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 <linux/swait.h>
29 #include <asm/signal.h>
31 #include <linux/kvm.h>
32 #include <linux/kvm_para.h>
34 #include <linux/kvm_types.h>
36 #include <asm/kvm_host.h>
39 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
40 * in kvm, other bits are visible for userspace which are defined in
41 * include/linux/kvm_h.
43 #define KVM_MEMSLOT_INVALID (1UL << 16)
44 #define KVM_MEMSLOT_INCOHERENT (1UL << 17)
46 /* Two fragments for cross MMIO pages. */
47 #define KVM_MAX_MMIO_FRAGMENTS 2
49 #ifndef KVM_ADDRESS_SPACE_NUM
50 #define KVM_ADDRESS_SPACE_NUM 1
54 * For the normal pfn, the highest 12 bits should be zero,
55 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
56 * mask bit 63 to indicate the noslot pfn.
58 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
59 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
60 #define KVM_PFN_NOSLOT (0x1ULL << 63)
62 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
63 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
64 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
67 * error pfns indicate that the gfn is in slot but faild to
68 * translate it to pfn on host.
70 static inline bool is_error_pfn(kvm_pfn_t pfn)
72 return !!(pfn & KVM_PFN_ERR_MASK);
76 * error_noslot pfns indicate that the gfn can not be
77 * translated to pfn - it is not in slot or failed to
78 * translate it to pfn.
80 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn)
82 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
85 /* noslot pfn indicates that the gfn is not in slot. */
86 static inline bool is_noslot_pfn(kvm_pfn_t pfn)
88 return pfn == KVM_PFN_NOSLOT;
92 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
93 * provide own defines and kvm_is_error_hva
95 #ifndef KVM_HVA_ERR_BAD
97 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
98 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
100 static inline bool kvm_is_error_hva(unsigned long addr)
102 return addr >= PAGE_OFFSET;
107 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
109 static inline bool is_error_page(struct page *page)
115 * Architecture-independent vcpu->requests bit members
116 * Bits 4-7 are reserved for more arch-independent bits.
118 #define KVM_REQ_TLB_FLUSH 0
119 #define KVM_REQ_MMU_RELOAD 1
120 #define KVM_REQ_PENDING_TIMER 2
121 #define KVM_REQ_UNHALT 3
123 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
124 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
126 extern struct kmem_cache *kvm_vcpu_cache;
128 extern spinlock_t kvm_lock;
129 extern struct list_head vm_list;
131 struct kvm_io_range {
134 struct kvm_io_device *dev;
137 #define NR_IOBUS_DEVS 1000
142 struct kvm_io_range range[];
148 KVM_VIRTIO_CCW_NOTIFY_BUS,
153 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
154 int len, const void *val);
155 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
156 gpa_t addr, int len, const void *val, long cookie);
157 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
159 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
160 int len, struct kvm_io_device *dev);
161 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
162 struct kvm_io_device *dev);
164 #ifdef CONFIG_KVM_ASYNC_PF
165 struct kvm_async_pf {
166 struct work_struct work;
167 struct list_head link;
168 struct list_head queue;
169 struct kvm_vcpu *vcpu;
170 struct mm_struct *mm;
173 struct kvm_arch_async_pf arch;
177 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
178 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
179 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
180 struct kvm_arch_async_pf *arch);
181 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
188 READING_SHADOW_PAGE_TABLES,
192 * Sometimes a large or cross-page mmio needs to be broken up into separate
193 * exits for userspace servicing.
195 struct kvm_mmio_fragment {
203 #ifdef CONFIG_PREEMPT_NOTIFIERS
204 struct preempt_notifier preempt_notifier;
210 unsigned long requests;
211 unsigned long guest_debug;
214 struct list_head blocked_vcpu_list;
220 int guest_fpu_loaded, guest_xcr0_loaded;
221 unsigned char fpu_counter;
222 struct swait_queue_head wq;
226 struct kvm_vcpu_stat stat;
227 unsigned int halt_poll_ns;
229 #ifdef CONFIG_HAS_IOMEM
231 int mmio_read_completed;
233 int mmio_cur_fragment;
234 int mmio_nr_fragments;
235 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
238 #ifdef CONFIG_KVM_ASYNC_PF
241 struct list_head queue;
242 struct list_head done;
247 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
249 * Cpu relax intercept or pause loop exit optimization
250 * in_spin_loop: set when a vcpu does a pause loop exit
251 * or cpu relax intercepted.
252 * dy_eligible: indicates whether vcpu is eligible for directed yield.
260 struct kvm_vcpu_arch arch;
263 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
265 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
269 * Some of the bitops functions do not support too long bitmaps.
270 * This number must be determined not to exceed such limits.
272 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
274 struct kvm_memory_slot {
276 unsigned long npages;
277 unsigned long *dirty_bitmap;
278 struct kvm_arch_memory_slot arch;
279 unsigned long userspace_addr;
284 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
286 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
289 struct kvm_s390_adapter_int {
302 struct kvm_kernel_irq_routing_entry {
305 int (*set)(struct kvm_kernel_irq_routing_entry *e,
306 struct kvm *kvm, int irq_source_id, int level,
314 struct kvm_s390_adapter_int adapter;
315 struct kvm_hv_sint hv_sint;
317 struct hlist_node link;
320 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
321 struct kvm_irq_routing_table {
322 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
325 * Array indexed by gsi. Each entry contains list of irq chips
326 * the gsi is connected to.
328 struct hlist_head map[0];
332 #ifndef KVM_PRIVATE_MEM_SLOTS
333 #define KVM_PRIVATE_MEM_SLOTS 0
336 #ifndef KVM_MEM_SLOTS_NUM
337 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
340 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
341 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
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[KVM_ADDRESS_SPACE_NUM];
366 struct srcu_struct srcu;
367 struct srcu_struct irq_srcu;
368 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
369 atomic_t online_vcpus;
370 int last_boosted_vcpu;
371 struct list_head vm_list;
373 struct kvm_io_bus *buses[KVM_NR_BUSES];
374 #ifdef CONFIG_HAVE_KVM_EVENTFD
377 struct list_head items;
378 struct list_head resampler_list;
379 struct mutex resampler_lock;
381 struct list_head ioeventfds;
383 struct kvm_vm_stat stat;
384 struct kvm_arch arch;
385 atomic_t users_count;
386 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
387 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
388 spinlock_t ring_lock;
389 struct list_head coalesced_zones;
392 struct mutex irq_lock;
393 #ifdef CONFIG_HAVE_KVM_IRQCHIP
395 * Update side is protected by irq_lock.
397 struct kvm_irq_routing_table __rcu *irq_routing;
399 #ifdef CONFIG_HAVE_KVM_IRQFD
400 struct hlist_head irq_ack_notifier_list;
403 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
404 struct mmu_notifier mmu_notifier;
405 unsigned long mmu_notifier_seq;
406 long mmu_notifier_count;
409 struct list_head devices;
412 #define kvm_err(fmt, ...) \
413 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
414 #define kvm_info(fmt, ...) \
415 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
416 #define kvm_debug(fmt, ...) \
417 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
418 #define kvm_pr_unimpl(fmt, ...) \
419 pr_err_ratelimited("kvm [%i]: " fmt, \
420 task_tgid_nr(current), ## __VA_ARGS__)
422 /* The guest did something we don't support. */
423 #define vcpu_unimpl(vcpu, fmt, ...) \
424 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
425 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
427 #define vcpu_debug(vcpu, fmt, ...) \
428 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
429 #define vcpu_err(vcpu, fmt, ...) \
430 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
432 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
434 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
435 * the caller has read kvm->online_vcpus before (as is the case
436 * for kvm_for_each_vcpu, for example).
439 return kvm->vcpus[i];
442 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
444 idx < atomic_read(&kvm->online_vcpus) && \
445 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
448 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
450 struct kvm_vcpu *vcpu = NULL;
455 if (id < KVM_MAX_VCPUS)
456 vcpu = kvm_get_vcpu(kvm, id);
457 if (vcpu && vcpu->vcpu_id == id)
459 kvm_for_each_vcpu(i, vcpu, kvm)
460 if (vcpu->vcpu_id == id)
465 #define kvm_for_each_memslot(memslot, slots) \
466 for (memslot = &slots->memslots[0]; \
467 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
470 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
471 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
473 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
474 void vcpu_put(struct kvm_vcpu *vcpu);
476 #ifdef __KVM_HAVE_IOAPIC
477 void kvm_vcpu_request_scan_ioapic(struct kvm *kvm);
478 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
480 static inline void kvm_vcpu_request_scan_ioapic(struct kvm *kvm)
483 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
488 #ifdef CONFIG_HAVE_KVM_IRQFD
489 int kvm_irqfd_init(void);
490 void kvm_irqfd_exit(void);
492 static inline int kvm_irqfd_init(void)
497 static inline void kvm_irqfd_exit(void)
501 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
502 struct module *module);
505 void kvm_get_kvm(struct kvm *kvm);
506 void kvm_put_kvm(struct kvm *kvm);
508 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
510 return rcu_dereference_check(kvm->memslots[as_id],
511 srcu_read_lock_held(&kvm->srcu)
512 || lockdep_is_held(&kvm->slots_lock));
515 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
517 return __kvm_memslots(kvm, 0);
520 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
522 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
524 return __kvm_memslots(vcpu->kvm, as_id);
527 static inline struct kvm_memory_slot *
528 id_to_memslot(struct kvm_memslots *slots, int id)
530 int index = slots->id_to_index[id];
531 struct kvm_memory_slot *slot;
533 slot = &slots->memslots[index];
535 WARN_ON(slot->id != id);
540 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
541 * - create a new memory slot
542 * - delete an existing memory slot
543 * - modify an existing memory slot
544 * -- move it in the guest physical memory space
545 * -- just change its flags
547 * Since flags can be changed by some of these operations, the following
548 * differentiation is the best we can do for __kvm_set_memory_region():
557 int kvm_set_memory_region(struct kvm *kvm,
558 const struct kvm_userspace_memory_region *mem);
559 int __kvm_set_memory_region(struct kvm *kvm,
560 const struct kvm_userspace_memory_region *mem);
561 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
562 struct kvm_memory_slot *dont);
563 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
564 unsigned long npages);
565 void kvm_arch_memslots_updated(struct kvm *kvm, struct kvm_memslots *slots);
566 int kvm_arch_prepare_memory_region(struct kvm *kvm,
567 struct kvm_memory_slot *memslot,
568 const struct kvm_userspace_memory_region *mem,
569 enum kvm_mr_change change);
570 void kvm_arch_commit_memory_region(struct kvm *kvm,
571 const struct kvm_userspace_memory_region *mem,
572 const struct kvm_memory_slot *old,
573 const struct kvm_memory_slot *new,
574 enum kvm_mr_change change);
575 bool kvm_largepages_enabled(void);
576 void kvm_disable_largepages(void);
577 /* flush all memory translations */
578 void kvm_arch_flush_shadow_all(struct kvm *kvm);
579 /* flush memory translations pointing to 'slot' */
580 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
581 struct kvm_memory_slot *slot);
583 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
584 struct page **pages, int nr_pages);
586 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
587 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
588 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
589 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
590 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
592 void kvm_release_page_clean(struct page *page);
593 void kvm_release_page_dirty(struct page *page);
594 void kvm_set_page_accessed(struct page *page);
596 kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
597 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
598 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
600 kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
601 kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
602 kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
603 bool atomic, bool *async, bool write_fault,
606 void kvm_release_pfn_clean(kvm_pfn_t pfn);
607 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
608 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
609 void kvm_get_pfn(kvm_pfn_t pfn);
611 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
613 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
615 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
616 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
617 void *data, unsigned long len);
618 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
619 int offset, int len);
620 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
622 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
623 void *data, unsigned long len);
624 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
625 gpa_t gpa, unsigned long len);
626 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
627 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
628 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
629 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
630 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
631 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
633 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
634 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
635 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
636 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
637 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
638 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
639 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
640 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
642 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
644 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
646 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
647 int offset, int len);
648 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
650 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
652 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
653 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
654 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
655 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
656 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
657 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
658 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
659 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
661 void kvm_flush_remote_tlbs(struct kvm *kvm);
662 void kvm_reload_remote_mmus(struct kvm *kvm);
663 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
665 long kvm_arch_dev_ioctl(struct file *filp,
666 unsigned int ioctl, unsigned long arg);
667 long kvm_arch_vcpu_ioctl(struct file *filp,
668 unsigned int ioctl, unsigned long arg);
669 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
671 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
673 int kvm_get_dirty_log(struct kvm *kvm,
674 struct kvm_dirty_log *log, int *is_dirty);
676 int kvm_get_dirty_log_protect(struct kvm *kvm,
677 struct kvm_dirty_log *log, bool *is_dirty);
679 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
680 struct kvm_memory_slot *slot,
684 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
685 struct kvm_dirty_log *log);
687 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
689 long kvm_arch_vm_ioctl(struct file *filp,
690 unsigned int ioctl, unsigned long arg);
692 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
693 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
695 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
696 struct kvm_translation *tr);
698 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
699 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
700 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
701 struct kvm_sregs *sregs);
702 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
703 struct kvm_sregs *sregs);
704 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
705 struct kvm_mp_state *mp_state);
706 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
707 struct kvm_mp_state *mp_state);
708 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
709 struct kvm_guest_debug *dbg);
710 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
712 int kvm_arch_init(void *opaque);
713 void kvm_arch_exit(void);
715 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
716 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
718 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
720 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
721 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
722 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
723 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
724 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
725 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
726 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
728 int kvm_arch_hardware_enable(void);
729 void kvm_arch_hardware_disable(void);
730 int kvm_arch_hardware_setup(void);
731 void kvm_arch_hardware_unsetup(void);
732 void kvm_arch_check_processor_compat(void *rtn);
733 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
734 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
736 void *kvm_kvzalloc(unsigned long size);
738 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
739 static inline struct kvm *kvm_arch_alloc_vm(void)
741 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
744 static inline void kvm_arch_free_vm(struct kvm *kvm)
750 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
751 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
752 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
753 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
755 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
759 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
763 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
768 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
769 void kvm_arch_start_assignment(struct kvm *kvm);
770 void kvm_arch_end_assignment(struct kvm *kvm);
771 bool kvm_arch_has_assigned_device(struct kvm *kvm);
773 static inline void kvm_arch_start_assignment(struct kvm *kvm)
777 static inline void kvm_arch_end_assignment(struct kvm *kvm)
781 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
787 static inline struct swait_queue_head *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
789 #ifdef __KVM_HAVE_ARCH_WQP
790 return vcpu->arch.wqp;
796 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
798 * returns true if the virtual interrupt controller is initialized and
799 * ready to accept virtual IRQ. On some architectures the virtual interrupt
800 * controller is dynamically instantiated and this is not always true.
802 bool kvm_arch_intc_initialized(struct kvm *kvm);
804 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
810 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
811 void kvm_arch_destroy_vm(struct kvm *kvm);
812 void kvm_arch_sync_events(struct kvm *kvm);
814 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
815 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
817 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
819 struct kvm_irq_ack_notifier {
820 struct hlist_node link;
822 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
825 int kvm_irq_map_gsi(struct kvm *kvm,
826 struct kvm_kernel_irq_routing_entry *entries, int gsi);
827 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
829 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
831 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
832 int irq_source_id, int level, bool line_status);
833 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
834 struct kvm *kvm, int irq_source_id,
835 int level, bool line_status);
836 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
837 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
838 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
839 void kvm_register_irq_ack_notifier(struct kvm *kvm,
840 struct kvm_irq_ack_notifier *kian);
841 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
842 struct kvm_irq_ack_notifier *kian);
843 int kvm_request_irq_source_id(struct kvm *kvm);
844 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
846 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
847 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
848 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
850 static inline int kvm_iommu_map_pages(struct kvm *kvm,
851 struct kvm_memory_slot *slot)
856 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
857 struct kvm_memory_slot *slot)
862 /* must be called with irqs disabled */
863 static inline void __kvm_guest_enter(void)
866 /* KVM does not hold any references to rcu protected data when it
867 * switches CPU into a guest mode. In fact switching to a guest mode
868 * is very similar to exiting to userspace from rcu point of view. In
869 * addition CPU may stay in a guest mode for quite a long time (up to
870 * one time slice). Lets treat guest mode as quiescent state, just like
871 * we do with user-mode execution.
873 if (!context_tracking_cpu_is_enabled())
874 rcu_virt_note_context_switch(smp_processor_id());
877 /* must be called with irqs disabled */
878 static inline void __kvm_guest_exit(void)
883 static inline void kvm_guest_enter(void)
887 local_irq_save(flags);
889 local_irq_restore(flags);
892 static inline void kvm_guest_exit(void)
896 local_irq_save(flags);
898 local_irq_restore(flags);
902 * search_memslots() and __gfn_to_memslot() are here because they are
903 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
904 * gfn_to_memslot() itself isn't here as an inline because that would
905 * bloat other code too much.
907 static inline struct kvm_memory_slot *
908 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
910 int start = 0, end = slots->used_slots;
911 int slot = atomic_read(&slots->lru_slot);
912 struct kvm_memory_slot *memslots = slots->memslots;
914 if (gfn >= memslots[slot].base_gfn &&
915 gfn < memslots[slot].base_gfn + memslots[slot].npages)
916 return &memslots[slot];
918 while (start < end) {
919 slot = start + (end - start) / 2;
921 if (gfn >= memslots[slot].base_gfn)
927 if (gfn >= memslots[start].base_gfn &&
928 gfn < memslots[start].base_gfn + memslots[start].npages) {
929 atomic_set(&slots->lru_slot, start);
930 return &memslots[start];
936 static inline struct kvm_memory_slot *
937 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
939 return search_memslots(slots, gfn);
942 static inline unsigned long
943 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
945 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
948 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
950 return gfn_to_memslot(kvm, gfn)->id;
954 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
956 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
958 return slot->base_gfn + gfn_offset;
961 static inline gpa_t gfn_to_gpa(gfn_t gfn)
963 return (gpa_t)gfn << PAGE_SHIFT;
966 static inline gfn_t gpa_to_gfn(gpa_t gpa)
968 return (gfn_t)(gpa >> PAGE_SHIFT);
971 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
973 return (hpa_t)pfn << PAGE_SHIFT;
976 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
978 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
980 return kvm_is_error_hva(hva);
988 struct kvm_stats_debugfs_item {
991 enum kvm_stat_kind kind;
993 extern struct kvm_stats_debugfs_item debugfs_entries[];
994 extern struct dentry *kvm_debugfs_dir;
996 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
997 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
999 if (unlikely(kvm->mmu_notifier_count))
1002 * Ensure the read of mmu_notifier_count happens before the read
1003 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1004 * mmu_notifier_invalidate_range_end to make sure that the caller
1005 * either sees the old (non-zero) value of mmu_notifier_count or
1006 * the new (incremented) value of mmu_notifier_seq.
1007 * PowerPC Book3s HV KVM calls this under a per-page lock
1008 * rather than under kvm->mmu_lock, for scalability, so
1009 * can't rely on kvm->mmu_lock to keep things ordered.
1012 if (kvm->mmu_notifier_seq != mmu_seq)
1018 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1021 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1023 #define KVM_MAX_IRQ_ROUTES 1024
1026 int kvm_setup_default_irq_routing(struct kvm *kvm);
1027 int kvm_setup_empty_irq_routing(struct kvm *kvm);
1028 int kvm_set_irq_routing(struct kvm *kvm,
1029 const struct kvm_irq_routing_entry *entries,
1032 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
1033 const struct kvm_irq_routing_entry *ue);
1034 void kvm_free_irq_routing(struct kvm *kvm);
1038 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1042 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1044 #ifdef CONFIG_HAVE_KVM_EVENTFD
1046 void kvm_eventfd_init(struct kvm *kvm);
1047 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1049 #ifdef CONFIG_HAVE_KVM_IRQFD
1050 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1051 void kvm_irqfd_release(struct kvm *kvm);
1052 void kvm_irq_routing_update(struct kvm *);
1054 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1059 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1064 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1066 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1071 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1073 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1074 static inline void kvm_irq_routing_update(struct kvm *kvm)
1078 void kvm_arch_irq_routing_update(struct kvm *kvm);
1080 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1085 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1087 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
1088 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
1090 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
1093 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1096 * Ensure the rest of the request is published to kvm_check_request's
1097 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1100 set_bit(req, &vcpu->requests);
1103 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1105 if (test_bit(req, &vcpu->requests)) {
1106 clear_bit(req, &vcpu->requests);
1109 * Ensure the rest of the request is visible to kvm_check_request's
1110 * caller. Paired with the smp_wmb in kvm_make_request.
1112 smp_mb__after_atomic();
1119 extern bool kvm_rebooting;
1122 struct kvm_device_ops *ops;
1125 struct list_head vm_node;
1128 /* create, destroy, and name are mandatory */
1129 struct kvm_device_ops {
1131 int (*create)(struct kvm_device *dev, u32 type);
1134 * Destroy is responsible for freeing dev.
1136 * Destroy may be called before or after destructors are called
1137 * on emulated I/O regions, depending on whether a reference is
1138 * held by a vcpu or other kvm component that gets destroyed
1139 * after the emulated I/O.
1141 void (*destroy)(struct kvm_device *dev);
1143 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1144 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1145 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1146 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1150 void kvm_device_get(struct kvm_device *dev);
1151 void kvm_device_put(struct kvm_device *dev);
1152 struct kvm_device *kvm_device_from_filp(struct file *filp);
1153 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1154 void kvm_unregister_device_ops(u32 type);
1156 extern struct kvm_device_ops kvm_mpic_ops;
1157 extern struct kvm_device_ops kvm_xics_ops;
1158 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1159 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1161 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1163 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1165 vcpu->spin_loop.in_spin_loop = val;
1167 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1169 vcpu->spin_loop.dy_eligible = val;
1172 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1174 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1178 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1181 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1183 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1184 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1185 struct irq_bypass_producer *);
1186 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1187 struct irq_bypass_producer *);
1188 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1189 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1190 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1191 uint32_t guest_irq, bool set);
1192 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */