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 <asm/signal.h>
30 #include <linux/kvm.h>
31 #include <linux/kvm_para.h>
33 #include <linux/kvm_types.h>
35 #include <asm/kvm_host.h>
38 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
39 * in kvm, other bits are visible for userspace which are defined in
40 * include/linux/kvm_h.
42 #define KVM_MEMSLOT_INVALID (1UL << 16)
43 #define KVM_MEMSLOT_INCOHERENT (1UL << 17)
45 /* Two fragments for cross MMIO pages. */
46 #define KVM_MAX_MMIO_FRAGMENTS 2
48 #ifndef KVM_ADDRESS_SPACE_NUM
49 #define KVM_ADDRESS_SPACE_NUM 1
53 * For the normal pfn, the highest 12 bits should be zero,
54 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
55 * mask bit 63 to indicate the noslot pfn.
57 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
58 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
59 #define KVM_PFN_NOSLOT (0x1ULL << 63)
61 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
62 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
63 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
66 * error pfns indicate that the gfn is in slot but faild to
67 * translate it to pfn on host.
69 static inline bool is_error_pfn(pfn_t pfn
)
71 return !!(pfn
& KVM_PFN_ERR_MASK
);
75 * error_noslot pfns indicate that the gfn can not be
76 * translated to pfn - it is not in slot or failed to
77 * translate it to pfn.
79 static inline bool is_error_noslot_pfn(pfn_t pfn
)
81 return !!(pfn
& KVM_PFN_ERR_NOSLOT_MASK
);
84 /* noslot pfn indicates that the gfn is not in slot. */
85 static inline bool is_noslot_pfn(pfn_t pfn
)
87 return pfn
== KVM_PFN_NOSLOT
;
91 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
92 * provide own defines and kvm_is_error_hva
94 #ifndef KVM_HVA_ERR_BAD
96 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
97 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
99 static inline bool kvm_is_error_hva(unsigned long addr
)
101 return addr
>= PAGE_OFFSET
;
106 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
108 static inline bool is_error_page(struct page
*page
)
114 * vcpu->requests bit members
116 #define KVM_REQ_TLB_FLUSH 0
117 #define KVM_REQ_MIGRATE_TIMER 1
118 #define KVM_REQ_REPORT_TPR_ACCESS 2
119 #define KVM_REQ_MMU_RELOAD 3
120 #define KVM_REQ_TRIPLE_FAULT 4
121 #define KVM_REQ_PENDING_TIMER 5
122 #define KVM_REQ_UNHALT 6
123 #define KVM_REQ_MMU_SYNC 7
124 #define KVM_REQ_CLOCK_UPDATE 8
125 #define KVM_REQ_KICK 9
126 #define KVM_REQ_DEACTIVATE_FPU 10
127 #define KVM_REQ_EVENT 11
128 #define KVM_REQ_APF_HALT 12
129 #define KVM_REQ_STEAL_UPDATE 13
130 #define KVM_REQ_NMI 14
131 #define KVM_REQ_PMU 15
132 #define KVM_REQ_PMI 16
133 #define KVM_REQ_WATCHDOG 17
134 #define KVM_REQ_MASTERCLOCK_UPDATE 18
135 #define KVM_REQ_MCLOCK_INPROGRESS 19
136 #define KVM_REQ_EPR_EXIT 20
137 #define KVM_REQ_SCAN_IOAPIC 21
138 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 22
139 #define KVM_REQ_ENABLE_IBS 23
140 #define KVM_REQ_DISABLE_IBS 24
141 #define KVM_REQ_APIC_PAGE_RELOAD 25
142 #define KVM_REQ_SMI 26
143 #define KVM_REQ_HV_CRASH 27
144 #define KVM_REQ_IOAPIC_EOI_EXIT 28
145 #define KVM_REQ_HV_RESET 29
146 #define KVM_REQ_HV_EXIT 30
147 #define KVM_REQ_HV_STIMER 31
149 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
150 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
152 extern struct kmem_cache
*kvm_vcpu_cache
;
154 extern spinlock_t kvm_lock
;
155 extern struct list_head vm_list
;
157 struct kvm_io_range
{
160 struct kvm_io_device
*dev
;
163 #define NR_IOBUS_DEVS 1000
168 struct kvm_io_range range
[];
174 KVM_VIRTIO_CCW_NOTIFY_BUS
,
179 int kvm_io_bus_write(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
180 int len
, const void *val
);
181 int kvm_io_bus_write_cookie(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
,
182 gpa_t addr
, int len
, const void *val
, long cookie
);
183 int kvm_io_bus_read(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
185 int kvm_io_bus_register_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
,
186 int len
, struct kvm_io_device
*dev
);
187 void kvm_io_bus_unregister_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
188 struct kvm_io_device
*dev
);
190 #ifdef CONFIG_KVM_ASYNC_PF
191 struct kvm_async_pf
{
192 struct work_struct work
;
193 struct list_head link
;
194 struct list_head queue
;
195 struct kvm_vcpu
*vcpu
;
196 struct mm_struct
*mm
;
199 struct kvm_arch_async_pf arch
;
203 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu
*vcpu
);
204 void kvm_check_async_pf_completion(struct kvm_vcpu
*vcpu
);
205 int kvm_setup_async_pf(struct kvm_vcpu
*vcpu
, gva_t gva
, unsigned long hva
,
206 struct kvm_arch_async_pf
*arch
);
207 int kvm_async_pf_wakeup_all(struct kvm_vcpu
*vcpu
);
214 READING_SHADOW_PAGE_TABLES
,
218 * Sometimes a large or cross-page mmio needs to be broken up into separate
219 * exits for userspace servicing.
221 struct kvm_mmio_fragment
{
229 #ifdef CONFIG_PREEMPT_NOTIFIERS
230 struct preempt_notifier preempt_notifier
;
236 unsigned long requests
;
237 unsigned long guest_debug
;
240 struct list_head blocked_vcpu_list
;
246 int guest_fpu_loaded
, guest_xcr0_loaded
;
247 unsigned char fpu_counter
;
248 wait_queue_head_t wq
;
252 struct kvm_vcpu_stat stat
;
253 unsigned int halt_poll_ns
;
255 #ifdef CONFIG_HAS_IOMEM
257 int mmio_read_completed
;
259 int mmio_cur_fragment
;
260 int mmio_nr_fragments
;
261 struct kvm_mmio_fragment mmio_fragments
[KVM_MAX_MMIO_FRAGMENTS
];
264 #ifdef CONFIG_KVM_ASYNC_PF
267 struct list_head queue
;
268 struct list_head done
;
273 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
275 * Cpu relax intercept or pause loop exit optimization
276 * in_spin_loop: set when a vcpu does a pause loop exit
277 * or cpu relax intercepted.
278 * dy_eligible: indicates whether vcpu is eligible for directed yield.
286 struct kvm_vcpu_arch arch
;
289 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu
*vcpu
)
291 return cmpxchg(&vcpu
->mode
, IN_GUEST_MODE
, EXITING_GUEST_MODE
);
295 * Some of the bitops functions do not support too long bitmaps.
296 * This number must be determined not to exceed such limits.
298 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
300 struct kvm_memory_slot
{
302 unsigned long npages
;
303 unsigned long *dirty_bitmap
;
304 struct kvm_arch_memory_slot arch
;
305 unsigned long userspace_addr
;
310 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot
*memslot
)
312 return ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
315 struct kvm_s390_adapter_int
{
328 struct kvm_kernel_irq_routing_entry
{
331 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
332 struct kvm
*kvm
, int irq_source_id
, int level
,
340 struct kvm_s390_adapter_int adapter
;
341 struct kvm_hv_sint hv_sint
;
343 struct hlist_node link
;
346 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
347 struct kvm_irq_routing_table
{
348 int chip
[KVM_NR_IRQCHIPS
][KVM_IRQCHIP_NUM_PINS
];
351 * Array indexed by gsi. Each entry contains list of irq chips
352 * the gsi is connected to.
354 struct hlist_head map
[0];
358 #ifndef KVM_PRIVATE_MEM_SLOTS
359 #define KVM_PRIVATE_MEM_SLOTS 0
362 #ifndef KVM_MEM_SLOTS_NUM
363 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
366 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
367 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu
*vcpu
)
375 * memslots are not sorted by id anymore, please use id_to_memslot()
376 * to get the memslot by its id.
378 struct kvm_memslots
{
380 struct kvm_memory_slot memslots
[KVM_MEM_SLOTS_NUM
];
381 /* The mapping table from slot id to the index in memslots[]. */
382 short id_to_index
[KVM_MEM_SLOTS_NUM
];
389 struct mutex slots_lock
;
390 struct mm_struct
*mm
; /* userspace tied to this vm */
391 struct kvm_memslots
*memslots
[KVM_ADDRESS_SPACE_NUM
];
392 struct srcu_struct srcu
;
393 struct srcu_struct irq_srcu
;
394 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
395 atomic_t online_vcpus
;
396 int last_boosted_vcpu
;
397 struct list_head vm_list
;
399 struct kvm_io_bus
*buses
[KVM_NR_BUSES
];
400 #ifdef CONFIG_HAVE_KVM_EVENTFD
403 struct list_head items
;
404 struct list_head resampler_list
;
405 struct mutex resampler_lock
;
407 struct list_head ioeventfds
;
409 struct kvm_vm_stat stat
;
410 struct kvm_arch arch
;
411 atomic_t users_count
;
412 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
413 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
414 spinlock_t ring_lock
;
415 struct list_head coalesced_zones
;
418 struct mutex irq_lock
;
419 #ifdef CONFIG_HAVE_KVM_IRQCHIP
421 * Update side is protected by irq_lock.
423 struct kvm_irq_routing_table __rcu
*irq_routing
;
425 #ifdef CONFIG_HAVE_KVM_IRQFD
426 struct hlist_head irq_ack_notifier_list
;
429 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
430 struct mmu_notifier mmu_notifier
;
431 unsigned long mmu_notifier_seq
;
432 long mmu_notifier_count
;
435 struct list_head devices
;
438 #define kvm_err(fmt, ...) \
439 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
440 #define kvm_info(fmt, ...) \
441 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
442 #define kvm_debug(fmt, ...) \
443 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
444 #define kvm_pr_unimpl(fmt, ...) \
445 pr_err_ratelimited("kvm [%i]: " fmt, \
446 task_tgid_nr(current), ## __VA_ARGS__)
448 /* The guest did something we don't support. */
449 #define vcpu_unimpl(vcpu, fmt, ...) \
450 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
452 #define vcpu_debug(vcpu, fmt, ...) \
453 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
454 #define vcpu_err(vcpu, fmt, ...) \
455 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
457 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
459 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
460 * the caller has read kvm->online_vcpus before (as is the case
461 * for kvm_for_each_vcpu, for example).
464 return kvm
->vcpus
[i
];
467 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
469 idx < atomic_read(&kvm->online_vcpus) && \
470 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
473 static inline struct kvm_vcpu
*kvm_get_vcpu_by_id(struct kvm
*kvm
, int id
)
475 struct kvm_vcpu
*vcpu
;
478 kvm_for_each_vcpu(i
, vcpu
, kvm
)
479 if (vcpu
->vcpu_id
== id
)
484 #define kvm_for_each_memslot(memslot, slots) \
485 for (memslot = &slots->memslots[0]; \
486 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
489 int kvm_vcpu_init(struct kvm_vcpu
*vcpu
, struct kvm
*kvm
, unsigned id
);
490 void kvm_vcpu_uninit(struct kvm_vcpu
*vcpu
);
492 int __must_check
vcpu_load(struct kvm_vcpu
*vcpu
);
493 void vcpu_put(struct kvm_vcpu
*vcpu
);
495 #ifdef __KVM_HAVE_IOAPIC
496 void kvm_vcpu_request_scan_ioapic(struct kvm
*kvm
);
497 void kvm_arch_post_irq_routing_update(struct kvm
*kvm
);
499 static inline void kvm_vcpu_request_scan_ioapic(struct kvm
*kvm
)
502 static inline void kvm_arch_post_irq_routing_update(struct kvm
*kvm
)
507 #ifdef CONFIG_HAVE_KVM_IRQFD
508 int kvm_irqfd_init(void);
509 void kvm_irqfd_exit(void);
511 static inline int kvm_irqfd_init(void)
516 static inline void kvm_irqfd_exit(void)
520 int kvm_init(void *opaque
, unsigned vcpu_size
, unsigned vcpu_align
,
521 struct module
*module
);
524 void kvm_get_kvm(struct kvm
*kvm
);
525 void kvm_put_kvm(struct kvm
*kvm
);
527 static inline struct kvm_memslots
*__kvm_memslots(struct kvm
*kvm
, int as_id
)
529 return rcu_dereference_check(kvm
->memslots
[as_id
],
530 srcu_read_lock_held(&kvm
->srcu
)
531 || lockdep_is_held(&kvm
->slots_lock
));
534 static inline struct kvm_memslots
*kvm_memslots(struct kvm
*kvm
)
536 return __kvm_memslots(kvm
, 0);
539 static inline struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
)
541 int as_id
= kvm_arch_vcpu_memslots_id(vcpu
);
543 return __kvm_memslots(vcpu
->kvm
, as_id
);
546 static inline struct kvm_memory_slot
*
547 id_to_memslot(struct kvm_memslots
*slots
, int id
)
549 int index
= slots
->id_to_index
[id
];
550 struct kvm_memory_slot
*slot
;
552 slot
= &slots
->memslots
[index
];
554 WARN_ON(slot
->id
!= id
);
559 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
560 * - create a new memory slot
561 * - delete an existing memory slot
562 * - modify an existing memory slot
563 * -- move it in the guest physical memory space
564 * -- just change its flags
566 * Since flags can be changed by some of these operations, the following
567 * differentiation is the best we can do for __kvm_set_memory_region():
576 int kvm_set_memory_region(struct kvm
*kvm
,
577 const struct kvm_userspace_memory_region
*mem
);
578 int __kvm_set_memory_region(struct kvm
*kvm
,
579 const struct kvm_userspace_memory_region
*mem
);
580 void kvm_arch_free_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*free
,
581 struct kvm_memory_slot
*dont
);
582 int kvm_arch_create_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*slot
,
583 unsigned long npages
);
584 void kvm_arch_memslots_updated(struct kvm
*kvm
, struct kvm_memslots
*slots
);
585 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
586 struct kvm_memory_slot
*memslot
,
587 const struct kvm_userspace_memory_region
*mem
,
588 enum kvm_mr_change change
);
589 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
590 const struct kvm_userspace_memory_region
*mem
,
591 const struct kvm_memory_slot
*old
,
592 const struct kvm_memory_slot
*new,
593 enum kvm_mr_change change
);
594 bool kvm_largepages_enabled(void);
595 void kvm_disable_largepages(void);
596 /* flush all memory translations */
597 void kvm_arch_flush_shadow_all(struct kvm
*kvm
);
598 /* flush memory translations pointing to 'slot' */
599 void kvm_arch_flush_shadow_memslot(struct kvm
*kvm
,
600 struct kvm_memory_slot
*slot
);
602 int gfn_to_page_many_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
,
603 struct page
**pages
, int nr_pages
);
605 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
606 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
607 unsigned long gfn_to_hva_prot(struct kvm
*kvm
, gfn_t gfn
, bool *writable
);
608 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
609 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
611 void kvm_release_page_clean(struct page
*page
);
612 void kvm_release_page_dirty(struct page
*page
);
613 void kvm_set_page_accessed(struct page
*page
);
615 pfn_t
gfn_to_pfn_atomic(struct kvm
*kvm
, gfn_t gfn
);
616 pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
617 pfn_t
gfn_to_pfn_prot(struct kvm
*kvm
, gfn_t gfn
, bool write_fault
,
619 pfn_t
gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
620 pfn_t
gfn_to_pfn_memslot_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
);
621 pfn_t
__gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
, bool atomic
,
622 bool *async
, bool write_fault
, bool *writable
);
624 void kvm_release_pfn_clean(pfn_t pfn
);
625 void kvm_set_pfn_dirty(pfn_t pfn
);
626 void kvm_set_pfn_accessed(pfn_t pfn
);
627 void kvm_get_pfn(pfn_t pfn
);
629 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
631 int kvm_read_guest_atomic(struct kvm
*kvm
, gpa_t gpa
, void *data
,
633 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
634 int kvm_read_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
635 void *data
, unsigned long len
);
636 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
637 int offset
, int len
);
638 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
640 int kvm_write_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
641 void *data
, unsigned long len
);
642 int kvm_gfn_to_hva_cache_init(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
643 gpa_t gpa
, unsigned long len
);
644 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
645 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
646 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
647 int kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
648 unsigned long kvm_host_page_size(struct kvm
*kvm
, gfn_t gfn
);
649 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
651 struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
);
652 struct kvm_memory_slot
*kvm_vcpu_gfn_to_memslot(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
653 pfn_t
kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
654 pfn_t
kvm_vcpu_gfn_to_pfn(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
655 struct page
*kvm_vcpu_gfn_to_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
656 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
657 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu
*vcpu
, gfn_t gfn
, bool *writable
);
658 int kvm_vcpu_read_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, void *data
, int offset
,
660 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
662 int kvm_vcpu_read_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
664 int kvm_vcpu_write_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, const void *data
,
665 int offset
, int len
);
666 int kvm_vcpu_write_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, const void *data
,
668 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
670 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
671 void kvm_arch_vcpu_blocking(struct kvm_vcpu
*vcpu
);
672 void kvm_arch_vcpu_unblocking(struct kvm_vcpu
*vcpu
);
673 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
674 int kvm_vcpu_yield_to(struct kvm_vcpu
*target
);
675 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
);
676 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
677 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
679 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
680 void kvm_reload_remote_mmus(struct kvm
*kvm
);
681 void kvm_make_mclock_inprogress_request(struct kvm
*kvm
);
682 void kvm_make_scan_ioapic_request(struct kvm
*kvm
);
683 bool kvm_make_all_cpus_request(struct kvm
*kvm
, unsigned int req
);
685 long kvm_arch_dev_ioctl(struct file
*filp
,
686 unsigned int ioctl
, unsigned long arg
);
687 long kvm_arch_vcpu_ioctl(struct file
*filp
,
688 unsigned int ioctl
, unsigned long arg
);
689 int kvm_arch_vcpu_fault(struct kvm_vcpu
*vcpu
, struct vm_fault
*vmf
);
691 int kvm_vm_ioctl_check_extension(struct kvm
*kvm
, long ext
);
693 int kvm_get_dirty_log(struct kvm
*kvm
,
694 struct kvm_dirty_log
*log
, int *is_dirty
);
696 int kvm_get_dirty_log_protect(struct kvm
*kvm
,
697 struct kvm_dirty_log
*log
, bool *is_dirty
);
699 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm
*kvm
,
700 struct kvm_memory_slot
*slot
,
704 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
705 struct kvm_dirty_log
*log
);
707 int kvm_vm_ioctl_irq_line(struct kvm
*kvm
, struct kvm_irq_level
*irq_level
,
709 long kvm_arch_vm_ioctl(struct file
*filp
,
710 unsigned int ioctl
, unsigned long arg
);
712 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
713 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
715 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
716 struct kvm_translation
*tr
);
718 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
719 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
720 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
721 struct kvm_sregs
*sregs
);
722 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
723 struct kvm_sregs
*sregs
);
724 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
725 struct kvm_mp_state
*mp_state
);
726 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
727 struct kvm_mp_state
*mp_state
);
728 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
729 struct kvm_guest_debug
*dbg
);
730 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
732 int kvm_arch_init(void *opaque
);
733 void kvm_arch_exit(void);
735 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
);
736 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
);
738 void kvm_arch_sched_in(struct kvm_vcpu
*vcpu
, int cpu
);
740 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
);
741 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
742 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
743 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
);
744 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
);
745 void kvm_arch_vcpu_postcreate(struct kvm_vcpu
*vcpu
);
746 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
748 int kvm_arch_hardware_enable(void);
749 void kvm_arch_hardware_disable(void);
750 int kvm_arch_hardware_setup(void);
751 void kvm_arch_hardware_unsetup(void);
752 void kvm_arch_check_processor_compat(void *rtn
);
753 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
754 int kvm_arch_vcpu_should_kick(struct kvm_vcpu
*vcpu
);
756 void *kvm_kvzalloc(unsigned long size
);
758 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
759 static inline struct kvm
*kvm_arch_alloc_vm(void)
761 return kzalloc(sizeof(struct kvm
), GFP_KERNEL
);
764 static inline void kvm_arch_free_vm(struct kvm
*kvm
)
770 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
771 void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
);
772 void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
);
773 bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
);
775 static inline void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
)
779 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
)
783 static inline bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
)
788 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
789 void kvm_arch_start_assignment(struct kvm
*kvm
);
790 void kvm_arch_end_assignment(struct kvm
*kvm
);
791 bool kvm_arch_has_assigned_device(struct kvm
*kvm
);
793 static inline void kvm_arch_start_assignment(struct kvm
*kvm
)
797 static inline void kvm_arch_end_assignment(struct kvm
*kvm
)
801 static inline bool kvm_arch_has_assigned_device(struct kvm
*kvm
)
807 static inline wait_queue_head_t
*kvm_arch_vcpu_wq(struct kvm_vcpu
*vcpu
)
809 #ifdef __KVM_HAVE_ARCH_WQP
810 return vcpu
->arch
.wqp
;
816 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
818 * returns true if the virtual interrupt controller is initialized and
819 * ready to accept virtual IRQ. On some architectures the virtual interrupt
820 * controller is dynamically instantiated and this is not always true.
822 bool kvm_arch_intc_initialized(struct kvm
*kvm
);
824 static inline bool kvm_arch_intc_initialized(struct kvm
*kvm
)
830 int kvm_arch_init_vm(struct kvm
*kvm
, unsigned long type
);
831 void kvm_arch_destroy_vm(struct kvm
*kvm
);
832 void kvm_arch_sync_events(struct kvm
*kvm
);
834 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
835 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
837 bool kvm_is_reserved_pfn(pfn_t pfn
);
839 struct kvm_irq_ack_notifier
{
840 struct hlist_node link
;
842 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
845 int kvm_irq_map_gsi(struct kvm
*kvm
,
846 struct kvm_kernel_irq_routing_entry
*entries
, int gsi
);
847 int kvm_irq_map_chip_pin(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
849 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
,
851 int kvm_set_msi(struct kvm_kernel_irq_routing_entry
*irq_entry
, struct kvm
*kvm
,
852 int irq_source_id
, int level
, bool line_status
);
853 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry
*e
,
854 struct kvm
*kvm
, int irq_source_id
,
855 int level
, bool line_status
);
856 bool kvm_irq_has_notifier(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
857 void kvm_notify_acked_gsi(struct kvm
*kvm
, int gsi
);
858 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
859 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
860 struct kvm_irq_ack_notifier
*kian
);
861 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
862 struct kvm_irq_ack_notifier
*kian
);
863 int kvm_request_irq_source_id(struct kvm
*kvm
);
864 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
866 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
867 int kvm_iommu_map_pages(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
868 void kvm_iommu_unmap_pages(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
870 static inline int kvm_iommu_map_pages(struct kvm
*kvm
,
871 struct kvm_memory_slot
*slot
)
876 static inline void kvm_iommu_unmap_pages(struct kvm
*kvm
,
877 struct kvm_memory_slot
*slot
)
882 /* must be called with irqs disabled */
883 static inline void __kvm_guest_enter(void)
886 /* KVM does not hold any references to rcu protected data when it
887 * switches CPU into a guest mode. In fact switching to a guest mode
888 * is very similar to exiting to userspace from rcu point of view. In
889 * addition CPU may stay in a guest mode for quite a long time (up to
890 * one time slice). Lets treat guest mode as quiescent state, just like
891 * we do with user-mode execution.
893 if (!context_tracking_cpu_is_enabled())
894 rcu_virt_note_context_switch(smp_processor_id());
897 /* must be called with irqs disabled */
898 static inline void __kvm_guest_exit(void)
903 static inline void kvm_guest_enter(void)
907 local_irq_save(flags
);
909 local_irq_restore(flags
);
912 static inline void kvm_guest_exit(void)
916 local_irq_save(flags
);
918 local_irq_restore(flags
);
922 * search_memslots() and __gfn_to_memslot() are here because they are
923 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
924 * gfn_to_memslot() itself isn't here as an inline because that would
925 * bloat other code too much.
927 static inline struct kvm_memory_slot
*
928 search_memslots(struct kvm_memslots
*slots
, gfn_t gfn
)
930 int start
= 0, end
= slots
->used_slots
;
931 int slot
= atomic_read(&slots
->lru_slot
);
932 struct kvm_memory_slot
*memslots
= slots
->memslots
;
934 if (gfn
>= memslots
[slot
].base_gfn
&&
935 gfn
< memslots
[slot
].base_gfn
+ memslots
[slot
].npages
)
936 return &memslots
[slot
];
938 while (start
< end
) {
939 slot
= start
+ (end
- start
) / 2;
941 if (gfn
>= memslots
[slot
].base_gfn
)
947 if (gfn
>= memslots
[start
].base_gfn
&&
948 gfn
< memslots
[start
].base_gfn
+ memslots
[start
].npages
) {
949 atomic_set(&slots
->lru_slot
, start
);
950 return &memslots
[start
];
956 static inline struct kvm_memory_slot
*
957 __gfn_to_memslot(struct kvm_memslots
*slots
, gfn_t gfn
)
959 return search_memslots(slots
, gfn
);
962 static inline unsigned long
963 __gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
)
965 return slot
->userspace_addr
+ (gfn
- slot
->base_gfn
) * PAGE_SIZE
;
968 static inline int memslot_id(struct kvm
*kvm
, gfn_t gfn
)
970 return gfn_to_memslot(kvm
, gfn
)->id
;
974 hva_to_gfn_memslot(unsigned long hva
, struct kvm_memory_slot
*slot
)
976 gfn_t gfn_offset
= (hva
- slot
->userspace_addr
) >> PAGE_SHIFT
;
978 return slot
->base_gfn
+ gfn_offset
;
981 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
983 return (gpa_t
)gfn
<< PAGE_SHIFT
;
986 static inline gfn_t
gpa_to_gfn(gpa_t gpa
)
988 return (gfn_t
)(gpa
>> PAGE_SHIFT
);
991 static inline hpa_t
pfn_to_hpa(pfn_t pfn
)
993 return (hpa_t
)pfn
<< PAGE_SHIFT
;
996 static inline bool kvm_is_error_gpa(struct kvm
*kvm
, gpa_t gpa
)
998 unsigned long hva
= gfn_to_hva(kvm
, gpa_to_gfn(gpa
));
1000 return kvm_is_error_hva(hva
);
1003 static inline void kvm_migrate_timers(struct kvm_vcpu
*vcpu
)
1005 set_bit(KVM_REQ_MIGRATE_TIMER
, &vcpu
->requests
);
1008 enum kvm_stat_kind
{
1013 struct kvm_stats_debugfs_item
{
1016 enum kvm_stat_kind kind
;
1017 struct dentry
*dentry
;
1019 extern struct kvm_stats_debugfs_item debugfs_entries
[];
1020 extern struct dentry
*kvm_debugfs_dir
;
1022 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1023 static inline int mmu_notifier_retry(struct kvm
*kvm
, unsigned long mmu_seq
)
1025 if (unlikely(kvm
->mmu_notifier_count
))
1028 * Ensure the read of mmu_notifier_count happens before the read
1029 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1030 * mmu_notifier_invalidate_range_end to make sure that the caller
1031 * either sees the old (non-zero) value of mmu_notifier_count or
1032 * the new (incremented) value of mmu_notifier_seq.
1033 * PowerPC Book3s HV KVM calls this under a per-page lock
1034 * rather than under kvm->mmu_lock, for scalability, so
1035 * can't rely on kvm->mmu_lock to keep things ordered.
1038 if (kvm
->mmu_notifier_seq
!= mmu_seq
)
1044 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1047 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1049 #define KVM_MAX_IRQ_ROUTES 1024
1052 int kvm_setup_default_irq_routing(struct kvm
*kvm
);
1053 int kvm_setup_empty_irq_routing(struct kvm
*kvm
);
1054 int kvm_set_irq_routing(struct kvm
*kvm
,
1055 const struct kvm_irq_routing_entry
*entries
,
1058 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry
*e
,
1059 const struct kvm_irq_routing_entry
*ue
);
1060 void kvm_free_irq_routing(struct kvm
*kvm
);
1064 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
1068 int kvm_send_userspace_msi(struct kvm
*kvm
, struct kvm_msi
*msi
);
1070 #ifdef CONFIG_HAVE_KVM_EVENTFD
1072 void kvm_eventfd_init(struct kvm
*kvm
);
1073 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
1075 #ifdef CONFIG_HAVE_KVM_IRQFD
1076 int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
);
1077 void kvm_irqfd_release(struct kvm
*kvm
);
1078 void kvm_irq_routing_update(struct kvm
*);
1080 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1085 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1090 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
1092 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1097 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1099 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1100 static inline void kvm_irq_routing_update(struct kvm
*kvm
)
1104 void kvm_arch_irq_routing_update(struct kvm
*kvm
);
1106 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
1111 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1113 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
1114 bool kvm_vcpu_compatible(struct kvm_vcpu
*vcpu
);
1116 static inline bool kvm_vcpu_compatible(struct kvm_vcpu
*vcpu
) { return true; }
1119 static inline void kvm_make_request(int req
, struct kvm_vcpu
*vcpu
)
1121 set_bit(req
, &vcpu
->requests
);
1124 static inline bool kvm_check_request(int req
, struct kvm_vcpu
*vcpu
)
1126 if (test_bit(req
, &vcpu
->requests
)) {
1127 clear_bit(req
, &vcpu
->requests
);
1134 extern bool kvm_rebooting
;
1137 struct kvm_device_ops
*ops
;
1140 struct list_head vm_node
;
1143 /* create, destroy, and name are mandatory */
1144 struct kvm_device_ops
{
1146 int (*create
)(struct kvm_device
*dev
, u32 type
);
1149 * Destroy is responsible for freeing dev.
1151 * Destroy may be called before or after destructors are called
1152 * on emulated I/O regions, depending on whether a reference is
1153 * held by a vcpu or other kvm component that gets destroyed
1154 * after the emulated I/O.
1156 void (*destroy
)(struct kvm_device
*dev
);
1158 int (*set_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1159 int (*get_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1160 int (*has_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1161 long (*ioctl
)(struct kvm_device
*dev
, unsigned int ioctl
,
1165 void kvm_device_get(struct kvm_device
*dev
);
1166 void kvm_device_put(struct kvm_device
*dev
);
1167 struct kvm_device
*kvm_device_from_filp(struct file
*filp
);
1168 int kvm_register_device_ops(struct kvm_device_ops
*ops
, u32 type
);
1169 void kvm_unregister_device_ops(u32 type
);
1171 extern struct kvm_device_ops kvm_mpic_ops
;
1172 extern struct kvm_device_ops kvm_xics_ops
;
1173 extern struct kvm_device_ops kvm_arm_vgic_v2_ops
;
1174 extern struct kvm_device_ops kvm_arm_vgic_v3_ops
;
1176 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1178 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1180 vcpu
->spin_loop
.in_spin_loop
= val
;
1182 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1184 vcpu
->spin_loop
.dy_eligible
= val
;
1187 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1189 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1193 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1196 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1198 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1199 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer
*,
1200 struct irq_bypass_producer
*);
1201 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer
*,
1202 struct irq_bypass_producer
*);
1203 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer
*);
1204 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer
*);
1205 int kvm_arch_update_irqfd_routing(struct kvm
*kvm
, unsigned int host_irq
,
1206 uint32_t guest_irq
, bool set
);
1207 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */