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
;
85 unsigned long requests
;
86 unsigned long guest_debug
;
94 struct kvm_vcpu_stat stat
;
96 #ifdef CONFIG_HAS_IOMEM
98 int mmio_read_completed
;
101 unsigned char mmio_data
[8];
102 gpa_t mmio_phys_addr
;
105 struct kvm_vcpu_arch arch
;
108 struct kvm_memory_slot
{
110 unsigned long npages
;
113 unsigned long *dirty_bitmap
;
115 unsigned long rmap_pde
;
117 } *lpage_info
[KVM_NR_PAGE_SIZES
- 1];
118 unsigned long userspace_addr
;
122 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot
*memslot
)
124 return ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
127 struct kvm_kernel_irq_routing_entry
{
130 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
131 struct kvm
*kvm
, int irq_source_id
, int level
);
139 struct hlist_node link
;
142 #ifdef __KVM_HAVE_IOAPIC
144 struct kvm_irq_routing_table
{
145 int chip
[KVM_NR_IRQCHIPS
][KVM_IOAPIC_NUM_PINS
];
146 struct kvm_kernel_irq_routing_entry
*rt_entries
;
149 * Array indexed by gsi. Each entry contains list of irq chips
150 * the gsi is connected to.
152 struct hlist_head map
[0];
157 struct kvm_irq_routing_table
{};
161 struct kvm_memslots
{
163 struct kvm_memory_slot memslots
[KVM_MEMORY_SLOTS
+
164 KVM_PRIVATE_MEM_SLOTS
];
169 raw_spinlock_t requests_lock
;
170 struct mutex slots_lock
;
171 struct mm_struct
*mm
; /* userspace tied to this vm */
172 struct kvm_memslots
*memslots
;
173 struct srcu_struct srcu
;
174 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
176 struct kvm_vcpu
*bsp_vcpu
;
178 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
179 atomic_t online_vcpus
;
180 struct list_head vm_list
;
182 struct kvm_io_bus
*buses
[KVM_NR_BUSES
];
183 #ifdef CONFIG_HAVE_KVM_EVENTFD
186 struct list_head items
;
188 struct list_head ioeventfds
;
190 struct kvm_vm_stat stat
;
191 struct kvm_arch arch
;
192 atomic_t users_count
;
193 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
194 struct kvm_coalesced_mmio_dev
*coalesced_mmio_dev
;
195 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
198 struct mutex irq_lock
;
199 #ifdef CONFIG_HAVE_KVM_IRQCHIP
200 struct kvm_irq_routing_table
*irq_routing
;
201 struct hlist_head mask_notifier_list
;
202 struct hlist_head irq_ack_notifier_list
;
205 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
206 struct mmu_notifier mmu_notifier
;
207 unsigned long mmu_notifier_seq
;
208 long mmu_notifier_count
;
212 /* The guest did something we don't support. */
213 #define pr_unimpl(vcpu, fmt, ...) \
215 if (printk_ratelimit()) \
216 printk(KERN_ERR "kvm: %i: cpu%i " fmt, \
217 current->tgid, (vcpu)->vcpu_id , ## __VA_ARGS__); \
220 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
221 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
223 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
226 return kvm
->vcpus
[i
];
229 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
230 for (idx = 0, vcpup = kvm_get_vcpu(kvm, idx); \
231 idx < atomic_read(&kvm->online_vcpus) && vcpup; \
232 vcpup = kvm_get_vcpu(kvm, ++idx))
234 int kvm_vcpu_init(struct kvm_vcpu
*vcpu
, struct kvm
*kvm
, unsigned id
);
235 void kvm_vcpu_uninit(struct kvm_vcpu
*vcpu
);
237 void vcpu_load(struct kvm_vcpu
*vcpu
);
238 void vcpu_put(struct kvm_vcpu
*vcpu
);
240 int kvm_init(void *opaque
, unsigned int vcpu_size
,
241 struct module
*module
);
244 void kvm_get_kvm(struct kvm
*kvm
);
245 void kvm_put_kvm(struct kvm
*kvm
);
247 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
248 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
249 static inline int is_error_hpa(hpa_t hpa
) { return hpa
>> HPA_MSB
; }
250 struct page
*gva_to_page(struct kvm_vcpu
*vcpu
, gva_t gva
);
252 extern struct page
*bad_page
;
253 extern pfn_t bad_pfn
;
255 int is_error_page(struct page
*page
);
256 int is_error_pfn(pfn_t pfn
);
257 int kvm_is_error_hva(unsigned long addr
);
258 int kvm_set_memory_region(struct kvm
*kvm
,
259 struct kvm_userspace_memory_region
*mem
,
261 int __kvm_set_memory_region(struct kvm
*kvm
,
262 struct kvm_userspace_memory_region
*mem
,
264 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
265 struct kvm_memory_slot
*memslot
,
266 struct kvm_memory_slot old
,
267 struct kvm_userspace_memory_region
*mem
,
269 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
270 struct kvm_userspace_memory_region
*mem
,
271 struct kvm_memory_slot old
,
273 void kvm_disable_largepages(void);
274 void kvm_arch_flush_shadow(struct kvm
*kvm
);
275 gfn_t
unalias_gfn(struct kvm
*kvm
, gfn_t gfn
);
276 gfn_t
unalias_gfn_instantiation(struct kvm
*kvm
, gfn_t gfn
);
278 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
279 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
280 void kvm_release_page_clean(struct page
*page
);
281 void kvm_release_page_dirty(struct page
*page
);
282 void kvm_set_page_dirty(struct page
*page
);
283 void kvm_set_page_accessed(struct page
*page
);
285 pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
286 pfn_t
gfn_to_pfn_memslot(struct kvm
*kvm
,
287 struct kvm_memory_slot
*slot
, gfn_t gfn
);
288 int memslot_id(struct kvm
*kvm
, gfn_t gfn
);
289 void kvm_release_pfn_dirty(pfn_t
);
290 void kvm_release_pfn_clean(pfn_t pfn
);
291 void kvm_set_pfn_dirty(pfn_t pfn
);
292 void kvm_set_pfn_accessed(pfn_t pfn
);
293 void kvm_get_pfn(pfn_t pfn
);
295 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
297 int kvm_read_guest_atomic(struct kvm
*kvm
, gpa_t gpa
, void *data
,
299 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
300 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
301 int offset
, int len
);
302 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
304 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
305 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
306 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
307 int kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
308 unsigned long kvm_host_page_size(struct kvm
*kvm
, gfn_t gfn
);
309 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
311 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
312 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
);
313 void kvm_resched(struct kvm_vcpu
*vcpu
);
314 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
315 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
316 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
317 void kvm_reload_remote_mmus(struct kvm
*kvm
);
319 long kvm_arch_dev_ioctl(struct file
*filp
,
320 unsigned int ioctl
, unsigned long arg
);
321 long kvm_arch_vcpu_ioctl(struct file
*filp
,
322 unsigned int ioctl
, unsigned long arg
);
324 int kvm_dev_ioctl_check_extension(long ext
);
326 int kvm_get_dirty_log(struct kvm
*kvm
,
327 struct kvm_dirty_log
*log
, int *is_dirty
);
328 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
329 struct kvm_dirty_log
*log
);
331 int kvm_vm_ioctl_set_memory_region(struct kvm
*kvm
,
333 kvm_userspace_memory_region
*mem
,
335 long kvm_arch_vm_ioctl(struct file
*filp
,
336 unsigned int ioctl
, unsigned long arg
);
338 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
339 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
341 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
342 struct kvm_translation
*tr
);
344 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
345 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
346 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
347 struct kvm_sregs
*sregs
);
348 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
349 struct kvm_sregs
*sregs
);
350 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
351 struct kvm_mp_state
*mp_state
);
352 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
353 struct kvm_mp_state
*mp_state
);
354 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
355 struct kvm_guest_debug
*dbg
);
356 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
358 int kvm_arch_init(void *opaque
);
359 void kvm_arch_exit(void);
361 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
);
362 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
);
364 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
);
365 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
366 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
367 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
);
368 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
);
369 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
371 int kvm_arch_vcpu_reset(struct kvm_vcpu
*vcpu
);
372 int kvm_arch_hardware_enable(void *garbage
);
373 void kvm_arch_hardware_disable(void *garbage
);
374 int kvm_arch_hardware_setup(void);
375 void kvm_arch_hardware_unsetup(void);
376 void kvm_arch_check_processor_compat(void *rtn
);
377 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
379 void kvm_free_physmem(struct kvm
*kvm
);
381 struct kvm
*kvm_arch_create_vm(void);
382 void kvm_arch_destroy_vm(struct kvm
*kvm
);
383 void kvm_free_all_assigned_devices(struct kvm
*kvm
);
384 void kvm_arch_sync_events(struct kvm
*kvm
);
386 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
387 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
389 int kvm_is_mmio_pfn(pfn_t pfn
);
391 struct kvm_irq_ack_notifier
{
392 struct hlist_node link
;
394 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
397 #define KVM_ASSIGNED_MSIX_PENDING 0x1
398 struct kvm_guest_msix_entry
{
404 struct kvm_assigned_dev_kernel
{
405 struct kvm_irq_ack_notifier ack_notifier
;
406 struct work_struct interrupt_work
;
407 struct list_head list
;
412 unsigned int entries_nr
;
414 bool host_irq_disabled
;
415 struct msix_entry
*host_msix_entries
;
417 struct kvm_guest_msix_entry
*guest_msix_entries
;
418 unsigned long irq_requested_type
;
423 spinlock_t assigned_dev_lock
;
426 struct kvm_irq_mask_notifier
{
427 void (*func
)(struct kvm_irq_mask_notifier
*kimn
, bool masked
);
429 struct hlist_node link
;
432 void kvm_register_irq_mask_notifier(struct kvm
*kvm
, int irq
,
433 struct kvm_irq_mask_notifier
*kimn
);
434 void kvm_unregister_irq_mask_notifier(struct kvm
*kvm
, int irq
,
435 struct kvm_irq_mask_notifier
*kimn
);
436 void kvm_fire_mask_notifiers(struct kvm
*kvm
, int irq
, bool mask
);
438 #ifdef __KVM_HAVE_IOAPIC
439 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic
*ioapic
,
440 union kvm_ioapic_redirect_entry
*entry
,
441 unsigned long *deliver_bitmask
);
443 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
);
444 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
445 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
446 struct kvm_irq_ack_notifier
*kian
);
447 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
448 struct kvm_irq_ack_notifier
*kian
);
449 int kvm_request_irq_source_id(struct kvm
*kvm
);
450 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
452 /* For vcpu->arch.iommu_flags */
453 #define KVM_IOMMU_CACHE_COHERENCY 0x1
455 #ifdef CONFIG_IOMMU_API
456 int kvm_iommu_map_pages(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
457 int kvm_iommu_map_guest(struct kvm
*kvm
);
458 int kvm_iommu_unmap_guest(struct kvm
*kvm
);
459 int kvm_assign_device(struct kvm
*kvm
,
460 struct kvm_assigned_dev_kernel
*assigned_dev
);
461 int kvm_deassign_device(struct kvm
*kvm
,
462 struct kvm_assigned_dev_kernel
*assigned_dev
);
463 #else /* CONFIG_IOMMU_API */
464 static inline int kvm_iommu_map_pages(struct kvm
*kvm
,
466 unsigned long npages
)
471 static inline int kvm_iommu_map_guest(struct kvm
*kvm
)
476 static inline int kvm_iommu_unmap_guest(struct kvm
*kvm
)
481 static inline int kvm_assign_device(struct kvm
*kvm
,
482 struct kvm_assigned_dev_kernel
*assigned_dev
)
487 static inline int kvm_deassign_device(struct kvm
*kvm
,
488 struct kvm_assigned_dev_kernel
*assigned_dev
)
492 #endif /* CONFIG_IOMMU_API */
494 static inline void kvm_guest_enter(void)
496 account_system_vtime(current
);
497 current
->flags
|= PF_VCPU
;
500 static inline void kvm_guest_exit(void)
502 account_system_vtime(current
);
503 current
->flags
&= ~PF_VCPU
;
506 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
508 return (gpa_t
)gfn
<< PAGE_SHIFT
;
511 static inline hpa_t
pfn_to_hpa(pfn_t pfn
)
513 return (hpa_t
)pfn
<< PAGE_SHIFT
;
516 static inline void kvm_migrate_timers(struct kvm_vcpu
*vcpu
)
518 set_bit(KVM_REQ_MIGRATE_TIMER
, &vcpu
->requests
);
526 struct kvm_stats_debugfs_item
{
529 enum kvm_stat_kind kind
;
530 struct dentry
*dentry
;
532 extern struct kvm_stats_debugfs_item debugfs_entries
[];
533 extern struct dentry
*kvm_debugfs_dir
;
535 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
536 static inline int mmu_notifier_retry(struct kvm_vcpu
*vcpu
, unsigned long mmu_seq
)
538 if (unlikely(vcpu
->kvm
->mmu_notifier_count
))
541 * Both reads happen under the mmu_lock and both values are
542 * modified under mmu_lock, so there's no need of smb_rmb()
543 * here in between, otherwise mmu_notifier_count should be
544 * read before mmu_notifier_seq, see
545 * mmu_notifier_invalidate_range_end write side.
547 if (vcpu
->kvm
->mmu_notifier_seq
!= mmu_seq
)
553 #ifndef KVM_ARCH_HAS_UNALIAS_INSTANTIATION
554 #define unalias_gfn_instantiation unalias_gfn
557 #ifdef CONFIG_HAVE_KVM_IRQCHIP
559 #define KVM_MAX_IRQ_ROUTES 1024
561 int kvm_setup_default_irq_routing(struct kvm
*kvm
);
562 int kvm_set_irq_routing(struct kvm
*kvm
,
563 const struct kvm_irq_routing_entry
*entries
,
566 void kvm_free_irq_routing(struct kvm
*kvm
);
570 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
574 #ifdef CONFIG_HAVE_KVM_EVENTFD
576 void kvm_eventfd_init(struct kvm
*kvm
);
577 int kvm_irqfd(struct kvm
*kvm
, int fd
, int gsi
, int flags
);
578 void kvm_irqfd_release(struct kvm
*kvm
);
579 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
583 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
584 static inline int kvm_irqfd(struct kvm
*kvm
, int fd
, int gsi
, int flags
)
589 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
590 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
595 #endif /* CONFIG_HAVE_KVM_EVENTFD */
597 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
598 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu
*vcpu
)
600 return vcpu
->kvm
->bsp_vcpu_id
== vcpu
->vcpu_id
;
604 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
606 long kvm_vm_ioctl_assigned_device(struct kvm
*kvm
, unsigned ioctl
,
611 static inline long kvm_vm_ioctl_assigned_device(struct kvm
*kvm
, unsigned ioctl
,