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
42 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
46 extern struct kmem_cache
*kvm_vcpu_cache
;
49 * It would be nice to use something smarter than a linear search, TBD...
50 * Thankfully we dont expect many devices to register (famous last words :),
51 * so until then it will suffice. At least its abstracted so we can change
56 #define NR_IOBUS_DEVS 200
57 struct kvm_io_device
*devs
[NR_IOBUS_DEVS
];
60 void kvm_io_bus_init(struct kvm_io_bus
*bus
);
61 void kvm_io_bus_destroy(struct kvm_io_bus
*bus
);
62 int kvm_io_bus_write(struct kvm_io_bus
*bus
, gpa_t addr
, int len
,
64 int kvm_io_bus_read(struct kvm_io_bus
*bus
, gpa_t addr
, int len
,
66 int __kvm_io_bus_register_dev(struct kvm_io_bus
*bus
,
67 struct kvm_io_device
*dev
);
68 int kvm_io_bus_register_dev(struct kvm
*kvm
, struct kvm_io_bus
*bus
,
69 struct kvm_io_device
*dev
);
70 void __kvm_io_bus_unregister_dev(struct kvm_io_bus
*bus
,
71 struct kvm_io_device
*dev
);
72 void kvm_io_bus_unregister_dev(struct kvm
*kvm
, struct kvm_io_bus
*bus
,
73 struct kvm_io_device
*dev
);
77 #ifdef CONFIG_PREEMPT_NOTIFIERS
78 struct preempt_notifier preempt_notifier
;
84 unsigned long requests
;
85 unsigned long guest_debug
;
91 struct kvm_vcpu_stat stat
;
93 #ifdef CONFIG_HAS_IOMEM
95 int mmio_read_completed
;
98 unsigned char mmio_data
[8];
102 struct kvm_vcpu_arch arch
;
105 struct kvm_memory_slot
{
107 unsigned long npages
;
110 unsigned long *dirty_bitmap
;
112 unsigned long rmap_pde
;
114 } *lpage_info
[KVM_NR_PAGE_SIZES
- 1];
115 unsigned long userspace_addr
;
119 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot
*memslot
)
121 return ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
124 struct kvm_kernel_irq_routing_entry
{
127 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
128 struct kvm
*kvm
, int irq_source_id
, int level
);
136 struct hlist_node link
;
139 #ifdef __KVM_HAVE_IOAPIC
141 struct kvm_irq_routing_table
{
142 int chip
[KVM_NR_IRQCHIPS
][KVM_IOAPIC_NUM_PINS
];
143 struct kvm_kernel_irq_routing_entry
*rt_entries
;
146 * Array indexed by gsi. Each entry contains list of irq chips
147 * the gsi is connected to.
149 struct hlist_head map
[0];
154 struct kvm_irq_routing_table
{};
160 spinlock_t requests_lock
;
161 struct rw_semaphore slots_lock
;
162 struct mm_struct
*mm
; /* userspace tied to this vm */
164 struct kvm_memory_slot memslots
[KVM_MEMORY_SLOTS
+
165 KVM_PRIVATE_MEM_SLOTS
];
166 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
168 struct kvm_vcpu
*bsp_vcpu
;
170 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
171 atomic_t online_vcpus
;
172 struct list_head vm_list
;
174 struct kvm_io_bus mmio_bus
;
175 struct kvm_io_bus pio_bus
;
176 #ifdef CONFIG_HAVE_KVM_EVENTFD
179 struct list_head items
;
181 struct list_head ioeventfds
;
183 struct kvm_vm_stat stat
;
184 struct kvm_arch arch
;
185 atomic_t users_count
;
186 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
187 struct kvm_coalesced_mmio_dev
*coalesced_mmio_dev
;
188 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
191 struct mutex irq_lock
;
192 #ifdef CONFIG_HAVE_KVM_IRQCHIP
193 struct kvm_irq_routing_table
*irq_routing
;
194 struct hlist_head mask_notifier_list
;
195 struct hlist_head irq_ack_notifier_list
;
198 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
199 struct mmu_notifier mmu_notifier
;
200 unsigned long mmu_notifier_seq
;
201 long mmu_notifier_count
;
205 /* The guest did something we don't support. */
206 #define pr_unimpl(vcpu, fmt, ...) \
208 if (printk_ratelimit()) \
209 printk(KERN_ERR "kvm: %i: cpu%i " fmt, \
210 current->tgid, (vcpu)->vcpu_id , ## __VA_ARGS__); \
213 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
214 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
216 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
219 return kvm
->vcpus
[i
];
222 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
223 for (idx = 0, vcpup = kvm_get_vcpu(kvm, idx); \
224 idx < atomic_read(&kvm->online_vcpus) && vcpup; \
225 vcpup = kvm_get_vcpu(kvm, ++idx))
227 int kvm_vcpu_init(struct kvm_vcpu
*vcpu
, struct kvm
*kvm
, unsigned id
);
228 void kvm_vcpu_uninit(struct kvm_vcpu
*vcpu
);
230 void vcpu_load(struct kvm_vcpu
*vcpu
);
231 void vcpu_put(struct kvm_vcpu
*vcpu
);
233 int kvm_init(void *opaque
, unsigned int vcpu_size
,
234 struct module
*module
);
237 void kvm_get_kvm(struct kvm
*kvm
);
238 void kvm_put_kvm(struct kvm
*kvm
);
240 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
241 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
242 static inline int is_error_hpa(hpa_t hpa
) { return hpa
>> HPA_MSB
; }
243 struct page
*gva_to_page(struct kvm_vcpu
*vcpu
, gva_t gva
);
245 extern struct page
*bad_page
;
246 extern pfn_t bad_pfn
;
248 int is_error_page(struct page
*page
);
249 int is_error_pfn(pfn_t pfn
);
250 int kvm_is_error_hva(unsigned long addr
);
251 int kvm_set_memory_region(struct kvm
*kvm
,
252 struct kvm_userspace_memory_region
*mem
,
254 int __kvm_set_memory_region(struct kvm
*kvm
,
255 struct kvm_userspace_memory_region
*mem
,
257 int kvm_arch_set_memory_region(struct kvm
*kvm
,
258 struct kvm_userspace_memory_region
*mem
,
259 struct kvm_memory_slot old
,
261 void kvm_disable_largepages(void);
262 void kvm_arch_flush_shadow(struct kvm
*kvm
);
263 gfn_t
unalias_gfn(struct kvm
*kvm
, gfn_t gfn
);
264 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
265 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
266 void kvm_release_page_clean(struct page
*page
);
267 void kvm_release_page_dirty(struct page
*page
);
268 void kvm_set_page_dirty(struct page
*page
);
269 void kvm_set_page_accessed(struct page
*page
);
271 pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
272 void kvm_release_pfn_dirty(pfn_t
);
273 void kvm_release_pfn_clean(pfn_t pfn
);
274 void kvm_set_pfn_dirty(pfn_t pfn
);
275 void kvm_set_pfn_accessed(pfn_t pfn
);
276 void kvm_get_pfn(pfn_t pfn
);
278 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
280 int kvm_read_guest_atomic(struct kvm
*kvm
, gpa_t gpa
, void *data
,
282 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
283 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
284 int offset
, int len
);
285 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
287 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
288 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
289 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
290 int kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
291 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
293 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
294 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
);
295 void kvm_resched(struct kvm_vcpu
*vcpu
);
296 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
297 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
298 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
299 void kvm_reload_remote_mmus(struct kvm
*kvm
);
301 long kvm_arch_dev_ioctl(struct file
*filp
,
302 unsigned int ioctl
, unsigned long arg
);
303 long kvm_arch_vcpu_ioctl(struct file
*filp
,
304 unsigned int ioctl
, unsigned long arg
);
306 int kvm_dev_ioctl_check_extension(long ext
);
308 int kvm_get_dirty_log(struct kvm
*kvm
,
309 struct kvm_dirty_log
*log
, int *is_dirty
);
310 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
311 struct kvm_dirty_log
*log
);
313 int kvm_vm_ioctl_set_memory_region(struct kvm
*kvm
,
315 kvm_userspace_memory_region
*mem
,
317 long kvm_arch_vm_ioctl(struct file
*filp
,
318 unsigned int ioctl
, unsigned long arg
);
320 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
321 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
323 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
324 struct kvm_translation
*tr
);
326 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
327 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
328 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
329 struct kvm_sregs
*sregs
);
330 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
331 struct kvm_sregs
*sregs
);
332 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
333 struct kvm_mp_state
*mp_state
);
334 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
335 struct kvm_mp_state
*mp_state
);
336 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
337 struct kvm_guest_debug
*dbg
);
338 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
340 int kvm_arch_init(void *opaque
);
341 void kvm_arch_exit(void);
343 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
);
344 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
);
346 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
);
347 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
348 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
349 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
);
350 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
);
351 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
353 int kvm_arch_vcpu_reset(struct kvm_vcpu
*vcpu
);
354 int kvm_arch_hardware_enable(void *garbage
);
355 void kvm_arch_hardware_disable(void *garbage
);
356 int kvm_arch_hardware_setup(void);
357 void kvm_arch_hardware_unsetup(void);
358 void kvm_arch_check_processor_compat(void *rtn
);
359 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
361 void kvm_free_physmem(struct kvm
*kvm
);
363 struct kvm
*kvm_arch_create_vm(void);
364 void kvm_arch_destroy_vm(struct kvm
*kvm
);
365 void kvm_free_all_assigned_devices(struct kvm
*kvm
);
366 void kvm_arch_sync_events(struct kvm
*kvm
);
368 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
369 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
371 int kvm_is_mmio_pfn(pfn_t pfn
);
373 struct kvm_irq_ack_notifier
{
374 struct hlist_node link
;
376 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
379 #define KVM_ASSIGNED_MSIX_PENDING 0x1
380 struct kvm_guest_msix_entry
{
386 struct kvm_assigned_dev_kernel
{
387 struct kvm_irq_ack_notifier ack_notifier
;
388 struct work_struct interrupt_work
;
389 struct list_head list
;
393 unsigned int entries_nr
;
395 bool host_irq_disabled
;
396 struct msix_entry
*host_msix_entries
;
398 struct kvm_guest_msix_entry
*guest_msix_entries
;
399 unsigned long irq_requested_type
;
404 spinlock_t assigned_dev_lock
;
407 struct kvm_irq_mask_notifier
{
408 void (*func
)(struct kvm_irq_mask_notifier
*kimn
, bool masked
);
410 struct hlist_node link
;
413 void kvm_register_irq_mask_notifier(struct kvm
*kvm
, int irq
,
414 struct kvm_irq_mask_notifier
*kimn
);
415 void kvm_unregister_irq_mask_notifier(struct kvm
*kvm
, int irq
,
416 struct kvm_irq_mask_notifier
*kimn
);
417 void kvm_fire_mask_notifiers(struct kvm
*kvm
, int irq
, bool mask
);
419 #ifdef __KVM_HAVE_IOAPIC
420 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic
*ioapic
,
421 union kvm_ioapic_redirect_entry
*entry
,
422 unsigned long *deliver_bitmask
);
424 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
);
425 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
426 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
427 struct kvm_irq_ack_notifier
*kian
);
428 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
429 struct kvm_irq_ack_notifier
*kian
);
430 int kvm_request_irq_source_id(struct kvm
*kvm
);
431 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
433 /* For vcpu->arch.iommu_flags */
434 #define KVM_IOMMU_CACHE_COHERENCY 0x1
436 #ifdef CONFIG_IOMMU_API
437 int kvm_iommu_map_pages(struct kvm
*kvm
, gfn_t base_gfn
,
438 unsigned long npages
);
439 int kvm_iommu_map_guest(struct kvm
*kvm
);
440 int kvm_iommu_unmap_guest(struct kvm
*kvm
);
441 int kvm_assign_device(struct kvm
*kvm
,
442 struct kvm_assigned_dev_kernel
*assigned_dev
);
443 int kvm_deassign_device(struct kvm
*kvm
,
444 struct kvm_assigned_dev_kernel
*assigned_dev
);
445 #else /* CONFIG_IOMMU_API */
446 static inline int kvm_iommu_map_pages(struct kvm
*kvm
,
448 unsigned long npages
)
453 static inline int kvm_iommu_map_guest(struct kvm
*kvm
)
458 static inline int kvm_iommu_unmap_guest(struct kvm
*kvm
)
463 static inline int kvm_assign_device(struct kvm
*kvm
,
464 struct kvm_assigned_dev_kernel
*assigned_dev
)
469 static inline int kvm_deassign_device(struct kvm
*kvm
,
470 struct kvm_assigned_dev_kernel
*assigned_dev
)
474 #endif /* CONFIG_IOMMU_API */
476 static inline void kvm_guest_enter(void)
478 account_system_vtime(current
);
479 current
->flags
|= PF_VCPU
;
482 static inline void kvm_guest_exit(void)
484 account_system_vtime(current
);
485 current
->flags
&= ~PF_VCPU
;
488 static inline int memslot_id(struct kvm
*kvm
, struct kvm_memory_slot
*slot
)
490 return slot
- kvm
->memslots
;
493 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
495 return (gpa_t
)gfn
<< PAGE_SHIFT
;
498 static inline hpa_t
pfn_to_hpa(pfn_t pfn
)
500 return (hpa_t
)pfn
<< PAGE_SHIFT
;
503 static inline void kvm_migrate_timers(struct kvm_vcpu
*vcpu
)
505 set_bit(KVM_REQ_MIGRATE_TIMER
, &vcpu
->requests
);
513 struct kvm_stats_debugfs_item
{
516 enum kvm_stat_kind kind
;
517 struct dentry
*dentry
;
519 extern struct kvm_stats_debugfs_item debugfs_entries
[];
520 extern struct dentry
*kvm_debugfs_dir
;
522 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
523 static inline int mmu_notifier_retry(struct kvm_vcpu
*vcpu
, unsigned long mmu_seq
)
525 if (unlikely(vcpu
->kvm
->mmu_notifier_count
))
528 * Both reads happen under the mmu_lock and both values are
529 * modified under mmu_lock, so there's no need of smb_rmb()
530 * here in between, otherwise mmu_notifier_count should be
531 * read before mmu_notifier_seq, see
532 * mmu_notifier_invalidate_range_end write side.
534 if (vcpu
->kvm
->mmu_notifier_seq
!= mmu_seq
)
540 #ifdef CONFIG_HAVE_KVM_IRQCHIP
542 #define KVM_MAX_IRQ_ROUTES 1024
544 int kvm_setup_default_irq_routing(struct kvm
*kvm
);
545 int kvm_set_irq_routing(struct kvm
*kvm
,
546 const struct kvm_irq_routing_entry
*entries
,
549 void kvm_free_irq_routing(struct kvm
*kvm
);
553 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
557 #ifdef CONFIG_HAVE_KVM_EVENTFD
559 void kvm_eventfd_init(struct kvm
*kvm
);
560 int kvm_irqfd(struct kvm
*kvm
, int fd
, int gsi
, int flags
);
561 void kvm_irqfd_release(struct kvm
*kvm
);
562 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
566 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
567 static inline int kvm_irqfd(struct kvm
*kvm
, int fd
, int gsi
, int flags
)
572 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
573 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
578 #endif /* CONFIG_HAVE_KVM_EVENTFD */
580 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
581 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu
*vcpu
)
583 return vcpu
->kvm
->bsp_vcpu_id
== vcpu
->vcpu_id
;
587 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
589 long kvm_vm_ioctl_assigned_device(struct kvm
*kvm
, unsigned ioctl
,
594 static inline long kvm_vm_ioctl_assigned_device(struct kvm
*kvm
, unsigned ioctl
,