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 #define KVM_MMIO_SIZE 8
41 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
42 * in kvm, other bits are visible for userspace which are defined in
43 * include/linux/kvm_h.
45 #define KVM_MEMSLOT_INVALID (1UL << 16)
47 /* Two fragments for cross MMIO pages. */
48 #define KVM_MAX_MMIO_FRAGMENTS 2
51 * For the normal pfn, the highest 12 bits should be zero,
52 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
53 * mask bit 63 to indicate the noslot pfn.
55 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
56 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
57 #define KVM_PFN_NOSLOT (0x1ULL << 63)
59 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
60 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
61 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
64 * error pfns indicate that the gfn is in slot but faild to
65 * translate it to pfn on host.
67 static inline bool is_error_pfn(pfn_t pfn
)
69 return !!(pfn
& KVM_PFN_ERR_MASK
);
73 * error_noslot pfns indicate that the gfn can not be
74 * translated to pfn - it is not in slot or failed to
75 * translate it to pfn.
77 static inline bool is_error_noslot_pfn(pfn_t pfn
)
79 return !!(pfn
& KVM_PFN_ERR_NOSLOT_MASK
);
82 /* noslot pfn indicates that the gfn is not in slot. */
83 static inline bool is_noslot_pfn(pfn_t pfn
)
85 return pfn
== KVM_PFN_NOSLOT
;
89 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
90 * provide own defines and kvm_is_error_hva
92 #ifndef KVM_HVA_ERR_BAD
94 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
95 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
97 static inline bool kvm_is_error_hva(unsigned long addr
)
99 return addr
>= PAGE_OFFSET
;
104 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
106 static inline bool is_error_page(struct page
*page
)
112 * vcpu->requests bit members
114 #define KVM_REQ_TLB_FLUSH 0
115 #define KVM_REQ_MIGRATE_TIMER 1
116 #define KVM_REQ_REPORT_TPR_ACCESS 2
117 #define KVM_REQ_MMU_RELOAD 3
118 #define KVM_REQ_TRIPLE_FAULT 4
119 #define KVM_REQ_PENDING_TIMER 5
120 #define KVM_REQ_UNHALT 6
121 #define KVM_REQ_MMU_SYNC 7
122 #define KVM_REQ_CLOCK_UPDATE 8
123 #define KVM_REQ_KICK 9
124 #define KVM_REQ_DEACTIVATE_FPU 10
125 #define KVM_REQ_EVENT 11
126 #define KVM_REQ_APF_HALT 12
127 #define KVM_REQ_STEAL_UPDATE 13
128 #define KVM_REQ_NMI 14
129 #define KVM_REQ_PMU 15
130 #define KVM_REQ_PMI 16
131 #define KVM_REQ_WATCHDOG 17
132 #define KVM_REQ_MASTERCLOCK_UPDATE 18
133 #define KVM_REQ_MCLOCK_INPROGRESS 19
134 #define KVM_REQ_EPR_EXIT 20
135 #define KVM_REQ_SCAN_IOAPIC 21
136 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 22
138 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
139 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
143 extern struct kmem_cache
*kvm_vcpu_cache
;
145 extern spinlock_t kvm_lock
;
146 extern struct list_head vm_list
;
148 struct kvm_io_range
{
151 struct kvm_io_device
*dev
;
154 #define NR_IOBUS_DEVS 1000
159 struct kvm_io_range range
[];
165 KVM_VIRTIO_CCW_NOTIFY_BUS
,
169 int kvm_io_bus_write(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
,
170 int len
, const void *val
);
171 int kvm_io_bus_write_cookie(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
,
172 int len
, const void *val
, long cookie
);
173 int kvm_io_bus_read(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
, int len
,
175 int kvm_io_bus_register_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
,
176 int len
, struct kvm_io_device
*dev
);
177 int kvm_io_bus_unregister_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
178 struct kvm_io_device
*dev
);
180 #ifdef CONFIG_KVM_ASYNC_PF
181 struct kvm_async_pf
{
182 struct work_struct work
;
183 struct list_head link
;
184 struct list_head queue
;
185 struct kvm_vcpu
*vcpu
;
186 struct mm_struct
*mm
;
189 struct kvm_arch_async_pf arch
;
193 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu
*vcpu
);
194 void kvm_check_async_pf_completion(struct kvm_vcpu
*vcpu
);
195 int kvm_setup_async_pf(struct kvm_vcpu
*vcpu
, gva_t gva
, unsigned long hva
,
196 struct kvm_arch_async_pf
*arch
);
197 int kvm_async_pf_wakeup_all(struct kvm_vcpu
*vcpu
);
204 READING_SHADOW_PAGE_TABLES
,
208 * Sometimes a large or cross-page mmio needs to be broken up into separate
209 * exits for userspace servicing.
211 struct kvm_mmio_fragment
{
219 #ifdef CONFIG_PREEMPT_NOTIFIERS
220 struct preempt_notifier preempt_notifier
;
226 unsigned long requests
;
227 unsigned long guest_debug
;
233 int guest_fpu_loaded
, guest_xcr0_loaded
;
234 wait_queue_head_t wq
;
238 struct kvm_vcpu_stat stat
;
240 #ifdef CONFIG_HAS_IOMEM
242 int mmio_read_completed
;
244 int mmio_cur_fragment
;
245 int mmio_nr_fragments
;
246 struct kvm_mmio_fragment mmio_fragments
[KVM_MAX_MMIO_FRAGMENTS
];
249 #ifdef CONFIG_KVM_ASYNC_PF
252 struct list_head queue
;
253 struct list_head done
;
258 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
260 * Cpu relax intercept or pause loop exit optimization
261 * in_spin_loop: set when a vcpu does a pause loop exit
262 * or cpu relax intercepted.
263 * dy_eligible: indicates whether vcpu is eligible for directed yield.
271 struct kvm_vcpu_arch arch
;
274 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu
*vcpu
)
276 return cmpxchg(&vcpu
->mode
, IN_GUEST_MODE
, EXITING_GUEST_MODE
);
280 * Some of the bitops functions do not support too long bitmaps.
281 * This number must be determined not to exceed such limits.
283 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
285 struct kvm_memory_slot
{
287 unsigned long npages
;
288 unsigned long *dirty_bitmap
;
289 struct kvm_arch_memory_slot arch
;
290 unsigned long userspace_addr
;
295 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot
*memslot
)
297 return ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
300 struct kvm_s390_adapter_int
{
308 struct kvm_kernel_irq_routing_entry
{
311 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
312 struct kvm
*kvm
, int irq_source_id
, int level
,
320 struct kvm_s390_adapter_int adapter
;
322 struct hlist_node link
;
325 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
327 struct kvm_irq_routing_table
{
328 int chip
[KVM_NR_IRQCHIPS
][KVM_IRQCHIP_NUM_PINS
];
329 struct kvm_kernel_irq_routing_entry
*rt_entries
;
332 * Array indexed by gsi. Each entry contains list of irq chips
333 * the gsi is connected to.
335 struct hlist_head map
[0];
340 struct kvm_irq_routing_table
{};
344 #ifndef KVM_PRIVATE_MEM_SLOTS
345 #define KVM_PRIVATE_MEM_SLOTS 0
348 #ifndef KVM_MEM_SLOTS_NUM
349 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
354 * memslots are not sorted by id anymore, please use id_to_memslot()
355 * to get the memslot by its id.
357 struct kvm_memslots
{
359 struct kvm_memory_slot memslots
[KVM_MEM_SLOTS_NUM
];
360 /* The mapping table from slot id to the index in memslots[]. */
361 short id_to_index
[KVM_MEM_SLOTS_NUM
];
366 struct mutex slots_lock
;
367 struct mm_struct
*mm
; /* userspace tied to this vm */
368 struct kvm_memslots
*memslots
;
369 struct srcu_struct srcu
;
370 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
373 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
374 atomic_t online_vcpus
;
375 int last_boosted_vcpu
;
376 struct list_head vm_list
;
378 struct kvm_io_bus
*buses
[KVM_NR_BUSES
];
379 #ifdef CONFIG_HAVE_KVM_EVENTFD
382 struct list_head items
;
383 struct list_head resampler_list
;
384 struct mutex resampler_lock
;
386 struct list_head ioeventfds
;
388 struct kvm_vm_stat stat
;
389 struct kvm_arch arch
;
390 atomic_t users_count
;
391 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
392 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
393 spinlock_t ring_lock
;
394 struct list_head coalesced_zones
;
397 struct mutex irq_lock
;
398 #ifdef CONFIG_HAVE_KVM_IRQCHIP
400 * Update side is protected by irq_lock and,
401 * if configured, irqfds.lock.
403 struct kvm_irq_routing_table __rcu
*irq_routing
;
404 struct hlist_head mask_notifier_list
;
405 struct hlist_head irq_ack_notifier_list
;
408 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
409 struct mmu_notifier mmu_notifier
;
410 unsigned long mmu_notifier_seq
;
411 long mmu_notifier_count
;
413 /* Protected by mmu_lock */
416 struct list_head devices
;
419 #define kvm_err(fmt, ...) \
420 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
421 #define kvm_info(fmt, ...) \
422 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
423 #define kvm_debug(fmt, ...) \
424 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
425 #define kvm_pr_unimpl(fmt, ...) \
426 pr_err_ratelimited("kvm [%i]: " fmt, \
427 task_tgid_nr(current), ## __VA_ARGS__)
429 /* The guest did something we don't support. */
430 #define vcpu_unimpl(vcpu, fmt, ...) \
431 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
433 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
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 CONFIG_HAVE_KVM_IRQ_ROUTING
457 int kvm_irqfd_init(void);
458 void kvm_irqfd_exit(void);
460 static inline int kvm_irqfd_init(void)
465 static inline void kvm_irqfd_exit(void)
469 int kvm_init(void *opaque
, unsigned vcpu_size
, unsigned vcpu_align
,
470 struct module
*module
);
473 void kvm_get_kvm(struct kvm
*kvm
);
474 void kvm_put_kvm(struct kvm
*kvm
);
476 static inline struct kvm_memslots
*kvm_memslots(struct kvm
*kvm
)
478 return rcu_dereference_check(kvm
->memslots
,
479 srcu_read_lock_held(&kvm
->srcu
)
480 || lockdep_is_held(&kvm
->slots_lock
));
483 static inline struct kvm_memory_slot
*
484 id_to_memslot(struct kvm_memslots
*slots
, int id
)
486 int index
= slots
->id_to_index
[id
];
487 struct kvm_memory_slot
*slot
;
489 slot
= &slots
->memslots
[index
];
491 WARN_ON(slot
->id
!= id
);
496 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
497 * - create a new memory slot
498 * - delete an existing memory slot
499 * - modify an existing memory slot
500 * -- move it in the guest physical memory space
501 * -- just change its flags
503 * Since flags can be changed by some of these operations, the following
504 * differentiation is the best we can do for __kvm_set_memory_region():
513 int kvm_set_memory_region(struct kvm
*kvm
,
514 struct kvm_userspace_memory_region
*mem
);
515 int __kvm_set_memory_region(struct kvm
*kvm
,
516 struct kvm_userspace_memory_region
*mem
);
517 void kvm_arch_free_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*free
,
518 struct kvm_memory_slot
*dont
);
519 int kvm_arch_create_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*slot
,
520 unsigned long npages
);
521 void kvm_arch_memslots_updated(struct kvm
*kvm
);
522 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
523 struct kvm_memory_slot
*memslot
,
524 struct kvm_userspace_memory_region
*mem
,
525 enum kvm_mr_change change
);
526 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
527 struct kvm_userspace_memory_region
*mem
,
528 const struct kvm_memory_slot
*old
,
529 enum kvm_mr_change change
);
530 bool kvm_largepages_enabled(void);
531 void kvm_disable_largepages(void);
532 /* flush all memory translations */
533 void kvm_arch_flush_shadow_all(struct kvm
*kvm
);
534 /* flush memory translations pointing to 'slot' */
535 void kvm_arch_flush_shadow_memslot(struct kvm
*kvm
,
536 struct kvm_memory_slot
*slot
);
538 int gfn_to_page_many_atomic(struct kvm
*kvm
, gfn_t gfn
, struct page
**pages
,
541 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
542 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
543 unsigned long gfn_to_hva_prot(struct kvm
*kvm
, gfn_t gfn
, bool *writable
);
544 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
545 void kvm_release_page_clean(struct page
*page
);
546 void kvm_release_page_dirty(struct page
*page
);
547 void kvm_set_page_accessed(struct page
*page
);
549 pfn_t
gfn_to_pfn_atomic(struct kvm
*kvm
, gfn_t gfn
);
550 pfn_t
gfn_to_pfn_async(struct kvm
*kvm
, gfn_t gfn
, bool *async
,
551 bool write_fault
, bool *writable
);
552 pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
553 pfn_t
gfn_to_pfn_prot(struct kvm
*kvm
, gfn_t gfn
, bool write_fault
,
555 pfn_t
gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
556 pfn_t
gfn_to_pfn_memslot_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
);
558 void kvm_release_pfn_clean(pfn_t pfn
);
559 void kvm_set_pfn_dirty(pfn_t pfn
);
560 void kvm_set_pfn_accessed(pfn_t pfn
);
561 void kvm_get_pfn(pfn_t pfn
);
563 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
565 int kvm_read_guest_atomic(struct kvm
*kvm
, gpa_t gpa
, void *data
,
567 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
568 int kvm_read_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
569 void *data
, unsigned long len
);
570 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
571 int offset
, int len
);
572 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
574 int kvm_write_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
575 void *data
, unsigned long len
);
576 int kvm_gfn_to_hva_cache_init(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
577 gpa_t gpa
, unsigned long len
);
578 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
579 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
580 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
581 int kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
582 unsigned long kvm_host_page_size(struct kvm
*kvm
, gfn_t gfn
);
583 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
585 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
586 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
587 bool kvm_vcpu_yield_to(struct kvm_vcpu
*target
);
588 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
);
589 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
590 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
592 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
593 void kvm_reload_remote_mmus(struct kvm
*kvm
);
594 void kvm_make_mclock_inprogress_request(struct kvm
*kvm
);
595 void kvm_make_scan_ioapic_request(struct kvm
*kvm
);
597 long kvm_arch_dev_ioctl(struct file
*filp
,
598 unsigned int ioctl
, unsigned long arg
);
599 long kvm_arch_vcpu_ioctl(struct file
*filp
,
600 unsigned int ioctl
, unsigned long arg
);
601 int kvm_arch_vcpu_fault(struct kvm_vcpu
*vcpu
, struct vm_fault
*vmf
);
603 int kvm_dev_ioctl_check_extension(long ext
);
605 int kvm_get_dirty_log(struct kvm
*kvm
,
606 struct kvm_dirty_log
*log
, int *is_dirty
);
607 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
608 struct kvm_dirty_log
*log
);
610 int kvm_vm_ioctl_irq_line(struct kvm
*kvm
, struct kvm_irq_level
*irq_level
,
612 long kvm_arch_vm_ioctl(struct file
*filp
,
613 unsigned int ioctl
, unsigned long arg
);
615 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
616 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
618 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
619 struct kvm_translation
*tr
);
621 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
622 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
623 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
624 struct kvm_sregs
*sregs
);
625 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
626 struct kvm_sregs
*sregs
);
627 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
628 struct kvm_mp_state
*mp_state
);
629 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
630 struct kvm_mp_state
*mp_state
);
631 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
632 struct kvm_guest_debug
*dbg
);
633 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
635 int kvm_arch_init(void *opaque
);
636 void kvm_arch_exit(void);
638 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
);
639 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
);
641 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
);
642 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
643 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
644 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
);
645 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
);
646 int kvm_arch_vcpu_postcreate(struct kvm_vcpu
*vcpu
);
647 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
649 int kvm_arch_hardware_enable(void *garbage
);
650 void kvm_arch_hardware_disable(void *garbage
);
651 int kvm_arch_hardware_setup(void);
652 void kvm_arch_hardware_unsetup(void);
653 void kvm_arch_check_processor_compat(void *rtn
);
654 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
655 int kvm_arch_vcpu_should_kick(struct kvm_vcpu
*vcpu
);
657 void *kvm_kvzalloc(unsigned long size
);
658 void kvm_kvfree(const void *addr
);
660 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
661 static inline struct kvm
*kvm_arch_alloc_vm(void)
663 return kzalloc(sizeof(struct kvm
), GFP_KERNEL
);
666 static inline void kvm_arch_free_vm(struct kvm
*kvm
)
672 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
673 void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
);
674 void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
);
675 bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
);
677 static inline void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
)
681 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
)
685 static inline bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
)
691 static inline wait_queue_head_t
*kvm_arch_vcpu_wq(struct kvm_vcpu
*vcpu
)
693 #ifdef __KVM_HAVE_ARCH_WQP
694 return vcpu
->arch
.wqp
;
700 int kvm_arch_init_vm(struct kvm
*kvm
, unsigned long type
);
701 void kvm_arch_destroy_vm(struct kvm
*kvm
);
702 void kvm_arch_sync_events(struct kvm
*kvm
);
704 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
705 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
707 bool kvm_is_mmio_pfn(pfn_t pfn
);
709 struct kvm_irq_ack_notifier
{
710 struct hlist_node link
;
712 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
715 struct kvm_assigned_dev_kernel
{
716 struct kvm_irq_ack_notifier ack_notifier
;
717 struct list_head list
;
722 unsigned int entries_nr
;
724 bool host_irq_disabled
;
726 struct msix_entry
*host_msix_entries
;
728 struct msix_entry
*guest_msix_entries
;
729 unsigned long irq_requested_type
;
734 spinlock_t intx_lock
;
735 spinlock_t intx_mask_lock
;
737 struct pci_saved_state
*pci_saved_state
;
740 struct kvm_irq_mask_notifier
{
741 void (*func
)(struct kvm_irq_mask_notifier
*kimn
, bool masked
);
743 struct hlist_node link
;
746 void kvm_register_irq_mask_notifier(struct kvm
*kvm
, int irq
,
747 struct kvm_irq_mask_notifier
*kimn
);
748 void kvm_unregister_irq_mask_notifier(struct kvm
*kvm
, int irq
,
749 struct kvm_irq_mask_notifier
*kimn
);
750 void kvm_fire_mask_notifiers(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
,
753 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
,
755 int kvm_set_irq_inatomic(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
);
756 int kvm_set_msi(struct kvm_kernel_irq_routing_entry
*irq_entry
, struct kvm
*kvm
,
757 int irq_source_id
, int level
, bool line_status
);
758 bool kvm_irq_has_notifier(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
759 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
760 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
761 struct kvm_irq_ack_notifier
*kian
);
762 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
763 struct kvm_irq_ack_notifier
*kian
);
764 int kvm_request_irq_source_id(struct kvm
*kvm
);
765 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
767 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
768 int kvm_iommu_map_pages(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
769 void kvm_iommu_unmap_pages(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
770 int kvm_iommu_map_guest(struct kvm
*kvm
);
771 int kvm_iommu_unmap_guest(struct kvm
*kvm
);
772 int kvm_assign_device(struct kvm
*kvm
,
773 struct kvm_assigned_dev_kernel
*assigned_dev
);
774 int kvm_deassign_device(struct kvm
*kvm
,
775 struct kvm_assigned_dev_kernel
*assigned_dev
);
777 static inline int kvm_iommu_map_pages(struct kvm
*kvm
,
778 struct kvm_memory_slot
*slot
)
783 static inline void kvm_iommu_unmap_pages(struct kvm
*kvm
,
784 struct kvm_memory_slot
*slot
)
788 static inline int kvm_iommu_unmap_guest(struct kvm
*kvm
)
794 static inline void kvm_guest_enter(void)
798 BUG_ON(preemptible());
800 local_irq_save(flags
);
802 local_irq_restore(flags
);
804 /* KVM does not hold any references to rcu protected data when it
805 * switches CPU into a guest mode. In fact switching to a guest mode
806 * is very similar to exiting to userspace from rcu point of view. In
807 * addition CPU may stay in a guest mode for quite a long time (up to
808 * one time slice). Lets treat guest mode as quiescent state, just like
809 * we do with user-mode execution.
811 rcu_virt_note_context_switch(smp_processor_id());
814 static inline void kvm_guest_exit(void)
818 local_irq_save(flags
);
820 local_irq_restore(flags
);
824 * search_memslots() and __gfn_to_memslot() are here because they are
825 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
826 * gfn_to_memslot() itself isn't here as an inline because that would
827 * bloat other code too much.
829 static inline struct kvm_memory_slot
*
830 search_memslots(struct kvm_memslots
*slots
, gfn_t gfn
)
832 struct kvm_memory_slot
*memslot
;
834 kvm_for_each_memslot(memslot
, slots
)
835 if (gfn
>= memslot
->base_gfn
&&
836 gfn
< memslot
->base_gfn
+ memslot
->npages
)
842 static inline struct kvm_memory_slot
*
843 __gfn_to_memslot(struct kvm_memslots
*slots
, gfn_t gfn
)
845 return search_memslots(slots
, gfn
);
848 static inline unsigned long
849 __gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
)
851 return slot
->userspace_addr
+ (gfn
- slot
->base_gfn
) * PAGE_SIZE
;
854 static inline int memslot_id(struct kvm
*kvm
, gfn_t gfn
)
856 return gfn_to_memslot(kvm
, gfn
)->id
;
860 hva_to_gfn_memslot(unsigned long hva
, struct kvm_memory_slot
*slot
)
862 gfn_t gfn_offset
= (hva
- slot
->userspace_addr
) >> PAGE_SHIFT
;
864 return slot
->base_gfn
+ gfn_offset
;
867 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
869 return (gpa_t
)gfn
<< PAGE_SHIFT
;
872 static inline gfn_t
gpa_to_gfn(gpa_t gpa
)
874 return (gfn_t
)(gpa
>> PAGE_SHIFT
);
877 static inline hpa_t
pfn_to_hpa(pfn_t pfn
)
879 return (hpa_t
)pfn
<< PAGE_SHIFT
;
882 static inline void kvm_migrate_timers(struct kvm_vcpu
*vcpu
)
884 set_bit(KVM_REQ_MIGRATE_TIMER
, &vcpu
->requests
);
892 struct kvm_stats_debugfs_item
{
895 enum kvm_stat_kind kind
;
896 struct dentry
*dentry
;
898 extern struct kvm_stats_debugfs_item debugfs_entries
[];
899 extern struct dentry
*kvm_debugfs_dir
;
901 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
902 static inline int mmu_notifier_retry(struct kvm
*kvm
, unsigned long mmu_seq
)
904 if (unlikely(kvm
->mmu_notifier_count
))
907 * Ensure the read of mmu_notifier_count happens before the read
908 * of mmu_notifier_seq. This interacts with the smp_wmb() in
909 * mmu_notifier_invalidate_range_end to make sure that the caller
910 * either sees the old (non-zero) value of mmu_notifier_count or
911 * the new (incremented) value of mmu_notifier_seq.
912 * PowerPC Book3s HV KVM calls this under a per-page lock
913 * rather than under kvm->mmu_lock, for scalability, so
914 * can't rely on kvm->mmu_lock to keep things ordered.
917 if (kvm
->mmu_notifier_seq
!= mmu_seq
)
923 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
926 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
928 #define KVM_MAX_IRQ_ROUTES 1024
931 int kvm_setup_default_irq_routing(struct kvm
*kvm
);
932 int kvm_set_irq_routing(struct kvm
*kvm
,
933 const struct kvm_irq_routing_entry
*entries
,
936 int kvm_set_routing_entry(struct kvm_irq_routing_table
*rt
,
937 struct kvm_kernel_irq_routing_entry
*e
,
938 const struct kvm_irq_routing_entry
*ue
);
939 void kvm_free_irq_routing(struct kvm
*kvm
);
941 int kvm_send_userspace_msi(struct kvm
*kvm
, struct kvm_msi
*msi
);
945 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
949 #ifdef CONFIG_HAVE_KVM_EVENTFD
951 void kvm_eventfd_init(struct kvm
*kvm
);
952 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
954 #ifdef CONFIG_HAVE_KVM_IRQCHIP
955 int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
);
956 void kvm_irqfd_release(struct kvm
*kvm
);
957 void kvm_irq_routing_update(struct kvm
*, struct kvm_irq_routing_table
*);
959 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
964 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
969 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
971 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
976 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
978 #ifdef CONFIG_HAVE_KVM_IRQCHIP
979 static inline void kvm_irq_routing_update(struct kvm
*kvm
,
980 struct kvm_irq_routing_table
*irq_rt
)
982 rcu_assign_pointer(kvm
->irq_routing
, irq_rt
);
986 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
991 #endif /* CONFIG_HAVE_KVM_EVENTFD */
993 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
994 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu
*vcpu
)
996 return vcpu
->kvm
->bsp_vcpu_id
== vcpu
->vcpu_id
;
999 bool kvm_vcpu_compatible(struct kvm_vcpu
*vcpu
);
1003 static inline bool kvm_vcpu_compatible(struct kvm_vcpu
*vcpu
) { return true; }
1007 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
1009 long kvm_vm_ioctl_assigned_device(struct kvm
*kvm
, unsigned ioctl
,
1012 void kvm_free_all_assigned_devices(struct kvm
*kvm
);
1016 static inline long kvm_vm_ioctl_assigned_device(struct kvm
*kvm
, unsigned ioctl
,
1022 static inline void kvm_free_all_assigned_devices(struct kvm
*kvm
) {}
1026 static inline void kvm_make_request(int req
, struct kvm_vcpu
*vcpu
)
1028 set_bit(req
, &vcpu
->requests
);
1031 static inline bool kvm_check_request(int req
, struct kvm_vcpu
*vcpu
)
1033 if (test_bit(req
, &vcpu
->requests
)) {
1034 clear_bit(req
, &vcpu
->requests
);
1041 extern bool kvm_rebooting
;
1043 struct kvm_device_ops
;
1046 struct kvm_device_ops
*ops
;
1049 struct list_head vm_node
;
1052 /* create, destroy, and name are mandatory */
1053 struct kvm_device_ops
{
1055 int (*create
)(struct kvm_device
*dev
, u32 type
);
1058 * Destroy is responsible for freeing dev.
1060 * Destroy may be called before or after destructors are called
1061 * on emulated I/O regions, depending on whether a reference is
1062 * held by a vcpu or other kvm component that gets destroyed
1063 * after the emulated I/O.
1065 void (*destroy
)(struct kvm_device
*dev
);
1067 int (*set_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1068 int (*get_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1069 int (*has_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1070 long (*ioctl
)(struct kvm_device
*dev
, unsigned int ioctl
,
1074 void kvm_device_get(struct kvm_device
*dev
);
1075 void kvm_device_put(struct kvm_device
*dev
);
1076 struct kvm_device
*kvm_device_from_filp(struct file
*filp
);
1078 extern struct kvm_device_ops kvm_mpic_ops
;
1079 extern struct kvm_device_ops kvm_xics_ops
;
1080 extern struct kvm_device_ops kvm_vfio_ops
;
1081 extern struct kvm_device_ops kvm_arm_vgic_v2_ops
;
1082 extern struct kvm_device_ops kvm_flic_ops
;
1084 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1086 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1088 vcpu
->spin_loop
.in_spin_loop
= val
;
1090 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1092 vcpu
->spin_loop
.dy_eligible
= val
;
1095 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1097 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1101 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1104 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */