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
137 #define KVM_REQ_ENABLE_IBS 23
138 #define KVM_REQ_DISABLE_IBS 24
139 #define KVM_REQ_APIC_PAGE_RELOAD 25
141 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
142 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
144 extern struct kmem_cache *kvm_vcpu_cache;
146 extern spinlock_t kvm_lock;
147 extern struct list_head vm_list;
149 struct kvm_io_range {
152 struct kvm_io_device *dev;
155 #define NR_IOBUS_DEVS 1000
160 struct kvm_io_range range[];
166 KVM_VIRTIO_CCW_NOTIFY_BUS,
171 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
172 int len, const void *val);
173 int kvm_io_bus_write_cookie(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
174 int len, const void *val, long cookie);
175 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
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, unsigned long hva,
198 struct kvm_arch_async_pf *arch);
199 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
203 * Carry out a gup that requires IO. Allow the mm to relinquish the mmap
204 * semaphore if the filemap/swap has to wait on a page lock. pagep == NULL
205 * controls whether we retry the gup one more time to completion in that case.
206 * Typically this is called after a FAULT_FLAG_RETRY_NOWAIT in the main tdp
209 int kvm_get_user_page_io(struct task_struct *tsk, struct mm_struct *mm,
210 unsigned long addr, bool write_fault,
211 struct page **pagep);
217 READING_SHADOW_PAGE_TABLES,
221 * Sometimes a large or cross-page mmio needs to be broken up into separate
222 * exits for userspace servicing.
224 struct kvm_mmio_fragment {
232 #ifdef CONFIG_PREEMPT_NOTIFIERS
233 struct preempt_notifier preempt_notifier;
239 unsigned long requests;
240 unsigned long guest_debug;
246 int guest_fpu_loaded, guest_xcr0_loaded;
247 wait_queue_head_t wq;
251 struct kvm_vcpu_stat stat;
253 #ifdef CONFIG_HAS_IOMEM
255 int mmio_read_completed;
257 int mmio_cur_fragment;
258 int mmio_nr_fragments;
259 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
262 #ifdef CONFIG_KVM_ASYNC_PF
265 struct list_head queue;
266 struct list_head done;
271 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
273 * Cpu relax intercept or pause loop exit optimization
274 * in_spin_loop: set when a vcpu does a pause loop exit
275 * or cpu relax intercepted.
276 * dy_eligible: indicates whether vcpu is eligible for directed yield.
284 struct kvm_vcpu_arch arch;
287 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
289 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
293 * Some of the bitops functions do not support too long bitmaps.
294 * This number must be determined not to exceed such limits.
296 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
298 struct kvm_memory_slot {
300 unsigned long npages;
301 unsigned long *dirty_bitmap;
302 struct kvm_arch_memory_slot arch;
303 unsigned long userspace_addr;
308 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
310 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
313 struct kvm_s390_adapter_int {
321 struct kvm_kernel_irq_routing_entry {
324 int (*set)(struct kvm_kernel_irq_routing_entry *e,
325 struct kvm *kvm, int irq_source_id, int level,
333 struct kvm_s390_adapter_int adapter;
335 struct hlist_node link;
338 #ifndef KVM_PRIVATE_MEM_SLOTS
339 #define KVM_PRIVATE_MEM_SLOTS 0
342 #ifndef KVM_MEM_SLOTS_NUM
343 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
348 * memslots are not sorted by id anymore, please use id_to_memslot()
349 * to get the memslot by its id.
351 struct kvm_memslots {
353 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
354 /* The mapping table from slot id to the index in memslots[]. */
355 short id_to_index[KVM_MEM_SLOTS_NUM];
360 struct mutex slots_lock;
361 struct mm_struct *mm; /* userspace tied to this vm */
362 struct kvm_memslots *memslots;
363 struct srcu_struct srcu;
364 struct srcu_struct irq_srcu;
365 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
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 " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
426 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
429 return kvm->vcpus[i];
432 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
434 idx < atomic_read(&kvm->online_vcpus) && \
435 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
438 #define kvm_for_each_memslot(memslot, slots) \
439 for (memslot = &slots->memslots[0]; \
440 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
443 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
444 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
446 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
447 void vcpu_put(struct kvm_vcpu *vcpu);
449 #ifdef __KVM_HAVE_IOAPIC
450 void kvm_vcpu_request_scan_ioapic(struct kvm *kvm);
452 static inline void kvm_vcpu_request_scan_ioapic(struct kvm *kvm)
457 #ifdef CONFIG_HAVE_KVM_IRQFD
458 int kvm_irqfd_init(void);
459 void kvm_irqfd_exit(void);
461 static inline int kvm_irqfd_init(void)
466 static inline void kvm_irqfd_exit(void)
470 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
471 struct module *module);
474 void kvm_get_kvm(struct kvm *kvm);
475 void kvm_put_kvm(struct kvm *kvm);
477 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
479 return rcu_dereference_check(kvm->memslots,
480 srcu_read_lock_held(&kvm->srcu)
481 || lockdep_is_held(&kvm->slots_lock));
484 static inline struct kvm_memory_slot *
485 id_to_memslot(struct kvm_memslots *slots, int id)
487 int index = slots->id_to_index[id];
488 struct kvm_memory_slot *slot;
490 slot = &slots->memslots[index];
492 WARN_ON(slot->id != id);
497 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
498 * - create a new memory slot
499 * - delete an existing memory slot
500 * - modify an existing memory slot
501 * -- move it in the guest physical memory space
502 * -- just change its flags
504 * Since flags can be changed by some of these operations, the following
505 * differentiation is the best we can do for __kvm_set_memory_region():
514 int kvm_set_memory_region(struct kvm *kvm,
515 struct kvm_userspace_memory_region *mem);
516 int __kvm_set_memory_region(struct kvm *kvm,
517 struct kvm_userspace_memory_region *mem);
518 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
519 struct kvm_memory_slot *dont);
520 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
521 unsigned long npages);
522 void kvm_arch_memslots_updated(struct kvm *kvm);
523 int kvm_arch_prepare_memory_region(struct kvm *kvm,
524 struct kvm_memory_slot *memslot,
525 struct kvm_userspace_memory_region *mem,
526 enum kvm_mr_change change);
527 void kvm_arch_commit_memory_region(struct kvm *kvm,
528 struct kvm_userspace_memory_region *mem,
529 const struct kvm_memory_slot *old,
530 enum kvm_mr_change change);
531 bool kvm_largepages_enabled(void);
532 void kvm_disable_largepages(void);
533 /* flush all memory translations */
534 void kvm_arch_flush_shadow_all(struct kvm *kvm);
535 /* flush memory translations pointing to 'slot' */
536 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
537 struct kvm_memory_slot *slot);
539 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
542 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
543 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
544 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
545 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
546 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
548 void kvm_release_page_clean(struct page *page);
549 void kvm_release_page_dirty(struct page *page);
550 void kvm_set_page_accessed(struct page *page);
552 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
553 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
554 bool write_fault, bool *writable);
555 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
556 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
558 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
559 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
561 void kvm_release_pfn_clean(pfn_t pfn);
562 void kvm_set_pfn_dirty(pfn_t pfn);
563 void kvm_set_pfn_accessed(pfn_t pfn);
564 void kvm_get_pfn(pfn_t pfn);
566 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
568 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
570 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
571 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
572 void *data, unsigned long len);
573 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
574 int offset, int len);
575 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
577 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
578 void *data, unsigned long len);
579 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
580 gpa_t gpa, unsigned long len);
581 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
582 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
583 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
584 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
585 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
586 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
588 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
589 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
590 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
591 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
592 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
593 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
595 void kvm_flush_remote_tlbs(struct kvm *kvm);
596 void kvm_reload_remote_mmus(struct kvm *kvm);
597 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
598 void kvm_make_scan_ioapic_request(struct kvm *kvm);
599 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
601 long kvm_arch_dev_ioctl(struct file *filp,
602 unsigned int ioctl, unsigned long arg);
603 long kvm_arch_vcpu_ioctl(struct file *filp,
604 unsigned int ioctl, unsigned long arg);
605 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
607 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
609 int kvm_get_dirty_log(struct kvm *kvm,
610 struct kvm_dirty_log *log, int *is_dirty);
611 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
612 struct kvm_dirty_log *log);
614 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
616 long kvm_arch_vm_ioctl(struct file *filp,
617 unsigned int ioctl, unsigned long arg);
619 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
620 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
622 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
623 struct kvm_translation *tr);
625 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
626 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
627 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
628 struct kvm_sregs *sregs);
629 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
630 struct kvm_sregs *sregs);
631 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
632 struct kvm_mp_state *mp_state);
633 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
634 struct kvm_mp_state *mp_state);
635 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
636 struct kvm_guest_debug *dbg);
637 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
639 int kvm_arch_init(void *opaque);
640 void kvm_arch_exit(void);
642 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
643 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
645 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
647 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
648 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
649 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
650 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
651 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
652 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
653 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
655 int kvm_arch_hardware_enable(void);
656 void kvm_arch_hardware_disable(void);
657 int kvm_arch_hardware_setup(void);
658 void kvm_arch_hardware_unsetup(void);
659 void kvm_arch_check_processor_compat(void *rtn);
660 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
661 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
663 void *kvm_kvzalloc(unsigned long size);
664 void kvm_kvfree(const void *addr);
666 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
667 static inline struct kvm *kvm_arch_alloc_vm(void)
669 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
672 static inline void kvm_arch_free_vm(struct kvm *kvm)
678 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
679 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
680 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
681 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
683 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
687 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
691 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
697 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
699 #ifdef __KVM_HAVE_ARCH_WQP
700 return vcpu->arch.wqp;
706 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
707 void kvm_arch_destroy_vm(struct kvm *kvm);
708 void kvm_arch_sync_events(struct kvm *kvm);
710 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
711 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
713 bool kvm_is_mmio_pfn(pfn_t pfn);
715 struct kvm_irq_ack_notifier {
716 struct hlist_node link;
718 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
721 struct kvm_assigned_dev_kernel {
722 struct kvm_irq_ack_notifier ack_notifier;
723 struct list_head list;
728 unsigned int entries_nr;
730 bool host_irq_disabled;
732 struct msix_entry *host_msix_entries;
734 struct msix_entry *guest_msix_entries;
735 unsigned long irq_requested_type;
740 spinlock_t intx_lock;
741 spinlock_t intx_mask_lock;
743 struct pci_saved_state *pci_saved_state;
746 int kvm_irq_map_gsi(struct kvm *kvm,
747 struct kvm_kernel_irq_routing_entry *entries, int gsi);
748 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
750 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
752 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
753 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
754 int irq_source_id, int level, bool line_status);
755 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
756 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
757 void kvm_register_irq_ack_notifier(struct kvm *kvm,
758 struct kvm_irq_ack_notifier *kian);
759 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
760 struct kvm_irq_ack_notifier *kian);
761 int kvm_request_irq_source_id(struct kvm *kvm);
762 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
764 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
765 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
766 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
767 int kvm_assign_device(struct kvm *kvm,
768 struct kvm_assigned_dev_kernel *assigned_dev);
769 int kvm_deassign_device(struct kvm *kvm,
770 struct kvm_assigned_dev_kernel *assigned_dev);
772 static inline int kvm_iommu_map_pages(struct kvm *kvm,
773 struct kvm_memory_slot *slot)
778 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
779 struct kvm_memory_slot *slot)
784 static inline void kvm_guest_enter(void)
788 BUG_ON(preemptible());
790 local_irq_save(flags);
792 local_irq_restore(flags);
794 /* KVM does not hold any references to rcu protected data when it
795 * switches CPU into a guest mode. In fact switching to a guest mode
796 * is very similar to exiting to userspace from rcu point of view. In
797 * addition CPU may stay in a guest mode for quite a long time (up to
798 * one time slice). Lets treat guest mode as quiescent state, just like
799 * we do with user-mode execution.
801 rcu_virt_note_context_switch(smp_processor_id());
804 static inline void kvm_guest_exit(void)
808 local_irq_save(flags);
810 local_irq_restore(flags);
814 * search_memslots() and __gfn_to_memslot() are here because they are
815 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
816 * gfn_to_memslot() itself isn't here as an inline because that would
817 * bloat other code too much.
819 static inline struct kvm_memory_slot *
820 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
822 struct kvm_memory_slot *memslot;
824 kvm_for_each_memslot(memslot, slots)
825 if (gfn >= memslot->base_gfn &&
826 gfn < memslot->base_gfn + memslot->npages)
832 static inline struct kvm_memory_slot *
833 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
835 return search_memslots(slots, gfn);
838 static inline unsigned long
839 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
841 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
844 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
846 return gfn_to_memslot(kvm, gfn)->id;
850 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
852 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
854 return slot->base_gfn + gfn_offset;
857 static inline gpa_t gfn_to_gpa(gfn_t gfn)
859 return (gpa_t)gfn << PAGE_SHIFT;
862 static inline gfn_t gpa_to_gfn(gpa_t gpa)
864 return (gfn_t)(gpa >> PAGE_SHIFT);
867 static inline hpa_t pfn_to_hpa(pfn_t pfn)
869 return (hpa_t)pfn << PAGE_SHIFT;
872 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
874 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
876 return kvm_is_error_hva(hva);
879 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
881 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
889 struct kvm_stats_debugfs_item {
892 enum kvm_stat_kind kind;
893 struct dentry *dentry;
895 extern struct kvm_stats_debugfs_item debugfs_entries[];
896 extern struct dentry *kvm_debugfs_dir;
898 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
899 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
901 if (unlikely(kvm->mmu_notifier_count))
904 * Ensure the read of mmu_notifier_count happens before the read
905 * of mmu_notifier_seq. This interacts with the smp_wmb() in
906 * mmu_notifier_invalidate_range_end to make sure that the caller
907 * either sees the old (non-zero) value of mmu_notifier_count or
908 * the new (incremented) value of mmu_notifier_seq.
909 * PowerPC Book3s HV KVM calls this under a per-page lock
910 * rather than under kvm->mmu_lock, for scalability, so
911 * can't rely on kvm->mmu_lock to keep things ordered.
914 if (kvm->mmu_notifier_seq != mmu_seq)
920 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
923 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
925 #define KVM_MAX_IRQ_ROUTES 1024
928 int kvm_setup_default_irq_routing(struct kvm *kvm);
929 int kvm_set_irq_routing(struct kvm *kvm,
930 const struct kvm_irq_routing_entry *entries,
933 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
934 const struct kvm_irq_routing_entry *ue);
935 void kvm_free_irq_routing(struct kvm *kvm);
939 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
943 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
945 #ifdef CONFIG_HAVE_KVM_EVENTFD
947 void kvm_eventfd_init(struct kvm *kvm);
948 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
950 #ifdef CONFIG_HAVE_KVM_IRQFD
951 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
952 void kvm_irqfd_release(struct kvm *kvm);
953 void kvm_irq_routing_update(struct kvm *);
955 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
960 static inline void kvm_irqfd_release(struct kvm *kvm) {}
965 static inline void kvm_eventfd_init(struct kvm *kvm) {}
967 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
972 static inline void kvm_irqfd_release(struct kvm *kvm) {}
974 #ifdef CONFIG_HAVE_KVM_IRQCHIP
975 static inline void kvm_irq_routing_update(struct kvm *kvm)
980 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
985 #endif /* CONFIG_HAVE_KVM_EVENTFD */
987 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
988 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
990 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
993 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
997 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
1001 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1003 set_bit(req, &vcpu->requests);
1006 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1008 if (test_bit(req, &vcpu->requests)) {
1009 clear_bit(req, &vcpu->requests);
1016 extern bool kvm_rebooting;
1019 struct kvm_device_ops *ops;
1022 struct list_head vm_node;
1025 /* create, destroy, and name are mandatory */
1026 struct kvm_device_ops {
1028 int (*create)(struct kvm_device *dev, u32 type);
1031 * Destroy is responsible for freeing dev.
1033 * Destroy may be called before or after destructors are called
1034 * on emulated I/O regions, depending on whether a reference is
1035 * held by a vcpu or other kvm component that gets destroyed
1036 * after the emulated I/O.
1038 void (*destroy)(struct kvm_device *dev);
1040 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1041 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1042 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1043 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1047 void kvm_device_get(struct kvm_device *dev);
1048 void kvm_device_put(struct kvm_device *dev);
1049 struct kvm_device *kvm_device_from_filp(struct file *filp);
1050 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1051 void kvm_unregister_device_ops(u32 type);
1053 extern struct kvm_device_ops kvm_mpic_ops;
1054 extern struct kvm_device_ops kvm_xics_ops;
1056 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1058 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1060 vcpu->spin_loop.in_spin_loop = val;
1062 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1064 vcpu->spin_loop.dy_eligible = val;
1067 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1069 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1073 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1076 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */