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)
47 /* Two fragments for cross MMIO pages. */
48 #define KVM_MAX_MMIO_FRAGMENTS 2
51 * For the normal pfn, the highest 12 bits should be zero,
52 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
53 * mask bit 63 to indicate the noslot pfn.
55 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
56 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
57 #define KVM_PFN_NOSLOT (0x1ULL << 63)
59 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
60 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
61 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
64 * error pfns indicate that the gfn is in slot but faild to
65 * translate it to pfn on host.
67 static inline bool is_error_pfn(pfn_t pfn)
69 return !!(pfn & KVM_PFN_ERR_MASK);
73 * error_noslot pfns indicate that the gfn can not be
74 * translated to pfn - it is not in slot or failed to
75 * translate it to pfn.
77 static inline bool is_error_noslot_pfn(pfn_t pfn)
79 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
82 /* noslot pfn indicates that the gfn is not in slot. */
83 static inline bool is_noslot_pfn(pfn_t pfn)
85 return pfn == KVM_PFN_NOSLOT;
89 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
90 * provide own defines and kvm_is_error_hva
92 #ifndef KVM_HVA_ERR_BAD
94 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
95 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
97 static inline bool kvm_is_error_hva(unsigned long addr)
99 return addr >= PAGE_OFFSET;
104 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
106 static inline bool is_error_page(struct page *page)
112 * vcpu->requests bit members
114 #define KVM_REQ_TLB_FLUSH 0
115 #define KVM_REQ_MIGRATE_TIMER 1
116 #define KVM_REQ_REPORT_TPR_ACCESS 2
117 #define KVM_REQ_MMU_RELOAD 3
118 #define KVM_REQ_TRIPLE_FAULT 4
119 #define KVM_REQ_PENDING_TIMER 5
120 #define KVM_REQ_UNHALT 6
121 #define KVM_REQ_MMU_SYNC 7
122 #define KVM_REQ_CLOCK_UPDATE 8
123 #define KVM_REQ_KICK 9
124 #define KVM_REQ_DEACTIVATE_FPU 10
125 #define KVM_REQ_EVENT 11
126 #define KVM_REQ_APF_HALT 12
127 #define KVM_REQ_STEAL_UPDATE 13
128 #define KVM_REQ_NMI 14
129 #define KVM_REQ_PMU 15
130 #define KVM_REQ_PMI 16
131 #define KVM_REQ_WATCHDOG 17
132 #define KVM_REQ_MASTERCLOCK_UPDATE 18
133 #define KVM_REQ_MCLOCK_INPROGRESS 19
134 #define KVM_REQ_EPR_EXIT 20
135 #define KVM_REQ_SCAN_IOAPIC 21
136 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 22
138 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
139 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
143 extern struct kmem_cache *kvm_vcpu_cache;
145 extern spinlock_t kvm_lock;
146 extern struct list_head vm_list;
148 struct kvm_io_range {
151 struct kvm_io_device *dev;
154 #define NR_IOBUS_DEVS 1000
159 struct kvm_io_range range[];
165 KVM_VIRTIO_CCW_NOTIFY_BUS,
169 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
170 int len, const void *val);
171 int kvm_io_bus_write_cookie(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
172 int len, const void *val, long cookie);
173 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
175 int kvm_io_bus_read_cookie(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
176 int len, void *val, long cookie);
177 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
178 int len, struct kvm_io_device *dev);
179 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
180 struct kvm_io_device *dev);
182 #ifdef CONFIG_KVM_ASYNC_PF
183 struct kvm_async_pf {
184 struct work_struct work;
185 struct list_head link;
186 struct list_head queue;
187 struct kvm_vcpu *vcpu;
188 struct mm_struct *mm;
191 struct kvm_arch_async_pf arch;
195 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
196 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
197 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
198 struct kvm_arch_async_pf *arch);
199 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
206 READING_SHADOW_PAGE_TABLES,
210 * Sometimes a large or cross-page mmio needs to be broken up into separate
211 * exits for userspace servicing.
213 struct kvm_mmio_fragment {
221 #ifdef CONFIG_PREEMPT_NOTIFIERS
222 struct preempt_notifier preempt_notifier;
228 unsigned long requests;
229 unsigned long guest_debug;
235 int guest_fpu_loaded, guest_xcr0_loaded;
236 wait_queue_head_t wq;
240 struct kvm_vcpu_stat stat;
242 #ifdef CONFIG_HAS_IOMEM
244 int mmio_read_completed;
246 int mmio_cur_fragment;
247 int mmio_nr_fragments;
248 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
251 #ifdef CONFIG_KVM_ASYNC_PF
254 struct list_head queue;
255 struct list_head done;
260 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
262 * Cpu relax intercept or pause loop exit optimization
263 * in_spin_loop: set when a vcpu does a pause loop exit
264 * or cpu relax intercepted.
265 * dy_eligible: indicates whether vcpu is eligible for directed yield.
273 struct kvm_vcpu_arch arch;
276 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
278 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
282 * Some of the bitops functions do not support too long bitmaps.
283 * This number must be determined not to exceed such limits.
285 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
287 struct kvm_memory_slot {
289 unsigned long npages;
290 unsigned long *dirty_bitmap;
291 struct kvm_arch_memory_slot arch;
292 unsigned long userspace_addr;
297 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
299 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
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,
315 struct hlist_node link;
318 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
320 struct kvm_irq_routing_table {
321 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
322 struct kvm_kernel_irq_routing_entry *rt_entries;
325 * Array indexed by gsi. Each entry contains list of irq chips
326 * the gsi is connected to.
328 struct hlist_head map[0];
333 struct kvm_irq_routing_table {};
337 #ifndef KVM_PRIVATE_MEM_SLOTS
338 #define KVM_PRIVATE_MEM_SLOTS 0
341 #ifndef KVM_MEM_SLOTS_NUM
342 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
347 * memslots are not sorted by id anymore, please use id_to_memslot()
348 * to get the memslot by its id.
350 struct kvm_memslots {
352 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
353 /* The mapping table from slot id to the index in memslots[]. */
354 short id_to_index[KVM_MEM_SLOTS_NUM];
359 struct mutex slots_lock;
360 struct mm_struct *mm; /* userspace tied to this vm */
361 struct kvm_memslots *memslots;
362 struct srcu_struct srcu;
363 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
366 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
367 atomic_t online_vcpus;
368 int last_boosted_vcpu;
369 struct list_head vm_list;
371 struct kvm_io_bus *buses[KVM_NR_BUSES];
372 #ifdef CONFIG_HAVE_KVM_EVENTFD
375 struct list_head items;
376 struct list_head resampler_list;
377 struct mutex resampler_lock;
379 struct list_head ioeventfds;
381 struct kvm_vm_stat stat;
382 struct kvm_arch arch;
383 atomic_t users_count;
384 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
385 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
386 spinlock_t ring_lock;
387 struct list_head coalesced_zones;
390 struct mutex irq_lock;
391 #ifdef CONFIG_HAVE_KVM_IRQCHIP
393 * Update side is protected by irq_lock and,
394 * if configured, irqfds.lock.
396 struct kvm_irq_routing_table __rcu *irq_routing;
397 struct hlist_head mask_notifier_list;
398 struct hlist_head irq_ack_notifier_list;
401 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
402 struct mmu_notifier mmu_notifier;
403 unsigned long mmu_notifier_seq;
404 long mmu_notifier_count;
407 struct list_head devices;
410 #define kvm_err(fmt, ...) \
411 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
412 #define kvm_info(fmt, ...) \
413 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
414 #define kvm_debug(fmt, ...) \
415 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
416 #define kvm_pr_unimpl(fmt, ...) \
417 pr_err_ratelimited("kvm [%i]: " fmt, \
418 task_tgid_nr(current), ## __VA_ARGS__)
420 /* The guest did something we don't support. */
421 #define vcpu_unimpl(vcpu, fmt, ...) \
422 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
424 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
427 return kvm->vcpus[i];
430 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
432 idx < atomic_read(&kvm->online_vcpus) && \
433 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
436 #define kvm_for_each_memslot(memslot, slots) \
437 for (memslot = &slots->memslots[0]; \
438 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
441 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
442 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
444 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
445 void vcpu_put(struct kvm_vcpu *vcpu);
447 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
448 int kvm_irqfd_init(void);
449 void kvm_irqfd_exit(void);
451 static inline int kvm_irqfd_init(void)
456 static inline void kvm_irqfd_exit(void)
460 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
461 struct module *module);
464 void kvm_get_kvm(struct kvm *kvm);
465 void kvm_put_kvm(struct kvm *kvm);
466 void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new,
467 u64 last_generation);
469 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
471 return rcu_dereference_check(kvm->memslots,
472 srcu_read_lock_held(&kvm->srcu)
473 || lockdep_is_held(&kvm->slots_lock));
476 static inline struct kvm_memory_slot *
477 id_to_memslot(struct kvm_memslots *slots, int id)
479 int index = slots->id_to_index[id];
480 struct kvm_memory_slot *slot;
482 slot = &slots->memslots[index];
484 WARN_ON(slot->id != id);
489 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
490 * - create a new memory slot
491 * - delete an existing memory slot
492 * - modify an existing memory slot
493 * -- move it in the guest physical memory space
494 * -- just change its flags
496 * Since flags can be changed by some of these operations, the following
497 * differentiation is the best we can do for __kvm_set_memory_region():
506 int kvm_set_memory_region(struct kvm *kvm,
507 struct kvm_userspace_memory_region *mem);
508 int __kvm_set_memory_region(struct kvm *kvm,
509 struct kvm_userspace_memory_region *mem);
510 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
511 struct kvm_memory_slot *dont);
512 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
513 unsigned long npages);
514 void kvm_arch_memslots_updated(struct kvm *kvm);
515 int kvm_arch_prepare_memory_region(struct kvm *kvm,
516 struct kvm_memory_slot *memslot,
517 struct kvm_userspace_memory_region *mem,
518 enum kvm_mr_change change);
519 void kvm_arch_commit_memory_region(struct kvm *kvm,
520 struct kvm_userspace_memory_region *mem,
521 const struct kvm_memory_slot *old,
522 enum kvm_mr_change change);
523 bool kvm_largepages_enabled(void);
524 void kvm_disable_largepages(void);
525 /* flush all memory translations */
526 void kvm_arch_flush_shadow_all(struct kvm *kvm);
527 /* flush memory translations pointing to 'slot' */
528 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
529 struct kvm_memory_slot *slot);
531 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
534 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
535 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
536 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
537 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
538 void kvm_release_page_clean(struct page *page);
539 void kvm_release_page_dirty(struct page *page);
540 void kvm_set_page_dirty(struct page *page);
541 void kvm_set_page_accessed(struct page *page);
543 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
544 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
545 bool write_fault, bool *writable);
546 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
547 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
549 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
550 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
552 void kvm_release_pfn_dirty(pfn_t pfn);
553 void kvm_release_pfn_clean(pfn_t pfn);
554 void kvm_set_pfn_dirty(pfn_t pfn);
555 void kvm_set_pfn_accessed(pfn_t pfn);
556 void kvm_get_pfn(pfn_t pfn);
558 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
560 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
562 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
563 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
564 void *data, unsigned long len);
565 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
566 int offset, int len);
567 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
569 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
570 void *data, unsigned long len);
571 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
572 gpa_t gpa, unsigned long len);
573 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
574 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
575 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
576 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
577 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
578 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
579 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
582 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
583 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
584 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
585 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
586 void kvm_resched(struct kvm_vcpu *vcpu);
587 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
588 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
590 void kvm_flush_remote_tlbs(struct kvm *kvm);
591 void kvm_reload_remote_mmus(struct kvm *kvm);
592 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
593 void kvm_make_scan_ioapic_request(struct kvm *kvm);
595 long kvm_arch_dev_ioctl(struct file *filp,
596 unsigned int ioctl, unsigned long arg);
597 long kvm_arch_vcpu_ioctl(struct file *filp,
598 unsigned int ioctl, unsigned long arg);
599 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
601 int kvm_dev_ioctl_check_extension(long ext);
603 int kvm_get_dirty_log(struct kvm *kvm,
604 struct kvm_dirty_log *log, int *is_dirty);
605 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
606 struct kvm_dirty_log *log);
608 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
609 struct kvm_userspace_memory_region *mem);
610 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
612 long kvm_arch_vm_ioctl(struct file *filp,
613 unsigned int ioctl, unsigned long arg);
615 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
616 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
618 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
619 struct kvm_translation *tr);
621 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
622 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
623 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
624 struct kvm_sregs *sregs);
625 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
626 struct kvm_sregs *sregs);
627 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
628 struct kvm_mp_state *mp_state);
629 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
630 struct kvm_mp_state *mp_state);
631 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
632 struct kvm_guest_debug *dbg);
633 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
635 int kvm_arch_init(void *opaque);
636 void kvm_arch_exit(void);
638 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
639 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
641 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
642 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
643 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
644 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
645 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
646 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
647 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
649 int kvm_arch_hardware_enable(void *garbage);
650 void kvm_arch_hardware_disable(void *garbage);
651 int kvm_arch_hardware_setup(void);
652 void kvm_arch_hardware_unsetup(void);
653 void kvm_arch_check_processor_compat(void *rtn);
654 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
655 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
657 void kvm_free_physmem(struct kvm *kvm);
659 void *kvm_kvzalloc(unsigned long size);
660 void kvm_kvfree(const void *addr);
662 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
663 static inline struct kvm *kvm_arch_alloc_vm(void)
665 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
668 static inline void kvm_arch_free_vm(struct kvm *kvm)
674 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
675 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
676 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
677 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
679 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
683 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
687 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
693 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
695 #ifdef __KVM_HAVE_ARCH_WQP
696 return vcpu->arch.wqp;
702 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
703 void kvm_arch_destroy_vm(struct kvm *kvm);
704 void kvm_arch_sync_events(struct kvm *kvm);
706 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
707 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
709 bool kvm_is_mmio_pfn(pfn_t pfn);
711 struct kvm_irq_ack_notifier {
712 struct hlist_node link;
714 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
717 struct kvm_assigned_dev_kernel {
718 struct kvm_irq_ack_notifier ack_notifier;
719 struct list_head list;
724 unsigned int entries_nr;
726 bool host_irq_disabled;
728 struct msix_entry *host_msix_entries;
730 struct msix_entry *guest_msix_entries;
731 unsigned long irq_requested_type;
736 spinlock_t intx_lock;
737 spinlock_t intx_mask_lock;
739 struct pci_saved_state *pci_saved_state;
742 struct kvm_irq_mask_notifier {
743 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
745 struct hlist_node link;
748 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
749 struct kvm_irq_mask_notifier *kimn);
750 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
751 struct kvm_irq_mask_notifier *kimn);
752 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
755 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
757 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
758 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
759 int irq_source_id, int level, bool line_status);
760 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
761 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
762 void kvm_register_irq_ack_notifier(struct kvm *kvm,
763 struct kvm_irq_ack_notifier *kian);
764 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
765 struct kvm_irq_ack_notifier *kian);
766 int kvm_request_irq_source_id(struct kvm *kvm);
767 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
769 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
770 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
771 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
772 int kvm_iommu_map_guest(struct kvm *kvm);
773 int kvm_iommu_unmap_guest(struct kvm *kvm);
774 int kvm_assign_device(struct kvm *kvm,
775 struct kvm_assigned_dev_kernel *assigned_dev);
776 int kvm_deassign_device(struct kvm *kvm,
777 struct kvm_assigned_dev_kernel *assigned_dev);
779 static inline int kvm_iommu_map_pages(struct kvm *kvm,
780 struct kvm_memory_slot *slot)
785 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
786 struct kvm_memory_slot *slot)
790 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
796 static inline void kvm_guest_enter(void)
800 BUG_ON(preemptible());
802 local_irq_save(flags);
804 local_irq_restore(flags);
806 /* KVM does not hold any references to rcu protected data when it
807 * switches CPU into a guest mode. In fact switching to a guest mode
808 * is very similar to exiting to userspace from rcu point of view. In
809 * addition CPU may stay in a guest mode for quite a long time (up to
810 * one time slice). Lets treat guest mode as quiescent state, just like
811 * we do with user-mode execution.
813 rcu_virt_note_context_switch(smp_processor_id());
816 static inline void kvm_guest_exit(void)
820 local_irq_save(flags);
822 local_irq_restore(flags);
826 * search_memslots() and __gfn_to_memslot() are here because they are
827 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
828 * gfn_to_memslot() itself isn't here as an inline because that would
829 * bloat other code too much.
831 static inline struct kvm_memory_slot *
832 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
834 struct kvm_memory_slot *memslot;
836 kvm_for_each_memslot(memslot, slots)
837 if (gfn >= memslot->base_gfn &&
838 gfn < memslot->base_gfn + memslot->npages)
844 static inline struct kvm_memory_slot *
845 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
847 return search_memslots(slots, gfn);
850 static inline unsigned long
851 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
853 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
856 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
858 return gfn_to_memslot(kvm, gfn)->id;
862 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
864 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
866 return slot->base_gfn + gfn_offset;
869 static inline gpa_t gfn_to_gpa(gfn_t gfn)
871 return (gpa_t)gfn << PAGE_SHIFT;
874 static inline gfn_t gpa_to_gfn(gpa_t gpa)
876 return (gfn_t)(gpa >> PAGE_SHIFT);
879 static inline hpa_t pfn_to_hpa(pfn_t pfn)
881 return (hpa_t)pfn << PAGE_SHIFT;
884 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
886 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
894 struct kvm_stats_debugfs_item {
897 enum kvm_stat_kind kind;
898 struct dentry *dentry;
900 extern struct kvm_stats_debugfs_item debugfs_entries[];
901 extern struct dentry *kvm_debugfs_dir;
903 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
904 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
906 if (unlikely(kvm->mmu_notifier_count))
909 * Ensure the read of mmu_notifier_count happens before the read
910 * of mmu_notifier_seq. This interacts with the smp_wmb() in
911 * mmu_notifier_invalidate_range_end to make sure that the caller
912 * either sees the old (non-zero) value of mmu_notifier_count or
913 * the new (incremented) value of mmu_notifier_seq.
914 * PowerPC Book3s HV KVM calls this under a per-page lock
915 * rather than under kvm->mmu_lock, for scalability, so
916 * can't rely on kvm->mmu_lock to keep things ordered.
919 if (kvm->mmu_notifier_seq != mmu_seq)
925 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
927 #define KVM_MAX_IRQ_ROUTES 1024
929 int kvm_setup_default_irq_routing(struct kvm *kvm);
930 int kvm_set_irq_routing(struct kvm *kvm,
931 const struct kvm_irq_routing_entry *entries,
934 int kvm_set_routing_entry(struct kvm_irq_routing_table *rt,
935 struct kvm_kernel_irq_routing_entry *e,
936 const struct kvm_irq_routing_entry *ue);
937 void kvm_free_irq_routing(struct kvm *kvm);
939 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
943 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
947 #ifdef CONFIG_HAVE_KVM_EVENTFD
949 void kvm_eventfd_init(struct kvm *kvm);
950 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
952 #ifdef CONFIG_HAVE_KVM_IRQCHIP
953 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
954 void kvm_irqfd_release(struct kvm *kvm);
955 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
957 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
962 static inline void kvm_irqfd_release(struct kvm *kvm) {}
967 static inline void kvm_eventfd_init(struct kvm *kvm) {}
969 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
974 static inline void kvm_irqfd_release(struct kvm *kvm) {}
976 #ifdef CONFIG_HAVE_KVM_IRQCHIP
977 static inline void kvm_irq_routing_update(struct kvm *kvm,
978 struct kvm_irq_routing_table *irq_rt)
980 rcu_assign_pointer(kvm->irq_routing, irq_rt);
984 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
989 #endif /* CONFIG_HAVE_KVM_EVENTFD */
991 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
992 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
994 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
997 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
1001 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
1005 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
1007 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
1010 void kvm_free_all_assigned_devices(struct kvm *kvm);
1014 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
1020 static inline void kvm_free_all_assigned_devices(struct kvm *kvm) {}
1024 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1026 set_bit(req, &vcpu->requests);
1029 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1031 if (test_bit(req, &vcpu->requests)) {
1032 clear_bit(req, &vcpu->requests);
1039 extern bool kvm_rebooting;
1041 struct kvm_device_ops;
1044 struct kvm_device_ops *ops;
1047 struct list_head vm_node;
1050 /* create, destroy, and name are mandatory */
1051 struct kvm_device_ops {
1053 int (*create)(struct kvm_device *dev, u32 type);
1056 * Destroy is responsible for freeing dev.
1058 * Destroy may be called before or after destructors are called
1059 * on emulated I/O regions, depending on whether a reference is
1060 * held by a vcpu or other kvm component that gets destroyed
1061 * after the emulated I/O.
1063 void (*destroy)(struct kvm_device *dev);
1065 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1066 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1067 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1068 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1072 void kvm_device_get(struct kvm_device *dev);
1073 void kvm_device_put(struct kvm_device *dev);
1074 struct kvm_device *kvm_device_from_filp(struct file *filp);
1076 extern struct kvm_device_ops kvm_mpic_ops;
1077 extern struct kvm_device_ops kvm_xics_ops;
1078 extern struct kvm_device_ops kvm_vfio_ops;
1080 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1082 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1084 vcpu->spin_loop.in_spin_loop = val;
1086 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1088 vcpu->spin_loop.dy_eligible = val;
1091 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1093 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1097 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1101 static inline bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
1106 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */