifdef CONFIG_HAVE_KVM_IRQFD
941 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
942 void kvm_irqfd_release(struct kvm *kvm);
943 void kvm_irq_routing_update(struct kvm *);
945 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
950 static inline void kvm_irqfd_release(struct kvm *kvm) {}
955 static inline void kvm_eventfd_init(struct kvm *kvm) {}
957 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
962 static inline void kvm_irqfd_release(struct kvm *kvm) {}
964 #ifdef CONFIG_HAVE_KVM_IRQCHIP
965 static inline void kvm_irq_routing_update(struct kvm *kvm)
970 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
975 #endif /* CONFIG_HAVE_KVM_EVENTFD */
977 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
978 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
980 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
983 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
987 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
991 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
993 set_bit(req, &vcpu->requests);
996 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
998 if (test_bit(req, &vcpu->requests)) {
999 clear_bit(req, &vcpu->requests);
1006 extern bool kvm_rebooting;
1009 struct kvm_device_ops *ops;
1012 struct list_head vm_node;
1015 /* create, destroy, and name are mandatory */
1016 struct kvm_device_ops {
1018 int (*create)(struct kvm_device *dev, u32 type);
1021 * Destroy is responsible for freeing dev.
1023 * Destroy may be called before or after destructors are called
1024 * on emulated I/O regions, depending on whether a reference is
1025 * held by a vcpu or other kvm component that gets destroyed
1026 * after the emulated I/O.
1028 void (*destroy)(struct kvm_device *dev);
1030 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1031 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1032 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1033 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1037 void kvm_device_get(struct kvm_device *dev);
1038 void kvm_device_put(struct kvm_device *dev);
1039 struct kvm_device *kvm_device_from_filp(struct file *filp);
1040 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1041 void kvm_unregister_device_ops(u32 type);
1043 extern struct kvm_device_ops kvm_mpic_ops;
1044 extern struct kvm_device_ops kvm_xics_ops;
1046 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1048 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1050 vcpu->spin_loop.in_spin_loop = val;
1052 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1054 vcpu->spin_loop.dy_eligible = val;
1057 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1059 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1063 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1066 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */