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/nospec.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(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(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(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 * vcpu->requests bit members
117 #define KVM_REQ_TLB_FLUSH 0
118 #define KVM_REQ_MIGRATE_TIMER 1
119 #define KVM_REQ_REPORT_TPR_ACCESS 2
120 #define KVM_REQ_MMU_RELOAD 3
121 #define KVM_REQ_TRIPLE_FAULT 4
122 #define KVM_REQ_PENDING_TIMER 5
123 #define KVM_REQ_UNHALT 6
124 #define KVM_REQ_MMU_SYNC 7
125 #define KVM_REQ_CLOCK_UPDATE 8
126 #define KVM_REQ_KICK 9
127 #define KVM_REQ_DEACTIVATE_FPU 10
128 #define KVM_REQ_EVENT 11
129 #define KVM_REQ_APF_HALT 12
130 #define KVM_REQ_STEAL_UPDATE 13
131 #define KVM_REQ_NMI 14
132 #define KVM_REQ_PMU 15
133 #define KVM_REQ_PMI 16
134 #define KVM_REQ_WATCHDOG 17
135 #define KVM_REQ_MASTERCLOCK_UPDATE 18
136 #define KVM_REQ_MCLOCK_INPROGRESS 19
137 #define KVM_REQ_EPR_EXIT 20
138 #define KVM_REQ_SCAN_IOAPIC 21
139 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 22
140 #define KVM_REQ_ENABLE_IBS 23
141 #define KVM_REQ_DISABLE_IBS 24
142 #define KVM_REQ_APIC_PAGE_RELOAD 25
143 #define KVM_REQ_SMI 26
144 #define KVM_REQ_HV_CRASH 27
145 #define KVM_REQ_IOAPIC_EOI_EXIT 28
146 #define KVM_REQ_HV_RESET 29
148 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
149 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
151 extern struct kmem_cache *kvm_vcpu_cache;
153 extern spinlock_t kvm_lock;
154 extern struct list_head vm_list;
156 struct kvm_io_range {
159 struct kvm_io_device *dev;
162 #define NR_IOBUS_DEVS 1000
167 struct kvm_io_range range[];
173 KVM_VIRTIO_CCW_NOTIFY_BUS,
178 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
179 int len, const void *val);
180 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
181 gpa_t addr, int len, const void *val, long cookie);
182 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
184 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
185 int len, struct kvm_io_device *dev);
186 void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
187 struct kvm_io_device *dev);
189 #ifdef CONFIG_KVM_ASYNC_PF
190 struct kvm_async_pf {
191 struct work_struct work;
192 struct list_head link;
193 struct list_head queue;
194 struct kvm_vcpu *vcpu;
195 struct mm_struct *mm;
198 struct kvm_arch_async_pf arch;
202 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
203 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
204 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
205 struct kvm_arch_async_pf *arch);
206 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
213 READING_SHADOW_PAGE_TABLES,
217 * Sometimes a large or cross-page mmio needs to be broken up into separate
218 * exits for userspace servicing.
220 struct kvm_mmio_fragment {
228 #ifdef CONFIG_PREEMPT_NOTIFIERS
229 struct preempt_notifier preempt_notifier;
235 unsigned long requests;
236 unsigned long guest_debug;
239 struct list_head blocked_vcpu_list;
245 int guest_fpu_loaded, guest_xcr0_loaded;
246 unsigned char fpu_counter;
247 wait_queue_head_t wq;
251 struct kvm_vcpu_stat stat;
252 unsigned int halt_poll_ns;
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 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
340 struct kvm_irq_routing_table {
341 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
344 * Array indexed by gsi. Each entry contains list of irq chips
345 * the gsi is connected to.
347 struct hlist_head map[0];
351 #ifndef KVM_PRIVATE_MEM_SLOTS
352 #define KVM_PRIVATE_MEM_SLOTS 0
355 #ifndef KVM_MEM_SLOTS_NUM
356 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
359 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
360 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
368 * memslots are not sorted by id anymore, please use id_to_memslot()
369 * to get the memslot by its id.
371 struct kvm_memslots {
373 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
374 /* The mapping table from slot id to the index in memslots[]. */
375 short id_to_index[KVM_MEM_SLOTS_NUM];
382 struct mutex slots_lock;
383 struct mm_struct *mm; /* userspace tied to this vm */
384 struct kvm_memslots *memslots[KVM_ADDRESS_SPACE_NUM];
385 struct srcu_struct srcu;
386 struct srcu_struct irq_srcu;
387 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
388 atomic_t online_vcpus;
389 int last_boosted_vcpu;
390 struct list_head vm_list;
392 struct kvm_io_bus *buses[KVM_NR_BUSES];
393 #ifdef CONFIG_HAVE_KVM_EVENTFD
396 struct list_head items;
397 struct list_head resampler_list;
398 struct mutex resampler_lock;
400 struct list_head ioeventfds;
402 struct kvm_vm_stat stat;
403 struct kvm_arch arch;
404 atomic_t users_count;
405 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
406 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
407 spinlock_t ring_lock;
408 struct list_head coalesced_zones;
411 struct mutex irq_lock;
412 #ifdef CONFIG_HAVE_KVM_IRQCHIP
414 * Update side is protected by irq_lock.
416 struct kvm_irq_routing_table __rcu *irq_routing;
418 #ifdef CONFIG_HAVE_KVM_IRQFD
419 struct hlist_head irq_ack_notifier_list;
422 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
423 struct mmu_notifier mmu_notifier;
424 unsigned long mmu_notifier_seq;
425 long mmu_notifier_count;
428 struct list_head devices;
431 #define kvm_err(fmt, ...) \
432 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
433 #define kvm_info(fmt, ...) \
434 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
435 #define kvm_debug(fmt, ...) \
436 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
437 #define kvm_pr_unimpl(fmt, ...) \
438 pr_err_ratelimited("kvm [%i]: " fmt, \
439 task_tgid_nr(current), ## __VA_ARGS__)
441 /* The guest did something we don't support. */
442 #define vcpu_unimpl(vcpu, fmt, ...) \
443 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
445 #define vcpu_debug(vcpu, fmt, ...) \
446 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
448 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
450 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
451 * the caller has read kvm->online_vcpus before (as is the case
452 * for kvm_for_each_vcpu, for example).
455 return kvm->vcpus[i];
458 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
460 idx < atomic_read(&kvm->online_vcpus) && \
461 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
464 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
466 struct kvm_vcpu *vcpu;
469 kvm_for_each_vcpu(i, vcpu, kvm)
470 if (vcpu->vcpu_id == id)
475 #define kvm_for_each_memslot(memslot, slots) \
476 for (memslot = &slots->memslots[0]; \
477 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
480 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
481 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
483 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
484 void vcpu_put(struct kvm_vcpu *vcpu);
486 #ifdef __KVM_HAVE_IOAPIC
487 void kvm_vcpu_request_scan_ioapic(struct kvm *kvm);
488 void kvm_arch_irq_routing_update(struct kvm *kvm);
490 static inline void kvm_vcpu_request_scan_ioapic(struct kvm *kvm)
493 static inline void kvm_arch_irq_routing_update(struct kvm *kvm)
498 #ifdef CONFIG_HAVE_KVM_IRQFD
499 int kvm_irqfd_init(void);
500 void kvm_irqfd_exit(void);
502 static inline int kvm_irqfd_init(void)
507 static inline void kvm_irqfd_exit(void)
511 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
512 struct module *module);
515 void kvm_get_kvm(struct kvm *kvm);
516 void kvm_put_kvm(struct kvm *kvm);
518 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
520 return rcu_dereference_check(kvm->memslots[as_id],
521 srcu_read_lock_held(&kvm->srcu)
522 || lockdep_is_held(&kvm->slots_lock));
525 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
527 return __kvm_memslots(kvm, 0);
530 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
532 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
534 return __kvm_memslots(vcpu->kvm, as_id);
537 static inline struct kvm_memory_slot *
538 id_to_memslot(struct kvm_memslots *slots, int id)
540 int index = slots->id_to_index[id];
541 struct kvm_memory_slot *slot;
543 slot = &slots->memslots[index];
545 WARN_ON(slot->id != id);
550 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
551 * - create a new memory slot
552 * - delete an existing memory slot
553 * - modify an existing memory slot
554 * -- move it in the guest physical memory space
555 * -- just change its flags
557 * Since flags can be changed by some of these operations, the following
558 * differentiation is the best we can do for __kvm_set_memory_region():
567 int kvm_set_memory_region(struct kvm *kvm,
568 const struct kvm_userspace_memory_region *mem);
569 int __kvm_set_memory_region(struct kvm *kvm,
570 const struct kvm_userspace_memory_region *mem);
571 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
572 struct kvm_memory_slot *dont);
573 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
574 unsigned long npages);
575 void kvm_arch_memslots_updated(struct kvm *kvm, struct kvm_memslots *slots);
576 int kvm_arch_prepare_memory_region(struct kvm *kvm,
577 struct kvm_memory_slot *memslot,
578 const struct kvm_userspace_memory_region *mem,
579 enum kvm_mr_change change);
580 void kvm_arch_commit_memory_region(struct kvm *kvm,
581 const struct kvm_userspace_memory_region *mem,
582 const struct kvm_memory_slot *old,
583 const struct kvm_memory_slot *new,
584 enum kvm_mr_change change);
585 bool kvm_largepages_enabled(void);
586 void kvm_disable_largepages(void);
587 /* flush all memory translations */
588 void kvm_arch_flush_shadow_all(struct kvm *kvm);
589 /* flush memory translations pointing to 'slot' */
590 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
591 struct kvm_memory_slot *slot);
593 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
594 struct page **pages, int nr_pages);
596 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
597 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
598 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
599 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
600 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
602 void kvm_release_page_clean(struct page *page);
603 void kvm_release_page_dirty(struct page *page);
604 void kvm_set_page_accessed(struct page *page);
606 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
607 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
608 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
610 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
611 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
612 pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn, bool atomic,
613 bool *async, bool write_fault, bool *writable);
615 void kvm_release_pfn_clean(pfn_t pfn);
616 void kvm_set_pfn_dirty(pfn_t pfn);
617 void kvm_set_pfn_accessed(pfn_t pfn);
618 void kvm_get_pfn(pfn_t pfn);
620 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
622 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
624 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
625 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
626 void *data, unsigned long len);
627 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
628 int offset, int len);
629 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
631 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
632 void *data, unsigned long len);
633 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
634 gpa_t gpa, unsigned long len);
635 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
636 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
637 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
638 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
639 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
640 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
642 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
643 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
644 pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
645 pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
646 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
647 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
648 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
649 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
651 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
653 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
655 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
656 int offset, int len);
657 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
659 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
661 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
662 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
663 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
664 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
665 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
666 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
667 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
668 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
670 void kvm_flush_remote_tlbs(struct kvm *kvm);
671 void kvm_reload_remote_mmus(struct kvm *kvm);
672 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
673 void kvm_make_scan_ioapic_request(struct kvm *kvm);
674 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
676 long kvm_arch_dev_ioctl(struct file *filp,
677 unsigned int ioctl, unsigned long arg);
678 long kvm_arch_vcpu_ioctl(struct file *filp,
679 unsigned int ioctl, unsigned long arg);
680 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
682 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
684 int kvm_get_dirty_log(struct kvm *kvm,
685 struct kvm_dirty_log *log, int *is_dirty);
687 int kvm_get_dirty_log_protect(struct kvm *kvm,
688 struct kvm_dirty_log *log, bool *is_dirty);
690 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
691 struct kvm_memory_slot *slot,
695 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
696 struct kvm_dirty_log *log);
698 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
700 long kvm_arch_vm_ioctl(struct file *filp,
701 unsigned int ioctl, unsigned long arg);
703 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
704 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
706 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
707 struct kvm_translation *tr);
709 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
710 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
711 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
712 struct kvm_sregs *sregs);
713 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
714 struct kvm_sregs *sregs);
715 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
716 struct kvm_mp_state *mp_state);
717 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
718 struct kvm_mp_state *mp_state);
719 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
720 struct kvm_guest_debug *dbg);
721 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
723 int kvm_arch_init(void *opaque);
724 void kvm_arch_exit(void);
726 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
727 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
729 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
731 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
732 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
733 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
734 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
735 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
736 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
737 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
739 int kvm_arch_hardware_enable(void);
740 void kvm_arch_hardware_disable(void);
741 int kvm_arch_hardware_setup(void);
742 void kvm_arch_hardware_unsetup(void);
743 void kvm_arch_check_processor_compat(void *rtn);
744 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
745 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
747 void *kvm_kvzalloc(unsigned long size);
749 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
750 static inline struct kvm *kvm_arch_alloc_vm(void)
752 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
755 static inline void kvm_arch_free_vm(struct kvm *kvm)
761 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
762 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
763 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
764 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
766 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
770 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
774 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
779 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
780 void kvm_arch_start_assignment(struct kvm *kvm);
781 void kvm_arch_end_assignment(struct kvm *kvm);
782 bool kvm_arch_has_assigned_device(struct kvm *kvm);
784 static inline void kvm_arch_start_assignment(struct kvm *kvm)
788 static inline void kvm_arch_end_assignment(struct kvm *kvm)
792 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
798 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
800 #ifdef __KVM_HAVE_ARCH_WQP
801 return vcpu->arch.wqp;
807 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
809 * returns true if the virtual interrupt controller is initialized and
810 * ready to accept virtual IRQ. On some architectures the virtual interrupt
811 * controller is dynamically instantiated and this is not always true.
813 bool kvm_arch_intc_initialized(struct kvm *kvm);
815 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
821 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
822 void kvm_arch_destroy_vm(struct kvm *kvm);
823 void kvm_arch_sync_events(struct kvm *kvm);
825 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
826 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
828 bool kvm_is_reserved_pfn(pfn_t pfn);
830 struct kvm_irq_ack_notifier {
831 struct hlist_node link;
833 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
836 int kvm_irq_map_gsi(struct kvm *kvm,
837 struct kvm_kernel_irq_routing_entry *entries, int gsi);
838 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
840 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
842 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
843 int irq_source_id, int level, bool line_status);
844 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
845 struct kvm *kvm, int irq_source_id,
846 int level, bool line_status);
847 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
848 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
849 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
850 void kvm_register_irq_ack_notifier(struct kvm *kvm,
851 struct kvm_irq_ack_notifier *kian);
852 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
853 struct kvm_irq_ack_notifier *kian);
854 int kvm_request_irq_source_id(struct kvm *kvm);
855 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
857 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
858 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
859 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
861 static inline int kvm_iommu_map_pages(struct kvm *kvm,
862 struct kvm_memory_slot *slot)
867 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
868 struct kvm_memory_slot *slot)
873 /* must be called with irqs disabled */
874 static inline void __kvm_guest_enter(void)
877 /* KVM does not hold any references to rcu protected data when it
878 * switches CPU into a guest mode. In fact switching to a guest mode
879 * is very similar to exiting to userspace from rcu point of view. In
880 * addition CPU may stay in a guest mode for quite a long time (up to
881 * one time slice). Lets treat guest mode as quiescent state, just like
882 * we do with user-mode execution.
884 if (!context_tracking_cpu_is_enabled())
885 rcu_virt_note_context_switch(smp_processor_id());
888 /* must be called with irqs disabled */
889 static inline void __kvm_guest_exit(void)
894 static inline void kvm_guest_enter(void)
898 local_irq_save(flags);
900 local_irq_restore(flags);
903 static inline void kvm_guest_exit(void)
907 local_irq_save(flags);
909 local_irq_restore(flags);
913 * search_memslots() and __gfn_to_memslot() are here because they are
914 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
915 * gfn_to_memslot() itself isn't here as an inline because that would
916 * bloat other code too much.
918 static inline struct kvm_memory_slot *
919 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
921 int start = 0, end = slots->used_slots;
922 int slot = atomic_read(&slots->lru_slot);
923 struct kvm_memory_slot *memslots = slots->memslots;
925 if (gfn >= memslots[slot].base_gfn &&
926 gfn < memslots[slot].base_gfn + memslots[slot].npages)
927 return &memslots[slot];
929 while (start < end) {
930 slot = start + (end - start) / 2;
932 if (gfn >= memslots[slot].base_gfn)
938 if (start < slots->used_slots && gfn >= memslots[start].base_gfn &&
939 gfn < memslots[start].base_gfn + memslots[start].npages) {
940 atomic_set(&slots->lru_slot, start);
941 return &memslots[start];
947 static inline struct kvm_memory_slot *
948 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
950 return search_memslots(slots, gfn);
953 static inline unsigned long
954 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
957 * The index was checked originally in search_memslots. To avoid
958 * that a malicious guest builds a Spectre gadget out of e.g. page
959 * table walks, do not let the processor speculate loads outside
960 * the guest's registered memslots.
962 unsigned long offset = gfn - slot->base_gfn;
963 offset = array_index_nospec(offset, slot->npages);
964 return slot->userspace_addr + offset * PAGE_SIZE;
967 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
969 return gfn_to_memslot(kvm, gfn)->id;
973 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
975 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
977 return slot->base_gfn + gfn_offset;
980 static inline gpa_t gfn_to_gpa(gfn_t gfn)
982 return (gpa_t)gfn << PAGE_SHIFT;
985 static inline gfn_t gpa_to_gfn(gpa_t gpa)
987 return (gfn_t)(gpa >> PAGE_SHIFT);
990 static inline hpa_t pfn_to_hpa(pfn_t pfn)
992 return (hpa_t)pfn << PAGE_SHIFT;
995 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
997 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
999 return kvm_is_error_hva(hva);
1002 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
1004 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
1007 enum kvm_stat_kind {
1012 struct kvm_stats_debugfs_item {
1015 enum kvm_stat_kind kind;
1016 struct dentry *dentry;
1018 extern struct kvm_stats_debugfs_item debugfs_entries[];
1019 extern struct dentry *kvm_debugfs_dir;
1021 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1022 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1024 if (unlikely(kvm->mmu_notifier_count))
1027 * Ensure the read of mmu_notifier_count happens before the read
1028 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1029 * mmu_notifier_invalidate_range_end to make sure that the caller
1030 * either sees the old (non-zero) value of mmu_notifier_count or
1031 * the new (incremented) value of mmu_notifier_seq.
1032 * PowerPC Book3s HV KVM calls this under a per-page lock
1033 * rather than under kvm->mmu_lock, for scalability, so
1034 * can't rely on kvm->mmu_lock to keep things ordered.
1037 if (kvm->mmu_notifier_seq != mmu_seq)
1043 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1046 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1048 #define KVM_MAX_IRQ_ROUTES 1024
1051 int kvm_setup_default_irq_routing(struct kvm *kvm);
1052 int kvm_setup_empty_irq_routing(struct kvm *kvm);
1053 int kvm_set_irq_routing(struct kvm *kvm,
1054 const struct kvm_irq_routing_entry *entries,
1057 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
1058 const struct kvm_irq_routing_entry *ue);
1059 void kvm_free_irq_routing(struct kvm *kvm);
1063 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1067 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1069 #ifdef CONFIG_HAVE_KVM_EVENTFD
1071 void kvm_eventfd_init(struct kvm *kvm);
1072 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1074 #ifdef CONFIG_HAVE_KVM_IRQFD
1075 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1076 void kvm_irqfd_release(struct kvm *kvm);
1077 void kvm_irq_routing_update(struct kvm *);
1079 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1084 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1089 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1091 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1096 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1098 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1099 static inline void kvm_irq_routing_update(struct kvm *kvm)
1104 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1109 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1111 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
1112 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
1114 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
1117 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1119 set_bit(req, &vcpu->requests);
1122 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1124 if (test_bit(req, &vcpu->requests)) {
1125 clear_bit(req, &vcpu->requests);
1132 extern bool kvm_rebooting;
1135 struct kvm_device_ops *ops;
1138 struct list_head vm_node;
1141 /* create, destroy, and name are mandatory */
1142 struct kvm_device_ops {
1144 int (*create)(struct kvm_device *dev, u32 type);
1147 * Destroy is responsible for freeing dev.
1149 * Destroy may be called before or after destructors are called
1150 * on emulated I/O regions, depending on whether a reference is
1151 * held by a vcpu or other kvm component that gets destroyed
1152 * after the emulated I/O.
1154 void (*destroy)(struct kvm_device *dev);
1156 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1157 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1158 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1159 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1163 void kvm_device_get(struct kvm_device *dev);
1164 void kvm_device_put(struct kvm_device *dev);
1165 struct kvm_device *kvm_device_from_filp(struct file *filp);
1166 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1167 void kvm_unregister_device_ops(u32 type);
1169 extern struct kvm_device_ops kvm_mpic_ops;
1170 extern struct kvm_device_ops kvm_xics_ops;
1171 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1172 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1174 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1176 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1178 vcpu->spin_loop.in_spin_loop = val;
1180 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1182 vcpu->spin_loop.dy_eligible = val;
1185 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1187 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1191 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1194 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1196 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1197 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1198 struct irq_bypass_producer *);
1199 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1200 struct irq_bypass_producer *);
1201 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1202 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1203 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1204 uint32_t guest_irq, bool set);
1205 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */