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(kvm_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(kvm_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(kvm_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 * Architecture-independent vcpu->requests bit members
115 * Bits 4-7 are reserved for more arch-independent bits.
117 #define KVM_REQ_TLB_FLUSH 0
118 #define KVM_REQ_MMU_RELOAD 1
119 #define KVM_REQ_PENDING_TIMER 2
120 #define KVM_REQ_UNHALT 3
122 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
123 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
125 extern struct kmem_cache
*kvm_vcpu_cache
;
127 extern spinlock_t kvm_lock
;
128 extern struct list_head vm_list
;
130 struct kvm_io_range
{
133 struct kvm_io_device
*dev
;
136 #define NR_IOBUS_DEVS 1000
141 struct kvm_io_range range
[];
147 KVM_VIRTIO_CCW_NOTIFY_BUS
,
152 int kvm_io_bus_write(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
153 int len
, const void *val
);
154 int kvm_io_bus_write_cookie(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
,
155 gpa_t addr
, int len
, const void *val
, long cookie
);
156 int kvm_io_bus_read(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
158 int kvm_io_bus_register_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
,
159 int len
, struct kvm_io_device
*dev
);
160 int kvm_io_bus_unregister_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
161 struct kvm_io_device
*dev
);
163 #ifdef CONFIG_KVM_ASYNC_PF
164 struct kvm_async_pf
{
165 struct work_struct work
;
166 struct list_head link
;
167 struct list_head queue
;
168 struct kvm_vcpu
*vcpu
;
169 struct mm_struct
*mm
;
172 struct kvm_arch_async_pf arch
;
176 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu
*vcpu
);
177 void kvm_check_async_pf_completion(struct kvm_vcpu
*vcpu
);
178 int kvm_setup_async_pf(struct kvm_vcpu
*vcpu
, gva_t gva
, unsigned long hva
,
179 struct kvm_arch_async_pf
*arch
);
180 int kvm_async_pf_wakeup_all(struct kvm_vcpu
*vcpu
);
187 READING_SHADOW_PAGE_TABLES
,
191 * Sometimes a large or cross-page mmio needs to be broken up into separate
192 * exits for userspace servicing.
194 struct kvm_mmio_fragment
{
202 #ifdef CONFIG_PREEMPT_NOTIFIERS
203 struct preempt_notifier preempt_notifier
;
209 unsigned long requests
;
210 unsigned long guest_debug
;
213 struct list_head blocked_vcpu_list
;
219 int guest_fpu_loaded
, guest_xcr0_loaded
;
220 unsigned char fpu_counter
;
221 wait_queue_head_t wq
;
225 struct kvm_vcpu_stat stat
;
226 unsigned int halt_poll_ns
;
228 #ifdef CONFIG_HAS_IOMEM
230 int mmio_read_completed
;
232 int mmio_cur_fragment
;
233 int mmio_nr_fragments
;
234 struct kvm_mmio_fragment mmio_fragments
[KVM_MAX_MMIO_FRAGMENTS
];
237 #ifdef CONFIG_KVM_ASYNC_PF
240 struct list_head queue
;
241 struct list_head done
;
246 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
248 * Cpu relax intercept or pause loop exit optimization
249 * in_spin_loop: set when a vcpu does a pause loop exit
250 * or cpu relax intercepted.
251 * dy_eligible: indicates whether vcpu is eligible for directed yield.
259 struct kvm_vcpu_arch arch
;
262 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu
*vcpu
)
264 return cmpxchg(&vcpu
->mode
, IN_GUEST_MODE
, EXITING_GUEST_MODE
);
268 * Some of the bitops functions do not support too long bitmaps.
269 * This number must be determined not to exceed such limits.
271 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
273 struct kvm_memory_slot
{
275 unsigned long npages
;
276 unsigned long *dirty_bitmap
;
277 struct kvm_arch_memory_slot arch
;
278 unsigned long userspace_addr
;
283 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot
*memslot
)
285 return ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
288 struct kvm_s390_adapter_int
{
301 struct kvm_kernel_irq_routing_entry
{
304 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
305 struct kvm
*kvm
, int irq_source_id
, int level
,
313 struct kvm_s390_adapter_int adapter
;
314 struct kvm_hv_sint hv_sint
;
316 struct hlist_node link
;
319 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
320 struct kvm_irq_routing_table
{
321 int chip
[KVM_NR_IRQCHIPS
][KVM_IRQCHIP_NUM_PINS
];
324 * Array indexed by gsi. Each entry contains list of irq chips
325 * the gsi is connected to.
327 struct hlist_head map
[0];
331 #ifndef KVM_PRIVATE_MEM_SLOTS
332 #define KVM_PRIVATE_MEM_SLOTS 0
335 #ifndef KVM_MEM_SLOTS_NUM
336 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
339 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
340 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu
*vcpu
)
348 * memslots are not sorted by id anymore, please use id_to_memslot()
349 * to get the memslot by its id.
351 struct kvm_memslots
{
353 struct kvm_memory_slot memslots
[KVM_MEM_SLOTS_NUM
];
354 /* The mapping table from slot id to the index in memslots[]. */
355 short id_to_index
[KVM_MEM_SLOTS_NUM
];
362 struct mutex slots_lock
;
363 struct mm_struct
*mm
; /* userspace tied to this vm */
364 struct kvm_memslots
*memslots
[KVM_ADDRESS_SPACE_NUM
];
365 struct srcu_struct srcu
;
366 struct srcu_struct irq_srcu
;
367 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
368 atomic_t online_vcpus
;
369 int last_boosted_vcpu
;
370 struct list_head vm_list
;
372 struct kvm_io_bus
*buses
[KVM_NR_BUSES
];
373 #ifdef CONFIG_HAVE_KVM_EVENTFD
376 struct list_head items
;
377 struct list_head resampler_list
;
378 struct mutex resampler_lock
;
380 struct list_head ioeventfds
;
382 struct kvm_vm_stat stat
;
383 struct kvm_arch arch
;
384 atomic_t users_count
;
385 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
386 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
387 spinlock_t ring_lock
;
388 struct list_head coalesced_zones
;
391 struct mutex irq_lock
;
392 #ifdef CONFIG_HAVE_KVM_IRQCHIP
394 * Update side is protected by irq_lock.
396 struct kvm_irq_routing_table __rcu
*irq_routing
;
398 #ifdef CONFIG_HAVE_KVM_IRQFD
399 struct hlist_head irq_ack_notifier_list
;
402 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
403 struct mmu_notifier mmu_notifier
;
404 unsigned long mmu_notifier_seq
;
405 long mmu_notifier_count
;
408 struct list_head devices
;
411 #define kvm_err(fmt, ...) \
412 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
413 #define kvm_info(fmt, ...) \
414 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
415 #define kvm_debug(fmt, ...) \
416 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
417 #define kvm_pr_unimpl(fmt, ...) \
418 pr_err_ratelimited("kvm [%i]: " fmt, \
419 task_tgid_nr(current), ## __VA_ARGS__)
421 /* The guest did something we don't support. */
422 #define vcpu_unimpl(vcpu, fmt, ...) \
423 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
424 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
426 #define vcpu_debug(vcpu, fmt, ...) \
427 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
428 #define vcpu_err(vcpu, fmt, ...) \
429 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
431 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
433 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
434 * the caller has read kvm->online_vcpus before (as is the case
435 * for kvm_for_each_vcpu, for example).
438 return kvm
->vcpus
[i
];
441 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
443 idx < atomic_read(&kvm->online_vcpus) && \
444 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
447 static inline struct kvm_vcpu
*kvm_get_vcpu_by_id(struct kvm
*kvm
, int id
)
449 struct kvm_vcpu
*vcpu
;
452 if (id
< 0 || id
>= KVM_MAX_VCPUS
)
454 vcpu
= kvm_get_vcpu(kvm
, id
);
455 if (vcpu
&& vcpu
->vcpu_id
== id
)
457 kvm_for_each_vcpu(i
, vcpu
, kvm
)
458 if (vcpu
->vcpu_id
== id
)
463 #define kvm_for_each_memslot(memslot, slots) \
464 for (memslot = &slots->memslots[0]; \
465 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
468 int kvm_vcpu_init(struct kvm_vcpu
*vcpu
, struct kvm
*kvm
, unsigned id
);
469 void kvm_vcpu_uninit(struct kvm_vcpu
*vcpu
);
471 int __must_check
vcpu_load(struct kvm_vcpu
*vcpu
);
472 void vcpu_put(struct kvm_vcpu
*vcpu
);
474 #ifdef __KVM_HAVE_IOAPIC
475 void kvm_vcpu_request_scan_ioapic(struct kvm
*kvm
);
476 void kvm_arch_post_irq_routing_update(struct kvm
*kvm
);
478 static inline void kvm_vcpu_request_scan_ioapic(struct kvm
*kvm
)
481 static inline void kvm_arch_post_irq_routing_update(struct kvm
*kvm
)
486 #ifdef CONFIG_HAVE_KVM_IRQFD
487 int kvm_irqfd_init(void);
488 void kvm_irqfd_exit(void);
490 static inline int kvm_irqfd_init(void)
495 static inline void kvm_irqfd_exit(void)
499 int kvm_init(void *opaque
, unsigned vcpu_size
, unsigned vcpu_align
,
500 struct module
*module
);
503 void kvm_get_kvm(struct kvm
*kvm
);
504 void kvm_put_kvm(struct kvm
*kvm
);
506 static inline struct kvm_memslots
*__kvm_memslots(struct kvm
*kvm
, int as_id
)
508 return rcu_dereference_check(kvm
->memslots
[as_id
],
509 srcu_read_lock_held(&kvm
->srcu
)
510 || lockdep_is_held(&kvm
->slots_lock
));
513 static inline struct kvm_memslots
*kvm_memslots(struct kvm
*kvm
)
515 return __kvm_memslots(kvm
, 0);
518 static inline struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
)
520 int as_id
= kvm_arch_vcpu_memslots_id(vcpu
);
522 return __kvm_memslots(vcpu
->kvm
, as_id
);
525 static inline struct kvm_memory_slot
*
526 id_to_memslot(struct kvm_memslots
*slots
, int id
)
528 int index
= slots
->id_to_index
[id
];
529 struct kvm_memory_slot
*slot
;
531 slot
= &slots
->memslots
[index
];
533 WARN_ON(slot
->id
!= id
);
538 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
539 * - create a new memory slot
540 * - delete an existing memory slot
541 * - modify an existing memory slot
542 * -- move it in the guest physical memory space
543 * -- just change its flags
545 * Since flags can be changed by some of these operations, the following
546 * differentiation is the best we can do for __kvm_set_memory_region():
555 int kvm_set_memory_region(struct kvm
*kvm
,
556 const struct kvm_userspace_memory_region
*mem
);
557 int __kvm_set_memory_region(struct kvm
*kvm
,
558 const struct kvm_userspace_memory_region
*mem
);
559 void kvm_arch_free_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*free
,
560 struct kvm_memory_slot
*dont
);
561 int kvm_arch_create_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*slot
,
562 unsigned long npages
);
563 void kvm_arch_memslots_updated(struct kvm
*kvm
, struct kvm_memslots
*slots
);
564 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
565 struct kvm_memory_slot
*memslot
,
566 const struct kvm_userspace_memory_region
*mem
,
567 enum kvm_mr_change change
);
568 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
569 const struct kvm_userspace_memory_region
*mem
,
570 const struct kvm_memory_slot
*old
,
571 const struct kvm_memory_slot
*new,
572 enum kvm_mr_change change
);
573 bool kvm_largepages_enabled(void);
574 void kvm_disable_largepages(void);
575 /* flush all memory translations */
576 void kvm_arch_flush_shadow_all(struct kvm
*kvm
);
577 /* flush memory translations pointing to 'slot' */
578 void kvm_arch_flush_shadow_memslot(struct kvm
*kvm
,
579 struct kvm_memory_slot
*slot
);
581 int gfn_to_page_many_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
,
582 struct page
**pages
, int nr_pages
);
584 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
585 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
586 unsigned long gfn_to_hva_prot(struct kvm
*kvm
, gfn_t gfn
, bool *writable
);
587 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
588 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
590 void kvm_release_page_clean(struct page
*page
);
591 void kvm_release_page_dirty(struct page
*page
);
592 void kvm_set_page_accessed(struct page
*page
);
594 kvm_pfn_t
gfn_to_pfn_atomic(struct kvm
*kvm
, gfn_t gfn
);
595 kvm_pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
596 kvm_pfn_t
gfn_to_pfn_prot(struct kvm
*kvm
, gfn_t gfn
, bool write_fault
,
598 kvm_pfn_t
gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
599 kvm_pfn_t
gfn_to_pfn_memslot_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
);
600 kvm_pfn_t
__gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
601 bool atomic
, bool *async
, bool write_fault
,
604 void kvm_release_pfn_clean(kvm_pfn_t pfn
);
605 void kvm_set_pfn_dirty(kvm_pfn_t pfn
);
606 void kvm_set_pfn_accessed(kvm_pfn_t pfn
);
607 void kvm_get_pfn(kvm_pfn_t pfn
);
609 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
611 int kvm_read_guest_atomic(struct kvm
*kvm
, gpa_t gpa
, void *data
,
613 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
614 int kvm_read_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
615 void *data
, unsigned long len
);
616 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
617 int offset
, int len
);
618 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
620 int kvm_write_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
621 void *data
, unsigned long len
);
622 int kvm_gfn_to_hva_cache_init(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
623 gpa_t gpa
, unsigned long len
);
624 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
625 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
626 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
627 bool kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
628 unsigned long kvm_host_page_size(struct kvm
*kvm
, gfn_t gfn
);
629 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
631 struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
);
632 struct kvm_memory_slot
*kvm_vcpu_gfn_to_memslot(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
633 kvm_pfn_t
kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
634 kvm_pfn_t
kvm_vcpu_gfn_to_pfn(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
635 struct page
*kvm_vcpu_gfn_to_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
636 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
637 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu
*vcpu
, gfn_t gfn
, bool *writable
);
638 int kvm_vcpu_read_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, void *data
, int offset
,
640 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
642 int kvm_vcpu_read_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
644 int kvm_vcpu_write_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, const void *data
,
645 int offset
, int len
);
646 int kvm_vcpu_write_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, const void *data
,
648 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
650 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
651 void kvm_arch_vcpu_blocking(struct kvm_vcpu
*vcpu
);
652 void kvm_arch_vcpu_unblocking(struct kvm_vcpu
*vcpu
);
653 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
654 int kvm_vcpu_yield_to(struct kvm_vcpu
*target
);
655 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
);
656 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
657 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
659 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
660 void kvm_reload_remote_mmus(struct kvm
*kvm
);
661 bool kvm_make_all_cpus_request(struct kvm
*kvm
, unsigned int req
);
663 long kvm_arch_dev_ioctl(struct file
*filp
,
664 unsigned int ioctl
, unsigned long arg
);
665 long kvm_arch_vcpu_ioctl(struct file
*filp
,
666 unsigned int ioctl
, unsigned long arg
);
667 int kvm_arch_vcpu_fault(struct kvm_vcpu
*vcpu
, struct vm_fault
*vmf
);
669 int kvm_vm_ioctl_check_extension(struct kvm
*kvm
, long ext
);
671 int kvm_get_dirty_log(struct kvm
*kvm
,
672 struct kvm_dirty_log
*log
, int *is_dirty
);
674 int kvm_get_dirty_log_protect(struct kvm
*kvm
,
675 struct kvm_dirty_log
*log
, bool *is_dirty
);
677 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm
*kvm
,
678 struct kvm_memory_slot
*slot
,
682 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
683 struct kvm_dirty_log
*log
);
685 int kvm_vm_ioctl_irq_line(struct kvm
*kvm
, struct kvm_irq_level
*irq_level
,
687 long kvm_arch_vm_ioctl(struct file
*filp
,
688 unsigned int ioctl
, unsigned long arg
);
690 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
691 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
693 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
694 struct kvm_translation
*tr
);
696 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
697 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
698 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
699 struct kvm_sregs
*sregs
);
700 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
701 struct kvm_sregs
*sregs
);
702 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
703 struct kvm_mp_state
*mp_state
);
704 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
705 struct kvm_mp_state
*mp_state
);
706 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
707 struct kvm_guest_debug
*dbg
);
708 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
710 int kvm_arch_init(void *opaque
);
711 void kvm_arch_exit(void);
713 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
);
714 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
);
716 void kvm_arch_sched_in(struct kvm_vcpu
*vcpu
, int cpu
);
718 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
);
719 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
720 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
721 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
);
722 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
);
723 void kvm_arch_vcpu_postcreate(struct kvm_vcpu
*vcpu
);
724 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
726 int kvm_arch_hardware_enable(void);
727 void kvm_arch_hardware_disable(void);
728 int kvm_arch_hardware_setup(void);
729 void kvm_arch_hardware_unsetup(void);
730 void kvm_arch_check_processor_compat(void *rtn
);
731 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
732 int kvm_arch_vcpu_should_kick(struct kvm_vcpu
*vcpu
);
734 void *kvm_kvzalloc(unsigned long size
);
736 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
737 static inline struct kvm
*kvm_arch_alloc_vm(void)
739 return kzalloc(sizeof(struct kvm
), GFP_KERNEL
);
742 static inline void kvm_arch_free_vm(struct kvm
*kvm
)
748 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
749 void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
);
750 void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
);
751 bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
);
753 static inline void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
)
757 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
)
761 static inline bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
)
766 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
767 void kvm_arch_start_assignment(struct kvm
*kvm
);
768 void kvm_arch_end_assignment(struct kvm
*kvm
);
769 bool kvm_arch_has_assigned_device(struct kvm
*kvm
);
771 static inline void kvm_arch_start_assignment(struct kvm
*kvm
)
775 static inline void kvm_arch_end_assignment(struct kvm
*kvm
)
779 static inline bool kvm_arch_has_assigned_device(struct kvm
*kvm
)
785 static inline wait_queue_head_t
*kvm_arch_vcpu_wq(struct kvm_vcpu
*vcpu
)
787 #ifdef __KVM_HAVE_ARCH_WQP
788 return vcpu
->arch
.wqp
;
794 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
796 * returns true if the virtual interrupt controller is initialized and
797 * ready to accept virtual IRQ. On some architectures the virtual interrupt
798 * controller is dynamically instantiated and this is not always true.
800 bool kvm_arch_intc_initialized(struct kvm
*kvm
);
802 static inline bool kvm_arch_intc_initialized(struct kvm
*kvm
)
808 int kvm_arch_init_vm(struct kvm
*kvm
, unsigned long type
);
809 void kvm_arch_destroy_vm(struct kvm
*kvm
);
810 void kvm_arch_sync_events(struct kvm
*kvm
);
812 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
813 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
815 bool kvm_is_reserved_pfn(kvm_pfn_t pfn
);
817 struct kvm_irq_ack_notifier
{
818 struct hlist_node link
;
820 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
823 int kvm_irq_map_gsi(struct kvm
*kvm
,
824 struct kvm_kernel_irq_routing_entry
*entries
, int gsi
);
825 int kvm_irq_map_chip_pin(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
827 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
,
829 int kvm_set_msi(struct kvm_kernel_irq_routing_entry
*irq_entry
, struct kvm
*kvm
,
830 int irq_source_id
, int level
, bool line_status
);
831 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry
*e
,
832 struct kvm
*kvm
, int irq_source_id
,
833 int level
, bool line_status
);
834 bool kvm_irq_has_notifier(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
835 void kvm_notify_acked_gsi(struct kvm
*kvm
, int gsi
);
836 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
837 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
838 struct kvm_irq_ack_notifier
*kian
);
839 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
840 struct kvm_irq_ack_notifier
*kian
);
841 int kvm_request_irq_source_id(struct kvm
*kvm
);
842 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
844 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
845 int kvm_iommu_map_pages(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
846 void kvm_iommu_unmap_pages(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
848 static inline int kvm_iommu_map_pages(struct kvm
*kvm
,
849 struct kvm_memory_slot
*slot
)
854 static inline void kvm_iommu_unmap_pages(struct kvm
*kvm
,
855 struct kvm_memory_slot
*slot
)
860 /* must be called with irqs disabled */
861 static inline void __kvm_guest_enter(void)
864 /* KVM does not hold any references to rcu protected data when it
865 * switches CPU into a guest mode. In fact switching to a guest mode
866 * is very similar to exiting to userspace from rcu point of view. In
867 * addition CPU may stay in a guest mode for quite a long time (up to
868 * one time slice). Lets treat guest mode as quiescent state, just like
869 * we do with user-mode execution.
871 if (!context_tracking_cpu_is_enabled())
872 rcu_virt_note_context_switch(smp_processor_id());
875 /* must be called with irqs disabled */
876 static inline void __kvm_guest_exit(void)
881 static inline void kvm_guest_enter(void)
885 local_irq_save(flags
);
887 local_irq_restore(flags
);
890 static inline void kvm_guest_exit(void)
894 local_irq_save(flags
);
896 local_irq_restore(flags
);
900 * search_memslots() and __gfn_to_memslot() are here because they are
901 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
902 * gfn_to_memslot() itself isn't here as an inline because that would
903 * bloat other code too much.
905 static inline struct kvm_memory_slot
*
906 search_memslots(struct kvm_memslots
*slots
, gfn_t gfn
)
908 int start
= 0, end
= slots
->used_slots
;
909 int slot
= atomic_read(&slots
->lru_slot
);
910 struct kvm_memory_slot
*memslots
= slots
->memslots
;
912 if (gfn
>= memslots
[slot
].base_gfn
&&
913 gfn
< memslots
[slot
].base_gfn
+ memslots
[slot
].npages
)
914 return &memslots
[slot
];
916 while (start
< end
) {
917 slot
= start
+ (end
- start
) / 2;
919 if (gfn
>= memslots
[slot
].base_gfn
)
925 if (gfn
>= memslots
[start
].base_gfn
&&
926 gfn
< memslots
[start
].base_gfn
+ memslots
[start
].npages
) {
927 atomic_set(&slots
->lru_slot
, start
);
928 return &memslots
[start
];
934 static inline struct kvm_memory_slot
*
935 __gfn_to_memslot(struct kvm_memslots
*slots
, gfn_t gfn
)
937 return search_memslots(slots
, gfn
);
940 static inline unsigned long
941 __gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
)
943 return slot
->userspace_addr
+ (gfn
- slot
->base_gfn
) * PAGE_SIZE
;
946 static inline int memslot_id(struct kvm
*kvm
, gfn_t gfn
)
948 return gfn_to_memslot(kvm
, gfn
)->id
;
952 hva_to_gfn_memslot(unsigned long hva
, struct kvm_memory_slot
*slot
)
954 gfn_t gfn_offset
= (hva
- slot
->userspace_addr
) >> PAGE_SHIFT
;
956 return slot
->base_gfn
+ gfn_offset
;
959 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
961 return (gpa_t
)gfn
<< PAGE_SHIFT
;
964 static inline gfn_t
gpa_to_gfn(gpa_t gpa
)
966 return (gfn_t
)(gpa
>> PAGE_SHIFT
);
969 static inline hpa_t
pfn_to_hpa(kvm_pfn_t pfn
)
971 return (hpa_t
)pfn
<< PAGE_SHIFT
;
974 static inline bool kvm_is_error_gpa(struct kvm
*kvm
, gpa_t gpa
)
976 unsigned long hva
= gfn_to_hva(kvm
, gpa_to_gfn(gpa
));
978 return kvm_is_error_hva(hva
);
986 struct kvm_stats_debugfs_item
{
989 enum kvm_stat_kind kind
;
991 extern struct kvm_stats_debugfs_item debugfs_entries
[];
992 extern struct dentry
*kvm_debugfs_dir
;
994 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
995 static inline int mmu_notifier_retry(struct kvm
*kvm
, unsigned long mmu_seq
)
997 if (unlikely(kvm
->mmu_notifier_count
))
1000 * Ensure the read of mmu_notifier_count happens before the read
1001 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1002 * mmu_notifier_invalidate_range_end to make sure that the caller
1003 * either sees the old (non-zero) value of mmu_notifier_count or
1004 * the new (incremented) value of mmu_notifier_seq.
1005 * PowerPC Book3s HV KVM calls this under a per-page lock
1006 * rather than under kvm->mmu_lock, for scalability, so
1007 * can't rely on kvm->mmu_lock to keep things ordered.
1010 if (kvm
->mmu_notifier_seq
!= mmu_seq
)
1016 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1019 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1021 #define KVM_MAX_IRQ_ROUTES 1024
1024 int kvm_setup_default_irq_routing(struct kvm
*kvm
);
1025 int kvm_setup_empty_irq_routing(struct kvm
*kvm
);
1026 int kvm_set_irq_routing(struct kvm
*kvm
,
1027 const struct kvm_irq_routing_entry
*entries
,
1030 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry
*e
,
1031 const struct kvm_irq_routing_entry
*ue
);
1032 void kvm_free_irq_routing(struct kvm
*kvm
);
1036 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
1040 int kvm_send_userspace_msi(struct kvm
*kvm
, struct kvm_msi
*msi
);
1042 #ifdef CONFIG_HAVE_KVM_EVENTFD
1044 void kvm_eventfd_init(struct kvm
*kvm
);
1045 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
1047 #ifdef CONFIG_HAVE_KVM_IRQFD
1048 int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
);
1049 void kvm_irqfd_release(struct kvm
*kvm
);
1050 void kvm_irq_routing_update(struct kvm
*);
1052 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1057 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1062 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
1064 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1069 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1071 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1072 static inline void kvm_irq_routing_update(struct kvm
*kvm
)
1076 void kvm_arch_irq_routing_update(struct kvm
*kvm
);
1078 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
1083 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1085 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
1086 bool kvm_vcpu_compatible(struct kvm_vcpu
*vcpu
);
1088 static inline bool kvm_vcpu_compatible(struct kvm_vcpu
*vcpu
) { return true; }
1091 static inline void kvm_make_request(int req
, struct kvm_vcpu
*vcpu
)
1093 set_bit(req
, &vcpu
->requests
);
1096 static inline bool kvm_check_request(int req
, struct kvm_vcpu
*vcpu
)
1098 if (test_bit(req
, &vcpu
->requests
)) {
1099 clear_bit(req
, &vcpu
->requests
);
1106 extern bool kvm_rebooting
;
1109 struct kvm_device_ops
*ops
;
1112 struct list_head vm_node
;
1115 /* create, destroy, and name are mandatory */
1116 struct kvm_device_ops
{
1118 int (*create
)(struct kvm_device
*dev
, u32 type
);
1121 * Destroy is responsible for freeing dev.
1123 * Destroy may be called before or after destructors are called
1124 * on emulated I/O regions, depending on whether a reference is
1125 * held by a vcpu or other kvm component that gets destroyed
1126 * after the emulated I/O.
1128 void (*destroy
)(struct kvm_device
*dev
);
1130 int (*set_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1131 int (*get_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1132 int (*has_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1133 long (*ioctl
)(struct kvm_device
*dev
, unsigned int ioctl
,
1137 void kvm_device_get(struct kvm_device
*dev
);
1138 void kvm_device_put(struct kvm_device
*dev
);
1139 struct kvm_device
*kvm_device_from_filp(struct file
*filp
);
1140 int kvm_register_device_ops(struct kvm_device_ops
*ops
, u32 type
);
1141 void kvm_unregister_device_ops(u32 type
);
1143 extern struct kvm_device_ops kvm_mpic_ops
;
1144 extern struct kvm_device_ops kvm_xics_ops
;
1145 extern struct kvm_device_ops kvm_arm_vgic_v2_ops
;
1146 extern struct kvm_device_ops kvm_arm_vgic_v3_ops
;
1148 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1150 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1152 vcpu
->spin_loop
.in_spin_loop
= val
;
1154 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1156 vcpu
->spin_loop
.dy_eligible
= val
;
1159 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1161 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1165 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1168 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1170 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1171 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer
*,
1172 struct irq_bypass_producer
*);
1173 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer
*,
1174 struct irq_bypass_producer
*);
1175 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer
*);
1176 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer
*);
1177 int kvm_arch_update_irqfd_routing(struct kvm
*kvm
, unsigned int host_irq
,
1178 uint32_t guest_irq
, bool set
);
1179 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */