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_IDS
43 #define KVM_MAX_VCPU_IDS 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 unsigned long *vcpu_bitmap);
164 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
165 bool kvm_make_all_cpus_request_except(struct kvm *kvm, unsigned int req,
166 struct kvm_vcpu *except);
167 bool kvm_make_cpus_request_mask(struct kvm *kvm, unsigned int req,
168 unsigned long *vcpu_bitmap);
170 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
171 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
173 extern struct mutex kvm_lock;
174 extern struct list_head vm_list;
176 struct kvm_io_range {
179 struct kvm_io_device *dev;
182 #define NR_IOBUS_DEVS 1000
187 struct kvm_io_range range[];
193 KVM_VIRTIO_CCW_NOTIFY_BUS,
198 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
199 int len, const void *val);
200 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
201 gpa_t addr, int len, const void *val, long cookie);
202 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
204 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
205 int len, struct kvm_io_device *dev);
206 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
207 struct kvm_io_device *dev);
208 struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx,
211 #ifdef CONFIG_KVM_ASYNC_PF
212 struct kvm_async_pf {
213 struct work_struct work;
214 struct list_head link;
215 struct list_head queue;
216 struct kvm_vcpu *vcpu;
217 struct mm_struct *mm;
220 struct kvm_arch_async_pf arch;
222 bool notpresent_injected;
225 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
226 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
227 bool kvm_setup_async_pf(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
228 unsigned long hva, struct kvm_arch_async_pf *arch);
229 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
232 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
233 struct kvm_gfn_range {
234 struct kvm_memory_slot *slot;
240 bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range);
241 bool kvm_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
242 bool kvm_test_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
243 bool kvm_set_spte_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
250 READING_SHADOW_PAGE_TABLES,
253 #define KVM_UNMAPPED_PAGE ((void *) 0x500 + POISON_POINTER_DELTA)
255 struct kvm_host_map {
257 * Only valid if the 'pfn' is managed by the host kernel (i.e. There is
258 * a 'struct page' for it. When using mem= kernel parameter some memory
259 * can be used as guest memory but they are not managed by host
261 * If 'pfn' is not managed by the host kernel, this field is
262 * initialized to KVM_UNMAPPED_PAGE.
271 * Used to check if the mapping is valid or not. Never use 'kvm_host_map'
272 * directly to check for that.
274 static inline bool kvm_vcpu_mapped(struct kvm_host_map *map)
279 static inline bool kvm_vcpu_can_poll(ktime_t cur, ktime_t stop)
281 return single_task_running() && !need_resched() && ktime_before(cur, stop);
285 * Sometimes a large or cross-page mmio needs to be broken up into separate
286 * exits for userspace servicing.
288 struct kvm_mmio_fragment {
296 #ifdef CONFIG_PREEMPT_NOTIFIERS
297 struct preempt_notifier preempt_notifier;
300 int vcpu_id; /* id given by userspace at creation */
301 int vcpu_idx; /* index in kvm->vcpus array */
305 unsigned long guest_debug;
308 struct list_head blocked_vcpu_list;
314 struct pid __rcu *pid;
317 unsigned int halt_poll_ns;
320 #ifdef CONFIG_HAS_IOMEM
322 int mmio_read_completed;
324 int mmio_cur_fragment;
325 int mmio_nr_fragments;
326 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
329 #ifdef CONFIG_KVM_ASYNC_PF
332 struct list_head queue;
333 struct list_head done;
338 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
340 * Cpu relax intercept or pause loop exit optimization
341 * in_spin_loop: set when a vcpu does a pause loop exit
342 * or cpu relax intercepted.
343 * dy_eligible: indicates whether vcpu is eligible for directed yield.
352 struct kvm_vcpu_arch arch;
353 struct kvm_vcpu_stat stat;
354 char stats_id[KVM_STATS_NAME_SIZE];
355 struct kvm_dirty_ring dirty_ring;
358 * The index of the most recently used memslot by this vCPU. It's ok
359 * if this becomes stale due to memslot changes since we always check
360 * it is a valid slot.
365 /* must be called with irqs disabled */
366 static __always_inline void guest_enter_irqoff(void)
369 * This is running in ioctl context so its safe to assume that it's the
370 * stime pending cputime to flush.
372 instrumentation_begin();
373 vtime_account_guest_enter();
374 instrumentation_end();
377 * KVM does not hold any references to rcu protected data when it
378 * switches CPU into a guest mode. In fact switching to a guest mode
379 * is very similar to exiting to userspace from rcu point of view. In
380 * addition CPU may stay in a guest mode for quite a long time (up to
381 * one time slice). Lets treat guest mode as quiescent state, just like
382 * we do with user-mode execution.
384 if (!context_tracking_guest_enter()) {
385 instrumentation_begin();
386 rcu_virt_note_context_switch(smp_processor_id());
387 instrumentation_end();
391 static __always_inline void guest_exit_irqoff(void)
393 context_tracking_guest_exit();
395 instrumentation_begin();
396 /* Flush the guest cputime we spent on the guest */
397 vtime_account_guest_exit();
398 instrumentation_end();
401 static inline void guest_exit(void)
405 local_irq_save(flags);
407 local_irq_restore(flags);
410 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
413 * The memory barrier ensures a previous write to vcpu->requests cannot
414 * be reordered with the read of vcpu->mode. It pairs with the general
415 * memory barrier following the write of vcpu->mode in VCPU RUN.
417 smp_mb__before_atomic();
418 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
422 * Some of the bitops functions do not support too long bitmaps.
423 * This number must be determined not to exceed such limits.
425 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
427 struct kvm_memory_slot {
429 unsigned long npages;
430 unsigned long *dirty_bitmap;
431 struct kvm_arch_memory_slot arch;
432 unsigned long userspace_addr;
438 static inline bool kvm_slot_dirty_track_enabled(struct kvm_memory_slot *slot)
440 return slot->flags & KVM_MEM_LOG_DIRTY_PAGES;
443 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
445 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
448 static inline unsigned long *kvm_second_dirty_bitmap(struct kvm_memory_slot *memslot)
450 unsigned long len = kvm_dirty_bitmap_bytes(memslot);
452 return memslot->dirty_bitmap + len / sizeof(*memslot->dirty_bitmap);
455 #ifndef KVM_DIRTY_LOG_MANUAL_CAPS
456 #define KVM_DIRTY_LOG_MANUAL_CAPS KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE
459 struct kvm_s390_adapter_int {
472 struct kvm_kernel_irq_routing_entry {
475 int (*set)(struct kvm_kernel_irq_routing_entry *e,
476 struct kvm *kvm, int irq_source_id, int level,
490 struct kvm_s390_adapter_int adapter;
491 struct kvm_hv_sint hv_sint;
493 struct hlist_node link;
496 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
497 struct kvm_irq_routing_table {
498 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
501 * Array indexed by gsi. Each entry contains list of irq chips
502 * the gsi is connected to.
504 struct hlist_head map[];
508 #ifndef KVM_PRIVATE_MEM_SLOTS
509 #define KVM_PRIVATE_MEM_SLOTS 0
512 #define KVM_MEM_SLOTS_NUM SHRT_MAX
513 #define KVM_USER_MEM_SLOTS (KVM_MEM_SLOTS_NUM - KVM_PRIVATE_MEM_SLOTS)
515 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
516 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
524 * memslots are not sorted by id anymore, please use id_to_memslot()
525 * to get the memslot by its id.
527 struct kvm_memslots {
529 /* The mapping table from slot id to the index in memslots[]. */
530 short id_to_index[KVM_MEM_SLOTS_NUM];
531 atomic_t last_used_slot;
533 struct kvm_memory_slot memslots[];
537 #ifdef KVM_HAVE_MMU_RWLOCK
541 #endif /* KVM_HAVE_MMU_RWLOCK */
543 struct mutex slots_lock;
546 * Protects the arch-specific fields of struct kvm_memory_slots in
547 * use by the VM. To be used under the slots_lock (above) or in a
548 * kvm->srcu critical section where acquiring the slots_lock would
549 * lead to deadlock with the synchronize_srcu in
550 * install_new_memslots.
552 struct mutex slots_arch_lock;
553 struct mm_struct *mm; /* userspace tied to this vm */
554 struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
555 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
557 /* Used to wait for completion of MMU notifiers. */
558 spinlock_t mn_invalidate_lock;
559 unsigned long mn_active_invalidate_count;
560 struct rcuwait mn_memslots_update_rcuwait;
563 * created_vcpus is protected by kvm->lock, and is incremented
564 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
565 * incremented after storing the kvm_vcpu pointer in vcpus,
566 * and is accessed atomically.
568 atomic_t online_vcpus;
570 int last_boosted_vcpu;
571 struct list_head vm_list;
573 struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
574 #ifdef CONFIG_HAVE_KVM_EVENTFD
577 struct list_head items;
578 struct list_head resampler_list;
579 struct mutex resampler_lock;
581 struct list_head ioeventfds;
583 struct kvm_vm_stat stat;
584 struct kvm_arch arch;
585 refcount_t users_count;
586 #ifdef CONFIG_KVM_MMIO
587 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
588 spinlock_t ring_lock;
589 struct list_head coalesced_zones;
592 struct mutex irq_lock;
593 #ifdef CONFIG_HAVE_KVM_IRQCHIP
595 * Update side is protected by irq_lock.
597 struct kvm_irq_routing_table __rcu *irq_routing;
599 #ifdef CONFIG_HAVE_KVM_IRQFD
600 struct hlist_head irq_ack_notifier_list;
603 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
604 struct mmu_notifier mmu_notifier;
605 unsigned long mmu_notifier_seq;
606 long mmu_notifier_count;
607 unsigned long mmu_notifier_range_start;
608 unsigned long mmu_notifier_range_end;
610 struct list_head devices;
611 u64 manual_dirty_log_protect;
612 struct dentry *debugfs_dentry;
613 struct kvm_stat_data **debugfs_stat_data;
614 struct srcu_struct srcu;
615 struct srcu_struct irq_srcu;
617 unsigned int max_halt_poll_ns;
621 #ifdef CONFIG_HAVE_KVM_PM_NOTIFIER
622 struct notifier_block pm_notifier;
624 char stats_id[KVM_STATS_NAME_SIZE];
627 #define kvm_err(fmt, ...) \
628 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
629 #define kvm_info(fmt, ...) \
630 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
631 #define kvm_debug(fmt, ...) \
632 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
633 #define kvm_debug_ratelimited(fmt, ...) \
634 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
636 #define kvm_pr_unimpl(fmt, ...) \
637 pr_err_ratelimited("kvm [%i]: " fmt, \
638 task_tgid_nr(current), ## __VA_ARGS__)
640 /* The guest did something we don't support. */
641 #define vcpu_unimpl(vcpu, fmt, ...) \
642 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
643 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
645 #define vcpu_debug(vcpu, fmt, ...) \
646 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
647 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
648 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
650 #define vcpu_err(vcpu, fmt, ...) \
651 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
653 static inline void kvm_vm_bugged(struct kvm *kvm)
655 kvm->vm_bugged = true;
656 kvm_make_all_cpus_request(kvm, KVM_REQ_VM_BUGGED);
659 #define KVM_BUG(cond, kvm, fmt...) \
661 int __ret = (cond); \
663 if (WARN_ONCE(__ret && !(kvm)->vm_bugged, fmt)) \
664 kvm_vm_bugged(kvm); \
668 #define KVM_BUG_ON(cond, kvm) \
670 int __ret = (cond); \
672 if (WARN_ON_ONCE(__ret && !(kvm)->vm_bugged)) \
673 kvm_vm_bugged(kvm); \
677 static inline bool kvm_dirty_log_manual_protect_and_init_set(struct kvm *kvm)
679 return !!(kvm->manual_dirty_log_protect & KVM_DIRTY_LOG_INITIALLY_SET);
682 static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx)
684 return srcu_dereference_check(kvm->buses[idx], &kvm->srcu,
685 lockdep_is_held(&kvm->slots_lock) ||
686 !refcount_read(&kvm->users_count));
689 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
691 int num_vcpus = atomic_read(&kvm->online_vcpus);
692 i = array_index_nospec(i, num_vcpus);
694 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. */
696 return kvm->vcpus[i];
699 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
701 idx < atomic_read(&kvm->online_vcpus) && \
702 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
705 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
707 struct kvm_vcpu *vcpu = NULL;
712 if (id < KVM_MAX_VCPUS)
713 vcpu = kvm_get_vcpu(kvm, id);
714 if (vcpu && vcpu->vcpu_id == id)
716 kvm_for_each_vcpu(i, vcpu, kvm)
717 if (vcpu->vcpu_id == id)
722 #define kvm_for_each_memslot(memslot, slots) \
723 for (memslot = &slots->memslots[0]; \
724 memslot < slots->memslots + slots->used_slots; memslot++) \
725 if (WARN_ON_ONCE(!memslot->npages)) { \
728 void kvm_vcpu_destroy(struct kvm_vcpu *vcpu);
730 void vcpu_load(struct kvm_vcpu *vcpu);
731 void vcpu_put(struct kvm_vcpu *vcpu);
733 #ifdef __KVM_HAVE_IOAPIC
734 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
735 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
737 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
740 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
745 #ifdef CONFIG_HAVE_KVM_IRQFD
746 int kvm_irqfd_init(void);
747 void kvm_irqfd_exit(void);
749 static inline int kvm_irqfd_init(void)
754 static inline void kvm_irqfd_exit(void)
758 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
759 struct module *module);
762 void kvm_get_kvm(struct kvm *kvm);
763 bool kvm_get_kvm_safe(struct kvm *kvm);
764 void kvm_put_kvm(struct kvm *kvm);
765 bool file_is_kvm(struct file *file);
766 void kvm_put_kvm_no_destroy(struct kvm *kvm);
768 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
770 as_id = array_index_nospec(as_id, KVM_ADDRESS_SPACE_NUM);
771 return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu,
772 lockdep_is_held(&kvm->slots_lock) ||
773 !refcount_read(&kvm->users_count));
776 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
778 return __kvm_memslots(kvm, 0);
781 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
783 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
785 return __kvm_memslots(vcpu->kvm, as_id);
789 struct kvm_memory_slot *id_to_memslot(struct kvm_memslots *slots, int id)
791 int index = slots->id_to_index[id];
792 struct kvm_memory_slot *slot;
797 slot = &slots->memslots[index];
799 WARN_ON(slot->id != id);
804 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
805 * - create a new memory slot
806 * - delete an existing memory slot
807 * - modify an existing memory slot
808 * -- move it in the guest physical memory space
809 * -- just change its flags
811 * Since flags can be changed by some of these operations, the following
812 * differentiation is the best we can do for __kvm_set_memory_region():
821 int kvm_set_memory_region(struct kvm *kvm,
822 const struct kvm_userspace_memory_region *mem);
823 int __kvm_set_memory_region(struct kvm *kvm,
824 const struct kvm_userspace_memory_region *mem);
825 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot);
826 void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen);
827 int kvm_arch_prepare_memory_region(struct kvm *kvm,
828 struct kvm_memory_slot *memslot,
829 const struct kvm_userspace_memory_region *mem,
830 enum kvm_mr_change change);
831 void kvm_arch_commit_memory_region(struct kvm *kvm,
832 const struct kvm_userspace_memory_region *mem,
833 struct kvm_memory_slot *old,
834 const struct kvm_memory_slot *new,
835 enum kvm_mr_change change);
836 /* flush all memory translations */
837 void kvm_arch_flush_shadow_all(struct kvm *kvm);
838 /* flush memory translations pointing to 'slot' */
839 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
840 struct kvm_memory_slot *slot);
842 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
843 struct page **pages, int nr_pages);
845 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
846 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
847 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
848 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
849 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
851 void kvm_release_page_clean(struct page *page);
852 void kvm_release_page_dirty(struct page *page);
853 void kvm_set_page_accessed(struct page *page);
855 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
856 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
858 kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
859 kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
860 kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
861 bool atomic, bool *async, bool write_fault,
862 bool *writable, hva_t *hva);
864 void kvm_release_pfn_clean(kvm_pfn_t pfn);
865 void kvm_release_pfn_dirty(kvm_pfn_t pfn);
866 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
867 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
869 void kvm_release_pfn(kvm_pfn_t pfn, bool dirty, struct gfn_to_pfn_cache *cache);
870 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
872 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
873 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
874 void *data, unsigned long len);
875 int kvm_read_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
876 void *data, unsigned int offset,
878 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
879 int offset, int len);
880 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
882 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
883 void *data, unsigned long len);
884 int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
885 void *data, unsigned int offset,
887 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
888 gpa_t gpa, unsigned long len);
890 #define __kvm_get_guest(kvm, gfn, offset, v) \
892 unsigned long __addr = gfn_to_hva(kvm, gfn); \
893 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
894 int __ret = -EFAULT; \
896 if (!kvm_is_error_hva(__addr)) \
897 __ret = get_user(v, __uaddr); \
901 #define kvm_get_guest(kvm, gpa, v) \
904 struct kvm *__kvm = kvm; \
906 __kvm_get_guest(__kvm, __gpa >> PAGE_SHIFT, \
907 offset_in_page(__gpa), v); \
910 #define __kvm_put_guest(kvm, gfn, offset, v) \
912 unsigned long __addr = gfn_to_hva(kvm, gfn); \
913 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
914 int __ret = -EFAULT; \
916 if (!kvm_is_error_hva(__addr)) \
917 __ret = put_user(v, __uaddr); \
919 mark_page_dirty(kvm, gfn); \
923 #define kvm_put_guest(kvm, gpa, v) \
926 struct kvm *__kvm = kvm; \
928 __kvm_put_guest(__kvm, __gpa >> PAGE_SHIFT, \
929 offset_in_page(__gpa), v); \
932 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
933 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
934 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
935 bool kvm_vcpu_is_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
936 unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn);
937 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot, gfn_t gfn);
938 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
940 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
941 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
942 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
943 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
944 int kvm_vcpu_map(struct kvm_vcpu *vcpu, gpa_t gpa, struct kvm_host_map *map);
945 int kvm_map_gfn(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map,
946 struct gfn_to_pfn_cache *cache, bool atomic);
947 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
948 void kvm_vcpu_unmap(struct kvm_vcpu *vcpu, struct kvm_host_map *map, bool dirty);
949 int kvm_unmap_gfn(struct kvm_vcpu *vcpu, struct kvm_host_map *map,
950 struct gfn_to_pfn_cache *cache, bool dirty, bool atomic);
951 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
952 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
953 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
955 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
957 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
959 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
960 int offset, int len);
961 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
963 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
965 void kvm_sigset_activate(struct kvm_vcpu *vcpu);
966 void kvm_sigset_deactivate(struct kvm_vcpu *vcpu);
968 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
969 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
970 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
971 bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
972 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
973 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
974 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool usermode_vcpu_not_eligible);
976 void kvm_flush_remote_tlbs(struct kvm *kvm);
977 void kvm_reload_remote_mmus(struct kvm *kvm);
979 #ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE
980 int kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache *mc, int min);
981 int kvm_mmu_memory_cache_nr_free_objects(struct kvm_mmu_memory_cache *mc);
982 void kvm_mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc);
983 void *kvm_mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc);
986 void kvm_inc_notifier_count(struct kvm *kvm, unsigned long start,
988 void kvm_dec_notifier_count(struct kvm *kvm, unsigned long start,
991 long kvm_arch_dev_ioctl(struct file *filp,
992 unsigned int ioctl, unsigned long arg);
993 long kvm_arch_vcpu_ioctl(struct file *filp,
994 unsigned int ioctl, unsigned long arg);
995 vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
997 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
999 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
1000 struct kvm_memory_slot *slot,
1002 unsigned long mask);
1003 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot);
1005 #ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
1006 void kvm_arch_flush_remote_tlbs_memslot(struct kvm *kvm,
1007 const struct kvm_memory_slot *memslot);
1008 #else /* !CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
1009 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log);
1010 int kvm_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log,
1011 int *is_dirty, struct kvm_memory_slot **memslot);
1014 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
1016 int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
1017 struct kvm_enable_cap *cap);
1018 long kvm_arch_vm_ioctl(struct file *filp,
1019 unsigned int ioctl, unsigned long arg);
1021 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
1022 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
1024 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1025 struct kvm_translation *tr);
1027 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
1028 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
1029 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1030 struct kvm_sregs *sregs);
1031 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1032 struct kvm_sregs *sregs);
1033 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
1034 struct kvm_mp_state *mp_state);
1035 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
1036 struct kvm_mp_state *mp_state);
1037 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
1038 struct kvm_guest_debug *dbg);
1039 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu);
1041 int kvm_arch_init(void *opaque);
1042 void kvm_arch_exit(void);
1044 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
1046 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
1047 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
1048 int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id);
1049 int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu);
1050 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
1051 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
1053 #ifdef CONFIG_HAVE_KVM_PM_NOTIFIER
1054 int kvm_arch_pm_notifier(struct kvm *kvm, unsigned long state);
1057 #ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
1058 void kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu, struct dentry *debugfs_dentry);
1061 int kvm_arch_hardware_enable(void);
1062 void kvm_arch_hardware_disable(void);
1063 int kvm_arch_hardware_setup(void *opaque);
1064 void kvm_arch_hardware_unsetup(void);
1065 int kvm_arch_check_processor_compat(void *opaque);
1066 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
1067 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
1068 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
1069 bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu);
1070 bool kvm_arch_dy_has_pending_interrupt(struct kvm_vcpu *vcpu);
1071 int kvm_arch_post_init_vm(struct kvm *kvm);
1072 void kvm_arch_pre_destroy_vm(struct kvm *kvm);
1073 int kvm_arch_create_vm_debugfs(struct kvm *kvm);
1075 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
1077 * All architectures that want to use vzalloc currently also
1078 * need their own kvm_arch_alloc_vm implementation.
1080 static inline struct kvm *kvm_arch_alloc_vm(void)
1082 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
1086 static inline void __kvm_arch_free_vm(struct kvm *kvm)
1091 #ifndef __KVM_HAVE_ARCH_VM_FREE
1092 static inline void kvm_arch_free_vm(struct kvm *kvm)
1094 __kvm_arch_free_vm(kvm);
1098 #ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1099 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1105 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
1106 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
1107 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
1108 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
1110 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
1114 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
1118 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
1123 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
1124 void kvm_arch_start_assignment(struct kvm *kvm);
1125 void kvm_arch_end_assignment(struct kvm *kvm);
1126 bool kvm_arch_has_assigned_device(struct kvm *kvm);
1128 static inline void kvm_arch_start_assignment(struct kvm *kvm)
1132 static inline void kvm_arch_end_assignment(struct kvm *kvm)
1136 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
1142 static inline struct rcuwait *kvm_arch_vcpu_get_wait(struct kvm_vcpu *vcpu)
1144 #ifdef __KVM_HAVE_ARCH_WQP
1145 return vcpu->arch.waitp;
1151 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
1153 * returns true if the virtual interrupt controller is initialized and
1154 * ready to accept virtual IRQ. On some architectures the virtual interrupt
1155 * controller is dynamically instantiated and this is not always true.
1157 bool kvm_arch_intc_initialized(struct kvm *kvm);
1159 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
1165 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
1166 void kvm_arch_destroy_vm(struct kvm *kvm);
1167 void kvm_arch_sync_events(struct kvm *kvm);
1169 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
1171 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
1172 bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
1173 bool kvm_is_transparent_hugepage(kvm_pfn_t pfn);
1175 struct kvm_irq_ack_notifier {
1176 struct hlist_node link;
1178 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
1181 int kvm_irq_map_gsi(struct kvm *kvm,
1182 struct kvm_kernel_irq_routing_entry *entries, int gsi);
1183 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
1185 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
1187 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
1188 int irq_source_id, int level, bool line_status);
1189 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
1190 struct kvm *kvm, int irq_source_id,
1191 int level, bool line_status);
1192 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
1193 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
1194 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
1195 void kvm_register_irq_ack_notifier(struct kvm *kvm,
1196 struct kvm_irq_ack_notifier *kian);
1197 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
1198 struct kvm_irq_ack_notifier *kian);
1199 int kvm_request_irq_source_id(struct kvm *kvm);
1200 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
1201 bool kvm_arch_irqfd_allowed(struct kvm *kvm, struct kvm_irqfd *args);
1204 * Returns a pointer to the memslot at slot_index if it contains gfn.
1205 * Otherwise returns NULL.
1207 static inline struct kvm_memory_slot *
1208 try_get_memslot(struct kvm_memslots *slots, int slot_index, gfn_t gfn)
1210 struct kvm_memory_slot *slot;
1212 if (slot_index < 0 || slot_index >= slots->used_slots)
1216 * slot_index can come from vcpu->last_used_slot which is not kept
1217 * in sync with userspace-controllable memslot deletion. So use nospec
1218 * to prevent the CPU from speculating past the end of memslots[].
1220 slot_index = array_index_nospec(slot_index, slots->used_slots);
1221 slot = &slots->memslots[slot_index];
1223 if (gfn >= slot->base_gfn && gfn < slot->base_gfn + slot->npages)
1230 * Returns a pointer to the memslot that contains gfn and records the index of
1231 * the slot in index. Otherwise returns NULL.
1233 * IMPORTANT: Slots are sorted from highest GFN to lowest GFN!
1235 static inline struct kvm_memory_slot *
1236 search_memslots(struct kvm_memslots *slots, gfn_t gfn, int *index)
1238 int start = 0, end = slots->used_slots;
1239 struct kvm_memory_slot *memslots = slots->memslots;
1240 struct kvm_memory_slot *slot;
1242 if (unlikely(!slots->used_slots))
1245 while (start < end) {
1246 int slot = start + (end - start) / 2;
1248 if (gfn >= memslots[slot].base_gfn)
1254 slot = try_get_memslot(slots, start, gfn);
1264 * __gfn_to_memslot() and its descendants are here because it is called from
1265 * non-modular code in arch/powerpc/kvm/book3s_64_vio{,_hv}.c. gfn_to_memslot()
1266 * itself isn't here as an inline because that would bloat other code too much.
1268 static inline struct kvm_memory_slot *
1269 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
1271 struct kvm_memory_slot *slot;
1272 int slot_index = atomic_read(&slots->last_used_slot);
1274 slot = try_get_memslot(slots, slot_index, gfn);
1278 slot = search_memslots(slots, gfn, &slot_index);
1280 atomic_set(&slots->last_used_slot, slot_index);
1287 static inline unsigned long
1288 __gfn_to_hva_memslot(const struct kvm_memory_slot *slot, gfn_t gfn)
1291 * The index was checked originally in search_memslots. To avoid
1292 * that a malicious guest builds a Spectre gadget out of e.g. page
1293 * table walks, do not let the processor speculate loads outside
1294 * the guest's registered memslots.
1296 unsigned long offset = gfn - slot->base_gfn;
1297 offset = array_index_nospec(offset, slot->npages);
1298 return slot->userspace_addr + offset * PAGE_SIZE;
1301 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
1303 return gfn_to_memslot(kvm, gfn)->id;
1307 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
1309 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
1311 return slot->base_gfn + gfn_offset;
1314 static inline gpa_t gfn_to_gpa(gfn_t gfn)
1316 return (gpa_t)gfn << PAGE_SHIFT;
1319 static inline gfn_t gpa_to_gfn(gpa_t gpa)
1321 return (gfn_t)(gpa >> PAGE_SHIFT);
1324 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
1326 return (hpa_t)pfn << PAGE_SHIFT;
1329 static inline struct page *kvm_vcpu_gpa_to_page(struct kvm_vcpu *vcpu,
1332 return kvm_vcpu_gfn_to_page(vcpu, gpa_to_gfn(gpa));
1335 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
1337 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
1339 return kvm_is_error_hva(hva);
1342 enum kvm_stat_kind {
1347 struct kvm_stat_data {
1349 const struct _kvm_stats_desc *desc;
1350 enum kvm_stat_kind kind;
1353 struct _kvm_stats_desc {
1354 struct kvm_stats_desc desc;
1355 char name[KVM_STATS_NAME_SIZE];
1358 #define STATS_DESC_COMMON(type, unit, base, exp, sz, bsz) \
1359 .flags = type | unit | base | \
1360 BUILD_BUG_ON_ZERO(type & ~KVM_STATS_TYPE_MASK) | \
1361 BUILD_BUG_ON_ZERO(unit & ~KVM_STATS_UNIT_MASK) | \
1362 BUILD_BUG_ON_ZERO(base & ~KVM_STATS_BASE_MASK), \
1367 #define VM_GENERIC_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1370 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1371 .offset = offsetof(struct kvm_vm_stat, generic.stat) \
1375 #define VCPU_GENERIC_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1378 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1379 .offset = offsetof(struct kvm_vcpu_stat, generic.stat) \
1383 #define VM_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1386 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1387 .offset = offsetof(struct kvm_vm_stat, stat) \
1391 #define VCPU_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1394 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1395 .offset = offsetof(struct kvm_vcpu_stat, stat) \
1399 /* SCOPE: VM, VM_GENERIC, VCPU, VCPU_GENERIC */
1400 #define STATS_DESC(SCOPE, stat, type, unit, base, exp, sz, bsz) \
1401 SCOPE##_STATS_DESC(stat, type, unit, base, exp, sz, bsz)
1403 #define STATS_DESC_CUMULATIVE(SCOPE, name, unit, base, exponent) \
1404 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_CUMULATIVE, \
1405 unit, base, exponent, 1, 0)
1406 #define STATS_DESC_INSTANT(SCOPE, name, unit, base, exponent) \
1407 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_INSTANT, \
1408 unit, base, exponent, 1, 0)
1409 #define STATS_DESC_PEAK(SCOPE, name, unit, base, exponent) \
1410 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_PEAK, \
1411 unit, base, exponent, 1, 0)
1412 #define STATS_DESC_LINEAR_HIST(SCOPE, name, unit, base, exponent, sz, bsz) \
1413 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_LINEAR_HIST, \
1414 unit, base, exponent, sz, bsz)
1415 #define STATS_DESC_LOG_HIST(SCOPE, name, unit, base, exponent, sz) \
1416 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_LOG_HIST, \
1417 unit, base, exponent, sz, 0)
1419 /* Cumulative counter, read/write */
1420 #define STATS_DESC_COUNTER(SCOPE, name) \
1421 STATS_DESC_CUMULATIVE(SCOPE, name, KVM_STATS_UNIT_NONE, \
1422 KVM_STATS_BASE_POW10, 0)
1423 /* Instantaneous counter, read only */
1424 #define STATS_DESC_ICOUNTER(SCOPE, name) \
1425 STATS_DESC_INSTANT(SCOPE, name, KVM_STATS_UNIT_NONE, \
1426 KVM_STATS_BASE_POW10, 0)
1427 /* Peak counter, read/write */
1428 #define STATS_DESC_PCOUNTER(SCOPE, name) \
1429 STATS_DESC_PEAK(SCOPE, name, KVM_STATS_UNIT_NONE, \
1430 KVM_STATS_BASE_POW10, 0)
1432 /* Cumulative time in nanosecond */
1433 #define STATS_DESC_TIME_NSEC(SCOPE, name) \
1434 STATS_DESC_CUMULATIVE(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
1435 KVM_STATS_BASE_POW10, -9)
1436 /* Linear histogram for time in nanosecond */
1437 #define STATS_DESC_LINHIST_TIME_NSEC(SCOPE, name, sz, bsz) \
1438 STATS_DESC_LINEAR_HIST(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
1439 KVM_STATS_BASE_POW10, -9, sz, bsz)
1440 /* Logarithmic histogram for time in nanosecond */
1441 #define STATS_DESC_LOGHIST_TIME_NSEC(SCOPE, name, sz) \
1442 STATS_DESC_LOG_HIST(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
1443 KVM_STATS_BASE_POW10, -9, sz)
1445 #define KVM_GENERIC_VM_STATS() \
1446 STATS_DESC_COUNTER(VM_GENERIC, remote_tlb_flush), \
1447 STATS_DESC_COUNTER(VM_GENERIC, remote_tlb_flush_requests)
1449 #define KVM_GENERIC_VCPU_STATS() \
1450 STATS_DESC_COUNTER(VCPU_GENERIC, halt_successful_poll), \
1451 STATS_DESC_COUNTER(VCPU_GENERIC, halt_attempted_poll), \
1452 STATS_DESC_COUNTER(VCPU_GENERIC, halt_poll_invalid), \
1453 STATS_DESC_COUNTER(VCPU_GENERIC, halt_wakeup), \
1454 STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_poll_success_ns), \
1455 STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_poll_fail_ns), \
1456 STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_wait_ns), \
1457 STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_poll_success_hist, \
1458 HALT_POLL_HIST_COUNT), \
1459 STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_poll_fail_hist, \
1460 HALT_POLL_HIST_COUNT), \
1461 STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_wait_hist, \
1462 HALT_POLL_HIST_COUNT)
1464 extern struct dentry *kvm_debugfs_dir;
1466 ssize_t kvm_stats_read(char *id, const struct kvm_stats_header *header,
1467 const struct _kvm_stats_desc *desc,
1468 void *stats, size_t size_stats,
1469 char __user *user_buffer, size_t size, loff_t *offset);
1472 * kvm_stats_linear_hist_update() - Update bucket value for linear histogram
1475 * @data: start address of the stats data
1476 * @size: the number of bucket of the stats data
1477 * @value: the new value used to update the linear histogram's bucket
1478 * @bucket_size: the size (width) of a bucket
1480 static inline void kvm_stats_linear_hist_update(u64 *data, size_t size,
1481 u64 value, size_t bucket_size)
1483 size_t index = div64_u64(value, bucket_size);
1485 index = min(index, size - 1);
1490 * kvm_stats_log_hist_update() - Update bucket value for logarithmic histogram
1493 * @data: start address of the stats data
1494 * @size: the number of bucket of the stats data
1495 * @value: the new value used to update the logarithmic histogram's bucket
1497 static inline void kvm_stats_log_hist_update(u64 *data, size_t size, u64 value)
1499 size_t index = fls64(value);
1501 index = min(index, size - 1);
1505 #define KVM_STATS_LINEAR_HIST_UPDATE(array, value, bsize) \
1506 kvm_stats_linear_hist_update(array, ARRAY_SIZE(array), value, bsize)
1507 #define KVM_STATS_LOG_HIST_UPDATE(array, value) \
1508 kvm_stats_log_hist_update(array, ARRAY_SIZE(array), value)
1511 extern const struct kvm_stats_header kvm_vm_stats_header;
1512 extern const struct _kvm_stats_desc kvm_vm_stats_desc[];
1513 extern const struct kvm_stats_header kvm_vcpu_stats_header;
1514 extern const struct _kvm_stats_desc kvm_vcpu_stats_desc[];
1516 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1517 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1519 if (unlikely(kvm->mmu_notifier_count))
1522 * Ensure the read of mmu_notifier_count happens before the read
1523 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1524 * mmu_notifier_invalidate_range_end to make sure that the caller
1525 * either sees the old (non-zero) value of mmu_notifier_count or
1526 * the new (incremented) value of mmu_notifier_seq.
1527 * PowerPC Book3s HV KVM calls this under a per-page lock
1528 * rather than under kvm->mmu_lock, for scalability, so
1529 * can't rely on kvm->mmu_lock to keep things ordered.
1532 if (kvm->mmu_notifier_seq != mmu_seq)
1537 static inline int mmu_notifier_retry_hva(struct kvm *kvm,
1538 unsigned long mmu_seq,
1541 lockdep_assert_held(&kvm->mmu_lock);
1543 * If mmu_notifier_count is non-zero, then the range maintained by
1544 * kvm_mmu_notifier_invalidate_range_start contains all addresses that
1545 * might be being invalidated. Note that it may include some false
1546 * positives, due to shortcuts when handing concurrent invalidations.
1548 if (unlikely(kvm->mmu_notifier_count) &&
1549 hva >= kvm->mmu_notifier_range_start &&
1550 hva < kvm->mmu_notifier_range_end)
1552 if (kvm->mmu_notifier_seq != mmu_seq)
1558 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1560 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1562 bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
1563 int kvm_set_irq_routing(struct kvm *kvm,
1564 const struct kvm_irq_routing_entry *entries,
1567 int kvm_set_routing_entry(struct kvm *kvm,
1568 struct kvm_kernel_irq_routing_entry *e,
1569 const struct kvm_irq_routing_entry *ue);
1570 void kvm_free_irq_routing(struct kvm *kvm);
1574 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1578 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1580 #ifdef CONFIG_HAVE_KVM_EVENTFD
1582 void kvm_eventfd_init(struct kvm *kvm);
1583 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1585 #ifdef CONFIG_HAVE_KVM_IRQFD
1586 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1587 void kvm_irqfd_release(struct kvm *kvm);
1588 void kvm_irq_routing_update(struct kvm *);
1590 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1595 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1600 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1602 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1607 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1609 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1610 static inline void kvm_irq_routing_update(struct kvm *kvm)
1615 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1620 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1622 void kvm_arch_irq_routing_update(struct kvm *kvm);
1624 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1627 * Ensure the rest of the request is published to kvm_check_request's
1628 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1631 set_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1634 static inline bool kvm_request_pending(struct kvm_vcpu *vcpu)
1636 return READ_ONCE(vcpu->requests);
1639 static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)
1641 return test_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1644 static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu)
1646 clear_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1649 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1651 if (kvm_test_request(req, vcpu)) {
1652 kvm_clear_request(req, vcpu);
1655 * Ensure the rest of the request is visible to kvm_check_request's
1656 * caller. Paired with the smp_wmb in kvm_make_request.
1658 smp_mb__after_atomic();
1665 extern bool kvm_rebooting;
1667 extern unsigned int halt_poll_ns;
1668 extern unsigned int halt_poll_ns_grow;
1669 extern unsigned int halt_poll_ns_grow_start;
1670 extern unsigned int halt_poll_ns_shrink;
1673 const struct kvm_device_ops *ops;
1676 struct list_head vm_node;
1679 /* create, destroy, and name are mandatory */
1680 struct kvm_device_ops {
1684 * create is called holding kvm->lock and any operations not suitable
1685 * to do while holding the lock should be deferred to init (see
1688 int (*create)(struct kvm_device *dev, u32 type);
1691 * init is called after create if create is successful and is called
1692 * outside of holding kvm->lock.
1694 void (*init)(struct kvm_device *dev);
1697 * Destroy is responsible for freeing dev.
1699 * Destroy may be called before or after destructors are called
1700 * on emulated I/O regions, depending on whether a reference is
1701 * held by a vcpu or other kvm component that gets destroyed
1702 * after the emulated I/O.
1704 void (*destroy)(struct kvm_device *dev);
1707 * Release is an alternative method to free the device. It is
1708 * called when the device file descriptor is closed. Once
1709 * release is called, the destroy method will not be called
1710 * anymore as the device is removed from the device list of
1711 * the VM. kvm->lock is held.
1713 void (*release)(struct kvm_device *dev);
1715 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1716 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1717 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1718 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1720 int (*mmap)(struct kvm_device *dev, struct vm_area_struct *vma);
1723 void kvm_device_get(struct kvm_device *dev);
1724 void kvm_device_put(struct kvm_device *dev);
1725 struct kvm_device *kvm_device_from_filp(struct file *filp);
1726 int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type);
1727 void kvm_unregister_device_ops(u32 type);
1729 extern struct kvm_device_ops kvm_mpic_ops;
1730 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1731 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1733 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1735 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1737 vcpu->spin_loop.in_spin_loop = val;
1739 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1741 vcpu->spin_loop.dy_eligible = val;
1744 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1746 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1750 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1753 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1755 static inline bool kvm_is_visible_memslot(struct kvm_memory_slot *memslot)
1757 return (memslot && memslot->id < KVM_USER_MEM_SLOTS &&
1758 !(memslot->flags & KVM_MEMSLOT_INVALID));
1761 struct kvm_vcpu *kvm_get_running_vcpu(void);
1762 struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void);
1764 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1765 bool kvm_arch_has_irq_bypass(void);
1766 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1767 struct irq_bypass_producer *);
1768 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1769 struct irq_bypass_producer *);
1770 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1771 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1772 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1773 uint32_t guest_irq, bool set);
1774 bool kvm_arch_irqfd_route_changed(struct kvm_kernel_irq_routing_entry *,
1775 struct kvm_kernel_irq_routing_entry *);
1776 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1778 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1779 /* If we wakeup during the poll time, was it a sucessful poll? */
1780 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1782 return vcpu->valid_wakeup;
1786 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1790 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
1792 #ifdef CONFIG_HAVE_KVM_NO_POLL
1793 /* Callback that tells if we must not poll */
1794 bool kvm_arch_no_poll(struct kvm_vcpu *vcpu);
1796 static inline bool kvm_arch_no_poll(struct kvm_vcpu *vcpu)
1800 #endif /* CONFIG_HAVE_KVM_NO_POLL */
1802 #ifdef CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL
1803 long kvm_arch_vcpu_async_ioctl(struct file *filp,
1804 unsigned int ioctl, unsigned long arg);
1806 static inline long kvm_arch_vcpu_async_ioctl(struct file *filp,
1810 return -ENOIOCTLCMD;
1812 #endif /* CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL */
1814 void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
1815 unsigned long start, unsigned long end);
1817 #ifdef CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE
1818 int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu);
1820 static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
1824 #endif /* CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE */
1826 typedef int (*kvm_vm_thread_fn_t)(struct kvm *kvm, uintptr_t data);
1828 int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn,
1829 uintptr_t data, const char *name,
1830 struct task_struct **thread_ptr);
1832 #ifdef CONFIG_KVM_XFER_TO_GUEST_WORK
1833 static inline void kvm_handle_signal_exit(struct kvm_vcpu *vcpu)
1835 vcpu->run->exit_reason = KVM_EXIT_INTR;
1836 vcpu->stat.signal_exits++;
1838 #endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */
1841 * This defines how many reserved entries we want to keep before we
1842 * kick the vcpu to the userspace to avoid dirty ring full. This
1843 * value can be tuned to higher if e.g. PML is enabled on the host.
1845 #define KVM_DIRTY_RING_RSVD_ENTRIES 64
1847 /* Max number of entries allowed for each kvm dirty ring */
1848 #define KVM_DIRTY_RING_MAX_ENTRIES 65536