5 * This work is licensed under the terms of the GNU GPL, version 2. See
6 * the COPYING file in the top-level directory.
9 #include <linux/types.h>
10 #include <linux/hardirq.h>
11 #include <linux/list.h>
12 #include <linux/mutex.h>
13 #include <linux/spinlock.h>
14 #include <linux/signal.h>
15 #include <linux/sched.h>
16 #include <linux/bug.h>
18 #include <linux/mmu_notifier.h>
19 #include <linux/preempt.h>
20 #include <linux/msi.h>
21 #include <linux/slab.h>
22 #include <linux/rcupdate.h>
23 #include <linux/ratelimit.h>
24 #include <linux/err.h>
25 #include <linux/irqflags.h>
26 #include <linux/context_tracking.h>
27 #include <asm/signal.h>
29 #include <linux/kvm.h>
30 #include <linux/kvm_para.h>
32 #include <linux/kvm_types.h>
34 #include <asm/kvm_host.h>
37 #define KVM_MMIO_SIZE 8
41 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
42 * in kvm, other bits are visible for userspace which are defined in
43 * include/linux/kvm_h.
45 #define KVM_MEMSLOT_INVALID (1UL << 16)
47 /* Two fragments for cross MMIO pages. */
48 #define KVM_MAX_MMIO_FRAGMENTS 2
51 * For the normal pfn, the highest 12 bits should be zero,
52 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
53 * mask bit 63 to indicate the noslot pfn.
55 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
56 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
57 #define KVM_PFN_NOSLOT (0x1ULL << 63)
59 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
60 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
61 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
64 * error pfns indicate that the gfn is in slot but faild to
65 * translate it to pfn on host.
67 static inline bool is_error_pfn(pfn_t pfn
)
69 return !!(pfn
& KVM_PFN_ERR_MASK
);
73 * error_noslot pfns indicate that the gfn can not be
74 * translated to pfn - it is not in slot or failed to
75 * translate it to pfn.
77 static inline bool is_error_noslot_pfn(pfn_t pfn
)
79 return !!(pfn
& KVM_PFN_ERR_NOSLOT_MASK
);
82 /* noslot pfn indicates that the gfn is not in slot. */
83 static inline bool is_noslot_pfn(pfn_t pfn
)
85 return pfn
== KVM_PFN_NOSLOT
;
89 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
90 * provide own defines and kvm_is_error_hva
92 #ifndef KVM_HVA_ERR_BAD
94 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
95 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
97 static inline bool kvm_is_error_hva(unsigned long addr
)
99 return addr
>= PAGE_OFFSET
;
104 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
106 static inline bool is_error_page(struct page
*page
)
112 * vcpu->requests bit members
114 #define KVM_REQ_TLB_FLUSH 0
115 #define KVM_REQ_MIGRATE_TIMER 1
116 #define KVM_REQ_REPORT_TPR_ACCESS 2
117 #define KVM_REQ_MMU_RELOAD 3
118 #define KVM_REQ_TRIPLE_FAULT 4
119 #define KVM_REQ_PENDING_TIMER 5
120 #define KVM_REQ_UNHALT 6
121 #define KVM_REQ_MMU_SYNC 7
122 #define KVM_REQ_CLOCK_UPDATE 8
123 #define KVM_REQ_KICK 9
124 #define KVM_REQ_DEACTIVATE_FPU 10
125 #define KVM_REQ_EVENT 11
126 #define KVM_REQ_APF_HALT 12
127 #define KVM_REQ_STEAL_UPDATE 13
128 #define KVM_REQ_NMI 14
129 #define KVM_REQ_PMU 15
130 #define KVM_REQ_PMI 16
131 #define KVM_REQ_WATCHDOG 17
132 #define KVM_REQ_MASTERCLOCK_UPDATE 18
133 #define KVM_REQ_MCLOCK_INPROGRESS 19
134 #define KVM_REQ_EPR_EXIT 20
135 #define KVM_REQ_SCAN_IOAPIC 21
136 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 22
137 #define KVM_REQ_ENABLE_IBS 23
138 #define KVM_REQ_DISABLE_IBS 24
140 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
141 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
145 extern struct kmem_cache
*kvm_vcpu_cache
;
147 extern spinlock_t kvm_lock
;
148 extern struct list_head vm_list
;
150 struct kvm_io_range
{
153 struct kvm_io_device
*dev
;
156 #define NR_IOBUS_DEVS 1000
161 struct kvm_io_range range
[];
167 KVM_VIRTIO_CCW_NOTIFY_BUS
,
172 int kvm_io_bus_write(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
,
173 int len
, const void *val
);
174 int kvm_io_bus_write_cookie(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
,
175 int len
, const void *val
, long cookie
);
176 int kvm_io_bus_read(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
, int len
,
178 int kvm_io_bus_register_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
,
179 int len
, struct kvm_io_device
*dev
);
180 int kvm_io_bus_unregister_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
181 struct kvm_io_device
*dev
);
183 #ifdef CONFIG_KVM_ASYNC_PF
184 struct kvm_async_pf
{
185 struct work_struct work
;
186 struct list_head link
;
187 struct list_head queue
;
188 struct kvm_vcpu
*vcpu
;
189 struct mm_struct
*mm
;
192 struct kvm_arch_async_pf arch
;
196 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu
*vcpu
);
197 void kvm_check_async_pf_completion(struct kvm_vcpu
*vcpu
);
198 int kvm_setup_async_pf(struct kvm_vcpu
*vcpu
, gva_t gva
, unsigned long hva
,
199 struct kvm_arch_async_pf
*arch
);
200 int kvm_async_pf_wakeup_all(struct kvm_vcpu
*vcpu
);
207 READING_SHADOW_PAGE_TABLES
,
211 * Sometimes a large or cross-page mmio needs to be broken up into separate
212 * exits for userspace servicing.
214 struct kvm_mmio_fragment
{
222 #ifdef CONFIG_PREEMPT_NOTIFIERS
223 struct preempt_notifier preempt_notifier
;
229 unsigned long requests
;
230 unsigned long guest_debug
;
236 int guest_fpu_loaded
, guest_xcr0_loaded
;
237 wait_queue_head_t wq
;
241 struct kvm_vcpu_stat stat
;
243 #ifdef CONFIG_HAS_IOMEM
245 int mmio_read_completed
;
247 int mmio_cur_fragment
;
248 int mmio_nr_fragments
;
249 struct kvm_mmio_fragment mmio_fragments
[KVM_MAX_MMIO_FRAGMENTS
];
252 #ifdef CONFIG_KVM_ASYNC_PF
255 struct list_head queue
;
256 struct list_head done
;
261 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
263 * Cpu relax intercept or pause loop exit optimization
264 * in_spin_loop: set when a vcpu does a pause loop exit
265 * or cpu relax intercepted.
266 * dy_eligible: indicates whether vcpu is eligible for directed yield.
274 struct kvm_vcpu_arch arch
;
277 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu
*vcpu
)
279 return cmpxchg(&vcpu
->mode
, IN_GUEST_MODE
, EXITING_GUEST_MODE
);
283 * Some of the bitops functions do not support too long bitmaps.
284 * This number must be determined not to exceed such limits.
286 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
288 struct kvm_memory_slot
{
290 unsigned long npages
;
291 unsigned long *dirty_bitmap
;
292 struct kvm_arch_memory_slot arch
;
293 unsigned long userspace_addr
;
298 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot
*memslot
)
300 return ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
303 struct kvm_s390_adapter_int
{
311 struct kvm_kernel_irq_routing_entry
{
314 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
315 struct kvm
*kvm
, int irq_source_id
, int level
,
323 struct kvm_s390_adapter_int adapter
;
325 struct hlist_node link
;
328 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
330 struct kvm_irq_routing_table
{
331 int chip
[KVM_NR_IRQCHIPS
][KVM_IRQCHIP_NUM_PINS
];
332 struct kvm_kernel_irq_routing_entry
*rt_entries
;
335 * Array indexed by gsi. Each entry contains list of irq chips
336 * the gsi is connected to.
338 struct hlist_head map
[0];
343 struct kvm_irq_routing_table
{};
347 #ifndef KVM_PRIVATE_MEM_SLOTS
348 #define KVM_PRIVATE_MEM_SLOTS 0
351 #ifndef KVM_MEM_SLOTS_NUM
352 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
357 * memslots are not sorted by id anymore, please use id_to_memslot()
358 * to get the memslot by its id.
360 struct kvm_memslots
{
362 struct kvm_memory_slot memslots
[KVM_MEM_SLOTS_NUM
];
363 /* The mapping table from slot id to the index in memslots[]. */
364 short id_to_index
[KVM_MEM_SLOTS_NUM
];
369 struct mutex slots_lock
;
370 struct mm_struct
*mm
; /* userspace tied to this vm */
371 struct kvm_memslots
*memslots
;
372 struct srcu_struct srcu
;
373 struct srcu_struct irq_srcu
;
374 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
377 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
378 atomic_t online_vcpus
;
379 int last_boosted_vcpu
;
380 struct list_head vm_list
;
382 struct kvm_io_bus
*buses
[KVM_NR_BUSES
];
383 #ifdef CONFIG_HAVE_KVM_EVENTFD
386 struct list_head items
;
387 struct list_head resampler_list
;
388 struct mutex resampler_lock
;
390 struct list_head ioeventfds
;
392 struct kvm_vm_stat stat
;
393 struct kvm_arch arch
;
394 atomic_t users_count
;
395 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
396 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
397 spinlock_t ring_lock
;
398 struct list_head coalesced_zones
;
401 struct mutex irq_lock
;
402 #ifdef CONFIG_HAVE_KVM_IRQCHIP
404 * Update side is protected by irq_lock and,
405 * if configured, irqfds.lock.
407 struct kvm_irq_routing_table __rcu
*irq_routing
;
408 struct hlist_head mask_notifier_list
;
409 struct hlist_head irq_ack_notifier_list
;
412 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
413 struct mmu_notifier mmu_notifier
;
414 unsigned long mmu_notifier_seq
;
415 long mmu_notifier_count
;
418 struct list_head devices
;
421 #define kvm_err(fmt, ...) \
422 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
423 #define kvm_info(fmt, ...) \
424 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
425 #define kvm_debug(fmt, ...) \
426 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
427 #define kvm_pr_unimpl(fmt, ...) \
428 pr_err_ratelimited("kvm [%i]: " fmt, \
429 task_tgid_nr(current), ## __VA_ARGS__)
431 /* The guest did something we don't support. */
432 #define vcpu_unimpl(vcpu, fmt, ...) \
433 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
435 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
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 #define kvm_for_each_memslot(memslot, slots) \
448 for (memslot = &slots->memslots[0]; \
449 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
452 int kvm_vcpu_init(struct kvm_vcpu
*vcpu
, struct kvm
*kvm
, unsigned id
);
453 void kvm_vcpu_uninit(struct kvm_vcpu
*vcpu
);
455 int __must_check
vcpu_load(struct kvm_vcpu
*vcpu
);
456 void vcpu_put(struct kvm_vcpu
*vcpu
);
458 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
459 int kvm_irqfd_init(void);
460 void kvm_irqfd_exit(void);
462 static inline int kvm_irqfd_init(void)
467 static inline void kvm_irqfd_exit(void)
471 int kvm_init(void *opaque
, unsigned vcpu_size
, unsigned vcpu_align
,
472 struct module
*module
);
475 void kvm_get_kvm(struct kvm
*kvm
);
476 void kvm_put_kvm(struct kvm
*kvm
);
478 static inline struct kvm_memslots
*kvm_memslots(struct kvm
*kvm
)
480 return rcu_dereference_check(kvm
->memslots
,
481 srcu_read_lock_held(&kvm
->srcu
)
482 || lockdep_is_held(&kvm
->slots_lock
));
485 static inline struct kvm_memory_slot
*
486 id_to_memslot(struct kvm_memslots
*slots
, int id
)
488 int index
= slots
->id_to_index
[id
];
489 struct kvm_memory_slot
*slot
;
491 slot
= &slots
->memslots
[index
];
493 WARN_ON(slot
->id
!= id
);
498 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
499 * - create a new memory slot
500 * - delete an existing memory slot
501 * - modify an existing memory slot
502 * -- move it in the guest physical memory space
503 * -- just change its flags
505 * Since flags can be changed by some of these operations, the following
506 * differentiation is the best we can do for __kvm_set_memory_region():
515 int kvm_set_memory_region(struct kvm
*kvm
,
516 struct kvm_userspace_memory_region
*mem
);
517 int __kvm_set_memory_region(struct kvm
*kvm
,
518 struct kvm_userspace_memory_region
*mem
);
519 void kvm_arch_free_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*free
,
520 struct kvm_memory_slot
*dont
);
521 int kvm_arch_create_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*slot
,
522 unsigned long npages
);
523 void kvm_arch_memslots_updated(struct kvm
*kvm
);
524 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
525 struct kvm_memory_slot
*memslot
,
526 struct kvm_userspace_memory_region
*mem
,
527 enum kvm_mr_change change
);
528 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
529 struct kvm_userspace_memory_region
*mem
,
530 const struct kvm_memory_slot
*old
,
531 enum kvm_mr_change change
);
532 bool kvm_largepages_enabled(void);
533 void kvm_disable_largepages(void);
534 /* flush all memory translations */
535 void kvm_arch_flush_shadow_all(struct kvm
*kvm
);
536 /* flush memory translations pointing to 'slot' */
537 void kvm_arch_flush_shadow_memslot(struct kvm
*kvm
,
538 struct kvm_memory_slot
*slot
);
540 int gfn_to_page_many_atomic(struct kvm
*kvm
, gfn_t gfn
, struct page
**pages
,
543 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
544 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
545 unsigned long gfn_to_hva_prot(struct kvm
*kvm
, gfn_t gfn
, bool *writable
);
546 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
547 void kvm_release_page_clean(struct page
*page
);
548 void kvm_release_page_dirty(struct page
*page
);
549 void kvm_set_page_accessed(struct page
*page
);
551 pfn_t
gfn_to_pfn_atomic(struct kvm
*kvm
, gfn_t gfn
);
552 pfn_t
gfn_to_pfn_async(struct kvm
*kvm
, gfn_t gfn
, bool *async
,
553 bool write_fault
, bool *writable
);
554 pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
555 pfn_t
gfn_to_pfn_prot(struct kvm
*kvm
, gfn_t gfn
, bool write_fault
,
557 pfn_t
gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
558 pfn_t
gfn_to_pfn_memslot_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
);
560 void kvm_release_pfn_clean(pfn_t pfn
);
561 void kvm_set_pfn_dirty(pfn_t pfn
);
562 void kvm_set_pfn_accessed(pfn_t pfn
);
563 void kvm_get_pfn(pfn_t pfn
);
565 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
567 int kvm_read_guest_atomic(struct kvm
*kvm
, gpa_t gpa
, void *data
,
569 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
570 int kvm_read_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
571 void *data
, unsigned long len
);
572 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
573 int offset
, int len
);
574 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
576 int kvm_write_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
577 void *data
, unsigned long len
);
578 int kvm_gfn_to_hva_cache_init(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
579 gpa_t gpa
, unsigned long len
);
580 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
581 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
582 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
583 int kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
584 unsigned long kvm_host_page_size(struct kvm
*kvm
, gfn_t gfn
);
585 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
587 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
588 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
589 int kvm_vcpu_yield_to(struct kvm_vcpu
*target
);
590 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
);
591 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
592 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
594 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
595 void kvm_reload_remote_mmus(struct kvm
*kvm
);
596 void kvm_make_mclock_inprogress_request(struct kvm
*kvm
);
597 void kvm_make_scan_ioapic_request(struct kvm
*kvm
);
599 long kvm_arch_dev_ioctl(struct file
*filp
,
600 unsigned int ioctl
, unsigned long arg
);
601 long kvm_arch_vcpu_ioctl(struct file
*filp
,
602 unsigned int ioctl
, unsigned long arg
);
603 int kvm_arch_vcpu_fault(struct kvm_vcpu
*vcpu
, struct vm_fault
*vmf
);
605 int kvm_dev_ioctl_check_extension(long ext
);
607 int kvm_get_dirty_log(struct kvm
*kvm
,
608 struct kvm_dirty_log
*log
, int *is_dirty
);
609 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
610 struct kvm_dirty_log
*log
);
612 int kvm_vm_ioctl_irq_line(struct kvm
*kvm
, struct kvm_irq_level
*irq_level
,
614 long kvm_arch_vm_ioctl(struct file
*filp
,
615 unsigned int ioctl
, unsigned long arg
);
617 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
618 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
620 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
621 struct kvm_translation
*tr
);
623 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
624 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
625 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
626 struct kvm_sregs
*sregs
);
627 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
628 struct kvm_sregs
*sregs
);
629 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
630 struct kvm_mp_state
*mp_state
);
631 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
632 struct kvm_mp_state
*mp_state
);
633 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
634 struct kvm_guest_debug
*dbg
);
635 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
637 int kvm_arch_init(void *opaque
);
638 void kvm_arch_exit(void);
640 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
);
641 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
);
643 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
);
644 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
645 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
646 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
);
647 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
);
648 int kvm_arch_vcpu_postcreate(struct kvm_vcpu
*vcpu
);
649 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
651 int kvm_arch_hardware_enable(void *garbage
);
652 void kvm_arch_hardware_disable(void *garbage
);
653 int kvm_arch_hardware_setup(void);
654 void kvm_arch_hardware_unsetup(void);
655 void kvm_arch_check_processor_compat(void *rtn
);
656 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
657 int kvm_arch_vcpu_should_kick(struct kvm_vcpu
*vcpu
);
659 void *kvm_kvzalloc(unsigned long size
);
660 void kvm_kvfree(const void *addr
);
662 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
663 static inline struct kvm
*kvm_arch_alloc_vm(void)
665 return kzalloc(sizeof(struct kvm
), GFP_KERNEL
);
668 static inline void kvm_arch_free_vm(struct kvm
*kvm
)
674 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
675 void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
);
676 void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
);
677 bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
);
679 static inline void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
)
683 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
)
687 static inline bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
)
693 static inline wait_queue_head_t
*kvm_arch_vcpu_wq(struct kvm_vcpu
*vcpu
)
695 #ifdef __KVM_HAVE_ARCH_WQP
696 return vcpu
->arch
.wqp
;
702 int kvm_arch_init_vm(struct kvm
*kvm
, unsigned long type
);
703 void kvm_arch_destroy_vm(struct kvm
*kvm
);
704 void kvm_arch_sync_events(struct kvm
*kvm
);
706 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
707 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
709 bool kvm_is_mmio_pfn(pfn_t pfn
);
711 struct kvm_irq_ack_notifier
{
712 struct hlist_node link
;
714 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
717 struct kvm_assigned_dev_kernel
{
718 struct kvm_irq_ack_notifier ack_notifier
;
719 struct list_head list
;
724 unsigned int entries_nr
;
726 bool host_irq_disabled
;
728 struct msix_entry
*host_msix_entries
;
730 struct msix_entry
*guest_msix_entries
;
731 unsigned long irq_requested_type
;
736 spinlock_t intx_lock
;
737 spinlock_t intx_mask_lock
;
739 struct pci_saved_state
*pci_saved_state
;
742 struct kvm_irq_mask_notifier
{
743 void (*func
)(struct kvm_irq_mask_notifier
*kimn
, bool masked
);
745 struct hlist_node link
;
748 void kvm_register_irq_mask_notifier(struct kvm
*kvm
, int irq
,
749 struct kvm_irq_mask_notifier
*kimn
);
750 void kvm_unregister_irq_mask_notifier(struct kvm
*kvm
, int irq
,
751 struct kvm_irq_mask_notifier
*kimn
);
752 void kvm_fire_mask_notifiers(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
,
755 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
,
757 int kvm_set_irq_inatomic(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
);
758 int kvm_set_msi(struct kvm_kernel_irq_routing_entry
*irq_entry
, struct kvm
*kvm
,
759 int irq_source_id
, int level
, bool line_status
);
760 bool kvm_irq_has_notifier(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
761 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
762 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
763 struct kvm_irq_ack_notifier
*kian
);
764 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
765 struct kvm_irq_ack_notifier
*kian
);
766 int kvm_request_irq_source_id(struct kvm
*kvm
);
767 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
769 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
770 int kvm_iommu_map_pages(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
771 void kvm_iommu_unmap_pages(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
772 int kvm_iommu_map_guest(struct kvm
*kvm
);
773 int kvm_iommu_unmap_guest(struct kvm
*kvm
);
774 int kvm_assign_device(struct kvm
*kvm
,
775 struct kvm_assigned_dev_kernel
*assigned_dev
);
776 int kvm_deassign_device(struct kvm
*kvm
,
777 struct kvm_assigned_dev_kernel
*assigned_dev
);
779 static inline int kvm_iommu_map_pages(struct kvm
*kvm
,
780 struct kvm_memory_slot
*slot
)
785 static inline void kvm_iommu_unmap_pages(struct kvm
*kvm
,
786 struct kvm_memory_slot
*slot
)
790 static inline int kvm_iommu_unmap_guest(struct kvm
*kvm
)
796 static inline void kvm_guest_enter(void)
800 BUG_ON(preemptible());
802 local_irq_save(flags
);
804 local_irq_restore(flags
);
806 /* KVM does not hold any references to rcu protected data when it
807 * switches CPU into a guest mode. In fact switching to a guest mode
808 * is very similar to exiting to userspace from rcu point of view. In
809 * addition CPU may stay in a guest mode for quite a long time (up to
810 * one time slice). Lets treat guest mode as quiescent state, just like
811 * we do with user-mode execution.
813 rcu_virt_note_context_switch(smp_processor_id());
816 static inline void kvm_guest_exit(void)
820 local_irq_save(flags
);
822 local_irq_restore(flags
);
826 * search_memslots() and __gfn_to_memslot() are here because they are
827 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
828 * gfn_to_memslot() itself isn't here as an inline because that would
829 * bloat other code too much.
831 static inline struct kvm_memory_slot
*
832 search_memslots(struct kvm_memslots
*slots
, gfn_t gfn
)
834 struct kvm_memory_slot
*memslot
;
836 kvm_for_each_memslot(memslot
, slots
)
837 if (gfn
>= memslot
->base_gfn
&&
838 gfn
< memslot
->base_gfn
+ memslot
->npages
)
844 static inline struct kvm_memory_slot
*
845 __gfn_to_memslot(struct kvm_memslots
*slots
, gfn_t gfn
)
847 return search_memslots(slots
, gfn
);
850 static inline unsigned long
851 __gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
)
853 return slot
->userspace_addr
+ (gfn
- slot
->base_gfn
) * PAGE_SIZE
;
856 static inline int memslot_id(struct kvm
*kvm
, gfn_t gfn
)
858 return gfn_to_memslot(kvm
, gfn
)->id
;
862 hva_to_gfn_memslot(unsigned long hva
, struct kvm_memory_slot
*slot
)
864 gfn_t gfn_offset
= (hva
- slot
->userspace_addr
) >> PAGE_SHIFT
;
866 return slot
->base_gfn
+ gfn_offset
;
869 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
871 return (gpa_t
)gfn
<< PAGE_SHIFT
;
874 static inline gfn_t
gpa_to_gfn(gpa_t gpa
)
876 return (gfn_t
)(gpa
>> PAGE_SHIFT
);
879 static inline hpa_t
pfn_to_hpa(pfn_t pfn
)
881 return (hpa_t
)pfn
<< PAGE_SHIFT
;
884 static inline bool kvm_is_error_gpa(struct kvm
*kvm
, gpa_t gpa
)
886 unsigned long hva
= gfn_to_hva(kvm
, gpa_to_gfn(gpa
));
888 return kvm_is_error_hva(hva
);
891 static inline void kvm_migrate_timers(struct kvm_vcpu
*vcpu
)
893 set_bit(KVM_REQ_MIGRATE_TIMER
, &vcpu
->requests
);
901 struct kvm_stats_debugfs_item
{
904 enum kvm_stat_kind kind
;
905 struct dentry
*dentry
;
907 extern struct kvm_stats_debugfs_item debugfs_entries
[];
908 extern struct dentry
*kvm_debugfs_dir
;
910 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
911 static inline int mmu_notifier_retry(struct kvm
*kvm
, unsigned long mmu_seq
)
913 if (unlikely(kvm
->mmu_notifier_count
))
916 * Ensure the read of mmu_notifier_count happens before the read
917 * of mmu_notifier_seq. This interacts with the smp_wmb() in
918 * mmu_notifier_invalidate_range_end to make sure that the caller
919 * either sees the old (non-zero) value of mmu_notifier_count or
920 * the new (incremented) value of mmu_notifier_seq.
921 * PowerPC Book3s HV KVM calls this under a per-page lock
922 * rather than under kvm->mmu_lock, for scalability, so
923 * can't rely on kvm->mmu_lock to keep things ordered.
926 if (kvm
->mmu_notifier_seq
!= mmu_seq
)
932 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
935 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
937 #define KVM_MAX_IRQ_ROUTES 1024
940 int kvm_setup_default_irq_routing(struct kvm
*kvm
);
941 int kvm_set_irq_routing(struct kvm
*kvm
,
942 const struct kvm_irq_routing_entry
*entries
,
945 int kvm_set_routing_entry(struct kvm_irq_routing_table
*rt
,
946 struct kvm_kernel_irq_routing_entry
*e
,
947 const struct kvm_irq_routing_entry
*ue
);
948 void kvm_free_irq_routing(struct kvm
*kvm
);
950 int kvm_send_userspace_msi(struct kvm
*kvm
, struct kvm_msi
*msi
);
954 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
958 #ifdef CONFIG_HAVE_KVM_EVENTFD
960 void kvm_eventfd_init(struct kvm
*kvm
);
961 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
963 #ifdef CONFIG_HAVE_KVM_IRQCHIP
964 int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
);
965 void kvm_irqfd_release(struct kvm
*kvm
);
966 void kvm_irq_routing_update(struct kvm
*, struct kvm_irq_routing_table
*);
968 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
973 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
978 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
980 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
985 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
987 #ifdef CONFIG_HAVE_KVM_IRQCHIP
988 static inline void kvm_irq_routing_update(struct kvm
*kvm
,
989 struct kvm_irq_routing_table
*irq_rt
)
991 rcu_assign_pointer(kvm
->irq_routing
, irq_rt
);
995 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
1000 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1002 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
1003 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu
*vcpu
)
1005 return vcpu
->kvm
->bsp_vcpu_id
== vcpu
->vcpu_id
;
1008 bool kvm_vcpu_compatible(struct kvm_vcpu
*vcpu
);
1012 static inline bool kvm_vcpu_compatible(struct kvm_vcpu
*vcpu
) { return true; }
1016 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
1018 long kvm_vm_ioctl_assigned_device(struct kvm
*kvm
, unsigned ioctl
,
1021 void kvm_free_all_assigned_devices(struct kvm
*kvm
);
1025 static inline long kvm_vm_ioctl_assigned_device(struct kvm
*kvm
, unsigned ioctl
,
1031 static inline void kvm_free_all_assigned_devices(struct kvm
*kvm
) {}
1035 static inline void kvm_make_request(int req
, struct kvm_vcpu
*vcpu
)
1037 set_bit(req
, &vcpu
->requests
);
1040 static inline bool kvm_check_request(int req
, struct kvm_vcpu
*vcpu
)
1042 if (test_bit(req
, &vcpu
->requests
)) {
1043 clear_bit(req
, &vcpu
->requests
);
1050 extern bool kvm_rebooting
;
1052 struct kvm_device_ops
;
1055 struct kvm_device_ops
*ops
;
1058 struct list_head vm_node
;
1061 /* create, destroy, and name are mandatory */
1062 struct kvm_device_ops
{
1064 int (*create
)(struct kvm_device
*dev
, u32 type
);
1067 * Destroy is responsible for freeing dev.
1069 * Destroy may be called before or after destructors are called
1070 * on emulated I/O regions, depending on whether a reference is
1071 * held by a vcpu or other kvm component that gets destroyed
1072 * after the emulated I/O.
1074 void (*destroy
)(struct kvm_device
*dev
);
1076 int (*set_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1077 int (*get_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1078 int (*has_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1079 long (*ioctl
)(struct kvm_device
*dev
, unsigned int ioctl
,
1083 void kvm_device_get(struct kvm_device
*dev
);
1084 void kvm_device_put(struct kvm_device
*dev
);
1085 struct kvm_device
*kvm_device_from_filp(struct file
*filp
);
1087 extern struct kvm_device_ops kvm_mpic_ops
;
1088 extern struct kvm_device_ops kvm_xics_ops
;
1089 extern struct kvm_device_ops kvm_vfio_ops
;
1090 extern struct kvm_device_ops kvm_arm_vgic_v2_ops
;
1091 extern struct kvm_device_ops kvm_flic_ops
;
1093 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1095 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1097 vcpu
->spin_loop
.in_spin_loop
= val
;
1099 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1101 vcpu
->spin_loop
.dy_eligible
= val
;
1104 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1106 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1110 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1113 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */