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
)
118 #define KVM_REQUEST_MASK GENMASK(7,0)
119 #define KVM_REQUEST_NO_WAKEUP BIT(8)
120 #define KVM_REQUEST_WAIT BIT(9)
122 * Architecture-independent vcpu->requests bit members
123 * Bits 4-7 are reserved for more arch-independent bits.
125 #define KVM_REQ_TLB_FLUSH (0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
126 #define KVM_REQ_MMU_RELOAD (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
127 #define KVM_REQ_PENDING_TIMER 2
128 #define KVM_REQ_UNHALT 3
129 #define KVM_REQUEST_ARCH_BASE 8
131 #define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
132 BUILD_BUG_ON((unsigned)(nr) >= 32 - KVM_REQUEST_ARCH_BASE); \
133 (unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) | (flags)); \
135 #define KVM_ARCH_REQ(nr) KVM_ARCH_REQ_FLAGS(nr, 0)
137 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
138 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
140 extern struct kmem_cache
*kvm_vcpu_cache
;
142 extern spinlock_t kvm_lock
;
143 extern struct list_head vm_list
;
145 struct kvm_io_range
{
148 struct kvm_io_device
*dev
;
151 #define NR_IOBUS_DEVS 1000
156 struct kvm_io_range range
[];
162 KVM_VIRTIO_CCW_NOTIFY_BUS
,
167 int kvm_io_bus_write(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
168 int len
, const void *val
);
169 int kvm_io_bus_write_cookie(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
,
170 gpa_t addr
, int len
, const void *val
, long cookie
);
171 int kvm_io_bus_read(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
173 int kvm_io_bus_register_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
,
174 int len
, struct kvm_io_device
*dev
);
175 void kvm_io_bus_unregister_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
176 struct kvm_io_device
*dev
);
177 struct kvm_io_device
*kvm_io_bus_get_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
180 #ifdef CONFIG_KVM_ASYNC_PF
181 struct kvm_async_pf
{
182 struct work_struct work
;
183 struct list_head link
;
184 struct list_head queue
;
185 struct kvm_vcpu
*vcpu
;
186 struct mm_struct
*mm
;
189 struct kvm_arch_async_pf arch
;
193 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu
*vcpu
);
194 void kvm_check_async_pf_completion(struct kvm_vcpu
*vcpu
);
195 int kvm_setup_async_pf(struct kvm_vcpu
*vcpu
, gva_t gva
, unsigned long hva
,
196 struct kvm_arch_async_pf
*arch
);
197 int kvm_async_pf_wakeup_all(struct kvm_vcpu
*vcpu
);
204 READING_SHADOW_PAGE_TABLES
,
208 * Sometimes a large or cross-page mmio needs to be broken up into separate
209 * exits for userspace servicing.
211 struct kvm_mmio_fragment
{
219 #ifdef CONFIG_PREEMPT_NOTIFIERS
220 struct preempt_notifier preempt_notifier
;
226 unsigned long requests
;
227 unsigned long guest_debug
;
230 struct list_head blocked_vcpu_list
;
235 int guest_fpu_loaded
, guest_xcr0_loaded
;
236 struct swait_queue_head wq
;
240 struct kvm_vcpu_stat stat
;
241 unsigned int halt_poll_ns
;
244 #ifdef CONFIG_HAS_IOMEM
246 int mmio_read_completed
;
248 int mmio_cur_fragment
;
249 int mmio_nr_fragments
;
250 struct kvm_mmio_fragment mmio_fragments
[KVM_MAX_MMIO_FRAGMENTS
];
253 #ifdef CONFIG_KVM_ASYNC_PF
256 struct list_head queue
;
257 struct list_head done
;
262 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
264 * Cpu relax intercept or pause loop exit optimization
265 * in_spin_loop: set when a vcpu does a pause loop exit
266 * or cpu relax intercepted.
267 * dy_eligible: indicates whether vcpu is eligible for directed yield.
275 struct kvm_vcpu_arch arch
;
276 struct dentry
*debugfs_dentry
;
279 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu
*vcpu
)
282 * The memory barrier ensures a previous write to vcpu->requests cannot
283 * be reordered with the read of vcpu->mode. It pairs with the general
284 * memory barrier following the write of vcpu->mode in VCPU RUN.
286 smp_mb__before_atomic();
287 return cmpxchg(&vcpu
->mode
, IN_GUEST_MODE
, EXITING_GUEST_MODE
);
291 * Some of the bitops functions do not support too long bitmaps.
292 * This number must be determined not to exceed such limits.
294 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
296 struct kvm_memory_slot
{
298 unsigned long npages
;
299 unsigned long *dirty_bitmap
;
300 struct kvm_arch_memory_slot arch
;
301 unsigned long userspace_addr
;
306 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot
*memslot
)
308 return ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
311 struct kvm_s390_adapter_int
{
324 struct kvm_kernel_irq_routing_entry
{
327 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
328 struct kvm
*kvm
, int irq_source_id
, int level
,
342 struct kvm_s390_adapter_int adapter
;
343 struct kvm_hv_sint hv_sint
;
345 struct hlist_node link
;
348 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
349 struct kvm_irq_routing_table
{
350 int chip
[KVM_NR_IRQCHIPS
][KVM_IRQCHIP_NUM_PINS
];
353 * Array indexed by gsi. Each entry contains list of irq chips
354 * the gsi is connected to.
356 struct hlist_head map
[0];
360 #ifndef KVM_PRIVATE_MEM_SLOTS
361 #define KVM_PRIVATE_MEM_SLOTS 0
364 #ifndef KVM_MEM_SLOTS_NUM
365 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
368 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
369 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu
*vcpu
)
377 * memslots are not sorted by id anymore, please use id_to_memslot()
378 * to get the memslot by its id.
380 struct kvm_memslots
{
382 struct kvm_memory_slot memslots
[KVM_MEM_SLOTS_NUM
];
383 /* The mapping table from slot id to the index in memslots[]. */
384 short id_to_index
[KVM_MEM_SLOTS_NUM
];
391 struct mutex slots_lock
;
392 struct mm_struct
*mm
; /* userspace tied to this vm */
393 struct kvm_memslots
*memslots
[KVM_ADDRESS_SPACE_NUM
];
394 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
397 * created_vcpus is protected by kvm->lock, and is incremented
398 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
399 * incremented after storing the kvm_vcpu pointer in vcpus,
400 * and is accessed atomically.
402 atomic_t online_vcpus
;
404 int last_boosted_vcpu
;
405 struct list_head vm_list
;
407 struct kvm_io_bus
*buses
[KVM_NR_BUSES
];
408 #ifdef CONFIG_HAVE_KVM_EVENTFD
411 struct list_head items
;
412 struct list_head resampler_list
;
413 struct mutex resampler_lock
;
415 struct list_head ioeventfds
;
417 struct kvm_vm_stat stat
;
418 struct kvm_arch arch
;
419 refcount_t users_count
;
420 #ifdef CONFIG_KVM_MMIO
421 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
422 spinlock_t ring_lock
;
423 struct list_head coalesced_zones
;
426 struct mutex irq_lock
;
427 #ifdef CONFIG_HAVE_KVM_IRQCHIP
429 * Update side is protected by irq_lock.
431 struct kvm_irq_routing_table __rcu
*irq_routing
;
433 #ifdef CONFIG_HAVE_KVM_IRQFD
434 struct hlist_head irq_ack_notifier_list
;
437 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
438 struct mmu_notifier mmu_notifier
;
439 unsigned long mmu_notifier_seq
;
440 long mmu_notifier_count
;
443 struct list_head devices
;
444 struct dentry
*debugfs_dentry
;
445 struct kvm_stat_data
**debugfs_stat_data
;
446 struct srcu_struct srcu
;
447 struct srcu_struct irq_srcu
;
450 #define kvm_err(fmt, ...) \
451 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
452 #define kvm_info(fmt, ...) \
453 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
454 #define kvm_debug(fmt, ...) \
455 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
456 #define kvm_debug_ratelimited(fmt, ...) \
457 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
459 #define kvm_pr_unimpl(fmt, ...) \
460 pr_err_ratelimited("kvm [%i]: " fmt, \
461 task_tgid_nr(current), ## __VA_ARGS__)
463 /* The guest did something we don't support. */
464 #define vcpu_unimpl(vcpu, fmt, ...) \
465 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
466 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
468 #define vcpu_debug(vcpu, fmt, ...) \
469 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
470 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
471 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
473 #define vcpu_err(vcpu, fmt, ...) \
474 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
476 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
478 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
479 * the caller has read kvm->online_vcpus before (as is the case
480 * for kvm_for_each_vcpu, for example).
483 return kvm
->vcpus
[i
];
486 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
488 idx < atomic_read(&kvm->online_vcpus) && \
489 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
492 static inline struct kvm_vcpu
*kvm_get_vcpu_by_id(struct kvm
*kvm
, int id
)
494 struct kvm_vcpu
*vcpu
= NULL
;
499 if (id
< KVM_MAX_VCPUS
)
500 vcpu
= kvm_get_vcpu(kvm
, id
);
501 if (vcpu
&& vcpu
->vcpu_id
== id
)
503 kvm_for_each_vcpu(i
, vcpu
, kvm
)
504 if (vcpu
->vcpu_id
== id
)
509 static inline int kvm_vcpu_get_idx(struct kvm_vcpu
*vcpu
)
511 struct kvm_vcpu
*tmp
;
514 kvm_for_each_vcpu(idx
, tmp
, vcpu
->kvm
)
520 #define kvm_for_each_memslot(memslot, slots) \
521 for (memslot = &slots->memslots[0]; \
522 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
525 int kvm_vcpu_init(struct kvm_vcpu
*vcpu
, struct kvm
*kvm
, unsigned id
);
526 void kvm_vcpu_uninit(struct kvm_vcpu
*vcpu
);
528 int __must_check
vcpu_load(struct kvm_vcpu
*vcpu
);
529 void vcpu_put(struct kvm_vcpu
*vcpu
);
531 #ifdef __KVM_HAVE_IOAPIC
532 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm
*kvm
);
533 void kvm_arch_post_irq_routing_update(struct kvm
*kvm
);
535 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm
*kvm
)
538 static inline void kvm_arch_post_irq_routing_update(struct kvm
*kvm
)
543 #ifdef CONFIG_HAVE_KVM_IRQFD
544 int kvm_irqfd_init(void);
545 void kvm_irqfd_exit(void);
547 static inline int kvm_irqfd_init(void)
552 static inline void kvm_irqfd_exit(void)
556 int kvm_init(void *opaque
, unsigned vcpu_size
, unsigned vcpu_align
,
557 struct module
*module
);
560 void kvm_get_kvm(struct kvm
*kvm
);
561 void kvm_put_kvm(struct kvm
*kvm
);
563 static inline struct kvm_memslots
*__kvm_memslots(struct kvm
*kvm
, int as_id
)
565 return rcu_dereference_check(kvm
->memslots
[as_id
],
566 srcu_read_lock_held(&kvm
->srcu
)
567 || lockdep_is_held(&kvm
->slots_lock
));
570 static inline struct kvm_memslots
*kvm_memslots(struct kvm
*kvm
)
572 return __kvm_memslots(kvm
, 0);
575 static inline struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
)
577 int as_id
= kvm_arch_vcpu_memslots_id(vcpu
);
579 return __kvm_memslots(vcpu
->kvm
, as_id
);
582 static inline struct kvm_memory_slot
*
583 id_to_memslot(struct kvm_memslots
*slots
, int id
)
585 int index
= slots
->id_to_index
[id
];
586 struct kvm_memory_slot
*slot
;
588 slot
= &slots
->memslots
[index
];
590 WARN_ON(slot
->id
!= id
);
595 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
596 * - create a new memory slot
597 * - delete an existing memory slot
598 * - modify an existing memory slot
599 * -- move it in the guest physical memory space
600 * -- just change its flags
602 * Since flags can be changed by some of these operations, the following
603 * differentiation is the best we can do for __kvm_set_memory_region():
612 int kvm_set_memory_region(struct kvm
*kvm
,
613 const struct kvm_userspace_memory_region
*mem
);
614 int __kvm_set_memory_region(struct kvm
*kvm
,
615 const struct kvm_userspace_memory_region
*mem
);
616 void kvm_arch_free_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*free
,
617 struct kvm_memory_slot
*dont
);
618 int kvm_arch_create_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*slot
,
619 unsigned long npages
);
620 void kvm_arch_memslots_updated(struct kvm
*kvm
, struct kvm_memslots
*slots
);
621 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
622 struct kvm_memory_slot
*memslot
,
623 const struct kvm_userspace_memory_region
*mem
,
624 enum kvm_mr_change change
);
625 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
626 const struct kvm_userspace_memory_region
*mem
,
627 const struct kvm_memory_slot
*old
,
628 const struct kvm_memory_slot
*new,
629 enum kvm_mr_change change
);
630 bool kvm_largepages_enabled(void);
631 void kvm_disable_largepages(void);
632 /* flush all memory translations */
633 void kvm_arch_flush_shadow_all(struct kvm
*kvm
);
634 /* flush memory translations pointing to 'slot' */
635 void kvm_arch_flush_shadow_memslot(struct kvm
*kvm
,
636 struct kvm_memory_slot
*slot
);
638 int gfn_to_page_many_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
,
639 struct page
**pages
, int nr_pages
);
641 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
642 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
643 unsigned long gfn_to_hva_prot(struct kvm
*kvm
, gfn_t gfn
, bool *writable
);
644 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
645 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
647 void kvm_release_page_clean(struct page
*page
);
648 void kvm_release_page_dirty(struct page
*page
);
649 void kvm_set_page_accessed(struct page
*page
);
651 kvm_pfn_t
gfn_to_pfn_atomic(struct kvm
*kvm
, gfn_t gfn
);
652 kvm_pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
653 kvm_pfn_t
gfn_to_pfn_prot(struct kvm
*kvm
, gfn_t gfn
, bool write_fault
,
655 kvm_pfn_t
gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
656 kvm_pfn_t
gfn_to_pfn_memslot_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
);
657 kvm_pfn_t
__gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
658 bool atomic
, bool *async
, bool write_fault
,
661 void kvm_release_pfn_clean(kvm_pfn_t pfn
);
662 void kvm_set_pfn_dirty(kvm_pfn_t pfn
);
663 void kvm_set_pfn_accessed(kvm_pfn_t pfn
);
664 void kvm_get_pfn(kvm_pfn_t pfn
);
666 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
668 int kvm_read_guest_atomic(struct kvm
*kvm
, gpa_t gpa
, void *data
,
670 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
671 int kvm_read_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
672 void *data
, unsigned long len
);
673 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
674 int offset
, int len
);
675 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
677 int kvm_write_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
678 void *data
, unsigned long len
);
679 int kvm_write_guest_offset_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
680 void *data
, int offset
, unsigned long len
);
681 int kvm_gfn_to_hva_cache_init(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
682 gpa_t gpa
, unsigned long len
);
683 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
684 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
685 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
686 bool kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
687 unsigned long kvm_host_page_size(struct kvm
*kvm
, gfn_t gfn
);
688 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
690 struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
);
691 struct kvm_memory_slot
*kvm_vcpu_gfn_to_memslot(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
692 kvm_pfn_t
kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
693 kvm_pfn_t
kvm_vcpu_gfn_to_pfn(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
694 struct page
*kvm_vcpu_gfn_to_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
695 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
696 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu
*vcpu
, gfn_t gfn
, bool *writable
);
697 int kvm_vcpu_read_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, void *data
, int offset
,
699 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
701 int kvm_vcpu_read_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
703 int kvm_vcpu_write_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, const void *data
,
704 int offset
, int len
);
705 int kvm_vcpu_write_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, const void *data
,
707 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
709 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
710 void kvm_arch_vcpu_blocking(struct kvm_vcpu
*vcpu
);
711 void kvm_arch_vcpu_unblocking(struct kvm_vcpu
*vcpu
);
712 bool kvm_vcpu_wake_up(struct kvm_vcpu
*vcpu
);
713 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
714 int kvm_vcpu_yield_to(struct kvm_vcpu
*target
);
715 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
);
716 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
717 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
719 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
720 void kvm_reload_remote_mmus(struct kvm
*kvm
);
721 bool kvm_make_all_cpus_request(struct kvm
*kvm
, unsigned int req
);
723 long kvm_arch_dev_ioctl(struct file
*filp
,
724 unsigned int ioctl
, unsigned long arg
);
725 long kvm_arch_vcpu_ioctl(struct file
*filp
,
726 unsigned int ioctl
, unsigned long arg
);
727 int kvm_arch_vcpu_fault(struct kvm_vcpu
*vcpu
, struct vm_fault
*vmf
);
729 int kvm_vm_ioctl_check_extension(struct kvm
*kvm
, long ext
);
731 int kvm_get_dirty_log(struct kvm
*kvm
,
732 struct kvm_dirty_log
*log
, int *is_dirty
);
734 int kvm_get_dirty_log_protect(struct kvm
*kvm
,
735 struct kvm_dirty_log
*log
, bool *is_dirty
);
737 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm
*kvm
,
738 struct kvm_memory_slot
*slot
,
742 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
743 struct kvm_dirty_log
*log
);
745 int kvm_vm_ioctl_irq_line(struct kvm
*kvm
, struct kvm_irq_level
*irq_level
,
747 long kvm_arch_vm_ioctl(struct file
*filp
,
748 unsigned int ioctl
, unsigned long arg
);
750 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
751 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
753 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
754 struct kvm_translation
*tr
);
756 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
757 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
758 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
759 struct kvm_sregs
*sregs
);
760 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
761 struct kvm_sregs
*sregs
);
762 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
763 struct kvm_mp_state
*mp_state
);
764 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
765 struct kvm_mp_state
*mp_state
);
766 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
767 struct kvm_guest_debug
*dbg
);
768 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
770 int kvm_arch_init(void *opaque
);
771 void kvm_arch_exit(void);
773 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
);
774 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
);
776 void kvm_arch_sched_in(struct kvm_vcpu
*vcpu
, int cpu
);
778 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
);
779 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
780 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
781 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
);
782 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
);
783 void kvm_arch_vcpu_postcreate(struct kvm_vcpu
*vcpu
);
784 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
786 bool kvm_arch_has_vcpu_debugfs(void);
787 int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu
*vcpu
);
789 int kvm_arch_hardware_enable(void);
790 void kvm_arch_hardware_disable(void);
791 int kvm_arch_hardware_setup(void);
792 void kvm_arch_hardware_unsetup(void);
793 void kvm_arch_check_processor_compat(void *rtn
);
794 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
795 int kvm_arch_vcpu_should_kick(struct kvm_vcpu
*vcpu
);
797 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
798 static inline struct kvm
*kvm_arch_alloc_vm(void)
800 return kzalloc(sizeof(struct kvm
), GFP_KERNEL
);
803 static inline void kvm_arch_free_vm(struct kvm
*kvm
)
809 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
810 void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
);
811 void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
);
812 bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
);
814 static inline void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
)
818 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
)
822 static inline bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
)
827 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
828 void kvm_arch_start_assignment(struct kvm
*kvm
);
829 void kvm_arch_end_assignment(struct kvm
*kvm
);
830 bool kvm_arch_has_assigned_device(struct kvm
*kvm
);
832 static inline void kvm_arch_start_assignment(struct kvm
*kvm
)
836 static inline void kvm_arch_end_assignment(struct kvm
*kvm
)
840 static inline bool kvm_arch_has_assigned_device(struct kvm
*kvm
)
846 static inline struct swait_queue_head
*kvm_arch_vcpu_wq(struct kvm_vcpu
*vcpu
)
848 #ifdef __KVM_HAVE_ARCH_WQP
849 return vcpu
->arch
.wqp
;
855 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
857 * returns true if the virtual interrupt controller is initialized and
858 * ready to accept virtual IRQ. On some architectures the virtual interrupt
859 * controller is dynamically instantiated and this is not always true.
861 bool kvm_arch_intc_initialized(struct kvm
*kvm
);
863 static inline bool kvm_arch_intc_initialized(struct kvm
*kvm
)
869 int kvm_arch_init_vm(struct kvm
*kvm
, unsigned long type
);
870 void kvm_arch_destroy_vm(struct kvm
*kvm
);
871 void kvm_arch_sync_events(struct kvm
*kvm
);
873 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
874 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
876 bool kvm_is_reserved_pfn(kvm_pfn_t pfn
);
878 struct kvm_irq_ack_notifier
{
879 struct hlist_node link
;
881 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
884 int kvm_irq_map_gsi(struct kvm
*kvm
,
885 struct kvm_kernel_irq_routing_entry
*entries
, int gsi
);
886 int kvm_irq_map_chip_pin(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
888 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
,
890 int kvm_set_msi(struct kvm_kernel_irq_routing_entry
*irq_entry
, struct kvm
*kvm
,
891 int irq_source_id
, int level
, bool line_status
);
892 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry
*e
,
893 struct kvm
*kvm
, int irq_source_id
,
894 int level
, bool line_status
);
895 bool kvm_irq_has_notifier(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
896 void kvm_notify_acked_gsi(struct kvm
*kvm
, int gsi
);
897 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
898 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
899 struct kvm_irq_ack_notifier
*kian
);
900 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
901 struct kvm_irq_ack_notifier
*kian
);
902 int kvm_request_irq_source_id(struct kvm
*kvm
);
903 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
906 * search_memslots() and __gfn_to_memslot() are here because they are
907 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
908 * gfn_to_memslot() itself isn't here as an inline because that would
909 * bloat other code too much.
911 static inline struct kvm_memory_slot
*
912 search_memslots(struct kvm_memslots
*slots
, gfn_t gfn
)
914 int start
= 0, end
= slots
->used_slots
;
915 int slot
= atomic_read(&slots
->lru_slot
);
916 struct kvm_memory_slot
*memslots
= slots
->memslots
;
918 if (gfn
>= memslots
[slot
].base_gfn
&&
919 gfn
< memslots
[slot
].base_gfn
+ memslots
[slot
].npages
)
920 return &memslots
[slot
];
922 while (start
< end
) {
923 slot
= start
+ (end
- start
) / 2;
925 if (gfn
>= memslots
[slot
].base_gfn
)
931 if (gfn
>= memslots
[start
].base_gfn
&&
932 gfn
< memslots
[start
].base_gfn
+ memslots
[start
].npages
) {
933 atomic_set(&slots
->lru_slot
, start
);
934 return &memslots
[start
];
940 static inline struct kvm_memory_slot
*
941 __gfn_to_memslot(struct kvm_memslots
*slots
, gfn_t gfn
)
943 return search_memslots(slots
, gfn
);
946 static inline unsigned long
947 __gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
)
949 return slot
->userspace_addr
+ (gfn
- slot
->base_gfn
) * PAGE_SIZE
;
952 static inline int memslot_id(struct kvm
*kvm
, gfn_t gfn
)
954 return gfn_to_memslot(kvm
, gfn
)->id
;
958 hva_to_gfn_memslot(unsigned long hva
, struct kvm_memory_slot
*slot
)
960 gfn_t gfn_offset
= (hva
- slot
->userspace_addr
) >> PAGE_SHIFT
;
962 return slot
->base_gfn
+ gfn_offset
;
965 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
967 return (gpa_t
)gfn
<< PAGE_SHIFT
;
970 static inline gfn_t
gpa_to_gfn(gpa_t gpa
)
972 return (gfn_t
)(gpa
>> PAGE_SHIFT
);
975 static inline hpa_t
pfn_to_hpa(kvm_pfn_t pfn
)
977 return (hpa_t
)pfn
<< PAGE_SHIFT
;
980 static inline bool kvm_is_error_gpa(struct kvm
*kvm
, gpa_t gpa
)
982 unsigned long hva
= gfn_to_hva(kvm
, gpa_to_gfn(gpa
));
984 return kvm_is_error_hva(hva
);
992 struct kvm_stat_data
{
997 struct kvm_stats_debugfs_item
{
1000 enum kvm_stat_kind kind
;
1002 extern struct kvm_stats_debugfs_item debugfs_entries
[];
1003 extern struct dentry
*kvm_debugfs_dir
;
1005 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1006 static inline int mmu_notifier_retry(struct kvm
*kvm
, unsigned long mmu_seq
)
1008 if (unlikely(kvm
->mmu_notifier_count
))
1011 * Ensure the read of mmu_notifier_count happens before the read
1012 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1013 * mmu_notifier_invalidate_range_end to make sure that the caller
1014 * either sees the old (non-zero) value of mmu_notifier_count or
1015 * the new (incremented) value of mmu_notifier_seq.
1016 * PowerPC Book3s HV KVM calls this under a per-page lock
1017 * rather than under kvm->mmu_lock, for scalability, so
1018 * can't rely on kvm->mmu_lock to keep things ordered.
1021 if (kvm
->mmu_notifier_seq
!= mmu_seq
)
1027 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1030 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1031 #elif defined(CONFIG_ARM64)
1032 #define KVM_MAX_IRQ_ROUTES 4096
1034 #define KVM_MAX_IRQ_ROUTES 1024
1037 bool kvm_arch_can_set_irq_routing(struct kvm
*kvm
);
1038 int kvm_set_irq_routing(struct kvm
*kvm
,
1039 const struct kvm_irq_routing_entry
*entries
,
1042 int kvm_set_routing_entry(struct kvm
*kvm
,
1043 struct kvm_kernel_irq_routing_entry
*e
,
1044 const struct kvm_irq_routing_entry
*ue
);
1045 void kvm_free_irq_routing(struct kvm
*kvm
);
1049 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
1053 int kvm_send_userspace_msi(struct kvm
*kvm
, struct kvm_msi
*msi
);
1055 #ifdef CONFIG_HAVE_KVM_EVENTFD
1057 void kvm_eventfd_init(struct kvm
*kvm
);
1058 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
1060 #ifdef CONFIG_HAVE_KVM_IRQFD
1061 int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
);
1062 void kvm_irqfd_release(struct kvm
*kvm
);
1063 void kvm_irq_routing_update(struct kvm
*);
1065 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1070 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1075 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
1077 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1082 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1084 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1085 static inline void kvm_irq_routing_update(struct kvm
*kvm
)
1089 void kvm_arch_irq_routing_update(struct kvm
*kvm
);
1091 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
1096 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1098 static inline void kvm_make_request(int req
, struct kvm_vcpu
*vcpu
)
1101 * Ensure the rest of the request is published to kvm_check_request's
1102 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1105 set_bit(req
& KVM_REQUEST_MASK
, &vcpu
->requests
);
1108 static inline bool kvm_request_pending(struct kvm_vcpu
*vcpu
)
1110 return READ_ONCE(vcpu
->requests
);
1113 static inline bool kvm_test_request(int req
, struct kvm_vcpu
*vcpu
)
1115 return test_bit(req
& KVM_REQUEST_MASK
, &vcpu
->requests
);
1118 static inline void kvm_clear_request(int req
, struct kvm_vcpu
*vcpu
)
1120 clear_bit(req
& KVM_REQUEST_MASK
, &vcpu
->requests
);
1123 static inline bool kvm_check_request(int req
, struct kvm_vcpu
*vcpu
)
1125 if (kvm_test_request(req
, vcpu
)) {
1126 kvm_clear_request(req
, vcpu
);
1129 * Ensure the rest of the request is visible to kvm_check_request's
1130 * caller. Paired with the smp_wmb in kvm_make_request.
1132 smp_mb__after_atomic();
1139 extern bool kvm_rebooting
;
1141 extern unsigned int halt_poll_ns
;
1142 extern unsigned int halt_poll_ns_grow
;
1143 extern unsigned int halt_poll_ns_shrink
;
1146 struct kvm_device_ops
*ops
;
1149 struct list_head vm_node
;
1152 /* create, destroy, and name are mandatory */
1153 struct kvm_device_ops
{
1157 * create is called holding kvm->lock and any operations not suitable
1158 * to do while holding the lock should be deferred to init (see
1161 int (*create
)(struct kvm_device
*dev
, u32 type
);
1164 * init is called after create if create is successful and is called
1165 * outside of holding kvm->lock.
1167 void (*init
)(struct kvm_device
*dev
);
1170 * Destroy is responsible for freeing dev.
1172 * Destroy may be called before or after destructors are called
1173 * on emulated I/O regions, depending on whether a reference is
1174 * held by a vcpu or other kvm component that gets destroyed
1175 * after the emulated I/O.
1177 void (*destroy
)(struct kvm_device
*dev
);
1179 int (*set_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1180 int (*get_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1181 int (*has_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1182 long (*ioctl
)(struct kvm_device
*dev
, unsigned int ioctl
,
1186 void kvm_device_get(struct kvm_device
*dev
);
1187 void kvm_device_put(struct kvm_device
*dev
);
1188 struct kvm_device
*kvm_device_from_filp(struct file
*filp
);
1189 int kvm_register_device_ops(struct kvm_device_ops
*ops
, u32 type
);
1190 void kvm_unregister_device_ops(u32 type
);
1192 extern struct kvm_device_ops kvm_mpic_ops
;
1193 extern struct kvm_device_ops kvm_arm_vgic_v2_ops
;
1194 extern struct kvm_device_ops kvm_arm_vgic_v3_ops
;
1196 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1198 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1200 vcpu
->spin_loop
.in_spin_loop
= val
;
1202 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1204 vcpu
->spin_loop
.dy_eligible
= val
;
1207 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1209 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1213 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1216 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1218 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1219 bool kvm_arch_has_irq_bypass(void);
1220 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer
*,
1221 struct irq_bypass_producer
*);
1222 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer
*,
1223 struct irq_bypass_producer
*);
1224 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer
*);
1225 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer
*);
1226 int kvm_arch_update_irqfd_routing(struct kvm
*kvm
, unsigned int host_irq
,
1227 uint32_t guest_irq
, bool set
);
1228 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1230 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1231 /* If we wakeup during the poll time, was it a sucessful poll? */
1232 static inline bool vcpu_valid_wakeup(struct kvm_vcpu
*vcpu
)
1234 return vcpu
->valid_wakeup
;
1238 static inline bool vcpu_valid_wakeup(struct kvm_vcpu
*vcpu
)
1242 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */