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>
17 #include <linux/preempt.h>
18 #include <linux/msi.h>
19 #include <asm/signal.h>
21 #include <linux/kvm.h>
22 #include <linux/kvm_para.h>
24 #include <linux/kvm_types.h>
26 #include <asm/kvm_host.h>
29 * vcpu->requests bit members
31 #define KVM_REQ_TLB_FLUSH 0
32 #define KVM_REQ_MIGRATE_TIMER 1
33 #define KVM_REQ_REPORT_TPR_ACCESS 2
34 #define KVM_REQ_MMU_RELOAD 3
35 #define KVM_REQ_TRIPLE_FAULT 4
36 #define KVM_REQ_PENDING_TIMER 5
37 #define KVM_REQ_UNHALT 6
38 #define KVM_REQ_MMU_SYNC 7
39 #define KVM_REQ_KVMCLOCK_UPDATE 8
40 #define KVM_REQ_KICK 9
41 #define KVM_REQ_DEACTIVATE_FPU 10
43 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
47 extern struct kmem_cache
*kvm_vcpu_cache
;
50 * It would be nice to use something smarter than a linear search, TBD...
51 * Thankfully we dont expect many devices to register (famous last words :),
52 * so until then it will suffice. At least its abstracted so we can change
57 #define NR_IOBUS_DEVS 200
58 struct kvm_io_device
*devs
[NR_IOBUS_DEVS
];
67 int kvm_io_bus_write(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
,
68 int len
, const void *val
);
69 int kvm_io_bus_read(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
, int len
,
71 int kvm_io_bus_register_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
72 struct kvm_io_device
*dev
);
73 int kvm_io_bus_unregister_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
74 struct kvm_io_device
*dev
);
78 #ifdef CONFIG_PREEMPT_NOTIFIERS
79 struct preempt_notifier preempt_notifier
;
86 unsigned long requests
;
87 unsigned long guest_debug
;
91 int guest_fpu_loaded
, guest_xcr0_loaded
;
95 struct kvm_vcpu_stat stat
;
97 #ifdef CONFIG_HAS_IOMEM
99 int mmio_read_completed
;
102 unsigned char mmio_data
[8];
103 gpa_t mmio_phys_addr
;
106 struct kvm_vcpu_arch arch
;
110 * Some of the bitops functions do not support too long bitmaps.
111 * This number must be determined not to exceed such limits.
113 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
115 struct kvm_memory_slot
{
117 unsigned long npages
;
120 unsigned long *dirty_bitmap
;
122 unsigned long rmap_pde
;
124 } *lpage_info
[KVM_NR_PAGE_SIZES
- 1];
125 unsigned long userspace_addr
;
130 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot
*memslot
)
132 return ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
135 struct kvm_kernel_irq_routing_entry
{
138 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
139 struct kvm
*kvm
, int irq_source_id
, int level
);
147 struct hlist_node link
;
150 #ifdef __KVM_HAVE_IOAPIC
152 struct kvm_irq_routing_table
{
153 int chip
[KVM_NR_IRQCHIPS
][KVM_IOAPIC_NUM_PINS
];
154 struct kvm_kernel_irq_routing_entry
*rt_entries
;
157 * Array indexed by gsi. Each entry contains list of irq chips
158 * the gsi is connected to.
160 struct hlist_head map
[0];
165 struct kvm_irq_routing_table
{};
169 struct kvm_memslots
{
171 struct kvm_memory_slot memslots
[KVM_MEMORY_SLOTS
+
172 KVM_PRIVATE_MEM_SLOTS
];
177 raw_spinlock_t requests_lock
;
178 struct mutex slots_lock
;
179 struct mm_struct
*mm
; /* userspace tied to this vm */
180 struct kvm_memslots
*memslots
;
181 struct srcu_struct srcu
;
182 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
184 struct kvm_vcpu
*bsp_vcpu
;
186 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
187 atomic_t online_vcpus
;
188 struct list_head vm_list
;
190 struct kvm_io_bus
*buses
[KVM_NR_BUSES
];
191 #ifdef CONFIG_HAVE_KVM_EVENTFD
194 struct list_head items
;
196 struct list_head ioeventfds
;
198 struct kvm_vm_stat stat
;
199 struct kvm_arch arch
;
200 atomic_t users_count
;
201 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
202 struct kvm_coalesced_mmio_dev
*coalesced_mmio_dev
;
203 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
206 struct mutex irq_lock
;
207 #ifdef CONFIG_HAVE_KVM_IRQCHIP
208 struct kvm_irq_routing_table __rcu
*irq_routing
;
209 struct hlist_head mask_notifier_list
;
210 struct hlist_head irq_ack_notifier_list
;
213 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
214 struct mmu_notifier mmu_notifier
;
215 unsigned long mmu_notifier_seq
;
216 long mmu_notifier_count
;
220 /* The guest did something we don't support. */
221 #define pr_unimpl(vcpu, fmt, ...) \
223 if (printk_ratelimit()) \
224 printk(KERN_ERR "kvm: %i: cpu%i " fmt, \
225 current->tgid, (vcpu)->vcpu_id , ## __VA_ARGS__); \
228 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
229 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
231 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
234 return kvm
->vcpus
[i
];
237 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
238 for (idx = 0, vcpup = kvm_get_vcpu(kvm, idx); \
239 idx < atomic_read(&kvm->online_vcpus) && vcpup; \
240 vcpup = kvm_get_vcpu(kvm, ++idx))
242 int kvm_vcpu_init(struct kvm_vcpu
*vcpu
, struct kvm
*kvm
, unsigned id
);
243 void kvm_vcpu_uninit(struct kvm_vcpu
*vcpu
);
245 void vcpu_load(struct kvm_vcpu
*vcpu
);
246 void vcpu_put(struct kvm_vcpu
*vcpu
);
248 int kvm_init(void *opaque
, unsigned vcpu_size
, unsigned vcpu_align
,
249 struct module
*module
);
252 void kvm_get_kvm(struct kvm
*kvm
);
253 void kvm_put_kvm(struct kvm
*kvm
);
255 static inline struct kvm_memslots
*kvm_memslots(struct kvm
*kvm
)
257 return rcu_dereference_check(kvm
->memslots
,
258 srcu_read_lock_held(&kvm
->srcu
)
259 || lockdep_is_held(&kvm
->slots_lock
));
262 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
263 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
264 static inline int is_error_hpa(hpa_t hpa
) { return hpa
>> HPA_MSB
; }
266 extern struct page
*bad_page
;
267 extern pfn_t bad_pfn
;
269 int is_error_page(struct page
*page
);
270 int is_error_pfn(pfn_t pfn
);
271 int is_hwpoison_pfn(pfn_t pfn
);
272 int is_fault_pfn(pfn_t pfn
);
273 int kvm_is_error_hva(unsigned long addr
);
274 int kvm_set_memory_region(struct kvm
*kvm
,
275 struct kvm_userspace_memory_region
*mem
,
277 int __kvm_set_memory_region(struct kvm
*kvm
,
278 struct kvm_userspace_memory_region
*mem
,
280 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
281 struct kvm_memory_slot
*memslot
,
282 struct kvm_memory_slot old
,
283 struct kvm_userspace_memory_region
*mem
,
285 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
286 struct kvm_userspace_memory_region
*mem
,
287 struct kvm_memory_slot old
,
289 void kvm_disable_largepages(void);
290 void kvm_arch_flush_shadow(struct kvm
*kvm
);
292 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
293 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
294 void kvm_release_page_clean(struct page
*page
);
295 void kvm_release_page_dirty(struct page
*page
);
296 void kvm_set_page_dirty(struct page
*page
);
297 void kvm_set_page_accessed(struct page
*page
);
299 pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
300 pfn_t
gfn_to_pfn_memslot(struct kvm
*kvm
,
301 struct kvm_memory_slot
*slot
, gfn_t gfn
);
302 int memslot_id(struct kvm
*kvm
, gfn_t gfn
);
303 void kvm_release_pfn_dirty(pfn_t
);
304 void kvm_release_pfn_clean(pfn_t pfn
);
305 void kvm_set_pfn_dirty(pfn_t pfn
);
306 void kvm_set_pfn_accessed(pfn_t pfn
);
307 void kvm_get_pfn(pfn_t pfn
);
309 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
311 int kvm_read_guest_atomic(struct kvm
*kvm
, gpa_t gpa
, void *data
,
313 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
314 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
315 int offset
, int len
);
316 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
318 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
319 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
320 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
321 int kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
322 unsigned long kvm_host_page_size(struct kvm
*kvm
, gfn_t gfn
);
323 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
325 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
326 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
);
327 void kvm_resched(struct kvm_vcpu
*vcpu
);
328 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
329 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
330 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
331 void kvm_reload_remote_mmus(struct kvm
*kvm
);
333 long kvm_arch_dev_ioctl(struct file
*filp
,
334 unsigned int ioctl
, unsigned long arg
);
335 long kvm_arch_vcpu_ioctl(struct file
*filp
,
336 unsigned int ioctl
, unsigned long arg
);
338 int kvm_dev_ioctl_check_extension(long ext
);
340 int kvm_get_dirty_log(struct kvm
*kvm
,
341 struct kvm_dirty_log
*log
, int *is_dirty
);
342 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
343 struct kvm_dirty_log
*log
);
345 int kvm_vm_ioctl_set_memory_region(struct kvm
*kvm
,
347 kvm_userspace_memory_region
*mem
,
349 long kvm_arch_vm_ioctl(struct file
*filp
,
350 unsigned int ioctl
, unsigned long arg
);
352 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
353 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
355 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
356 struct kvm_translation
*tr
);
358 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
359 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
360 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
361 struct kvm_sregs
*sregs
);
362 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
363 struct kvm_sregs
*sregs
);
364 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
365 struct kvm_mp_state
*mp_state
);
366 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
367 struct kvm_mp_state
*mp_state
);
368 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
369 struct kvm_guest_debug
*dbg
);
370 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
372 int kvm_arch_init(void *opaque
);
373 void kvm_arch_exit(void);
375 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
);
376 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
);
378 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
);
379 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
380 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
381 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
);
382 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
);
383 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
385 int kvm_arch_vcpu_reset(struct kvm_vcpu
*vcpu
);
386 int kvm_arch_hardware_enable(void *garbage
);
387 void kvm_arch_hardware_disable(void *garbage
);
388 int kvm_arch_hardware_setup(void);
389 void kvm_arch_hardware_unsetup(void);
390 void kvm_arch_check_processor_compat(void *rtn
);
391 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
393 void kvm_free_physmem(struct kvm
*kvm
);
395 struct kvm
*kvm_arch_create_vm(void);
396 void kvm_arch_destroy_vm(struct kvm
*kvm
);
397 void kvm_free_all_assigned_devices(struct kvm
*kvm
);
398 void kvm_arch_sync_events(struct kvm
*kvm
);
400 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
401 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
403 int kvm_is_mmio_pfn(pfn_t pfn
);
405 struct kvm_irq_ack_notifier
{
406 struct hlist_node link
;
408 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
411 #define KVM_ASSIGNED_MSIX_PENDING 0x1
412 struct kvm_guest_msix_entry
{
418 struct kvm_assigned_dev_kernel
{
419 struct kvm_irq_ack_notifier ack_notifier
;
420 struct work_struct interrupt_work
;
421 struct list_head list
;
426 unsigned int entries_nr
;
428 bool host_irq_disabled
;
429 struct msix_entry
*host_msix_entries
;
431 struct kvm_guest_msix_entry
*guest_msix_entries
;
432 unsigned long irq_requested_type
;
437 spinlock_t assigned_dev_lock
;
440 struct kvm_irq_mask_notifier
{
441 void (*func
)(struct kvm_irq_mask_notifier
*kimn
, bool masked
);
443 struct hlist_node link
;
446 void kvm_register_irq_mask_notifier(struct kvm
*kvm
, int irq
,
447 struct kvm_irq_mask_notifier
*kimn
);
448 void kvm_unregister_irq_mask_notifier(struct kvm
*kvm
, int irq
,
449 struct kvm_irq_mask_notifier
*kimn
);
450 void kvm_fire_mask_notifiers(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
,
453 #ifdef __KVM_HAVE_IOAPIC
454 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic
*ioapic
,
455 union kvm_ioapic_redirect_entry
*entry
,
456 unsigned long *deliver_bitmask
);
458 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
);
459 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
460 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
461 struct kvm_irq_ack_notifier
*kian
);
462 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
463 struct kvm_irq_ack_notifier
*kian
);
464 int kvm_request_irq_source_id(struct kvm
*kvm
);
465 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
467 /* For vcpu->arch.iommu_flags */
468 #define KVM_IOMMU_CACHE_COHERENCY 0x1
470 #ifdef CONFIG_IOMMU_API
471 int kvm_iommu_map_pages(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
472 int kvm_iommu_map_guest(struct kvm
*kvm
);
473 int kvm_iommu_unmap_guest(struct kvm
*kvm
);
474 int kvm_assign_device(struct kvm
*kvm
,
475 struct kvm_assigned_dev_kernel
*assigned_dev
);
476 int kvm_deassign_device(struct kvm
*kvm
,
477 struct kvm_assigned_dev_kernel
*assigned_dev
);
478 #else /* CONFIG_IOMMU_API */
479 static inline int kvm_iommu_map_pages(struct kvm
*kvm
,
481 unsigned long npages
)
486 static inline int kvm_iommu_map_guest(struct kvm
*kvm
)
491 static inline int kvm_iommu_unmap_guest(struct kvm
*kvm
)
496 static inline int kvm_assign_device(struct kvm
*kvm
,
497 struct kvm_assigned_dev_kernel
*assigned_dev
)
502 static inline int kvm_deassign_device(struct kvm
*kvm
,
503 struct kvm_assigned_dev_kernel
*assigned_dev
)
507 #endif /* CONFIG_IOMMU_API */
509 static inline void kvm_guest_enter(void)
511 account_system_vtime(current
);
512 current
->flags
|= PF_VCPU
;
515 static inline void kvm_guest_exit(void)
517 account_system_vtime(current
);
518 current
->flags
&= ~PF_VCPU
;
521 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
523 return (gpa_t
)gfn
<< PAGE_SHIFT
;
526 static inline hpa_t
pfn_to_hpa(pfn_t pfn
)
528 return (hpa_t
)pfn
<< PAGE_SHIFT
;
531 static inline void kvm_migrate_timers(struct kvm_vcpu
*vcpu
)
533 set_bit(KVM_REQ_MIGRATE_TIMER
, &vcpu
->requests
);
541 struct kvm_stats_debugfs_item
{
544 enum kvm_stat_kind kind
;
545 struct dentry
*dentry
;
547 extern struct kvm_stats_debugfs_item debugfs_entries
[];
548 extern struct dentry
*kvm_debugfs_dir
;
550 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
551 static inline int mmu_notifier_retry(struct kvm_vcpu
*vcpu
, unsigned long mmu_seq
)
553 if (unlikely(vcpu
->kvm
->mmu_notifier_count
))
556 * Both reads happen under the mmu_lock and both values are
557 * modified under mmu_lock, so there's no need of smb_rmb()
558 * here in between, otherwise mmu_notifier_count should be
559 * read before mmu_notifier_seq, see
560 * mmu_notifier_invalidate_range_end write side.
562 if (vcpu
->kvm
->mmu_notifier_seq
!= mmu_seq
)
568 #ifdef CONFIG_HAVE_KVM_IRQCHIP
570 #define KVM_MAX_IRQ_ROUTES 1024
572 int kvm_setup_default_irq_routing(struct kvm
*kvm
);
573 int kvm_set_irq_routing(struct kvm
*kvm
,
574 const struct kvm_irq_routing_entry
*entries
,
577 void kvm_free_irq_routing(struct kvm
*kvm
);
581 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
585 #ifdef CONFIG_HAVE_KVM_EVENTFD
587 void kvm_eventfd_init(struct kvm
*kvm
);
588 int kvm_irqfd(struct kvm
*kvm
, int fd
, int gsi
, int flags
);
589 void kvm_irqfd_release(struct kvm
*kvm
);
590 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
594 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
595 static inline int kvm_irqfd(struct kvm
*kvm
, int fd
, int gsi
, int flags
)
600 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
601 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
606 #endif /* CONFIG_HAVE_KVM_EVENTFD */
608 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
609 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu
*vcpu
)
611 return vcpu
->kvm
->bsp_vcpu_id
== vcpu
->vcpu_id
;
615 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
617 long kvm_vm_ioctl_assigned_device(struct kvm
*kvm
, unsigned ioctl
,
622 static inline long kvm_vm_ioctl_assigned_device(struct kvm
*kvm
, unsigned ioctl
,
630 static inline void kvm_make_request(int req
, struct kvm_vcpu
*vcpu
)
632 set_bit(req
, &vcpu
->requests
);
635 static inline bool kvm_make_check_request(int req
, struct kvm_vcpu
*vcpu
)
637 return test_and_set_bit(req
, &vcpu
->requests
);
640 static inline bool kvm_check_request(int req
, struct kvm_vcpu
*vcpu
)
642 if (test_bit(req
, &vcpu
->requests
)) {
643 clear_bit(req
, &vcpu
->requests
);