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 <linux/nospec.h>
31 #include <asm/signal.h>
33 #include <linux/kvm.h>
34 #include <linux/kvm_para.h>
36 #include <linux/kvm_types.h>
38 #include <asm/kvm_host.h>
40 #ifndef KVM_MAX_VCPU_ID
41 #define KVM_MAX_VCPU_ID KVM_MAX_VCPUS
45 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
46 * in kvm, other bits are visible for userspace which are defined in
47 * include/linux/kvm_h.
49 #define KVM_MEMSLOT_INVALID (1UL << 16)
51 /* Two fragments for cross MMIO pages. */
52 #define KVM_MAX_MMIO_FRAGMENTS 2
54 #ifndef KVM_ADDRESS_SPACE_NUM
55 #define KVM_ADDRESS_SPACE_NUM 1
59 * For the normal pfn, the highest 12 bits should be zero,
60 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
61 * mask bit 63 to indicate the noslot pfn.
63 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
64 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
65 #define KVM_PFN_NOSLOT (0x1ULL << 63)
67 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
68 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
69 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
72 * error pfns indicate that the gfn is in slot but faild to
73 * translate it to pfn on host.
75 static inline bool is_error_pfn(kvm_pfn_t pfn
)
77 return !!(pfn
& KVM_PFN_ERR_MASK
);
81 * error_noslot pfns indicate that the gfn can not be
82 * translated to pfn - it is not in slot or failed to
83 * translate it to pfn.
85 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn
)
87 return !!(pfn
& KVM_PFN_ERR_NOSLOT_MASK
);
90 /* noslot pfn indicates that the gfn is not in slot. */
91 static inline bool is_noslot_pfn(kvm_pfn_t pfn
)
93 return pfn
== KVM_PFN_NOSLOT
;
97 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
98 * provide own defines and kvm_is_error_hva
100 #ifndef KVM_HVA_ERR_BAD
102 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
103 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
105 static inline bool kvm_is_error_hva(unsigned long addr
)
107 return addr
>= PAGE_OFFSET
;
112 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
114 static inline bool is_error_page(struct page
*page
)
119 #define KVM_REQUEST_MASK GENMASK(7,0)
120 #define KVM_REQUEST_NO_WAKEUP BIT(8)
121 #define KVM_REQUEST_WAIT BIT(9)
123 * Architecture-independent vcpu->requests bit members
124 * Bits 4-7 are reserved for more arch-independent bits.
126 #define KVM_REQ_TLB_FLUSH (0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
127 #define KVM_REQ_MMU_RELOAD (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
128 #define KVM_REQ_PENDING_TIMER 2
129 #define KVM_REQ_UNHALT 3
130 #define KVM_REQUEST_ARCH_BASE 8
132 #define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
133 BUILD_BUG_ON((unsigned)(nr) >= 32 - KVM_REQUEST_ARCH_BASE); \
134 (unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) | (flags)); \
136 #define KVM_ARCH_REQ(nr) KVM_ARCH_REQ_FLAGS(nr, 0)
138 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
139 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
141 extern struct kmem_cache
*kvm_vcpu_cache
;
143 extern spinlock_t kvm_lock
;
144 extern struct list_head vm_list
;
146 struct kvm_io_range
{
149 struct kvm_io_device
*dev
;
152 #define NR_IOBUS_DEVS 1000
157 struct kvm_io_range range
[];
163 KVM_VIRTIO_CCW_NOTIFY_BUS
,
168 int kvm_io_bus_write(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
169 int len
, const void *val
);
170 int kvm_io_bus_write_cookie(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
,
171 gpa_t addr
, int len
, const void *val
, long cookie
);
172 int kvm_io_bus_read(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
174 int kvm_io_bus_register_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
,
175 int len
, struct kvm_io_device
*dev
);
176 void kvm_io_bus_unregister_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
177 struct kvm_io_device
*dev
);
178 struct kvm_io_device
*kvm_io_bus_get_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
181 #ifdef CONFIG_KVM_ASYNC_PF
182 struct kvm_async_pf
{
183 struct work_struct work
;
184 struct list_head link
;
185 struct list_head queue
;
186 struct kvm_vcpu
*vcpu
;
187 struct mm_struct
*mm
;
190 struct kvm_arch_async_pf arch
;
194 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu
*vcpu
);
195 void kvm_check_async_pf_completion(struct kvm_vcpu
*vcpu
);
196 int kvm_setup_async_pf(struct kvm_vcpu
*vcpu
, gva_t gva
, unsigned long hva
,
197 struct kvm_arch_async_pf
*arch
);
198 int kvm_async_pf_wakeup_all(struct kvm_vcpu
*vcpu
);
205 READING_SHADOW_PAGE_TABLES
,
209 * Sometimes a large or cross-page mmio needs to be broken up into separate
210 * exits for userspace servicing.
212 struct kvm_mmio_fragment
{
220 #ifdef CONFIG_PREEMPT_NOTIFIERS
221 struct preempt_notifier preempt_notifier
;
227 unsigned long requests
;
228 unsigned long guest_debug
;
231 struct list_head blocked_vcpu_list
;
236 int guest_xcr0_loaded
;
237 struct swait_queue_head wq
;
238 struct pid __rcu
*pid
;
241 struct kvm_vcpu_stat stat
;
242 unsigned int halt_poll_ns
;
245 #ifdef CONFIG_HAS_IOMEM
247 int mmio_read_completed
;
249 int mmio_cur_fragment
;
250 int mmio_nr_fragments
;
251 struct kvm_mmio_fragment mmio_fragments
[KVM_MAX_MMIO_FRAGMENTS
];
254 #ifdef CONFIG_KVM_ASYNC_PF
257 struct list_head queue
;
258 struct list_head done
;
263 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
265 * Cpu relax intercept or pause loop exit optimization
266 * in_spin_loop: set when a vcpu does a pause loop exit
267 * or cpu relax intercepted.
268 * dy_eligible: indicates whether vcpu is eligible for directed yield.
276 struct kvm_vcpu_arch arch
;
277 struct dentry
*debugfs_dentry
;
280 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu
*vcpu
)
283 * The memory barrier ensures a previous write to vcpu->requests cannot
284 * be reordered with the read of vcpu->mode. It pairs with the general
285 * memory barrier following the write of vcpu->mode in VCPU RUN.
287 smp_mb__before_atomic();
288 return cmpxchg(&vcpu
->mode
, IN_GUEST_MODE
, EXITING_GUEST_MODE
);
292 * Some of the bitops functions do not support too long bitmaps.
293 * This number must be determined not to exceed such limits.
295 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
297 struct kvm_memory_slot
{
299 unsigned long npages
;
300 unsigned long *dirty_bitmap
;
301 struct kvm_arch_memory_slot arch
;
302 unsigned long userspace_addr
;
307 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot
*memslot
)
309 return ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
312 struct kvm_s390_adapter_int
{
325 struct kvm_kernel_irq_routing_entry
{
328 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
329 struct kvm
*kvm
, int irq_source_id
, int level
,
343 struct kvm_s390_adapter_int adapter
;
344 struct kvm_hv_sint hv_sint
;
346 struct hlist_node link
;
349 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
350 struct kvm_irq_routing_table
{
351 int chip
[KVM_NR_IRQCHIPS
][KVM_IRQCHIP_NUM_PINS
];
354 * Array indexed by gsi. Each entry contains list of irq chips
355 * the gsi is connected to.
357 struct hlist_head map
[0];
361 #ifndef KVM_PRIVATE_MEM_SLOTS
362 #define KVM_PRIVATE_MEM_SLOTS 0
365 #ifndef KVM_MEM_SLOTS_NUM
366 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
369 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
370 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu
*vcpu
)
378 * memslots are not sorted by id anymore, please use id_to_memslot()
379 * to get the memslot by its id.
381 struct kvm_memslots
{
383 struct kvm_memory_slot memslots
[KVM_MEM_SLOTS_NUM
];
384 /* The mapping table from slot id to the index in memslots[]. */
385 short id_to_index
[KVM_MEM_SLOTS_NUM
];
392 struct mutex slots_lock
;
393 struct mm_struct
*mm
; /* userspace tied to this vm */
394 struct kvm_memslots __rcu
*memslots
[KVM_ADDRESS_SPACE_NUM
];
395 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
398 * created_vcpus is protected by kvm->lock, and is incremented
399 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
400 * incremented after storing the kvm_vcpu pointer in vcpus,
401 * and is accessed atomically.
403 atomic_t online_vcpus
;
405 int last_boosted_vcpu
;
406 struct list_head vm_list
;
408 struct kvm_io_bus __rcu
*buses
[KVM_NR_BUSES
];
409 #ifdef CONFIG_HAVE_KVM_EVENTFD
412 struct list_head items
;
413 struct list_head resampler_list
;
414 struct mutex resampler_lock
;
416 struct list_head ioeventfds
;
418 struct kvm_vm_stat stat
;
419 struct kvm_arch arch
;
420 refcount_t users_count
;
421 #ifdef CONFIG_KVM_MMIO
422 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
423 spinlock_t ring_lock
;
424 struct list_head coalesced_zones
;
427 struct mutex irq_lock
;
428 #ifdef CONFIG_HAVE_KVM_IRQCHIP
430 * Update side is protected by irq_lock.
432 struct kvm_irq_routing_table __rcu
*irq_routing
;
434 #ifdef CONFIG_HAVE_KVM_IRQFD
435 struct hlist_head irq_ack_notifier_list
;
438 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
439 struct mmu_notifier mmu_notifier
;
440 unsigned long mmu_notifier_seq
;
441 long mmu_notifier_count
;
444 struct list_head devices
;
445 struct dentry
*debugfs_dentry
;
446 struct kvm_stat_data
**debugfs_stat_data
;
447 struct srcu_struct srcu
;
448 struct srcu_struct irq_srcu
;
452 #define kvm_err(fmt, ...) \
453 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
454 #define kvm_info(fmt, ...) \
455 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
456 #define kvm_debug(fmt, ...) \
457 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
458 #define kvm_debug_ratelimited(fmt, ...) \
459 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
461 #define kvm_pr_unimpl(fmt, ...) \
462 pr_err_ratelimited("kvm [%i]: " fmt, \
463 task_tgid_nr(current), ## __VA_ARGS__)
465 /* The guest did something we don't support. */
466 #define vcpu_unimpl(vcpu, fmt, ...) \
467 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
468 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
470 #define vcpu_debug(vcpu, fmt, ...) \
471 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
472 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
473 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
475 #define vcpu_err(vcpu, fmt, ...) \
476 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
478 static inline struct kvm_io_bus
*kvm_get_bus(struct kvm
*kvm
, enum kvm_bus idx
)
480 return srcu_dereference_check(kvm
->buses
[idx
], &kvm
->srcu
,
481 lockdep_is_held(&kvm
->slots_lock
) ||
482 !refcount_read(&kvm
->users_count
));
485 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
487 int num_vcpus
= atomic_read(&kvm
->online_vcpus
);
488 i
= array_index_nospec(i
, num_vcpus
);
490 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. */
492 return kvm
->vcpus
[i
];
495 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
497 idx < atomic_read(&kvm->online_vcpus) && \
498 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
501 static inline struct kvm_vcpu
*kvm_get_vcpu_by_id(struct kvm
*kvm
, int id
)
503 struct kvm_vcpu
*vcpu
= NULL
;
508 if (id
< KVM_MAX_VCPUS
)
509 vcpu
= kvm_get_vcpu(kvm
, id
);
510 if (vcpu
&& vcpu
->vcpu_id
== id
)
512 kvm_for_each_vcpu(i
, vcpu
, kvm
)
513 if (vcpu
->vcpu_id
== id
)
518 static inline int kvm_vcpu_get_idx(struct kvm_vcpu
*vcpu
)
520 struct kvm_vcpu
*tmp
;
523 kvm_for_each_vcpu(idx
, tmp
, vcpu
->kvm
)
529 #define kvm_for_each_memslot(memslot, slots) \
530 for (memslot = &slots->memslots[0]; \
531 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
534 int kvm_vcpu_init(struct kvm_vcpu
*vcpu
, struct kvm
*kvm
, unsigned id
);
535 void kvm_vcpu_uninit(struct kvm_vcpu
*vcpu
);
537 int __must_check
vcpu_load(struct kvm_vcpu
*vcpu
);
538 void vcpu_put(struct kvm_vcpu
*vcpu
);
540 #ifdef __KVM_HAVE_IOAPIC
541 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm
*kvm
);
542 void kvm_arch_post_irq_routing_update(struct kvm
*kvm
);
544 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm
*kvm
)
547 static inline void kvm_arch_post_irq_routing_update(struct kvm
*kvm
)
552 #ifdef CONFIG_HAVE_KVM_IRQFD
553 int kvm_irqfd_init(void);
554 void kvm_irqfd_exit(void);
556 static inline int kvm_irqfd_init(void)
561 static inline void kvm_irqfd_exit(void)
565 int kvm_init(void *opaque
, unsigned vcpu_size
, unsigned vcpu_align
,
566 struct module
*module
);
569 void kvm_get_kvm(struct kvm
*kvm
);
570 void kvm_put_kvm(struct kvm
*kvm
);
572 static inline struct kvm_memslots
*__kvm_memslots(struct kvm
*kvm
, int as_id
)
574 as_id
= array_index_nospec(as_id
, KVM_ADDRESS_SPACE_NUM
);
575 return srcu_dereference_check(kvm
->memslots
[as_id
], &kvm
->srcu
,
576 lockdep_is_held(&kvm
->slots_lock
) ||
577 !refcount_read(&kvm
->users_count
));
580 static inline struct kvm_memslots
*kvm_memslots(struct kvm
*kvm
)
582 return __kvm_memslots(kvm
, 0);
585 static inline struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
)
587 int as_id
= kvm_arch_vcpu_memslots_id(vcpu
);
589 return __kvm_memslots(vcpu
->kvm
, as_id
);
592 static inline struct kvm_memory_slot
*
593 id_to_memslot(struct kvm_memslots
*slots
, int id
)
595 int index
= slots
->id_to_index
[id
];
596 struct kvm_memory_slot
*slot
;
598 slot
= &slots
->memslots
[index
];
600 WARN_ON(slot
->id
!= id
);
605 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
606 * - create a new memory slot
607 * - delete an existing memory slot
608 * - modify an existing memory slot
609 * -- move it in the guest physical memory space
610 * -- just change its flags
612 * Since flags can be changed by some of these operations, the following
613 * differentiation is the best we can do for __kvm_set_memory_region():
622 int kvm_set_memory_region(struct kvm
*kvm
,
623 const struct kvm_userspace_memory_region
*mem
);
624 int __kvm_set_memory_region(struct kvm
*kvm
,
625 const struct kvm_userspace_memory_region
*mem
);
626 void kvm_arch_free_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*free
,
627 struct kvm_memory_slot
*dont
);
628 int kvm_arch_create_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*slot
,
629 unsigned long npages
);
630 void kvm_arch_memslots_updated(struct kvm
*kvm
, u64 gen
);
631 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
632 struct kvm_memory_slot
*memslot
,
633 const struct kvm_userspace_memory_region
*mem
,
634 enum kvm_mr_change change
);
635 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
636 const struct kvm_userspace_memory_region
*mem
,
637 const struct kvm_memory_slot
*old
,
638 const struct kvm_memory_slot
*new,
639 enum kvm_mr_change change
);
640 bool kvm_largepages_enabled(void);
641 void kvm_disable_largepages(void);
642 /* flush all memory translations */
643 void kvm_arch_flush_shadow_all(struct kvm
*kvm
);
644 /* flush memory translations pointing to 'slot' */
645 void kvm_arch_flush_shadow_memslot(struct kvm
*kvm
,
646 struct kvm_memory_slot
*slot
);
648 int gfn_to_page_many_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
,
649 struct page
**pages
, int nr_pages
);
651 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
652 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
653 unsigned long gfn_to_hva_prot(struct kvm
*kvm
, gfn_t gfn
, bool *writable
);
654 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
655 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
657 void kvm_release_page_clean(struct page
*page
);
658 void kvm_release_page_dirty(struct page
*page
);
659 void kvm_set_page_accessed(struct page
*page
);
661 kvm_pfn_t
gfn_to_pfn_atomic(struct kvm
*kvm
, gfn_t gfn
);
662 kvm_pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
663 kvm_pfn_t
gfn_to_pfn_prot(struct kvm
*kvm
, gfn_t gfn
, bool write_fault
,
665 kvm_pfn_t
gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
666 kvm_pfn_t
gfn_to_pfn_memslot_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
);
667 kvm_pfn_t
__gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
668 bool atomic
, bool *async
, bool write_fault
,
671 void kvm_release_pfn_clean(kvm_pfn_t pfn
);
672 void kvm_release_pfn_dirty(kvm_pfn_t pfn
);
673 void kvm_set_pfn_dirty(kvm_pfn_t pfn
);
674 void kvm_set_pfn_accessed(kvm_pfn_t pfn
);
675 void kvm_get_pfn(kvm_pfn_t pfn
);
677 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
679 int kvm_read_guest_atomic(struct kvm
*kvm
, gpa_t gpa
, void *data
,
681 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
682 int kvm_read_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
683 void *data
, unsigned long len
);
684 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
685 int offset
, int len
);
686 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
688 int kvm_write_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
689 void *data
, unsigned long len
);
690 int kvm_write_guest_offset_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
691 void *data
, unsigned int offset
,
693 int kvm_gfn_to_hva_cache_init(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
694 gpa_t gpa
, unsigned long len
);
695 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
696 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
697 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
698 bool kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
699 unsigned long kvm_host_page_size(struct kvm
*kvm
, gfn_t gfn
);
700 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
702 struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
);
703 struct kvm_memory_slot
*kvm_vcpu_gfn_to_memslot(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
704 kvm_pfn_t
kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
705 kvm_pfn_t
kvm_vcpu_gfn_to_pfn(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
706 struct page
*kvm_vcpu_gfn_to_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
707 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
708 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu
*vcpu
, gfn_t gfn
, bool *writable
);
709 int kvm_vcpu_read_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, void *data
, int offset
,
711 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
713 int kvm_vcpu_read_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
715 int kvm_vcpu_write_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, const void *data
,
716 int offset
, int len
);
717 int kvm_vcpu_write_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, const void *data
,
719 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
721 void kvm_sigset_activate(struct kvm_vcpu
*vcpu
);
722 void kvm_sigset_deactivate(struct kvm_vcpu
*vcpu
);
724 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
725 void kvm_arch_vcpu_blocking(struct kvm_vcpu
*vcpu
);
726 void kvm_arch_vcpu_unblocking(struct kvm_vcpu
*vcpu
);
727 bool kvm_vcpu_wake_up(struct kvm_vcpu
*vcpu
);
728 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
729 int kvm_vcpu_yield_to(struct kvm_vcpu
*target
);
730 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
, bool usermode_vcpu_not_eligible
);
731 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
732 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
734 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
735 void kvm_reload_remote_mmus(struct kvm
*kvm
);
736 bool kvm_make_all_cpus_request(struct kvm
*kvm
, unsigned int req
);
738 long kvm_arch_dev_ioctl(struct file
*filp
,
739 unsigned int ioctl
, unsigned long arg
);
740 long kvm_arch_vcpu_ioctl(struct file
*filp
,
741 unsigned int ioctl
, unsigned long arg
);
742 int kvm_arch_vcpu_fault(struct kvm_vcpu
*vcpu
, struct vm_fault
*vmf
);
744 int kvm_vm_ioctl_check_extension(struct kvm
*kvm
, long ext
);
746 int kvm_get_dirty_log(struct kvm
*kvm
,
747 struct kvm_dirty_log
*log
, int *is_dirty
);
749 int kvm_get_dirty_log_protect(struct kvm
*kvm
,
750 struct kvm_dirty_log
*log
, bool *is_dirty
);
752 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm
*kvm
,
753 struct kvm_memory_slot
*slot
,
757 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
758 struct kvm_dirty_log
*log
);
760 int kvm_vm_ioctl_irq_line(struct kvm
*kvm
, struct kvm_irq_level
*irq_level
,
762 long kvm_arch_vm_ioctl(struct file
*filp
,
763 unsigned int ioctl
, unsigned long arg
);
765 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
766 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
768 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
769 struct kvm_translation
*tr
);
771 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
772 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
773 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
774 struct kvm_sregs
*sregs
);
775 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
776 struct kvm_sregs
*sregs
);
777 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
778 struct kvm_mp_state
*mp_state
);
779 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
780 struct kvm_mp_state
*mp_state
);
781 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
782 struct kvm_guest_debug
*dbg
);
783 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
785 int kvm_arch_init(void *opaque
);
786 void kvm_arch_exit(void);
788 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
);
789 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
);
791 void kvm_arch_sched_in(struct kvm_vcpu
*vcpu
, int cpu
);
793 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
);
794 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
795 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
796 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
);
797 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
);
798 void kvm_arch_vcpu_postcreate(struct kvm_vcpu
*vcpu
);
799 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
801 bool kvm_arch_has_vcpu_debugfs(void);
802 int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu
*vcpu
);
804 int kvm_arch_hardware_enable(void);
805 void kvm_arch_hardware_disable(void);
806 int kvm_arch_hardware_setup(void);
807 void kvm_arch_hardware_unsetup(void);
808 void kvm_arch_check_processor_compat(void *rtn
);
809 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
810 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu
*vcpu
);
811 int kvm_arch_vcpu_should_kick(struct kvm_vcpu
*vcpu
);
812 bool kvm_arch_dy_runnable(struct kvm_vcpu
*vcpu
);
814 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
815 static inline struct kvm
*kvm_arch_alloc_vm(void)
817 return kzalloc(sizeof(struct kvm
), GFP_KERNEL
);
820 static inline void kvm_arch_free_vm(struct kvm
*kvm
)
826 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
827 void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
);
828 void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
);
829 bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
);
831 static inline void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
)
835 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
)
839 static inline bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
)
844 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
845 void kvm_arch_start_assignment(struct kvm
*kvm
);
846 void kvm_arch_end_assignment(struct kvm
*kvm
);
847 bool kvm_arch_has_assigned_device(struct kvm
*kvm
);
849 static inline void kvm_arch_start_assignment(struct kvm
*kvm
)
853 static inline void kvm_arch_end_assignment(struct kvm
*kvm
)
857 static inline bool kvm_arch_has_assigned_device(struct kvm
*kvm
)
863 static inline struct swait_queue_head
*kvm_arch_vcpu_wq(struct kvm_vcpu
*vcpu
)
865 #ifdef __KVM_HAVE_ARCH_WQP
866 return vcpu
->arch
.wqp
;
872 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
874 * returns true if the virtual interrupt controller is initialized and
875 * ready to accept virtual IRQ. On some architectures the virtual interrupt
876 * controller is dynamically instantiated and this is not always true.
878 bool kvm_arch_intc_initialized(struct kvm
*kvm
);
880 static inline bool kvm_arch_intc_initialized(struct kvm
*kvm
)
886 int kvm_arch_init_vm(struct kvm
*kvm
, unsigned long type
);
887 void kvm_arch_destroy_vm(struct kvm
*kvm
);
888 void kvm_arch_sync_events(struct kvm
*kvm
);
890 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
891 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
893 bool kvm_is_reserved_pfn(kvm_pfn_t pfn
);
895 struct kvm_irq_ack_notifier
{
896 struct hlist_node link
;
898 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
901 int kvm_irq_map_gsi(struct kvm
*kvm
,
902 struct kvm_kernel_irq_routing_entry
*entries
, int gsi
);
903 int kvm_irq_map_chip_pin(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
905 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
,
907 int kvm_set_msi(struct kvm_kernel_irq_routing_entry
*irq_entry
, struct kvm
*kvm
,
908 int irq_source_id
, int level
, bool line_status
);
909 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry
*e
,
910 struct kvm
*kvm
, int irq_source_id
,
911 int level
, bool line_status
);
912 bool kvm_irq_has_notifier(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
913 void kvm_notify_acked_gsi(struct kvm
*kvm
, int gsi
);
914 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
915 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
916 struct kvm_irq_ack_notifier
*kian
);
917 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
918 struct kvm_irq_ack_notifier
*kian
);
919 int kvm_request_irq_source_id(struct kvm
*kvm
);
920 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
923 * search_memslots() and __gfn_to_memslot() are here because they are
924 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
925 * gfn_to_memslot() itself isn't here as an inline because that would
926 * bloat other code too much.
928 static inline struct kvm_memory_slot
*
929 search_memslots(struct kvm_memslots
*slots
, gfn_t gfn
)
931 int start
= 0, end
= slots
->used_slots
;
932 int slot
= atomic_read(&slots
->lru_slot
);
933 struct kvm_memory_slot
*memslots
= slots
->memslots
;
935 if (gfn
>= memslots
[slot
].base_gfn
&&
936 gfn
< memslots
[slot
].base_gfn
+ memslots
[slot
].npages
)
937 return &memslots
[slot
];
939 while (start
< end
) {
940 slot
= start
+ (end
- start
) / 2;
942 if (gfn
>= memslots
[slot
].base_gfn
)
948 if (gfn
>= memslots
[start
].base_gfn
&&
949 gfn
< memslots
[start
].base_gfn
+ memslots
[start
].npages
) {
950 atomic_set(&slots
->lru_slot
, start
);
951 return &memslots
[start
];
957 static inline struct kvm_memory_slot
*
958 __gfn_to_memslot(struct kvm_memslots
*slots
, gfn_t gfn
)
960 return search_memslots(slots
, gfn
);
963 static inline unsigned long
964 __gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
)
966 return slot
->userspace_addr
+ (gfn
- slot
->base_gfn
) * PAGE_SIZE
;
969 static inline int memslot_id(struct kvm
*kvm
, gfn_t gfn
)
971 return gfn_to_memslot(kvm
, gfn
)->id
;
975 hva_to_gfn_memslot(unsigned long hva
, struct kvm_memory_slot
*slot
)
977 gfn_t gfn_offset
= (hva
- slot
->userspace_addr
) >> PAGE_SHIFT
;
979 return slot
->base_gfn
+ gfn_offset
;
982 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
984 return (gpa_t
)gfn
<< PAGE_SHIFT
;
987 static inline gfn_t
gpa_to_gfn(gpa_t gpa
)
989 return (gfn_t
)(gpa
>> PAGE_SHIFT
);
992 static inline hpa_t
pfn_to_hpa(kvm_pfn_t pfn
)
994 return (hpa_t
)pfn
<< PAGE_SHIFT
;
997 static inline struct page
*kvm_vcpu_gpa_to_page(struct kvm_vcpu
*vcpu
,
1000 return kvm_vcpu_gfn_to_page(vcpu
, gpa_to_gfn(gpa
));
1003 static inline bool kvm_is_error_gpa(struct kvm
*kvm
, gpa_t gpa
)
1005 unsigned long hva
= gfn_to_hva(kvm
, gpa_to_gfn(gpa
));
1007 return kvm_is_error_hva(hva
);
1010 enum kvm_stat_kind
{
1015 struct kvm_stat_data
{
1020 struct kvm_stats_debugfs_item
{
1023 enum kvm_stat_kind kind
;
1025 extern struct kvm_stats_debugfs_item debugfs_entries
[];
1026 extern struct dentry
*kvm_debugfs_dir
;
1028 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1029 static inline int mmu_notifier_retry(struct kvm
*kvm
, unsigned long mmu_seq
)
1031 if (unlikely(kvm
->mmu_notifier_count
))
1034 * Ensure the read of mmu_notifier_count happens before the read
1035 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1036 * mmu_notifier_invalidate_range_end to make sure that the caller
1037 * either sees the old (non-zero) value of mmu_notifier_count or
1038 * the new (incremented) value of mmu_notifier_seq.
1039 * PowerPC Book3s HV KVM calls this under a per-page lock
1040 * rather than under kvm->mmu_lock, for scalability, so
1041 * can't rely on kvm->mmu_lock to keep things ordered.
1044 if (kvm
->mmu_notifier_seq
!= mmu_seq
)
1050 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1052 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1054 bool kvm_arch_can_set_irq_routing(struct kvm
*kvm
);
1055 int kvm_set_irq_routing(struct kvm
*kvm
,
1056 const struct kvm_irq_routing_entry
*entries
,
1059 int kvm_set_routing_entry(struct kvm
*kvm
,
1060 struct kvm_kernel_irq_routing_entry
*e
,
1061 const struct kvm_irq_routing_entry
*ue
);
1062 void kvm_free_irq_routing(struct kvm
*kvm
);
1066 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
1070 int kvm_send_userspace_msi(struct kvm
*kvm
, struct kvm_msi
*msi
);
1072 #ifdef CONFIG_HAVE_KVM_EVENTFD
1074 void kvm_eventfd_init(struct kvm
*kvm
);
1075 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
1077 #ifdef CONFIG_HAVE_KVM_IRQFD
1078 int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
);
1079 void kvm_irqfd_release(struct kvm
*kvm
);
1080 void kvm_irq_routing_update(struct kvm
*);
1082 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1087 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1092 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
1094 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1099 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1101 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1102 static inline void kvm_irq_routing_update(struct kvm
*kvm
)
1107 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
1112 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1114 void kvm_arch_irq_routing_update(struct kvm
*kvm
);
1116 static inline void kvm_make_request(int req
, struct kvm_vcpu
*vcpu
)
1119 * Ensure the rest of the request is published to kvm_check_request's
1120 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1123 set_bit(req
& KVM_REQUEST_MASK
, &vcpu
->requests
);
1126 static inline bool kvm_request_pending(struct kvm_vcpu
*vcpu
)
1128 return READ_ONCE(vcpu
->requests
);
1131 static inline bool kvm_test_request(int req
, struct kvm_vcpu
*vcpu
)
1133 return test_bit(req
& KVM_REQUEST_MASK
, &vcpu
->requests
);
1136 static inline void kvm_clear_request(int req
, struct kvm_vcpu
*vcpu
)
1138 clear_bit(req
& KVM_REQUEST_MASK
, &vcpu
->requests
);
1141 static inline bool kvm_check_request(int req
, struct kvm_vcpu
*vcpu
)
1143 if (kvm_test_request(req
, vcpu
)) {
1144 kvm_clear_request(req
, vcpu
);
1147 * Ensure the rest of the request is visible to kvm_check_request's
1148 * caller. Paired with the smp_wmb in kvm_make_request.
1150 smp_mb__after_atomic();
1157 extern bool kvm_rebooting
;
1159 extern unsigned int halt_poll_ns
;
1160 extern unsigned int halt_poll_ns_grow
;
1161 extern unsigned int halt_poll_ns_shrink
;
1164 struct kvm_device_ops
*ops
;
1167 struct list_head vm_node
;
1170 /* create, destroy, and name are mandatory */
1171 struct kvm_device_ops
{
1175 * create is called holding kvm->lock and any operations not suitable
1176 * to do while holding the lock should be deferred to init (see
1179 int (*create
)(struct kvm_device
*dev
, u32 type
);
1182 * init is called after create if create is successful and is called
1183 * outside of holding kvm->lock.
1185 void (*init
)(struct kvm_device
*dev
);
1188 * Destroy is responsible for freeing dev.
1190 * Destroy may be called before or after destructors are called
1191 * on emulated I/O regions, depending on whether a reference is
1192 * held by a vcpu or other kvm component that gets destroyed
1193 * after the emulated I/O.
1195 void (*destroy
)(struct kvm_device
*dev
);
1197 int (*set_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1198 int (*get_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1199 int (*has_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1200 long (*ioctl
)(struct kvm_device
*dev
, unsigned int ioctl
,
1204 void kvm_device_get(struct kvm_device
*dev
);
1205 void kvm_device_put(struct kvm_device
*dev
);
1206 struct kvm_device
*kvm_device_from_filp(struct file
*filp
);
1207 int kvm_register_device_ops(struct kvm_device_ops
*ops
, u32 type
);
1208 void kvm_unregister_device_ops(u32 type
);
1210 extern struct kvm_device_ops kvm_mpic_ops
;
1211 extern struct kvm_device_ops kvm_arm_vgic_v2_ops
;
1212 extern struct kvm_device_ops kvm_arm_vgic_v3_ops
;
1214 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1216 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1218 vcpu
->spin_loop
.in_spin_loop
= val
;
1220 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1222 vcpu
->spin_loop
.dy_eligible
= val
;
1225 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1227 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1231 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1234 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1236 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1237 bool kvm_arch_has_irq_bypass(void);
1238 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer
*,
1239 struct irq_bypass_producer
*);
1240 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer
*,
1241 struct irq_bypass_producer
*);
1242 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer
*);
1243 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer
*);
1244 int kvm_arch_update_irqfd_routing(struct kvm
*kvm
, unsigned int host_irq
,
1245 uint32_t guest_irq
, bool set
);
1246 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1248 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1249 /* If we wakeup during the poll time, was it a sucessful poll? */
1250 static inline bool vcpu_valid_wakeup(struct kvm_vcpu
*vcpu
)
1252 return vcpu
->valid_wakeup
;
1256 static inline bool vcpu_valid_wakeup(struct kvm_vcpu
*vcpu
)
1260 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */