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 6
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 struct kvm_kernel_irq_routing_entry
{
122 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
123 struct kvm
*kvm
, int irq_source_id
, int level
);
131 struct hlist_node link
;
134 #ifdef __KVM_HAVE_IOAPIC
136 struct kvm_irq_routing_table
{
137 int chip
[KVM_NR_IRQCHIPS
][KVM_IOAPIC_NUM_PINS
];
138 struct kvm_kernel_irq_routing_entry
*rt_entries
;
141 * Array indexed by gsi. Each entry contains list of irq chips
142 * the gsi is connected to.
144 struct hlist_head map
[0];
149 struct kvm_irq_routing_table
{};
155 spinlock_t requests_lock
;
156 struct rw_semaphore slots_lock
;
157 struct mm_struct
*mm
; /* userspace tied to this vm */
159 struct kvm_memory_slot memslots
[KVM_MEMORY_SLOTS
+
160 KVM_PRIVATE_MEM_SLOTS
];
161 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
163 struct kvm_vcpu
*bsp_vcpu
;
165 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
166 atomic_t online_vcpus
;
167 struct list_head vm_list
;
169 struct kvm_io_bus mmio_bus
;
170 struct kvm_io_bus pio_bus
;
171 #ifdef CONFIG_HAVE_KVM_EVENTFD
174 struct list_head items
;
176 struct list_head ioeventfds
;
178 struct kvm_vm_stat stat
;
179 struct kvm_arch arch
;
180 atomic_t users_count
;
181 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
182 struct kvm_coalesced_mmio_dev
*coalesced_mmio_dev
;
183 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
186 struct mutex irq_lock
;
187 #ifdef CONFIG_HAVE_KVM_IRQCHIP
188 struct kvm_irq_routing_table
*irq_routing
;
189 struct hlist_head mask_notifier_list
;
190 struct hlist_head irq_ack_notifier_list
;
193 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
194 struct mmu_notifier mmu_notifier
;
195 unsigned long mmu_notifier_seq
;
196 long mmu_notifier_count
;
200 /* The guest did something we don't support. */
201 #define pr_unimpl(vcpu, fmt, ...) \
203 if (printk_ratelimit()) \
204 printk(KERN_ERR "kvm: %i: cpu%i " fmt, \
205 current->tgid, (vcpu)->vcpu_id , ## __VA_ARGS__); \
208 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
209 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
211 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
214 return kvm
->vcpus
[i
];
217 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
218 for (idx = 0, vcpup = kvm_get_vcpu(kvm, idx); \
219 idx < atomic_read(&kvm->online_vcpus) && vcpup; \
220 vcpup = kvm_get_vcpu(kvm, ++idx))
222 int kvm_vcpu_init(struct kvm_vcpu
*vcpu
, struct kvm
*kvm
, unsigned id
);
223 void kvm_vcpu_uninit(struct kvm_vcpu
*vcpu
);
225 void vcpu_load(struct kvm_vcpu
*vcpu
);
226 void vcpu_put(struct kvm_vcpu
*vcpu
);
228 int kvm_init(void *opaque
, unsigned int vcpu_size
,
229 struct module
*module
);
232 void kvm_get_kvm(struct kvm
*kvm
);
233 void kvm_put_kvm(struct kvm
*kvm
);
235 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
236 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
237 static inline int is_error_hpa(hpa_t hpa
) { return hpa
>> HPA_MSB
; }
238 struct page
*gva_to_page(struct kvm_vcpu
*vcpu
, gva_t gva
);
240 extern struct page
*bad_page
;
241 extern pfn_t bad_pfn
;
243 int is_error_page(struct page
*page
);
244 int is_error_pfn(pfn_t pfn
);
245 int kvm_is_error_hva(unsigned long addr
);
246 int kvm_set_memory_region(struct kvm
*kvm
,
247 struct kvm_userspace_memory_region
*mem
,
249 int __kvm_set_memory_region(struct kvm
*kvm
,
250 struct kvm_userspace_memory_region
*mem
,
252 int kvm_arch_set_memory_region(struct kvm
*kvm
,
253 struct kvm_userspace_memory_region
*mem
,
254 struct kvm_memory_slot old
,
256 void kvm_disable_largepages(void);
257 void kvm_arch_flush_shadow(struct kvm
*kvm
);
258 gfn_t
unalias_gfn(struct kvm
*kvm
, gfn_t gfn
);
259 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
260 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
261 void kvm_release_page_clean(struct page
*page
);
262 void kvm_release_page_dirty(struct page
*page
);
263 void kvm_set_page_dirty(struct page
*page
);
264 void kvm_set_page_accessed(struct page
*page
);
266 pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
267 void kvm_release_pfn_dirty(pfn_t
);
268 void kvm_release_pfn_clean(pfn_t pfn
);
269 void kvm_set_pfn_dirty(pfn_t pfn
);
270 void kvm_set_pfn_accessed(pfn_t pfn
);
271 void kvm_get_pfn(pfn_t pfn
);
273 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
275 int kvm_read_guest_atomic(struct kvm
*kvm
, gpa_t gpa
, void *data
,
277 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
278 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
279 int offset
, int len
);
280 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
282 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
283 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
284 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
285 int kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
286 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
288 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
289 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
);
290 void kvm_resched(struct kvm_vcpu
*vcpu
);
291 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
292 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
293 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
294 void kvm_reload_remote_mmus(struct kvm
*kvm
);
296 long kvm_arch_dev_ioctl(struct file
*filp
,
297 unsigned int ioctl
, unsigned long arg
);
298 long kvm_arch_vcpu_ioctl(struct file
*filp
,
299 unsigned int ioctl
, unsigned long arg
);
301 int kvm_dev_ioctl_check_extension(long ext
);
303 int kvm_get_dirty_log(struct kvm
*kvm
,
304 struct kvm_dirty_log
*log
, int *is_dirty
);
305 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
306 struct kvm_dirty_log
*log
);
308 int kvm_vm_ioctl_set_memory_region(struct kvm
*kvm
,
310 kvm_userspace_memory_region
*mem
,
312 long kvm_arch_vm_ioctl(struct file
*filp
,
313 unsigned int ioctl
, unsigned long arg
);
315 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
316 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
318 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
319 struct kvm_translation
*tr
);
321 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
322 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
323 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
324 struct kvm_sregs
*sregs
);
325 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
326 struct kvm_sregs
*sregs
);
327 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
328 struct kvm_mp_state
*mp_state
);
329 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
330 struct kvm_mp_state
*mp_state
);
331 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
332 struct kvm_guest_debug
*dbg
);
333 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
335 int kvm_arch_init(void *opaque
);
336 void kvm_arch_exit(void);
338 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
);
339 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
);
341 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
);
342 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
343 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
344 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
);
345 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
);
346 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
348 int kvm_arch_vcpu_reset(struct kvm_vcpu
*vcpu
);
349 int kvm_arch_hardware_enable(void *garbage
);
350 void kvm_arch_hardware_disable(void *garbage
);
351 int kvm_arch_hardware_setup(void);
352 void kvm_arch_hardware_unsetup(void);
353 void kvm_arch_check_processor_compat(void *rtn
);
354 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
356 void kvm_free_physmem(struct kvm
*kvm
);
358 struct kvm
*kvm_arch_create_vm(void);
359 void kvm_arch_destroy_vm(struct kvm
*kvm
);
360 void kvm_free_all_assigned_devices(struct kvm
*kvm
);
361 void kvm_arch_sync_events(struct kvm
*kvm
);
363 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
364 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
366 int kvm_is_mmio_pfn(pfn_t pfn
);
368 struct kvm_irq_ack_notifier
{
369 struct hlist_node link
;
371 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
374 #define KVM_ASSIGNED_MSIX_PENDING 0x1
375 struct kvm_guest_msix_entry
{
381 struct kvm_assigned_dev_kernel
{
382 struct kvm_irq_ack_notifier ack_notifier
;
383 struct work_struct interrupt_work
;
384 struct list_head list
;
388 unsigned int entries_nr
;
390 bool host_irq_disabled
;
391 struct msix_entry
*host_msix_entries
;
393 struct kvm_guest_msix_entry
*guest_msix_entries
;
394 unsigned long irq_requested_type
;
399 spinlock_t assigned_dev_lock
;
402 struct kvm_irq_mask_notifier
{
403 void (*func
)(struct kvm_irq_mask_notifier
*kimn
, bool masked
);
405 struct hlist_node link
;
408 void kvm_register_irq_mask_notifier(struct kvm
*kvm
, int irq
,
409 struct kvm_irq_mask_notifier
*kimn
);
410 void kvm_unregister_irq_mask_notifier(struct kvm
*kvm
, int irq
,
411 struct kvm_irq_mask_notifier
*kimn
);
412 void kvm_fire_mask_notifiers(struct kvm
*kvm
, int irq
, bool mask
);
414 #ifdef __KVM_HAVE_IOAPIC
415 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic
*ioapic
,
416 union kvm_ioapic_redirect_entry
*entry
,
417 unsigned long *deliver_bitmask
);
419 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
);
420 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
421 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
422 struct kvm_irq_ack_notifier
*kian
);
423 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
424 struct kvm_irq_ack_notifier
*kian
);
425 int kvm_request_irq_source_id(struct kvm
*kvm
);
426 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
428 /* For vcpu->arch.iommu_flags */
429 #define KVM_IOMMU_CACHE_COHERENCY 0x1
431 #ifdef CONFIG_IOMMU_API
432 int kvm_iommu_map_pages(struct kvm
*kvm
, gfn_t base_gfn
,
433 unsigned long npages
);
434 int kvm_iommu_map_guest(struct kvm
*kvm
);
435 int kvm_iommu_unmap_guest(struct kvm
*kvm
);
436 int kvm_assign_device(struct kvm
*kvm
,
437 struct kvm_assigned_dev_kernel
*assigned_dev
);
438 int kvm_deassign_device(struct kvm
*kvm
,
439 struct kvm_assigned_dev_kernel
*assigned_dev
);
440 #else /* CONFIG_IOMMU_API */
441 static inline int kvm_iommu_map_pages(struct kvm
*kvm
,
443 unsigned long npages
)
448 static inline int kvm_iommu_map_guest(struct kvm
*kvm
)
453 static inline int kvm_iommu_unmap_guest(struct kvm
*kvm
)
458 static inline int kvm_assign_device(struct kvm
*kvm
,
459 struct kvm_assigned_dev_kernel
*assigned_dev
)
464 static inline int kvm_deassign_device(struct kvm
*kvm
,
465 struct kvm_assigned_dev_kernel
*assigned_dev
)
469 #endif /* CONFIG_IOMMU_API */
471 static inline void kvm_guest_enter(void)
473 account_system_vtime(current
);
474 current
->flags
|= PF_VCPU
;
477 static inline void kvm_guest_exit(void)
479 account_system_vtime(current
);
480 current
->flags
&= ~PF_VCPU
;
483 static inline int memslot_id(struct kvm
*kvm
, struct kvm_memory_slot
*slot
)
485 return slot
- kvm
->memslots
;
488 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
490 return (gpa_t
)gfn
<< PAGE_SHIFT
;
493 static inline hpa_t
pfn_to_hpa(pfn_t pfn
)
495 return (hpa_t
)pfn
<< PAGE_SHIFT
;
498 static inline void kvm_migrate_timers(struct kvm_vcpu
*vcpu
)
500 set_bit(KVM_REQ_MIGRATE_TIMER
, &vcpu
->requests
);
508 struct kvm_stats_debugfs_item
{
511 enum kvm_stat_kind kind
;
512 struct dentry
*dentry
;
514 extern struct kvm_stats_debugfs_item debugfs_entries
[];
515 extern struct dentry
*kvm_debugfs_dir
;
517 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
518 static inline int mmu_notifier_retry(struct kvm_vcpu
*vcpu
, unsigned long mmu_seq
)
520 if (unlikely(vcpu
->kvm
->mmu_notifier_count
))
523 * Both reads happen under the mmu_lock and both values are
524 * modified under mmu_lock, so there's no need of smb_rmb()
525 * here in between, otherwise mmu_notifier_count should be
526 * read before mmu_notifier_seq, see
527 * mmu_notifier_invalidate_range_end write side.
529 if (vcpu
->kvm
->mmu_notifier_seq
!= mmu_seq
)
535 #ifdef CONFIG_HAVE_KVM_IRQCHIP
537 #define KVM_MAX_IRQ_ROUTES 1024
539 int kvm_setup_default_irq_routing(struct kvm
*kvm
);
540 int kvm_set_irq_routing(struct kvm
*kvm
,
541 const struct kvm_irq_routing_entry
*entries
,
544 void kvm_free_irq_routing(struct kvm
*kvm
);
548 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
552 #ifdef CONFIG_HAVE_KVM_EVENTFD
554 void kvm_eventfd_init(struct kvm
*kvm
);
555 int kvm_irqfd(struct kvm
*kvm
, int fd
, int gsi
, int flags
);
556 void kvm_irqfd_release(struct kvm
*kvm
);
557 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
561 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
562 static inline int kvm_irqfd(struct kvm
*kvm
, int fd
, int gsi
, int flags
)
567 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
568 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
573 #endif /* CONFIG_HAVE_KVM_EVENTFD */
575 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
576 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu
*vcpu
)
578 return vcpu
->kvm
->bsp_vcpu_id
== vcpu
->vcpu_id
;
582 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
584 long kvm_vm_ioctl_assigned_device(struct kvm
*kvm
, unsigned ioctl
,
589 static inline long kvm_vm_ioctl_assigned_device(struct kvm
*kvm
, unsigned ioctl
,