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
130 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
131 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
133 extern struct kmem_cache
*kvm_vcpu_cache
;
135 extern spinlock_t kvm_lock
;
136 extern struct list_head vm_list
;
138 struct kvm_io_range
{
141 struct kvm_io_device
*dev
;
144 #define NR_IOBUS_DEVS 1000
149 struct kvm_io_range range
[];
155 KVM_VIRTIO_CCW_NOTIFY_BUS
,
160 int kvm_io_bus_write(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
161 int len
, const void *val
);
162 int kvm_io_bus_write_cookie(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
,
163 gpa_t addr
, int len
, const void *val
, long cookie
);
164 int kvm_io_bus_read(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
166 int kvm_io_bus_register_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
,
167 int len
, struct kvm_io_device
*dev
);
168 void kvm_io_bus_unregister_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
169 struct kvm_io_device
*dev
);
170 struct kvm_io_device
*kvm_io_bus_get_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
173 #ifdef CONFIG_KVM_ASYNC_PF
174 struct kvm_async_pf
{
175 struct work_struct work
;
176 struct list_head link
;
177 struct list_head queue
;
178 struct kvm_vcpu
*vcpu
;
179 struct mm_struct
*mm
;
182 struct kvm_arch_async_pf arch
;
186 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu
*vcpu
);
187 void kvm_check_async_pf_completion(struct kvm_vcpu
*vcpu
);
188 int kvm_setup_async_pf(struct kvm_vcpu
*vcpu
, gva_t gva
, unsigned long hva
,
189 struct kvm_arch_async_pf
*arch
);
190 int kvm_async_pf_wakeup_all(struct kvm_vcpu
*vcpu
);
197 READING_SHADOW_PAGE_TABLES
,
201 * Sometimes a large or cross-page mmio needs to be broken up into separate
202 * exits for userspace servicing.
204 struct kvm_mmio_fragment
{
212 #ifdef CONFIG_PREEMPT_NOTIFIERS
213 struct preempt_notifier preempt_notifier
;
219 unsigned long requests
;
220 unsigned long guest_debug
;
223 struct list_head blocked_vcpu_list
;
228 int guest_fpu_loaded
, guest_xcr0_loaded
;
229 struct swait_queue_head wq
;
233 struct kvm_vcpu_stat stat
;
234 unsigned int halt_poll_ns
;
237 #ifdef CONFIG_HAS_IOMEM
239 int mmio_read_completed
;
241 int mmio_cur_fragment
;
242 int mmio_nr_fragments
;
243 struct kvm_mmio_fragment mmio_fragments
[KVM_MAX_MMIO_FRAGMENTS
];
246 #ifdef CONFIG_KVM_ASYNC_PF
249 struct list_head queue
;
250 struct list_head done
;
255 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
257 * Cpu relax intercept or pause loop exit optimization
258 * in_spin_loop: set when a vcpu does a pause loop exit
259 * or cpu relax intercepted.
260 * dy_eligible: indicates whether vcpu is eligible for directed yield.
268 struct kvm_vcpu_arch arch
;
269 struct dentry
*debugfs_dentry
;
272 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu
*vcpu
)
275 * The memory barrier ensures a previous write to vcpu->requests cannot
276 * be reordered with the read of vcpu->mode. It pairs with the general
277 * memory barrier following the write of vcpu->mode in VCPU RUN.
279 smp_mb__before_atomic();
280 return cmpxchg(&vcpu
->mode
, IN_GUEST_MODE
, EXITING_GUEST_MODE
);
284 * Some of the bitops functions do not support too long bitmaps.
285 * This number must be determined not to exceed such limits.
287 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
289 struct kvm_memory_slot
{
291 unsigned long npages
;
292 unsigned long *dirty_bitmap
;
293 struct kvm_arch_memory_slot arch
;
294 unsigned long userspace_addr
;
299 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot
*memslot
)
301 return ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
304 struct kvm_s390_adapter_int
{
317 struct kvm_kernel_irq_routing_entry
{
320 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
321 struct kvm
*kvm
, int irq_source_id
, int level
,
335 struct kvm_s390_adapter_int adapter
;
336 struct kvm_hv_sint hv_sint
;
338 struct hlist_node link
;
341 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
342 struct kvm_irq_routing_table
{
343 int chip
[KVM_NR_IRQCHIPS
][KVM_IRQCHIP_NUM_PINS
];
346 * Array indexed by gsi. Each entry contains list of irq chips
347 * the gsi is connected to.
349 struct hlist_head map
[0];
353 #ifndef KVM_PRIVATE_MEM_SLOTS
354 #define KVM_PRIVATE_MEM_SLOTS 0
357 #ifndef KVM_MEM_SLOTS_NUM
358 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
361 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
362 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu
*vcpu
)
370 * memslots are not sorted by id anymore, please use id_to_memslot()
371 * to get the memslot by its id.
373 struct kvm_memslots
{
375 struct kvm_memory_slot memslots
[KVM_MEM_SLOTS_NUM
];
376 /* The mapping table from slot id to the index in memslots[]. */
377 short id_to_index
[KVM_MEM_SLOTS_NUM
];
384 struct mutex slots_lock
;
385 struct mm_struct
*mm
; /* userspace tied to this vm */
386 struct kvm_memslots
*memslots
[KVM_ADDRESS_SPACE_NUM
];
387 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
390 * created_vcpus is protected by kvm->lock, and is incremented
391 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
392 * incremented after storing the kvm_vcpu pointer in vcpus,
393 * and is accessed atomically.
395 atomic_t online_vcpus
;
397 int last_boosted_vcpu
;
398 struct list_head vm_list
;
400 struct kvm_io_bus
*buses
[KVM_NR_BUSES
];
401 #ifdef CONFIG_HAVE_KVM_EVENTFD
404 struct list_head items
;
405 struct list_head resampler_list
;
406 struct mutex resampler_lock
;
408 struct list_head ioeventfds
;
410 struct kvm_vm_stat stat
;
411 struct kvm_arch arch
;
412 refcount_t users_count
;
413 #ifdef CONFIG_KVM_MMIO
414 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
415 spinlock_t ring_lock
;
416 struct list_head coalesced_zones
;
419 struct mutex irq_lock
;
420 #ifdef CONFIG_HAVE_KVM_IRQCHIP
422 * Update side is protected by irq_lock.
424 struct kvm_irq_routing_table __rcu
*irq_routing
;
426 #ifdef CONFIG_HAVE_KVM_IRQFD
427 struct hlist_head irq_ack_notifier_list
;
430 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
431 struct mmu_notifier mmu_notifier
;
432 unsigned long mmu_notifier_seq
;
433 long mmu_notifier_count
;
436 struct list_head devices
;
437 struct dentry
*debugfs_dentry
;
438 struct kvm_stat_data
**debugfs_stat_data
;
439 struct srcu_struct srcu
;
440 struct srcu_struct irq_srcu
;
443 #define kvm_err(fmt, ...) \
444 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
445 #define kvm_info(fmt, ...) \
446 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
447 #define kvm_debug(fmt, ...) \
448 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
449 #define kvm_debug_ratelimited(fmt, ...) \
450 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
452 #define kvm_pr_unimpl(fmt, ...) \
453 pr_err_ratelimited("kvm [%i]: " fmt, \
454 task_tgid_nr(current), ## __VA_ARGS__)
456 /* The guest did something we don't support. */
457 #define vcpu_unimpl(vcpu, fmt, ...) \
458 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
459 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
461 #define vcpu_debug(vcpu, fmt, ...) \
462 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
463 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
464 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
466 #define vcpu_err(vcpu, fmt, ...) \
467 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
469 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
471 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
472 * the caller has read kvm->online_vcpus before (as is the case
473 * for kvm_for_each_vcpu, for example).
476 return kvm
->vcpus
[i
];
479 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
481 idx < atomic_read(&kvm->online_vcpus) && \
482 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
485 static inline struct kvm_vcpu
*kvm_get_vcpu_by_id(struct kvm
*kvm
, int id
)
487 struct kvm_vcpu
*vcpu
= NULL
;
492 if (id
< KVM_MAX_VCPUS
)
493 vcpu
= kvm_get_vcpu(kvm
, id
);
494 if (vcpu
&& vcpu
->vcpu_id
== id
)
496 kvm_for_each_vcpu(i
, vcpu
, kvm
)
497 if (vcpu
->vcpu_id
== id
)
502 static inline int kvm_vcpu_get_idx(struct kvm_vcpu
*vcpu
)
504 struct kvm_vcpu
*tmp
;
507 kvm_for_each_vcpu(idx
, tmp
, vcpu
->kvm
)
513 #define kvm_for_each_memslot(memslot, slots) \
514 for (memslot = &slots->memslots[0]; \
515 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
518 int kvm_vcpu_init(struct kvm_vcpu
*vcpu
, struct kvm
*kvm
, unsigned id
);
519 void kvm_vcpu_uninit(struct kvm_vcpu
*vcpu
);
521 int __must_check
vcpu_load(struct kvm_vcpu
*vcpu
);
522 void vcpu_put(struct kvm_vcpu
*vcpu
);
524 #ifdef __KVM_HAVE_IOAPIC
525 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm
*kvm
);
526 void kvm_arch_post_irq_routing_update(struct kvm
*kvm
);
528 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm
*kvm
)
531 static inline void kvm_arch_post_irq_routing_update(struct kvm
*kvm
)
536 #ifdef CONFIG_HAVE_KVM_IRQFD
537 int kvm_irqfd_init(void);
538 void kvm_irqfd_exit(void);
540 static inline int kvm_irqfd_init(void)
545 static inline void kvm_irqfd_exit(void)
549 int kvm_init(void *opaque
, unsigned vcpu_size
, unsigned vcpu_align
,
550 struct module
*module
);
553 void kvm_get_kvm(struct kvm
*kvm
);
554 void kvm_put_kvm(struct kvm
*kvm
);
556 static inline struct kvm_memslots
*__kvm_memslots(struct kvm
*kvm
, int as_id
)
558 return rcu_dereference_check(kvm
->memslots
[as_id
],
559 srcu_read_lock_held(&kvm
->srcu
)
560 || lockdep_is_held(&kvm
->slots_lock
));
563 static inline struct kvm_memslots
*kvm_memslots(struct kvm
*kvm
)
565 return __kvm_memslots(kvm
, 0);
568 static inline struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
)
570 int as_id
= kvm_arch_vcpu_memslots_id(vcpu
);
572 return __kvm_memslots(vcpu
->kvm
, as_id
);
575 static inline struct kvm_memory_slot
*
576 id_to_memslot(struct kvm_memslots
*slots
, int id
)
578 int index
= slots
->id_to_index
[id
];
579 struct kvm_memory_slot
*slot
;
581 slot
= &slots
->memslots
[index
];
583 WARN_ON(slot
->id
!= id
);
588 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
589 * - create a new memory slot
590 * - delete an existing memory slot
591 * - modify an existing memory slot
592 * -- move it in the guest physical memory space
593 * -- just change its flags
595 * Since flags can be changed by some of these operations, the following
596 * differentiation is the best we can do for __kvm_set_memory_region():
605 int kvm_set_memory_region(struct kvm
*kvm
,
606 const struct kvm_userspace_memory_region
*mem
);
607 int __kvm_set_memory_region(struct kvm
*kvm
,
608 const struct kvm_userspace_memory_region
*mem
);
609 void kvm_arch_free_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*free
,
610 struct kvm_memory_slot
*dont
);
611 int kvm_arch_create_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*slot
,
612 unsigned long npages
);
613 void kvm_arch_memslots_updated(struct kvm
*kvm
, struct kvm_memslots
*slots
);
614 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
615 struct kvm_memory_slot
*memslot
,
616 const struct kvm_userspace_memory_region
*mem
,
617 enum kvm_mr_change change
);
618 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
619 const struct kvm_userspace_memory_region
*mem
,
620 const struct kvm_memory_slot
*old
,
621 const struct kvm_memory_slot
*new,
622 enum kvm_mr_change change
);
623 bool kvm_largepages_enabled(void);
624 void kvm_disable_largepages(void);
625 /* flush all memory translations */
626 void kvm_arch_flush_shadow_all(struct kvm
*kvm
);
627 /* flush memory translations pointing to 'slot' */
628 void kvm_arch_flush_shadow_memslot(struct kvm
*kvm
,
629 struct kvm_memory_slot
*slot
);
631 int gfn_to_page_many_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
,
632 struct page
**pages
, int nr_pages
);
634 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
635 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
636 unsigned long gfn_to_hva_prot(struct kvm
*kvm
, gfn_t gfn
, bool *writable
);
637 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
638 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
640 void kvm_release_page_clean(struct page
*page
);
641 void kvm_release_page_dirty(struct page
*page
);
642 void kvm_set_page_accessed(struct page
*page
);
644 kvm_pfn_t
gfn_to_pfn_atomic(struct kvm
*kvm
, gfn_t gfn
);
645 kvm_pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
646 kvm_pfn_t
gfn_to_pfn_prot(struct kvm
*kvm
, gfn_t gfn
, bool write_fault
,
648 kvm_pfn_t
gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
649 kvm_pfn_t
gfn_to_pfn_memslot_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
);
650 kvm_pfn_t
__gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
651 bool atomic
, bool *async
, bool write_fault
,
654 void kvm_release_pfn_clean(kvm_pfn_t pfn
);
655 void kvm_set_pfn_dirty(kvm_pfn_t pfn
);
656 void kvm_set_pfn_accessed(kvm_pfn_t pfn
);
657 void kvm_get_pfn(kvm_pfn_t pfn
);
659 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
661 int kvm_read_guest_atomic(struct kvm
*kvm
, gpa_t gpa
, void *data
,
663 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
664 int kvm_read_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
665 void *data
, unsigned long len
);
666 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
667 int offset
, int len
);
668 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
670 int kvm_write_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
671 void *data
, unsigned long len
);
672 int kvm_write_guest_offset_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
673 void *data
, int offset
, unsigned long len
);
674 int kvm_gfn_to_hva_cache_init(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
675 gpa_t gpa
, unsigned long len
);
676 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
677 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
678 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
679 bool kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
680 unsigned long kvm_host_page_size(struct kvm
*kvm
, gfn_t gfn
);
681 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
683 struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
);
684 struct kvm_memory_slot
*kvm_vcpu_gfn_to_memslot(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
685 kvm_pfn_t
kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
686 kvm_pfn_t
kvm_vcpu_gfn_to_pfn(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
687 struct page
*kvm_vcpu_gfn_to_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
688 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
689 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu
*vcpu
, gfn_t gfn
, bool *writable
);
690 int kvm_vcpu_read_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, void *data
, int offset
,
692 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
694 int kvm_vcpu_read_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
696 int kvm_vcpu_write_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, const void *data
,
697 int offset
, int len
);
698 int kvm_vcpu_write_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, const void *data
,
700 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
702 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
703 void kvm_arch_vcpu_blocking(struct kvm_vcpu
*vcpu
);
704 void kvm_arch_vcpu_unblocking(struct kvm_vcpu
*vcpu
);
705 bool kvm_vcpu_wake_up(struct kvm_vcpu
*vcpu
);
706 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
707 int kvm_vcpu_yield_to(struct kvm_vcpu
*target
);
708 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
);
709 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
710 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
712 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
713 void kvm_reload_remote_mmus(struct kvm
*kvm
);
714 bool kvm_make_all_cpus_request(struct kvm
*kvm
, unsigned int req
);
716 long kvm_arch_dev_ioctl(struct file
*filp
,
717 unsigned int ioctl
, unsigned long arg
);
718 long kvm_arch_vcpu_ioctl(struct file
*filp
,
719 unsigned int ioctl
, unsigned long arg
);
720 int kvm_arch_vcpu_fault(struct kvm_vcpu
*vcpu
, struct vm_fault
*vmf
);
722 int kvm_vm_ioctl_check_extension(struct kvm
*kvm
, long ext
);
724 int kvm_get_dirty_log(struct kvm
*kvm
,
725 struct kvm_dirty_log
*log
, int *is_dirty
);
727 int kvm_get_dirty_log_protect(struct kvm
*kvm
,
728 struct kvm_dirty_log
*log
, bool *is_dirty
);
730 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm
*kvm
,
731 struct kvm_memory_slot
*slot
,
735 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
736 struct kvm_dirty_log
*log
);
738 int kvm_vm_ioctl_irq_line(struct kvm
*kvm
, struct kvm_irq_level
*irq_level
,
740 long kvm_arch_vm_ioctl(struct file
*filp
,
741 unsigned int ioctl
, unsigned long arg
);
743 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
744 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
746 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
747 struct kvm_translation
*tr
);
749 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
750 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
751 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
752 struct kvm_sregs
*sregs
);
753 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
754 struct kvm_sregs
*sregs
);
755 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
756 struct kvm_mp_state
*mp_state
);
757 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
758 struct kvm_mp_state
*mp_state
);
759 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
760 struct kvm_guest_debug
*dbg
);
761 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
763 int kvm_arch_init(void *opaque
);
764 void kvm_arch_exit(void);
766 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
);
767 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
);
769 void kvm_arch_sched_in(struct kvm_vcpu
*vcpu
, int cpu
);
771 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
);
772 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
773 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
774 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
);
775 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
);
776 void kvm_arch_vcpu_postcreate(struct kvm_vcpu
*vcpu
);
777 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
779 bool kvm_arch_has_vcpu_debugfs(void);
780 int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu
*vcpu
);
782 int kvm_arch_hardware_enable(void);
783 void kvm_arch_hardware_disable(void);
784 int kvm_arch_hardware_setup(void);
785 void kvm_arch_hardware_unsetup(void);
786 void kvm_arch_check_processor_compat(void *rtn
);
787 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
788 int kvm_arch_vcpu_should_kick(struct kvm_vcpu
*vcpu
);
790 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
791 static inline struct kvm
*kvm_arch_alloc_vm(void)
793 return kzalloc(sizeof(struct kvm
), GFP_KERNEL
);
796 static inline void kvm_arch_free_vm(struct kvm
*kvm
)
802 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
803 void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
);
804 void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
);
805 bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
);
807 static inline void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
)
811 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
)
815 static inline bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
)
820 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
821 void kvm_arch_start_assignment(struct kvm
*kvm
);
822 void kvm_arch_end_assignment(struct kvm
*kvm
);
823 bool kvm_arch_has_assigned_device(struct kvm
*kvm
);
825 static inline void kvm_arch_start_assignment(struct kvm
*kvm
)
829 static inline void kvm_arch_end_assignment(struct kvm
*kvm
)
833 static inline bool kvm_arch_has_assigned_device(struct kvm
*kvm
)
839 static inline struct swait_queue_head
*kvm_arch_vcpu_wq(struct kvm_vcpu
*vcpu
)
841 #ifdef __KVM_HAVE_ARCH_WQP
842 return vcpu
->arch
.wqp
;
848 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
850 * returns true if the virtual interrupt controller is initialized and
851 * ready to accept virtual IRQ. On some architectures the virtual interrupt
852 * controller is dynamically instantiated and this is not always true.
854 bool kvm_arch_intc_initialized(struct kvm
*kvm
);
856 static inline bool kvm_arch_intc_initialized(struct kvm
*kvm
)
862 int kvm_arch_init_vm(struct kvm
*kvm
, unsigned long type
);
863 void kvm_arch_destroy_vm(struct kvm
*kvm
);
864 void kvm_arch_sync_events(struct kvm
*kvm
);
866 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
867 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
869 bool kvm_is_reserved_pfn(kvm_pfn_t pfn
);
871 struct kvm_irq_ack_notifier
{
872 struct hlist_node link
;
874 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
877 int kvm_irq_map_gsi(struct kvm
*kvm
,
878 struct kvm_kernel_irq_routing_entry
*entries
, int gsi
);
879 int kvm_irq_map_chip_pin(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
881 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
,
883 int kvm_set_msi(struct kvm_kernel_irq_routing_entry
*irq_entry
, struct kvm
*kvm
,
884 int irq_source_id
, int level
, bool line_status
);
885 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry
*e
,
886 struct kvm
*kvm
, int irq_source_id
,
887 int level
, bool line_status
);
888 bool kvm_irq_has_notifier(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
889 void kvm_notify_acked_gsi(struct kvm
*kvm
, int gsi
);
890 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
891 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
892 struct kvm_irq_ack_notifier
*kian
);
893 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
894 struct kvm_irq_ack_notifier
*kian
);
895 int kvm_request_irq_source_id(struct kvm
*kvm
);
896 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
899 * search_memslots() and __gfn_to_memslot() are here because they are
900 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
901 * gfn_to_memslot() itself isn't here as an inline because that would
902 * bloat other code too much.
904 static inline struct kvm_memory_slot
*
905 search_memslots(struct kvm_memslots
*slots
, gfn_t gfn
)
907 int start
= 0, end
= slots
->used_slots
;
908 int slot
= atomic_read(&slots
->lru_slot
);
909 struct kvm_memory_slot
*memslots
= slots
->memslots
;
911 if (gfn
>= memslots
[slot
].base_gfn
&&
912 gfn
< memslots
[slot
].base_gfn
+ memslots
[slot
].npages
)
913 return &memslots
[slot
];
915 while (start
< end
) {
916 slot
= start
+ (end
- start
) / 2;
918 if (gfn
>= memslots
[slot
].base_gfn
)
924 if (gfn
>= memslots
[start
].base_gfn
&&
925 gfn
< memslots
[start
].base_gfn
+ memslots
[start
].npages
) {
926 atomic_set(&slots
->lru_slot
, start
);
927 return &memslots
[start
];
933 static inline struct kvm_memory_slot
*
934 __gfn_to_memslot(struct kvm_memslots
*slots
, gfn_t gfn
)
936 return search_memslots(slots
, gfn
);
939 static inline unsigned long
940 __gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
)
942 return slot
->userspace_addr
+ (gfn
- slot
->base_gfn
) * PAGE_SIZE
;
945 static inline int memslot_id(struct kvm
*kvm
, gfn_t gfn
)
947 return gfn_to_memslot(kvm
, gfn
)->id
;
951 hva_to_gfn_memslot(unsigned long hva
, struct kvm_memory_slot
*slot
)
953 gfn_t gfn_offset
= (hva
- slot
->userspace_addr
) >> PAGE_SHIFT
;
955 return slot
->base_gfn
+ gfn_offset
;
958 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
960 return (gpa_t
)gfn
<< PAGE_SHIFT
;
963 static inline gfn_t
gpa_to_gfn(gpa_t gpa
)
965 return (gfn_t
)(gpa
>> PAGE_SHIFT
);
968 static inline hpa_t
pfn_to_hpa(kvm_pfn_t pfn
)
970 return (hpa_t
)pfn
<< PAGE_SHIFT
;
973 static inline bool kvm_is_error_gpa(struct kvm
*kvm
, gpa_t gpa
)
975 unsigned long hva
= gfn_to_hva(kvm
, gpa_to_gfn(gpa
));
977 return kvm_is_error_hva(hva
);
985 struct kvm_stat_data
{
990 struct kvm_stats_debugfs_item
{
993 enum kvm_stat_kind kind
;
995 extern struct kvm_stats_debugfs_item debugfs_entries
[];
996 extern struct dentry
*kvm_debugfs_dir
;
998 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
999 static inline int mmu_notifier_retry(struct kvm
*kvm
, unsigned long mmu_seq
)
1001 if (unlikely(kvm
->mmu_notifier_count
))
1004 * Ensure the read of mmu_notifier_count happens before the read
1005 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1006 * mmu_notifier_invalidate_range_end to make sure that the caller
1007 * either sees the old (non-zero) value of mmu_notifier_count or
1008 * the new (incremented) value of mmu_notifier_seq.
1009 * PowerPC Book3s HV KVM calls this under a per-page lock
1010 * rather than under kvm->mmu_lock, for scalability, so
1011 * can't rely on kvm->mmu_lock to keep things ordered.
1014 if (kvm
->mmu_notifier_seq
!= mmu_seq
)
1020 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1023 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1024 #elif defined(CONFIG_ARM64)
1025 #define KVM_MAX_IRQ_ROUTES 4096
1027 #define KVM_MAX_IRQ_ROUTES 1024
1030 bool kvm_arch_can_set_irq_routing(struct kvm
*kvm
);
1031 int kvm_set_irq_routing(struct kvm
*kvm
,
1032 const struct kvm_irq_routing_entry
*entries
,
1035 int kvm_set_routing_entry(struct kvm
*kvm
,
1036 struct kvm_kernel_irq_routing_entry
*e
,
1037 const struct kvm_irq_routing_entry
*ue
);
1038 void kvm_free_irq_routing(struct kvm
*kvm
);
1042 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
1046 int kvm_send_userspace_msi(struct kvm
*kvm
, struct kvm_msi
*msi
);
1048 #ifdef CONFIG_HAVE_KVM_EVENTFD
1050 void kvm_eventfd_init(struct kvm
*kvm
);
1051 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
1053 #ifdef CONFIG_HAVE_KVM_IRQFD
1054 int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
);
1055 void kvm_irqfd_release(struct kvm
*kvm
);
1056 void kvm_irq_routing_update(struct kvm
*);
1058 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1063 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1068 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
1070 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1075 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1077 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1078 static inline void kvm_irq_routing_update(struct kvm
*kvm
)
1082 void kvm_arch_irq_routing_update(struct kvm
*kvm
);
1084 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
1089 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1091 static inline void kvm_make_request(int req
, struct kvm_vcpu
*vcpu
)
1094 * Ensure the rest of the request is published to kvm_check_request's
1095 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1098 set_bit(req
& KVM_REQUEST_MASK
, &vcpu
->requests
);
1101 static inline bool kvm_test_request(int req
, struct kvm_vcpu
*vcpu
)
1103 return test_bit(req
& KVM_REQUEST_MASK
, &vcpu
->requests
);
1106 static inline void kvm_clear_request(int req
, struct kvm_vcpu
*vcpu
)
1108 clear_bit(req
& KVM_REQUEST_MASK
, &vcpu
->requests
);
1111 static inline bool kvm_check_request(int req
, struct kvm_vcpu
*vcpu
)
1113 if (kvm_test_request(req
, vcpu
)) {
1114 kvm_clear_request(req
, vcpu
);
1117 * Ensure the rest of the request is visible to kvm_check_request's
1118 * caller. Paired with the smp_wmb in kvm_make_request.
1120 smp_mb__after_atomic();
1127 extern bool kvm_rebooting
;
1129 extern unsigned int halt_poll_ns
;
1130 extern unsigned int halt_poll_ns_grow
;
1131 extern unsigned int halt_poll_ns_shrink
;
1134 struct kvm_device_ops
*ops
;
1137 struct list_head vm_node
;
1140 /* create, destroy, and name are mandatory */
1141 struct kvm_device_ops
{
1145 * create is called holding kvm->lock and any operations not suitable
1146 * to do while holding the lock should be deferred to init (see
1149 int (*create
)(struct kvm_device
*dev
, u32 type
);
1152 * init is called after create if create is successful and is called
1153 * outside of holding kvm->lock.
1155 void (*init
)(struct kvm_device
*dev
);
1158 * Destroy is responsible for freeing dev.
1160 * Destroy may be called before or after destructors are called
1161 * on emulated I/O regions, depending on whether a reference is
1162 * held by a vcpu or other kvm component that gets destroyed
1163 * after the emulated I/O.
1165 void (*destroy
)(struct kvm_device
*dev
);
1167 int (*set_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1168 int (*get_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1169 int (*has_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1170 long (*ioctl
)(struct kvm_device
*dev
, unsigned int ioctl
,
1174 void kvm_device_get(struct kvm_device
*dev
);
1175 void kvm_device_put(struct kvm_device
*dev
);
1176 struct kvm_device
*kvm_device_from_filp(struct file
*filp
);
1177 int kvm_register_device_ops(struct kvm_device_ops
*ops
, u32 type
);
1178 void kvm_unregister_device_ops(u32 type
);
1180 extern struct kvm_device_ops kvm_mpic_ops
;
1181 extern struct kvm_device_ops kvm_arm_vgic_v2_ops
;
1182 extern struct kvm_device_ops kvm_arm_vgic_v3_ops
;
1184 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1186 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1188 vcpu
->spin_loop
.in_spin_loop
= val
;
1190 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1192 vcpu
->spin_loop
.dy_eligible
= val
;
1195 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1197 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1201 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1204 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1206 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1207 bool kvm_arch_has_irq_bypass(void);
1208 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer
*,
1209 struct irq_bypass_producer
*);
1210 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer
*,
1211 struct irq_bypass_producer
*);
1212 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer
*);
1213 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer
*);
1214 int kvm_arch_update_irqfd_routing(struct kvm
*kvm
, unsigned int host_irq
,
1215 uint32_t guest_irq
, bool set
);
1216 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1218 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1219 /* If we wakeup during the poll time, was it a sucessful poll? */
1220 static inline bool vcpu_valid_wakeup(struct kvm_vcpu
*vcpu
)
1222 return vcpu
->valid_wakeup
;
1226 static inline bool vcpu_valid_wakeup(struct kvm_vcpu
*vcpu
)
1230 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */