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 <linux/refcount.h>
30 #include <asm/signal.h>
32 #include <linux/kvm.h>
33 #include <linux/kvm_para.h>
35 #include <linux/kvm_types.h>
37 #include <asm/kvm_host.h>
39 #ifndef KVM_MAX_VCPU_ID
40 #define KVM_MAX_VCPU_ID KVM_MAX_VCPUS
44 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
45 * in kvm, other bits are visible for userspace which are defined in
46 * include/linux/kvm_h.
48 #define KVM_MEMSLOT_INVALID (1UL << 16)
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 void 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
;
225 int guest_fpu_loaded
, guest_xcr0_loaded
;
226 struct swait_queue_head wq
;
230 struct kvm_vcpu_stat stat
;
231 unsigned int halt_poll_ns
;
234 #ifdef CONFIG_HAS_IOMEM
236 int mmio_read_completed
;
238 int mmio_cur_fragment
;
239 int mmio_nr_fragments
;
240 struct kvm_mmio_fragment mmio_fragments
[KVM_MAX_MMIO_FRAGMENTS
];
243 #ifdef CONFIG_KVM_ASYNC_PF
246 struct list_head queue
;
247 struct list_head done
;
252 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
254 * Cpu relax intercept or pause loop exit optimization
255 * in_spin_loop: set when a vcpu does a pause loop exit
256 * or cpu relax intercepted.
257 * dy_eligible: indicates whether vcpu is eligible for directed yield.
265 struct kvm_vcpu_arch arch
;
266 struct dentry
*debugfs_dentry
;
269 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu
*vcpu
)
271 return cmpxchg(&vcpu
->mode
, IN_GUEST_MODE
, EXITING_GUEST_MODE
);
275 * Some of the bitops functions do not support too long bitmaps.
276 * This number must be determined not to exceed such limits.
278 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
280 struct kvm_memory_slot
{
282 unsigned long npages
;
283 unsigned long *dirty_bitmap
;
284 struct kvm_arch_memory_slot arch
;
285 unsigned long userspace_addr
;
290 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot
*memslot
)
292 return ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
295 struct kvm_s390_adapter_int
{
308 struct kvm_kernel_irq_routing_entry
{
311 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
312 struct kvm
*kvm
, int irq_source_id
, int level
,
326 struct kvm_s390_adapter_int adapter
;
327 struct kvm_hv_sint hv_sint
;
329 struct hlist_node link
;
332 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
333 struct kvm_irq_routing_table
{
334 int chip
[KVM_NR_IRQCHIPS
][KVM_IRQCHIP_NUM_PINS
];
337 * Array indexed by gsi. Each entry contains list of irq chips
338 * the gsi is connected to.
340 struct hlist_head map
[0];
344 #ifndef KVM_PRIVATE_MEM_SLOTS
345 #define KVM_PRIVATE_MEM_SLOTS 0
348 #ifndef KVM_MEM_SLOTS_NUM
349 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
352 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
353 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu
*vcpu
)
361 * memslots are not sorted by id anymore, please use id_to_memslot()
362 * to get the memslot by its id.
364 struct kvm_memslots
{
366 struct kvm_memory_slot memslots
[KVM_MEM_SLOTS_NUM
];
367 /* The mapping table from slot id to the index in memslots[]. */
368 short id_to_index
[KVM_MEM_SLOTS_NUM
];
375 struct mutex slots_lock
;
376 struct mm_struct
*mm
; /* userspace tied to this vm */
377 struct kvm_memslots
*memslots
[KVM_ADDRESS_SPACE_NUM
];
378 struct srcu_struct srcu
;
379 struct srcu_struct irq_srcu
;
380 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
383 * created_vcpus is protected by kvm->lock, and is incremented
384 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
385 * incremented after storing the kvm_vcpu pointer in vcpus,
386 * and is accessed atomically.
388 atomic_t online_vcpus
;
390 int last_boosted_vcpu
;
391 struct list_head vm_list
;
393 struct kvm_io_bus
*buses
[KVM_NR_BUSES
];
394 #ifdef CONFIG_HAVE_KVM_EVENTFD
397 struct list_head items
;
398 struct list_head resampler_list
;
399 struct mutex resampler_lock
;
401 struct list_head ioeventfds
;
403 struct kvm_vm_stat stat
;
404 struct kvm_arch arch
;
405 refcount_t users_count
;
406 #ifdef CONFIG_KVM_MMIO
407 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
408 spinlock_t ring_lock
;
409 struct list_head coalesced_zones
;
412 struct mutex irq_lock
;
413 #ifdef CONFIG_HAVE_KVM_IRQCHIP
415 * Update side is protected by irq_lock.
417 struct kvm_irq_routing_table __rcu
*irq_routing
;
419 #ifdef CONFIG_HAVE_KVM_IRQFD
420 struct hlist_head irq_ack_notifier_list
;
423 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
424 struct mmu_notifier mmu_notifier
;
425 unsigned long mmu_notifier_seq
;
426 long mmu_notifier_count
;
429 struct list_head devices
;
430 struct dentry
*debugfs_dentry
;
431 struct kvm_stat_data
**debugfs_stat_data
;
434 #define kvm_err(fmt, ...) \
435 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
436 #define kvm_info(fmt, ...) \
437 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
438 #define kvm_debug(fmt, ...) \
439 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
440 #define kvm_debug_ratelimited(fmt, ...) \
441 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
443 #define kvm_pr_unimpl(fmt, ...) \
444 pr_err_ratelimited("kvm [%i]: " fmt, \
445 task_tgid_nr(current), ## __VA_ARGS__)
447 /* The guest did something we don't support. */
448 #define vcpu_unimpl(vcpu, fmt, ...) \
449 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
450 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
452 #define vcpu_debug(vcpu, fmt, ...) \
453 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
454 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
455 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
457 #define vcpu_err(vcpu, fmt, ...) \
458 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
460 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
462 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
463 * the caller has read kvm->online_vcpus before (as is the case
464 * for kvm_for_each_vcpu, for example).
467 return kvm
->vcpus
[i
];
470 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
472 idx < atomic_read(&kvm->online_vcpus) && \
473 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
476 static inline struct kvm_vcpu
*kvm_get_vcpu_by_id(struct kvm
*kvm
, int id
)
478 struct kvm_vcpu
*vcpu
= NULL
;
483 if (id
< KVM_MAX_VCPUS
)
484 vcpu
= kvm_get_vcpu(kvm
, id
);
485 if (vcpu
&& vcpu
->vcpu_id
== id
)
487 kvm_for_each_vcpu(i
, vcpu
, kvm
)
488 if (vcpu
->vcpu_id
== id
)
493 #define kvm_for_each_memslot(memslot, slots) \
494 for (memslot = &slots->memslots[0]; \
495 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
498 int kvm_vcpu_init(struct kvm_vcpu
*vcpu
, struct kvm
*kvm
, unsigned id
);
499 void kvm_vcpu_uninit(struct kvm_vcpu
*vcpu
);
501 int __must_check
vcpu_load(struct kvm_vcpu
*vcpu
);
502 void vcpu_put(struct kvm_vcpu
*vcpu
);
504 #ifdef __KVM_HAVE_IOAPIC
505 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm
*kvm
);
506 void kvm_arch_post_irq_routing_update(struct kvm
*kvm
);
508 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm
*kvm
)
511 static inline void kvm_arch_post_irq_routing_update(struct kvm
*kvm
)
516 #ifdef CONFIG_HAVE_KVM_IRQFD
517 int kvm_irqfd_init(void);
518 void kvm_irqfd_exit(void);
520 static inline int kvm_irqfd_init(void)
525 static inline void kvm_irqfd_exit(void)
529 int kvm_init(void *opaque
, unsigned vcpu_size
, unsigned vcpu_align
,
530 struct module
*module
);
533 void kvm_get_kvm(struct kvm
*kvm
);
534 void kvm_put_kvm(struct kvm
*kvm
);
536 static inline struct kvm_memslots
*__kvm_memslots(struct kvm
*kvm
, int as_id
)
538 return rcu_dereference_check(kvm
->memslots
[as_id
],
539 srcu_read_lock_held(&kvm
->srcu
)
540 || lockdep_is_held(&kvm
->slots_lock
));
543 static inline struct kvm_memslots
*kvm_memslots(struct kvm
*kvm
)
545 return __kvm_memslots(kvm
, 0);
548 static inline struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
)
550 int as_id
= kvm_arch_vcpu_memslots_id(vcpu
);
552 return __kvm_memslots(vcpu
->kvm
, as_id
);
555 static inline struct kvm_memory_slot
*
556 id_to_memslot(struct kvm_memslots
*slots
, int id
)
558 int index
= slots
->id_to_index
[id
];
559 struct kvm_memory_slot
*slot
;
561 slot
= &slots
->memslots
[index
];
563 WARN_ON(slot
->id
!= id
);
568 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
569 * - create a new memory slot
570 * - delete an existing memory slot
571 * - modify an existing memory slot
572 * -- move it in the guest physical memory space
573 * -- just change its flags
575 * Since flags can be changed by some of these operations, the following
576 * differentiation is the best we can do for __kvm_set_memory_region():
585 int kvm_set_memory_region(struct kvm
*kvm
,
586 const struct kvm_userspace_memory_region
*mem
);
587 int __kvm_set_memory_region(struct kvm
*kvm
,
588 const struct kvm_userspace_memory_region
*mem
);
589 void kvm_arch_free_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*free
,
590 struct kvm_memory_slot
*dont
);
591 int kvm_arch_create_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*slot
,
592 unsigned long npages
);
593 void kvm_arch_memslots_updated(struct kvm
*kvm
, struct kvm_memslots
*slots
);
594 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
595 struct kvm_memory_slot
*memslot
,
596 const struct kvm_userspace_memory_region
*mem
,
597 enum kvm_mr_change change
);
598 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
599 const struct kvm_userspace_memory_region
*mem
,
600 const struct kvm_memory_slot
*old
,
601 const struct kvm_memory_slot
*new,
602 enum kvm_mr_change change
);
603 bool kvm_largepages_enabled(void);
604 void kvm_disable_largepages(void);
605 /* flush all memory translations */
606 void kvm_arch_flush_shadow_all(struct kvm
*kvm
);
607 /* flush memory translations pointing to 'slot' */
608 void kvm_arch_flush_shadow_memslot(struct kvm
*kvm
,
609 struct kvm_memory_slot
*slot
);
611 int gfn_to_page_many_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
,
612 struct page
**pages
, int nr_pages
);
614 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
615 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
616 unsigned long gfn_to_hva_prot(struct kvm
*kvm
, gfn_t gfn
, bool *writable
);
617 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
618 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
620 void kvm_release_page_clean(struct page
*page
);
621 void kvm_release_page_dirty(struct page
*page
);
622 void kvm_set_page_accessed(struct page
*page
);
624 kvm_pfn_t
gfn_to_pfn_atomic(struct kvm
*kvm
, gfn_t gfn
);
625 kvm_pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
626 kvm_pfn_t
gfn_to_pfn_prot(struct kvm
*kvm
, gfn_t gfn
, bool write_fault
,
628 kvm_pfn_t
gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
629 kvm_pfn_t
gfn_to_pfn_memslot_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
);
630 kvm_pfn_t
__gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
631 bool atomic
, bool *async
, bool write_fault
,
634 void kvm_release_pfn_clean(kvm_pfn_t pfn
);
635 void kvm_set_pfn_dirty(kvm_pfn_t pfn
);
636 void kvm_set_pfn_accessed(kvm_pfn_t pfn
);
637 void kvm_get_pfn(kvm_pfn_t pfn
);
639 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
641 int kvm_read_guest_atomic(struct kvm
*kvm
, gpa_t gpa
, void *data
,
643 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
644 int kvm_vcpu_read_guest_cached(struct kvm_vcpu
*vcpu
, struct gfn_to_hva_cache
*ghc
,
645 void *data
, unsigned long len
);
646 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
647 int offset
, int len
);
648 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
650 int kvm_vcpu_write_guest_cached(struct kvm_vcpu
*v
, struct gfn_to_hva_cache
*ghc
,
651 void *data
, unsigned long len
);
652 int kvm_vcpu_write_guest_offset_cached(struct kvm_vcpu
*v
, struct gfn_to_hva_cache
*ghc
,
653 void *data
, int offset
, unsigned long len
);
654 int kvm_vcpu_gfn_to_hva_cache_init(struct kvm_vcpu
*v
, struct gfn_to_hva_cache
*ghc
,
655 gpa_t gpa
, unsigned long len
);
656 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
657 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
658 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
659 bool kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
660 unsigned long kvm_host_page_size(struct kvm
*kvm
, gfn_t gfn
);
661 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
663 struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
);
664 struct kvm_memory_slot
*kvm_vcpu_gfn_to_memslot(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
665 kvm_pfn_t
kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
666 kvm_pfn_t
kvm_vcpu_gfn_to_pfn(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
667 struct page
*kvm_vcpu_gfn_to_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
668 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
669 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu
*vcpu
, gfn_t gfn
, bool *writable
);
670 int kvm_vcpu_read_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, void *data
, int offset
,
672 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
674 int kvm_vcpu_read_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
676 int kvm_vcpu_write_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, const void *data
,
677 int offset
, int len
);
678 int kvm_vcpu_write_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, const void *data
,
680 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
682 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
683 void kvm_arch_vcpu_blocking(struct kvm_vcpu
*vcpu
);
684 void kvm_arch_vcpu_unblocking(struct kvm_vcpu
*vcpu
);
685 void kvm_vcpu_wake_up(struct kvm_vcpu
*vcpu
);
686 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
687 int kvm_vcpu_yield_to(struct kvm_vcpu
*target
);
688 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
);
689 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
690 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
692 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
693 void kvm_reload_remote_mmus(struct kvm
*kvm
);
694 bool kvm_make_all_cpus_request(struct kvm
*kvm
, unsigned int req
);
696 long kvm_arch_dev_ioctl(struct file
*filp
,
697 unsigned int ioctl
, unsigned long arg
);
698 long kvm_arch_vcpu_ioctl(struct file
*filp
,
699 unsigned int ioctl
, unsigned long arg
);
700 int kvm_arch_vcpu_fault(struct kvm_vcpu
*vcpu
, struct vm_fault
*vmf
);
702 int kvm_vm_ioctl_check_extension(struct kvm
*kvm
, long ext
);
704 int kvm_get_dirty_log(struct kvm
*kvm
,
705 struct kvm_dirty_log
*log
, int *is_dirty
);
707 int kvm_get_dirty_log_protect(struct kvm
*kvm
,
708 struct kvm_dirty_log
*log
, bool *is_dirty
);
710 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm
*kvm
,
711 struct kvm_memory_slot
*slot
,
715 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
716 struct kvm_dirty_log
*log
);
718 int kvm_vm_ioctl_irq_line(struct kvm
*kvm
, struct kvm_irq_level
*irq_level
,
720 long kvm_arch_vm_ioctl(struct file
*filp
,
721 unsigned int ioctl
, unsigned long arg
);
723 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
724 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
726 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
727 struct kvm_translation
*tr
);
729 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
730 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
731 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
732 struct kvm_sregs
*sregs
);
733 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
734 struct kvm_sregs
*sregs
);
735 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
736 struct kvm_mp_state
*mp_state
);
737 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
738 struct kvm_mp_state
*mp_state
);
739 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
740 struct kvm_guest_debug
*dbg
);
741 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
743 int kvm_arch_init(void *opaque
);
744 void kvm_arch_exit(void);
746 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
);
747 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
);
749 void kvm_arch_sched_in(struct kvm_vcpu
*vcpu
, int cpu
);
751 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
);
752 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
753 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
754 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
);
755 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
);
756 void kvm_arch_vcpu_postcreate(struct kvm_vcpu
*vcpu
);
757 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
759 bool kvm_arch_has_vcpu_debugfs(void);
760 int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu
*vcpu
);
762 int kvm_arch_hardware_enable(void);
763 void kvm_arch_hardware_disable(void);
764 int kvm_arch_hardware_setup(void);
765 void kvm_arch_hardware_unsetup(void);
766 void kvm_arch_check_processor_compat(void *rtn
);
767 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
768 int kvm_arch_vcpu_should_kick(struct kvm_vcpu
*vcpu
);
770 void *kvm_kvzalloc(unsigned long size
);
772 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
773 static inline struct kvm
*kvm_arch_alloc_vm(void)
775 return kzalloc(sizeof(struct kvm
), GFP_KERNEL
);
778 static inline void kvm_arch_free_vm(struct kvm
*kvm
)
784 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
785 void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
);
786 void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
);
787 bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
);
789 static inline void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
)
793 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
)
797 static inline bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
)
802 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
803 void kvm_arch_start_assignment(struct kvm
*kvm
);
804 void kvm_arch_end_assignment(struct kvm
*kvm
);
805 bool kvm_arch_has_assigned_device(struct kvm
*kvm
);
807 static inline void kvm_arch_start_assignment(struct kvm
*kvm
)
811 static inline void kvm_arch_end_assignment(struct kvm
*kvm
)
815 static inline bool kvm_arch_has_assigned_device(struct kvm
*kvm
)
821 static inline struct swait_queue_head
*kvm_arch_vcpu_wq(struct kvm_vcpu
*vcpu
)
823 #ifdef __KVM_HAVE_ARCH_WQP
824 return vcpu
->arch
.wqp
;
830 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
832 * returns true if the virtual interrupt controller is initialized and
833 * ready to accept virtual IRQ. On some architectures the virtual interrupt
834 * controller is dynamically instantiated and this is not always true.
836 bool kvm_arch_intc_initialized(struct kvm
*kvm
);
838 static inline bool kvm_arch_intc_initialized(struct kvm
*kvm
)
844 int kvm_arch_init_vm(struct kvm
*kvm
, unsigned long type
);
845 void kvm_arch_destroy_vm(struct kvm
*kvm
);
846 void kvm_arch_sync_events(struct kvm
*kvm
);
848 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
849 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
851 bool kvm_is_reserved_pfn(kvm_pfn_t pfn
);
853 struct kvm_irq_ack_notifier
{
854 struct hlist_node link
;
856 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
859 int kvm_irq_map_gsi(struct kvm
*kvm
,
860 struct kvm_kernel_irq_routing_entry
*entries
, int gsi
);
861 int kvm_irq_map_chip_pin(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
863 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
,
865 int kvm_set_msi(struct kvm_kernel_irq_routing_entry
*irq_entry
, struct kvm
*kvm
,
866 int irq_source_id
, int level
, bool line_status
);
867 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry
*e
,
868 struct kvm
*kvm
, int irq_source_id
,
869 int level
, bool line_status
);
870 bool kvm_irq_has_notifier(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
871 void kvm_notify_acked_gsi(struct kvm
*kvm
, int gsi
);
872 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
873 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
874 struct kvm_irq_ack_notifier
*kian
);
875 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
876 struct kvm_irq_ack_notifier
*kian
);
877 int kvm_request_irq_source_id(struct kvm
*kvm
);
878 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
881 * search_memslots() and __gfn_to_memslot() are here because they are
882 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
883 * gfn_to_memslot() itself isn't here as an inline because that would
884 * bloat other code too much.
886 static inline struct kvm_memory_slot
*
887 search_memslots(struct kvm_memslots
*slots
, gfn_t gfn
)
889 int start
= 0, end
= slots
->used_slots
;
890 int slot
= atomic_read(&slots
->lru_slot
);
891 struct kvm_memory_slot
*memslots
= slots
->memslots
;
893 if (gfn
>= memslots
[slot
].base_gfn
&&
894 gfn
< memslots
[slot
].base_gfn
+ memslots
[slot
].npages
)
895 return &memslots
[slot
];
897 while (start
< end
) {
898 slot
= start
+ (end
- start
) / 2;
900 if (gfn
>= memslots
[slot
].base_gfn
)
906 if (gfn
>= memslots
[start
].base_gfn
&&
907 gfn
< memslots
[start
].base_gfn
+ memslots
[start
].npages
) {
908 atomic_set(&slots
->lru_slot
, start
);
909 return &memslots
[start
];
915 static inline struct kvm_memory_slot
*
916 __gfn_to_memslot(struct kvm_memslots
*slots
, gfn_t gfn
)
918 return search_memslots(slots
, gfn
);
921 static inline unsigned long
922 __gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
)
924 return slot
->userspace_addr
+ (gfn
- slot
->base_gfn
) * PAGE_SIZE
;
927 static inline int memslot_id(struct kvm
*kvm
, gfn_t gfn
)
929 return gfn_to_memslot(kvm
, gfn
)->id
;
933 hva_to_gfn_memslot(unsigned long hva
, struct kvm_memory_slot
*slot
)
935 gfn_t gfn_offset
= (hva
- slot
->userspace_addr
) >> PAGE_SHIFT
;
937 return slot
->base_gfn
+ gfn_offset
;
940 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
942 return (gpa_t
)gfn
<< PAGE_SHIFT
;
945 static inline gfn_t
gpa_to_gfn(gpa_t gpa
)
947 return (gfn_t
)(gpa
>> PAGE_SHIFT
);
950 static inline hpa_t
pfn_to_hpa(kvm_pfn_t pfn
)
952 return (hpa_t
)pfn
<< PAGE_SHIFT
;
955 static inline bool kvm_is_error_gpa(struct kvm
*kvm
, gpa_t gpa
)
957 unsigned long hva
= gfn_to_hva(kvm
, gpa_to_gfn(gpa
));
959 return kvm_is_error_hva(hva
);
967 struct kvm_stat_data
{
972 struct kvm_stats_debugfs_item
{
975 enum kvm_stat_kind kind
;
977 extern struct kvm_stats_debugfs_item debugfs_entries
[];
978 extern struct dentry
*kvm_debugfs_dir
;
980 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
981 static inline int mmu_notifier_retry(struct kvm
*kvm
, unsigned long mmu_seq
)
983 if (unlikely(kvm
->mmu_notifier_count
))
986 * Ensure the read of mmu_notifier_count happens before the read
987 * of mmu_notifier_seq. This interacts with the smp_wmb() in
988 * mmu_notifier_invalidate_range_end to make sure that the caller
989 * either sees the old (non-zero) value of mmu_notifier_count or
990 * the new (incremented) value of mmu_notifier_seq.
991 * PowerPC Book3s HV KVM calls this under a per-page lock
992 * rather than under kvm->mmu_lock, for scalability, so
993 * can't rely on kvm->mmu_lock to keep things ordered.
996 if (kvm
->mmu_notifier_seq
!= mmu_seq
)
1002 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1005 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1006 #elif defined(CONFIG_ARM64)
1007 #define KVM_MAX_IRQ_ROUTES 4096
1009 #define KVM_MAX_IRQ_ROUTES 1024
1012 int kvm_set_irq_routing(struct kvm
*kvm
,
1013 const struct kvm_irq_routing_entry
*entries
,
1016 int kvm_set_routing_entry(struct kvm
*kvm
,
1017 struct kvm_kernel_irq_routing_entry
*e
,
1018 const struct kvm_irq_routing_entry
*ue
);
1019 void kvm_free_irq_routing(struct kvm
*kvm
);
1023 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
1027 int kvm_send_userspace_msi(struct kvm
*kvm
, struct kvm_msi
*msi
);
1029 #ifdef CONFIG_HAVE_KVM_EVENTFD
1031 void kvm_eventfd_init(struct kvm
*kvm
);
1032 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
1034 #ifdef CONFIG_HAVE_KVM_IRQFD
1035 int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
);
1036 void kvm_irqfd_release(struct kvm
*kvm
);
1037 void kvm_irq_routing_update(struct kvm
*);
1039 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1044 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1049 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
1051 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1056 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1058 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1059 static inline void kvm_irq_routing_update(struct kvm
*kvm
)
1063 void kvm_arch_irq_routing_update(struct kvm
*kvm
);
1065 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
1070 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1072 static inline void kvm_make_request(int req
, struct kvm_vcpu
*vcpu
)
1075 * Ensure the rest of the request is published to kvm_check_request's
1076 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1079 set_bit(req
, &vcpu
->requests
);
1082 static inline bool kvm_check_request(int req
, struct kvm_vcpu
*vcpu
)
1084 if (test_bit(req
, &vcpu
->requests
)) {
1085 clear_bit(req
, &vcpu
->requests
);
1088 * Ensure the rest of the request is visible to kvm_check_request's
1089 * caller. Paired with the smp_wmb in kvm_make_request.
1091 smp_mb__after_atomic();
1098 extern bool kvm_rebooting
;
1100 extern unsigned int halt_poll_ns
;
1101 extern unsigned int halt_poll_ns_grow
;
1102 extern unsigned int halt_poll_ns_shrink
;
1105 struct kvm_device_ops
*ops
;
1108 struct list_head vm_node
;
1111 /* create, destroy, and name are mandatory */
1112 struct kvm_device_ops
{
1116 * create is called holding kvm->lock and any operations not suitable
1117 * to do while holding the lock should be deferred to init (see
1120 int (*create
)(struct kvm_device
*dev
, u32 type
);
1123 * init is called after create if create is successful and is called
1124 * outside of holding kvm->lock.
1126 void (*init
)(struct kvm_device
*dev
);
1129 * Destroy is responsible for freeing dev.
1131 * Destroy may be called before or after destructors are called
1132 * on emulated I/O regions, depending on whether a reference is
1133 * held by a vcpu or other kvm component that gets destroyed
1134 * after the emulated I/O.
1136 void (*destroy
)(struct kvm_device
*dev
);
1138 int (*set_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1139 int (*get_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1140 int (*has_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1141 long (*ioctl
)(struct kvm_device
*dev
, unsigned int ioctl
,
1145 void kvm_device_get(struct kvm_device
*dev
);
1146 void kvm_device_put(struct kvm_device
*dev
);
1147 struct kvm_device
*kvm_device_from_filp(struct file
*filp
);
1148 int kvm_register_device_ops(struct kvm_device_ops
*ops
, u32 type
);
1149 void kvm_unregister_device_ops(u32 type
);
1151 extern struct kvm_device_ops kvm_mpic_ops
;
1152 extern struct kvm_device_ops kvm_xics_ops
;
1153 extern struct kvm_device_ops kvm_arm_vgic_v2_ops
;
1154 extern struct kvm_device_ops kvm_arm_vgic_v3_ops
;
1156 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1158 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1160 vcpu
->spin_loop
.in_spin_loop
= val
;
1162 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1164 vcpu
->spin_loop
.dy_eligible
= val
;
1167 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1169 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1173 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1176 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1178 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1179 bool kvm_arch_has_irq_bypass(void);
1180 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer
*,
1181 struct irq_bypass_producer
*);
1182 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer
*,
1183 struct irq_bypass_producer
*);
1184 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer
*);
1185 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer
*);
1186 int kvm_arch_update_irqfd_routing(struct kvm
*kvm
, unsigned int host_irq
,
1187 uint32_t guest_irq
, bool set
);
1188 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1190 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1191 /* If we wakeup during the poll time, was it a sucessful poll? */
1192 static inline bool vcpu_valid_wakeup(struct kvm_vcpu
*vcpu
)
1194 return vcpu
->valid_wakeup
;
1198 static inline bool vcpu_valid_wakeup(struct kvm_vcpu
*vcpu
)
1202 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */