1 /* SPDX-License-Identifier: GPL-2.0-only */
6 #include <linux/types.h>
7 #include <linux/hardirq.h>
8 #include <linux/list.h>
9 #include <linux/mutex.h>
10 #include <linux/spinlock.h>
11 #include <linux/signal.h>
12 #include <linux/sched.h>
13 #include <linux/sched/stat.h>
14 #include <linux/bug.h>
15 #include <linux/minmax.h>
17 #include <linux/mmu_notifier.h>
18 #include <linux/preempt.h>
19 #include <linux/msi.h>
20 #include <linux/slab.h>
21 #include <linux/vmalloc.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/rcuwait.h>
29 #include <linux/refcount.h>
30 #include <linux/nospec.h>
31 #include <linux/notifier.h>
32 #include <asm/signal.h>
34 #include <linux/kvm.h>
35 #include <linux/kvm_para.h>
37 #include <linux/kvm_types.h>
39 #include <asm/kvm_host.h>
40 #include <linux/kvm_dirty_ring.h>
42 #ifndef KVM_MAX_VCPU_ID
43 #define KVM_MAX_VCPU_ID KVM_MAX_VCPUS
47 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
48 * in kvm, other bits are visible for userspace which are defined in
49 * include/linux/kvm_h.
51 #define KVM_MEMSLOT_INVALID (1UL << 16)
54 * Bit 63 of the memslot generation number is an "update in-progress flag",
55 * e.g. is temporarily set for the duration of install_new_memslots().
56 * This flag effectively creates a unique generation number that is used to
57 * mark cached memslot data, e.g. MMIO accesses, as potentially being stale,
58 * i.e. may (or may not) have come from the previous memslots generation.
60 * This is necessary because the actual memslots update is not atomic with
61 * respect to the generation number update. Updating the generation number
62 * first would allow a vCPU to cache a spte from the old memslots using the
63 * new generation number, and updating the generation number after switching
64 * to the new memslots would allow cache hits using the old generation number
65 * to reference the defunct memslots.
67 * This mechanism is used to prevent getting hits in KVM's caches while a
68 * memslot update is in-progress, and to prevent cache hits *after* updating
69 * the actual generation number against accesses that were inserted into the
70 * cache *before* the memslots were updated.
72 #define KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS BIT_ULL(63)
74 /* Two fragments for cross MMIO pages. */
75 #define KVM_MAX_MMIO_FRAGMENTS 2
77 #ifndef KVM_ADDRESS_SPACE_NUM
78 #define KVM_ADDRESS_SPACE_NUM 1
82 * For the normal pfn, the highest 12 bits should be zero,
83 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
84 * mask bit 63 to indicate the noslot pfn.
86 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
87 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
88 #define KVM_PFN_NOSLOT (0x1ULL << 63)
90 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
91 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
92 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
95 * error pfns indicate that the gfn is in slot but faild to
96 * translate it to pfn on host.
98 static inline bool is_error_pfn(kvm_pfn_t pfn)
100 return !!(pfn & KVM_PFN_ERR_MASK);
104 * error_noslot pfns indicate that the gfn can not be
105 * translated to pfn - it is not in slot or failed to
106 * translate it to pfn.
108 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn)
110 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
113 /* noslot pfn indicates that the gfn is not in slot. */
114 static inline bool is_noslot_pfn(kvm_pfn_t pfn)
116 return pfn == KVM_PFN_NOSLOT;
120 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
121 * provide own defines and kvm_is_error_hva
123 #ifndef KVM_HVA_ERR_BAD
125 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
126 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
128 static inline bool kvm_is_error_hva(unsigned long addr)
130 return addr >= PAGE_OFFSET;
135 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
137 static inline bool is_error_page(struct page *page)
142 #define KVM_REQUEST_MASK GENMASK(7,0)
143 #define KVM_REQUEST_NO_WAKEUP BIT(8)
144 #define KVM_REQUEST_WAIT BIT(9)
146 * Architecture-independent vcpu->requests bit members
147 * Bits 4-7 are reserved for more arch-independent bits.
149 #define KVM_REQ_TLB_FLUSH (0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
150 #define KVM_REQ_MMU_RELOAD (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
151 #define KVM_REQ_UNBLOCK 2
152 #define KVM_REQ_UNHALT 3
153 #define KVM_REQ_VM_BUGGED (4 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
154 #define KVM_REQUEST_ARCH_BASE 8
156 #define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
157 BUILD_BUG_ON((unsigned)(nr) >= (sizeof_field(struct kvm_vcpu, requests) * 8) - KVM_REQUEST_ARCH_BASE); \
158 (unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) | (flags)); \
160 #define KVM_ARCH_REQ(nr) KVM_ARCH_REQ_FLAGS(nr, 0)
162 bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
163 struct kvm_vcpu *except,
164 unsigned long *vcpu_bitmap, cpumask_var_t tmp);
165 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
166 bool kvm_make_all_cpus_request_except(struct kvm *kvm, unsigned int req,
167 struct kvm_vcpu *except);
168 bool kvm_make_cpus_request_mask(struct kvm *kvm, unsigned int req,
169 unsigned long *vcpu_bitmap);
171 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
172 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
174 extern struct mutex kvm_lock;
175 extern struct list_head vm_list;
177 struct kvm_io_range {
180 struct kvm_io_device *dev;
183 #define NR_IOBUS_DEVS 1000
188 struct kvm_io_range range[];
194 KVM_VIRTIO_CCW_NOTIFY_BUS,
199 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
200 int len, const void *val);
201 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
202 gpa_t addr, int len, const void *val, long cookie);
203 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
205 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
206 int len, struct kvm_io_device *dev);
207 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
208 struct kvm_io_device *dev);
209 struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx,
212 #ifdef CONFIG_KVM_ASYNC_PF
213 struct kvm_async_pf {
214 struct work_struct work;
215 struct list_head link;
216 struct list_head queue;
217 struct kvm_vcpu *vcpu;
218 struct mm_struct *mm;
221 struct kvm_arch_async_pf arch;
223 bool notpresent_injected;
226 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
227 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
228 bool kvm_setup_async_pf(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
229 unsigned long hva, struct kvm_arch_async_pf *arch);
230 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
233 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
234 struct kvm_gfn_range {
235 struct kvm_memory_slot *slot;
241 bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range);
242 bool kvm_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
243 bool kvm_test_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
244 bool kvm_set_spte_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
251 READING_SHADOW_PAGE_TABLES,
254 #define KVM_UNMAPPED_PAGE ((void *) 0x500 + POISON_POINTER_DELTA)
256 struct kvm_host_map {
258 * Only valid if the 'pfn' is managed by the host kernel (i.e. There is
259 * a 'struct page' for it. When using mem= kernel parameter some memory
260 * can be used as guest memory but they are not managed by host
262 * If 'pfn' is not managed by the host kernel, this field is
263 * initialized to KVM_UNMAPPED_PAGE.
272 * Used to check if the mapping is valid or not. Never use 'kvm_host_map'
273 * directly to check for that.
275 static inline bool kvm_vcpu_mapped(struct kvm_host_map *map)
280 static inline bool kvm_vcpu_can_poll(ktime_t cur, ktime_t stop)
282 return single_task_running() && !need_resched() && ktime_before(cur, stop);
286 * Sometimes a large or cross-page mmio needs to be broken up into separate
287 * exits for userspace servicing.
289 struct kvm_mmio_fragment {
297 #ifdef CONFIG_PREEMPT_NOTIFIERS
298 struct preempt_notifier preempt_notifier;
301 int vcpu_id; /* id given by userspace at creation */
302 int vcpu_idx; /* index in kvm->vcpus array */
306 unsigned long guest_debug;
309 struct list_head blocked_vcpu_list;
315 struct pid __rcu *pid;
318 unsigned int halt_poll_ns;
321 #ifdef CONFIG_HAS_IOMEM
323 int mmio_read_completed;
325 int mmio_cur_fragment;
326 int mmio_nr_fragments;
327 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
330 #ifdef CONFIG_KVM_ASYNC_PF
333 struct list_head queue;
334 struct list_head done;
339 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
341 * Cpu relax intercept or pause loop exit optimization
342 * in_spin_loop: set when a vcpu does a pause loop exit
343 * or cpu relax intercepted.
344 * dy_eligible: indicates whether vcpu is eligible for directed yield.
353 struct kvm_vcpu_arch arch;
354 struct kvm_vcpu_stat stat;
355 char stats_id[KVM_STATS_NAME_SIZE];
356 struct kvm_dirty_ring dirty_ring;
359 * The index of the most recently used memslot by this vCPU. It's ok
360 * if this becomes stale due to memslot changes since we always check
361 * it is a valid slot.
366 /* must be called with irqs disabled */
367 static __always_inline void guest_enter_irqoff(void)
370 * This is running in ioctl context so its safe to assume that it's the
371 * stime pending cputime to flush.
373 instrumentation_begin();
374 vtime_account_guest_enter();
375 instrumentation_end();
378 * KVM does not hold any references to rcu protected data when it
379 * switches CPU into a guest mode. In fact switching to a guest mode
380 * is very similar to exiting to userspace from rcu point of view. In
381 * addition CPU may stay in a guest mode for quite a long time (up to
382 * one time slice). Lets treat guest mode as quiescent state, just like
383 * we do with user-mode execution.
385 if (!context_tracking_guest_enter()) {
386 instrumentation_begin();
387 rcu_virt_note_context_switch(smp_processor_id());
388 instrumentation_end();
392 static __always_inline void guest_exit_irqoff(void)
394 context_tracking_guest_exit();
396 instrumentation_begin();
397 /* Flush the guest cputime we spent on the guest */
398 vtime_account_guest_exit();
399 instrumentation_end();
402 static inline void guest_exit(void)
406 local_irq_save(flags);
408 local_irq_restore(flags);
411 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
414 * The memory barrier ensures a previous write to vcpu->requests cannot
415 * be reordered with the read of vcpu->mode. It pairs with the general
416 * memory barrier following the write of vcpu->mode in VCPU RUN.
418 smp_mb__before_atomic();
419 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
423 * Some of the bitops functions do not support too long bitmaps.
424 * This number must be determined not to exceed such limits.
426 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
428 struct kvm_memory_slot {
430 unsigned long npages;
431 unsigned long *dirty_bitmap;
432 struct kvm_arch_memory_slot arch;
433 unsigned long userspace_addr;
439 static inline bool kvm_slot_dirty_track_enabled(struct kvm_memory_slot *slot)
441 return slot->flags & KVM_MEM_LOG_DIRTY_PAGES;
444 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
446 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
449 static inline unsigned long *kvm_second_dirty_bitmap(struct kvm_memory_slot *memslot)
451 unsigned long len = kvm_dirty_bitmap_bytes(memslot);
453 return memslot->dirty_bitmap + len / sizeof(*memslot->dirty_bitmap);
456 #ifndef KVM_DIRTY_LOG_MANUAL_CAPS
457 #define KVM_DIRTY_LOG_MANUAL_CAPS KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE
460 struct kvm_s390_adapter_int {
473 struct kvm_kernel_irq_routing_entry {
476 int (*set)(struct kvm_kernel_irq_routing_entry *e,
477 struct kvm *kvm, int irq_source_id, int level,
491 struct kvm_s390_adapter_int adapter;
492 struct kvm_hv_sint hv_sint;
494 struct hlist_node link;
497 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
498 struct kvm_irq_routing_table {
499 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
502 * Array indexed by gsi. Each entry contains list of irq chips
503 * the gsi is connected to.
505 struct hlist_head map[];
509 #ifndef KVM_PRIVATE_MEM_SLOTS
510 #define KVM_PRIVATE_MEM_SLOTS 0
513 #define KVM_MEM_SLOTS_NUM SHRT_MAX
514 #define KVM_USER_MEM_SLOTS (KVM_MEM_SLOTS_NUM - KVM_PRIVATE_MEM_SLOTS)
516 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
517 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
525 * memslots are not sorted by id anymore, please use id_to_memslot()
526 * to get the memslot by its id.
528 struct kvm_memslots {
530 /* The mapping table from slot id to the index in memslots[]. */
531 short id_to_index[KVM_MEM_SLOTS_NUM];
532 atomic_t last_used_slot;
534 struct kvm_memory_slot memslots[];
538 #ifdef KVM_HAVE_MMU_RWLOCK
542 #endif /* KVM_HAVE_MMU_RWLOCK */
544 struct mutex slots_lock;
547 * Protects the arch-specific fields of struct kvm_memory_slots in
548 * use by the VM. To be used under the slots_lock (above) or in a
549 * kvm->srcu critical section where acquiring the slots_lock would
550 * lead to deadlock with the synchronize_srcu in
551 * install_new_memslots.
553 struct mutex slots_arch_lock;
554 struct mm_struct *mm; /* userspace tied to this vm */
555 struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
556 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
558 /* Used to wait for completion of MMU notifiers. */
559 spinlock_t mn_invalidate_lock;
560 unsigned long mn_active_invalidate_count;
561 struct rcuwait mn_memslots_update_rcuwait;
564 * created_vcpus is protected by kvm->lock, and is incremented
565 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
566 * incremented after storing the kvm_vcpu pointer in vcpus,
567 * and is accessed atomically.
569 atomic_t online_vcpus;
571 int last_boosted_vcpu;
572 struct list_head vm_list;
574 struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
575 #ifdef CONFIG_HAVE_KVM_EVENTFD
578 struct list_head items;
579 struct list_head resampler_list;
580 struct mutex resampler_lock;
582 struct list_head ioeventfds;
584 struct kvm_vm_stat stat;
585 struct kvm_arch arch;
586 refcount_t users_count;
587 #ifdef CONFIG_KVM_MMIO
588 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
589 spinlock_t ring_lock;
590 struct list_head coalesced_zones;
593 struct mutex irq_lock;
594 #ifdef CONFIG_HAVE_KVM_IRQCHIP
596 * Update side is protected by irq_lock.
598 struct kvm_irq_routing_table __rcu *irq_routing;
600 #ifdef CONFIG_HAVE_KVM_IRQFD
601 struct hlist_head irq_ack_notifier_list;
604 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
605 struct mmu_notifier mmu_notifier;
606 unsigned long mmu_notifier_seq;
607 long mmu_notifier_count;
608 unsigned long mmu_notifier_range_start;
609 unsigned long mmu_notifier_range_end;
611 struct list_head devices;
612 u64 manual_dirty_log_protect;
613 struct dentry *debugfs_dentry;
614 struct kvm_stat_data **debugfs_stat_data;
615 struct srcu_struct srcu;
616 struct srcu_struct irq_srcu;
618 unsigned int max_halt_poll_ns;
622 #ifdef CONFIG_HAVE_KVM_PM_NOTIFIER
623 struct notifier_block pm_notifier;
625 char stats_id[KVM_STATS_NAME_SIZE];
628 #define kvm_err(fmt, ...) \
629 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
630 #define kvm_info(fmt, ...) \
631 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
632 #define kvm_debug(fmt, ...) \
633 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
634 #define kvm_debug_ratelimited(fmt, ...) \
635 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
637 #define kvm_pr_unimpl(fmt, ...) \
638 pr_err_ratelimited("kvm [%i]: " fmt, \
639 task_tgid_nr(current), ## __VA_ARGS__)
641 /* The guest did something we don't support. */
642 #define vcpu_unimpl(vcpu, fmt, ...) \
643 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
644 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
646 #define vcpu_debug(vcpu, fmt, ...) \
647 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
648 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
649 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
651 #define vcpu_err(vcpu, fmt, ...) \
652 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
654 static inline void kvm_vm_bugged(struct kvm *kvm)
656 kvm->vm_bugged = true;
657 kvm_make_all_cpus_request(kvm, KVM_REQ_VM_BUGGED);
660 #define KVM_BUG(cond, kvm, fmt...) \
662 int __ret = (cond); \
664 if (WARN_ONCE(__ret && !(kvm)->vm_bugged, fmt)) \
665 kvm_vm_bugged(kvm); \
669 #define KVM_BUG_ON(cond, kvm) \
671 int __ret = (cond); \
673 if (WARN_ON_ONCE(__ret && !(kvm)->vm_bugged)) \
674 kvm_vm_bugged(kvm); \
678 static inline bool kvm_dirty_log_manual_protect_and_init_set(struct kvm *kvm)
680 return !!(kvm->manual_dirty_log_protect & KVM_DIRTY_LOG_INITIALLY_SET);
683 static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx)
685 return srcu_dereference_check(kvm->buses[idx], &kvm->srcu,
686 lockdep_is_held(&kvm->slots_lock) ||
687 !refcount_read(&kvm->users_count));
690 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
692 int num_vcpus = atomic_read(&kvm->online_vcpus);
693 i = array_index_nospec(i, num_vcpus);
695 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. */
697 return kvm->vcpus[i];
700 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
702 idx < atomic_read(&kvm->online_vcpus) && \
703 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
706 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
708 struct kvm_vcpu *vcpu = NULL;
713 if (id < KVM_MAX_VCPUS)
714 vcpu = kvm_get_vcpu(kvm, id);
715 if (vcpu && vcpu->vcpu_id == id)
717 kvm_for_each_vcpu(i, vcpu, kvm)
718 if (vcpu->vcpu_id == id)
723 #define kvm_for_each_memslot(memslot, slots) \
724 for (memslot = &slots->memslots[0]; \
725 memslot < slots->memslots + slots->used_slots; memslot++) \
726 if (WARN_ON_ONCE(!memslot->npages)) { \
729 void kvm_vcpu_destroy(struct kvm_vcpu *vcpu);
731 void vcpu_load(struct kvm_vcpu *vcpu);
732 void vcpu_put(struct kvm_vcpu *vcpu);
734 #ifdef __KVM_HAVE_IOAPIC
735 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
736 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
738 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
741 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
746 #ifdef CONFIG_HAVE_KVM_IRQFD
747 int kvm_irqfd_init(void);
748 void kvm_irqfd_exit(void);
750 static inline int kvm_irqfd_init(void)
755 static inline void kvm_irqfd_exit(void)
759 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
760 struct module *module);
763 void kvm_get_kvm(struct kvm *kvm);
764 bool kvm_get_kvm_safe(struct kvm *kvm);
765 void kvm_put_kvm(struct kvm *kvm);
766 bool file_is_kvm(struct file *file);
767 void kvm_put_kvm_no_destroy(struct kvm *kvm);
769 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
771 as_id = array_index_nospec(as_id, KVM_ADDRESS_SPACE_NUM);
772 return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu,
773 lockdep_is_held(&kvm->slots_lock) ||
774 !refcount_read(&kvm->users_count));
777 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
779 return __kvm_memslots(kvm, 0);
782 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
784 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
786 return __kvm_memslots(vcpu->kvm, as_id);
790 struct kvm_memory_slot *id_to_memslot(struct kvm_memslots *slots, int id)
792 int index = slots->id_to_index[id];
793 struct kvm_memory_slot *slot;
798 slot = &slots->memslots[index];
800 WARN_ON(slot->id != id);
805 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
806 * - create a new memory slot
807 * - delete an existing memory slot
808 * - modify an existing memory slot
809 * -- move it in the guest physical memory space
810 * -- just change its flags
812 * Since flags can be changed by some of these operations, the following
813 * differentiation is the best we can do for __kvm_set_memory_region():
822 int kvm_set_memory_region(struct kvm *kvm,
823 const struct kvm_userspace_memory_region *mem);
824 int __kvm_set_memory_region(struct kvm *kvm,
825 const struct kvm_userspace_memory_region *mem);
826 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot);
827 void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen);
828 int kvm_arch_prepare_memory_region(struct kvm *kvm,
829 struct kvm_memory_slot *memslot,
830 const struct kvm_userspace_memory_region *mem,
831 enum kvm_mr_change change);
832 void kvm_arch_commit_memory_region(struct kvm *kvm,
833 const struct kvm_userspace_memory_region *mem,
834 struct kvm_memory_slot *old,
835 const struct kvm_memory_slot *new,
836 enum kvm_mr_change change);
837 /* flush all memory translations */
838 void kvm_arch_flush_shadow_all(struct kvm *kvm);
839 /* flush memory translations pointing to 'slot' */
840 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
841 struct kvm_memory_slot *slot);
843 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
844 struct page **pages, int nr_pages);
846 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
847 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
848 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
849 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
850 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
852 void kvm_release_page_clean(struct page *page);
853 void kvm_release_page_dirty(struct page *page);
854 void kvm_set_page_accessed(struct page *page);
856 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
857 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
859 kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
860 kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
861 kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
862 bool atomic, bool *async, bool write_fault,
863 bool *writable, hva_t *hva);
865 void kvm_release_pfn_clean(kvm_pfn_t pfn);
866 void kvm_release_pfn_dirty(kvm_pfn_t pfn);
867 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
868 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
870 void kvm_release_pfn(kvm_pfn_t pfn, bool dirty, struct gfn_to_pfn_cache *cache);
871 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
873 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
874 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
875 void *data, unsigned long len);
876 int kvm_read_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
877 void *data, unsigned int offset,
879 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
880 int offset, int len);
881 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
883 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
884 void *data, unsigned long len);
885 int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
886 void *data, unsigned int offset,
888 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
889 gpa_t gpa, unsigned long len);
891 #define __kvm_get_guest(kvm, gfn, offset, v) \
893 unsigned long __addr = gfn_to_hva(kvm, gfn); \
894 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
895 int __ret = -EFAULT; \
897 if (!kvm_is_error_hva(__addr)) \
898 __ret = get_user(v, __uaddr); \
902 #define kvm_get_guest(kvm, gpa, v) \
905 struct kvm *__kvm = kvm; \
907 __kvm_get_guest(__kvm, __gpa >> PAGE_SHIFT, \
908 offset_in_page(__gpa), v); \
911 #define __kvm_put_guest(kvm, gfn, offset, v) \
913 unsigned long __addr = gfn_to_hva(kvm, gfn); \
914 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
915 int __ret = -EFAULT; \
917 if (!kvm_is_error_hva(__addr)) \
918 __ret = put_user(v, __uaddr); \
920 mark_page_dirty(kvm, gfn); \
924 #define kvm_put_guest(kvm, gpa, v) \
927 struct kvm *__kvm = kvm; \
929 __kvm_put_guest(__kvm, __gpa >> PAGE_SHIFT, \
930 offset_in_page(__gpa), v); \
933 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
934 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
935 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
936 bool kvm_vcpu_is_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
937 unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn);
938 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot, gfn_t gfn);
939 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
941 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
942 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
943 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
944 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
945 int kvm_vcpu_map(struct kvm_vcpu *vcpu, gpa_t gpa, struct kvm_host_map *map);
946 int kvm_map_gfn(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map,
947 struct gfn_to_pfn_cache *cache, bool atomic);
948 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
949 void kvm_vcpu_unmap(struct kvm_vcpu *vcpu, struct kvm_host_map *map, bool dirty);
950 int kvm_unmap_gfn(struct kvm_vcpu *vcpu, struct kvm_host_map *map,
951 struct gfn_to_pfn_cache *cache, bool dirty, bool atomic);
952 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
953 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
954 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
956 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
958 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
960 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
961 int offset, int len);
962 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
964 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
966 void kvm_sigset_activate(struct kvm_vcpu *vcpu);
967 void kvm_sigset_deactivate(struct kvm_vcpu *vcpu);
969 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
970 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
971 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
972 bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
973 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
974 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
975 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool usermode_vcpu_not_eligible);
977 void kvm_flush_remote_tlbs(struct kvm *kvm);
978 void kvm_reload_remote_mmus(struct kvm *kvm);
980 #ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE
981 int kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache *mc, int min);
982 int kvm_mmu_memory_cache_nr_free_objects(struct kvm_mmu_memory_cache *mc);
983 void kvm_mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc);
984 void *kvm_mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc);
987 void kvm_inc_notifier_count(struct kvm *kvm, unsigned long start,
989 void kvm_dec_notifier_count(struct kvm *kvm, unsigned long start,
992 long kvm_arch_dev_ioctl(struct file *filp,
993 unsigned int ioctl, unsigned long arg);
994 long kvm_arch_vcpu_ioctl(struct file *filp,
995 unsigned int ioctl, unsigned long arg);
996 vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
998 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
1000 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
1001 struct kvm_memory_slot *slot,
1003 unsigned long mask);
1004 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot);
1006 #ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
1007 void kvm_arch_flush_remote_tlbs_memslot(struct kvm *kvm,
1008 const struct kvm_memory_slot *memslot);
1009 #else /* !CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
1010 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log);
1011 int kvm_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log,
1012 int *is_dirty, struct kvm_memory_slot **memslot);
1015 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
1017 int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
1018 struct kvm_enable_cap *cap);
1019 long kvm_arch_vm_ioctl(struct file *filp,
1020 unsigned int ioctl, unsigned long arg);
1022 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
1023 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
1025 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1026 struct kvm_translation *tr);
1028 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
1029 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
1030 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1031 struct kvm_sregs *sregs);
1032 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1033 struct kvm_sregs *sregs);
1034 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
1035 struct kvm_mp_state *mp_state);
1036 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
1037 struct kvm_mp_state *mp_state);
1038 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
1039 struct kvm_guest_debug *dbg);
1040 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu);
1042 int kvm_arch_init(void *opaque);
1043 void kvm_arch_exit(void);
1045 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
1047 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
1048 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
1049 int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id);
1050 int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu);
1051 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
1052 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
1054 #ifdef CONFIG_HAVE_KVM_PM_NOTIFIER
1055 int kvm_arch_pm_notifier(struct kvm *kvm, unsigned long state);
1058 #ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
1059 void kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu, struct dentry *debugfs_dentry);
1062 int kvm_arch_hardware_enable(void);
1063 void kvm_arch_hardware_disable(void);
1064 int kvm_arch_hardware_setup(void *opaque);
1065 void kvm_arch_hardware_unsetup(void);
1066 int kvm_arch_check_processor_compat(void *opaque);
1067 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
1068 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
1069 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
1070 bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu);
1071 bool kvm_arch_dy_has_pending_interrupt(struct kvm_vcpu *vcpu);
1072 int kvm_arch_post_init_vm(struct kvm *kvm);
1073 void kvm_arch_pre_destroy_vm(struct kvm *kvm);
1074 int kvm_arch_create_vm_debugfs(struct kvm *kvm);
1076 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
1078 * All architectures that want to use vzalloc currently also
1079 * need their own kvm_arch_alloc_vm implementation.
1081 static inline struct kvm *kvm_arch_alloc_vm(void)
1083 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
1086 static inline void kvm_arch_free_vm(struct kvm *kvm)
1092 #ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1093 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1099 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
1100 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
1101 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
1102 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
1104 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
1108 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
1112 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
1117 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
1118 void kvm_arch_start_assignment(struct kvm *kvm);
1119 void kvm_arch_end_assignment(struct kvm *kvm);
1120 bool kvm_arch_has_assigned_device(struct kvm *kvm);
1122 static inline void kvm_arch_start_assignment(struct kvm *kvm)
1126 static inline void kvm_arch_end_assignment(struct kvm *kvm)
1130 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
1136 static inline struct rcuwait *kvm_arch_vcpu_get_wait(struct kvm_vcpu *vcpu)
1138 #ifdef __KVM_HAVE_ARCH_WQP
1139 return vcpu->arch.waitp;
1145 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
1147 * returns true if the virtual interrupt controller is initialized and
1148 * ready to accept virtual IRQ. On some architectures the virtual interrupt
1149 * controller is dynamically instantiated and this is not always true.
1151 bool kvm_arch_intc_initialized(struct kvm *kvm);
1153 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
1159 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
1160 void kvm_arch_destroy_vm(struct kvm *kvm);
1161 void kvm_arch_sync_events(struct kvm *kvm);
1163 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
1165 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
1166 bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
1167 bool kvm_is_transparent_hugepage(kvm_pfn_t pfn);
1169 struct kvm_irq_ack_notifier {
1170 struct hlist_node link;
1172 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
1175 int kvm_irq_map_gsi(struct kvm *kvm,
1176 struct kvm_kernel_irq_routing_entry *entries, int gsi);
1177 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
1179 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
1181 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
1182 int irq_source_id, int level, bool line_status);
1183 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
1184 struct kvm *kvm, int irq_source_id,
1185 int level, bool line_status);
1186 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
1187 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
1188 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
1189 void kvm_register_irq_ack_notifier(struct kvm *kvm,
1190 struct kvm_irq_ack_notifier *kian);
1191 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
1192 struct kvm_irq_ack_notifier *kian);
1193 int kvm_request_irq_source_id(struct kvm *kvm);
1194 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
1195 bool kvm_arch_irqfd_allowed(struct kvm *kvm, struct kvm_irqfd *args);
1198 * Returns a pointer to the memslot at slot_index if it contains gfn.
1199 * Otherwise returns NULL.
1201 static inline struct kvm_memory_slot *
1202 try_get_memslot(struct kvm_memslots *slots, int slot_index, gfn_t gfn)
1204 struct kvm_memory_slot *slot;
1206 if (slot_index < 0 || slot_index >= slots->used_slots)
1210 * slot_index can come from vcpu->last_used_slot which is not kept
1211 * in sync with userspace-controllable memslot deletion. So use nospec
1212 * to prevent the CPU from speculating past the end of memslots[].
1214 slot_index = array_index_nospec(slot_index, slots->used_slots);
1215 slot = &slots->memslots[slot_index];
1217 if (gfn >= slot->base_gfn && gfn < slot->base_gfn + slot->npages)
1224 * Returns a pointer to the memslot that contains gfn and records the index of
1225 * the slot in index. Otherwise returns NULL.
1227 * IMPORTANT: Slots are sorted from highest GFN to lowest GFN!
1229 static inline struct kvm_memory_slot *
1230 search_memslots(struct kvm_memslots *slots, gfn_t gfn, int *index)
1232 int start = 0, end = slots->used_slots;
1233 struct kvm_memory_slot *memslots = slots->memslots;
1234 struct kvm_memory_slot *slot;
1236 if (unlikely(!slots->used_slots))
1239 while (start < end) {
1240 int slot = start + (end - start) / 2;
1242 if (gfn >= memslots[slot].base_gfn)
1248 slot = try_get_memslot(slots, start, gfn);
1258 * __gfn_to_memslot() and its descendants are here because it is called from
1259 * non-modular code in arch/powerpc/kvm/book3s_64_vio{,_hv}.c. gfn_to_memslot()
1260 * itself isn't here as an inline because that would bloat other code too much.
1262 static inline struct kvm_memory_slot *
1263 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
1265 struct kvm_memory_slot *slot;
1266 int slot_index = atomic_read(&slots->last_used_slot);
1268 slot = try_get_memslot(slots, slot_index, gfn);
1272 slot = search_memslots(slots, gfn, &slot_index);
1274 atomic_set(&slots->last_used_slot, slot_index);
1281 static inline unsigned long
1282 __gfn_to_hva_memslot(const struct kvm_memory_slot *slot, gfn_t gfn)
1285 * The index was checked originally in search_memslots. To avoid
1286 * that a malicious guest builds a Spectre gadget out of e.g. page
1287 * table walks, do not let the processor speculate loads outside
1288 * the guest's registered memslots.
1290 unsigned long offset = gfn - slot->base_gfn;
1291 offset = array_index_nospec(offset, slot->npages);
1292 return slot->userspace_addr + offset * PAGE_SIZE;
1295 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
1297 return gfn_to_memslot(kvm, gfn)->id;
1301 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
1303 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
1305 return slot->base_gfn + gfn_offset;
1308 static inline gpa_t gfn_to_gpa(gfn_t gfn)
1310 return (gpa_t)gfn << PAGE_SHIFT;
1313 static inline gfn_t gpa_to_gfn(gpa_t gpa)
1315 return (gfn_t)(gpa >> PAGE_SHIFT);
1318 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
1320 return (hpa_t)pfn << PAGE_SHIFT;
1323 static inline struct page *kvm_vcpu_gpa_to_page(struct kvm_vcpu *vcpu,
1326 return kvm_vcpu_gfn_to_page(vcpu, gpa_to_gfn(gpa));
1329 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
1331 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
1333 return kvm_is_error_hva(hva);
1336 enum kvm_stat_kind {
1341 struct kvm_stat_data {
1343 const struct _kvm_stats_desc *desc;
1344 enum kvm_stat_kind kind;
1347 struct _kvm_stats_desc {
1348 struct kvm_stats_desc desc;
1349 char name[KVM_STATS_NAME_SIZE];
1352 #define STATS_DESC_COMMON(type, unit, base, exp, sz, bsz) \
1353 .flags = type | unit | base | \
1354 BUILD_BUG_ON_ZERO(type & ~KVM_STATS_TYPE_MASK) | \
1355 BUILD_BUG_ON_ZERO(unit & ~KVM_STATS_UNIT_MASK) | \
1356 BUILD_BUG_ON_ZERO(base & ~KVM_STATS_BASE_MASK), \
1361 #define VM_GENERIC_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1364 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1365 .offset = offsetof(struct kvm_vm_stat, generic.stat) \
1369 #define VCPU_GENERIC_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1372 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1373 .offset = offsetof(struct kvm_vcpu_stat, generic.stat) \
1377 #define VM_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1380 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1381 .offset = offsetof(struct kvm_vm_stat, stat) \
1385 #define VCPU_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1388 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1389 .offset = offsetof(struct kvm_vcpu_stat, stat) \
1393 /* SCOPE: VM, VM_GENERIC, VCPU, VCPU_GENERIC */
1394 #define STATS_DESC(SCOPE, stat, type, unit, base, exp, sz, bsz) \
1395 SCOPE##_STATS_DESC(stat, type, unit, base, exp, sz, bsz)
1397 #define STATS_DESC_CUMULATIVE(SCOPE, name, unit, base, exponent) \
1398 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_CUMULATIVE, \
1399 unit, base, exponent, 1, 0)
1400 #define STATS_DESC_INSTANT(SCOPE, name, unit, base, exponent) \
1401 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_INSTANT, \
1402 unit, base, exponent, 1, 0)
1403 #define STATS_DESC_PEAK(SCOPE, name, unit, base, exponent) \
1404 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_PEAK, \
1405 unit, base, exponent, 1, 0)
1406 #define STATS_DESC_LINEAR_HIST(SCOPE, name, unit, base, exponent, sz, bsz) \
1407 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_LINEAR_HIST, \
1408 unit, base, exponent, sz, bsz)
1409 #define STATS_DESC_LOG_HIST(SCOPE, name, unit, base, exponent, sz) \
1410 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_LOG_HIST, \
1411 unit, base, exponent, sz, 0)
1413 /* Cumulative counter, read/write */
1414 #define STATS_DESC_COUNTER(SCOPE, name) \
1415 STATS_DESC_CUMULATIVE(SCOPE, name, KVM_STATS_UNIT_NONE, \
1416 KVM_STATS_BASE_POW10, 0)
1417 /* Instantaneous counter, read only */
1418 #define STATS_DESC_ICOUNTER(SCOPE, name) \
1419 STATS_DESC_INSTANT(SCOPE, name, KVM_STATS_UNIT_NONE, \
1420 KVM_STATS_BASE_POW10, 0)
1421 /* Peak counter, read/write */
1422 #define STATS_DESC_PCOUNTER(SCOPE, name) \
1423 STATS_DESC_PEAK(SCOPE, name, KVM_STATS_UNIT_NONE, \
1424 KVM_STATS_BASE_POW10, 0)
1426 /* Cumulative time in nanosecond */
1427 #define STATS_DESC_TIME_NSEC(SCOPE, name) \
1428 STATS_DESC_CUMULATIVE(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
1429 KVM_STATS_BASE_POW10, -9)
1430 /* Linear histogram for time in nanosecond */
1431 #define STATS_DESC_LINHIST_TIME_NSEC(SCOPE, name, sz, bsz) \
1432 STATS_DESC_LINEAR_HIST(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
1433 KVM_STATS_BASE_POW10, -9, sz, bsz)
1434 /* Logarithmic histogram for time in nanosecond */
1435 #define STATS_DESC_LOGHIST_TIME_NSEC(SCOPE, name, sz) \
1436 STATS_DESC_LOG_HIST(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
1437 KVM_STATS_BASE_POW10, -9, sz)
1439 #define KVM_GENERIC_VM_STATS() \
1440 STATS_DESC_COUNTER(VM_GENERIC, remote_tlb_flush), \
1441 STATS_DESC_COUNTER(VM_GENERIC, remote_tlb_flush_requests)
1443 #define KVM_GENERIC_VCPU_STATS() \
1444 STATS_DESC_COUNTER(VCPU_GENERIC, halt_successful_poll), \
1445 STATS_DESC_COUNTER(VCPU_GENERIC, halt_attempted_poll), \
1446 STATS_DESC_COUNTER(VCPU_GENERIC, halt_poll_invalid), \
1447 STATS_DESC_COUNTER(VCPU_GENERIC, halt_wakeup), \
1448 STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_poll_success_ns), \
1449 STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_poll_fail_ns), \
1450 STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_wait_ns), \
1451 STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_poll_success_hist, \
1452 HALT_POLL_HIST_COUNT), \
1453 STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_poll_fail_hist, \
1454 HALT_POLL_HIST_COUNT), \
1455 STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_wait_hist, \
1456 HALT_POLL_HIST_COUNT)
1458 extern struct dentry *kvm_debugfs_dir;
1460 ssize_t kvm_stats_read(char *id, const struct kvm_stats_header *header,
1461 const struct _kvm_stats_desc *desc,
1462 void *stats, size_t size_stats,
1463 char __user *user_buffer, size_t size, loff_t *offset);
1466 * kvm_stats_linear_hist_update() - Update bucket value for linear histogram
1469 * @data: start address of the stats data
1470 * @size: the number of bucket of the stats data
1471 * @value: the new value used to update the linear histogram's bucket
1472 * @bucket_size: the size (width) of a bucket
1474 static inline void kvm_stats_linear_hist_update(u64 *data, size_t size,
1475 u64 value, size_t bucket_size)
1477 size_t index = div64_u64(value, bucket_size);
1479 index = min(index, size - 1);
1484 * kvm_stats_log_hist_update() - Update bucket value for logarithmic histogram
1487 * @data: start address of the stats data
1488 * @size: the number of bucket of the stats data
1489 * @value: the new value used to update the logarithmic histogram's bucket
1491 static inline void kvm_stats_log_hist_update(u64 *data, size_t size, u64 value)
1493 size_t index = fls64(value);
1495 index = min(index, size - 1);
1499 #define KVM_STATS_LINEAR_HIST_UPDATE(array, value, bsize) \
1500 kvm_stats_linear_hist_update(array, ARRAY_SIZE(array), value, bsize)
1501 #define KVM_STATS_LOG_HIST_UPDATE(array, value) \
1502 kvm_stats_log_hist_update(array, ARRAY_SIZE(array), value)
1505 extern const struct kvm_stats_header kvm_vm_stats_header;
1506 extern const struct _kvm_stats_desc kvm_vm_stats_desc[];
1507 extern const struct kvm_stats_header kvm_vcpu_stats_header;
1508 extern const struct _kvm_stats_desc kvm_vcpu_stats_desc[];
1510 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1511 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1513 if (unlikely(kvm->mmu_notifier_count))
1516 * Ensure the read of mmu_notifier_count happens before the read
1517 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1518 * mmu_notifier_invalidate_range_end to make sure that the caller
1519 * either sees the old (non-zero) value of mmu_notifier_count or
1520 * the new (incremented) value of mmu_notifier_seq.
1521 * PowerPC Book3s HV KVM calls this under a per-page lock
1522 * rather than under kvm->mmu_lock, for scalability, so
1523 * can't rely on kvm->mmu_lock to keep things ordered.
1526 if (kvm->mmu_notifier_seq != mmu_seq)
1531 static inline int mmu_notifier_retry_hva(struct kvm *kvm,
1532 unsigned long mmu_seq,
1535 lockdep_assert_held(&kvm->mmu_lock);
1537 * If mmu_notifier_count is non-zero, then the range maintained by
1538 * kvm_mmu_notifier_invalidate_range_start contains all addresses that
1539 * might be being invalidated. Note that it may include some false
1540 * positives, due to shortcuts when handing concurrent invalidations.
1542 if (unlikely(kvm->mmu_notifier_count) &&
1543 hva >= kvm->mmu_notifier_range_start &&
1544 hva < kvm->mmu_notifier_range_end)
1546 if (kvm->mmu_notifier_seq != mmu_seq)
1552 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1554 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1556 bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
1557 int kvm_set_irq_routing(struct kvm *kvm,
1558 const struct kvm_irq_routing_entry *entries,
1561 int kvm_set_routing_entry(struct kvm *kvm,
1562 struct kvm_kernel_irq_routing_entry *e,
1563 const struct kvm_irq_routing_entry *ue);
1564 void kvm_free_irq_routing(struct kvm *kvm);
1568 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1572 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1574 #ifdef CONFIG_HAVE_KVM_EVENTFD
1576 void kvm_eventfd_init(struct kvm *kvm);
1577 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1579 #ifdef CONFIG_HAVE_KVM_IRQFD
1580 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1581 void kvm_irqfd_release(struct kvm *kvm);
1582 void kvm_irq_routing_update(struct kvm *);
1584 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1589 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1594 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1596 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1601 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1603 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1604 static inline void kvm_irq_routing_update(struct kvm *kvm)
1609 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1614 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1616 void kvm_arch_irq_routing_update(struct kvm *kvm);
1618 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1621 * Ensure the rest of the request is published to kvm_check_request's
1622 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1625 set_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1628 static inline bool kvm_request_pending(struct kvm_vcpu *vcpu)
1630 return READ_ONCE(vcpu->requests);
1633 static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)
1635 return test_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1638 static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu)
1640 clear_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1643 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1645 if (kvm_test_request(req, vcpu)) {
1646 kvm_clear_request(req, vcpu);
1649 * Ensure the rest of the request is visible to kvm_check_request's
1650 * caller. Paired with the smp_wmb in kvm_make_request.
1652 smp_mb__after_atomic();
1659 extern bool kvm_rebooting;
1661 extern unsigned int halt_poll_ns;
1662 extern unsigned int halt_poll_ns_grow;
1663 extern unsigned int halt_poll_ns_grow_start;
1664 extern unsigned int halt_poll_ns_shrink;
1667 const struct kvm_device_ops *ops;
1670 struct list_head vm_node;
1673 /* create, destroy, and name are mandatory */
1674 struct kvm_device_ops {
1678 * create is called holding kvm->lock and any operations not suitable
1679 * to do while holding the lock should be deferred to init (see
1682 int (*create)(struct kvm_device *dev, u32 type);
1685 * init is called after create if create is successful and is called
1686 * outside of holding kvm->lock.
1688 void (*init)(struct kvm_device *dev);
1691 * Destroy is responsible for freeing dev.
1693 * Destroy may be called before or after destructors are called
1694 * on emulated I/O regions, depending on whether a reference is
1695 * held by a vcpu or other kvm component that gets destroyed
1696 * after the emulated I/O.
1698 void (*destroy)(struct kvm_device *dev);
1701 * Release is an alternative method to free the device. It is
1702 * called when the device file descriptor is closed. Once
1703 * release is called, the destroy method will not be called
1704 * anymore as the device is removed from the device list of
1705 * the VM. kvm->lock is held.
1707 void (*release)(struct kvm_device *dev);
1709 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1710 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1711 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1712 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1714 int (*mmap)(struct kvm_device *dev, struct vm_area_struct *vma);
1717 void kvm_device_get(struct kvm_device *dev);
1718 void kvm_device_put(struct kvm_device *dev);
1719 struct kvm_device *kvm_device_from_filp(struct file *filp);
1720 int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type);
1721 void kvm_unregister_device_ops(u32 type);
1723 extern struct kvm_device_ops kvm_mpic_ops;
1724 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1725 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1727 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1729 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1731 vcpu->spin_loop.in_spin_loop = val;
1733 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1735 vcpu->spin_loop.dy_eligible = val;
1738 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1740 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1744 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1747 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1749 static inline bool kvm_is_visible_memslot(struct kvm_memory_slot *memslot)
1751 return (memslot && memslot->id < KVM_USER_MEM_SLOTS &&
1752 !(memslot->flags & KVM_MEMSLOT_INVALID));
1755 struct kvm_vcpu *kvm_get_running_vcpu(void);
1756 struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void);
1758 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1759 bool kvm_arch_has_irq_bypass(void);
1760 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1761 struct irq_bypass_producer *);
1762 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1763 struct irq_bypass_producer *);
1764 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1765 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1766 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1767 uint32_t guest_irq, bool set);
1768 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1770 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1771 /* If we wakeup during the poll time, was it a sucessful poll? */
1772 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1774 return vcpu->valid_wakeup;
1778 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1782 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
1784 #ifdef CONFIG_HAVE_KVM_NO_POLL
1785 /* Callback that tells if we must not poll */
1786 bool kvm_arch_no_poll(struct kvm_vcpu *vcpu);
1788 static inline bool kvm_arch_no_poll(struct kvm_vcpu *vcpu)
1792 #endif /* CONFIG_HAVE_KVM_NO_POLL */
1794 #ifdef CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL
1795 long kvm_arch_vcpu_async_ioctl(struct file *filp,
1796 unsigned int ioctl, unsigned long arg);
1798 static inline long kvm_arch_vcpu_async_ioctl(struct file *filp,
1802 return -ENOIOCTLCMD;
1804 #endif /* CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL */
1806 void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
1807 unsigned long start, unsigned long end);
1809 #ifdef CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE
1810 int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu);
1812 static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
1816 #endif /* CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE */
1818 typedef int (*kvm_vm_thread_fn_t)(struct kvm *kvm, uintptr_t data);
1820 int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn,
1821 uintptr_t data, const char *name,
1822 struct task_struct **thread_ptr);
1824 #ifdef CONFIG_KVM_XFER_TO_GUEST_WORK
1825 static inline void kvm_handle_signal_exit(struct kvm_vcpu *vcpu)
1827 vcpu->run->exit_reason = KVM_EXIT_INTR;
1828 vcpu->stat.signal_exits++;
1830 #endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */
1833 * This defines how many reserved entries we want to keep before we
1834 * kick the vcpu to the userspace to avoid dirty ring full. This
1835 * value can be tuned to higher if e.g. PML is enabled on the host.
1837 #define KVM_DIRTY_RING_RSVD_ENTRIES 64
1839 /* Max number of entries allowed for each kvm dirty ring */
1840 #define KVM_DIRTY_RING_MAX_ENTRIES 65536