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 <linux/ftrace.h>
33 #include <linux/hashtable.h>
34 #include <linux/instrumentation.h>
35 #include <linux/interval_tree.h>
36 #include <linux/rbtree.h>
37 #include <linux/xarray.h>
38 #include <asm/signal.h>
40 #include <linux/kvm.h>
41 #include <linux/kvm_para.h>
43 #include <linux/kvm_types.h>
45 #include <asm/kvm_host.h>
46 #include <linux/kvm_dirty_ring.h>
48 #ifndef KVM_MAX_VCPU_IDS
49 #define KVM_MAX_VCPU_IDS KVM_MAX_VCPUS
53 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
54 * in kvm, other bits are visible for userspace which are defined in
55 * include/linux/kvm_h.
57 #define KVM_MEMSLOT_INVALID (1UL << 16)
60 * Bit 63 of the memslot generation number is an "update in-progress flag",
61 * e.g. is temporarily set for the duration of install_new_memslots().
62 * This flag effectively creates a unique generation number that is used to
63 * mark cached memslot data, e.g. MMIO accesses, as potentially being stale,
64 * i.e. may (or may not) have come from the previous memslots generation.
66 * This is necessary because the actual memslots update is not atomic with
67 * respect to the generation number update. Updating the generation number
68 * first would allow a vCPU to cache a spte from the old memslots using the
69 * new generation number, and updating the generation number after switching
70 * to the new memslots would allow cache hits using the old generation number
71 * to reference the defunct memslots.
73 * This mechanism is used to prevent getting hits in KVM's caches while a
74 * memslot update is in-progress, and to prevent cache hits *after* updating
75 * the actual generation number against accesses that were inserted into the
76 * cache *before* the memslots were updated.
78 #define KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS BIT_ULL(63)
80 /* Two fragments for cross MMIO pages. */
81 #define KVM_MAX_MMIO_FRAGMENTS 2
83 #ifndef KVM_ADDRESS_SPACE_NUM
84 #define KVM_ADDRESS_SPACE_NUM 1
88 * For the normal pfn, the highest 12 bits should be zero,
89 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
90 * mask bit 63 to indicate the noslot pfn.
92 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
93 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
94 #define KVM_PFN_NOSLOT (0x1ULL << 63)
96 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
97 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
98 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
101 * error pfns indicate that the gfn is in slot but faild to
102 * translate it to pfn on host.
104 static inline bool is_error_pfn(kvm_pfn_t pfn)
106 return !!(pfn & KVM_PFN_ERR_MASK);
110 * error_noslot pfns indicate that the gfn can not be
111 * translated to pfn - it is not in slot or failed to
112 * translate it to pfn.
114 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn)
116 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
119 /* noslot pfn indicates that the gfn is not in slot. */
120 static inline bool is_noslot_pfn(kvm_pfn_t pfn)
122 return pfn == KVM_PFN_NOSLOT;
126 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
127 * provide own defines and kvm_is_error_hva
129 #ifndef KVM_HVA_ERR_BAD
131 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
132 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
134 static inline bool kvm_is_error_hva(unsigned long addr)
136 return addr >= PAGE_OFFSET;
141 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
143 static inline bool is_error_page(struct page *page)
148 #define KVM_REQUEST_MASK GENMASK(7,0)
149 #define KVM_REQUEST_NO_WAKEUP BIT(8)
150 #define KVM_REQUEST_WAIT BIT(9)
152 * Architecture-independent vcpu->requests bit members
153 * Bits 4-7 are reserved for more arch-independent bits.
155 #define KVM_REQ_TLB_FLUSH (0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
156 #define KVM_REQ_VM_DEAD (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
157 #define KVM_REQ_UNBLOCK 2
158 #define KVM_REQ_UNHALT 3
159 #define KVM_REQ_GPC_INVALIDATE (5 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
160 #define KVM_REQUEST_ARCH_BASE 8
162 #define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
163 BUILD_BUG_ON((unsigned)(nr) >= (sizeof_field(struct kvm_vcpu, requests) * 8) - KVM_REQUEST_ARCH_BASE); \
164 (unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) | (flags)); \
166 #define KVM_ARCH_REQ(nr) KVM_ARCH_REQ_FLAGS(nr, 0)
168 bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
169 unsigned long *vcpu_bitmap);
170 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
171 bool kvm_make_all_cpus_request_except(struct kvm *kvm, unsigned int req,
172 struct kvm_vcpu *except);
173 bool kvm_make_cpus_request_mask(struct kvm *kvm, unsigned int req,
174 unsigned long *vcpu_bitmap);
176 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
177 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
179 extern struct mutex kvm_lock;
180 extern struct list_head vm_list;
182 struct kvm_io_range {
185 struct kvm_io_device *dev;
188 #define NR_IOBUS_DEVS 1000
193 struct kvm_io_range range[];
199 KVM_VIRTIO_CCW_NOTIFY_BUS,
204 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
205 int len, const void *val);
206 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
207 gpa_t addr, int len, const void *val, long cookie);
208 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
210 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
211 int len, struct kvm_io_device *dev);
212 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
213 struct kvm_io_device *dev);
214 struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx,
217 #ifdef CONFIG_KVM_ASYNC_PF
218 struct kvm_async_pf {
219 struct work_struct work;
220 struct list_head link;
221 struct list_head queue;
222 struct kvm_vcpu *vcpu;
223 struct mm_struct *mm;
226 struct kvm_arch_async_pf arch;
228 bool notpresent_injected;
231 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
232 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
233 bool kvm_setup_async_pf(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
234 unsigned long hva, struct kvm_arch_async_pf *arch);
235 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
238 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
239 struct kvm_gfn_range {
240 struct kvm_memory_slot *slot;
246 bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range);
247 bool kvm_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
248 bool kvm_test_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
249 bool kvm_set_spte_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
256 READING_SHADOW_PAGE_TABLES,
259 #define KVM_UNMAPPED_PAGE ((void *) 0x500 + POISON_POINTER_DELTA)
261 struct kvm_host_map {
263 * Only valid if the 'pfn' is managed by the host kernel (i.e. There is
264 * a 'struct page' for it. When using mem= kernel parameter some memory
265 * can be used as guest memory but they are not managed by host
267 * If 'pfn' is not managed by the host kernel, this field is
268 * initialized to KVM_UNMAPPED_PAGE.
277 * Used to check if the mapping is valid or not. Never use 'kvm_host_map'
278 * directly to check for that.
280 static inline bool kvm_vcpu_mapped(struct kvm_host_map *map)
285 static inline bool kvm_vcpu_can_poll(ktime_t cur, ktime_t stop)
287 return single_task_running() && !need_resched() && ktime_before(cur, stop);
291 * Sometimes a large or cross-page mmio needs to be broken up into separate
292 * exits for userspace servicing.
294 struct kvm_mmio_fragment {
302 #ifdef CONFIG_PREEMPT_NOTIFIERS
303 struct preempt_notifier preempt_notifier;
306 int vcpu_id; /* id given by userspace at creation */
307 int vcpu_idx; /* index in kvm->vcpus array */
311 unsigned long guest_debug;
316 #ifndef __KVM_HAVE_ARCH_WQP
319 struct pid __rcu *pid;
322 unsigned int halt_poll_ns;
325 #ifdef CONFIG_HAS_IOMEM
327 int mmio_read_completed;
329 int mmio_cur_fragment;
330 int mmio_nr_fragments;
331 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
334 #ifdef CONFIG_KVM_ASYNC_PF
337 struct list_head queue;
338 struct list_head done;
343 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
345 * Cpu relax intercept or pause loop exit optimization
346 * in_spin_loop: set when a vcpu does a pause loop exit
347 * or cpu relax intercepted.
348 * dy_eligible: indicates whether vcpu is eligible for directed yield.
357 struct kvm_vcpu_arch arch;
358 struct kvm_vcpu_stat stat;
359 char stats_id[KVM_STATS_NAME_SIZE];
360 struct kvm_dirty_ring dirty_ring;
363 * The most recently used memslot by this vCPU and the slots generation
364 * for which it is valid.
365 * No wraparound protection is needed since generations won't overflow in
366 * thousands of years, even assuming 1M memslot operations per second.
368 struct kvm_memory_slot *last_used_slot;
369 u64 last_used_slot_gen;
373 * Start accounting time towards a guest.
374 * Must be called before entering guest context.
376 static __always_inline void guest_timing_enter_irqoff(void)
379 * This is running in ioctl context so its safe to assume that it's the
380 * stime pending cputime to flush.
382 instrumentation_begin();
383 vtime_account_guest_enter();
384 instrumentation_end();
388 * Enter guest context and enter an RCU extended quiescent state.
390 * Between guest_context_enter_irqoff() and guest_context_exit_irqoff() it is
391 * unsafe to use any code which may directly or indirectly use RCU, tracing
392 * (including IRQ flag tracing), or lockdep. All code in this period must be
393 * non-instrumentable.
395 static __always_inline void guest_context_enter_irqoff(void)
398 * KVM does not hold any references to rcu protected data when it
399 * switches CPU into a guest mode. In fact switching to a guest mode
400 * is very similar to exiting to userspace from rcu point of view. In
401 * addition CPU may stay in a guest mode for quite a long time (up to
402 * one time slice). Lets treat guest mode as quiescent state, just like
403 * we do with user-mode execution.
405 if (!context_tracking_guest_enter()) {
406 instrumentation_begin();
407 rcu_virt_note_context_switch(smp_processor_id());
408 instrumentation_end();
413 * Deprecated. Architectures should move to guest_timing_enter_irqoff() and
414 * guest_state_enter_irqoff().
416 static __always_inline void guest_enter_irqoff(void)
418 guest_timing_enter_irqoff();
419 guest_context_enter_irqoff();
423 * guest_state_enter_irqoff - Fixup state when entering a guest
425 * Entry to a guest will enable interrupts, but the kernel state is interrupts
426 * disabled when this is invoked. Also tell RCU about it.
428 * 1) Trace interrupts on state
429 * 2) Invoke context tracking if enabled to adjust RCU state
430 * 3) Tell lockdep that interrupts are enabled
432 * Invoked from architecture specific code before entering a guest.
433 * Must be called with interrupts disabled and the caller must be
434 * non-instrumentable.
435 * The caller has to invoke guest_timing_enter_irqoff() before this.
437 * Note: this is analogous to exit_to_user_mode().
439 static __always_inline void guest_state_enter_irqoff(void)
441 instrumentation_begin();
442 trace_hardirqs_on_prepare();
443 lockdep_hardirqs_on_prepare(CALLER_ADDR0);
444 instrumentation_end();
446 guest_context_enter_irqoff();
447 lockdep_hardirqs_on(CALLER_ADDR0);
451 * Exit guest context and exit an RCU extended quiescent state.
453 * Between guest_context_enter_irqoff() and guest_context_exit_irqoff() it is
454 * unsafe to use any code which may directly or indirectly use RCU, tracing
455 * (including IRQ flag tracing), or lockdep. All code in this period must be
456 * non-instrumentable.
458 static __always_inline void guest_context_exit_irqoff(void)
460 context_tracking_guest_exit();
464 * Stop accounting time towards a guest.
465 * Must be called after exiting guest context.
467 static __always_inline void guest_timing_exit_irqoff(void)
469 instrumentation_begin();
470 /* Flush the guest cputime we spent on the guest */
471 vtime_account_guest_exit();
472 instrumentation_end();
476 * Deprecated. Architectures should move to guest_state_exit_irqoff() and
477 * guest_timing_exit_irqoff().
479 static __always_inline void guest_exit_irqoff(void)
481 guest_context_exit_irqoff();
482 guest_timing_exit_irqoff();
485 static inline void guest_exit(void)
489 local_irq_save(flags);
491 local_irq_restore(flags);
495 * guest_state_exit_irqoff - Establish state when returning from guest mode
497 * Entry from a guest disables interrupts, but guest mode is traced as
498 * interrupts enabled. Also with NO_HZ_FULL RCU might be idle.
500 * 1) Tell lockdep that interrupts are disabled
501 * 2) Invoke context tracking if enabled to reactivate RCU
502 * 3) Trace interrupts off state
504 * Invoked from architecture specific code after exiting a guest.
505 * Must be invoked with interrupts disabled and the caller must be
506 * non-instrumentable.
507 * The caller has to invoke guest_timing_exit_irqoff() after this.
509 * Note: this is analogous to enter_from_user_mode().
511 static __always_inline void guest_state_exit_irqoff(void)
513 lockdep_hardirqs_off(CALLER_ADDR0);
514 guest_context_exit_irqoff();
516 instrumentation_begin();
517 trace_hardirqs_off_finish();
518 instrumentation_end();
521 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
524 * The memory barrier ensures a previous write to vcpu->requests cannot
525 * be reordered with the read of vcpu->mode. It pairs with the general
526 * memory barrier following the write of vcpu->mode in VCPU RUN.
528 smp_mb__before_atomic();
529 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
533 * Some of the bitops functions do not support too long bitmaps.
534 * This number must be determined not to exceed such limits.
536 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
539 * Since at idle each memslot belongs to two memslot sets it has to contain
540 * two embedded nodes for each data structure that it forms a part of.
542 * Two memslot sets (one active and one inactive) are necessary so the VM
543 * continues to run on one memslot set while the other is being modified.
545 * These two memslot sets normally point to the same set of memslots.
546 * They can, however, be desynchronized when performing a memslot management
547 * operation by replacing the memslot to be modified by its copy.
548 * After the operation is complete, both memslot sets once again point to
549 * the same, common set of memslot data.
551 * The memslots themselves are independent of each other so they can be
552 * individually added or deleted.
554 struct kvm_memory_slot {
555 struct hlist_node id_node[2];
556 struct interval_tree_node hva_node[2];
557 struct rb_node gfn_node[2];
559 unsigned long npages;
560 unsigned long *dirty_bitmap;
561 struct kvm_arch_memory_slot arch;
562 unsigned long userspace_addr;
568 static inline bool kvm_slot_dirty_track_enabled(const struct kvm_memory_slot *slot)
570 return slot->flags & KVM_MEM_LOG_DIRTY_PAGES;
573 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
575 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
578 static inline unsigned long *kvm_second_dirty_bitmap(struct kvm_memory_slot *memslot)
580 unsigned long len = kvm_dirty_bitmap_bytes(memslot);
582 return memslot->dirty_bitmap + len / sizeof(*memslot->dirty_bitmap);
585 #ifndef KVM_DIRTY_LOG_MANUAL_CAPS
586 #define KVM_DIRTY_LOG_MANUAL_CAPS KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE
589 struct kvm_s390_adapter_int {
602 struct kvm_xen_evtchn {
608 struct kvm_kernel_irq_routing_entry {
611 int (*set)(struct kvm_kernel_irq_routing_entry *e,
612 struct kvm *kvm, int irq_source_id, int level,
626 struct kvm_s390_adapter_int adapter;
627 struct kvm_hv_sint hv_sint;
628 struct kvm_xen_evtchn xen_evtchn;
630 struct hlist_node link;
633 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
634 struct kvm_irq_routing_table {
635 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
638 * Array indexed by gsi. Each entry contains list of irq chips
639 * the gsi is connected to.
641 struct hlist_head map[];
645 #ifndef KVM_PRIVATE_MEM_SLOTS
646 #define KVM_PRIVATE_MEM_SLOTS 0
649 #define KVM_MEM_SLOTS_NUM SHRT_MAX
650 #define KVM_USER_MEM_SLOTS (KVM_MEM_SLOTS_NUM - KVM_PRIVATE_MEM_SLOTS)
652 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
653 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
659 struct kvm_memslots {
661 atomic_long_t last_used_slot;
662 struct rb_root_cached hva_tree;
663 struct rb_root gfn_tree;
665 * The mapping table from slot id to memslot.
667 * 7-bit bucket count matches the size of the old id to index array for
668 * 512 slots, while giving good performance with this slot count.
669 * Higher bucket counts bring only small performance improvements but
670 * always result in higher memory usage (even for lower memslot counts).
672 DECLARE_HASHTABLE(id_hash, 7);
677 #ifdef KVM_HAVE_MMU_RWLOCK
681 #endif /* KVM_HAVE_MMU_RWLOCK */
683 struct mutex slots_lock;
686 * Protects the arch-specific fields of struct kvm_memory_slots in
687 * use by the VM. To be used under the slots_lock (above) or in a
688 * kvm->srcu critical section where acquiring the slots_lock would
689 * lead to deadlock with the synchronize_srcu in
690 * install_new_memslots.
692 struct mutex slots_arch_lock;
693 struct mm_struct *mm; /* userspace tied to this vm */
694 unsigned long nr_memslot_pages;
695 /* The two memslot sets - active and inactive (per address space) */
696 struct kvm_memslots __memslots[KVM_ADDRESS_SPACE_NUM][2];
697 /* The current active memslot set for each address space */
698 struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
699 struct xarray vcpu_array;
701 /* Used to wait for completion of MMU notifiers. */
702 spinlock_t mn_invalidate_lock;
703 unsigned long mn_active_invalidate_count;
704 struct rcuwait mn_memslots_update_rcuwait;
706 /* For management / invalidation of gfn_to_pfn_caches */
708 struct list_head gpc_list;
711 * created_vcpus is protected by kvm->lock, and is incremented
712 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
713 * incremented after storing the kvm_vcpu pointer in vcpus,
714 * and is accessed atomically.
716 atomic_t online_vcpus;
718 int last_boosted_vcpu;
719 struct list_head vm_list;
721 struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
722 #ifdef CONFIG_HAVE_KVM_EVENTFD
725 struct list_head items;
726 struct list_head resampler_list;
727 struct mutex resampler_lock;
729 struct list_head ioeventfds;
731 struct kvm_vm_stat stat;
732 struct kvm_arch arch;
733 refcount_t users_count;
734 #ifdef CONFIG_KVM_MMIO
735 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
736 spinlock_t ring_lock;
737 struct list_head coalesced_zones;
740 struct mutex irq_lock;
741 #ifdef CONFIG_HAVE_KVM_IRQCHIP
743 * Update side is protected by irq_lock.
745 struct kvm_irq_routing_table __rcu *irq_routing;
747 #ifdef CONFIG_HAVE_KVM_IRQFD
748 struct hlist_head irq_ack_notifier_list;
751 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
752 struct mmu_notifier mmu_notifier;
753 unsigned long mmu_notifier_seq;
754 long mmu_notifier_count;
755 unsigned long mmu_notifier_range_start;
756 unsigned long mmu_notifier_range_end;
758 struct list_head devices;
759 u64 manual_dirty_log_protect;
760 struct dentry *debugfs_dentry;
761 struct kvm_stat_data **debugfs_stat_data;
762 struct srcu_struct srcu;
763 struct srcu_struct irq_srcu;
765 unsigned int max_halt_poll_ns;
770 #ifdef CONFIG_HAVE_KVM_PM_NOTIFIER
771 struct notifier_block pm_notifier;
773 char stats_id[KVM_STATS_NAME_SIZE];
776 #define kvm_err(fmt, ...) \
777 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
778 #define kvm_info(fmt, ...) \
779 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
780 #define kvm_debug(fmt, ...) \
781 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
782 #define kvm_debug_ratelimited(fmt, ...) \
783 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
785 #define kvm_pr_unimpl(fmt, ...) \
786 pr_err_ratelimited("kvm [%i]: " fmt, \
787 task_tgid_nr(current), ## __VA_ARGS__)
789 /* The guest did something we don't support. */
790 #define vcpu_unimpl(vcpu, fmt, ...) \
791 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
792 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
794 #define vcpu_debug(vcpu, fmt, ...) \
795 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
796 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
797 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
799 #define vcpu_err(vcpu, fmt, ...) \
800 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
802 static inline void kvm_vm_dead(struct kvm *kvm)
805 kvm_make_all_cpus_request(kvm, KVM_REQ_VM_DEAD);
808 static inline void kvm_vm_bugged(struct kvm *kvm)
810 kvm->vm_bugged = true;
815 #define KVM_BUG(cond, kvm, fmt...) \
817 int __ret = (cond); \
819 if (WARN_ONCE(__ret && !(kvm)->vm_bugged, fmt)) \
820 kvm_vm_bugged(kvm); \
824 #define KVM_BUG_ON(cond, kvm) \
826 int __ret = (cond); \
828 if (WARN_ON_ONCE(__ret && !(kvm)->vm_bugged)) \
829 kvm_vm_bugged(kvm); \
833 static inline bool kvm_dirty_log_manual_protect_and_init_set(struct kvm *kvm)
835 return !!(kvm->manual_dirty_log_protect & KVM_DIRTY_LOG_INITIALLY_SET);
838 static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx)
840 return srcu_dereference_check(kvm->buses[idx], &kvm->srcu,
841 lockdep_is_held(&kvm->slots_lock) ||
842 !refcount_read(&kvm->users_count));
845 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
847 int num_vcpus = atomic_read(&kvm->online_vcpus);
848 i = array_index_nospec(i, num_vcpus);
850 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. */
852 return xa_load(&kvm->vcpu_array, i);
855 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
856 xa_for_each_range(&kvm->vcpu_array, idx, vcpup, 0, \
857 (atomic_read(&kvm->online_vcpus) - 1))
859 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
861 struct kvm_vcpu *vcpu = NULL;
866 if (id < KVM_MAX_VCPUS)
867 vcpu = kvm_get_vcpu(kvm, id);
868 if (vcpu && vcpu->vcpu_id == id)
870 kvm_for_each_vcpu(i, vcpu, kvm)
871 if (vcpu->vcpu_id == id)
876 static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu)
878 return vcpu->vcpu_idx;
881 void kvm_destroy_vcpus(struct kvm *kvm);
883 void vcpu_load(struct kvm_vcpu *vcpu);
884 void vcpu_put(struct kvm_vcpu *vcpu);
886 #ifdef __KVM_HAVE_IOAPIC
887 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
888 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
890 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
893 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
898 #ifdef CONFIG_HAVE_KVM_IRQFD
899 int kvm_irqfd_init(void);
900 void kvm_irqfd_exit(void);
902 static inline int kvm_irqfd_init(void)
907 static inline void kvm_irqfd_exit(void)
911 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
912 struct module *module);
915 void kvm_get_kvm(struct kvm *kvm);
916 bool kvm_get_kvm_safe(struct kvm *kvm);
917 void kvm_put_kvm(struct kvm *kvm);
918 bool file_is_kvm(struct file *file);
919 void kvm_put_kvm_no_destroy(struct kvm *kvm);
921 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
923 as_id = array_index_nospec(as_id, KVM_ADDRESS_SPACE_NUM);
924 return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu,
925 lockdep_is_held(&kvm->slots_lock) ||
926 !refcount_read(&kvm->users_count));
929 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
931 return __kvm_memslots(kvm, 0);
934 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
936 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
938 return __kvm_memslots(vcpu->kvm, as_id);
941 static inline bool kvm_memslots_empty(struct kvm_memslots *slots)
943 return RB_EMPTY_ROOT(&slots->gfn_tree);
946 #define kvm_for_each_memslot(memslot, bkt, slots) \
947 hash_for_each(slots->id_hash, bkt, memslot, id_node[slots->node_idx]) \
948 if (WARN_ON_ONCE(!memslot->npages)) { \
952 struct kvm_memory_slot *id_to_memslot(struct kvm_memslots *slots, int id)
954 struct kvm_memory_slot *slot;
955 int idx = slots->node_idx;
957 hash_for_each_possible(slots->id_hash, slot, id_node[idx], id) {
965 /* Iterator used for walking memslots that overlap a gfn range. */
966 struct kvm_memslot_iter {
967 struct kvm_memslots *slots;
968 struct rb_node *node;
969 struct kvm_memory_slot *slot;
972 static inline void kvm_memslot_iter_next(struct kvm_memslot_iter *iter)
974 iter->node = rb_next(iter->node);
978 iter->slot = container_of(iter->node, struct kvm_memory_slot, gfn_node[iter->slots->node_idx]);
981 static inline void kvm_memslot_iter_start(struct kvm_memslot_iter *iter,
982 struct kvm_memslots *slots,
985 int idx = slots->node_idx;
987 struct kvm_memory_slot *slot;
992 * Find the so called "upper bound" of a key - the first node that has
993 * its key strictly greater than the searched one (the start gfn in our case).
996 for (tmp = slots->gfn_tree.rb_node; tmp; ) {
997 slot = container_of(tmp, struct kvm_memory_slot, gfn_node[idx]);
998 if (start < slot->base_gfn) {
1002 tmp = tmp->rb_right;
1007 * Find the slot with the lowest gfn that can possibly intersect with
1008 * the range, so we'll ideally have slot start <= range start
1012 * A NULL previous node means that the very first slot
1013 * already has a higher start gfn.
1014 * In this case slot start > range start.
1016 tmp = rb_prev(iter->node);
1020 /* a NULL node below means no slots */
1021 iter->node = rb_last(&slots->gfn_tree);
1025 iter->slot = container_of(iter->node, struct kvm_memory_slot, gfn_node[idx]);
1028 * It is possible in the slot start < range start case that the
1029 * found slot ends before or at range start (slot end <= range start)
1030 * and so it does not overlap the requested range.
1032 * In such non-overlapping case the next slot (if it exists) will
1033 * already have slot start > range start, otherwise the logic above
1034 * would have found it instead of the current slot.
1036 if (iter->slot->base_gfn + iter->slot->npages <= start)
1037 kvm_memslot_iter_next(iter);
1041 static inline bool kvm_memslot_iter_is_valid(struct kvm_memslot_iter *iter, gfn_t end)
1047 * If this slot starts beyond or at the end of the range so does
1050 return iter->slot->base_gfn < end;
1053 /* Iterate over each memslot at least partially intersecting [start, end) range */
1054 #define kvm_for_each_memslot_in_gfn_range(iter, slots, start, end) \
1055 for (kvm_memslot_iter_start(iter, slots, start); \
1056 kvm_memslot_iter_is_valid(iter, end); \
1057 kvm_memslot_iter_next(iter))
1060 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
1061 * - create a new memory slot
1062 * - delete an existing memory slot
1063 * - modify an existing memory slot
1064 * -- move it in the guest physical memory space
1065 * -- just change its flags
1067 * Since flags can be changed by some of these operations, the following
1068 * differentiation is the best we can do for __kvm_set_memory_region():
1070 enum kvm_mr_change {
1077 int kvm_set_memory_region(struct kvm *kvm,
1078 const struct kvm_userspace_memory_region *mem);
1079 int __kvm_set_memory_region(struct kvm *kvm,
1080 const struct kvm_userspace_memory_region *mem);
1081 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot);
1082 void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen);
1083 int kvm_arch_prepare_memory_region(struct kvm *kvm,
1084 const struct kvm_memory_slot *old,
1085 struct kvm_memory_slot *new,
1086 enum kvm_mr_change change);
1087 void kvm_arch_commit_memory_region(struct kvm *kvm,
1088 struct kvm_memory_slot *old,
1089 const struct kvm_memory_slot *new,
1090 enum kvm_mr_change change);
1091 /* flush all memory translations */
1092 void kvm_arch_flush_shadow_all(struct kvm *kvm);
1093 /* flush memory translations pointing to 'slot' */
1094 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
1095 struct kvm_memory_slot *slot);
1097 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
1098 struct page **pages, int nr_pages);
1100 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
1101 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
1102 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
1103 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
1104 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
1106 void kvm_release_page_clean(struct page *page);
1107 void kvm_release_page_dirty(struct page *page);
1108 void kvm_set_page_accessed(struct page *page);
1110 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
1111 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1113 kvm_pfn_t gfn_to_pfn_memslot(const struct kvm_memory_slot *slot, gfn_t gfn);
1114 kvm_pfn_t gfn_to_pfn_memslot_atomic(const struct kvm_memory_slot *slot, gfn_t gfn);
1115 kvm_pfn_t __gfn_to_pfn_memslot(const struct kvm_memory_slot *slot, gfn_t gfn,
1116 bool atomic, bool *async, bool write_fault,
1117 bool *writable, hva_t *hva);
1119 void kvm_release_pfn_clean(kvm_pfn_t pfn);
1120 void kvm_release_pfn_dirty(kvm_pfn_t pfn);
1121 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
1122 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
1124 void kvm_release_pfn(kvm_pfn_t pfn, bool dirty);
1125 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
1127 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
1128 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1129 void *data, unsigned long len);
1130 int kvm_read_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1131 void *data, unsigned int offset,
1133 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
1134 int offset, int len);
1135 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
1137 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1138 void *data, unsigned long len);
1139 int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1140 void *data, unsigned int offset,
1142 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1143 gpa_t gpa, unsigned long len);
1145 #define __kvm_get_guest(kvm, gfn, offset, v) \
1147 unsigned long __addr = gfn_to_hva(kvm, gfn); \
1148 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
1149 int __ret = -EFAULT; \
1151 if (!kvm_is_error_hva(__addr)) \
1152 __ret = get_user(v, __uaddr); \
1156 #define kvm_get_guest(kvm, gpa, v) \
1158 gpa_t __gpa = gpa; \
1159 struct kvm *__kvm = kvm; \
1161 __kvm_get_guest(__kvm, __gpa >> PAGE_SHIFT, \
1162 offset_in_page(__gpa), v); \
1165 #define __kvm_put_guest(kvm, gfn, offset, v) \
1167 unsigned long __addr = gfn_to_hva(kvm, gfn); \
1168 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
1169 int __ret = -EFAULT; \
1171 if (!kvm_is_error_hva(__addr)) \
1172 __ret = put_user(v, __uaddr); \
1174 mark_page_dirty(kvm, gfn); \
1178 #define kvm_put_guest(kvm, gpa, v) \
1180 gpa_t __gpa = gpa; \
1181 struct kvm *__kvm = kvm; \
1183 __kvm_put_guest(__kvm, __gpa >> PAGE_SHIFT, \
1184 offset_in_page(__gpa), v); \
1187 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
1188 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
1189 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
1190 bool kvm_vcpu_is_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1191 unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn);
1192 void mark_page_dirty_in_slot(struct kvm *kvm, const struct kvm_memory_slot *memslot, gfn_t gfn);
1193 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
1195 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
1196 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
1197 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
1198 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1199 int kvm_vcpu_map(struct kvm_vcpu *vcpu, gpa_t gpa, struct kvm_host_map *map);
1200 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
1201 void kvm_vcpu_unmap(struct kvm_vcpu *vcpu, struct kvm_host_map *map, bool dirty);
1202 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
1203 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
1204 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
1206 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
1208 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
1210 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
1211 int offset, int len);
1212 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
1214 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
1217 * kvm_gfn_to_pfn_cache_init - prepare a cached kernel mapping and HPA for a
1218 * given guest physical address.
1220 * @kvm: pointer to kvm instance.
1221 * @gpc: struct gfn_to_pfn_cache object.
1222 * @vcpu: vCPU to be used for marking pages dirty and to be woken on
1224 * @guest_uses_pa: indicates that the resulting host physical PFN is used while
1225 * @vcpu is IN_GUEST_MODE so invalidations should wake it.
1226 * @kernel_map: requests a kernel virtual mapping (kmap / memremap).
1227 * @gpa: guest physical address to map.
1228 * @len: sanity check; the range being access must fit a single page.
1229 * @dirty: mark the cache dirty immediately.
1231 * @return: 0 for success.
1232 * -EINVAL for a mapping which would cross a page boundary.
1233 * -EFAULT for an untranslatable guest physical address.
1235 * This primes a gfn_to_pfn_cache and links it into the @kvm's list for
1236 * invalidations to be processed. Invalidation callbacks to @vcpu using
1237 * %KVM_REQ_GPC_INVALIDATE will occur only for MMU notifiers, not for KVM
1238 * memslot changes. Callers are required to use kvm_gfn_to_pfn_cache_check()
1239 * to ensure that the cache is valid before accessing the target page.
1241 int kvm_gfn_to_pfn_cache_init(struct kvm *kvm, struct gfn_to_pfn_cache *gpc,
1242 struct kvm_vcpu *vcpu, bool guest_uses_pa,
1243 bool kernel_map, gpa_t gpa, unsigned long len,
1247 * kvm_gfn_to_pfn_cache_check - check validity of a gfn_to_pfn_cache.
1249 * @kvm: pointer to kvm instance.
1250 * @gpc: struct gfn_to_pfn_cache object.
1251 * @gpa: current guest physical address to map.
1252 * @len: sanity check; the range being access must fit a single page.
1253 * @dirty: mark the cache dirty immediately.
1255 * @return: %true if the cache is still valid and the address matches.
1256 * %false if the cache is not valid.
1258 * Callers outside IN_GUEST_MODE context should hold a read lock on @gpc->lock
1259 * while calling this function, and then continue to hold the lock until the
1260 * access is complete.
1262 * Callers in IN_GUEST_MODE may do so without locking, although they should
1263 * still hold a read lock on kvm->scru for the memslot checks.
1265 bool kvm_gfn_to_pfn_cache_check(struct kvm *kvm, struct gfn_to_pfn_cache *gpc,
1266 gpa_t gpa, unsigned long len);
1269 * kvm_gfn_to_pfn_cache_refresh - update a previously initialized cache.
1271 * @kvm: pointer to kvm instance.
1272 * @gpc: struct gfn_to_pfn_cache object.
1273 * @gpa: updated guest physical address to map.
1274 * @len: sanity check; the range being access must fit a single page.
1275 * @dirty: mark the cache dirty immediately.
1277 * @return: 0 for success.
1278 * -EINVAL for a mapping which would cross a page boundary.
1279 * -EFAULT for an untranslatable guest physical address.
1281 * This will attempt to refresh a gfn_to_pfn_cache. Note that a successful
1282 * returm from this function does not mean the page can be immediately
1283 * accessed because it may have raced with an invalidation. Callers must
1284 * still lock and check the cache status, as this function does not return
1285 * with the lock still held to permit access.
1287 int kvm_gfn_to_pfn_cache_refresh(struct kvm *kvm, struct gfn_to_pfn_cache *gpc,
1288 gpa_t gpa, unsigned long len, bool dirty);
1291 * kvm_gfn_to_pfn_cache_unmap - temporarily unmap a gfn_to_pfn_cache.
1293 * @kvm: pointer to kvm instance.
1294 * @gpc: struct gfn_to_pfn_cache object.
1296 * This unmaps the referenced page and marks it dirty, if appropriate. The
1297 * cache is left in the invalid state but at least the mapping from GPA to
1298 * userspace HVA will remain cached and can be reused on a subsequent
1301 void kvm_gfn_to_pfn_cache_unmap(struct kvm *kvm, struct gfn_to_pfn_cache *gpc);
1304 * kvm_gfn_to_pfn_cache_destroy - destroy and unlink a gfn_to_pfn_cache.
1306 * @kvm: pointer to kvm instance.
1307 * @gpc: struct gfn_to_pfn_cache object.
1309 * This removes a cache from the @kvm's list to be processed on MMU notifier
1312 void kvm_gfn_to_pfn_cache_destroy(struct kvm *kvm, struct gfn_to_pfn_cache *gpc);
1314 void kvm_sigset_activate(struct kvm_vcpu *vcpu);
1315 void kvm_sigset_deactivate(struct kvm_vcpu *vcpu);
1317 void kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1318 bool kvm_vcpu_block(struct kvm_vcpu *vcpu);
1319 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
1320 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
1321 bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
1322 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
1323 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
1324 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool usermode_vcpu_not_eligible);
1326 void kvm_flush_remote_tlbs(struct kvm *kvm);
1328 #ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE
1329 int kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache *mc, int min);
1330 int kvm_mmu_memory_cache_nr_free_objects(struct kvm_mmu_memory_cache *mc);
1331 void kvm_mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc);
1332 void *kvm_mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc);
1335 void kvm_inc_notifier_count(struct kvm *kvm, unsigned long start,
1337 void kvm_dec_notifier_count(struct kvm *kvm, unsigned long start,
1340 long kvm_arch_dev_ioctl(struct file *filp,
1341 unsigned int ioctl, unsigned long arg);
1342 long kvm_arch_vcpu_ioctl(struct file *filp,
1343 unsigned int ioctl, unsigned long arg);
1344 vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
1346 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
1348 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
1349 struct kvm_memory_slot *slot,
1351 unsigned long mask);
1352 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot);
1354 #ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
1355 void kvm_arch_flush_remote_tlbs_memslot(struct kvm *kvm,
1356 const struct kvm_memory_slot *memslot);
1357 #else /* !CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
1358 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log);
1359 int kvm_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log,
1360 int *is_dirty, struct kvm_memory_slot **memslot);
1363 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
1365 int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
1366 struct kvm_enable_cap *cap);
1367 long kvm_arch_vm_ioctl(struct file *filp,
1368 unsigned int ioctl, unsigned long arg);
1370 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
1371 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
1373 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1374 struct kvm_translation *tr);
1376 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
1377 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
1378 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1379 struct kvm_sregs *sregs);
1380 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1381 struct kvm_sregs *sregs);
1382 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
1383 struct kvm_mp_state *mp_state);
1384 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
1385 struct kvm_mp_state *mp_state);
1386 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
1387 struct kvm_guest_debug *dbg);
1388 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu);
1390 int kvm_arch_init(void *opaque);
1391 void kvm_arch_exit(void);
1393 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
1395 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
1396 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
1397 int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id);
1398 int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu);
1399 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
1400 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
1402 #ifdef CONFIG_HAVE_KVM_PM_NOTIFIER
1403 int kvm_arch_pm_notifier(struct kvm *kvm, unsigned long state);
1406 #ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
1407 void kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu, struct dentry *debugfs_dentry);
1410 int kvm_arch_hardware_enable(void);
1411 void kvm_arch_hardware_disable(void);
1412 int kvm_arch_hardware_setup(void *opaque);
1413 void kvm_arch_hardware_unsetup(void);
1414 int kvm_arch_check_processor_compat(void *opaque);
1415 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
1416 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
1417 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
1418 bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu);
1419 bool kvm_arch_dy_has_pending_interrupt(struct kvm_vcpu *vcpu);
1420 int kvm_arch_post_init_vm(struct kvm *kvm);
1421 void kvm_arch_pre_destroy_vm(struct kvm *kvm);
1422 int kvm_arch_create_vm_debugfs(struct kvm *kvm);
1424 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
1426 * All architectures that want to use vzalloc currently also
1427 * need their own kvm_arch_alloc_vm implementation.
1429 static inline struct kvm *kvm_arch_alloc_vm(void)
1431 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
1435 static inline void __kvm_arch_free_vm(struct kvm *kvm)
1440 #ifndef __KVM_HAVE_ARCH_VM_FREE
1441 static inline void kvm_arch_free_vm(struct kvm *kvm)
1443 __kvm_arch_free_vm(kvm);
1447 #ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1448 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1454 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
1455 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
1456 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
1457 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
1459 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
1463 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
1467 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
1472 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
1473 void kvm_arch_start_assignment(struct kvm *kvm);
1474 void kvm_arch_end_assignment(struct kvm *kvm);
1475 bool kvm_arch_has_assigned_device(struct kvm *kvm);
1477 static inline void kvm_arch_start_assignment(struct kvm *kvm)
1481 static inline void kvm_arch_end_assignment(struct kvm *kvm)
1485 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
1491 static inline struct rcuwait *kvm_arch_vcpu_get_wait(struct kvm_vcpu *vcpu)
1493 #ifdef __KVM_HAVE_ARCH_WQP
1494 return vcpu->arch.waitp;
1501 * Wake a vCPU if necessary, but don't do any stats/metadata updates. Returns
1502 * true if the vCPU was blocking and was awakened, false otherwise.
1504 static inline bool __kvm_vcpu_wake_up(struct kvm_vcpu *vcpu)
1506 return !!rcuwait_wake_up(kvm_arch_vcpu_get_wait(vcpu));
1509 static inline bool kvm_vcpu_is_blocking(struct kvm_vcpu *vcpu)
1511 return rcuwait_active(kvm_arch_vcpu_get_wait(vcpu));
1514 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
1516 * returns true if the virtual interrupt controller is initialized and
1517 * ready to accept virtual IRQ. On some architectures the virtual interrupt
1518 * controller is dynamically instantiated and this is not always true.
1520 bool kvm_arch_intc_initialized(struct kvm *kvm);
1522 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
1528 #ifdef CONFIG_GUEST_PERF_EVENTS
1529 unsigned long kvm_arch_vcpu_get_ip(struct kvm_vcpu *vcpu);
1531 void kvm_register_perf_callbacks(unsigned int (*pt_intr_handler)(void));
1532 void kvm_unregister_perf_callbacks(void);
1534 static inline void kvm_register_perf_callbacks(void *ign) {}
1535 static inline void kvm_unregister_perf_callbacks(void) {}
1536 #endif /* CONFIG_GUEST_PERF_EVENTS */
1538 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
1539 void kvm_arch_destroy_vm(struct kvm *kvm);
1540 void kvm_arch_sync_events(struct kvm *kvm);
1542 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
1544 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
1545 bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
1547 struct kvm_irq_ack_notifier {
1548 struct hlist_node link;
1550 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
1553 int kvm_irq_map_gsi(struct kvm *kvm,
1554 struct kvm_kernel_irq_routing_entry *entries, int gsi);
1555 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
1557 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
1559 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
1560 int irq_source_id, int level, bool line_status);
1561 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
1562 struct kvm *kvm, int irq_source_id,
1563 int level, bool line_status);
1564 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
1565 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
1566 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
1567 void kvm_register_irq_ack_notifier(struct kvm *kvm,
1568 struct kvm_irq_ack_notifier *kian);
1569 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
1570 struct kvm_irq_ack_notifier *kian);
1571 int kvm_request_irq_source_id(struct kvm *kvm);
1572 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
1573 bool kvm_arch_irqfd_allowed(struct kvm *kvm, struct kvm_irqfd *args);
1576 * Returns a pointer to the memslot if it contains gfn.
1577 * Otherwise returns NULL.
1579 static inline struct kvm_memory_slot *
1580 try_get_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1585 if (gfn >= slot->base_gfn && gfn < slot->base_gfn + slot->npages)
1592 * Returns a pointer to the memslot that contains gfn. Otherwise returns NULL.
1594 * With "approx" set returns the memslot also when the address falls
1595 * in a hole. In that case one of the memslots bordering the hole is
1598 static inline struct kvm_memory_slot *
1599 search_memslots(struct kvm_memslots *slots, gfn_t gfn, bool approx)
1601 struct kvm_memory_slot *slot;
1602 struct rb_node *node;
1603 int idx = slots->node_idx;
1606 for (node = slots->gfn_tree.rb_node; node; ) {
1607 slot = container_of(node, struct kvm_memory_slot, gfn_node[idx]);
1608 if (gfn >= slot->base_gfn) {
1609 if (gfn < slot->base_gfn + slot->npages)
1611 node = node->rb_right;
1613 node = node->rb_left;
1616 return approx ? slot : NULL;
1619 static inline struct kvm_memory_slot *
1620 ____gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn, bool approx)
1622 struct kvm_memory_slot *slot;
1624 slot = (struct kvm_memory_slot *)atomic_long_read(&slots->last_used_slot);
1625 slot = try_get_memslot(slot, gfn);
1629 slot = search_memslots(slots, gfn, approx);
1631 atomic_long_set(&slots->last_used_slot, (unsigned long)slot);
1639 * __gfn_to_memslot() and its descendants are here to allow arch code to inline
1640 * the lookups in hot paths. gfn_to_memslot() itself isn't here as an inline
1641 * because that would bloat other code too much.
1643 static inline struct kvm_memory_slot *
1644 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
1646 return ____gfn_to_memslot(slots, gfn, false);
1649 static inline unsigned long
1650 __gfn_to_hva_memslot(const struct kvm_memory_slot *slot, gfn_t gfn)
1653 * The index was checked originally in search_memslots. To avoid
1654 * that a malicious guest builds a Spectre gadget out of e.g. page
1655 * table walks, do not let the processor speculate loads outside
1656 * the guest's registered memslots.
1658 unsigned long offset = gfn - slot->base_gfn;
1659 offset = array_index_nospec(offset, slot->npages);
1660 return slot->userspace_addr + offset * PAGE_SIZE;
1663 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
1665 return gfn_to_memslot(kvm, gfn)->id;
1669 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
1671 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
1673 return slot->base_gfn + gfn_offset;
1676 static inline gpa_t gfn_to_gpa(gfn_t gfn)
1678 return (gpa_t)gfn << PAGE_SHIFT;
1681 static inline gfn_t gpa_to_gfn(gpa_t gpa)
1683 return (gfn_t)(gpa >> PAGE_SHIFT);
1686 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
1688 return (hpa_t)pfn << PAGE_SHIFT;
1691 static inline struct page *kvm_vcpu_gpa_to_page(struct kvm_vcpu *vcpu,
1694 return kvm_vcpu_gfn_to_page(vcpu, gpa_to_gfn(gpa));
1697 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
1699 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
1701 return kvm_is_error_hva(hva);
1704 enum kvm_stat_kind {
1709 struct kvm_stat_data {
1711 const struct _kvm_stats_desc *desc;
1712 enum kvm_stat_kind kind;
1715 struct _kvm_stats_desc {
1716 struct kvm_stats_desc desc;
1717 char name[KVM_STATS_NAME_SIZE];
1720 #define STATS_DESC_COMMON(type, unit, base, exp, sz, bsz) \
1721 .flags = type | unit | base | \
1722 BUILD_BUG_ON_ZERO(type & ~KVM_STATS_TYPE_MASK) | \
1723 BUILD_BUG_ON_ZERO(unit & ~KVM_STATS_UNIT_MASK) | \
1724 BUILD_BUG_ON_ZERO(base & ~KVM_STATS_BASE_MASK), \
1729 #define VM_GENERIC_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1732 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1733 .offset = offsetof(struct kvm_vm_stat, generic.stat) \
1737 #define VCPU_GENERIC_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1740 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1741 .offset = offsetof(struct kvm_vcpu_stat, generic.stat) \
1745 #define VM_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1748 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1749 .offset = offsetof(struct kvm_vm_stat, stat) \
1753 #define VCPU_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1756 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1757 .offset = offsetof(struct kvm_vcpu_stat, stat) \
1761 /* SCOPE: VM, VM_GENERIC, VCPU, VCPU_GENERIC */
1762 #define STATS_DESC(SCOPE, stat, type, unit, base, exp, sz, bsz) \
1763 SCOPE##_STATS_DESC(stat, type, unit, base, exp, sz, bsz)
1765 #define STATS_DESC_CUMULATIVE(SCOPE, name, unit, base, exponent) \
1766 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_CUMULATIVE, \
1767 unit, base, exponent, 1, 0)
1768 #define STATS_DESC_INSTANT(SCOPE, name, unit, base, exponent) \
1769 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_INSTANT, \
1770 unit, base, exponent, 1, 0)
1771 #define STATS_DESC_PEAK(SCOPE, name, unit, base, exponent) \
1772 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_PEAK, \
1773 unit, base, exponent, 1, 0)
1774 #define STATS_DESC_LINEAR_HIST(SCOPE, name, unit, base, exponent, sz, bsz) \
1775 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_LINEAR_HIST, \
1776 unit, base, exponent, sz, bsz)
1777 #define STATS_DESC_LOG_HIST(SCOPE, name, unit, base, exponent, sz) \
1778 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_LOG_HIST, \
1779 unit, base, exponent, sz, 0)
1781 /* Cumulative counter, read/write */
1782 #define STATS_DESC_COUNTER(SCOPE, name) \
1783 STATS_DESC_CUMULATIVE(SCOPE, name, KVM_STATS_UNIT_NONE, \
1784 KVM_STATS_BASE_POW10, 0)
1785 /* Instantaneous counter, read only */
1786 #define STATS_DESC_ICOUNTER(SCOPE, name) \
1787 STATS_DESC_INSTANT(SCOPE, name, KVM_STATS_UNIT_NONE, \
1788 KVM_STATS_BASE_POW10, 0)
1789 /* Peak counter, read/write */
1790 #define STATS_DESC_PCOUNTER(SCOPE, name) \
1791 STATS_DESC_PEAK(SCOPE, name, KVM_STATS_UNIT_NONE, \
1792 KVM_STATS_BASE_POW10, 0)
1794 /* Cumulative time in nanosecond */
1795 #define STATS_DESC_TIME_NSEC(SCOPE, name) \
1796 STATS_DESC_CUMULATIVE(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
1797 KVM_STATS_BASE_POW10, -9)
1798 /* Linear histogram for time in nanosecond */
1799 #define STATS_DESC_LINHIST_TIME_NSEC(SCOPE, name, sz, bsz) \
1800 STATS_DESC_LINEAR_HIST(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
1801 KVM_STATS_BASE_POW10, -9, sz, bsz)
1802 /* Logarithmic histogram for time in nanosecond */
1803 #define STATS_DESC_LOGHIST_TIME_NSEC(SCOPE, name, sz) \
1804 STATS_DESC_LOG_HIST(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
1805 KVM_STATS_BASE_POW10, -9, sz)
1807 #define KVM_GENERIC_VM_STATS() \
1808 STATS_DESC_COUNTER(VM_GENERIC, remote_tlb_flush), \
1809 STATS_DESC_COUNTER(VM_GENERIC, remote_tlb_flush_requests)
1811 #define KVM_GENERIC_VCPU_STATS() \
1812 STATS_DESC_COUNTER(VCPU_GENERIC, halt_successful_poll), \
1813 STATS_DESC_COUNTER(VCPU_GENERIC, halt_attempted_poll), \
1814 STATS_DESC_COUNTER(VCPU_GENERIC, halt_poll_invalid), \
1815 STATS_DESC_COUNTER(VCPU_GENERIC, halt_wakeup), \
1816 STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_poll_success_ns), \
1817 STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_poll_fail_ns), \
1818 STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_wait_ns), \
1819 STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_poll_success_hist, \
1820 HALT_POLL_HIST_COUNT), \
1821 STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_poll_fail_hist, \
1822 HALT_POLL_HIST_COUNT), \
1823 STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_wait_hist, \
1824 HALT_POLL_HIST_COUNT), \
1825 STATS_DESC_ICOUNTER(VCPU_GENERIC, blocking)
1827 extern struct dentry *kvm_debugfs_dir;
1829 ssize_t kvm_stats_read(char *id, const struct kvm_stats_header *header,
1830 const struct _kvm_stats_desc *desc,
1831 void *stats, size_t size_stats,
1832 char __user *user_buffer, size_t size, loff_t *offset);
1835 * kvm_stats_linear_hist_update() - Update bucket value for linear histogram
1838 * @data: start address of the stats data
1839 * @size: the number of bucket of the stats data
1840 * @value: the new value used to update the linear histogram's bucket
1841 * @bucket_size: the size (width) of a bucket
1843 static inline void kvm_stats_linear_hist_update(u64 *data, size_t size,
1844 u64 value, size_t bucket_size)
1846 size_t index = div64_u64(value, bucket_size);
1848 index = min(index, size - 1);
1853 * kvm_stats_log_hist_update() - Update bucket value for logarithmic histogram
1856 * @data: start address of the stats data
1857 * @size: the number of bucket of the stats data
1858 * @value: the new value used to update the logarithmic histogram's bucket
1860 static inline void kvm_stats_log_hist_update(u64 *data, size_t size, u64 value)
1862 size_t index = fls64(value);
1864 index = min(index, size - 1);
1868 #define KVM_STATS_LINEAR_HIST_UPDATE(array, value, bsize) \
1869 kvm_stats_linear_hist_update(array, ARRAY_SIZE(array), value, bsize)
1870 #define KVM_STATS_LOG_HIST_UPDATE(array, value) \
1871 kvm_stats_log_hist_update(array, ARRAY_SIZE(array), value)
1874 extern const struct kvm_stats_header kvm_vm_stats_header;
1875 extern const struct _kvm_stats_desc kvm_vm_stats_desc[];
1876 extern const struct kvm_stats_header kvm_vcpu_stats_header;
1877 extern const struct _kvm_stats_desc kvm_vcpu_stats_desc[];
1879 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1880 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1882 if (unlikely(kvm->mmu_notifier_count))
1885 * Ensure the read of mmu_notifier_count happens before the read
1886 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1887 * mmu_notifier_invalidate_range_end to make sure that the caller
1888 * either sees the old (non-zero) value of mmu_notifier_count or
1889 * the new (incremented) value of mmu_notifier_seq.
1890 * PowerPC Book3s HV KVM calls this under a per-page lock
1891 * rather than under kvm->mmu_lock, for scalability, so
1892 * can't rely on kvm->mmu_lock to keep things ordered.
1895 if (kvm->mmu_notifier_seq != mmu_seq)
1900 static inline int mmu_notifier_retry_hva(struct kvm *kvm,
1901 unsigned long mmu_seq,
1904 lockdep_assert_held(&kvm->mmu_lock);
1906 * If mmu_notifier_count is non-zero, then the range maintained by
1907 * kvm_mmu_notifier_invalidate_range_start contains all addresses that
1908 * might be being invalidated. Note that it may include some false
1909 * positives, due to shortcuts when handing concurrent invalidations.
1911 if (unlikely(kvm->mmu_notifier_count) &&
1912 hva >= kvm->mmu_notifier_range_start &&
1913 hva < kvm->mmu_notifier_range_end)
1915 if (kvm->mmu_notifier_seq != mmu_seq)
1921 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1923 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1925 bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
1926 int kvm_set_irq_routing(struct kvm *kvm,
1927 const struct kvm_irq_routing_entry *entries,
1930 int kvm_set_routing_entry(struct kvm *kvm,
1931 struct kvm_kernel_irq_routing_entry *e,
1932 const struct kvm_irq_routing_entry *ue);
1933 void kvm_free_irq_routing(struct kvm *kvm);
1937 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1941 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1943 #ifdef CONFIG_HAVE_KVM_EVENTFD
1945 void kvm_eventfd_init(struct kvm *kvm);
1946 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1948 #ifdef CONFIG_HAVE_KVM_IRQFD
1949 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1950 void kvm_irqfd_release(struct kvm *kvm);
1951 void kvm_irq_routing_update(struct kvm *);
1953 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1958 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1963 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1965 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1970 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1972 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1973 static inline void kvm_irq_routing_update(struct kvm *kvm)
1978 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1983 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1985 void kvm_arch_irq_routing_update(struct kvm *kvm);
1987 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1990 * Ensure the rest of the request is published to kvm_check_request's
1991 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1994 set_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1997 static inline bool kvm_request_pending(struct kvm_vcpu *vcpu)
1999 return READ_ONCE(vcpu->requests);
2002 static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)
2004 return test_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
2007 static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu)
2009 clear_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
2012 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
2014 if (kvm_test_request(req, vcpu)) {
2015 kvm_clear_request(req, vcpu);
2018 * Ensure the rest of the request is visible to kvm_check_request's
2019 * caller. Paired with the smp_wmb in kvm_make_request.
2021 smp_mb__after_atomic();
2028 extern bool kvm_rebooting;
2030 extern unsigned int halt_poll_ns;
2031 extern unsigned int halt_poll_ns_grow;
2032 extern unsigned int halt_poll_ns_grow_start;
2033 extern unsigned int halt_poll_ns_shrink;
2036 const struct kvm_device_ops *ops;
2039 struct list_head vm_node;
2042 /* create, destroy, and name are mandatory */
2043 struct kvm_device_ops {
2047 * create is called holding kvm->lock and any operations not suitable
2048 * to do while holding the lock should be deferred to init (see
2051 int (*create)(struct kvm_device *dev, u32 type);
2054 * init is called after create if create is successful and is called
2055 * outside of holding kvm->lock.
2057 void (*init)(struct kvm_device *dev);
2060 * Destroy is responsible for freeing dev.
2062 * Destroy may be called before or after destructors are called
2063 * on emulated I/O regions, depending on whether a reference is
2064 * held by a vcpu or other kvm component that gets destroyed
2065 * after the emulated I/O.
2067 void (*destroy)(struct kvm_device *dev);
2070 * Release is an alternative method to free the device. It is
2071 * called when the device file descriptor is closed. Once
2072 * release is called, the destroy method will not be called
2073 * anymore as the device is removed from the device list of
2074 * the VM. kvm->lock is held.
2076 void (*release)(struct kvm_device *dev);
2078 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
2079 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
2080 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
2081 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
2083 int (*mmap)(struct kvm_device *dev, struct vm_area_struct *vma);
2086 void kvm_device_get(struct kvm_device *dev);
2087 void kvm_device_put(struct kvm_device *dev);
2088 struct kvm_device *kvm_device_from_filp(struct file *filp);
2089 int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type);
2090 void kvm_unregister_device_ops(u32 type);
2092 extern struct kvm_device_ops kvm_mpic_ops;
2093 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
2094 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
2096 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
2098 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
2100 vcpu->spin_loop.in_spin_loop = val;
2102 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
2104 vcpu->spin_loop.dy_eligible = val;
2107 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
2109 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
2113 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
2116 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
2118 static inline bool kvm_is_visible_memslot(struct kvm_memory_slot *memslot)
2120 return (memslot && memslot->id < KVM_USER_MEM_SLOTS &&
2121 !(memslot->flags & KVM_MEMSLOT_INVALID));
2124 struct kvm_vcpu *kvm_get_running_vcpu(void);
2125 struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void);
2127 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
2128 bool kvm_arch_has_irq_bypass(void);
2129 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
2130 struct irq_bypass_producer *);
2131 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
2132 struct irq_bypass_producer *);
2133 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
2134 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
2135 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
2136 uint32_t guest_irq, bool set);
2137 bool kvm_arch_irqfd_route_changed(struct kvm_kernel_irq_routing_entry *,
2138 struct kvm_kernel_irq_routing_entry *);
2139 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
2141 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
2142 /* If we wakeup during the poll time, was it a sucessful poll? */
2143 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
2145 return vcpu->valid_wakeup;
2149 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
2153 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
2155 #ifdef CONFIG_HAVE_KVM_NO_POLL
2156 /* Callback that tells if we must not poll */
2157 bool kvm_arch_no_poll(struct kvm_vcpu *vcpu);
2159 static inline bool kvm_arch_no_poll(struct kvm_vcpu *vcpu)
2163 #endif /* CONFIG_HAVE_KVM_NO_POLL */
2165 #ifdef CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL
2166 long kvm_arch_vcpu_async_ioctl(struct file *filp,
2167 unsigned int ioctl, unsigned long arg);
2169 static inline long kvm_arch_vcpu_async_ioctl(struct file *filp,
2173 return -ENOIOCTLCMD;
2175 #endif /* CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL */
2177 void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
2178 unsigned long start, unsigned long end);
2180 #ifdef CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE
2181 int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu);
2183 static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
2187 #endif /* CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE */
2189 typedef int (*kvm_vm_thread_fn_t)(struct kvm *kvm, uintptr_t data);
2191 int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn,
2192 uintptr_t data, const char *name,
2193 struct task_struct **thread_ptr);
2195 #ifdef CONFIG_KVM_XFER_TO_GUEST_WORK
2196 static inline void kvm_handle_signal_exit(struct kvm_vcpu *vcpu)
2198 vcpu->run->exit_reason = KVM_EXIT_INTR;
2199 vcpu->stat.signal_exits++;
2201 #endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */
2204 * This defines how many reserved entries we want to keep before we
2205 * kick the vcpu to the userspace to avoid dirty ring full. This
2206 * value can be tuned to higher if e.g. PML is enabled on the host.
2208 #define KVM_DIRTY_RING_RSVD_ENTRIES 64
2210 /* Max number of entries allowed for each kvm dirty ring */
2211 #define KVM_DIRTY_RING_MAX_ENTRIES 65536