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 unsigned char fpu_counter
;
228 struct swait_queue_head wq
;
232 struct kvm_vcpu_stat stat
;
233 unsigned int halt_poll_ns
;
236 #ifdef CONFIG_HAS_IOMEM
238 int mmio_read_completed
;
240 int mmio_cur_fragment
;
241 int mmio_nr_fragments
;
242 struct kvm_mmio_fragment mmio_fragments
[KVM_MAX_MMIO_FRAGMENTS
];
245 #ifdef CONFIG_KVM_ASYNC_PF
248 struct list_head queue
;
249 struct list_head done
;
254 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
256 * Cpu relax intercept or pause loop exit optimization
257 * in_spin_loop: set when a vcpu does a pause loop exit
258 * or cpu relax intercepted.
259 * dy_eligible: indicates whether vcpu is eligible for directed yield.
267 struct kvm_vcpu_arch arch
;
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_pr_unimpl(fmt, ...) \
442 pr_err_ratelimited("kvm [%i]: " fmt, \
443 task_tgid_nr(current), ## __VA_ARGS__)
445 /* The guest did something we don't support. */
446 #define vcpu_unimpl(vcpu, fmt, ...) \
447 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
448 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
450 #define vcpu_debug(vcpu, fmt, ...) \
451 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
452 #define vcpu_err(vcpu, fmt, ...) \
453 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
455 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
457 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
458 * the caller has read kvm->online_vcpus before (as is the case
459 * for kvm_for_each_vcpu, for example).
462 return kvm
->vcpus
[i
];
465 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
467 idx < atomic_read(&kvm->online_vcpus) && \
468 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
471 static inline struct kvm_vcpu
*kvm_get_vcpu_by_id(struct kvm
*kvm
, int id
)
473 struct kvm_vcpu
*vcpu
= NULL
;
478 if (id
< KVM_MAX_VCPUS
)
479 vcpu
= kvm_get_vcpu(kvm
, id
);
480 if (vcpu
&& vcpu
->vcpu_id
== id
)
482 kvm_for_each_vcpu(i
, vcpu
, kvm
)
483 if (vcpu
->vcpu_id
== id
)
488 #define kvm_for_each_memslot(memslot, slots) \
489 for (memslot = &slots->memslots[0]; \
490 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
493 int kvm_vcpu_init(struct kvm_vcpu
*vcpu
, struct kvm
*kvm
, unsigned id
);
494 void kvm_vcpu_uninit(struct kvm_vcpu
*vcpu
);
496 int __must_check
vcpu_load(struct kvm_vcpu
*vcpu
);
497 void vcpu_put(struct kvm_vcpu
*vcpu
);
499 #ifdef __KVM_HAVE_IOAPIC
500 void kvm_vcpu_request_scan_ioapic(struct kvm
*kvm
);
501 void kvm_arch_post_irq_routing_update(struct kvm
*kvm
);
503 static inline void kvm_vcpu_request_scan_ioapic(struct kvm
*kvm
)
506 static inline void kvm_arch_post_irq_routing_update(struct kvm
*kvm
)
511 #ifdef CONFIG_HAVE_KVM_IRQFD
512 int kvm_irqfd_init(void);
513 void kvm_irqfd_exit(void);
515 static inline int kvm_irqfd_init(void)
520 static inline void kvm_irqfd_exit(void)
524 int kvm_init(void *opaque
, unsigned vcpu_size
, unsigned vcpu_align
,
525 struct module
*module
);
528 void kvm_get_kvm(struct kvm
*kvm
);
529 void kvm_put_kvm(struct kvm
*kvm
);
531 static inline struct kvm_memslots
*__kvm_memslots(struct kvm
*kvm
, int as_id
)
533 return rcu_dereference_check(kvm
->memslots
[as_id
],
534 srcu_read_lock_held(&kvm
->srcu
)
535 || lockdep_is_held(&kvm
->slots_lock
));
538 static inline struct kvm_memslots
*kvm_memslots(struct kvm
*kvm
)
540 return __kvm_memslots(kvm
, 0);
543 static inline struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
)
545 int as_id
= kvm_arch_vcpu_memslots_id(vcpu
);
547 return __kvm_memslots(vcpu
->kvm
, as_id
);
550 static inline struct kvm_memory_slot
*
551 id_to_memslot(struct kvm_memslots
*slots
, int id
)
553 int index
= slots
->id_to_index
[id
];
554 struct kvm_memory_slot
*slot
;
556 slot
= &slots
->memslots
[index
];
558 WARN_ON(slot
->id
!= id
);
563 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
564 * - create a new memory slot
565 * - delete an existing memory slot
566 * - modify an existing memory slot
567 * -- move it in the guest physical memory space
568 * -- just change its flags
570 * Since flags can be changed by some of these operations, the following
571 * differentiation is the best we can do for __kvm_set_memory_region():
580 int kvm_set_memory_region(struct kvm
*kvm
,
581 const struct kvm_userspace_memory_region
*mem
);
582 int __kvm_set_memory_region(struct kvm
*kvm
,
583 const struct kvm_userspace_memory_region
*mem
);
584 void kvm_arch_free_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*free
,
585 struct kvm_memory_slot
*dont
);
586 int kvm_arch_create_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*slot
,
587 unsigned long npages
);
588 void kvm_arch_memslots_updated(struct kvm
*kvm
, struct kvm_memslots
*slots
);
589 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
590 struct kvm_memory_slot
*memslot
,
591 const struct kvm_userspace_memory_region
*mem
,
592 enum kvm_mr_change change
);
593 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
594 const struct kvm_userspace_memory_region
*mem
,
595 const struct kvm_memory_slot
*old
,
596 const struct kvm_memory_slot
*new,
597 enum kvm_mr_change change
);
598 bool kvm_largepages_enabled(void);
599 void kvm_disable_largepages(void);
600 /* flush all memory translations */
601 void kvm_arch_flush_shadow_all(struct kvm
*kvm
);
602 /* flush memory translations pointing to 'slot' */
603 void kvm_arch_flush_shadow_memslot(struct kvm
*kvm
,
604 struct kvm_memory_slot
*slot
);
606 int gfn_to_page_many_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
,
607 struct page
**pages
, int nr_pages
);
609 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
610 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
611 unsigned long gfn_to_hva_prot(struct kvm
*kvm
, gfn_t gfn
, bool *writable
);
612 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
613 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
615 void kvm_release_page_clean(struct page
*page
);
616 void kvm_release_page_dirty(struct page
*page
);
617 void kvm_set_page_accessed(struct page
*page
);
619 kvm_pfn_t
gfn_to_pfn_atomic(struct kvm
*kvm
, gfn_t gfn
);
620 kvm_pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
621 kvm_pfn_t
gfn_to_pfn_prot(struct kvm
*kvm
, gfn_t gfn
, bool write_fault
,
623 kvm_pfn_t
gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
624 kvm_pfn_t
gfn_to_pfn_memslot_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
);
625 kvm_pfn_t
__gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
626 bool atomic
, bool *async
, bool write_fault
,
629 void kvm_release_pfn_clean(kvm_pfn_t pfn
);
630 void kvm_set_pfn_dirty(kvm_pfn_t pfn
);
631 void kvm_set_pfn_accessed(kvm_pfn_t pfn
);
632 void kvm_get_pfn(kvm_pfn_t pfn
);
634 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
636 int kvm_read_guest_atomic(struct kvm
*kvm
, gpa_t gpa
, void *data
,
638 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
639 int kvm_read_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
640 void *data
, unsigned long len
);
641 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
642 int offset
, int len
);
643 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
645 int kvm_write_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
646 void *data
, unsigned long len
);
647 int kvm_gfn_to_hva_cache_init(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
648 gpa_t gpa
, unsigned long len
);
649 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
650 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
651 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
652 bool kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
653 unsigned long kvm_host_page_size(struct kvm
*kvm
, gfn_t gfn
);
654 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
656 struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
);
657 struct kvm_memory_slot
*kvm_vcpu_gfn_to_memslot(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
658 kvm_pfn_t
kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
659 kvm_pfn_t
kvm_vcpu_gfn_to_pfn(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
660 struct page
*kvm_vcpu_gfn_to_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
661 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
662 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu
*vcpu
, gfn_t gfn
, bool *writable
);
663 int kvm_vcpu_read_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, void *data
, int offset
,
665 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
667 int kvm_vcpu_read_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
669 int kvm_vcpu_write_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, const void *data
,
670 int offset
, int len
);
671 int kvm_vcpu_write_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, const void *data
,
673 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
675 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
676 void kvm_arch_vcpu_blocking(struct kvm_vcpu
*vcpu
);
677 void kvm_arch_vcpu_unblocking(struct kvm_vcpu
*vcpu
);
678 void kvm_vcpu_wake_up(struct kvm_vcpu
*vcpu
);
679 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
680 int kvm_vcpu_yield_to(struct kvm_vcpu
*target
);
681 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
);
682 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
683 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
685 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
686 void kvm_reload_remote_mmus(struct kvm
*kvm
);
687 bool kvm_make_all_cpus_request(struct kvm
*kvm
, unsigned int req
);
689 long kvm_arch_dev_ioctl(struct file
*filp
,
690 unsigned int ioctl
, unsigned long arg
);
691 long kvm_arch_vcpu_ioctl(struct file
*filp
,
692 unsigned int ioctl
, unsigned long arg
);
693 int kvm_arch_vcpu_fault(struct kvm_vcpu
*vcpu
, struct vm_fault
*vmf
);
695 int kvm_vm_ioctl_check_extension(struct kvm
*kvm
, long ext
);
697 int kvm_get_dirty_log(struct kvm
*kvm
,
698 struct kvm_dirty_log
*log
, int *is_dirty
);
700 int kvm_get_dirty_log_protect(struct kvm
*kvm
,
701 struct kvm_dirty_log
*log
, bool *is_dirty
);
703 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm
*kvm
,
704 struct kvm_memory_slot
*slot
,
708 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
709 struct kvm_dirty_log
*log
);
711 int kvm_vm_ioctl_irq_line(struct kvm
*kvm
, struct kvm_irq_level
*irq_level
,
713 long kvm_arch_vm_ioctl(struct file
*filp
,
714 unsigned int ioctl
, unsigned long arg
);
716 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
717 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
719 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
720 struct kvm_translation
*tr
);
722 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
723 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
724 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
725 struct kvm_sregs
*sregs
);
726 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
727 struct kvm_sregs
*sregs
);
728 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
729 struct kvm_mp_state
*mp_state
);
730 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
731 struct kvm_mp_state
*mp_state
);
732 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
733 struct kvm_guest_debug
*dbg
);
734 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
736 int kvm_arch_init(void *opaque
);
737 void kvm_arch_exit(void);
739 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
);
740 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
);
742 void kvm_arch_sched_in(struct kvm_vcpu
*vcpu
, int cpu
);
744 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
);
745 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
746 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
747 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
);
748 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
);
749 void kvm_arch_vcpu_postcreate(struct kvm_vcpu
*vcpu
);
750 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
752 int kvm_arch_hardware_enable(void);
753 void kvm_arch_hardware_disable(void);
754 int kvm_arch_hardware_setup(void);
755 void kvm_arch_hardware_unsetup(void);
756 void kvm_arch_check_processor_compat(void *rtn
);
757 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
758 int kvm_arch_vcpu_should_kick(struct kvm_vcpu
*vcpu
);
760 void *kvm_kvzalloc(unsigned long size
);
762 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
763 static inline struct kvm
*kvm_arch_alloc_vm(void)
765 return kzalloc(sizeof(struct kvm
), GFP_KERNEL
);
768 static inline void kvm_arch_free_vm(struct kvm
*kvm
)
774 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
775 void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
);
776 void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
);
777 bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
);
779 static inline void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
)
783 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
)
787 static inline bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
)
792 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
793 void kvm_arch_start_assignment(struct kvm
*kvm
);
794 void kvm_arch_end_assignment(struct kvm
*kvm
);
795 bool kvm_arch_has_assigned_device(struct kvm
*kvm
);
797 static inline void kvm_arch_start_assignment(struct kvm
*kvm
)
801 static inline void kvm_arch_end_assignment(struct kvm
*kvm
)
805 static inline bool kvm_arch_has_assigned_device(struct kvm
*kvm
)
811 static inline struct swait_queue_head
*kvm_arch_vcpu_wq(struct kvm_vcpu
*vcpu
)
813 #ifdef __KVM_HAVE_ARCH_WQP
814 return vcpu
->arch
.wqp
;
820 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
822 * returns true if the virtual interrupt controller is initialized and
823 * ready to accept virtual IRQ. On some architectures the virtual interrupt
824 * controller is dynamically instantiated and this is not always true.
826 bool kvm_arch_intc_initialized(struct kvm
*kvm
);
828 static inline bool kvm_arch_intc_initialized(struct kvm
*kvm
)
834 int kvm_arch_init_vm(struct kvm
*kvm
, unsigned long type
);
835 void kvm_arch_destroy_vm(struct kvm
*kvm
);
836 void kvm_arch_sync_events(struct kvm
*kvm
);
838 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
839 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
841 bool kvm_is_reserved_pfn(kvm_pfn_t pfn
);
843 struct kvm_irq_ack_notifier
{
844 struct hlist_node link
;
846 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
849 int kvm_irq_map_gsi(struct kvm
*kvm
,
850 struct kvm_kernel_irq_routing_entry
*entries
, int gsi
);
851 int kvm_irq_map_chip_pin(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
853 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
,
855 int kvm_set_msi(struct kvm_kernel_irq_routing_entry
*irq_entry
, struct kvm
*kvm
,
856 int irq_source_id
, int level
, bool line_status
);
857 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry
*e
,
858 struct kvm
*kvm
, int irq_source_id
,
859 int level
, bool line_status
);
860 bool kvm_irq_has_notifier(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
861 void kvm_notify_acked_gsi(struct kvm
*kvm
, int gsi
);
862 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
863 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
864 struct kvm_irq_ack_notifier
*kian
);
865 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
866 struct kvm_irq_ack_notifier
*kian
);
867 int kvm_request_irq_source_id(struct kvm
*kvm
);
868 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
870 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
871 int kvm_iommu_map_pages(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
872 void kvm_iommu_unmap_pages(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
874 static inline int kvm_iommu_map_pages(struct kvm
*kvm
,
875 struct kvm_memory_slot
*slot
)
880 static inline void kvm_iommu_unmap_pages(struct kvm
*kvm
,
881 struct kvm_memory_slot
*slot
)
887 * search_memslots() and __gfn_to_memslot() are here because they are
888 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
889 * gfn_to_memslot() itself isn't here as an inline because that would
890 * bloat other code too much.
892 static inline struct kvm_memory_slot
*
893 search_memslots(struct kvm_memslots
*slots
, gfn_t gfn
)
895 int start
= 0, end
= slots
->used_slots
;
896 int slot
= atomic_read(&slots
->lru_slot
);
897 struct kvm_memory_slot
*memslots
= slots
->memslots
;
899 if (gfn
>= memslots
[slot
].base_gfn
&&
900 gfn
< memslots
[slot
].base_gfn
+ memslots
[slot
].npages
)
901 return &memslots
[slot
];
903 while (start
< end
) {
904 slot
= start
+ (end
- start
) / 2;
906 if (gfn
>= memslots
[slot
].base_gfn
)
912 if (gfn
>= memslots
[start
].base_gfn
&&
913 gfn
< memslots
[start
].base_gfn
+ memslots
[start
].npages
) {
914 atomic_set(&slots
->lru_slot
, start
);
915 return &memslots
[start
];
921 static inline struct kvm_memory_slot
*
922 __gfn_to_memslot(struct kvm_memslots
*slots
, gfn_t gfn
)
924 return search_memslots(slots
, gfn
);
927 static inline unsigned long
928 __gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
)
930 return slot
->userspace_addr
+ (gfn
- slot
->base_gfn
) * PAGE_SIZE
;
933 static inline int memslot_id(struct kvm
*kvm
, gfn_t gfn
)
935 return gfn_to_memslot(kvm
, gfn
)->id
;
939 hva_to_gfn_memslot(unsigned long hva
, struct kvm_memory_slot
*slot
)
941 gfn_t gfn_offset
= (hva
- slot
->userspace_addr
) >> PAGE_SHIFT
;
943 return slot
->base_gfn
+ gfn_offset
;
946 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
948 return (gpa_t
)gfn
<< PAGE_SHIFT
;
951 static inline gfn_t
gpa_to_gfn(gpa_t gpa
)
953 return (gfn_t
)(gpa
>> PAGE_SHIFT
);
956 static inline hpa_t
pfn_to_hpa(kvm_pfn_t pfn
)
958 return (hpa_t
)pfn
<< PAGE_SHIFT
;
961 static inline bool kvm_is_error_gpa(struct kvm
*kvm
, gpa_t gpa
)
963 unsigned long hva
= gfn_to_hva(kvm
, gpa_to_gfn(gpa
));
965 return kvm_is_error_hva(hva
);
973 struct kvm_stat_data
{
978 struct kvm_stats_debugfs_item
{
981 enum kvm_stat_kind kind
;
983 extern struct kvm_stats_debugfs_item debugfs_entries
[];
984 extern struct dentry
*kvm_debugfs_dir
;
986 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
987 static inline int mmu_notifier_retry(struct kvm
*kvm
, unsigned long mmu_seq
)
989 if (unlikely(kvm
->mmu_notifier_count
))
992 * Ensure the read of mmu_notifier_count happens before the read
993 * of mmu_notifier_seq. This interacts with the smp_wmb() in
994 * mmu_notifier_invalidate_range_end to make sure that the caller
995 * either sees the old (non-zero) value of mmu_notifier_count or
996 * the new (incremented) value of mmu_notifier_seq.
997 * PowerPC Book3s HV KVM calls this under a per-page lock
998 * rather than under kvm->mmu_lock, for scalability, so
999 * can't rely on kvm->mmu_lock to keep things ordered.
1002 if (kvm
->mmu_notifier_seq
!= mmu_seq
)
1008 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1011 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1012 #elif defined(CONFIG_ARM64)
1013 #define KVM_MAX_IRQ_ROUTES 4096
1015 #define KVM_MAX_IRQ_ROUTES 1024
1018 int kvm_set_irq_routing(struct kvm
*kvm
,
1019 const struct kvm_irq_routing_entry
*entries
,
1022 int kvm_set_routing_entry(struct kvm
*kvm
,
1023 struct kvm_kernel_irq_routing_entry
*e
,
1024 const struct kvm_irq_routing_entry
*ue
);
1025 void kvm_free_irq_routing(struct kvm
*kvm
);
1029 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
1033 int kvm_send_userspace_msi(struct kvm
*kvm
, struct kvm_msi
*msi
);
1035 #ifdef CONFIG_HAVE_KVM_EVENTFD
1037 void kvm_eventfd_init(struct kvm
*kvm
);
1038 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
1040 #ifdef CONFIG_HAVE_KVM_IRQFD
1041 int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
);
1042 void kvm_irqfd_release(struct kvm
*kvm
);
1043 void kvm_irq_routing_update(struct kvm
*);
1045 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1050 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1055 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
1057 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1062 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1064 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1065 static inline void kvm_irq_routing_update(struct kvm
*kvm
)
1069 void kvm_arch_irq_routing_update(struct kvm
*kvm
);
1071 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
1076 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1078 static inline void kvm_make_request(int req
, struct kvm_vcpu
*vcpu
)
1081 * Ensure the rest of the request is published to kvm_check_request's
1082 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1085 set_bit(req
, &vcpu
->requests
);
1088 static inline bool kvm_check_request(int req
, struct kvm_vcpu
*vcpu
)
1090 if (test_bit(req
, &vcpu
->requests
)) {
1091 clear_bit(req
, &vcpu
->requests
);
1094 * Ensure the rest of the request is visible to kvm_check_request's
1095 * caller. Paired with the smp_wmb in kvm_make_request.
1097 smp_mb__after_atomic();
1104 extern bool kvm_rebooting
;
1107 struct kvm_device_ops
*ops
;
1110 struct list_head vm_node
;
1113 /* create, destroy, and name are mandatory */
1114 struct kvm_device_ops
{
1116 int (*create
)(struct kvm_device
*dev
, u32 type
);
1119 * Destroy is responsible for freeing dev.
1121 * Destroy may be called before or after destructors are called
1122 * on emulated I/O regions, depending on whether a reference is
1123 * held by a vcpu or other kvm component that gets destroyed
1124 * after the emulated I/O.
1126 void (*destroy
)(struct kvm_device
*dev
);
1128 int (*set_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1129 int (*get_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1130 int (*has_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1131 long (*ioctl
)(struct kvm_device
*dev
, unsigned int ioctl
,
1135 void kvm_device_get(struct kvm_device
*dev
);
1136 void kvm_device_put(struct kvm_device
*dev
);
1137 struct kvm_device
*kvm_device_from_filp(struct file
*filp
);
1138 int kvm_register_device_ops(struct kvm_device_ops
*ops
, u32 type
);
1139 void kvm_unregister_device_ops(u32 type
);
1141 extern struct kvm_device_ops kvm_mpic_ops
;
1142 extern struct kvm_device_ops kvm_xics_ops
;
1143 extern struct kvm_device_ops kvm_arm_vgic_v2_ops
;
1144 extern struct kvm_device_ops kvm_arm_vgic_v3_ops
;
1146 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1148 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1150 vcpu
->spin_loop
.in_spin_loop
= val
;
1152 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1154 vcpu
->spin_loop
.dy_eligible
= val
;
1157 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1159 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1163 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1166 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1168 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1169 bool kvm_arch_has_irq_bypass(void);
1170 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer
*,
1171 struct irq_bypass_producer
*);
1172 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer
*,
1173 struct irq_bypass_producer
*);
1174 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer
*);
1175 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer
*);
1176 int kvm_arch_update_irqfd_routing(struct kvm
*kvm
, unsigned int host_irq
,
1177 uint32_t guest_irq
, bool set
);
1178 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1180 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1181 /* If we wakeup during the poll time, was it a sucessful poll? */
1182 static inline bool vcpu_valid_wakeup(struct kvm_vcpu
*vcpu
)
1184 return vcpu
->valid_wakeup
;
1188 static inline bool vcpu_valid_wakeup(struct kvm_vcpu
*vcpu
)
1192 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */