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/bug.h>
15 #include <linux/mmu_notifier.h>
16 #include <linux/preempt.h>
17 #include <linux/msi.h>
18 #include <linux/slab.h>
19 #include <linux/vmalloc.h>
20 #include <linux/rcupdate.h>
21 #include <linux/ratelimit.h>
22 #include <linux/err.h>
23 #include <linux/irqflags.h>
24 #include <linux/context_tracking.h>
25 #include <linux/irqbypass.h>
26 #include <linux/swait.h>
27 #include <linux/refcount.h>
28 #include <linux/nospec.h>
29 #include <asm/signal.h>
31 #include <linux/kvm.h>
32 #include <linux/kvm_para.h>
34 #include <linux/kvm_types.h>
36 #include <asm/kvm_host.h>
38 #ifndef KVM_MAX_VCPU_ID
39 #define KVM_MAX_VCPU_ID KVM_MAX_VCPUS
43 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
44 * in kvm, other bits are visible for userspace which are defined in
45 * include/linux/kvm_h.
47 #define KVM_MEMSLOT_INVALID (1UL << 16)
50 * Bit 63 of the memslot generation number is an "update in-progress flag",
51 * e.g. is temporarily set for the duration of install_new_memslots().
52 * This flag effectively creates a unique generation number that is used to
53 * mark cached memslot data, e.g. MMIO accesses, as potentially being stale,
54 * i.e. may (or may not) have come from the previous memslots generation.
56 * This is necessary because the actual memslots update is not atomic with
57 * respect to the generation number update. Updating the generation number
58 * first would allow a vCPU to cache a spte from the old memslots using the
59 * new generation number, and updating the generation number after switching
60 * to the new memslots would allow cache hits using the old generation number
61 * to reference the defunct memslots.
63 * This mechanism is used to prevent getting hits in KVM's caches while a
64 * memslot update is in-progress, and to prevent cache hits *after* updating
65 * the actual generation number against accesses that were inserted into the
66 * cache *before* the memslots were updated.
68 #define KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS BIT_ULL(63)
70 /* Two fragments for cross MMIO pages. */
71 #define KVM_MAX_MMIO_FRAGMENTS 2
73 #ifndef KVM_ADDRESS_SPACE_NUM
74 #define KVM_ADDRESS_SPACE_NUM 1
78 * For the normal pfn, the highest 12 bits should be zero,
79 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
80 * mask bit 63 to indicate the noslot pfn.
82 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
83 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
84 #define KVM_PFN_NOSLOT (0x1ULL << 63)
86 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
87 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
88 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
91 * error pfns indicate that the gfn is in slot but faild to
92 * translate it to pfn on host.
94 static inline bool is_error_pfn(kvm_pfn_t pfn
)
96 return !!(pfn
& KVM_PFN_ERR_MASK
);
100 * error_noslot pfns indicate that the gfn can not be
101 * translated to pfn - it is not in slot or failed to
102 * translate it to pfn.
104 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn
)
106 return !!(pfn
& KVM_PFN_ERR_NOSLOT_MASK
);
109 /* noslot pfn indicates that the gfn is not in slot. */
110 static inline bool is_noslot_pfn(kvm_pfn_t pfn
)
112 return pfn
== KVM_PFN_NOSLOT
;
116 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
117 * provide own defines and kvm_is_error_hva
119 #ifndef KVM_HVA_ERR_BAD
121 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
122 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
124 static inline bool kvm_is_error_hva(unsigned long addr
)
126 return addr
>= PAGE_OFFSET
;
131 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
133 static inline bool is_error_page(struct page
*page
)
138 #define KVM_REQUEST_MASK GENMASK(7,0)
139 #define KVM_REQUEST_NO_WAKEUP BIT(8)
140 #define KVM_REQUEST_WAIT BIT(9)
142 * Architecture-independent vcpu->requests bit members
143 * Bits 4-7 are reserved for more arch-independent bits.
145 #define KVM_REQ_TLB_FLUSH (0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
146 #define KVM_REQ_MMU_RELOAD (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
147 #define KVM_REQ_PENDING_TIMER 2
148 #define KVM_REQ_UNHALT 3
149 #define KVM_REQUEST_ARCH_BASE 8
151 #define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
152 BUILD_BUG_ON((unsigned)(nr) >= (sizeof_field(struct kvm_vcpu, requests) * 8) - KVM_REQUEST_ARCH_BASE); \
153 (unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) | (flags)); \
155 #define KVM_ARCH_REQ(nr) KVM_ARCH_REQ_FLAGS(nr, 0)
157 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
158 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
160 extern struct mutex kvm_lock
;
161 extern struct list_head vm_list
;
163 struct kvm_io_range
{
166 struct kvm_io_device
*dev
;
169 #define NR_IOBUS_DEVS 1000
174 struct kvm_io_range range
[];
180 KVM_VIRTIO_CCW_NOTIFY_BUS
,
185 int kvm_io_bus_write(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
186 int len
, const void *val
);
187 int kvm_io_bus_write_cookie(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
,
188 gpa_t addr
, int len
, const void *val
, long cookie
);
189 int kvm_io_bus_read(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
191 int kvm_io_bus_register_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
,
192 int len
, struct kvm_io_device
*dev
);
193 void kvm_io_bus_unregister_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
194 struct kvm_io_device
*dev
);
195 struct kvm_io_device
*kvm_io_bus_get_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
198 #ifdef CONFIG_KVM_ASYNC_PF
199 struct kvm_async_pf
{
200 struct work_struct work
;
201 struct list_head link
;
202 struct list_head queue
;
203 struct kvm_vcpu
*vcpu
;
204 struct mm_struct
*mm
;
207 struct kvm_arch_async_pf arch
;
211 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu
*vcpu
);
212 void kvm_check_async_pf_completion(struct kvm_vcpu
*vcpu
);
213 int kvm_setup_async_pf(struct kvm_vcpu
*vcpu
, gpa_t cr2_or_gpa
,
214 unsigned long hva
, struct kvm_arch_async_pf
*arch
);
215 int kvm_async_pf_wakeup_all(struct kvm_vcpu
*vcpu
);
222 READING_SHADOW_PAGE_TABLES
,
225 #define KVM_UNMAPPED_PAGE ((void *) 0x500 + POISON_POINTER_DELTA)
227 struct kvm_host_map
{
229 * Only valid if the 'pfn' is managed by the host kernel (i.e. There is
230 * a 'struct page' for it. When using mem= kernel parameter some memory
231 * can be used as guest memory but they are not managed by host
233 * If 'pfn' is not managed by the host kernel, this field is
234 * initialized to KVM_UNMAPPED_PAGE.
243 * Used to check if the mapping is valid or not. Never use 'kvm_host_map'
244 * directly to check for that.
246 static inline bool kvm_vcpu_mapped(struct kvm_host_map
*map
)
252 * Sometimes a large or cross-page mmio needs to be broken up into separate
253 * exits for userspace servicing.
255 struct kvm_mmio_fragment
{
263 #ifdef CONFIG_PREEMPT_NOTIFIERS
264 struct preempt_notifier preempt_notifier
;
267 int vcpu_id
; /* id given by userspace at creation */
268 int vcpu_idx
; /* index in kvm->vcpus array */
272 unsigned long guest_debug
;
275 struct list_head blocked_vcpu_list
;
280 struct swait_queue_head wq
;
281 struct pid __rcu
*pid
;
284 struct kvm_vcpu_stat stat
;
285 unsigned int halt_poll_ns
;
288 #ifdef CONFIG_HAS_IOMEM
290 int mmio_read_completed
;
292 int mmio_cur_fragment
;
293 int mmio_nr_fragments
;
294 struct kvm_mmio_fragment mmio_fragments
[KVM_MAX_MMIO_FRAGMENTS
];
297 #ifdef CONFIG_KVM_ASYNC_PF
300 struct list_head queue
;
301 struct list_head done
;
306 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
308 * Cpu relax intercept or pause loop exit optimization
309 * in_spin_loop: set when a vcpu does a pause loop exit
310 * or cpu relax intercepted.
311 * dy_eligible: indicates whether vcpu is eligible for directed yield.
320 struct kvm_vcpu_arch arch
;
321 struct dentry
*debugfs_dentry
;
324 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu
*vcpu
)
327 * The memory barrier ensures a previous write to vcpu->requests cannot
328 * be reordered with the read of vcpu->mode. It pairs with the general
329 * memory barrier following the write of vcpu->mode in VCPU RUN.
331 smp_mb__before_atomic();
332 return cmpxchg(&vcpu
->mode
, IN_GUEST_MODE
, EXITING_GUEST_MODE
);
336 * Some of the bitops functions do not support too long bitmaps.
337 * This number must be determined not to exceed such limits.
339 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
341 struct kvm_memory_slot
{
343 unsigned long npages
;
344 unsigned long *dirty_bitmap
;
345 struct kvm_arch_memory_slot arch
;
346 unsigned long userspace_addr
;
351 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot
*memslot
)
353 return ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
356 static inline unsigned long *kvm_second_dirty_bitmap(struct kvm_memory_slot
*memslot
)
358 unsigned long len
= kvm_dirty_bitmap_bytes(memslot
);
360 return memslot
->dirty_bitmap
+ len
/ sizeof(*memslot
->dirty_bitmap
);
363 #ifndef KVM_DIRTY_LOG_MANUAL_CAPS
364 #define KVM_DIRTY_LOG_MANUAL_CAPS KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE
367 struct kvm_s390_adapter_int
{
380 struct kvm_kernel_irq_routing_entry
{
383 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
384 struct kvm
*kvm
, int irq_source_id
, int level
,
398 struct kvm_s390_adapter_int adapter
;
399 struct kvm_hv_sint hv_sint
;
401 struct hlist_node link
;
404 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
405 struct kvm_irq_routing_table
{
406 int chip
[KVM_NR_IRQCHIPS
][KVM_IRQCHIP_NUM_PINS
];
409 * Array indexed by gsi. Each entry contains list of irq chips
410 * the gsi is connected to.
412 struct hlist_head map
[0];
416 #ifndef KVM_PRIVATE_MEM_SLOTS
417 #define KVM_PRIVATE_MEM_SLOTS 0
420 #ifndef KVM_MEM_SLOTS_NUM
421 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
424 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
425 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu
*vcpu
)
433 * memslots are not sorted by id anymore, please use id_to_memslot()
434 * to get the memslot by its id.
436 struct kvm_memslots
{
438 /* The mapping table from slot id to the index in memslots[]. */
439 short id_to_index
[KVM_MEM_SLOTS_NUM
];
442 struct kvm_memory_slot memslots
[];
447 struct mutex slots_lock
;
448 struct mm_struct
*mm
; /* userspace tied to this vm */
449 struct kvm_memslots __rcu
*memslots
[KVM_ADDRESS_SPACE_NUM
];
450 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
453 * created_vcpus is protected by kvm->lock, and is incremented
454 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
455 * incremented after storing the kvm_vcpu pointer in vcpus,
456 * and is accessed atomically.
458 atomic_t online_vcpus
;
460 int last_boosted_vcpu
;
461 struct list_head vm_list
;
463 struct kvm_io_bus __rcu
*buses
[KVM_NR_BUSES
];
464 #ifdef CONFIG_HAVE_KVM_EVENTFD
467 struct list_head items
;
468 struct list_head resampler_list
;
469 struct mutex resampler_lock
;
471 struct list_head ioeventfds
;
473 struct kvm_vm_stat stat
;
474 struct kvm_arch arch
;
475 refcount_t users_count
;
476 #ifdef CONFIG_KVM_MMIO
477 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
478 spinlock_t ring_lock
;
479 struct list_head coalesced_zones
;
482 struct mutex irq_lock
;
483 #ifdef CONFIG_HAVE_KVM_IRQCHIP
485 * Update side is protected by irq_lock.
487 struct kvm_irq_routing_table __rcu
*irq_routing
;
489 #ifdef CONFIG_HAVE_KVM_IRQFD
490 struct hlist_head irq_ack_notifier_list
;
493 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
494 struct mmu_notifier mmu_notifier
;
495 unsigned long mmu_notifier_seq
;
496 long mmu_notifier_count
;
499 struct list_head devices
;
500 u64 manual_dirty_log_protect
;
501 struct dentry
*debugfs_dentry
;
502 struct kvm_stat_data
**debugfs_stat_data
;
503 struct srcu_struct srcu
;
504 struct srcu_struct irq_srcu
;
508 #define kvm_err(fmt, ...) \
509 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
510 #define kvm_info(fmt, ...) \
511 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
512 #define kvm_debug(fmt, ...) \
513 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
514 #define kvm_debug_ratelimited(fmt, ...) \
515 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
517 #define kvm_pr_unimpl(fmt, ...) \
518 pr_err_ratelimited("kvm [%i]: " fmt, \
519 task_tgid_nr(current), ## __VA_ARGS__)
521 /* The guest did something we don't support. */
522 #define vcpu_unimpl(vcpu, fmt, ...) \
523 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
524 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
526 #define vcpu_debug(vcpu, fmt, ...) \
527 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
528 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
529 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
531 #define vcpu_err(vcpu, fmt, ...) \
532 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
534 static inline bool kvm_dirty_log_manual_protect_and_init_set(struct kvm
*kvm
)
536 return !!(kvm
->manual_dirty_log_protect
& KVM_DIRTY_LOG_INITIALLY_SET
);
539 static inline struct kvm_io_bus
*kvm_get_bus(struct kvm
*kvm
, enum kvm_bus idx
)
541 return srcu_dereference_check(kvm
->buses
[idx
], &kvm
->srcu
,
542 lockdep_is_held(&kvm
->slots_lock
) ||
543 !refcount_read(&kvm
->users_count
));
546 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
548 int num_vcpus
= atomic_read(&kvm
->online_vcpus
);
549 i
= array_index_nospec(i
, num_vcpus
);
551 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. */
553 return kvm
->vcpus
[i
];
556 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
558 idx < atomic_read(&kvm->online_vcpus) && \
559 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
562 static inline struct kvm_vcpu
*kvm_get_vcpu_by_id(struct kvm
*kvm
, int id
)
564 struct kvm_vcpu
*vcpu
= NULL
;
569 if (id
< KVM_MAX_VCPUS
)
570 vcpu
= kvm_get_vcpu(kvm
, id
);
571 if (vcpu
&& vcpu
->vcpu_id
== id
)
573 kvm_for_each_vcpu(i
, vcpu
, kvm
)
574 if (vcpu
->vcpu_id
== id
)
579 static inline int kvm_vcpu_get_idx(struct kvm_vcpu
*vcpu
)
581 return vcpu
->vcpu_idx
;
584 #define kvm_for_each_memslot(memslot, slots) \
585 for (memslot = &slots->memslots[0]; \
586 memslot < slots->memslots + slots->used_slots; memslot++) \
587 if (WARN_ON_ONCE(!memslot->npages)) { \
590 void kvm_vcpu_destroy(struct kvm_vcpu
*vcpu
);
592 void vcpu_load(struct kvm_vcpu
*vcpu
);
593 void vcpu_put(struct kvm_vcpu
*vcpu
);
595 #ifdef __KVM_HAVE_IOAPIC
596 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm
*kvm
);
597 void kvm_arch_post_irq_routing_update(struct kvm
*kvm
);
599 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm
*kvm
)
602 static inline void kvm_arch_post_irq_routing_update(struct kvm
*kvm
)
607 #ifdef CONFIG_HAVE_KVM_IRQFD
608 int kvm_irqfd_init(void);
609 void kvm_irqfd_exit(void);
611 static inline int kvm_irqfd_init(void)
616 static inline void kvm_irqfd_exit(void)
620 int kvm_init(void *opaque
, unsigned vcpu_size
, unsigned vcpu_align
,
621 struct module
*module
);
624 void kvm_get_kvm(struct kvm
*kvm
);
625 void kvm_put_kvm(struct kvm
*kvm
);
626 void kvm_put_kvm_no_destroy(struct kvm
*kvm
);
628 static inline struct kvm_memslots
*__kvm_memslots(struct kvm
*kvm
, int as_id
)
630 as_id
= array_index_nospec(as_id
, KVM_ADDRESS_SPACE_NUM
);
631 return srcu_dereference_check(kvm
->memslots
[as_id
], &kvm
->srcu
,
632 lockdep_is_held(&kvm
->slots_lock
) ||
633 !refcount_read(&kvm
->users_count
));
636 static inline struct kvm_memslots
*kvm_memslots(struct kvm
*kvm
)
638 return __kvm_memslots(kvm
, 0);
641 static inline struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
)
643 int as_id
= kvm_arch_vcpu_memslots_id(vcpu
);
645 return __kvm_memslots(vcpu
->kvm
, as_id
);
649 struct kvm_memory_slot
*id_to_memslot(struct kvm_memslots
*slots
, int id
)
651 int index
= slots
->id_to_index
[id
];
652 struct kvm_memory_slot
*slot
;
657 slot
= &slots
->memslots
[index
];
659 WARN_ON(slot
->id
!= id
);
664 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
665 * - create a new memory slot
666 * - delete an existing memory slot
667 * - modify an existing memory slot
668 * -- move it in the guest physical memory space
669 * -- just change its flags
671 * Since flags can be changed by some of these operations, the following
672 * differentiation is the best we can do for __kvm_set_memory_region():
681 int kvm_set_memory_region(struct kvm
*kvm
,
682 const struct kvm_userspace_memory_region
*mem
);
683 int __kvm_set_memory_region(struct kvm
*kvm
,
684 const struct kvm_userspace_memory_region
*mem
);
685 void kvm_arch_free_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
686 void kvm_arch_memslots_updated(struct kvm
*kvm
, u64 gen
);
687 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
688 struct kvm_memory_slot
*memslot
,
689 const struct kvm_userspace_memory_region
*mem
,
690 enum kvm_mr_change change
);
691 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
692 const struct kvm_userspace_memory_region
*mem
,
693 struct kvm_memory_slot
*old
,
694 const struct kvm_memory_slot
*new,
695 enum kvm_mr_change change
);
696 /* flush all memory translations */
697 void kvm_arch_flush_shadow_all(struct kvm
*kvm
);
698 /* flush memory translations pointing to 'slot' */
699 void kvm_arch_flush_shadow_memslot(struct kvm
*kvm
,
700 struct kvm_memory_slot
*slot
);
702 int gfn_to_page_many_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
,
703 struct page
**pages
, int nr_pages
);
705 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
706 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
707 unsigned long gfn_to_hva_prot(struct kvm
*kvm
, gfn_t gfn
, bool *writable
);
708 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
709 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
711 void kvm_release_page_clean(struct page
*page
);
712 void kvm_release_page_dirty(struct page
*page
);
713 void kvm_set_page_accessed(struct page
*page
);
715 kvm_pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
716 kvm_pfn_t
gfn_to_pfn_prot(struct kvm
*kvm
, gfn_t gfn
, bool write_fault
,
718 kvm_pfn_t
gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
719 kvm_pfn_t
gfn_to_pfn_memslot_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
);
720 kvm_pfn_t
__gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
721 bool atomic
, bool *async
, bool write_fault
,
724 void kvm_release_pfn_clean(kvm_pfn_t pfn
);
725 void kvm_release_pfn_dirty(kvm_pfn_t pfn
);
726 void kvm_set_pfn_dirty(kvm_pfn_t pfn
);
727 void kvm_set_pfn_accessed(kvm_pfn_t pfn
);
728 void kvm_get_pfn(kvm_pfn_t pfn
);
730 void kvm_release_pfn(kvm_pfn_t pfn
, bool dirty
, struct gfn_to_pfn_cache
*cache
);
731 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
733 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
734 int kvm_read_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
735 void *data
, unsigned long len
);
736 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
737 int offset
, int len
);
738 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
740 int kvm_write_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
741 void *data
, unsigned long len
);
742 int kvm_write_guest_offset_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
743 void *data
, unsigned int offset
,
745 int kvm_gfn_to_hva_cache_init(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
746 gpa_t gpa
, unsigned long len
);
748 #define __kvm_put_guest(kvm, gfn, offset, value, type) \
750 unsigned long __addr = gfn_to_hva(kvm, gfn); \
751 type __user *__uaddr = (type __user *)(__addr + offset); \
752 int __ret = -EFAULT; \
754 if (!kvm_is_error_hva(__addr)) \
755 __ret = put_user(value, __uaddr); \
757 mark_page_dirty(kvm, gfn); \
761 #define kvm_put_guest(kvm, gpa, value, type) \
764 struct kvm *__kvm = kvm; \
765 __kvm_put_guest(__kvm, __gpa >> PAGE_SHIFT, \
766 offset_in_page(__gpa), (value), type); \
769 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
770 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
771 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
772 bool kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
773 unsigned long kvm_host_page_size(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
774 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
776 struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
);
777 struct kvm_memory_slot
*kvm_vcpu_gfn_to_memslot(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
778 kvm_pfn_t
kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
779 kvm_pfn_t
kvm_vcpu_gfn_to_pfn(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
780 int kvm_vcpu_map(struct kvm_vcpu
*vcpu
, gpa_t gpa
, struct kvm_host_map
*map
);
781 int kvm_map_gfn(struct kvm_vcpu
*vcpu
, gfn_t gfn
, struct kvm_host_map
*map
,
782 struct gfn_to_pfn_cache
*cache
, bool atomic
);
783 struct page
*kvm_vcpu_gfn_to_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
784 void kvm_vcpu_unmap(struct kvm_vcpu
*vcpu
, struct kvm_host_map
*map
, bool dirty
);
785 int kvm_unmap_gfn(struct kvm_vcpu
*vcpu
, struct kvm_host_map
*map
,
786 struct gfn_to_pfn_cache
*cache
, bool dirty
, bool atomic
);
787 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
788 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu
*vcpu
, gfn_t gfn
, bool *writable
);
789 int kvm_vcpu_read_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, void *data
, int offset
,
791 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
793 int kvm_vcpu_read_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
795 int kvm_vcpu_write_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, const void *data
,
796 int offset
, int len
);
797 int kvm_vcpu_write_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, const void *data
,
799 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
801 void kvm_sigset_activate(struct kvm_vcpu
*vcpu
);
802 void kvm_sigset_deactivate(struct kvm_vcpu
*vcpu
);
804 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
805 void kvm_arch_vcpu_blocking(struct kvm_vcpu
*vcpu
);
806 void kvm_arch_vcpu_unblocking(struct kvm_vcpu
*vcpu
);
807 bool kvm_vcpu_wake_up(struct kvm_vcpu
*vcpu
);
808 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
809 int kvm_vcpu_yield_to(struct kvm_vcpu
*target
);
810 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
, bool usermode_vcpu_not_eligible
);
812 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
813 void kvm_reload_remote_mmus(struct kvm
*kvm
);
815 bool kvm_make_vcpus_request_mask(struct kvm
*kvm
, unsigned int req
,
816 struct kvm_vcpu
*except
,
817 unsigned long *vcpu_bitmap
, cpumask_var_t tmp
);
818 bool kvm_make_all_cpus_request(struct kvm
*kvm
, unsigned int req
);
819 bool kvm_make_all_cpus_request_except(struct kvm
*kvm
, unsigned int req
,
820 struct kvm_vcpu
*except
);
821 bool kvm_make_cpus_request_mask(struct kvm
*kvm
, unsigned int req
,
822 unsigned long *vcpu_bitmap
);
824 long kvm_arch_dev_ioctl(struct file
*filp
,
825 unsigned int ioctl
, unsigned long arg
);
826 long kvm_arch_vcpu_ioctl(struct file
*filp
,
827 unsigned int ioctl
, unsigned long arg
);
828 vm_fault_t
kvm_arch_vcpu_fault(struct kvm_vcpu
*vcpu
, struct vm_fault
*vmf
);
830 int kvm_vm_ioctl_check_extension(struct kvm
*kvm
, long ext
);
832 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm
*kvm
,
833 struct kvm_memory_slot
*slot
,
836 void kvm_arch_sync_dirty_log(struct kvm
*kvm
, struct kvm_memory_slot
*memslot
);
838 #ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
839 void kvm_arch_flush_remote_tlbs_memslot(struct kvm
*kvm
,
840 struct kvm_memory_slot
*memslot
);
841 #else /* !CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
842 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
, struct kvm_dirty_log
*log
);
843 int kvm_get_dirty_log(struct kvm
*kvm
, struct kvm_dirty_log
*log
,
844 int *is_dirty
, struct kvm_memory_slot
**memslot
);
847 int kvm_vm_ioctl_irq_line(struct kvm
*kvm
, struct kvm_irq_level
*irq_level
,
849 int kvm_vm_ioctl_enable_cap(struct kvm
*kvm
,
850 struct kvm_enable_cap
*cap
);
851 long kvm_arch_vm_ioctl(struct file
*filp
,
852 unsigned int ioctl
, unsigned long arg
);
854 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
855 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
857 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
858 struct kvm_translation
*tr
);
860 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
861 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
862 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
863 struct kvm_sregs
*sregs
);
864 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
865 struct kvm_sregs
*sregs
);
866 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
867 struct kvm_mp_state
*mp_state
);
868 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
869 struct kvm_mp_state
*mp_state
);
870 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
871 struct kvm_guest_debug
*dbg
);
872 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
874 int kvm_arch_init(void *opaque
);
875 void kvm_arch_exit(void);
877 void kvm_arch_sched_in(struct kvm_vcpu
*vcpu
, int cpu
);
879 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
880 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
881 int kvm_arch_vcpu_precreate(struct kvm
*kvm
, unsigned int id
);
882 int kvm_arch_vcpu_create(struct kvm_vcpu
*vcpu
);
883 void kvm_arch_vcpu_postcreate(struct kvm_vcpu
*vcpu
);
884 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
886 #ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
887 void kvm_arch_create_vcpu_debugfs(struct kvm_vcpu
*vcpu
);
890 int kvm_arch_hardware_enable(void);
891 void kvm_arch_hardware_disable(void);
892 int kvm_arch_hardware_setup(void *opaque
);
893 void kvm_arch_hardware_unsetup(void);
894 int kvm_arch_check_processor_compat(void *opaque
);
895 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
896 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu
*vcpu
);
897 int kvm_arch_vcpu_should_kick(struct kvm_vcpu
*vcpu
);
898 bool kvm_arch_dy_runnable(struct kvm_vcpu
*vcpu
);
899 int kvm_arch_post_init_vm(struct kvm
*kvm
);
900 void kvm_arch_pre_destroy_vm(struct kvm
*kvm
);
902 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
904 * All architectures that want to use vzalloc currently also
905 * need their own kvm_arch_alloc_vm implementation.
907 static inline struct kvm
*kvm_arch_alloc_vm(void)
909 return kzalloc(sizeof(struct kvm
), GFP_KERNEL
);
912 static inline void kvm_arch_free_vm(struct kvm
*kvm
)
918 #ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
919 static inline int kvm_arch_flush_remote_tlb(struct kvm
*kvm
)
925 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
926 void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
);
927 void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
);
928 bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
);
930 static inline void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
)
934 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
)
938 static inline bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
)
943 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
944 void kvm_arch_start_assignment(struct kvm
*kvm
);
945 void kvm_arch_end_assignment(struct kvm
*kvm
);
946 bool kvm_arch_has_assigned_device(struct kvm
*kvm
);
948 static inline void kvm_arch_start_assignment(struct kvm
*kvm
)
952 static inline void kvm_arch_end_assignment(struct kvm
*kvm
)
956 static inline bool kvm_arch_has_assigned_device(struct kvm
*kvm
)
962 static inline struct swait_queue_head
*kvm_arch_vcpu_wq(struct kvm_vcpu
*vcpu
)
964 #ifdef __KVM_HAVE_ARCH_WQP
965 return vcpu
->arch
.wqp
;
971 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
973 * returns true if the virtual interrupt controller is initialized and
974 * ready to accept virtual IRQ. On some architectures the virtual interrupt
975 * controller is dynamically instantiated and this is not always true.
977 bool kvm_arch_intc_initialized(struct kvm
*kvm
);
979 static inline bool kvm_arch_intc_initialized(struct kvm
*kvm
)
985 int kvm_arch_init_vm(struct kvm
*kvm
, unsigned long type
);
986 void kvm_arch_destroy_vm(struct kvm
*kvm
);
987 void kvm_arch_sync_events(struct kvm
*kvm
);
989 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
991 bool kvm_is_reserved_pfn(kvm_pfn_t pfn
);
992 bool kvm_is_zone_device_pfn(kvm_pfn_t pfn
);
993 bool kvm_is_transparent_hugepage(kvm_pfn_t pfn
);
995 struct kvm_irq_ack_notifier
{
996 struct hlist_node link
;
998 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
1001 int kvm_irq_map_gsi(struct kvm
*kvm
,
1002 struct kvm_kernel_irq_routing_entry
*entries
, int gsi
);
1003 int kvm_irq_map_chip_pin(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
1005 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
,
1007 int kvm_set_msi(struct kvm_kernel_irq_routing_entry
*irq_entry
, struct kvm
*kvm
,
1008 int irq_source_id
, int level
, bool line_status
);
1009 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry
*e
,
1010 struct kvm
*kvm
, int irq_source_id
,
1011 int level
, bool line_status
);
1012 bool kvm_irq_has_notifier(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
1013 void kvm_notify_acked_gsi(struct kvm
*kvm
, int gsi
);
1014 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
1015 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
1016 struct kvm_irq_ack_notifier
*kian
);
1017 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
1018 struct kvm_irq_ack_notifier
*kian
);
1019 int kvm_request_irq_source_id(struct kvm
*kvm
);
1020 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
1021 bool kvm_arch_irqfd_allowed(struct kvm
*kvm
, struct kvm_irqfd
*args
);
1024 * search_memslots() and __gfn_to_memslot() are here because they are
1025 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
1026 * gfn_to_memslot() itself isn't here as an inline because that would
1027 * bloat other code too much.
1029 * IMPORTANT: Slots are sorted from highest GFN to lowest GFN!
1031 static inline struct kvm_memory_slot
*
1032 search_memslots(struct kvm_memslots
*slots
, gfn_t gfn
)
1034 int start
= 0, end
= slots
->used_slots
;
1035 int slot
= atomic_read(&slots
->lru_slot
);
1036 struct kvm_memory_slot
*memslots
= slots
->memslots
;
1038 if (unlikely(!slots
->used_slots
))
1041 if (gfn
>= memslots
[slot
].base_gfn
&&
1042 gfn
< memslots
[slot
].base_gfn
+ memslots
[slot
].npages
)
1043 return &memslots
[slot
];
1045 while (start
< end
) {
1046 slot
= start
+ (end
- start
) / 2;
1048 if (gfn
>= memslots
[slot
].base_gfn
)
1054 if (start
< slots
->used_slots
&& gfn
>= memslots
[start
].base_gfn
&&
1055 gfn
< memslots
[start
].base_gfn
+ memslots
[start
].npages
) {
1056 atomic_set(&slots
->lru_slot
, start
);
1057 return &memslots
[start
];
1063 static inline struct kvm_memory_slot
*
1064 __gfn_to_memslot(struct kvm_memslots
*slots
, gfn_t gfn
)
1066 return search_memslots(slots
, gfn
);
1069 static inline unsigned long
1070 __gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
)
1072 return slot
->userspace_addr
+ (gfn
- slot
->base_gfn
) * PAGE_SIZE
;
1075 static inline int memslot_id(struct kvm
*kvm
, gfn_t gfn
)
1077 return gfn_to_memslot(kvm
, gfn
)->id
;
1081 hva_to_gfn_memslot(unsigned long hva
, struct kvm_memory_slot
*slot
)
1083 gfn_t gfn_offset
= (hva
- slot
->userspace_addr
) >> PAGE_SHIFT
;
1085 return slot
->base_gfn
+ gfn_offset
;
1088 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
1090 return (gpa_t
)gfn
<< PAGE_SHIFT
;
1093 static inline gfn_t
gpa_to_gfn(gpa_t gpa
)
1095 return (gfn_t
)(gpa
>> PAGE_SHIFT
);
1098 static inline hpa_t
pfn_to_hpa(kvm_pfn_t pfn
)
1100 return (hpa_t
)pfn
<< PAGE_SHIFT
;
1103 static inline struct page
*kvm_vcpu_gpa_to_page(struct kvm_vcpu
*vcpu
,
1106 return kvm_vcpu_gfn_to_page(vcpu
, gpa_to_gfn(gpa
));
1109 static inline bool kvm_is_error_gpa(struct kvm
*kvm
, gpa_t gpa
)
1111 unsigned long hva
= gfn_to_hva(kvm
, gpa_to_gfn(gpa
));
1113 return kvm_is_error_hva(hva
);
1116 enum kvm_stat_kind
{
1121 struct kvm_stat_data
{
1123 struct kvm_stats_debugfs_item
*dbgfs_item
;
1126 struct kvm_stats_debugfs_item
{
1129 enum kvm_stat_kind kind
;
1133 #define KVM_DBGFS_GET_MODE(dbgfs_item) \
1134 ((dbgfs_item)->mode ? (dbgfs_item)->mode : 0644)
1136 extern struct kvm_stats_debugfs_item debugfs_entries
[];
1137 extern struct dentry
*kvm_debugfs_dir
;
1139 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1140 static inline int mmu_notifier_retry(struct kvm
*kvm
, unsigned long mmu_seq
)
1142 if (unlikely(kvm
->mmu_notifier_count
))
1145 * Ensure the read of mmu_notifier_count happens before the read
1146 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1147 * mmu_notifier_invalidate_range_end to make sure that the caller
1148 * either sees the old (non-zero) value of mmu_notifier_count or
1149 * the new (incremented) value of mmu_notifier_seq.
1150 * PowerPC Book3s HV KVM calls this under a per-page lock
1151 * rather than under kvm->mmu_lock, for scalability, so
1152 * can't rely on kvm->mmu_lock to keep things ordered.
1155 if (kvm
->mmu_notifier_seq
!= mmu_seq
)
1161 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1163 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1165 bool kvm_arch_can_set_irq_routing(struct kvm
*kvm
);
1166 int kvm_set_irq_routing(struct kvm
*kvm
,
1167 const struct kvm_irq_routing_entry
*entries
,
1170 int kvm_set_routing_entry(struct kvm
*kvm
,
1171 struct kvm_kernel_irq_routing_entry
*e
,
1172 const struct kvm_irq_routing_entry
*ue
);
1173 void kvm_free_irq_routing(struct kvm
*kvm
);
1177 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
1181 int kvm_send_userspace_msi(struct kvm
*kvm
, struct kvm_msi
*msi
);
1183 #ifdef CONFIG_HAVE_KVM_EVENTFD
1185 void kvm_eventfd_init(struct kvm
*kvm
);
1186 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
1188 #ifdef CONFIG_HAVE_KVM_IRQFD
1189 int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
);
1190 void kvm_irqfd_release(struct kvm
*kvm
);
1191 void kvm_irq_routing_update(struct kvm
*);
1193 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1198 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1203 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
1205 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1210 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1212 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1213 static inline void kvm_irq_routing_update(struct kvm
*kvm
)
1218 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
1223 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1225 void kvm_arch_irq_routing_update(struct kvm
*kvm
);
1227 static inline void kvm_make_request(int req
, struct kvm_vcpu
*vcpu
)
1230 * Ensure the rest of the request is published to kvm_check_request's
1231 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1234 set_bit(req
& KVM_REQUEST_MASK
, (void *)&vcpu
->requests
);
1237 static inline bool kvm_request_pending(struct kvm_vcpu
*vcpu
)
1239 return READ_ONCE(vcpu
->requests
);
1242 static inline bool kvm_test_request(int req
, struct kvm_vcpu
*vcpu
)
1244 return test_bit(req
& KVM_REQUEST_MASK
, (void *)&vcpu
->requests
);
1247 static inline void kvm_clear_request(int req
, struct kvm_vcpu
*vcpu
)
1249 clear_bit(req
& KVM_REQUEST_MASK
, (void *)&vcpu
->requests
);
1252 static inline bool kvm_check_request(int req
, struct kvm_vcpu
*vcpu
)
1254 if (kvm_test_request(req
, vcpu
)) {
1255 kvm_clear_request(req
, vcpu
);
1258 * Ensure the rest of the request is visible to kvm_check_request's
1259 * caller. Paired with the smp_wmb in kvm_make_request.
1261 smp_mb__after_atomic();
1268 extern bool kvm_rebooting
;
1270 extern unsigned int halt_poll_ns
;
1271 extern unsigned int halt_poll_ns_grow
;
1272 extern unsigned int halt_poll_ns_grow_start
;
1273 extern unsigned int halt_poll_ns_shrink
;
1276 const struct kvm_device_ops
*ops
;
1279 struct list_head vm_node
;
1282 /* create, destroy, and name are mandatory */
1283 struct kvm_device_ops
{
1287 * create is called holding kvm->lock and any operations not suitable
1288 * to do while holding the lock should be deferred to init (see
1291 int (*create
)(struct kvm_device
*dev
, u32 type
);
1294 * init is called after create if create is successful and is called
1295 * outside of holding kvm->lock.
1297 void (*init
)(struct kvm_device
*dev
);
1300 * Destroy is responsible for freeing dev.
1302 * Destroy may be called before or after destructors are called
1303 * on emulated I/O regions, depending on whether a reference is
1304 * held by a vcpu or other kvm component that gets destroyed
1305 * after the emulated I/O.
1307 void (*destroy
)(struct kvm_device
*dev
);
1310 * Release is an alternative method to free the device. It is
1311 * called when the device file descriptor is closed. Once
1312 * release is called, the destroy method will not be called
1313 * anymore as the device is removed from the device list of
1314 * the VM. kvm->lock is held.
1316 void (*release
)(struct kvm_device
*dev
);
1318 int (*set_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1319 int (*get_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1320 int (*has_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1321 long (*ioctl
)(struct kvm_device
*dev
, unsigned int ioctl
,
1323 int (*mmap
)(struct kvm_device
*dev
, struct vm_area_struct
*vma
);
1326 void kvm_device_get(struct kvm_device
*dev
);
1327 void kvm_device_put(struct kvm_device
*dev
);
1328 struct kvm_device
*kvm_device_from_filp(struct file
*filp
);
1329 int kvm_register_device_ops(const struct kvm_device_ops
*ops
, u32 type
);
1330 void kvm_unregister_device_ops(u32 type
);
1332 extern struct kvm_device_ops kvm_mpic_ops
;
1333 extern struct kvm_device_ops kvm_arm_vgic_v2_ops
;
1334 extern struct kvm_device_ops kvm_arm_vgic_v3_ops
;
1336 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1338 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1340 vcpu
->spin_loop
.in_spin_loop
= val
;
1342 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1344 vcpu
->spin_loop
.dy_eligible
= val
;
1347 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1349 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1353 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1356 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1358 struct kvm_vcpu
*kvm_get_running_vcpu(void);
1359 struct kvm_vcpu
* __percpu
*kvm_get_running_vcpus(void);
1361 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1362 bool kvm_arch_has_irq_bypass(void);
1363 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer
*,
1364 struct irq_bypass_producer
*);
1365 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer
*,
1366 struct irq_bypass_producer
*);
1367 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer
*);
1368 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer
*);
1369 int kvm_arch_update_irqfd_routing(struct kvm
*kvm
, unsigned int host_irq
,
1370 uint32_t guest_irq
, bool set
);
1371 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1373 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1374 /* If we wakeup during the poll time, was it a sucessful poll? */
1375 static inline bool vcpu_valid_wakeup(struct kvm_vcpu
*vcpu
)
1377 return vcpu
->valid_wakeup
;
1381 static inline bool vcpu_valid_wakeup(struct kvm_vcpu
*vcpu
)
1385 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
1387 #ifdef CONFIG_HAVE_KVM_NO_POLL
1388 /* Callback that tells if we must not poll */
1389 bool kvm_arch_no_poll(struct kvm_vcpu
*vcpu
);
1391 static inline bool kvm_arch_no_poll(struct kvm_vcpu
*vcpu
)
1395 #endif /* CONFIG_HAVE_KVM_NO_POLL */
1397 #ifdef CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL
1398 long kvm_arch_vcpu_async_ioctl(struct file
*filp
,
1399 unsigned int ioctl
, unsigned long arg
);
1401 static inline long kvm_arch_vcpu_async_ioctl(struct file
*filp
,
1405 return -ENOIOCTLCMD
;
1407 #endif /* CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL */
1409 int kvm_arch_mmu_notifier_invalidate_range(struct kvm
*kvm
,
1410 unsigned long start
, unsigned long end
, bool blockable
);
1412 #ifdef CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE
1413 int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu
*vcpu
);
1415 static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu
*vcpu
)
1419 #endif /* CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE */
1421 typedef int (*kvm_vm_thread_fn_t
)(struct kvm
*kvm
, uintptr_t data
);
1423 int kvm_vm_create_worker_thread(struct kvm
*kvm
, kvm_vm_thread_fn_t thread_fn
,
1424 uintptr_t data
, const char *name
,
1425 struct task_struct
**thread_ptr
);