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 <linux/swait.h>
29 #include <linux/refcount.h>
30 #include <asm/signal.h>
32 #include <linux/kvm.h>
33 #include <linux/kvm_para.h>
35 #include <linux/kvm_types.h>
37 #include <asm/kvm_host.h>
39 #ifndef KVM_MAX_VCPU_ID
40 #define KVM_MAX_VCPU_ID KVM_MAX_VCPUS
44 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
45 * in kvm, other bits are visible for userspace which are defined in
46 * include/linux/kvm_h.
48 #define KVM_MEMSLOT_INVALID (1UL << 16)
50 /* Two fragments for cross MMIO pages. */
51 #define KVM_MAX_MMIO_FRAGMENTS 2
53 #ifndef KVM_ADDRESS_SPACE_NUM
54 #define KVM_ADDRESS_SPACE_NUM 1
58 * For the normal pfn, the highest 12 bits should be zero,
59 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
60 * mask bit 63 to indicate the noslot pfn.
62 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
63 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
64 #define KVM_PFN_NOSLOT (0x1ULL << 63)
66 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
67 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
68 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
71 * error pfns indicate that the gfn is in slot but faild to
72 * translate it to pfn on host.
74 static inline bool is_error_pfn(kvm_pfn_t pfn
)
76 return !!(pfn
& KVM_PFN_ERR_MASK
);
80 * error_noslot pfns indicate that the gfn can not be
81 * translated to pfn - it is not in slot or failed to
82 * translate it to pfn.
84 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn
)
86 return !!(pfn
& KVM_PFN_ERR_NOSLOT_MASK
);
89 /* noslot pfn indicates that the gfn is not in slot. */
90 static inline bool is_noslot_pfn(kvm_pfn_t pfn
)
92 return pfn
== KVM_PFN_NOSLOT
;
96 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
97 * provide own defines and kvm_is_error_hva
99 #ifndef KVM_HVA_ERR_BAD
101 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
102 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
104 static inline bool kvm_is_error_hva(unsigned long addr
)
106 return addr
>= PAGE_OFFSET
;
111 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
113 static inline bool is_error_page(struct page
*page
)
118 #define KVM_REQUEST_MASK GENMASK(7,0)
119 #define KVM_REQUEST_NO_WAKEUP BIT(8)
120 #define KVM_REQUEST_WAIT BIT(9)
122 * Architecture-independent vcpu->requests bit members
123 * Bits 4-7 are reserved for more arch-independent bits.
125 #define KVM_REQ_TLB_FLUSH (0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
126 #define KVM_REQ_MMU_RELOAD (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
127 #define KVM_REQ_PENDING_TIMER 2
128 #define KVM_REQ_UNHALT 3
130 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
131 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
133 extern struct kmem_cache
*kvm_vcpu_cache
;
135 extern spinlock_t kvm_lock
;
136 extern struct list_head vm_list
;
138 struct kvm_io_range
{
141 struct kvm_io_device
*dev
;
144 #define NR_IOBUS_DEVS 1000
149 struct kvm_io_range range
[];
155 KVM_VIRTIO_CCW_NOTIFY_BUS
,
160 int kvm_io_bus_write(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
161 int len
, const void *val
);
162 int kvm_io_bus_write_cookie(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
,
163 gpa_t addr
, int len
, const void *val
, long cookie
);
164 int kvm_io_bus_read(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
166 int kvm_io_bus_register_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
,
167 int len
, struct kvm_io_device
*dev
);
168 void kvm_io_bus_unregister_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
169 struct kvm_io_device
*dev
);
170 struct kvm_io_device
*kvm_io_bus_get_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
173 #ifdef CONFIG_KVM_ASYNC_PF
174 struct kvm_async_pf
{
175 struct work_struct work
;
176 struct list_head link
;
177 struct list_head queue
;
178 struct kvm_vcpu
*vcpu
;
179 struct mm_struct
*mm
;
182 struct kvm_arch_async_pf arch
;
186 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu
*vcpu
);
187 void kvm_check_async_pf_completion(struct kvm_vcpu
*vcpu
);
188 int kvm_setup_async_pf(struct kvm_vcpu
*vcpu
, gva_t gva
, unsigned long hva
,
189 struct kvm_arch_async_pf
*arch
);
190 int kvm_async_pf_wakeup_all(struct kvm_vcpu
*vcpu
);
197 READING_SHADOW_PAGE_TABLES
,
201 * Sometimes a large or cross-page mmio needs to be broken up into separate
202 * exits for userspace servicing.
204 struct kvm_mmio_fragment
{
212 #ifdef CONFIG_PREEMPT_NOTIFIERS
213 struct preempt_notifier preempt_notifier
;
219 unsigned long requests
;
220 unsigned long guest_debug
;
223 struct list_head blocked_vcpu_list
;
228 int guest_fpu_loaded
, guest_xcr0_loaded
;
229 struct swait_queue_head wq
;
233 struct kvm_vcpu_stat stat
;
234 unsigned int halt_poll_ns
;
237 #ifdef CONFIG_HAS_IOMEM
239 int mmio_read_completed
;
241 int mmio_cur_fragment
;
242 int mmio_nr_fragments
;
243 struct kvm_mmio_fragment mmio_fragments
[KVM_MAX_MMIO_FRAGMENTS
];
246 #ifdef CONFIG_KVM_ASYNC_PF
249 struct list_head queue
;
250 struct list_head done
;
255 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
257 * Cpu relax intercept or pause loop exit optimization
258 * in_spin_loop: set when a vcpu does a pause loop exit
259 * or cpu relax intercepted.
260 * dy_eligible: indicates whether vcpu is eligible for directed yield.
268 struct kvm_vcpu_arch arch
;
269 struct dentry
*debugfs_dentry
;
272 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu
*vcpu
)
275 * The memory barrier ensures a previous write to vcpu->requests cannot
276 * be reordered with the read of vcpu->mode. It pairs with the general
277 * memory barrier following the write of vcpu->mode in VCPU RUN.
279 smp_mb__before_atomic();
280 return cmpxchg(&vcpu
->mode
, IN_GUEST_MODE
, EXITING_GUEST_MODE
);
284 * Some of the bitops functions do not support too long bitmaps.
285 * This number must be determined not to exceed such limits.
287 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
289 struct kvm_memory_slot
{
291 unsigned long npages
;
292 unsigned long *dirty_bitmap
;
293 struct kvm_arch_memory_slot arch
;
294 unsigned long userspace_addr
;
299 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot
*memslot
)
301 return ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
304 struct kvm_s390_adapter_int
{
317 struct kvm_kernel_irq_routing_entry
{
320 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
321 struct kvm
*kvm
, int irq_source_id
, int level
,
335 struct kvm_s390_adapter_int adapter
;
336 struct kvm_hv_sint hv_sint
;
338 struct hlist_node link
;
341 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
342 struct kvm_irq_routing_table
{
343 int chip
[KVM_NR_IRQCHIPS
][KVM_IRQCHIP_NUM_PINS
];
346 * Array indexed by gsi. Each entry contains list of irq chips
347 * the gsi is connected to.
349 struct hlist_head map
[0];
353 #ifndef KVM_PRIVATE_MEM_SLOTS
354 #define KVM_PRIVATE_MEM_SLOTS 0
357 #ifndef KVM_MEM_SLOTS_NUM
358 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
361 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
362 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu
*vcpu
)
370 * memslots are not sorted by id anymore, please use id_to_memslot()
371 * to get the memslot by its id.
373 struct kvm_memslots
{
375 struct kvm_memory_slot memslots
[KVM_MEM_SLOTS_NUM
];
376 /* The mapping table from slot id to the index in memslots[]. */
377 short id_to_index
[KVM_MEM_SLOTS_NUM
];
384 struct mutex slots_lock
;
385 struct mm_struct
*mm
; /* userspace tied to this vm */
386 struct kvm_memslots
*memslots
[KVM_ADDRESS_SPACE_NUM
];
387 struct srcu_struct srcu
;
388 struct srcu_struct irq_srcu
;
389 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
392 * created_vcpus is protected by kvm->lock, and is incremented
393 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
394 * incremented after storing the kvm_vcpu pointer in vcpus,
395 * and is accessed atomically.
397 atomic_t online_vcpus
;
399 int last_boosted_vcpu
;
400 struct list_head vm_list
;
402 struct kvm_io_bus
*buses
[KVM_NR_BUSES
];
403 #ifdef CONFIG_HAVE_KVM_EVENTFD
406 struct list_head items
;
407 struct list_head resampler_list
;
408 struct mutex resampler_lock
;
410 struct list_head ioeventfds
;
412 struct kvm_vm_stat stat
;
413 struct kvm_arch arch
;
414 refcount_t users_count
;
415 #ifdef CONFIG_KVM_MMIO
416 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
417 spinlock_t ring_lock
;
418 struct list_head coalesced_zones
;
421 struct mutex irq_lock
;
422 #ifdef CONFIG_HAVE_KVM_IRQCHIP
424 * Update side is protected by irq_lock.
426 struct kvm_irq_routing_table __rcu
*irq_routing
;
428 #ifdef CONFIG_HAVE_KVM_IRQFD
429 struct hlist_head irq_ack_notifier_list
;
432 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
433 struct mmu_notifier mmu_notifier
;
434 unsigned long mmu_notifier_seq
;
435 long mmu_notifier_count
;
438 struct list_head devices
;
439 struct dentry
*debugfs_dentry
;
440 struct kvm_stat_data
**debugfs_stat_data
;
443 #define kvm_err(fmt, ...) \
444 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
445 #define kvm_info(fmt, ...) \
446 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
447 #define kvm_debug(fmt, ...) \
448 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
449 #define kvm_debug_ratelimited(fmt, ...) \
450 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
452 #define kvm_pr_unimpl(fmt, ...) \
453 pr_err_ratelimited("kvm [%i]: " fmt, \
454 task_tgid_nr(current), ## __VA_ARGS__)
456 /* The guest did something we don't support. */
457 #define vcpu_unimpl(vcpu, fmt, ...) \
458 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
459 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
461 #define vcpu_debug(vcpu, fmt, ...) \
462 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
463 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
464 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
466 #define vcpu_err(vcpu, fmt, ...) \
467 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
469 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
471 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
472 * the caller has read kvm->online_vcpus before (as is the case
473 * for kvm_for_each_vcpu, for example).
476 return kvm
->vcpus
[i
];
479 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
481 idx < atomic_read(&kvm->online_vcpus) && \
482 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
485 static inline struct kvm_vcpu
*kvm_get_vcpu_by_id(struct kvm
*kvm
, int id
)
487 struct kvm_vcpu
*vcpu
= NULL
;
492 if (id
< KVM_MAX_VCPUS
)
493 vcpu
= kvm_get_vcpu(kvm
, id
);
494 if (vcpu
&& vcpu
->vcpu_id
== id
)
496 kvm_for_each_vcpu(i
, vcpu
, kvm
)
497 if (vcpu
->vcpu_id
== id
)
502 #define kvm_for_each_memslot(memslot, slots) \
503 for (memslot = &slots->memslots[0]; \
504 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
507 int kvm_vcpu_init(struct kvm_vcpu
*vcpu
, struct kvm
*kvm
, unsigned id
);
508 void kvm_vcpu_uninit(struct kvm_vcpu
*vcpu
);
510 int __must_check
vcpu_load(struct kvm_vcpu
*vcpu
);
511 void vcpu_put(struct kvm_vcpu
*vcpu
);
513 #ifdef __KVM_HAVE_IOAPIC
514 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm
*kvm
);
515 void kvm_arch_post_irq_routing_update(struct kvm
*kvm
);
517 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm
*kvm
)
520 static inline void kvm_arch_post_irq_routing_update(struct kvm
*kvm
)
525 #ifdef CONFIG_HAVE_KVM_IRQFD
526 int kvm_irqfd_init(void);
527 void kvm_irqfd_exit(void);
529 static inline int kvm_irqfd_init(void)
534 static inline void kvm_irqfd_exit(void)
538 int kvm_init(void *opaque
, unsigned vcpu_size
, unsigned vcpu_align
,
539 struct module
*module
);
542 void kvm_get_kvm(struct kvm
*kvm
);
543 void kvm_put_kvm(struct kvm
*kvm
);
545 static inline struct kvm_memslots
*__kvm_memslots(struct kvm
*kvm
, int as_id
)
547 return rcu_dereference_check(kvm
->memslots
[as_id
],
548 srcu_read_lock_held(&kvm
->srcu
)
549 || lockdep_is_held(&kvm
->slots_lock
));
552 static inline struct kvm_memslots
*kvm_memslots(struct kvm
*kvm
)
554 return __kvm_memslots(kvm
, 0);
557 static inline struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
)
559 int as_id
= kvm_arch_vcpu_memslots_id(vcpu
);
561 return __kvm_memslots(vcpu
->kvm
, as_id
);
564 static inline struct kvm_memory_slot
*
565 id_to_memslot(struct kvm_memslots
*slots
, int id
)
567 int index
= slots
->id_to_index
[id
];
568 struct kvm_memory_slot
*slot
;
570 slot
= &slots
->memslots
[index
];
572 WARN_ON(slot
->id
!= id
);
577 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
578 * - create a new memory slot
579 * - delete an existing memory slot
580 * - modify an existing memory slot
581 * -- move it in the guest physical memory space
582 * -- just change its flags
584 * Since flags can be changed by some of these operations, the following
585 * differentiation is the best we can do for __kvm_set_memory_region():
594 int kvm_set_memory_region(struct kvm
*kvm
,
595 const struct kvm_userspace_memory_region
*mem
);
596 int __kvm_set_memory_region(struct kvm
*kvm
,
597 const struct kvm_userspace_memory_region
*mem
);
598 void kvm_arch_free_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*free
,
599 struct kvm_memory_slot
*dont
);
600 int kvm_arch_create_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*slot
,
601 unsigned long npages
);
602 void kvm_arch_memslots_updated(struct kvm
*kvm
, struct kvm_memslots
*slots
);
603 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
604 struct kvm_memory_slot
*memslot
,
605 const struct kvm_userspace_memory_region
*mem
,
606 enum kvm_mr_change change
);
607 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
608 const struct kvm_userspace_memory_region
*mem
,
609 const struct kvm_memory_slot
*old
,
610 const struct kvm_memory_slot
*new,
611 enum kvm_mr_change change
);
612 bool kvm_largepages_enabled(void);
613 void kvm_disable_largepages(void);
614 /* flush all memory translations */
615 void kvm_arch_flush_shadow_all(struct kvm
*kvm
);
616 /* flush memory translations pointing to 'slot' */
617 void kvm_arch_flush_shadow_memslot(struct kvm
*kvm
,
618 struct kvm_memory_slot
*slot
);
620 int gfn_to_page_many_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
,
621 struct page
**pages
, int nr_pages
);
623 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
624 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
625 unsigned long gfn_to_hva_prot(struct kvm
*kvm
, gfn_t gfn
, bool *writable
);
626 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
627 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
629 void kvm_release_page_clean(struct page
*page
);
630 void kvm_release_page_dirty(struct page
*page
);
631 void kvm_set_page_accessed(struct page
*page
);
633 kvm_pfn_t
gfn_to_pfn_atomic(struct kvm
*kvm
, gfn_t gfn
);
634 kvm_pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
635 kvm_pfn_t
gfn_to_pfn_prot(struct kvm
*kvm
, gfn_t gfn
, bool write_fault
,
637 kvm_pfn_t
gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
638 kvm_pfn_t
gfn_to_pfn_memslot_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
);
639 kvm_pfn_t
__gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
640 bool atomic
, bool *async
, bool write_fault
,
643 void kvm_release_pfn_clean(kvm_pfn_t pfn
);
644 void kvm_set_pfn_dirty(kvm_pfn_t pfn
);
645 void kvm_set_pfn_accessed(kvm_pfn_t pfn
);
646 void kvm_get_pfn(kvm_pfn_t pfn
);
648 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
650 int kvm_read_guest_atomic(struct kvm
*kvm
, gpa_t gpa
, void *data
,
652 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
653 int kvm_vcpu_read_guest_cached(struct kvm_vcpu
*vcpu
, struct gfn_to_hva_cache
*ghc
,
654 void *data
, unsigned long len
);
655 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
656 int offset
, int len
);
657 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
659 int kvm_vcpu_write_guest_cached(struct kvm_vcpu
*v
, struct gfn_to_hva_cache
*ghc
,
660 void *data
, unsigned long len
);
661 int kvm_vcpu_write_guest_offset_cached(struct kvm_vcpu
*v
, struct gfn_to_hva_cache
*ghc
,
662 void *data
, int offset
, unsigned long len
);
663 int kvm_vcpu_gfn_to_hva_cache_init(struct kvm_vcpu
*v
, struct gfn_to_hva_cache
*ghc
,
664 gpa_t gpa
, unsigned long len
);
665 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
666 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
667 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
668 bool kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
669 unsigned long kvm_host_page_size(struct kvm
*kvm
, gfn_t gfn
);
670 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
672 struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
);
673 struct kvm_memory_slot
*kvm_vcpu_gfn_to_memslot(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
674 kvm_pfn_t
kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
675 kvm_pfn_t
kvm_vcpu_gfn_to_pfn(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
676 struct page
*kvm_vcpu_gfn_to_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
677 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
678 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu
*vcpu
, gfn_t gfn
, bool *writable
);
679 int kvm_vcpu_read_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, void *data
, int offset
,
681 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
683 int kvm_vcpu_read_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
685 int kvm_vcpu_write_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, const void *data
,
686 int offset
, int len
);
687 int kvm_vcpu_write_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, const void *data
,
689 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
691 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
692 void kvm_arch_vcpu_blocking(struct kvm_vcpu
*vcpu
);
693 void kvm_arch_vcpu_unblocking(struct kvm_vcpu
*vcpu
);
694 bool kvm_vcpu_wake_up(struct kvm_vcpu
*vcpu
);
695 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
696 int kvm_vcpu_yield_to(struct kvm_vcpu
*target
);
697 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
);
698 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
699 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
701 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
702 void kvm_reload_remote_mmus(struct kvm
*kvm
);
703 bool kvm_make_all_cpus_request(struct kvm
*kvm
, unsigned int req
);
705 long kvm_arch_dev_ioctl(struct file
*filp
,
706 unsigned int ioctl
, unsigned long arg
);
707 long kvm_arch_vcpu_ioctl(struct file
*filp
,
708 unsigned int ioctl
, unsigned long arg
);
709 int kvm_arch_vcpu_fault(struct kvm_vcpu
*vcpu
, struct vm_fault
*vmf
);
711 int kvm_vm_ioctl_check_extension(struct kvm
*kvm
, long ext
);
713 int kvm_get_dirty_log(struct kvm
*kvm
,
714 struct kvm_dirty_log
*log
, int *is_dirty
);
716 int kvm_get_dirty_log_protect(struct kvm
*kvm
,
717 struct kvm_dirty_log
*log
, bool *is_dirty
);
719 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm
*kvm
,
720 struct kvm_memory_slot
*slot
,
724 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
725 struct kvm_dirty_log
*log
);
727 int kvm_vm_ioctl_irq_line(struct kvm
*kvm
, struct kvm_irq_level
*irq_level
,
729 long kvm_arch_vm_ioctl(struct file
*filp
,
730 unsigned int ioctl
, unsigned long arg
);
732 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
733 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
735 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
736 struct kvm_translation
*tr
);
738 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
739 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
740 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
741 struct kvm_sregs
*sregs
);
742 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
743 struct kvm_sregs
*sregs
);
744 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
745 struct kvm_mp_state
*mp_state
);
746 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
747 struct kvm_mp_state
*mp_state
);
748 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
749 struct kvm_guest_debug
*dbg
);
750 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
752 int kvm_arch_init(void *opaque
);
753 void kvm_arch_exit(void);
755 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
);
756 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
);
758 void kvm_arch_sched_in(struct kvm_vcpu
*vcpu
, int cpu
);
760 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
);
761 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
762 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
763 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
);
764 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
);
765 void kvm_arch_vcpu_postcreate(struct kvm_vcpu
*vcpu
);
766 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
768 bool kvm_arch_has_vcpu_debugfs(void);
769 int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu
*vcpu
);
771 int kvm_arch_hardware_enable(void);
772 void kvm_arch_hardware_disable(void);
773 int kvm_arch_hardware_setup(void);
774 void kvm_arch_hardware_unsetup(void);
775 void kvm_arch_check_processor_compat(void *rtn
);
776 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
777 int kvm_arch_vcpu_should_kick(struct kvm_vcpu
*vcpu
);
779 void *kvm_kvzalloc(unsigned long size
);
781 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
782 static inline struct kvm
*kvm_arch_alloc_vm(void)
784 return kzalloc(sizeof(struct kvm
), GFP_KERNEL
);
787 static inline void kvm_arch_free_vm(struct kvm
*kvm
)
793 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
794 void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
);
795 void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
);
796 bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
);
798 static inline void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
)
802 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
)
806 static inline bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
)
811 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
812 void kvm_arch_start_assignment(struct kvm
*kvm
);
813 void kvm_arch_end_assignment(struct kvm
*kvm
);
814 bool kvm_arch_has_assigned_device(struct kvm
*kvm
);
816 static inline void kvm_arch_start_assignment(struct kvm
*kvm
)
820 static inline void kvm_arch_end_assignment(struct kvm
*kvm
)
824 static inline bool kvm_arch_has_assigned_device(struct kvm
*kvm
)
830 static inline struct swait_queue_head
*kvm_arch_vcpu_wq(struct kvm_vcpu
*vcpu
)
832 #ifdef __KVM_HAVE_ARCH_WQP
833 return vcpu
->arch
.wqp
;
839 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
841 * returns true if the virtual interrupt controller is initialized and
842 * ready to accept virtual IRQ. On some architectures the virtual interrupt
843 * controller is dynamically instantiated and this is not always true.
845 bool kvm_arch_intc_initialized(struct kvm
*kvm
);
847 static inline bool kvm_arch_intc_initialized(struct kvm
*kvm
)
853 int kvm_arch_init_vm(struct kvm
*kvm
, unsigned long type
);
854 void kvm_arch_destroy_vm(struct kvm
*kvm
);
855 void kvm_arch_sync_events(struct kvm
*kvm
);
857 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
858 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
860 bool kvm_is_reserved_pfn(kvm_pfn_t pfn
);
862 struct kvm_irq_ack_notifier
{
863 struct hlist_node link
;
865 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
868 int kvm_irq_map_gsi(struct kvm
*kvm
,
869 struct kvm_kernel_irq_routing_entry
*entries
, int gsi
);
870 int kvm_irq_map_chip_pin(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
872 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
,
874 int kvm_set_msi(struct kvm_kernel_irq_routing_entry
*irq_entry
, struct kvm
*kvm
,
875 int irq_source_id
, int level
, bool line_status
);
876 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry
*e
,
877 struct kvm
*kvm
, int irq_source_id
,
878 int level
, bool line_status
);
879 bool kvm_irq_has_notifier(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
880 void kvm_notify_acked_gsi(struct kvm
*kvm
, int gsi
);
881 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
882 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
883 struct kvm_irq_ack_notifier
*kian
);
884 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
885 struct kvm_irq_ack_notifier
*kian
);
886 int kvm_request_irq_source_id(struct kvm
*kvm
);
887 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
890 * search_memslots() and __gfn_to_memslot() are here because they are
891 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
892 * gfn_to_memslot() itself isn't here as an inline because that would
893 * bloat other code too much.
895 static inline struct kvm_memory_slot
*
896 search_memslots(struct kvm_memslots
*slots
, gfn_t gfn
)
898 int start
= 0, end
= slots
->used_slots
;
899 int slot
= atomic_read(&slots
->lru_slot
);
900 struct kvm_memory_slot
*memslots
= slots
->memslots
;
902 if (gfn
>= memslots
[slot
].base_gfn
&&
903 gfn
< memslots
[slot
].base_gfn
+ memslots
[slot
].npages
)
904 return &memslots
[slot
];
906 while (start
< end
) {
907 slot
= start
+ (end
- start
) / 2;
909 if (gfn
>= memslots
[slot
].base_gfn
)
915 if (gfn
>= memslots
[start
].base_gfn
&&
916 gfn
< memslots
[start
].base_gfn
+ memslots
[start
].npages
) {
917 atomic_set(&slots
->lru_slot
, start
);
918 return &memslots
[start
];
924 static inline struct kvm_memory_slot
*
925 __gfn_to_memslot(struct kvm_memslots
*slots
, gfn_t gfn
)
927 return search_memslots(slots
, gfn
);
930 static inline unsigned long
931 __gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
)
933 return slot
->userspace_addr
+ (gfn
- slot
->base_gfn
) * PAGE_SIZE
;
936 static inline int memslot_id(struct kvm
*kvm
, gfn_t gfn
)
938 return gfn_to_memslot(kvm
, gfn
)->id
;
942 hva_to_gfn_memslot(unsigned long hva
, struct kvm_memory_slot
*slot
)
944 gfn_t gfn_offset
= (hva
- slot
->userspace_addr
) >> PAGE_SHIFT
;
946 return slot
->base_gfn
+ gfn_offset
;
949 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
951 return (gpa_t
)gfn
<< PAGE_SHIFT
;
954 static inline gfn_t
gpa_to_gfn(gpa_t gpa
)
956 return (gfn_t
)(gpa
>> PAGE_SHIFT
);
959 static inline hpa_t
pfn_to_hpa(kvm_pfn_t pfn
)
961 return (hpa_t
)pfn
<< PAGE_SHIFT
;
964 static inline bool kvm_is_error_gpa(struct kvm
*kvm
, gpa_t gpa
)
966 unsigned long hva
= gfn_to_hva(kvm
, gpa_to_gfn(gpa
));
968 return kvm_is_error_hva(hva
);
976 struct kvm_stat_data
{
981 struct kvm_stats_debugfs_item
{
984 enum kvm_stat_kind kind
;
986 extern struct kvm_stats_debugfs_item debugfs_entries
[];
987 extern struct dentry
*kvm_debugfs_dir
;
989 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
990 static inline int mmu_notifier_retry(struct kvm
*kvm
, unsigned long mmu_seq
)
992 if (unlikely(kvm
->mmu_notifier_count
))
995 * Ensure the read of mmu_notifier_count happens before the read
996 * of mmu_notifier_seq. This interacts with the smp_wmb() in
997 * mmu_notifier_invalidate_range_end to make sure that the caller
998 * either sees the old (non-zero) value of mmu_notifier_count or
999 * the new (incremented) value of mmu_notifier_seq.
1000 * PowerPC Book3s HV KVM calls this under a per-page lock
1001 * rather than under kvm->mmu_lock, for scalability, so
1002 * can't rely on kvm->mmu_lock to keep things ordered.
1005 if (kvm
->mmu_notifier_seq
!= mmu_seq
)
1011 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1014 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1015 #elif defined(CONFIG_ARM64)
1016 #define KVM_MAX_IRQ_ROUTES 4096
1018 #define KVM_MAX_IRQ_ROUTES 1024
1021 bool kvm_arch_can_set_irq_routing(struct kvm
*kvm
);
1022 int kvm_set_irq_routing(struct kvm
*kvm
,
1023 const struct kvm_irq_routing_entry
*entries
,
1026 int kvm_set_routing_entry(struct kvm
*kvm
,
1027 struct kvm_kernel_irq_routing_entry
*e
,
1028 const struct kvm_irq_routing_entry
*ue
);
1029 void kvm_free_irq_routing(struct kvm
*kvm
);
1033 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
1037 int kvm_send_userspace_msi(struct kvm
*kvm
, struct kvm_msi
*msi
);
1039 #ifdef CONFIG_HAVE_KVM_EVENTFD
1041 void kvm_eventfd_init(struct kvm
*kvm
);
1042 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
1044 #ifdef CONFIG_HAVE_KVM_IRQFD
1045 int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
);
1046 void kvm_irqfd_release(struct kvm
*kvm
);
1047 void kvm_irq_routing_update(struct kvm
*);
1049 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1054 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1059 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
1061 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1066 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1068 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1069 static inline void kvm_irq_routing_update(struct kvm
*kvm
)
1073 void kvm_arch_irq_routing_update(struct kvm
*kvm
);
1075 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
1080 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1082 static inline void kvm_make_request(int req
, struct kvm_vcpu
*vcpu
)
1085 * Ensure the rest of the request is published to kvm_check_request's
1086 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1089 set_bit(req
& KVM_REQUEST_MASK
, &vcpu
->requests
);
1092 static inline bool kvm_test_request(int req
, struct kvm_vcpu
*vcpu
)
1094 return test_bit(req
& KVM_REQUEST_MASK
, &vcpu
->requests
);
1097 static inline void kvm_clear_request(int req
, struct kvm_vcpu
*vcpu
)
1099 clear_bit(req
& KVM_REQUEST_MASK
, &vcpu
->requests
);
1102 static inline bool kvm_check_request(int req
, struct kvm_vcpu
*vcpu
)
1104 if (kvm_test_request(req
, vcpu
)) {
1105 kvm_clear_request(req
, vcpu
);
1108 * Ensure the rest of the request is visible to kvm_check_request's
1109 * caller. Paired with the smp_wmb in kvm_make_request.
1111 smp_mb__after_atomic();
1118 extern bool kvm_rebooting
;
1120 extern unsigned int halt_poll_ns
;
1121 extern unsigned int halt_poll_ns_grow
;
1122 extern unsigned int halt_poll_ns_shrink
;
1125 struct kvm_device_ops
*ops
;
1128 struct list_head vm_node
;
1131 /* create, destroy, and name are mandatory */
1132 struct kvm_device_ops
{
1136 * create is called holding kvm->lock and any operations not suitable
1137 * to do while holding the lock should be deferred to init (see
1140 int (*create
)(struct kvm_device
*dev
, u32 type
);
1143 * init is called after create if create is successful and is called
1144 * outside of holding kvm->lock.
1146 void (*init
)(struct kvm_device
*dev
);
1149 * Destroy is responsible for freeing dev.
1151 * Destroy may be called before or after destructors are called
1152 * on emulated I/O regions, depending on whether a reference is
1153 * held by a vcpu or other kvm component that gets destroyed
1154 * after the emulated I/O.
1156 void (*destroy
)(struct kvm_device
*dev
);
1158 int (*set_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1159 int (*get_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1160 int (*has_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1161 long (*ioctl
)(struct kvm_device
*dev
, unsigned int ioctl
,
1165 void kvm_device_get(struct kvm_device
*dev
);
1166 void kvm_device_put(struct kvm_device
*dev
);
1167 struct kvm_device
*kvm_device_from_filp(struct file
*filp
);
1168 int kvm_register_device_ops(struct kvm_device_ops
*ops
, u32 type
);
1169 void kvm_unregister_device_ops(u32 type
);
1171 extern struct kvm_device_ops kvm_mpic_ops
;
1172 extern struct kvm_device_ops kvm_xics_ops
;
1173 extern struct kvm_device_ops kvm_arm_vgic_v2_ops
;
1174 extern struct kvm_device_ops kvm_arm_vgic_v3_ops
;
1176 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1178 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1180 vcpu
->spin_loop
.in_spin_loop
= val
;
1182 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1184 vcpu
->spin_loop
.dy_eligible
= val
;
1187 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1189 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1193 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1196 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1198 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1199 bool kvm_arch_has_irq_bypass(void);
1200 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer
*,
1201 struct irq_bypass_producer
*);
1202 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer
*,
1203 struct irq_bypass_producer
*);
1204 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer
*);
1205 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer
*);
1206 int kvm_arch_update_irqfd_routing(struct kvm
*kvm
, unsigned int host_irq
,
1207 uint32_t guest_irq
, bool set
);
1208 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1210 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1211 /* If we wakeup during the poll time, was it a sucessful poll? */
1212 static inline bool vcpu_valid_wakeup(struct kvm_vcpu
*vcpu
)
1214 return vcpu
->valid_wakeup
;
1218 static inline bool vcpu_valid_wakeup(struct kvm_vcpu
*vcpu
)
1222 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */