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/marker.h>
19 #include <linux/msi.h>
20 #include <asm/signal.h>
22 #include <linux/kvm.h>
23 #include <linux/kvm_para.h>
25 #include <linux/kvm_types.h>
27 #include <asm/kvm_host.h>
30 * vcpu->requests bit members
32 #define KVM_REQ_TLB_FLUSH 0
33 #define KVM_REQ_MIGRATE_TIMER 1
34 #define KVM_REQ_REPORT_TPR_ACCESS 2
35 #define KVM_REQ_MMU_RELOAD 3
36 #define KVM_REQ_TRIPLE_FAULT 4
37 #define KVM_REQ_PENDING_TIMER 5
38 #define KVM_REQ_UNHALT 6
39 #define KVM_REQ_MMU_SYNC 7
40 #define KVM_REQ_KVMCLOCK_UPDATE 8
41 #define KVM_REQ_KICK 9
43 #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 struct kvm_io_device
*kvm_io_bus_find_dev(struct kvm_io_bus
*bus
,
63 gpa_t addr
, int len
, int is_write
);
64 void kvm_io_bus_register_dev(struct kvm_io_bus
*bus
,
65 struct kvm_io_device
*dev
);
69 #ifdef CONFIG_PREEMPT_NOTIFIERS
70 struct preempt_notifier preempt_notifier
;
76 unsigned long requests
;
77 unsigned long guest_debug
;
83 struct kvm_vcpu_stat stat
;
85 #ifdef CONFIG_HAS_IOMEM
87 int mmio_read_completed
;
90 unsigned char mmio_data
[8];
94 struct kvm_vcpu_arch arch
;
97 struct kvm_memory_slot
{
102 unsigned long *dirty_bitmap
;
104 unsigned long rmap_pde
;
107 unsigned long userspace_addr
;
111 struct kvm_kernel_irq_routing_entry
{
113 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
114 struct kvm
*kvm
, int level
);
122 struct list_head link
;
126 struct mutex lock
; /* protects the vcpus array and APIC accesses */
128 spinlock_t requests_lock
;
129 struct rw_semaphore slots_lock
;
130 struct mm_struct
*mm
; /* userspace tied to this vm */
132 struct kvm_memory_slot memslots
[KVM_MEMORY_SLOTS
+
133 KVM_PRIVATE_MEM_SLOTS
];
134 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
135 struct list_head vm_list
;
136 struct kvm_io_bus mmio_bus
;
137 struct kvm_io_bus pio_bus
;
138 struct kvm_vm_stat stat
;
139 struct kvm_arch arch
;
140 atomic_t users_count
;
141 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
142 struct kvm_coalesced_mmio_dev
*coalesced_mmio_dev
;
143 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
146 #ifdef CONFIG_HAVE_KVM_IRQCHIP
147 struct list_head irq_routing
; /* of kvm_kernel_irq_routing_entry */
148 struct hlist_head mask_notifier_list
;
151 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
152 struct mmu_notifier mmu_notifier
;
153 unsigned long mmu_notifier_seq
;
154 long mmu_notifier_count
;
158 /* The guest did something we don't support. */
159 #define pr_unimpl(vcpu, fmt, ...) \
161 if (printk_ratelimit()) \
162 printk(KERN_ERR "kvm: %i: cpu%i " fmt, \
163 current->tgid, (vcpu)->vcpu_id , ## __VA_ARGS__); \
166 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
167 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
169 int kvm_vcpu_init(struct kvm_vcpu
*vcpu
, struct kvm
*kvm
, unsigned id
);
170 void kvm_vcpu_uninit(struct kvm_vcpu
*vcpu
);
172 void vcpu_load(struct kvm_vcpu
*vcpu
);
173 void vcpu_put(struct kvm_vcpu
*vcpu
);
175 int kvm_init(void *opaque
, unsigned int vcpu_size
,
176 struct module
*module
);
179 void kvm_get_kvm(struct kvm
*kvm
);
180 void kvm_put_kvm(struct kvm
*kvm
);
182 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
183 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
184 static inline int is_error_hpa(hpa_t hpa
) { return hpa
>> HPA_MSB
; }
185 struct page
*gva_to_page(struct kvm_vcpu
*vcpu
, gva_t gva
);
187 extern struct page
*bad_page
;
188 extern pfn_t bad_pfn
;
190 int is_error_page(struct page
*page
);
191 int is_error_pfn(pfn_t pfn
);
192 int kvm_is_error_hva(unsigned long addr
);
193 int kvm_set_memory_region(struct kvm
*kvm
,
194 struct kvm_userspace_memory_region
*mem
,
196 int __kvm_set_memory_region(struct kvm
*kvm
,
197 struct kvm_userspace_memory_region
*mem
,
199 int kvm_arch_set_memory_region(struct kvm
*kvm
,
200 struct kvm_userspace_memory_region
*mem
,
201 struct kvm_memory_slot old
,
203 void kvm_arch_flush_shadow(struct kvm
*kvm
);
204 gfn_t
unalias_gfn(struct kvm
*kvm
, gfn_t gfn
);
205 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
206 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
207 void kvm_release_page_clean(struct page
*page
);
208 void kvm_release_page_dirty(struct page
*page
);
209 void kvm_set_page_dirty(struct page
*page
);
210 void kvm_set_page_accessed(struct page
*page
);
212 pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
213 void kvm_release_pfn_dirty(pfn_t
);
214 void kvm_release_pfn_clean(pfn_t pfn
);
215 void kvm_set_pfn_dirty(pfn_t pfn
);
216 void kvm_set_pfn_accessed(pfn_t pfn
);
217 void kvm_get_pfn(pfn_t pfn
);
219 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
221 int kvm_read_guest_atomic(struct kvm
*kvm
, gpa_t gpa
, void *data
,
223 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
224 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
225 int offset
, int len
);
226 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
228 int kvm_clear_guest_page(struct kvm
*kvm
, gfn_t gfn
, int offset
, int len
);
229 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
230 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
231 int kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
232 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
234 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
235 void kvm_resched(struct kvm_vcpu
*vcpu
);
236 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
237 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
238 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
239 void kvm_reload_remote_mmus(struct kvm
*kvm
);
241 long kvm_arch_dev_ioctl(struct file
*filp
,
242 unsigned int ioctl
, unsigned long arg
);
243 long kvm_arch_vcpu_ioctl(struct file
*filp
,
244 unsigned int ioctl
, unsigned long arg
);
245 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
246 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
248 int kvm_dev_ioctl_check_extension(long ext
);
250 int kvm_get_dirty_log(struct kvm
*kvm
,
251 struct kvm_dirty_log
*log
, int *is_dirty
);
252 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
253 struct kvm_dirty_log
*log
);
255 int kvm_vm_ioctl_set_memory_region(struct kvm
*kvm
,
257 kvm_userspace_memory_region
*mem
,
259 long kvm_arch_vm_ioctl(struct file
*filp
,
260 unsigned int ioctl
, unsigned long arg
);
262 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
263 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
265 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
266 struct kvm_translation
*tr
);
268 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
269 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
270 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
271 struct kvm_sregs
*sregs
);
272 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
273 struct kvm_sregs
*sregs
);
274 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
275 struct kvm_mp_state
*mp_state
);
276 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
277 struct kvm_mp_state
*mp_state
);
278 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
279 struct kvm_guest_debug
*dbg
);
280 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
);
282 int kvm_arch_init(void *opaque
);
283 void kvm_arch_exit(void);
285 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
);
286 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
);
288 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
);
289 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
290 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
291 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
);
292 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
);
293 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
295 int kvm_arch_vcpu_reset(struct kvm_vcpu
*vcpu
);
296 void kvm_arch_hardware_enable(void *garbage
);
297 void kvm_arch_hardware_disable(void *garbage
);
298 int kvm_arch_hardware_setup(void);
299 void kvm_arch_hardware_unsetup(void);
300 void kvm_arch_check_processor_compat(void *rtn
);
301 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
302 int kvm_arch_interrupt_allowed(struct kvm_vcpu
*vcpu
);
304 void kvm_free_physmem(struct kvm
*kvm
);
306 struct kvm
*kvm_arch_create_vm(void);
307 void kvm_arch_destroy_vm(struct kvm
*kvm
);
308 void kvm_free_all_assigned_devices(struct kvm
*kvm
);
309 void kvm_arch_sync_events(struct kvm
*kvm
);
311 int kvm_cpu_get_interrupt(struct kvm_vcpu
*v
);
312 int kvm_cpu_has_interrupt(struct kvm_vcpu
*v
);
313 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
314 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
316 int kvm_is_mmio_pfn(pfn_t pfn
);
318 struct kvm_irq_ack_notifier
{
319 struct hlist_node link
;
321 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
324 #define KVM_ASSIGNED_MSIX_PENDING 0x1
325 struct kvm_guest_msix_entry
{
331 struct kvm_assigned_dev_kernel
{
332 struct kvm_irq_ack_notifier ack_notifier
;
333 struct work_struct interrupt_work
;
334 struct list_head list
;
338 unsigned int entries_nr
;
340 bool host_irq_disabled
;
341 struct msix_entry
*host_msix_entries
;
343 struct kvm_guest_msix_entry
*guest_msix_entries
;
344 unsigned long irq_requested_type
;
349 spinlock_t assigned_dev_lock
;
352 struct kvm_irq_mask_notifier
{
353 void (*func
)(struct kvm_irq_mask_notifier
*kimn
, bool masked
);
355 struct hlist_node link
;
358 void kvm_register_irq_mask_notifier(struct kvm
*kvm
, int irq
,
359 struct kvm_irq_mask_notifier
*kimn
);
360 void kvm_unregister_irq_mask_notifier(struct kvm
*kvm
, int irq
,
361 struct kvm_irq_mask_notifier
*kimn
);
362 void kvm_fire_mask_notifiers(struct kvm
*kvm
, int irq
, bool mask
);
364 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, int irq
, int level
);
365 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
366 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
367 struct kvm_irq_ack_notifier
*kian
);
368 void kvm_unregister_irq_ack_notifier(struct kvm_irq_ack_notifier
*kian
);
369 int kvm_request_irq_source_id(struct kvm
*kvm
);
370 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
372 /* For vcpu->arch.iommu_flags */
373 #define KVM_IOMMU_CACHE_COHERENCY 0x1
375 #ifdef CONFIG_IOMMU_API
376 int kvm_iommu_map_pages(struct kvm
*kvm
, gfn_t base_gfn
,
377 unsigned long npages
);
378 int kvm_iommu_map_guest(struct kvm
*kvm
);
379 int kvm_iommu_unmap_guest(struct kvm
*kvm
);
380 int kvm_assign_device(struct kvm
*kvm
,
381 struct kvm_assigned_dev_kernel
*assigned_dev
);
382 int kvm_deassign_device(struct kvm
*kvm
,
383 struct kvm_assigned_dev_kernel
*assigned_dev
);
384 #else /* CONFIG_IOMMU_API */
385 static inline int kvm_iommu_map_pages(struct kvm
*kvm
,
387 unsigned long npages
)
392 static inline int kvm_iommu_map_guest(struct kvm
*kvm
)
397 static inline int kvm_iommu_unmap_guest(struct kvm
*kvm
)
402 static inline int kvm_assign_device(struct kvm
*kvm
,
403 struct kvm_assigned_dev_kernel
*assigned_dev
)
408 static inline int kvm_deassign_device(struct kvm
*kvm
,
409 struct kvm_assigned_dev_kernel
*assigned_dev
)
413 #endif /* CONFIG_IOMMU_API */
415 static inline void kvm_guest_enter(void)
417 account_system_vtime(current
);
418 current
->flags
|= PF_VCPU
;
421 static inline void kvm_guest_exit(void)
423 account_system_vtime(current
);
424 current
->flags
&= ~PF_VCPU
;
427 static inline int memslot_id(struct kvm
*kvm
, struct kvm_memory_slot
*slot
)
429 return slot
- kvm
->memslots
;
432 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
434 return (gpa_t
)gfn
<< PAGE_SHIFT
;
437 static inline hpa_t
pfn_to_hpa(pfn_t pfn
)
439 return (hpa_t
)pfn
<< PAGE_SHIFT
;
442 static inline void kvm_migrate_timers(struct kvm_vcpu
*vcpu
)
444 set_bit(KVM_REQ_MIGRATE_TIMER
, &vcpu
->requests
);
452 struct kvm_stats_debugfs_item
{
455 enum kvm_stat_kind kind
;
456 struct dentry
*dentry
;
458 extern struct kvm_stats_debugfs_item debugfs_entries
[];
459 extern struct dentry
*kvm_debugfs_dir
;
461 #define KVMTRACE_5D(evt, vcpu, d1, d2, d3, d4, d5, name) \
462 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
463 vcpu, 5, d1, d2, d3, d4, d5)
464 #define KVMTRACE_4D(evt, vcpu, d1, d2, d3, d4, name) \
465 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
466 vcpu, 4, d1, d2, d3, d4, 0)
467 #define KVMTRACE_3D(evt, vcpu, d1, d2, d3, name) \
468 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
469 vcpu, 3, d1, d2, d3, 0, 0)
470 #define KVMTRACE_2D(evt, vcpu, d1, d2, name) \
471 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
472 vcpu, 2, d1, d2, 0, 0, 0)
473 #define KVMTRACE_1D(evt, vcpu, d1, name) \
474 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
475 vcpu, 1, d1, 0, 0, 0, 0)
476 #define KVMTRACE_0D(evt, vcpu, name) \
477 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
478 vcpu, 0, 0, 0, 0, 0, 0)
480 #ifdef CONFIG_KVM_TRACE
481 int kvm_trace_ioctl(unsigned int ioctl
, unsigned long arg
);
482 void kvm_trace_cleanup(void);
485 int kvm_trace_ioctl(unsigned int ioctl
, unsigned long arg
)
489 #define kvm_trace_cleanup() ((void)0)
492 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
493 static inline int mmu_notifier_retry(struct kvm_vcpu
*vcpu
, unsigned long mmu_seq
)
495 if (unlikely(vcpu
->kvm
->mmu_notifier_count
))
498 * Both reads happen under the mmu_lock and both values are
499 * modified under mmu_lock, so there's no need of smb_rmb()
500 * here in between, otherwise mmu_notifier_count should be
501 * read before mmu_notifier_seq, see
502 * mmu_notifier_invalidate_range_end write side.
504 if (vcpu
->kvm
->mmu_notifier_seq
!= mmu_seq
)
510 #ifdef CONFIG_HAVE_KVM_IRQCHIP
512 #define KVM_MAX_IRQ_ROUTES 1024
514 int kvm_setup_default_irq_routing(struct kvm
*kvm
);
515 int kvm_set_irq_routing(struct kvm
*kvm
,
516 const struct kvm_irq_routing_entry
*entries
,
519 void kvm_free_irq_routing(struct kvm
*kvm
);
523 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}