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
147 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
148 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
150 extern struct kmem_cache
*kvm_vcpu_cache
;
152 extern spinlock_t kvm_lock
;
153 extern struct list_head vm_list
;
155 struct kvm_io_range
{
158 struct kvm_io_device
*dev
;
161 #define NR_IOBUS_DEVS 1000
166 struct kvm_io_range range
[];
172 KVM_VIRTIO_CCW_NOTIFY_BUS
,
177 int kvm_io_bus_write(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
178 int len
, const void *val
);
179 int kvm_io_bus_write_cookie(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
,
180 gpa_t addr
, int len
, const void *val
, long cookie
);
181 int kvm_io_bus_read(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
183 int kvm_io_bus_register_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
,
184 int len
, struct kvm_io_device
*dev
);
185 int kvm_io_bus_unregister_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
186 struct kvm_io_device
*dev
);
188 #ifdef CONFIG_KVM_ASYNC_PF
189 struct kvm_async_pf
{
190 struct work_struct work
;
191 struct list_head link
;
192 struct list_head queue
;
193 struct kvm_vcpu
*vcpu
;
194 struct mm_struct
*mm
;
197 struct kvm_arch_async_pf arch
;
201 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu
*vcpu
);
202 void kvm_check_async_pf_completion(struct kvm_vcpu
*vcpu
);
203 int kvm_setup_async_pf(struct kvm_vcpu
*vcpu
, gva_t gva
, unsigned long hva
,
204 struct kvm_arch_async_pf
*arch
);
205 int kvm_async_pf_wakeup_all(struct kvm_vcpu
*vcpu
);
212 READING_SHADOW_PAGE_TABLES
,
216 * Sometimes a large or cross-page mmio needs to be broken up into separate
217 * exits for userspace servicing.
219 struct kvm_mmio_fragment
{
227 #ifdef CONFIG_PREEMPT_NOTIFIERS
228 struct preempt_notifier preempt_notifier
;
234 unsigned long requests
;
235 unsigned long guest_debug
;
238 struct list_head blocked_vcpu_list
;
244 int guest_fpu_loaded
, guest_xcr0_loaded
;
245 unsigned char fpu_counter
;
246 wait_queue_head_t wq
;
250 struct kvm_vcpu_stat stat
;
251 unsigned int halt_poll_ns
;
253 #ifdef CONFIG_HAS_IOMEM
255 int mmio_read_completed
;
257 int mmio_cur_fragment
;
258 int mmio_nr_fragments
;
259 struct kvm_mmio_fragment mmio_fragments
[KVM_MAX_MMIO_FRAGMENTS
];
262 #ifdef CONFIG_KVM_ASYNC_PF
265 struct list_head queue
;
266 struct list_head done
;
271 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
273 * Cpu relax intercept or pause loop exit optimization
274 * in_spin_loop: set when a vcpu does a pause loop exit
275 * or cpu relax intercepted.
276 * dy_eligible: indicates whether vcpu is eligible for directed yield.
284 struct kvm_vcpu_arch arch
;
287 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu
*vcpu
)
289 return cmpxchg(&vcpu
->mode
, IN_GUEST_MODE
, EXITING_GUEST_MODE
);
293 * Some of the bitops functions do not support too long bitmaps.
294 * This number must be determined not to exceed such limits.
296 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
298 struct kvm_memory_slot
{
300 unsigned long npages
;
301 unsigned long *dirty_bitmap
;
302 struct kvm_arch_memory_slot arch
;
303 unsigned long userspace_addr
;
308 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot
*memslot
)
310 return ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
313 struct kvm_s390_adapter_int
{
321 struct kvm_kernel_irq_routing_entry
{
324 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
325 struct kvm
*kvm
, int irq_source_id
, int level
,
333 struct kvm_s390_adapter_int adapter
;
335 struct hlist_node link
;
338 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
339 struct kvm_irq_routing_table
{
340 int chip
[KVM_NR_IRQCHIPS
][KVM_IRQCHIP_NUM_PINS
];
343 * Array indexed by gsi. Each entry contains list of irq chips
344 * the gsi is connected to.
346 struct hlist_head map
[0];
350 #ifndef KVM_PRIVATE_MEM_SLOTS
351 #define KVM_PRIVATE_MEM_SLOTS 0
354 #ifndef KVM_MEM_SLOTS_NUM
355 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
358 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
359 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu
*vcpu
)
367 * memslots are not sorted by id anymore, please use id_to_memslot()
368 * to get the memslot by its id.
370 struct kvm_memslots
{
372 struct kvm_memory_slot memslots
[KVM_MEM_SLOTS_NUM
];
373 /* The mapping table from slot id to the index in memslots[]. */
374 short id_to_index
[KVM_MEM_SLOTS_NUM
];
381 struct mutex slots_lock
;
382 struct mm_struct
*mm
; /* userspace tied to this vm */
383 struct kvm_memslots
*memslots
[KVM_ADDRESS_SPACE_NUM
];
384 struct srcu_struct srcu
;
385 struct srcu_struct irq_srcu
;
386 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
387 atomic_t online_vcpus
;
388 int last_boosted_vcpu
;
389 struct list_head vm_list
;
391 struct kvm_io_bus
*buses
[KVM_NR_BUSES
];
392 #ifdef CONFIG_HAVE_KVM_EVENTFD
395 struct list_head items
;
396 struct list_head resampler_list
;
397 struct mutex resampler_lock
;
399 struct list_head ioeventfds
;
401 struct kvm_vm_stat stat
;
402 struct kvm_arch arch
;
403 atomic_t users_count
;
404 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
405 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
406 spinlock_t ring_lock
;
407 struct list_head coalesced_zones
;
410 struct mutex irq_lock
;
411 #ifdef CONFIG_HAVE_KVM_IRQCHIP
413 * Update side is protected by irq_lock.
415 struct kvm_irq_routing_table __rcu
*irq_routing
;
417 #ifdef CONFIG_HAVE_KVM_IRQFD
418 struct hlist_head irq_ack_notifier_list
;
421 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
422 struct mmu_notifier mmu_notifier
;
423 unsigned long mmu_notifier_seq
;
424 long mmu_notifier_count
;
427 struct list_head devices
;
430 #define kvm_err(fmt, ...) \
431 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
432 #define kvm_info(fmt, ...) \
433 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
434 #define kvm_debug(fmt, ...) \
435 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
436 #define kvm_pr_unimpl(fmt, ...) \
437 pr_err_ratelimited("kvm [%i]: " fmt, \
438 task_tgid_nr(current), ## __VA_ARGS__)
440 /* The guest did something we don't support. */
441 #define vcpu_unimpl(vcpu, fmt, ...) \
442 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
444 #define vcpu_debug(vcpu, fmt, ...) \
445 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
447 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
449 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
450 * the caller has read kvm->online_vcpus before (as is the case
451 * for kvm_for_each_vcpu, for example).
454 return kvm
->vcpus
[i
];
457 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
459 idx < atomic_read(&kvm->online_vcpus) && \
460 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
463 static inline struct kvm_vcpu
*kvm_get_vcpu_by_id(struct kvm
*kvm
, int id
)
465 struct kvm_vcpu
*vcpu
;
468 kvm_for_each_vcpu(i
, vcpu
, kvm
)
469 if (vcpu
->vcpu_id
== id
)
474 #define kvm_for_each_memslot(memslot, slots) \
475 for (memslot = &slots->memslots[0]; \
476 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
479 int kvm_vcpu_init(struct kvm_vcpu
*vcpu
, struct kvm
*kvm
, unsigned id
);
480 void kvm_vcpu_uninit(struct kvm_vcpu
*vcpu
);
482 int __must_check
vcpu_load(struct kvm_vcpu
*vcpu
);
483 void vcpu_put(struct kvm_vcpu
*vcpu
);
485 #ifdef __KVM_HAVE_IOAPIC
486 void kvm_vcpu_request_scan_ioapic(struct kvm
*kvm
);
487 void kvm_arch_irq_routing_update(struct kvm
*kvm
);
489 static inline void kvm_vcpu_request_scan_ioapic(struct kvm
*kvm
)
492 static inline void kvm_arch_irq_routing_update(struct kvm
*kvm
)
497 #ifdef CONFIG_HAVE_KVM_IRQFD
498 int kvm_irqfd_init(void);
499 void kvm_irqfd_exit(void);
501 static inline int kvm_irqfd_init(void)
506 static inline void kvm_irqfd_exit(void)
510 int kvm_init(void *opaque
, unsigned vcpu_size
, unsigned vcpu_align
,
511 struct module
*module
);
514 void kvm_get_kvm(struct kvm
*kvm
);
515 void kvm_put_kvm(struct kvm
*kvm
);
517 static inline struct kvm_memslots
*__kvm_memslots(struct kvm
*kvm
, int as_id
)
519 return rcu_dereference_check(kvm
->memslots
[as_id
],
520 srcu_read_lock_held(&kvm
->srcu
)
521 || lockdep_is_held(&kvm
->slots_lock
));
524 static inline struct kvm_memslots
*kvm_memslots(struct kvm
*kvm
)
526 return __kvm_memslots(kvm
, 0);
529 static inline struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
)
531 int as_id
= kvm_arch_vcpu_memslots_id(vcpu
);
533 return __kvm_memslots(vcpu
->kvm
, as_id
);
536 static inline struct kvm_memory_slot
*
537 id_to_memslot(struct kvm_memslots
*slots
, int id
)
539 int index
= slots
->id_to_index
[id
];
540 struct kvm_memory_slot
*slot
;
542 slot
= &slots
->memslots
[index
];
544 WARN_ON(slot
->id
!= id
);
549 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
550 * - create a new memory slot
551 * - delete an existing memory slot
552 * - modify an existing memory slot
553 * -- move it in the guest physical memory space
554 * -- just change its flags
556 * Since flags can be changed by some of these operations, the following
557 * differentiation is the best we can do for __kvm_set_memory_region():
566 int kvm_set_memory_region(struct kvm
*kvm
,
567 const struct kvm_userspace_memory_region
*mem
);
568 int __kvm_set_memory_region(struct kvm
*kvm
,
569 const struct kvm_userspace_memory_region
*mem
);
570 void kvm_arch_free_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*free
,
571 struct kvm_memory_slot
*dont
);
572 int kvm_arch_create_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*slot
,
573 unsigned long npages
);
574 void kvm_arch_memslots_updated(struct kvm
*kvm
, struct kvm_memslots
*slots
);
575 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
576 struct kvm_memory_slot
*memslot
,
577 const struct kvm_userspace_memory_region
*mem
,
578 enum kvm_mr_change change
);
579 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
580 const struct kvm_userspace_memory_region
*mem
,
581 const struct kvm_memory_slot
*old
,
582 const struct kvm_memory_slot
*new,
583 enum kvm_mr_change change
);
584 bool kvm_largepages_enabled(void);
585 void kvm_disable_largepages(void);
586 /* flush all memory translations */
587 void kvm_arch_flush_shadow_all(struct kvm
*kvm
);
588 /* flush memory translations pointing to 'slot' */
589 void kvm_arch_flush_shadow_memslot(struct kvm
*kvm
,
590 struct kvm_memory_slot
*slot
);
592 int gfn_to_page_many_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
,
593 struct page
**pages
, int nr_pages
);
595 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
596 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
597 unsigned long gfn_to_hva_prot(struct kvm
*kvm
, gfn_t gfn
, bool *writable
);
598 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
599 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
601 void kvm_release_page_clean(struct page
*page
);
602 void kvm_release_page_dirty(struct page
*page
);
603 void kvm_set_page_accessed(struct page
*page
);
605 pfn_t
gfn_to_pfn_atomic(struct kvm
*kvm
, gfn_t gfn
);
606 pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
607 pfn_t
gfn_to_pfn_prot(struct kvm
*kvm
, gfn_t gfn
, bool write_fault
,
609 pfn_t
gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
610 pfn_t
gfn_to_pfn_memslot_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
);
611 pfn_t
__gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
, bool atomic
,
612 bool *async
, bool write_fault
, bool *writable
);
614 void kvm_release_pfn_clean(pfn_t pfn
);
615 void kvm_set_pfn_dirty(pfn_t pfn
);
616 void kvm_set_pfn_accessed(pfn_t pfn
);
617 void kvm_get_pfn(pfn_t pfn
);
619 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
621 int kvm_read_guest_atomic(struct kvm
*kvm
, gpa_t gpa
, void *data
,
623 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
624 int kvm_read_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
625 void *data
, unsigned long len
);
626 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
627 int offset
, int len
);
628 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
630 int kvm_write_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
631 void *data
, unsigned long len
);
632 int kvm_gfn_to_hva_cache_init(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
633 gpa_t gpa
, unsigned long len
);
634 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
635 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
636 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
637 int kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
638 unsigned long kvm_host_page_size(struct kvm
*kvm
, gfn_t gfn
);
639 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
641 struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
);
642 struct kvm_memory_slot
*kvm_vcpu_gfn_to_memslot(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
643 pfn_t
kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
644 pfn_t
kvm_vcpu_gfn_to_pfn(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
645 struct page
*kvm_vcpu_gfn_to_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
646 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
647 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu
*vcpu
, gfn_t gfn
, bool *writable
);
648 int kvm_vcpu_read_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, void *data
, int offset
,
650 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
652 int kvm_vcpu_read_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
654 int kvm_vcpu_write_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, const void *data
,
655 int offset
, int len
);
656 int kvm_vcpu_write_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, const void *data
,
658 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
660 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
661 void kvm_arch_vcpu_blocking(struct kvm_vcpu
*vcpu
);
662 void kvm_arch_vcpu_unblocking(struct kvm_vcpu
*vcpu
);
663 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
664 int kvm_vcpu_yield_to(struct kvm_vcpu
*target
);
665 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
);
666 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
667 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
669 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
670 void kvm_reload_remote_mmus(struct kvm
*kvm
);
671 void kvm_make_mclock_inprogress_request(struct kvm
*kvm
);
672 void kvm_make_scan_ioapic_request(struct kvm
*kvm
);
673 bool kvm_make_all_cpus_request(struct kvm
*kvm
, unsigned int req
);
675 long kvm_arch_dev_ioctl(struct file
*filp
,
676 unsigned int ioctl
, unsigned long arg
);
677 long kvm_arch_vcpu_ioctl(struct file
*filp
,
678 unsigned int ioctl
, unsigned long arg
);
679 int kvm_arch_vcpu_fault(struct kvm_vcpu
*vcpu
, struct vm_fault
*vmf
);
681 int kvm_vm_ioctl_check_extension(struct kvm
*kvm
, long ext
);
683 int kvm_get_dirty_log(struct kvm
*kvm
,
684 struct kvm_dirty_log
*log
, int *is_dirty
);
686 int kvm_get_dirty_log_protect(struct kvm
*kvm
,
687 struct kvm_dirty_log
*log
, bool *is_dirty
);
689 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm
*kvm
,
690 struct kvm_memory_slot
*slot
,
694 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
695 struct kvm_dirty_log
*log
);
697 int kvm_vm_ioctl_irq_line(struct kvm
*kvm
, struct kvm_irq_level
*irq_level
,
699 long kvm_arch_vm_ioctl(struct file
*filp
,
700 unsigned int ioctl
, unsigned long arg
);
702 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
703 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
705 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
706 struct kvm_translation
*tr
);
708 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
709 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
710 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
711 struct kvm_sregs
*sregs
);
712 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
713 struct kvm_sregs
*sregs
);
714 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
715 struct kvm_mp_state
*mp_state
);
716 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
717 struct kvm_mp_state
*mp_state
);
718 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
719 struct kvm_guest_debug
*dbg
);
720 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
722 int kvm_arch_init(void *opaque
);
723 void kvm_arch_exit(void);
725 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
);
726 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
);
728 void kvm_arch_sched_in(struct kvm_vcpu
*vcpu
, int cpu
);
730 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
);
731 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
732 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
733 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
);
734 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
);
735 void kvm_arch_vcpu_postcreate(struct kvm_vcpu
*vcpu
);
736 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
738 int kvm_arch_hardware_enable(void);
739 void kvm_arch_hardware_disable(void);
740 int kvm_arch_hardware_setup(void);
741 void kvm_arch_hardware_unsetup(void);
742 void kvm_arch_check_processor_compat(void *rtn
);
743 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
744 int kvm_arch_vcpu_should_kick(struct kvm_vcpu
*vcpu
);
746 void *kvm_kvzalloc(unsigned long size
);
748 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
749 static inline struct kvm
*kvm_arch_alloc_vm(void)
751 return kzalloc(sizeof(struct kvm
), GFP_KERNEL
);
754 static inline void kvm_arch_free_vm(struct kvm
*kvm
)
760 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
761 void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
);
762 void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
);
763 bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
);
765 static inline void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
)
769 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
)
773 static inline bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
)
778 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
779 void kvm_arch_start_assignment(struct kvm
*kvm
);
780 void kvm_arch_end_assignment(struct kvm
*kvm
);
781 bool kvm_arch_has_assigned_device(struct kvm
*kvm
);
783 static inline void kvm_arch_start_assignment(struct kvm
*kvm
)
787 static inline void kvm_arch_end_assignment(struct kvm
*kvm
)
791 static inline bool kvm_arch_has_assigned_device(struct kvm
*kvm
)
797 static inline wait_queue_head_t
*kvm_arch_vcpu_wq(struct kvm_vcpu
*vcpu
)
799 #ifdef __KVM_HAVE_ARCH_WQP
800 return vcpu
->arch
.wqp
;
806 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
808 * returns true if the virtual interrupt controller is initialized and
809 * ready to accept virtual IRQ. On some architectures the virtual interrupt
810 * controller is dynamically instantiated and this is not always true.
812 bool kvm_arch_intc_initialized(struct kvm
*kvm
);
814 static inline bool kvm_arch_intc_initialized(struct kvm
*kvm
)
820 int kvm_arch_init_vm(struct kvm
*kvm
, unsigned long type
);
821 void kvm_arch_destroy_vm(struct kvm
*kvm
);
822 void kvm_arch_sync_events(struct kvm
*kvm
);
824 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
825 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
827 bool kvm_is_reserved_pfn(pfn_t pfn
);
829 struct kvm_irq_ack_notifier
{
830 struct hlist_node link
;
832 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
835 int kvm_irq_map_gsi(struct kvm
*kvm
,
836 struct kvm_kernel_irq_routing_entry
*entries
, int gsi
);
837 int kvm_irq_map_chip_pin(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
839 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
,
841 int kvm_set_msi(struct kvm_kernel_irq_routing_entry
*irq_entry
, struct kvm
*kvm
,
842 int irq_source_id
, int level
, bool line_status
);
843 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry
*e
,
844 struct kvm
*kvm
, int irq_source_id
,
845 int level
, bool line_status
);
846 bool kvm_irq_has_notifier(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
847 void kvm_notify_acked_gsi(struct kvm
*kvm
, int gsi
);
848 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
849 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
850 struct kvm_irq_ack_notifier
*kian
);
851 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
852 struct kvm_irq_ack_notifier
*kian
);
853 int kvm_request_irq_source_id(struct kvm
*kvm
);
854 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
856 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
857 int kvm_iommu_map_pages(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
858 void kvm_iommu_unmap_pages(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
860 static inline int kvm_iommu_map_pages(struct kvm
*kvm
,
861 struct kvm_memory_slot
*slot
)
866 static inline void kvm_iommu_unmap_pages(struct kvm
*kvm
,
867 struct kvm_memory_slot
*slot
)
872 /* must be called with irqs disabled */
873 static inline void __kvm_guest_enter(void)
876 /* KVM does not hold any references to rcu protected data when it
877 * switches CPU into a guest mode. In fact switching to a guest mode
878 * is very similar to exiting to userspace from rcu point of view. In
879 * addition CPU may stay in a guest mode for quite a long time (up to
880 * one time slice). Lets treat guest mode as quiescent state, just like
881 * we do with user-mode execution.
883 if (!context_tracking_cpu_is_enabled())
884 rcu_virt_note_context_switch(smp_processor_id());
887 /* must be called with irqs disabled */
888 static inline void __kvm_guest_exit(void)
893 static inline void kvm_guest_enter(void)
897 local_irq_save(flags
);
899 local_irq_restore(flags
);
902 static inline void kvm_guest_exit(void)
906 local_irq_save(flags
);
908 local_irq_restore(flags
);
912 * search_memslots() and __gfn_to_memslot() are here because they are
913 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
914 * gfn_to_memslot() itself isn't here as an inline because that would
915 * bloat other code too much.
917 static inline struct kvm_memory_slot
*
918 search_memslots(struct kvm_memslots
*slots
, gfn_t gfn
)
920 int start
= 0, end
= slots
->used_slots
;
921 int slot
= atomic_read(&slots
->lru_slot
);
922 struct kvm_memory_slot
*memslots
= slots
->memslots
;
924 if (gfn
>= memslots
[slot
].base_gfn
&&
925 gfn
< memslots
[slot
].base_gfn
+ memslots
[slot
].npages
)
926 return &memslots
[slot
];
928 while (start
< end
) {
929 slot
= start
+ (end
- start
) / 2;
931 if (gfn
>= memslots
[slot
].base_gfn
)
937 if (gfn
>= memslots
[start
].base_gfn
&&
938 gfn
< memslots
[start
].base_gfn
+ memslots
[start
].npages
) {
939 atomic_set(&slots
->lru_slot
, start
);
940 return &memslots
[start
];
946 static inline struct kvm_memory_slot
*
947 __gfn_to_memslot(struct kvm_memslots
*slots
, gfn_t gfn
)
949 return search_memslots(slots
, gfn
);
952 static inline unsigned long
953 __gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
)
955 return slot
->userspace_addr
+ (gfn
- slot
->base_gfn
) * PAGE_SIZE
;
958 static inline int memslot_id(struct kvm
*kvm
, gfn_t gfn
)
960 return gfn_to_memslot(kvm
, gfn
)->id
;
964 hva_to_gfn_memslot(unsigned long hva
, struct kvm_memory_slot
*slot
)
966 gfn_t gfn_offset
= (hva
- slot
->userspace_addr
) >> PAGE_SHIFT
;
968 return slot
->base_gfn
+ gfn_offset
;
971 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
973 return (gpa_t
)gfn
<< PAGE_SHIFT
;
976 static inline gfn_t
gpa_to_gfn(gpa_t gpa
)
978 return (gfn_t
)(gpa
>> PAGE_SHIFT
);
981 static inline hpa_t
pfn_to_hpa(pfn_t pfn
)
983 return (hpa_t
)pfn
<< PAGE_SHIFT
;
986 static inline bool kvm_is_error_gpa(struct kvm
*kvm
, gpa_t gpa
)
988 unsigned long hva
= gfn_to_hva(kvm
, gpa_to_gfn(gpa
));
990 return kvm_is_error_hva(hva
);
993 static inline void kvm_migrate_timers(struct kvm_vcpu
*vcpu
)
995 set_bit(KVM_REQ_MIGRATE_TIMER
, &vcpu
->requests
);
1003 struct kvm_stats_debugfs_item
{
1006 enum kvm_stat_kind kind
;
1007 struct dentry
*dentry
;
1009 extern struct kvm_stats_debugfs_item debugfs_entries
[];
1010 extern struct dentry
*kvm_debugfs_dir
;
1012 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1013 static inline int mmu_notifier_retry(struct kvm
*kvm
, unsigned long mmu_seq
)
1015 if (unlikely(kvm
->mmu_notifier_count
))
1018 * Ensure the read of mmu_notifier_count happens before the read
1019 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1020 * mmu_notifier_invalidate_range_end to make sure that the caller
1021 * either sees the old (non-zero) value of mmu_notifier_count or
1022 * the new (incremented) value of mmu_notifier_seq.
1023 * PowerPC Book3s HV KVM calls this under a per-page lock
1024 * rather than under kvm->mmu_lock, for scalability, so
1025 * can't rely on kvm->mmu_lock to keep things ordered.
1028 if (kvm
->mmu_notifier_seq
!= mmu_seq
)
1034 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1037 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1039 #define KVM_MAX_IRQ_ROUTES 1024
1042 int kvm_setup_default_irq_routing(struct kvm
*kvm
);
1043 int kvm_setup_empty_irq_routing(struct kvm
*kvm
);
1044 int kvm_set_irq_routing(struct kvm
*kvm
,
1045 const struct kvm_irq_routing_entry
*entries
,
1048 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry
*e
,
1049 const struct kvm_irq_routing_entry
*ue
);
1050 void kvm_free_irq_routing(struct kvm
*kvm
);
1054 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
1058 int kvm_send_userspace_msi(struct kvm
*kvm
, struct kvm_msi
*msi
);
1060 #ifdef CONFIG_HAVE_KVM_EVENTFD
1062 void kvm_eventfd_init(struct kvm
*kvm
);
1063 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
1065 #ifdef CONFIG_HAVE_KVM_IRQFD
1066 int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
);
1067 void kvm_irqfd_release(struct kvm
*kvm
);
1068 void kvm_irq_routing_update(struct kvm
*);
1070 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1075 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1080 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
1082 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1087 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1089 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1090 static inline void kvm_irq_routing_update(struct kvm
*kvm
)
1095 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
1100 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1102 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
1103 bool kvm_vcpu_compatible(struct kvm_vcpu
*vcpu
);
1105 static inline bool kvm_vcpu_compatible(struct kvm_vcpu
*vcpu
) { return true; }
1108 static inline void kvm_make_request(int req
, struct kvm_vcpu
*vcpu
)
1110 set_bit(req
, &vcpu
->requests
);
1113 static inline bool kvm_check_request(int req
, struct kvm_vcpu
*vcpu
)
1115 if (test_bit(req
, &vcpu
->requests
)) {
1116 clear_bit(req
, &vcpu
->requests
);
1123 extern bool kvm_rebooting
;
1126 struct kvm_device_ops
*ops
;
1129 struct list_head vm_node
;
1132 /* create, destroy, and name are mandatory */
1133 struct kvm_device_ops
{
1135 int (*create
)(struct kvm_device
*dev
, u32 type
);
1138 * Destroy is responsible for freeing dev.
1140 * Destroy may be called before or after destructors are called
1141 * on emulated I/O regions, depending on whether a reference is
1142 * held by a vcpu or other kvm component that gets destroyed
1143 * after the emulated I/O.
1145 void (*destroy
)(struct kvm_device
*dev
);
1147 int (*set_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1148 int (*get_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1149 int (*has_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1150 long (*ioctl
)(struct kvm_device
*dev
, unsigned int ioctl
,
1154 void kvm_device_get(struct kvm_device
*dev
);
1155 void kvm_device_put(struct kvm_device
*dev
);
1156 struct kvm_device
*kvm_device_from_filp(struct file
*filp
);
1157 int kvm_register_device_ops(struct kvm_device_ops
*ops
, u32 type
);
1158 void kvm_unregister_device_ops(u32 type
);
1160 extern struct kvm_device_ops kvm_mpic_ops
;
1161 extern struct kvm_device_ops kvm_xics_ops
;
1162 extern struct kvm_device_ops kvm_arm_vgic_v2_ops
;
1163 extern struct kvm_device_ops kvm_arm_vgic_v3_ops
;
1165 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1167 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1169 vcpu
->spin_loop
.in_spin_loop
= val
;
1171 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1173 vcpu
->spin_loop
.dy_eligible
= val
;
1176 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1178 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1182 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1185 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1187 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1188 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer
*,
1189 struct irq_bypass_producer
*);
1190 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer
*,
1191 struct irq_bypass_producer
*);
1192 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer
*);
1193 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer
*);
1194 int kvm_arch_update_irqfd_routing(struct kvm
*kvm
, unsigned int host_irq
,
1195 uint32_t guest_irq
, bool set
);
1196 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */