5 * This work is licensed under the terms of the GNU GPL, version 2. See
6 * the COPYING file in the top-level directory.
9 #include <linux/types.h>
10 #include <linux/hardirq.h>
11 #include <linux/list.h>
12 #include <linux/mutex.h>
13 #include <linux/spinlock.h>
14 #include <linux/signal.h>
15 #include <linux/sched.h>
16 #include <linux/bug.h>
18 #include <linux/mmu_notifier.h>
19 #include <linux/preempt.h>
20 #include <linux/msi.h>
21 #include <linux/slab.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/swait.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)
48 #define KVM_MEMSLOT_INCOHERENT (1UL << 17)
50 /* Two fragments for cross MMIO pages. */
51 #define KVM_MAX_MMIO_FRAGMENTS 2
53 #ifndef KVM_ADDRESS_SPACE_NUM
54 #define KVM_ADDRESS_SPACE_NUM 1
58 * For the normal pfn, the highest 12 bits should be zero,
59 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
60 * mask bit 63 to indicate the noslot pfn.
62 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
63 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
64 #define KVM_PFN_NOSLOT (0x1ULL << 63)
66 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
67 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
68 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
71 * error pfns indicate that the gfn is in slot but faild to
72 * translate it to pfn on host.
74 static inline bool is_error_pfn(kvm_pfn_t pfn
)
76 return !!(pfn
& KVM_PFN_ERR_MASK
);
80 * error_noslot pfns indicate that the gfn can not be
81 * translated to pfn - it is not in slot or failed to
82 * translate it to pfn.
84 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn
)
86 return !!(pfn
& KVM_PFN_ERR_NOSLOT_MASK
);
89 /* noslot pfn indicates that the gfn is not in slot. */
90 static inline bool is_noslot_pfn(kvm_pfn_t pfn
)
92 return pfn
== KVM_PFN_NOSLOT
;
96 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
97 * provide own defines and kvm_is_error_hva
99 #ifndef KVM_HVA_ERR_BAD
101 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
102 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
104 static inline bool kvm_is_error_hva(unsigned long addr
)
106 return addr
>= PAGE_OFFSET
;
111 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
113 static inline bool is_error_page(struct page
*page
)
119 * Architecture-independent vcpu->requests bit members
120 * Bits 4-7 are reserved for more arch-independent bits.
122 #define KVM_REQ_TLB_FLUSH 0
123 #define KVM_REQ_MMU_RELOAD 1
124 #define KVM_REQ_PENDING_TIMER 2
125 #define KVM_REQ_UNHALT 3
127 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
128 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
130 extern struct kmem_cache
*kvm_vcpu_cache
;
132 extern spinlock_t kvm_lock
;
133 extern struct list_head vm_list
;
135 struct kvm_io_range
{
138 struct kvm_io_device
*dev
;
141 #define NR_IOBUS_DEVS 1000
146 struct kvm_io_range range
[];
152 KVM_VIRTIO_CCW_NOTIFY_BUS
,
157 int kvm_io_bus_write(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
158 int len
, const void *val
);
159 int kvm_io_bus_write_cookie(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
,
160 gpa_t addr
, int len
, const void *val
, long cookie
);
161 int kvm_io_bus_read(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
163 int kvm_io_bus_register_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
,
164 int len
, struct kvm_io_device
*dev
);
165 int kvm_io_bus_unregister_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
166 struct kvm_io_device
*dev
);
167 struct kvm_io_device
*kvm_io_bus_get_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
170 #ifdef CONFIG_KVM_ASYNC_PF
171 struct kvm_async_pf
{
172 struct work_struct work
;
173 struct list_head link
;
174 struct list_head queue
;
175 struct kvm_vcpu
*vcpu
;
176 struct mm_struct
*mm
;
179 struct kvm_arch_async_pf arch
;
183 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu
*vcpu
);
184 void kvm_check_async_pf_completion(struct kvm_vcpu
*vcpu
);
185 int kvm_setup_async_pf(struct kvm_vcpu
*vcpu
, gva_t gva
, unsigned long hva
,
186 struct kvm_arch_async_pf
*arch
);
187 int kvm_async_pf_wakeup_all(struct kvm_vcpu
*vcpu
);
194 READING_SHADOW_PAGE_TABLES
,
198 * Sometimes a large or cross-page mmio needs to be broken up into separate
199 * exits for userspace servicing.
201 struct kvm_mmio_fragment
{
209 #ifdef CONFIG_PREEMPT_NOTIFIERS
210 struct preempt_notifier preempt_notifier
;
216 unsigned long requests
;
217 unsigned long guest_debug
;
220 struct list_head blocked_vcpu_list
;
226 int guest_fpu_loaded
, guest_xcr0_loaded
;
227 struct swait_queue_head wq
;
231 struct kvm_vcpu_stat stat
;
232 unsigned int halt_poll_ns
;
235 #ifdef CONFIG_HAS_IOMEM
237 int mmio_read_completed
;
239 int mmio_cur_fragment
;
240 int mmio_nr_fragments
;
241 struct kvm_mmio_fragment mmio_fragments
[KVM_MAX_MMIO_FRAGMENTS
];
244 #ifdef CONFIG_KVM_ASYNC_PF
247 struct list_head queue
;
248 struct list_head done
;
253 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
255 * Cpu relax intercept or pause loop exit optimization
256 * in_spin_loop: set when a vcpu does a pause loop exit
257 * or cpu relax intercepted.
258 * dy_eligible: indicates whether vcpu is eligible for directed yield.
266 struct kvm_vcpu_arch arch
;
267 struct dentry
*debugfs_dentry
;
270 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu
*vcpu
)
272 return cmpxchg(&vcpu
->mode
, IN_GUEST_MODE
, EXITING_GUEST_MODE
);
276 * Some of the bitops functions do not support too long bitmaps.
277 * This number must be determined not to exceed such limits.
279 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
281 struct kvm_memory_slot
{
283 unsigned long npages
;
284 unsigned long *dirty_bitmap
;
285 struct kvm_arch_memory_slot arch
;
286 unsigned long userspace_addr
;
291 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot
*memslot
)
293 return ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
296 struct kvm_s390_adapter_int
{
309 struct kvm_kernel_irq_routing_entry
{
312 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
313 struct kvm
*kvm
, int irq_source_id
, int level
,
327 struct kvm_s390_adapter_int adapter
;
328 struct kvm_hv_sint hv_sint
;
330 struct hlist_node link
;
333 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
334 struct kvm_irq_routing_table
{
335 int chip
[KVM_NR_IRQCHIPS
][KVM_IRQCHIP_NUM_PINS
];
338 * Array indexed by gsi. Each entry contains list of irq chips
339 * the gsi is connected to.
341 struct hlist_head map
[0];
345 #ifndef KVM_PRIVATE_MEM_SLOTS
346 #define KVM_PRIVATE_MEM_SLOTS 0
349 #ifndef KVM_MEM_SLOTS_NUM
350 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
353 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
354 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu
*vcpu
)
362 * memslots are not sorted by id anymore, please use id_to_memslot()
363 * to get the memslot by its id.
365 struct kvm_memslots
{
367 struct kvm_memory_slot memslots
[KVM_MEM_SLOTS_NUM
];
368 /* The mapping table from slot id to the index in memslots[]. */
369 short id_to_index
[KVM_MEM_SLOTS_NUM
];
376 struct mutex slots_lock
;
377 struct mm_struct
*mm
; /* userspace tied to this vm */
378 struct kvm_memslots
*memslots
[KVM_ADDRESS_SPACE_NUM
];
379 struct srcu_struct srcu
;
380 struct srcu_struct irq_srcu
;
381 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
384 * created_vcpus is protected by kvm->lock, and is incremented
385 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
386 * incremented after storing the kvm_vcpu pointer in vcpus,
387 * and is accessed atomically.
389 atomic_t online_vcpus
;
391 int last_boosted_vcpu
;
392 struct list_head vm_list
;
394 struct kvm_io_bus
*buses
[KVM_NR_BUSES
];
395 #ifdef CONFIG_HAVE_KVM_EVENTFD
398 struct list_head items
;
399 struct list_head resampler_list
;
400 struct mutex resampler_lock
;
402 struct list_head ioeventfds
;
404 struct kvm_vm_stat stat
;
405 struct kvm_arch arch
;
406 atomic_t users_count
;
407 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
408 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
409 spinlock_t ring_lock
;
410 struct list_head coalesced_zones
;
413 struct mutex irq_lock
;
414 #ifdef CONFIG_HAVE_KVM_IRQCHIP
416 * Update side is protected by irq_lock.
418 struct kvm_irq_routing_table __rcu
*irq_routing
;
420 #ifdef CONFIG_HAVE_KVM_IRQFD
421 struct hlist_head irq_ack_notifier_list
;
424 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
425 struct mmu_notifier mmu_notifier
;
426 unsigned long mmu_notifier_seq
;
427 long mmu_notifier_count
;
430 struct list_head devices
;
431 struct dentry
*debugfs_dentry
;
432 struct kvm_stat_data
**debugfs_stat_data
;
435 #define kvm_err(fmt, ...) \
436 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
437 #define kvm_info(fmt, ...) \
438 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
439 #define kvm_debug(fmt, ...) \
440 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
441 #define kvm_debug_ratelimited(fmt, ...) \
442 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
444 #define kvm_pr_unimpl(fmt, ...) \
445 pr_err_ratelimited("kvm [%i]: " fmt, \
446 task_tgid_nr(current), ## __VA_ARGS__)
448 /* The guest did something we don't support. */
449 #define vcpu_unimpl(vcpu, fmt, ...) \
450 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
451 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
453 #define vcpu_debug(vcpu, fmt, ...) \
454 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
455 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
456 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
458 #define vcpu_err(vcpu, fmt, ...) \
459 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
461 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
463 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
464 * the caller has read kvm->online_vcpus before (as is the case
465 * for kvm_for_each_vcpu, for example).
468 return kvm
->vcpus
[i
];
471 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
473 idx < atomic_read(&kvm->online_vcpus) && \
474 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
477 static inline struct kvm_vcpu
*kvm_get_vcpu_by_id(struct kvm
*kvm
, int id
)
479 struct kvm_vcpu
*vcpu
= NULL
;
484 if (id
< KVM_MAX_VCPUS
)
485 vcpu
= kvm_get_vcpu(kvm
, id
);
486 if (vcpu
&& vcpu
->vcpu_id
== id
)
488 kvm_for_each_vcpu(i
, vcpu
, kvm
)
489 if (vcpu
->vcpu_id
== id
)
494 #define kvm_for_each_memslot(memslot, slots) \
495 for (memslot = &slots->memslots[0]; \
496 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
499 int kvm_vcpu_init(struct kvm_vcpu
*vcpu
, struct kvm
*kvm
, unsigned id
);
500 void kvm_vcpu_uninit(struct kvm_vcpu
*vcpu
);
502 int __must_check
vcpu_load(struct kvm_vcpu
*vcpu
);
503 void vcpu_put(struct kvm_vcpu
*vcpu
);
505 #ifdef __KVM_HAVE_IOAPIC
506 void kvm_vcpu_request_scan_ioapic(struct kvm
*kvm
);
507 void kvm_arch_post_irq_routing_update(struct kvm
*kvm
);
509 static inline void kvm_vcpu_request_scan_ioapic(struct kvm
*kvm
)
512 static inline void kvm_arch_post_irq_routing_update(struct kvm
*kvm
)
517 #ifdef CONFIG_HAVE_KVM_IRQFD
518 int kvm_irqfd_init(void);
519 void kvm_irqfd_exit(void);
521 static inline int kvm_irqfd_init(void)
526 static inline void kvm_irqfd_exit(void)
530 int kvm_init(void *opaque
, unsigned vcpu_size
, unsigned vcpu_align
,
531 struct module
*module
);
534 void kvm_get_kvm(struct kvm
*kvm
);
535 void kvm_put_kvm(struct kvm
*kvm
);
537 static inline struct kvm_memslots
*__kvm_memslots(struct kvm
*kvm
, int as_id
)
539 return rcu_dereference_check(kvm
->memslots
[as_id
],
540 srcu_read_lock_held(&kvm
->srcu
)
541 || lockdep_is_held(&kvm
->slots_lock
));
544 static inline struct kvm_memslots
*kvm_memslots(struct kvm
*kvm
)
546 return __kvm_memslots(kvm
, 0);
549 static inline struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
)
551 int as_id
= kvm_arch_vcpu_memslots_id(vcpu
);
553 return __kvm_memslots(vcpu
->kvm
, as_id
);
556 static inline struct kvm_memory_slot
*
557 id_to_memslot(struct kvm_memslots
*slots
, int id
)
559 int index
= slots
->id_to_index
[id
];
560 struct kvm_memory_slot
*slot
;
562 slot
= &slots
->memslots
[index
];
564 WARN_ON(slot
->id
!= id
);
569 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
570 * - create a new memory slot
571 * - delete an existing memory slot
572 * - modify an existing memory slot
573 * -- move it in the guest physical memory space
574 * -- just change its flags
576 * Since flags can be changed by some of these operations, the following
577 * differentiation is the best we can do for __kvm_set_memory_region():
586 int kvm_set_memory_region(struct kvm
*kvm
,
587 const struct kvm_userspace_memory_region
*mem
);
588 int __kvm_set_memory_region(struct kvm
*kvm
,
589 const struct kvm_userspace_memory_region
*mem
);
590 void kvm_arch_free_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*free
,
591 struct kvm_memory_slot
*dont
);
592 int kvm_arch_create_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*slot
,
593 unsigned long npages
);
594 void kvm_arch_memslots_updated(struct kvm
*kvm
, struct kvm_memslots
*slots
);
595 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
596 struct kvm_memory_slot
*memslot
,
597 const struct kvm_userspace_memory_region
*mem
,
598 enum kvm_mr_change change
);
599 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
600 const struct kvm_userspace_memory_region
*mem
,
601 const struct kvm_memory_slot
*old
,
602 const struct kvm_memory_slot
*new,
603 enum kvm_mr_change change
);
604 bool kvm_largepages_enabled(void);
605 void kvm_disable_largepages(void);
606 /* flush all memory translations */
607 void kvm_arch_flush_shadow_all(struct kvm
*kvm
);
608 /* flush memory translations pointing to 'slot' */
609 void kvm_arch_flush_shadow_memslot(struct kvm
*kvm
,
610 struct kvm_memory_slot
*slot
);
612 int gfn_to_page_many_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
,
613 struct page
**pages
, int nr_pages
);
615 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
616 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
617 unsigned long gfn_to_hva_prot(struct kvm
*kvm
, gfn_t gfn
, bool *writable
);
618 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
619 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
621 void kvm_release_page_clean(struct page
*page
);
622 void kvm_release_page_dirty(struct page
*page
);
623 void kvm_set_page_accessed(struct page
*page
);
625 kvm_pfn_t
gfn_to_pfn_atomic(struct kvm
*kvm
, gfn_t gfn
);
626 kvm_pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
627 kvm_pfn_t
gfn_to_pfn_prot(struct kvm
*kvm
, gfn_t gfn
, bool write_fault
,
629 kvm_pfn_t
gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
630 kvm_pfn_t
gfn_to_pfn_memslot_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
);
631 kvm_pfn_t
__gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
632 bool atomic
, bool *async
, bool write_fault
,
635 void kvm_release_pfn_clean(kvm_pfn_t pfn
);
636 void kvm_set_pfn_dirty(kvm_pfn_t pfn
);
637 void kvm_set_pfn_accessed(kvm_pfn_t pfn
);
638 void kvm_get_pfn(kvm_pfn_t pfn
);
640 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
642 int kvm_read_guest_atomic(struct kvm
*kvm
, gpa_t gpa
, void *data
,
644 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
645 int kvm_read_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
646 void *data
, unsigned long len
);
647 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
648 int offset
, int len
);
649 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
651 int kvm_write_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
652 void *data
, unsigned long len
);
653 int kvm_write_guest_offset_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
654 void *data
, int offset
, unsigned long len
);
655 int kvm_gfn_to_hva_cache_init(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
656 gpa_t gpa
, unsigned long len
);
657 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
658 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
659 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
660 bool kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
661 unsigned long kvm_host_page_size(struct kvm
*kvm
, gfn_t gfn
);
662 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
664 struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
);
665 struct kvm_memory_slot
*kvm_vcpu_gfn_to_memslot(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
666 kvm_pfn_t
kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
667 kvm_pfn_t
kvm_vcpu_gfn_to_pfn(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
668 struct page
*kvm_vcpu_gfn_to_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
669 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
670 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu
*vcpu
, gfn_t gfn
, bool *writable
);
671 int kvm_vcpu_read_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, void *data
, int offset
,
673 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
675 int kvm_vcpu_read_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
677 int kvm_vcpu_write_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, const void *data
,
678 int offset
, int len
);
679 int kvm_vcpu_write_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, const void *data
,
681 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
683 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
684 void kvm_arch_vcpu_blocking(struct kvm_vcpu
*vcpu
);
685 void kvm_arch_vcpu_unblocking(struct kvm_vcpu
*vcpu
);
686 void kvm_vcpu_wake_up(struct kvm_vcpu
*vcpu
);
687 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
688 int kvm_vcpu_yield_to(struct kvm_vcpu
*target
);
689 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
);
690 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
691 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
693 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
694 void kvm_reload_remote_mmus(struct kvm
*kvm
);
695 bool kvm_make_all_cpus_request(struct kvm
*kvm
, unsigned int req
);
697 long kvm_arch_dev_ioctl(struct file
*filp
,
698 unsigned int ioctl
, unsigned long arg
);
699 long kvm_arch_vcpu_ioctl(struct file
*filp
,
700 unsigned int ioctl
, unsigned long arg
);
701 int kvm_arch_vcpu_fault(struct kvm_vcpu
*vcpu
, struct vm_fault
*vmf
);
703 int kvm_vm_ioctl_check_extension(struct kvm
*kvm
, long ext
);
705 int kvm_get_dirty_log(struct kvm
*kvm
,
706 struct kvm_dirty_log
*log
, int *is_dirty
);
708 int kvm_get_dirty_log_protect(struct kvm
*kvm
,
709 struct kvm_dirty_log
*log
, bool *is_dirty
);
711 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm
*kvm
,
712 struct kvm_memory_slot
*slot
,
716 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
717 struct kvm_dirty_log
*log
);
719 int kvm_vm_ioctl_irq_line(struct kvm
*kvm
, struct kvm_irq_level
*irq_level
,
721 long kvm_arch_vm_ioctl(struct file
*filp
,
722 unsigned int ioctl
, unsigned long arg
);
724 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
725 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
727 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
728 struct kvm_translation
*tr
);
730 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
731 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
732 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
733 struct kvm_sregs
*sregs
);
734 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
735 struct kvm_sregs
*sregs
);
736 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
737 struct kvm_mp_state
*mp_state
);
738 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
739 struct kvm_mp_state
*mp_state
);
740 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
741 struct kvm_guest_debug
*dbg
);
742 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
744 int kvm_arch_init(void *opaque
);
745 void kvm_arch_exit(void);
747 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
);
748 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
);
750 void kvm_arch_sched_in(struct kvm_vcpu
*vcpu
, int cpu
);
752 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
);
753 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
754 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
755 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
);
756 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
);
757 void kvm_arch_vcpu_postcreate(struct kvm_vcpu
*vcpu
);
758 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
760 bool kvm_arch_has_vcpu_debugfs(void);
761 int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu
*vcpu
);
763 int kvm_arch_hardware_enable(void);
764 void kvm_arch_hardware_disable(void);
765 int kvm_arch_hardware_setup(void);
766 void kvm_arch_hardware_unsetup(void);
767 void kvm_arch_check_processor_compat(void *rtn
);
768 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
769 int kvm_arch_vcpu_should_kick(struct kvm_vcpu
*vcpu
);
771 void *kvm_kvzalloc(unsigned long size
);
773 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
774 static inline struct kvm
*kvm_arch_alloc_vm(void)
776 return kzalloc(sizeof(struct kvm
), GFP_KERNEL
);
779 static inline void kvm_arch_free_vm(struct kvm
*kvm
)
785 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
786 void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
);
787 void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
);
788 bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
);
790 static inline void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
)
794 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
)
798 static inline bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
)
803 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
804 void kvm_arch_start_assignment(struct kvm
*kvm
);
805 void kvm_arch_end_assignment(struct kvm
*kvm
);
806 bool kvm_arch_has_assigned_device(struct kvm
*kvm
);
808 static inline void kvm_arch_start_assignment(struct kvm
*kvm
)
812 static inline void kvm_arch_end_assignment(struct kvm
*kvm
)
816 static inline bool kvm_arch_has_assigned_device(struct kvm
*kvm
)
822 static inline struct swait_queue_head
*kvm_arch_vcpu_wq(struct kvm_vcpu
*vcpu
)
824 #ifdef __KVM_HAVE_ARCH_WQP
825 return vcpu
->arch
.wqp
;
831 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
833 * returns true if the virtual interrupt controller is initialized and
834 * ready to accept virtual IRQ. On some architectures the virtual interrupt
835 * controller is dynamically instantiated and this is not always true.
837 bool kvm_arch_intc_initialized(struct kvm
*kvm
);
839 static inline bool kvm_arch_intc_initialized(struct kvm
*kvm
)
845 int kvm_arch_init_vm(struct kvm
*kvm
, unsigned long type
);
846 void kvm_arch_destroy_vm(struct kvm
*kvm
);
847 void kvm_arch_sync_events(struct kvm
*kvm
);
849 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
850 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
852 bool kvm_is_reserved_pfn(kvm_pfn_t pfn
);
854 struct kvm_irq_ack_notifier
{
855 struct hlist_node link
;
857 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
860 int kvm_irq_map_gsi(struct kvm
*kvm
,
861 struct kvm_kernel_irq_routing_entry
*entries
, int gsi
);
862 int kvm_irq_map_chip_pin(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
864 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
,
866 int kvm_set_msi(struct kvm_kernel_irq_routing_entry
*irq_entry
, struct kvm
*kvm
,
867 int irq_source_id
, int level
, bool line_status
);
868 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry
*e
,
869 struct kvm
*kvm
, int irq_source_id
,
870 int level
, bool line_status
);
871 bool kvm_irq_has_notifier(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
872 void kvm_notify_acked_gsi(struct kvm
*kvm
, int gsi
);
873 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
874 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
875 struct kvm_irq_ack_notifier
*kian
);
876 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
877 struct kvm_irq_ack_notifier
*kian
);
878 int kvm_request_irq_source_id(struct kvm
*kvm
);
879 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
881 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
882 int kvm_iommu_map_pages(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
883 void kvm_iommu_unmap_pages(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
885 static inline int kvm_iommu_map_pages(struct kvm
*kvm
,
886 struct kvm_memory_slot
*slot
)
891 static inline void kvm_iommu_unmap_pages(struct kvm
*kvm
,
892 struct kvm_memory_slot
*slot
)
898 * search_memslots() and __gfn_to_memslot() are here because they are
899 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
900 * gfn_to_memslot() itself isn't here as an inline because that would
901 * bloat other code too much.
903 static inline struct kvm_memory_slot
*
904 search_memslots(struct kvm_memslots
*slots
, gfn_t gfn
)
906 int start
= 0, end
= slots
->used_slots
;
907 int slot
= atomic_read(&slots
->lru_slot
);
908 struct kvm_memory_slot
*memslots
= slots
->memslots
;
910 if (gfn
>= memslots
[slot
].base_gfn
&&
911 gfn
< memslots
[slot
].base_gfn
+ memslots
[slot
].npages
)
912 return &memslots
[slot
];
914 while (start
< end
) {
915 slot
= start
+ (end
- start
) / 2;
917 if (gfn
>= memslots
[slot
].base_gfn
)
923 if (gfn
>= memslots
[start
].base_gfn
&&
924 gfn
< memslots
[start
].base_gfn
+ memslots
[start
].npages
) {
925 atomic_set(&slots
->lru_slot
, start
);
926 return &memslots
[start
];
932 static inline struct kvm_memory_slot
*
933 __gfn_to_memslot(struct kvm_memslots
*slots
, gfn_t gfn
)
935 return search_memslots(slots
, gfn
);
938 static inline unsigned long
939 __gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
)
941 return slot
->userspace_addr
+ (gfn
- slot
->base_gfn
) * PAGE_SIZE
;
944 static inline int memslot_id(struct kvm
*kvm
, gfn_t gfn
)
946 return gfn_to_memslot(kvm
, gfn
)->id
;
950 hva_to_gfn_memslot(unsigned long hva
, struct kvm_memory_slot
*slot
)
952 gfn_t gfn_offset
= (hva
- slot
->userspace_addr
) >> PAGE_SHIFT
;
954 return slot
->base_gfn
+ gfn_offset
;
957 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
959 return (gpa_t
)gfn
<< PAGE_SHIFT
;
962 static inline gfn_t
gpa_to_gfn(gpa_t gpa
)
964 return (gfn_t
)(gpa
>> PAGE_SHIFT
);
967 static inline hpa_t
pfn_to_hpa(kvm_pfn_t pfn
)
969 return (hpa_t
)pfn
<< PAGE_SHIFT
;
972 static inline bool kvm_is_error_gpa(struct kvm
*kvm
, gpa_t gpa
)
974 unsigned long hva
= gfn_to_hva(kvm
, gpa_to_gfn(gpa
));
976 return kvm_is_error_hva(hva
);
984 struct kvm_stat_data
{
989 struct kvm_stats_debugfs_item
{
992 enum kvm_stat_kind kind
;
994 extern struct kvm_stats_debugfs_item debugfs_entries
[];
995 extern struct dentry
*kvm_debugfs_dir
;
997 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
998 static inline int mmu_notifier_retry(struct kvm
*kvm
, unsigned long mmu_seq
)
1000 if (unlikely(kvm
->mmu_notifier_count
))
1003 * Ensure the read of mmu_notifier_count happens before the read
1004 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1005 * mmu_notifier_invalidate_range_end to make sure that the caller
1006 * either sees the old (non-zero) value of mmu_notifier_count or
1007 * the new (incremented) value of mmu_notifier_seq.
1008 * PowerPC Book3s HV KVM calls this under a per-page lock
1009 * rather than under kvm->mmu_lock, for scalability, so
1010 * can't rely on kvm->mmu_lock to keep things ordered.
1013 if (kvm
->mmu_notifier_seq
!= mmu_seq
)
1019 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1022 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1023 #elif defined(CONFIG_ARM64)
1024 #define KVM_MAX_IRQ_ROUTES 4096
1026 #define KVM_MAX_IRQ_ROUTES 1024
1029 int kvm_set_irq_routing(struct kvm
*kvm
,
1030 const struct kvm_irq_routing_entry
*entries
,
1033 int kvm_set_routing_entry(struct kvm
*kvm
,
1034 struct kvm_kernel_irq_routing_entry
*e
,
1035 const struct kvm_irq_routing_entry
*ue
);
1036 void kvm_free_irq_routing(struct kvm
*kvm
);
1040 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
1044 int kvm_send_userspace_msi(struct kvm
*kvm
, struct kvm_msi
*msi
);
1046 #ifdef CONFIG_HAVE_KVM_EVENTFD
1048 void kvm_eventfd_init(struct kvm
*kvm
);
1049 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
1051 #ifdef CONFIG_HAVE_KVM_IRQFD
1052 int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
);
1053 void kvm_irqfd_release(struct kvm
*kvm
);
1054 void kvm_irq_routing_update(struct kvm
*);
1056 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1061 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1066 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
1068 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1073 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1075 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1076 static inline void kvm_irq_routing_update(struct kvm
*kvm
)
1080 void kvm_arch_irq_routing_update(struct kvm
*kvm
);
1082 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
1087 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1089 static inline void kvm_make_request(int req
, struct kvm_vcpu
*vcpu
)
1092 * Ensure the rest of the request is published to kvm_check_request's
1093 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1096 set_bit(req
, &vcpu
->requests
);
1099 static inline bool kvm_check_request(int req
, struct kvm_vcpu
*vcpu
)
1101 if (test_bit(req
, &vcpu
->requests
)) {
1102 clear_bit(req
, &vcpu
->requests
);
1105 * Ensure the rest of the request is visible to kvm_check_request's
1106 * caller. Paired with the smp_wmb in kvm_make_request.
1108 smp_mb__after_atomic();
1115 extern bool kvm_rebooting
;
1117 extern unsigned int halt_poll_ns
;
1118 extern unsigned int halt_poll_ns_grow
;
1119 extern unsigned int halt_poll_ns_shrink
;
1122 struct kvm_device_ops
*ops
;
1125 struct list_head vm_node
;
1128 /* create, destroy, and name are mandatory */
1129 struct kvm_device_ops
{
1133 * create is called holding kvm->lock and any operations not suitable
1134 * to do while holding the lock should be deferred to init (see
1137 int (*create
)(struct kvm_device
*dev
, u32 type
);
1140 * init is called after create if create is successful and is called
1141 * outside of holding kvm->lock.
1143 void (*init
)(struct kvm_device
*dev
);
1146 * Destroy is responsible for freeing dev.
1148 * Destroy may be called before or after destructors are called
1149 * on emulated I/O regions, depending on whether a reference is
1150 * held by a vcpu or other kvm component that gets destroyed
1151 * after the emulated I/O.
1153 void (*destroy
)(struct kvm_device
*dev
);
1155 int (*set_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1156 int (*get_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1157 int (*has_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1158 long (*ioctl
)(struct kvm_device
*dev
, unsigned int ioctl
,
1162 void kvm_device_get(struct kvm_device
*dev
);
1163 void kvm_device_put(struct kvm_device
*dev
);
1164 struct kvm_device
*kvm_device_from_filp(struct file
*filp
);
1165 int kvm_register_device_ops(struct kvm_device_ops
*ops
, u32 type
);
1166 void kvm_unregister_device_ops(u32 type
);
1168 extern struct kvm_device_ops kvm_mpic_ops
;
1169 extern struct kvm_device_ops kvm_xics_ops
;
1170 extern struct kvm_device_ops kvm_arm_vgic_v2_ops
;
1171 extern struct kvm_device_ops kvm_arm_vgic_v3_ops
;
1173 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1175 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1177 vcpu
->spin_loop
.in_spin_loop
= val
;
1179 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1181 vcpu
->spin_loop
.dy_eligible
= val
;
1184 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1186 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1190 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1193 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1195 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1196 bool kvm_arch_has_irq_bypass(void);
1197 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer
*,
1198 struct irq_bypass_producer
*);
1199 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer
*,
1200 struct irq_bypass_producer
*);
1201 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer
*);
1202 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer
*);
1203 int kvm_arch_update_irqfd_routing(struct kvm
*kvm
, unsigned int host_irq
,
1204 uint32_t guest_irq
, bool set
);
1205 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1207 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1208 /* If we wakeup during the poll time, was it a sucessful poll? */
1209 static inline bool vcpu_valid_wakeup(struct kvm_vcpu
*vcpu
)
1211 return vcpu
->valid_wakeup
;
1215 static inline bool vcpu_valid_wakeup(struct kvm_vcpu
*vcpu
)
1219 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */