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_MMU_RELOAD (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
157 #define KVM_REQ_UNBLOCK 2
158 #define KVM_REQ_UNHALT 3
159 #define KVM_REQ_VM_DEAD (4 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
160 #define KVM_REQ_GPC_INVALIDATE (5 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
161 #define KVM_REQUEST_ARCH_BASE 8
163 #define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
164 BUILD_BUG_ON((unsigned)(nr) >= (sizeof_field(struct kvm_vcpu, requests) * 8) - KVM_REQUEST_ARCH_BASE); \
165 (unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) | (flags)); \
167 #define KVM_ARCH_REQ(nr) KVM_ARCH_REQ_FLAGS(nr, 0)
169 bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
170 unsigned long *vcpu_bitmap);
171 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
172 bool kvm_make_all_cpus_request_except(struct kvm *kvm, unsigned int req,
173 struct kvm_vcpu *except);
174 bool kvm_make_cpus_request_mask(struct kvm *kvm, unsigned int req,
175 unsigned long *vcpu_bitmap);
177 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
178 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
180 extern struct mutex kvm_lock;
181 extern struct list_head vm_list;
183 struct kvm_io_range {
186 struct kvm_io_device *dev;
189 #define NR_IOBUS_DEVS 1000
194 struct kvm_io_range range[];
200 KVM_VIRTIO_CCW_NOTIFY_BUS,
205 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
206 int len, const void *val);
207 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
208 gpa_t addr, int len, const void *val, long cookie);
209 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
211 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
212 int len, struct kvm_io_device *dev);
213 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
214 struct kvm_io_device *dev);
215 struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx,
218 #ifdef CONFIG_KVM_ASYNC_PF
219 struct kvm_async_pf {
220 struct work_struct work;
221 struct list_head link;
222 struct list_head queue;
223 struct kvm_vcpu *vcpu;
224 struct mm_struct *mm;
227 struct kvm_arch_async_pf arch;
229 bool notpresent_injected;
232 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
233 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
234 bool kvm_setup_async_pf(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
235 unsigned long hva, struct kvm_arch_async_pf *arch);
236 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
239 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
240 struct kvm_gfn_range {
241 struct kvm_memory_slot *slot;
247 bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range);
248 bool kvm_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
249 bool kvm_test_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
250 bool kvm_set_spte_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
257 READING_SHADOW_PAGE_TABLES,
260 #define KVM_UNMAPPED_PAGE ((void *) 0x500 + POISON_POINTER_DELTA)
262 struct kvm_host_map {
264 * Only valid if the 'pfn' is managed by the host kernel (i.e. There is
265 * a 'struct page' for it. When using mem= kernel parameter some memory
266 * can be used as guest memory but they are not managed by host
268 * If 'pfn' is not managed by the host kernel, this field is
269 * initialized to KVM_UNMAPPED_PAGE.
278 * Used to check if the mapping is valid or not. Never use 'kvm_host_map'
279 * directly to check for that.
281 static inline bool kvm_vcpu_mapped(struct kvm_host_map *map)
286 static inline bool kvm_vcpu_can_poll(ktime_t cur, ktime_t stop)
288 return single_task_running() && !need_resched() && ktime_before(cur, stop);
292 * Sometimes a large or cross-page mmio needs to be broken up into separate
293 * exits for userspace servicing.
295 struct kvm_mmio_fragment {
303 #ifdef CONFIG_PREEMPT_NOTIFIERS
304 struct preempt_notifier preempt_notifier;
307 int vcpu_id; /* id given by userspace at creation */
308 int vcpu_idx; /* index in kvm->vcpus array */
312 unsigned long guest_debug;
317 #ifndef __KVM_HAVE_ARCH_WQP
320 struct pid __rcu *pid;
323 unsigned int halt_poll_ns;
326 #ifdef CONFIG_HAS_IOMEM
328 int mmio_read_completed;
330 int mmio_cur_fragment;
331 int mmio_nr_fragments;
332 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
335 #ifdef CONFIG_KVM_ASYNC_PF
338 struct list_head queue;
339 struct list_head done;
344 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
346 * Cpu relax intercept or pause loop exit optimization
347 * in_spin_loop: set when a vcpu does a pause loop exit
348 * or cpu relax intercepted.
349 * dy_eligible: indicates whether vcpu is eligible for directed yield.
358 struct kvm_vcpu_arch arch;
359 struct kvm_vcpu_stat stat;
360 char stats_id[KVM_STATS_NAME_SIZE];
361 struct kvm_dirty_ring dirty_ring;
364 * The most recently used memslot by this vCPU and the slots generation
365 * for which it is valid.
366 * No wraparound protection is needed since generations won't overflow in
367 * thousands of years, even assuming 1M memslot operations per second.
369 struct kvm_memory_slot *last_used_slot;
370 u64 last_used_slot_gen;
374 * Start accounting time towards a guest.
375 * Must be called before entering guest context.
377 static __always_inline void guest_timing_enter_irqoff(void)
380 * This is running in ioctl context so its safe to assume that it's the
381 * stime pending cputime to flush.
383 instrumentation_begin();
384 vtime_account_guest_enter();
385 instrumentation_end();
389 * Enter guest context and enter an RCU extended quiescent state.
391 * Between guest_context_enter_irqoff() and guest_context_exit_irqoff() it is
392 * unsafe to use any code which may directly or indirectly use RCU, tracing
393 * (including IRQ flag tracing), or lockdep. All code in this period must be
394 * non-instrumentable.
396 static __always_inline void guest_context_enter_irqoff(void)
399 * KVM does not hold any references to rcu protected data when it
400 * switches CPU into a guest mode. In fact switching to a guest mode
401 * is very similar to exiting to userspace from rcu point of view. In
402 * addition CPU may stay in a guest mode for quite a long time (up to
403 * one time slice). Lets treat guest mode as quiescent state, just like
404 * we do with user-mode execution.
406 if (!context_tracking_guest_enter()) {
407 instrumentation_begin();
408 rcu_virt_note_context_switch(smp_processor_id());
409 instrumentation_end();
414 * Deprecated. Architectures should move to guest_timing_enter_irqoff() and
415 * guest_state_enter_irqoff().
417 static __always_inline void guest_enter_irqoff(void)
419 guest_timing_enter_irqoff();
420 guest_context_enter_irqoff();
424 * guest_state_enter_irqoff - Fixup state when entering a guest
426 * Entry to a guest will enable interrupts, but the kernel state is interrupts
427 * disabled when this is invoked. Also tell RCU about it.
429 * 1) Trace interrupts on state
430 * 2) Invoke context tracking if enabled to adjust RCU state
431 * 3) Tell lockdep that interrupts are enabled
433 * Invoked from architecture specific code before entering a guest.
434 * Must be called with interrupts disabled and the caller must be
435 * non-instrumentable.
436 * The caller has to invoke guest_timing_enter_irqoff() before this.
438 * Note: this is analogous to exit_to_user_mode().
440 static __always_inline void guest_state_enter_irqoff(void)
442 instrumentation_begin();
443 trace_hardirqs_on_prepare();
444 lockdep_hardirqs_on_prepare(CALLER_ADDR0);
445 instrumentation_end();
447 guest_context_enter_irqoff();
448 lockdep_hardirqs_on(CALLER_ADDR0);
452 * Exit guest context and exit an RCU extended quiescent state.
454 * Between guest_context_enter_irqoff() and guest_context_exit_irqoff() it is
455 * unsafe to use any code which may directly or indirectly use RCU, tracing
456 * (including IRQ flag tracing), or lockdep. All code in this period must be
457 * non-instrumentable.
459 static __always_inline void guest_context_exit_irqoff(void)
461 context_tracking_guest_exit();
465 * Stop accounting time towards a guest.
466 * Must be called after exiting guest context.
468 static __always_inline void guest_timing_exit_irqoff(void)
470 instrumentation_begin();
471 /* Flush the guest cputime we spent on the guest */
472 vtime_account_guest_exit();
473 instrumentation_end();
477 * Deprecated. Architectures should move to guest_state_exit_irqoff() and
478 * guest_timing_exit_irqoff().
480 static __always_inline void guest_exit_irqoff(void)
482 guest_context_exit_irqoff();
483 guest_timing_exit_irqoff();
486 static inline void guest_exit(void)
490 local_irq_save(flags);
492 local_irq_restore(flags);
496 * guest_state_exit_irqoff - Establish state when returning from guest mode
498 * Entry from a guest disables interrupts, but guest mode is traced as
499 * interrupts enabled. Also with NO_HZ_FULL RCU might be idle.
501 * 1) Tell lockdep that interrupts are disabled
502 * 2) Invoke context tracking if enabled to reactivate RCU
503 * 3) Trace interrupts off state
505 * Invoked from architecture specific code after exiting a guest.
506 * Must be invoked with interrupts disabled and the caller must be
507 * non-instrumentable.
508 * The caller has to invoke guest_timing_exit_irqoff() after this.
510 * Note: this is analogous to enter_from_user_mode().
512 static __always_inline void guest_state_exit_irqoff(void)
514 lockdep_hardirqs_off(CALLER_ADDR0);
515 guest_context_exit_irqoff();
517 instrumentation_begin();
518 trace_hardirqs_off_finish();
519 instrumentation_end();
522 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
525 * The memory barrier ensures a previous write to vcpu->requests cannot
526 * be reordered with the read of vcpu->mode. It pairs with the general
527 * memory barrier following the write of vcpu->mode in VCPU RUN.
529 smp_mb__before_atomic();
530 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
534 * Some of the bitops functions do not support too long bitmaps.
535 * This number must be determined not to exceed such limits.
537 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
540 * Since at idle each memslot belongs to two memslot sets it has to contain
541 * two embedded nodes for each data structure that it forms a part of.
543 * Two memslot sets (one active and one inactive) are necessary so the VM
544 * continues to run on one memslot set while the other is being modified.
546 * These two memslot sets normally point to the same set of memslots.
547 * They can, however, be desynchronized when performing a memslot management
548 * operation by replacing the memslot to be modified by its copy.
549 * After the operation is complete, both memslot sets once again point to
550 * the same, common set of memslot data.
552 * The memslots themselves are independent of each other so they can be
553 * individually added or deleted.
555 struct kvm_memory_slot {
556 struct hlist_node id_node[2];
557 struct interval_tree_node hva_node[2];
558 struct rb_node gfn_node[2];
560 unsigned long npages;
561 unsigned long *dirty_bitmap;
562 struct kvm_arch_memory_slot arch;
563 unsigned long userspace_addr;
569 static inline bool kvm_slot_dirty_track_enabled(const struct kvm_memory_slot *slot)
571 return slot->flags & KVM_MEM_LOG_DIRTY_PAGES;
574 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
576 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
579 static inline unsigned long *kvm_second_dirty_bitmap(struct kvm_memory_slot *memslot)
581 unsigned long len = kvm_dirty_bitmap_bytes(memslot);
583 return memslot->dirty_bitmap + len / sizeof(*memslot->dirty_bitmap);
586 #ifndef KVM_DIRTY_LOG_MANUAL_CAPS
587 #define KVM_DIRTY_LOG_MANUAL_CAPS KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE
590 struct kvm_s390_adapter_int {
603 struct kvm_xen_evtchn {
609 struct kvm_kernel_irq_routing_entry {
612 int (*set)(struct kvm_kernel_irq_routing_entry *e,
613 struct kvm *kvm, int irq_source_id, int level,
627 struct kvm_s390_adapter_int adapter;
628 struct kvm_hv_sint hv_sint;
629 struct kvm_xen_evtchn xen_evtchn;
631 struct hlist_node link;
634 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
635 struct kvm_irq_routing_table {
636 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
639 * Array indexed by gsi. Each entry contains list of irq chips
640 * the gsi is connected to.
642 struct hlist_head map[];
646 #ifndef KVM_PRIVATE_MEM_SLOTS
647 #define KVM_PRIVATE_MEM_SLOTS 0
650 #define KVM_MEM_SLOTS_NUM SHRT_MAX
651 #define KVM_USER_MEM_SLOTS (KVM_MEM_SLOTS_NUM - KVM_PRIVATE_MEM_SLOTS)
653 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
654 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
660 struct kvm_memslots {
662 atomic_long_t last_used_slot;
663 struct rb_root_cached hva_tree;
664 struct rb_root gfn_tree;
666 * The mapping table from slot id to memslot.
668 * 7-bit bucket count matches the size of the old id to index array for
669 * 512 slots, while giving good performance with this slot count.
670 * Higher bucket counts bring only small performance improvements but
671 * always result in higher memory usage (even for lower memslot counts).
673 DECLARE_HASHTABLE(id_hash, 7);
678 #ifdef KVM_HAVE_MMU_RWLOCK
682 #endif /* KVM_HAVE_MMU_RWLOCK */
684 struct mutex slots_lock;
687 * Protects the arch-specific fields of struct kvm_memory_slots in
688 * use by the VM. To be used under the slots_lock (above) or in a
689 * kvm->srcu critical section where acquiring the slots_lock would
690 * lead to deadlock with the synchronize_srcu in
691 * install_new_memslots.
693 struct mutex slots_arch_lock;
694 struct mm_struct *mm; /* userspace tied to this vm */
695 unsigned long nr_memslot_pages;
696 /* The two memslot sets - active and inactive (per address space) */
697 struct kvm_memslots __memslots[KVM_ADDRESS_SPACE_NUM][2];
698 /* The current active memslot set for each address space */
699 struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
700 struct xarray vcpu_array;
702 /* Used to wait for completion of MMU notifiers. */
703 spinlock_t mn_invalidate_lock;
704 unsigned long mn_active_invalidate_count;
705 struct rcuwait mn_memslots_update_rcuwait;
707 /* For management / invalidation of gfn_to_pfn_caches */
709 struct list_head gpc_list;
712 * created_vcpus is protected by kvm->lock, and is incremented
713 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
714 * incremented after storing the kvm_vcpu pointer in vcpus,
715 * and is accessed atomically.
717 atomic_t online_vcpus;
719 int last_boosted_vcpu;
720 struct list_head vm_list;
722 struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
723 #ifdef CONFIG_HAVE_KVM_EVENTFD
726 struct list_head items;
727 struct list_head resampler_list;
728 struct mutex resampler_lock;
730 struct list_head ioeventfds;
732 struct kvm_vm_stat stat;
733 struct kvm_arch arch;
734 refcount_t users_count;
735 #ifdef CONFIG_KVM_MMIO
736 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
737 spinlock_t ring_lock;
738 struct list_head coalesced_zones;
741 struct mutex irq_lock;
742 #ifdef CONFIG_HAVE_KVM_IRQCHIP
744 * Update side is protected by irq_lock.
746 struct kvm_irq_routing_table __rcu *irq_routing;
748 #ifdef CONFIG_HAVE_KVM_IRQFD
749 struct hlist_head irq_ack_notifier_list;
752 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
753 struct mmu_notifier mmu_notifier;
754 unsigned long mmu_notifier_seq;
755 long mmu_notifier_count;
756 unsigned long mmu_notifier_range_start;
757 unsigned long mmu_notifier_range_end;
759 struct list_head devices;
760 u64 manual_dirty_log_protect;
761 struct dentry *debugfs_dentry;
762 struct kvm_stat_data **debugfs_stat_data;
763 struct srcu_struct srcu;
764 struct srcu_struct irq_srcu;
766 unsigned int max_halt_poll_ns;
771 #ifdef CONFIG_HAVE_KVM_PM_NOTIFIER
772 struct notifier_block pm_notifier;
774 char stats_id[KVM_STATS_NAME_SIZE];
777 #define kvm_err(fmt, ...) \
778 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
779 #define kvm_info(fmt, ...) \
780 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
781 #define kvm_debug(fmt, ...) \
782 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
783 #define kvm_debug_ratelimited(fmt, ...) \
784 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
786 #define kvm_pr_unimpl(fmt, ...) \
787 pr_err_ratelimited("kvm [%i]: " fmt, \
788 task_tgid_nr(current), ## __VA_ARGS__)
790 /* The guest did something we don't support. */
791 #define vcpu_unimpl(vcpu, fmt, ...) \
792 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
793 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
795 #define vcpu_debug(vcpu, fmt, ...) \
796 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
797 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
798 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
800 #define vcpu_err(vcpu, fmt, ...) \
801 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
803 static inline void kvm_vm_dead(struct kvm *kvm)
806 kvm_make_all_cpus_request(kvm, KVM_REQ_VM_DEAD);
809 static inline void kvm_vm_bugged(struct kvm *kvm)
811 kvm->vm_bugged = true;
816 #define KVM_BUG(cond, kvm, fmt...) \
818 int __ret = (cond); \
820 if (WARN_ONCE(__ret && !(kvm)->vm_bugged, fmt)) \
821 kvm_vm_bugged(kvm); \
825 #define KVM_BUG_ON(cond, kvm) \
827 int __ret = (cond); \
829 if (WARN_ON_ONCE(__ret && !(kvm)->vm_bugged)) \
830 kvm_vm_bugged(kvm); \
834 static inline bool kvm_dirty_log_manual_protect_and_init_set(struct kvm *kvm)
836 return !!(kvm->manual_dirty_log_protect & KVM_DIRTY_LOG_INITIALLY_SET);
839 static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx)
841 return srcu_dereference_check(kvm->buses[idx], &kvm->srcu,
842 lockdep_is_held(&kvm->slots_lock) ||
843 !refcount_read(&kvm->users_count));
846 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
848 int num_vcpus = atomic_read(&kvm->online_vcpus);
849 i = array_index_nospec(i, num_vcpus);
851 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. */
853 return xa_load(&kvm->vcpu_array, i);
856 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
857 xa_for_each_range(&kvm->vcpu_array, idx, vcpup, 0, \
858 (atomic_read(&kvm->online_vcpus) - 1))
860 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
862 struct kvm_vcpu *vcpu = NULL;
867 if (id < KVM_MAX_VCPUS)
868 vcpu = kvm_get_vcpu(kvm, id);
869 if (vcpu && vcpu->vcpu_id == id)
871 kvm_for_each_vcpu(i, vcpu, kvm)
872 if (vcpu->vcpu_id == id)
877 static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu)
879 return vcpu->vcpu_idx;
882 void kvm_destroy_vcpus(struct kvm *kvm);
884 void vcpu_load(struct kvm_vcpu *vcpu);
885 void vcpu_put(struct kvm_vcpu *vcpu);
887 #ifdef __KVM_HAVE_IOAPIC
888 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
889 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
891 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
894 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
899 #ifdef CONFIG_HAVE_KVM_IRQFD
900 int kvm_irqfd_init(void);
901 void kvm_irqfd_exit(void);
903 static inline int kvm_irqfd_init(void)
908 static inline void kvm_irqfd_exit(void)
912 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
913 struct module *module);
916 void kvm_get_kvm(struct kvm *kvm);
917 bool kvm_get_kvm_safe(struct kvm *kvm);
918 void kvm_put_kvm(struct kvm *kvm);
919 bool file_is_kvm(struct file *file);
920 void kvm_put_kvm_no_destroy(struct kvm *kvm);
922 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
924 as_id = array_index_nospec(as_id, KVM_ADDRESS_SPACE_NUM);
925 return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu,
926 lockdep_is_held(&kvm->slots_lock) ||
927 !refcount_read(&kvm->users_count));
930 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
932 return __kvm_memslots(kvm, 0);
935 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
937 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
939 return __kvm_memslots(vcpu->kvm, as_id);
942 static inline bool kvm_memslots_empty(struct kvm_memslots *slots)
944 return RB_EMPTY_ROOT(&slots->gfn_tree);
947 #define kvm_for_each_memslot(memslot, bkt, slots) \
948 hash_for_each(slots->id_hash, bkt, memslot, id_node[slots->node_idx]) \
949 if (WARN_ON_ONCE(!memslot->npages)) { \
953 struct kvm_memory_slot *id_to_memslot(struct kvm_memslots *slots, int id)
955 struct kvm_memory_slot *slot;
956 int idx = slots->node_idx;
958 hash_for_each_possible(slots->id_hash, slot, id_node[idx], id) {
966 /* Iterator used for walking memslots that overlap a gfn range. */
967 struct kvm_memslot_iter {
968 struct kvm_memslots *slots;
969 struct rb_node *node;
970 struct kvm_memory_slot *slot;
973 static inline void kvm_memslot_iter_next(struct kvm_memslot_iter *iter)
975 iter->node = rb_next(iter->node);
979 iter->slot = container_of(iter->node, struct kvm_memory_slot, gfn_node[iter->slots->node_idx]);
982 static inline void kvm_memslot_iter_start(struct kvm_memslot_iter *iter,
983 struct kvm_memslots *slots,
986 int idx = slots->node_idx;
988 struct kvm_memory_slot *slot;
993 * Find the so called "upper bound" of a key - the first node that has
994 * its key strictly greater than the searched one (the start gfn in our case).
997 for (tmp = slots->gfn_tree.rb_node; tmp; ) {
998 slot = container_of(tmp, struct kvm_memory_slot, gfn_node[idx]);
999 if (start < slot->base_gfn) {
1003 tmp = tmp->rb_right;
1008 * Find the slot with the lowest gfn that can possibly intersect with
1009 * the range, so we'll ideally have slot start <= range start
1013 * A NULL previous node means that the very first slot
1014 * already has a higher start gfn.
1015 * In this case slot start > range start.
1017 tmp = rb_prev(iter->node);
1021 /* a NULL node below means no slots */
1022 iter->node = rb_last(&slots->gfn_tree);
1026 iter->slot = container_of(iter->node, struct kvm_memory_slot, gfn_node[idx]);
1029 * It is possible in the slot start < range start case that the
1030 * found slot ends before or at range start (slot end <= range start)
1031 * and so it does not overlap the requested range.
1033 * In such non-overlapping case the next slot (if it exists) will
1034 * already have slot start > range start, otherwise the logic above
1035 * would have found it instead of the current slot.
1037 if (iter->slot->base_gfn + iter->slot->npages <= start)
1038 kvm_memslot_iter_next(iter);
1042 static inline bool kvm_memslot_iter_is_valid(struct kvm_memslot_iter *iter, gfn_t end)
1048 * If this slot starts beyond or at the end of the range so does
1051 return iter->slot->base_gfn < end;
1054 /* Iterate over each memslot at least partially intersecting [start, end) range */
1055 #define kvm_for_each_memslot_in_gfn_range(iter, slots, start, end) \
1056 for (kvm_memslot_iter_start(iter, slots, start); \
1057 kvm_memslot_iter_is_valid(iter, end); \
1058 kvm_memslot_iter_next(iter))
1061 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
1062 * - create a new memory slot
1063 * - delete an existing memory slot
1064 * - modify an existing memory slot
1065 * -- move it in the guest physical memory space
1066 * -- just change its flags
1068 * Since flags can be changed by some of these operations, the following
1069 * differentiation is the best we can do for __kvm_set_memory_region():
1071 enum kvm_mr_change {
1078 int kvm_set_memory_region(struct kvm *kvm,
1079 const struct kvm_userspace_memory_region *mem);
1080 int __kvm_set_memory_region(struct kvm *kvm,
1081 const struct kvm_userspace_memory_region *mem);
1082 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot);
1083 void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen);
1084 int kvm_arch_prepare_memory_region(struct kvm *kvm,
1085 const struct kvm_memory_slot *old,
1086 struct kvm_memory_slot *new,
1087 enum kvm_mr_change change);
1088 void kvm_arch_commit_memory_region(struct kvm *kvm,
1089 struct kvm_memory_slot *old,
1090 const struct kvm_memory_slot *new,
1091 enum kvm_mr_change change);
1092 /* flush all memory translations */
1093 void kvm_arch_flush_shadow_all(struct kvm *kvm);
1094 /* flush memory translations pointing to 'slot' */
1095 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
1096 struct kvm_memory_slot *slot);
1098 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
1099 struct page **pages, int nr_pages);
1101 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
1102 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
1103 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
1104 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
1105 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
1107 void kvm_release_page_clean(struct page *page);
1108 void kvm_release_page_dirty(struct page *page);
1109 void kvm_set_page_accessed(struct page *page);
1111 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
1112 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1114 kvm_pfn_t gfn_to_pfn_memslot(const struct kvm_memory_slot *slot, gfn_t gfn);
1115 kvm_pfn_t gfn_to_pfn_memslot_atomic(const struct kvm_memory_slot *slot, gfn_t gfn);
1116 kvm_pfn_t __gfn_to_pfn_memslot(const struct kvm_memory_slot *slot, gfn_t gfn,
1117 bool atomic, bool *async, bool write_fault,
1118 bool *writable, hva_t *hva);
1120 void kvm_release_pfn_clean(kvm_pfn_t pfn);
1121 void kvm_release_pfn_dirty(kvm_pfn_t pfn);
1122 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
1123 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
1125 void kvm_release_pfn(kvm_pfn_t pfn, bool dirty);
1126 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
1128 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
1129 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1130 void *data, unsigned long len);
1131 int kvm_read_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1132 void *data, unsigned int offset,
1134 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
1135 int offset, int len);
1136 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
1138 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1139 void *data, unsigned long len);
1140 int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1141 void *data, unsigned int offset,
1143 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1144 gpa_t gpa, unsigned long len);
1146 #define __kvm_get_guest(kvm, gfn, offset, v) \
1148 unsigned long __addr = gfn_to_hva(kvm, gfn); \
1149 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
1150 int __ret = -EFAULT; \
1152 if (!kvm_is_error_hva(__addr)) \
1153 __ret = get_user(v, __uaddr); \
1157 #define kvm_get_guest(kvm, gpa, v) \
1159 gpa_t __gpa = gpa; \
1160 struct kvm *__kvm = kvm; \
1162 __kvm_get_guest(__kvm, __gpa >> PAGE_SHIFT, \
1163 offset_in_page(__gpa), v); \
1166 #define __kvm_put_guest(kvm, gfn, offset, v) \
1168 unsigned long __addr = gfn_to_hva(kvm, gfn); \
1169 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
1170 int __ret = -EFAULT; \
1172 if (!kvm_is_error_hva(__addr)) \
1173 __ret = put_user(v, __uaddr); \
1175 mark_page_dirty(kvm, gfn); \
1179 #define kvm_put_guest(kvm, gpa, v) \
1181 gpa_t __gpa = gpa; \
1182 struct kvm *__kvm = kvm; \
1184 __kvm_put_guest(__kvm, __gpa >> PAGE_SHIFT, \
1185 offset_in_page(__gpa), v); \
1188 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
1189 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
1190 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
1191 bool kvm_vcpu_is_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1192 unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn);
1193 void mark_page_dirty_in_slot(struct kvm *kvm, const struct kvm_memory_slot *memslot, gfn_t gfn);
1194 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
1196 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
1197 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
1198 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
1199 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1200 int kvm_vcpu_map(struct kvm_vcpu *vcpu, gpa_t gpa, struct kvm_host_map *map);
1201 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
1202 void kvm_vcpu_unmap(struct kvm_vcpu *vcpu, struct kvm_host_map *map, bool dirty);
1203 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
1204 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
1205 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
1207 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
1209 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
1211 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
1212 int offset, int len);
1213 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
1215 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
1218 * kvm_gfn_to_pfn_cache_init - prepare a cached kernel mapping and HPA for a
1219 * given guest physical address.
1221 * @kvm: pointer to kvm instance.
1222 * @gpc: struct gfn_to_pfn_cache object.
1223 * @vcpu: vCPU to be used for marking pages dirty and to be woken on
1225 * @guest_uses_pa: indicates that the resulting host physical PFN is used while
1226 * @vcpu is IN_GUEST_MODE so invalidations should wake it.
1227 * @kernel_map: requests a kernel virtual mapping (kmap / memremap).
1228 * @gpa: guest physical address to map.
1229 * @len: sanity check; the range being access must fit a single page.
1230 * @dirty: mark the cache dirty immediately.
1232 * @return: 0 for success.
1233 * -EINVAL for a mapping which would cross a page boundary.
1234 * -EFAULT for an untranslatable guest physical address.
1236 * This primes a gfn_to_pfn_cache and links it into the @kvm's list for
1237 * invalidations to be processed. Invalidation callbacks to @vcpu using
1238 * %KVM_REQ_GPC_INVALIDATE will occur only for MMU notifiers, not for KVM
1239 * memslot changes. Callers are required to use kvm_gfn_to_pfn_cache_check()
1240 * to ensure that the cache is valid before accessing the target page.
1242 int kvm_gfn_to_pfn_cache_init(struct kvm *kvm, struct gfn_to_pfn_cache *gpc,
1243 struct kvm_vcpu *vcpu, bool guest_uses_pa,
1244 bool kernel_map, gpa_t gpa, unsigned long len,
1248 * kvm_gfn_to_pfn_cache_check - check validity of a gfn_to_pfn_cache.
1250 * @kvm: pointer to kvm instance.
1251 * @gpc: struct gfn_to_pfn_cache object.
1252 * @gpa: current guest physical address to map.
1253 * @len: sanity check; the range being access must fit a single page.
1254 * @dirty: mark the cache dirty immediately.
1256 * @return: %true if the cache is still valid and the address matches.
1257 * %false if the cache is not valid.
1259 * Callers outside IN_GUEST_MODE context should hold a read lock on @gpc->lock
1260 * while calling this function, and then continue to hold the lock until the
1261 * access is complete.
1263 * Callers in IN_GUEST_MODE may do so without locking, although they should
1264 * still hold a read lock on kvm->scru for the memslot checks.
1266 bool kvm_gfn_to_pfn_cache_check(struct kvm *kvm, struct gfn_to_pfn_cache *gpc,
1267 gpa_t gpa, unsigned long len);
1270 * kvm_gfn_to_pfn_cache_refresh - update a previously initialized cache.
1272 * @kvm: pointer to kvm instance.
1273 * @gpc: struct gfn_to_pfn_cache object.
1274 * @gpa: updated guest physical address to map.
1275 * @len: sanity check; the range being access must fit a single page.
1276 * @dirty: mark the cache dirty immediately.
1278 * @return: 0 for success.
1279 * -EINVAL for a mapping which would cross a page boundary.
1280 * -EFAULT for an untranslatable guest physical address.
1282 * This will attempt to refresh a gfn_to_pfn_cache. Note that a successful
1283 * returm from this function does not mean the page can be immediately
1284 * accessed because it may have raced with an invalidation. Callers must
1285 * still lock and check the cache status, as this function does not return
1286 * with the lock still held to permit access.
1288 int kvm_gfn_to_pfn_cache_refresh(struct kvm *kvm, struct gfn_to_pfn_cache *gpc,
1289 gpa_t gpa, unsigned long len, bool dirty);
1292 * kvm_gfn_to_pfn_cache_unmap - temporarily unmap a gfn_to_pfn_cache.
1294 * @kvm: pointer to kvm instance.
1295 * @gpc: struct gfn_to_pfn_cache object.
1297 * This unmaps the referenced page and marks it dirty, if appropriate. The
1298 * cache is left in the invalid state but at least the mapping from GPA to
1299 * userspace HVA will remain cached and can be reused on a subsequent
1302 void kvm_gfn_to_pfn_cache_unmap(struct kvm *kvm, struct gfn_to_pfn_cache *gpc);
1305 * kvm_gfn_to_pfn_cache_destroy - destroy and unlink a gfn_to_pfn_cache.
1307 * @kvm: pointer to kvm instance.
1308 * @gpc: struct gfn_to_pfn_cache object.
1310 * This removes a cache from the @kvm's list to be processed on MMU notifier
1313 void kvm_gfn_to_pfn_cache_destroy(struct kvm *kvm, struct gfn_to_pfn_cache *gpc);
1315 void kvm_sigset_activate(struct kvm_vcpu *vcpu);
1316 void kvm_sigset_deactivate(struct kvm_vcpu *vcpu);
1318 void kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1319 bool kvm_vcpu_block(struct kvm_vcpu *vcpu);
1320 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
1321 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
1322 bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
1323 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
1324 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
1325 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool usermode_vcpu_not_eligible);
1327 void kvm_flush_remote_tlbs(struct kvm *kvm);
1328 void kvm_reload_remote_mmus(struct kvm *kvm);
1330 #ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE
1331 int kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache *mc, int min);
1332 int kvm_mmu_memory_cache_nr_free_objects(struct kvm_mmu_memory_cache *mc);
1333 void kvm_mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc);
1334 void *kvm_mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc);
1337 void kvm_inc_notifier_count(struct kvm *kvm, unsigned long start,
1339 void kvm_dec_notifier_count(struct kvm *kvm, unsigned long start,
1342 long kvm_arch_dev_ioctl(struct file *filp,
1343 unsigned int ioctl, unsigned long arg);
1344 long kvm_arch_vcpu_ioctl(struct file *filp,
1345 unsigned int ioctl, unsigned long arg);
1346 vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
1348 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
1350 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
1351 struct kvm_memory_slot *slot,
1353 unsigned long mask);
1354 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot);
1356 #ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
1357 void kvm_arch_flush_remote_tlbs_memslot(struct kvm *kvm,
1358 const struct kvm_memory_slot *memslot);
1359 #else /* !CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
1360 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log);
1361 int kvm_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log,
1362 int *is_dirty, struct kvm_memory_slot **memslot);
1365 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
1367 int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
1368 struct kvm_enable_cap *cap);
1369 long kvm_arch_vm_ioctl(struct file *filp,
1370 unsigned int ioctl, unsigned long arg);
1372 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
1373 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
1375 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1376 struct kvm_translation *tr);
1378 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
1379 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
1380 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1381 struct kvm_sregs *sregs);
1382 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1383 struct kvm_sregs *sregs);
1384 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
1385 struct kvm_mp_state *mp_state);
1386 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
1387 struct kvm_mp_state *mp_state);
1388 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
1389 struct kvm_guest_debug *dbg);
1390 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu);
1392 int kvm_arch_init(void *opaque);
1393 void kvm_arch_exit(void);
1395 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
1397 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
1398 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
1399 int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id);
1400 int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu);
1401 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
1402 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
1404 #ifdef CONFIG_HAVE_KVM_PM_NOTIFIER
1405 int kvm_arch_pm_notifier(struct kvm *kvm, unsigned long state);
1408 #ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
1409 void kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu, struct dentry *debugfs_dentry);
1412 int kvm_arch_hardware_enable(void);
1413 void kvm_arch_hardware_disable(void);
1414 int kvm_arch_hardware_setup(void *opaque);
1415 void kvm_arch_hardware_unsetup(void);
1416 int kvm_arch_check_processor_compat(void *opaque);
1417 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
1418 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
1419 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
1420 bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu);
1421 bool kvm_arch_dy_has_pending_interrupt(struct kvm_vcpu *vcpu);
1422 int kvm_arch_post_init_vm(struct kvm *kvm);
1423 void kvm_arch_pre_destroy_vm(struct kvm *kvm);
1424 int kvm_arch_create_vm_debugfs(struct kvm *kvm);
1426 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
1428 * All architectures that want to use vzalloc currently also
1429 * need their own kvm_arch_alloc_vm implementation.
1431 static inline struct kvm *kvm_arch_alloc_vm(void)
1433 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
1437 static inline void __kvm_arch_free_vm(struct kvm *kvm)
1442 #ifndef __KVM_HAVE_ARCH_VM_FREE
1443 static inline void kvm_arch_free_vm(struct kvm *kvm)
1445 __kvm_arch_free_vm(kvm);
1449 #ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1450 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1456 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
1457 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
1458 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
1459 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
1461 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
1465 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
1469 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
1474 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
1475 void kvm_arch_start_assignment(struct kvm *kvm);
1476 void kvm_arch_end_assignment(struct kvm *kvm);
1477 bool kvm_arch_has_assigned_device(struct kvm *kvm);
1479 static inline void kvm_arch_start_assignment(struct kvm *kvm)
1483 static inline void kvm_arch_end_assignment(struct kvm *kvm)
1487 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
1493 static inline struct rcuwait *kvm_arch_vcpu_get_wait(struct kvm_vcpu *vcpu)
1495 #ifdef __KVM_HAVE_ARCH_WQP
1496 return vcpu->arch.waitp;
1503 * Wake a vCPU if necessary, but don't do any stats/metadata updates. Returns
1504 * true if the vCPU was blocking and was awakened, false otherwise.
1506 static inline bool __kvm_vcpu_wake_up(struct kvm_vcpu *vcpu)
1508 return !!rcuwait_wake_up(kvm_arch_vcpu_get_wait(vcpu));
1511 static inline bool kvm_vcpu_is_blocking(struct kvm_vcpu *vcpu)
1513 return rcuwait_active(kvm_arch_vcpu_get_wait(vcpu));
1516 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
1518 * returns true if the virtual interrupt controller is initialized and
1519 * ready to accept virtual IRQ. On some architectures the virtual interrupt
1520 * controller is dynamically instantiated and this is not always true.
1522 bool kvm_arch_intc_initialized(struct kvm *kvm);
1524 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
1530 #ifdef CONFIG_GUEST_PERF_EVENTS
1531 unsigned long kvm_arch_vcpu_get_ip(struct kvm_vcpu *vcpu);
1533 void kvm_register_perf_callbacks(unsigned int (*pt_intr_handler)(void));
1534 void kvm_unregister_perf_callbacks(void);
1536 static inline void kvm_register_perf_callbacks(void *ign) {}
1537 static inline void kvm_unregister_perf_callbacks(void) {}
1538 #endif /* CONFIG_GUEST_PERF_EVENTS */
1540 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
1541 void kvm_arch_destroy_vm(struct kvm *kvm);
1542 void kvm_arch_sync_events(struct kvm *kvm);
1544 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
1546 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
1547 bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
1549 struct kvm_irq_ack_notifier {
1550 struct hlist_node link;
1552 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
1555 int kvm_irq_map_gsi(struct kvm *kvm,
1556 struct kvm_kernel_irq_routing_entry *entries, int gsi);
1557 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
1559 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
1561 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
1562 int irq_source_id, int level, bool line_status);
1563 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
1564 struct kvm *kvm, int irq_source_id,
1565 int level, bool line_status);
1566 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
1567 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
1568 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
1569 void kvm_register_irq_ack_notifier(struct kvm *kvm,
1570 struct kvm_irq_ack_notifier *kian);
1571 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
1572 struct kvm_irq_ack_notifier *kian);
1573 int kvm_request_irq_source_id(struct kvm *kvm);
1574 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
1575 bool kvm_arch_irqfd_allowed(struct kvm *kvm, struct kvm_irqfd *args);
1578 * Returns a pointer to the memslot if it contains gfn.
1579 * Otherwise returns NULL.
1581 static inline struct kvm_memory_slot *
1582 try_get_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1587 if (gfn >= slot->base_gfn && gfn < slot->base_gfn + slot->npages)
1594 * Returns a pointer to the memslot that contains gfn. Otherwise returns NULL.
1596 * With "approx" set returns the memslot also when the address falls
1597 * in a hole. In that case one of the memslots bordering the hole is
1600 static inline struct kvm_memory_slot *
1601 search_memslots(struct kvm_memslots *slots, gfn_t gfn, bool approx)
1603 struct kvm_memory_slot *slot;
1604 struct rb_node *node;
1605 int idx = slots->node_idx;
1608 for (node = slots->gfn_tree.rb_node; node; ) {
1609 slot = container_of(node, struct kvm_memory_slot, gfn_node[idx]);
1610 if (gfn >= slot->base_gfn) {
1611 if (gfn < slot->base_gfn + slot->npages)
1613 node = node->rb_right;
1615 node = node->rb_left;
1618 return approx ? slot : NULL;
1621 static inline struct kvm_memory_slot *
1622 ____gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn, bool approx)
1624 struct kvm_memory_slot *slot;
1626 slot = (struct kvm_memory_slot *)atomic_long_read(&slots->last_used_slot);
1627 slot = try_get_memslot(slot, gfn);
1631 slot = search_memslots(slots, gfn, approx);
1633 atomic_long_set(&slots->last_used_slot, (unsigned long)slot);
1641 * __gfn_to_memslot() and its descendants are here to allow arch code to inline
1642 * the lookups in hot paths. gfn_to_memslot() itself isn't here as an inline
1643 * because that would bloat other code too much.
1645 static inline struct kvm_memory_slot *
1646 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
1648 return ____gfn_to_memslot(slots, gfn, false);
1651 static inline unsigned long
1652 __gfn_to_hva_memslot(const struct kvm_memory_slot *slot, gfn_t gfn)
1655 * The index was checked originally in search_memslots. To avoid
1656 * that a malicious guest builds a Spectre gadget out of e.g. page
1657 * table walks, do not let the processor speculate loads outside
1658 * the guest's registered memslots.
1660 unsigned long offset = gfn - slot->base_gfn;
1661 offset = array_index_nospec(offset, slot->npages);
1662 return slot->userspace_addr + offset * PAGE_SIZE;
1665 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
1667 return gfn_to_memslot(kvm, gfn)->id;
1671 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
1673 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
1675 return slot->base_gfn + gfn_offset;
1678 static inline gpa_t gfn_to_gpa(gfn_t gfn)
1680 return (gpa_t)gfn << PAGE_SHIFT;
1683 static inline gfn_t gpa_to_gfn(gpa_t gpa)
1685 return (gfn_t)(gpa >> PAGE_SHIFT);
1688 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
1690 return (hpa_t)pfn << PAGE_SHIFT;
1693 static inline struct page *kvm_vcpu_gpa_to_page(struct kvm_vcpu *vcpu,
1696 return kvm_vcpu_gfn_to_page(vcpu, gpa_to_gfn(gpa));
1699 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
1701 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
1703 return kvm_is_error_hva(hva);
1706 enum kvm_stat_kind {
1711 struct kvm_stat_data {
1713 const struct _kvm_stats_desc *desc;
1714 enum kvm_stat_kind kind;
1717 struct _kvm_stats_desc {
1718 struct kvm_stats_desc desc;
1719 char name[KVM_STATS_NAME_SIZE];
1722 #define STATS_DESC_COMMON(type, unit, base, exp, sz, bsz) \
1723 .flags = type | unit | base | \
1724 BUILD_BUG_ON_ZERO(type & ~KVM_STATS_TYPE_MASK) | \
1725 BUILD_BUG_ON_ZERO(unit & ~KVM_STATS_UNIT_MASK) | \
1726 BUILD_BUG_ON_ZERO(base & ~KVM_STATS_BASE_MASK), \
1731 #define VM_GENERIC_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1734 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1735 .offset = offsetof(struct kvm_vm_stat, generic.stat) \
1739 #define VCPU_GENERIC_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1742 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1743 .offset = offsetof(struct kvm_vcpu_stat, generic.stat) \
1747 #define VM_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1750 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1751 .offset = offsetof(struct kvm_vm_stat, stat) \
1755 #define VCPU_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1758 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1759 .offset = offsetof(struct kvm_vcpu_stat, stat) \
1763 /* SCOPE: VM, VM_GENERIC, VCPU, VCPU_GENERIC */
1764 #define STATS_DESC(SCOPE, stat, type, unit, base, exp, sz, bsz) \
1765 SCOPE##_STATS_DESC(stat, type, unit, base, exp, sz, bsz)
1767 #define STATS_DESC_CUMULATIVE(SCOPE, name, unit, base, exponent) \
1768 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_CUMULATIVE, \
1769 unit, base, exponent, 1, 0)
1770 #define STATS_DESC_INSTANT(SCOPE, name, unit, base, exponent) \
1771 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_INSTANT, \
1772 unit, base, exponent, 1, 0)
1773 #define STATS_DESC_PEAK(SCOPE, name, unit, base, exponent) \
1774 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_PEAK, \
1775 unit, base, exponent, 1, 0)
1776 #define STATS_DESC_LINEAR_HIST(SCOPE, name, unit, base, exponent, sz, bsz) \
1777 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_LINEAR_HIST, \
1778 unit, base, exponent, sz, bsz)
1779 #define STATS_DESC_LOG_HIST(SCOPE, name, unit, base, exponent, sz) \
1780 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_LOG_HIST, \
1781 unit, base, exponent, sz, 0)
1783 /* Cumulative counter, read/write */
1784 #define STATS_DESC_COUNTER(SCOPE, name) \
1785 STATS_DESC_CUMULATIVE(SCOPE, name, KVM_STATS_UNIT_NONE, \
1786 KVM_STATS_BASE_POW10, 0)
1787 /* Instantaneous counter, read only */
1788 #define STATS_DESC_ICOUNTER(SCOPE, name) \
1789 STATS_DESC_INSTANT(SCOPE, name, KVM_STATS_UNIT_NONE, \
1790 KVM_STATS_BASE_POW10, 0)
1791 /* Peak counter, read/write */
1792 #define STATS_DESC_PCOUNTER(SCOPE, name) \
1793 STATS_DESC_PEAK(SCOPE, name, KVM_STATS_UNIT_NONE, \
1794 KVM_STATS_BASE_POW10, 0)
1796 /* Cumulative time in nanosecond */
1797 #define STATS_DESC_TIME_NSEC(SCOPE, name) \
1798 STATS_DESC_CUMULATIVE(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
1799 KVM_STATS_BASE_POW10, -9)
1800 /* Linear histogram for time in nanosecond */
1801 #define STATS_DESC_LINHIST_TIME_NSEC(SCOPE, name, sz, bsz) \
1802 STATS_DESC_LINEAR_HIST(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
1803 KVM_STATS_BASE_POW10, -9, sz, bsz)
1804 /* Logarithmic histogram for time in nanosecond */
1805 #define STATS_DESC_LOGHIST_TIME_NSEC(SCOPE, name, sz) \
1806 STATS_DESC_LOG_HIST(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
1807 KVM_STATS_BASE_POW10, -9, sz)
1809 #define KVM_GENERIC_VM_STATS() \
1810 STATS_DESC_COUNTER(VM_GENERIC, remote_tlb_flush), \
1811 STATS_DESC_COUNTER(VM_GENERIC, remote_tlb_flush_requests)
1813 #define KVM_GENERIC_VCPU_STATS() \
1814 STATS_DESC_COUNTER(VCPU_GENERIC, halt_successful_poll), \
1815 STATS_DESC_COUNTER(VCPU_GENERIC, halt_attempted_poll), \
1816 STATS_DESC_COUNTER(VCPU_GENERIC, halt_poll_invalid), \
1817 STATS_DESC_COUNTER(VCPU_GENERIC, halt_wakeup), \
1818 STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_poll_success_ns), \
1819 STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_poll_fail_ns), \
1820 STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_wait_ns), \
1821 STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_poll_success_hist, \
1822 HALT_POLL_HIST_COUNT), \
1823 STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_poll_fail_hist, \
1824 HALT_POLL_HIST_COUNT), \
1825 STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_wait_hist, \
1826 HALT_POLL_HIST_COUNT), \
1827 STATS_DESC_ICOUNTER(VCPU_GENERIC, blocking)
1829 extern struct dentry *kvm_debugfs_dir;
1831 ssize_t kvm_stats_read(char *id, const struct kvm_stats_header *header,
1832 const struct _kvm_stats_desc *desc,
1833 void *stats, size_t size_stats,
1834 char __user *user_buffer, size_t size, loff_t *offset);
1837 * kvm_stats_linear_hist_update() - Update bucket value for linear histogram
1840 * @data: start address of the stats data
1841 * @size: the number of bucket of the stats data
1842 * @value: the new value used to update the linear histogram's bucket
1843 * @bucket_size: the size (width) of a bucket
1845 static inline void kvm_stats_linear_hist_update(u64 *data, size_t size,
1846 u64 value, size_t bucket_size)
1848 size_t index = div64_u64(value, bucket_size);
1850 index = min(index, size - 1);
1855 * kvm_stats_log_hist_update() - Update bucket value for logarithmic histogram
1858 * @data: start address of the stats data
1859 * @size: the number of bucket of the stats data
1860 * @value: the new value used to update the logarithmic histogram's bucket
1862 static inline void kvm_stats_log_hist_update(u64 *data, size_t size, u64 value)
1864 size_t index = fls64(value);
1866 index = min(index, size - 1);
1870 #define KVM_STATS_LINEAR_HIST_UPDATE(array, value, bsize) \
1871 kvm_stats_linear_hist_update(array, ARRAY_SIZE(array), value, bsize)
1872 #define KVM_STATS_LOG_HIST_UPDATE(array, value) \
1873 kvm_stats_log_hist_update(array, ARRAY_SIZE(array), value)
1876 extern const struct kvm_stats_header kvm_vm_stats_header;
1877 extern const struct _kvm_stats_desc kvm_vm_stats_desc[];
1878 extern const struct kvm_stats_header kvm_vcpu_stats_header;
1879 extern const struct _kvm_stats_desc kvm_vcpu_stats_desc[];
1881 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1882 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1884 if (unlikely(kvm->mmu_notifier_count))
1887 * Ensure the read of mmu_notifier_count happens before the read
1888 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1889 * mmu_notifier_invalidate_range_end to make sure that the caller
1890 * either sees the old (non-zero) value of mmu_notifier_count or
1891 * the new (incremented) value of mmu_notifier_seq.
1892 * PowerPC Book3s HV KVM calls this under a per-page lock
1893 * rather than under kvm->mmu_lock, for scalability, so
1894 * can't rely on kvm->mmu_lock to keep things ordered.
1897 if (kvm->mmu_notifier_seq != mmu_seq)
1902 static inline int mmu_notifier_retry_hva(struct kvm *kvm,
1903 unsigned long mmu_seq,
1906 lockdep_assert_held(&kvm->mmu_lock);
1908 * If mmu_notifier_count is non-zero, then the range maintained by
1909 * kvm_mmu_notifier_invalidate_range_start contains all addresses that
1910 * might be being invalidated. Note that it may include some false
1911 * positives, due to shortcuts when handing concurrent invalidations.
1913 if (unlikely(kvm->mmu_notifier_count) &&
1914 hva >= kvm->mmu_notifier_range_start &&
1915 hva < kvm->mmu_notifier_range_end)
1917 if (kvm->mmu_notifier_seq != mmu_seq)
1923 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1925 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1927 bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
1928 int kvm_set_irq_routing(struct kvm *kvm,
1929 const struct kvm_irq_routing_entry *entries,
1932 int kvm_set_routing_entry(struct kvm *kvm,
1933 struct kvm_kernel_irq_routing_entry *e,
1934 const struct kvm_irq_routing_entry *ue);
1935 void kvm_free_irq_routing(struct kvm *kvm);
1939 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1943 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1945 #ifdef CONFIG_HAVE_KVM_EVENTFD
1947 void kvm_eventfd_init(struct kvm *kvm);
1948 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1950 #ifdef CONFIG_HAVE_KVM_IRQFD
1951 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1952 void kvm_irqfd_release(struct kvm *kvm);
1953 void kvm_irq_routing_update(struct kvm *);
1955 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1960 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1965 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1967 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1972 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1974 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1975 static inline void kvm_irq_routing_update(struct kvm *kvm)
1980 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1985 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1987 void kvm_arch_irq_routing_update(struct kvm *kvm);
1989 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1992 * Ensure the rest of the request is published to kvm_check_request's
1993 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1996 set_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1999 static inline bool kvm_request_pending(struct kvm_vcpu *vcpu)
2001 return READ_ONCE(vcpu->requests);
2004 static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)
2006 return test_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
2009 static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu)
2011 clear_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
2014 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
2016 if (kvm_test_request(req, vcpu)) {
2017 kvm_clear_request(req, vcpu);
2020 * Ensure the rest of the request is visible to kvm_check_request's
2021 * caller. Paired with the smp_wmb in kvm_make_request.
2023 smp_mb__after_atomic();
2030 extern bool kvm_rebooting;
2032 extern unsigned int halt_poll_ns;
2033 extern unsigned int halt_poll_ns_grow;
2034 extern unsigned int halt_poll_ns_grow_start;
2035 extern unsigned int halt_poll_ns_shrink;
2038 const struct kvm_device_ops *ops;
2041 struct list_head vm_node;
2044 /* create, destroy, and name are mandatory */
2045 struct kvm_device_ops {
2049 * create is called holding kvm->lock and any operations not suitable
2050 * to do while holding the lock should be deferred to init (see
2053 int (*create)(struct kvm_device *dev, u32 type);
2056 * init is called after create if create is successful and is called
2057 * outside of holding kvm->lock.
2059 void (*init)(struct kvm_device *dev);
2062 * Destroy is responsible for freeing dev.
2064 * Destroy may be called before or after destructors are called
2065 * on emulated I/O regions, depending on whether a reference is
2066 * held by a vcpu or other kvm component that gets destroyed
2067 * after the emulated I/O.
2069 void (*destroy)(struct kvm_device *dev);
2072 * Release is an alternative method to free the device. It is
2073 * called when the device file descriptor is closed. Once
2074 * release is called, the destroy method will not be called
2075 * anymore as the device is removed from the device list of
2076 * the VM. kvm->lock is held.
2078 void (*release)(struct kvm_device *dev);
2080 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
2081 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
2082 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
2083 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
2085 int (*mmap)(struct kvm_device *dev, struct vm_area_struct *vma);
2088 void kvm_device_get(struct kvm_device *dev);
2089 void kvm_device_put(struct kvm_device *dev);
2090 struct kvm_device *kvm_device_from_filp(struct file *filp);
2091 int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type);
2092 void kvm_unregister_device_ops(u32 type);
2094 extern struct kvm_device_ops kvm_mpic_ops;
2095 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
2096 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
2098 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
2100 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
2102 vcpu->spin_loop.in_spin_loop = val;
2104 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
2106 vcpu->spin_loop.dy_eligible = val;
2109 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
2111 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
2115 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
2118 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
2120 static inline bool kvm_is_visible_memslot(struct kvm_memory_slot *memslot)
2122 return (memslot && memslot->id < KVM_USER_MEM_SLOTS &&
2123 !(memslot->flags & KVM_MEMSLOT_INVALID));
2126 struct kvm_vcpu *kvm_get_running_vcpu(void);
2127 struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void);
2129 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
2130 bool kvm_arch_has_irq_bypass(void);
2131 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
2132 struct irq_bypass_producer *);
2133 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
2134 struct irq_bypass_producer *);
2135 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
2136 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
2137 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
2138 uint32_t guest_irq, bool set);
2139 bool kvm_arch_irqfd_route_changed(struct kvm_kernel_irq_routing_entry *,
2140 struct kvm_kernel_irq_routing_entry *);
2141 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
2143 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
2144 /* If we wakeup during the poll time, was it a sucessful poll? */
2145 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
2147 return vcpu->valid_wakeup;
2151 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
2155 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
2157 #ifdef CONFIG_HAVE_KVM_NO_POLL
2158 /* Callback that tells if we must not poll */
2159 bool kvm_arch_no_poll(struct kvm_vcpu *vcpu);
2161 static inline bool kvm_arch_no_poll(struct kvm_vcpu *vcpu)
2165 #endif /* CONFIG_HAVE_KVM_NO_POLL */
2167 #ifdef CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL
2168 long kvm_arch_vcpu_async_ioctl(struct file *filp,
2169 unsigned int ioctl, unsigned long arg);
2171 static inline long kvm_arch_vcpu_async_ioctl(struct file *filp,
2175 return -ENOIOCTLCMD;
2177 #endif /* CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL */
2179 void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
2180 unsigned long start, unsigned long end);
2182 #ifdef CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE
2183 int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu);
2185 static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
2189 #endif /* CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE */
2191 typedef int (*kvm_vm_thread_fn_t)(struct kvm *kvm, uintptr_t data);
2193 int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn,
2194 uintptr_t data, const char *name,
2195 struct task_struct **thread_ptr);
2197 #ifdef CONFIG_KVM_XFER_TO_GUEST_WORK
2198 static inline void kvm_handle_signal_exit(struct kvm_vcpu *vcpu)
2200 vcpu->run->exit_reason = KVM_EXIT_INTR;
2201 vcpu->stat.signal_exits++;
2203 #endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */
2206 * This defines how many reserved entries we want to keep before we
2207 * kick the vcpu to the userspace to avoid dirty ring full. This
2208 * value can be tuned to higher if e.g. PML is enabled on the host.
2210 #define KVM_DIRTY_RING_RSVD_ENTRIES 64
2212 /* Max number of entries allowed for each kvm dirty ring */
2213 #define KVM_DIRTY_RING_MAX_ENTRIES 65536