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 <asm/signal.h>
29 #include <linux/kvm.h>
30 #include <linux/kvm_para.h>
32 #include <linux/kvm_types.h>
34 #include <asm/kvm_host.h>
37 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
38 * in kvm, other bits are visible for userspace which are defined in
39 * include/linux/kvm_h.
41 #define KVM_MEMSLOT_INVALID (1UL << 16)
42 #define KVM_MEMSLOT_INCOHERENT (1UL << 17)
44 /* Two fragments for cross MMIO pages. */
45 #define KVM_MAX_MMIO_FRAGMENTS 2
47 #ifndef KVM_ADDRESS_SPACE_NUM
48 #define KVM_ADDRESS_SPACE_NUM 1
52 * For the normal pfn, the highest 12 bits should be zero,
53 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
54 * mask bit 63 to indicate the noslot pfn.
56 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
57 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
58 #define KVM_PFN_NOSLOT (0x1ULL << 63)
60 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
61 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
62 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
65 * error pfns indicate that the gfn is in slot but faild to
66 * translate it to pfn on host.
68 static inline bool is_error_pfn(pfn_t pfn
)
70 return !!(pfn
& KVM_PFN_ERR_MASK
);
74 * error_noslot pfns indicate that the gfn can not be
75 * translated to pfn - it is not in slot or failed to
76 * translate it to pfn.
78 static inline bool is_error_noslot_pfn(pfn_t pfn
)
80 return !!(pfn
& KVM_PFN_ERR_NOSLOT_MASK
);
83 /* noslot pfn indicates that the gfn is not in slot. */
84 static inline bool is_noslot_pfn(pfn_t pfn
)
86 return pfn
== KVM_PFN_NOSLOT
;
90 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
91 * provide own defines and kvm_is_error_hva
93 #ifndef KVM_HVA_ERR_BAD
95 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
96 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
98 static inline bool kvm_is_error_hva(unsigned long addr
)
100 return addr
>= PAGE_OFFSET
;
105 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
107 static inline bool is_error_page(struct page
*page
)
113 * vcpu->requests bit members
115 #define KVM_REQ_TLB_FLUSH 0
116 #define KVM_REQ_MIGRATE_TIMER 1
117 #define KVM_REQ_REPORT_TPR_ACCESS 2
118 #define KVM_REQ_MMU_RELOAD 3
119 #define KVM_REQ_TRIPLE_FAULT 4
120 #define KVM_REQ_PENDING_TIMER 5
121 #define KVM_REQ_UNHALT 6
122 #define KVM_REQ_MMU_SYNC 7
123 #define KVM_REQ_CLOCK_UPDATE 8
124 #define KVM_REQ_KICK 9
125 #define KVM_REQ_DEACTIVATE_FPU 10
126 #define KVM_REQ_EVENT 11
127 #define KVM_REQ_APF_HALT 12
128 #define KVM_REQ_STEAL_UPDATE 13
129 #define KVM_REQ_NMI 14
130 #define KVM_REQ_PMU 15
131 #define KVM_REQ_PMI 16
132 #define KVM_REQ_WATCHDOG 17
133 #define KVM_REQ_MASTERCLOCK_UPDATE 18
134 #define KVM_REQ_MCLOCK_INPROGRESS 19
135 #define KVM_REQ_EPR_EXIT 20
136 #define KVM_REQ_SCAN_IOAPIC 21
137 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 22
138 #define KVM_REQ_ENABLE_IBS 23
139 #define KVM_REQ_DISABLE_IBS 24
140 #define KVM_REQ_APIC_PAGE_RELOAD 25
141 #define KVM_REQ_SMI 26
142 #define KVM_REQ_HV_CRASH 27
144 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
145 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
147 extern struct kmem_cache
*kvm_vcpu_cache
;
149 extern spinlock_t kvm_lock
;
150 extern struct list_head vm_list
;
152 struct kvm_io_range
{
155 struct kvm_io_device
*dev
;
158 #define NR_IOBUS_DEVS 1000
163 struct kvm_io_range range
[];
169 KVM_VIRTIO_CCW_NOTIFY_BUS
,
174 int kvm_io_bus_write(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
175 int len
, const void *val
);
176 int kvm_io_bus_write_cookie(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
,
177 gpa_t addr
, int len
, const void *val
, long cookie
);
178 int kvm_io_bus_read(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
180 int kvm_io_bus_register_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
,
181 int len
, struct kvm_io_device
*dev
);
182 int kvm_io_bus_unregister_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
183 struct kvm_io_device
*dev
);
185 #ifdef CONFIG_KVM_ASYNC_PF
186 struct kvm_async_pf
{
187 struct work_struct work
;
188 struct list_head link
;
189 struct list_head queue
;
190 struct kvm_vcpu
*vcpu
;
191 struct mm_struct
*mm
;
194 struct kvm_arch_async_pf arch
;
198 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu
*vcpu
);
199 void kvm_check_async_pf_completion(struct kvm_vcpu
*vcpu
);
200 int kvm_setup_async_pf(struct kvm_vcpu
*vcpu
, gva_t gva
, unsigned long hva
,
201 struct kvm_arch_async_pf
*arch
);
202 int kvm_async_pf_wakeup_all(struct kvm_vcpu
*vcpu
);
209 READING_SHADOW_PAGE_TABLES
,
213 * Sometimes a large or cross-page mmio needs to be broken up into separate
214 * exits for userspace servicing.
216 struct kvm_mmio_fragment
{
224 #ifdef CONFIG_PREEMPT_NOTIFIERS
225 struct preempt_notifier preempt_notifier
;
231 unsigned long requests
;
232 unsigned long guest_debug
;
238 int guest_fpu_loaded
, guest_xcr0_loaded
;
239 unsigned char fpu_counter
;
240 wait_queue_head_t wq
;
244 struct kvm_vcpu_stat stat
;
245 unsigned int halt_poll_ns
;
247 #ifdef CONFIG_HAS_IOMEM
249 int mmio_read_completed
;
251 int mmio_cur_fragment
;
252 int mmio_nr_fragments
;
253 struct kvm_mmio_fragment mmio_fragments
[KVM_MAX_MMIO_FRAGMENTS
];
256 #ifdef CONFIG_KVM_ASYNC_PF
259 struct list_head queue
;
260 struct list_head done
;
265 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
267 * Cpu relax intercept or pause loop exit optimization
268 * in_spin_loop: set when a vcpu does a pause loop exit
269 * or cpu relax intercepted.
270 * dy_eligible: indicates whether vcpu is eligible for directed yield.
278 struct kvm_vcpu_arch arch
;
281 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu
*vcpu
)
283 return cmpxchg(&vcpu
->mode
, IN_GUEST_MODE
, EXITING_GUEST_MODE
);
287 * Some of the bitops functions do not support too long bitmaps.
288 * This number must be determined not to exceed such limits.
290 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
292 struct kvm_memory_slot
{
294 unsigned long npages
;
295 unsigned long *dirty_bitmap
;
296 struct kvm_arch_memory_slot arch
;
297 unsigned long userspace_addr
;
302 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot
*memslot
)
304 return ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
307 struct kvm_s390_adapter_int
{
315 struct kvm_kernel_irq_routing_entry
{
318 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
319 struct kvm
*kvm
, int irq_source_id
, int level
,
327 struct kvm_s390_adapter_int adapter
;
329 struct hlist_node link
;
332 #ifndef KVM_PRIVATE_MEM_SLOTS
333 #define KVM_PRIVATE_MEM_SLOTS 0
336 #ifndef KVM_MEM_SLOTS_NUM
337 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
340 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
341 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu
*vcpu
)
349 * memslots are not sorted by id anymore, please use id_to_memslot()
350 * to get the memslot by its id.
352 struct kvm_memslots
{
354 struct kvm_memory_slot memslots
[KVM_MEM_SLOTS_NUM
];
355 /* The mapping table from slot id to the index in memslots[]. */
356 short id_to_index
[KVM_MEM_SLOTS_NUM
];
363 struct mutex slots_lock
;
364 struct mm_struct
*mm
; /* userspace tied to this vm */
365 struct kvm_memslots
*memslots
[KVM_ADDRESS_SPACE_NUM
];
366 struct srcu_struct srcu
;
367 struct srcu_struct irq_srcu
;
368 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
369 atomic_t online_vcpus
;
370 int last_boosted_vcpu
;
371 struct list_head vm_list
;
373 struct kvm_io_bus
*buses
[KVM_NR_BUSES
];
374 #ifdef CONFIG_HAVE_KVM_EVENTFD
377 struct list_head items
;
378 struct list_head resampler_list
;
379 struct mutex resampler_lock
;
381 struct list_head ioeventfds
;
383 struct kvm_vm_stat stat
;
384 struct kvm_arch arch
;
385 atomic_t users_count
;
386 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
387 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
388 spinlock_t ring_lock
;
389 struct list_head coalesced_zones
;
392 struct mutex irq_lock
;
393 #ifdef CONFIG_HAVE_KVM_IRQCHIP
395 * Update side is protected by irq_lock.
397 struct kvm_irq_routing_table __rcu
*irq_routing
;
399 #ifdef CONFIG_HAVE_KVM_IRQFD
400 struct hlist_head irq_ack_notifier_list
;
403 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
404 struct mmu_notifier mmu_notifier
;
405 unsigned long mmu_notifier_seq
;
406 long mmu_notifier_count
;
409 struct list_head devices
;
412 #define kvm_err(fmt, ...) \
413 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
414 #define kvm_info(fmt, ...) \
415 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
416 #define kvm_debug(fmt, ...) \
417 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
418 #define kvm_pr_unimpl(fmt, ...) \
419 pr_err_ratelimited("kvm [%i]: " fmt, \
420 task_tgid_nr(current), ## __VA_ARGS__)
422 /* The guest did something we don't support. */
423 #define vcpu_unimpl(vcpu, fmt, ...) \
424 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
426 #define vcpu_debug(vcpu, fmt, ...) \
427 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
429 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
431 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
432 * the caller has read kvm->online_vcpus before (as is the case
433 * for kvm_for_each_vcpu, for example).
436 return kvm
->vcpus
[i
];
439 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
441 idx < atomic_read(&kvm->online_vcpus) && \
442 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
445 #define kvm_for_each_memslot(memslot, slots) \
446 for (memslot = &slots->memslots[0]; \
447 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
450 int kvm_vcpu_init(struct kvm_vcpu
*vcpu
, struct kvm
*kvm
, unsigned id
);
451 void kvm_vcpu_uninit(struct kvm_vcpu
*vcpu
);
453 int __must_check
vcpu_load(struct kvm_vcpu
*vcpu
);
454 void vcpu_put(struct kvm_vcpu
*vcpu
);
456 #ifdef __KVM_HAVE_IOAPIC
457 void kvm_vcpu_request_scan_ioapic(struct kvm
*kvm
);
459 static inline void kvm_vcpu_request_scan_ioapic(struct kvm
*kvm
)
464 #ifdef CONFIG_HAVE_KVM_IRQFD
465 int kvm_irqfd_init(void);
466 void kvm_irqfd_exit(void);
468 static inline int kvm_irqfd_init(void)
473 static inline void kvm_irqfd_exit(void)
477 int kvm_init(void *opaque
, unsigned vcpu_size
, unsigned vcpu_align
,
478 struct module
*module
);
481 void kvm_get_kvm(struct kvm
*kvm
);
482 void kvm_put_kvm(struct kvm
*kvm
);
484 static inline struct kvm_memslots
*__kvm_memslots(struct kvm
*kvm
, int as_id
)
486 return rcu_dereference_check(kvm
->memslots
[as_id
],
487 srcu_read_lock_held(&kvm
->srcu
)
488 || lockdep_is_held(&kvm
->slots_lock
));
491 static inline struct kvm_memslots
*kvm_memslots(struct kvm
*kvm
)
493 return __kvm_memslots(kvm
, 0);
496 static inline struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
)
498 int as_id
= kvm_arch_vcpu_memslots_id(vcpu
);
500 return __kvm_memslots(vcpu
->kvm
, as_id
);
503 static inline struct kvm_memory_slot
*
504 id_to_memslot(struct kvm_memslots
*slots
, int id
)
506 int index
= slots
->id_to_index
[id
];
507 struct kvm_memory_slot
*slot
;
509 slot
= &slots
->memslots
[index
];
511 WARN_ON(slot
->id
!= id
);
516 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
517 * - create a new memory slot
518 * - delete an existing memory slot
519 * - modify an existing memory slot
520 * -- move it in the guest physical memory space
521 * -- just change its flags
523 * Since flags can be changed by some of these operations, the following
524 * differentiation is the best we can do for __kvm_set_memory_region():
533 int kvm_set_memory_region(struct kvm
*kvm
,
534 const struct kvm_userspace_memory_region
*mem
);
535 int __kvm_set_memory_region(struct kvm
*kvm
,
536 const struct kvm_userspace_memory_region
*mem
);
537 void kvm_arch_free_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*free
,
538 struct kvm_memory_slot
*dont
);
539 int kvm_arch_create_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*slot
,
540 unsigned long npages
);
541 void kvm_arch_memslots_updated(struct kvm
*kvm
, struct kvm_memslots
*slots
);
542 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
543 struct kvm_memory_slot
*memslot
,
544 const struct kvm_userspace_memory_region
*mem
,
545 enum kvm_mr_change change
);
546 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
547 const struct kvm_userspace_memory_region
*mem
,
548 const struct kvm_memory_slot
*old
,
549 const struct kvm_memory_slot
*new,
550 enum kvm_mr_change change
);
551 bool kvm_largepages_enabled(void);
552 void kvm_disable_largepages(void);
553 /* flush all memory translations */
554 void kvm_arch_flush_shadow_all(struct kvm
*kvm
);
555 /* flush memory translations pointing to 'slot' */
556 void kvm_arch_flush_shadow_memslot(struct kvm
*kvm
,
557 struct kvm_memory_slot
*slot
);
559 int gfn_to_page_many_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
,
560 struct page
**pages
, int nr_pages
);
562 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
563 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
564 unsigned long gfn_to_hva_prot(struct kvm
*kvm
, gfn_t gfn
, bool *writable
);
565 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
566 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
568 void kvm_release_page_clean(struct page
*page
);
569 void kvm_release_page_dirty(struct page
*page
);
570 void kvm_set_page_accessed(struct page
*page
);
572 pfn_t
gfn_to_pfn_atomic(struct kvm
*kvm
, gfn_t gfn
);
573 pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
574 pfn_t
gfn_to_pfn_prot(struct kvm
*kvm
, gfn_t gfn
, bool write_fault
,
576 pfn_t
gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
577 pfn_t
gfn_to_pfn_memslot_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
);
578 pfn_t
__gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
, bool atomic
,
579 bool *async
, bool write_fault
, bool *writable
);
581 void kvm_release_pfn_clean(pfn_t pfn
);
582 void kvm_set_pfn_dirty(pfn_t pfn
);
583 void kvm_set_pfn_accessed(pfn_t pfn
);
584 void kvm_get_pfn(pfn_t pfn
);
586 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
588 int kvm_read_guest_atomic(struct kvm
*kvm
, gpa_t gpa
, void *data
,
590 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
591 int kvm_read_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
592 void *data
, unsigned long len
);
593 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
594 int offset
, int len
);
595 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
597 int kvm_write_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
598 void *data
, unsigned long len
);
599 int kvm_gfn_to_hva_cache_init(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
600 gpa_t gpa
, unsigned long len
);
601 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
602 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
603 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
604 int kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
605 unsigned long kvm_host_page_size(struct kvm
*kvm
, gfn_t gfn
);
606 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
608 struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
);
609 struct kvm_memory_slot
*kvm_vcpu_gfn_to_memslot(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
610 pfn_t
kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
611 pfn_t
kvm_vcpu_gfn_to_pfn(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
612 struct page
*kvm_vcpu_gfn_to_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
613 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
614 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu
*vcpu
, gfn_t gfn
, bool *writable
);
615 int kvm_vcpu_read_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, void *data
, int offset
,
617 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
619 int kvm_vcpu_read_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
621 int kvm_vcpu_write_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, const void *data
,
622 int offset
, int len
);
623 int kvm_vcpu_write_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, const void *data
,
625 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
627 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
628 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
629 int kvm_vcpu_yield_to(struct kvm_vcpu
*target
);
630 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
);
631 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
632 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
634 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
635 void kvm_reload_remote_mmus(struct kvm
*kvm
);
636 void kvm_make_mclock_inprogress_request(struct kvm
*kvm
);
637 void kvm_make_scan_ioapic_request(struct kvm
*kvm
);
638 bool kvm_make_all_cpus_request(struct kvm
*kvm
, unsigned int req
);
640 long kvm_arch_dev_ioctl(struct file
*filp
,
641 unsigned int ioctl
, unsigned long arg
);
642 long kvm_arch_vcpu_ioctl(struct file
*filp
,
643 unsigned int ioctl
, unsigned long arg
);
644 int kvm_arch_vcpu_fault(struct kvm_vcpu
*vcpu
, struct vm_fault
*vmf
);
646 int kvm_vm_ioctl_check_extension(struct kvm
*kvm
, long ext
);
648 int kvm_get_dirty_log(struct kvm
*kvm
,
649 struct kvm_dirty_log
*log
, int *is_dirty
);
651 int kvm_get_dirty_log_protect(struct kvm
*kvm
,
652 struct kvm_dirty_log
*log
, bool *is_dirty
);
654 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm
*kvm
,
655 struct kvm_memory_slot
*slot
,
659 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
660 struct kvm_dirty_log
*log
);
662 int kvm_vm_ioctl_irq_line(struct kvm
*kvm
, struct kvm_irq_level
*irq_level
,
664 long kvm_arch_vm_ioctl(struct file
*filp
,
665 unsigned int ioctl
, unsigned long arg
);
667 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
668 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
670 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
671 struct kvm_translation
*tr
);
673 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
674 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
675 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
676 struct kvm_sregs
*sregs
);
677 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
678 struct kvm_sregs
*sregs
);
679 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
680 struct kvm_mp_state
*mp_state
);
681 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
682 struct kvm_mp_state
*mp_state
);
683 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
684 struct kvm_guest_debug
*dbg
);
685 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
687 int kvm_arch_init(void *opaque
);
688 void kvm_arch_exit(void);
690 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
);
691 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
);
693 void kvm_arch_sched_in(struct kvm_vcpu
*vcpu
, int cpu
);
695 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
);
696 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
697 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
698 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
);
699 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
);
700 void kvm_arch_vcpu_postcreate(struct kvm_vcpu
*vcpu
);
701 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
703 int kvm_arch_hardware_enable(void);
704 void kvm_arch_hardware_disable(void);
705 int kvm_arch_hardware_setup(void);
706 void kvm_arch_hardware_unsetup(void);
707 void kvm_arch_check_processor_compat(void *rtn
);
708 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
709 int kvm_arch_vcpu_should_kick(struct kvm_vcpu
*vcpu
);
711 void *kvm_kvzalloc(unsigned long size
);
713 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
714 static inline struct kvm
*kvm_arch_alloc_vm(void)
716 return kzalloc(sizeof(struct kvm
), GFP_KERNEL
);
719 static inline void kvm_arch_free_vm(struct kvm
*kvm
)
725 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
726 void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
);
727 void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
);
728 bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
);
730 static inline void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
)
734 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
)
738 static inline bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
)
743 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
744 void kvm_arch_start_assignment(struct kvm
*kvm
);
745 void kvm_arch_end_assignment(struct kvm
*kvm
);
746 bool kvm_arch_has_assigned_device(struct kvm
*kvm
);
748 static inline void kvm_arch_start_assignment(struct kvm
*kvm
)
752 static inline void kvm_arch_end_assignment(struct kvm
*kvm
)
756 static inline bool kvm_arch_has_assigned_device(struct kvm
*kvm
)
762 static inline wait_queue_head_t
*kvm_arch_vcpu_wq(struct kvm_vcpu
*vcpu
)
764 #ifdef __KVM_HAVE_ARCH_WQP
765 return vcpu
->arch
.wqp
;
771 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
773 * returns true if the virtual interrupt controller is initialized and
774 * ready to accept virtual IRQ. On some architectures the virtual interrupt
775 * controller is dynamically instantiated and this is not always true.
777 bool kvm_arch_intc_initialized(struct kvm
*kvm
);
779 static inline bool kvm_arch_intc_initialized(struct kvm
*kvm
)
785 int kvm_arch_init_vm(struct kvm
*kvm
, unsigned long type
);
786 void kvm_arch_destroy_vm(struct kvm
*kvm
);
787 void kvm_arch_sync_events(struct kvm
*kvm
);
789 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
790 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
792 bool kvm_is_reserved_pfn(pfn_t pfn
);
794 struct kvm_irq_ack_notifier
{
795 struct hlist_node link
;
797 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
800 int kvm_irq_map_gsi(struct kvm
*kvm
,
801 struct kvm_kernel_irq_routing_entry
*entries
, int gsi
);
802 int kvm_irq_map_chip_pin(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
804 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
,
806 int kvm_set_irq_inatomic(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
);
807 int kvm_set_msi(struct kvm_kernel_irq_routing_entry
*irq_entry
, struct kvm
*kvm
,
808 int irq_source_id
, int level
, bool line_status
);
809 bool kvm_irq_has_notifier(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
810 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
811 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
812 struct kvm_irq_ack_notifier
*kian
);
813 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
814 struct kvm_irq_ack_notifier
*kian
);
815 int kvm_request_irq_source_id(struct kvm
*kvm
);
816 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
818 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
819 int kvm_iommu_map_pages(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
820 void kvm_iommu_unmap_pages(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
822 static inline int kvm_iommu_map_pages(struct kvm
*kvm
,
823 struct kvm_memory_slot
*slot
)
828 static inline void kvm_iommu_unmap_pages(struct kvm
*kvm
,
829 struct kvm_memory_slot
*slot
)
834 /* must be called with irqs disabled */
835 static inline void __kvm_guest_enter(void)
838 /* KVM does not hold any references to rcu protected data when it
839 * switches CPU into a guest mode. In fact switching to a guest mode
840 * is very similar to exiting to userspace from rcu point of view. In
841 * addition CPU may stay in a guest mode for quite a long time (up to
842 * one time slice). Lets treat guest mode as quiescent state, just like
843 * we do with user-mode execution.
845 if (!context_tracking_cpu_is_enabled())
846 rcu_virt_note_context_switch(smp_processor_id());
849 /* must be called with irqs disabled */
850 static inline void __kvm_guest_exit(void)
855 static inline void kvm_guest_enter(void)
859 local_irq_save(flags
);
861 local_irq_restore(flags
);
864 static inline void kvm_guest_exit(void)
868 local_irq_save(flags
);
870 local_irq_restore(flags
);
874 * search_memslots() and __gfn_to_memslot() are here because they are
875 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
876 * gfn_to_memslot() itself isn't here as an inline because that would
877 * bloat other code too much.
879 static inline struct kvm_memory_slot
*
880 search_memslots(struct kvm_memslots
*slots
, gfn_t gfn
)
882 int start
= 0, end
= slots
->used_slots
;
883 int slot
= atomic_read(&slots
->lru_slot
);
884 struct kvm_memory_slot
*memslots
= slots
->memslots
;
886 if (gfn
>= memslots
[slot
].base_gfn
&&
887 gfn
< memslots
[slot
].base_gfn
+ memslots
[slot
].npages
)
888 return &memslots
[slot
];
890 while (start
< end
) {
891 slot
= start
+ (end
- start
) / 2;
893 if (gfn
>= memslots
[slot
].base_gfn
)
899 if (gfn
>= memslots
[start
].base_gfn
&&
900 gfn
< memslots
[start
].base_gfn
+ memslots
[start
].npages
) {
901 atomic_set(&slots
->lru_slot
, start
);
902 return &memslots
[start
];
908 static inline struct kvm_memory_slot
*
909 __gfn_to_memslot(struct kvm_memslots
*slots
, gfn_t gfn
)
911 return search_memslots(slots
, gfn
);
914 static inline unsigned long
915 __gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
)
917 return slot
->userspace_addr
+ (gfn
- slot
->base_gfn
) * PAGE_SIZE
;
920 static inline int memslot_id(struct kvm
*kvm
, gfn_t gfn
)
922 return gfn_to_memslot(kvm
, gfn
)->id
;
926 hva_to_gfn_memslot(unsigned long hva
, struct kvm_memory_slot
*slot
)
928 gfn_t gfn_offset
= (hva
- slot
->userspace_addr
) >> PAGE_SHIFT
;
930 return slot
->base_gfn
+ gfn_offset
;
933 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
935 return (gpa_t
)gfn
<< PAGE_SHIFT
;
938 static inline gfn_t
gpa_to_gfn(gpa_t gpa
)
940 return (gfn_t
)(gpa
>> PAGE_SHIFT
);
943 static inline hpa_t
pfn_to_hpa(pfn_t pfn
)
945 return (hpa_t
)pfn
<< PAGE_SHIFT
;
948 static inline bool kvm_is_error_gpa(struct kvm
*kvm
, gpa_t gpa
)
950 unsigned long hva
= gfn_to_hva(kvm
, gpa_to_gfn(gpa
));
952 return kvm_is_error_hva(hva
);
955 static inline void kvm_migrate_timers(struct kvm_vcpu
*vcpu
)
957 set_bit(KVM_REQ_MIGRATE_TIMER
, &vcpu
->requests
);
965 struct kvm_stats_debugfs_item
{
968 enum kvm_stat_kind kind
;
969 struct dentry
*dentry
;
971 extern struct kvm_stats_debugfs_item debugfs_entries
[];
972 extern struct dentry
*kvm_debugfs_dir
;
974 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
975 static inline int mmu_notifier_retry(struct kvm
*kvm
, unsigned long mmu_seq
)
977 if (unlikely(kvm
->mmu_notifier_count
))
980 * Ensure the read of mmu_notifier_count happens before the read
981 * of mmu_notifier_seq. This interacts with the smp_wmb() in
982 * mmu_notifier_invalidate_range_end to make sure that the caller
983 * either sees the old (non-zero) value of mmu_notifier_count or
984 * the new (incremented) value of mmu_notifier_seq.
985 * PowerPC Book3s HV KVM calls this under a per-page lock
986 * rather than under kvm->mmu_lock, for scalability, so
987 * can't rely on kvm->mmu_lock to keep things ordered.
990 if (kvm
->mmu_notifier_seq
!= mmu_seq
)
996 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
999 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1001 #define KVM_MAX_IRQ_ROUTES 1024
1004 int kvm_setup_default_irq_routing(struct kvm
*kvm
);
1005 int kvm_set_irq_routing(struct kvm
*kvm
,
1006 const struct kvm_irq_routing_entry
*entries
,
1009 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry
*e
,
1010 const struct kvm_irq_routing_entry
*ue
);
1011 void kvm_free_irq_routing(struct kvm
*kvm
);
1015 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
1019 int kvm_send_userspace_msi(struct kvm
*kvm
, struct kvm_msi
*msi
);
1021 #ifdef CONFIG_HAVE_KVM_EVENTFD
1023 void kvm_eventfd_init(struct kvm
*kvm
);
1024 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
1026 #ifdef CONFIG_HAVE_KVM_IRQFD
1027 int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
);
1028 void kvm_irqfd_release(struct kvm
*kvm
);
1029 void kvm_irq_routing_update(struct kvm
*);
1031 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1036 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1041 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
1043 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1048 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1050 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1051 static inline void kvm_irq_routing_update(struct kvm
*kvm
)
1056 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
1061 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1063 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
1064 bool kvm_vcpu_compatible(struct kvm_vcpu
*vcpu
);
1066 static inline bool kvm_vcpu_compatible(struct kvm_vcpu
*vcpu
) { return true; }
1069 static inline void kvm_make_request(int req
, struct kvm_vcpu
*vcpu
)
1071 set_bit(req
, &vcpu
->requests
);
1074 static inline bool kvm_check_request(int req
, struct kvm_vcpu
*vcpu
)
1076 if (test_bit(req
, &vcpu
->requests
)) {
1077 clear_bit(req
, &vcpu
->requests
);
1084 extern bool kvm_rebooting
;
1087 struct kvm_device_ops
*ops
;
1090 struct list_head vm_node
;
1093 /* create, destroy, and name are mandatory */
1094 struct kvm_device_ops
{
1096 int (*create
)(struct kvm_device
*dev
, u32 type
);
1099 * Destroy is responsible for freeing dev.
1101 * Destroy may be called before or after destructors are called
1102 * on emulated I/O regions, depending on whether a reference is
1103 * held by a vcpu or other kvm component that gets destroyed
1104 * after the emulated I/O.
1106 void (*destroy
)(struct kvm_device
*dev
);
1108 int (*set_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1109 int (*get_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1110 int (*has_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1111 long (*ioctl
)(struct kvm_device
*dev
, unsigned int ioctl
,
1115 void kvm_device_get(struct kvm_device
*dev
);
1116 void kvm_device_put(struct kvm_device
*dev
);
1117 struct kvm_device
*kvm_device_from_filp(struct file
*filp
);
1118 int kvm_register_device_ops(struct kvm_device_ops
*ops
, u32 type
);
1119 void kvm_unregister_device_ops(u32 type
);
1121 extern struct kvm_device_ops kvm_mpic_ops
;
1122 extern struct kvm_device_ops kvm_xics_ops
;
1123 extern struct kvm_device_ops kvm_arm_vgic_v2_ops
;
1124 extern struct kvm_device_ops kvm_arm_vgic_v3_ops
;
1126 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1128 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1130 vcpu
->spin_loop
.in_spin_loop
= val
;
1132 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1134 vcpu
->spin_loop
.dy_eligible
= val
;
1137 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1139 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1143 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1146 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */