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
41 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
44 extern struct kmem_cache *kvm_vcpu_cache;
47 * It would be nice to use something smarter than a linear search, TBD...
48 * Thankfully we dont expect many devices to register (famous last words :),
49 * so until then it will suffice. At least its abstracted so we can change
54 #define NR_IOBUS_DEVS 6
55 struct kvm_io_device *devs[NR_IOBUS_DEVS];
58 void kvm_io_bus_init(struct kvm_io_bus *bus);
59 void kvm_io_bus_destroy(struct kvm_io_bus *bus);
60 struct kvm_io_device *kvm_io_bus_find_dev(struct kvm_io_bus *bus,
61 gpa_t addr, int len, int is_write);
62 void kvm_io_bus_register_dev(struct kvm_io_bus *bus,
63 struct kvm_io_device *dev);
67 #ifdef CONFIG_PREEMPT_NOTIFIERS
68 struct preempt_notifier preempt_notifier;
75 unsigned long requests;
76 struct kvm_guest_debug guest_debug;
82 struct kvm_vcpu_stat stat;
84 #ifdef CONFIG_HAS_IOMEM
86 int mmio_read_completed;
89 unsigned char mmio_data[8];
93 struct kvm_vcpu_arch arch;
96 struct kvm_memory_slot {
101 unsigned long *dirty_bitmap;
103 unsigned long rmap_pde;
106 unsigned long userspace_addr;
111 struct mutex lock; /* protects the vcpus array and APIC accesses */
113 struct rw_semaphore slots_lock;
114 struct mm_struct *mm; /* userspace tied to this vm */
116 struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS +
117 KVM_PRIVATE_MEM_SLOTS];
118 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
119 struct list_head vm_list;
120 struct kvm_io_bus mmio_bus;
121 struct kvm_io_bus pio_bus;
122 struct kvm_vm_stat stat;
123 struct kvm_arch arch;
124 atomic_t users_count;
125 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
126 struct kvm_coalesced_mmio_dev *coalesced_mmio_dev;
127 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
130 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
131 struct mmu_notifier mmu_notifier;
132 unsigned long mmu_notifier_seq;
133 long mmu_notifier_count;
137 /* The guest did something we don't support. */
138 #define pr_unimpl(vcpu, fmt, ...) \
140 if (printk_ratelimit()) \
141 printk(KERN_ERR "kvm: %i: cpu%i " fmt, \
142 current->tgid, (vcpu)->vcpu_id , ## __VA_ARGS__); \
145 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
146 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
148 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
149 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
151 void vcpu_load(struct kvm_vcpu *vcpu);
152 void vcpu_put(struct kvm_vcpu *vcpu);
154 int kvm_init(void *opaque, unsigned int vcpu_size,
155 struct module *module);
158 void kvm_get_kvm(struct kvm *kvm);
159 void kvm_put_kvm(struct kvm *kvm);
161 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
162 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
163 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
164 struct page *gva_to_page(struct kvm_vcpu *vcpu, gva_t gva);
166 extern struct page *bad_page;
167 extern pfn_t bad_pfn;
169 int is_error_page(struct page *page);
170 int is_error_pfn(pfn_t pfn);
171 int kvm_is_error_hva(unsigned long addr);
172 int kvm_set_memory_region(struct kvm *kvm,
173 struct kvm_userspace_memory_region *mem,
175 int __kvm_set_memory_region(struct kvm *kvm,
176 struct kvm_userspace_memory_region *mem,
178 int kvm_arch_set_memory_region(struct kvm *kvm,
179 struct kvm_userspace_memory_region *mem,
180 struct kvm_memory_slot old,
182 void kvm_arch_flush_shadow(struct kvm *kvm);
183 gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn);
184 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
185 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
186 void kvm_release_page_clean(struct page *page);
187 void kvm_release_page_dirty(struct page *page);
188 void kvm_set_page_dirty(struct page *page);
189 void kvm_set_page_accessed(struct page *page);
191 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
192 void kvm_release_pfn_dirty(pfn_t);
193 void kvm_release_pfn_clean(pfn_t pfn);
194 void kvm_set_pfn_dirty(pfn_t pfn);
195 void kvm_set_pfn_accessed(pfn_t pfn);
196 void kvm_get_pfn(pfn_t pfn);
198 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
200 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
202 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
203 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
204 int offset, int len);
205 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
207 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
208 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
209 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
210 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
211 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
213 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
214 void kvm_resched(struct kvm_vcpu *vcpu);
215 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
216 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
217 void kvm_flush_remote_tlbs(struct kvm *kvm);
218 void kvm_reload_remote_mmus(struct kvm *kvm);
220 long kvm_arch_dev_ioctl(struct file *filp,
221 unsigned int ioctl, unsigned long arg);
222 long kvm_arch_vcpu_ioctl(struct file *filp,
223 unsigned int ioctl, unsigned long arg);
224 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
225 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
227 int kvm_dev_ioctl_check_extension(long ext);
229 int kvm_get_dirty_log(struct kvm *kvm,
230 struct kvm_dirty_log *log, int *is_dirty);
231 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
232 struct kvm_dirty_log *log);
234 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
236 kvm_userspace_memory_region *mem,
238 long kvm_arch_vm_ioctl(struct file *filp,
239 unsigned int ioctl, unsigned long arg);
240 void kvm_arch_destroy_vm(struct kvm *kvm);
242 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
243 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
245 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
246 struct kvm_translation *tr);
248 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
249 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
250 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
251 struct kvm_sregs *sregs);
252 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
253 struct kvm_sregs *sregs);
254 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
255 struct kvm_mp_state *mp_state);
256 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
257 struct kvm_mp_state *mp_state);
258 int kvm_arch_vcpu_ioctl_debug_guest(struct kvm_vcpu *vcpu,
259 struct kvm_debug_guest *dbg);
260 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
262 int kvm_arch_init(void *opaque);
263 void kvm_arch_exit(void);
265 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
266 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
268 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
269 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
270 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
271 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
272 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
273 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
275 int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu);
276 void kvm_arch_hardware_enable(void *garbage);
277 void kvm_arch_hardware_disable(void *garbage);
278 int kvm_arch_hardware_setup(void);
279 void kvm_arch_hardware_unsetup(void);
280 void kvm_arch_check_processor_compat(void *rtn);
281 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
283 void kvm_free_physmem(struct kvm *kvm);
285 struct kvm *kvm_arch_create_vm(void);
286 void kvm_arch_destroy_vm(struct kvm *kvm);
287 void kvm_free_all_assigned_devices(struct kvm *kvm);
288 void kvm_arch_sync_events(struct kvm *kvm);
290 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
291 int kvm_cpu_has_interrupt(struct kvm_vcpu *v);
292 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
293 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
295 int kvm_is_mmio_pfn(pfn_t pfn);
297 struct kvm_irq_ack_notifier {
298 struct hlist_node link;
300 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
303 struct kvm_assigned_dev_kernel {
304 struct kvm_irq_ack_notifier ack_notifier;
305 struct work_struct interrupt_work;
306 struct list_head list;
311 bool host_irq_disabled;
313 struct msi_msg guest_msi;
314 #define KVM_ASSIGNED_DEV_GUEST_INTX (1 << 0)
315 #define KVM_ASSIGNED_DEV_GUEST_MSI (1 << 1)
316 #define KVM_ASSIGNED_DEV_HOST_INTX (1 << 8)
317 #define KVM_ASSIGNED_DEV_HOST_MSI (1 << 9)
318 unsigned long irq_requested_type;
324 void kvm_set_irq(struct kvm *kvm, int irq_source_id, int irq, int level);
325 void kvm_notify_acked_irq(struct kvm *kvm, unsigned gsi);
326 void kvm_register_irq_ack_notifier(struct kvm *kvm,
327 struct kvm_irq_ack_notifier *kian);
328 void kvm_unregister_irq_ack_notifier(struct kvm_irq_ack_notifier *kian);
329 int kvm_request_irq_source_id(struct kvm *kvm);
330 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
332 #ifdef CONFIG_IOMMU_API
333 int kvm_iommu_map_pages(struct kvm *kvm, gfn_t base_gfn,
334 unsigned long npages);
335 int kvm_iommu_map_guest(struct kvm *kvm);
336 int kvm_iommu_unmap_guest(struct kvm *kvm);
337 int kvm_assign_device(struct kvm *kvm,
338 struct kvm_assigned_dev_kernel *assigned_dev);
339 int kvm_deassign_device(struct kvm *kvm,
340 struct kvm_assigned_dev_kernel *assigned_dev);
341 #else /* CONFIG_IOMMU_API */
342 static inline int kvm_iommu_map_pages(struct kvm *kvm,
344 unsigned long npages)
349 static inline int kvm_iommu_map_guest(struct kvm *kvm)
354 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
359 static inline int kvm_assign_device(struct kvm *kvm,
360 struct kvm_assigned_dev_kernel *assigned_dev)
365 static inline int kvm_deassign_device(struct kvm *kvm,
366 struct kvm_assigned_dev_kernel *assigned_dev)
370 #endif /* CONFIG_IOMMU_API */
372 static inline void kvm_guest_enter(void)
374 account_system_vtime(current);
375 current->flags |= PF_VCPU;
378 static inline void kvm_guest_exit(void)
380 account_system_vtime(current);
381 current->flags &= ~PF_VCPU;
384 static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
386 return slot - kvm->memslots;
389 static inline gpa_t gfn_to_gpa(gfn_t gfn)
391 return (gpa_t)gfn << PAGE_SHIFT;
394 static inline hpa_t pfn_to_hpa(pfn_t pfn)
396 return (hpa_t)pfn << PAGE_SHIFT;
399 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
401 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
409 struct kvm_stats_debugfs_item {
412 enum kvm_stat_kind kind;
413 struct dentry *dentry;
415 extern struct kvm_stats_debugfs_item debugfs_entries[];
416 extern struct dentry *kvm_debugfs_dir;
418 #define KVMTRACE_5D(evt, vcpu, d1, d2, d3, d4, d5, name) \
419 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
420 vcpu, 5, d1, d2, d3, d4, d5)
421 #define KVMTRACE_4D(evt, vcpu, d1, d2, d3, d4, name) \
422 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
423 vcpu, 4, d1, d2, d3, d4, 0)
424 #define KVMTRACE_3D(evt, vcpu, d1, d2, d3, name) \
425 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
426 vcpu, 3, d1, d2, d3, 0, 0)
427 #define KVMTRACE_2D(evt, vcpu, d1, d2, name) \
428 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
429 vcpu, 2, d1, d2, 0, 0, 0)
430 #define KVMTRACE_1D(evt, vcpu, d1, name) \
431 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
432 vcpu, 1, d1, 0, 0, 0, 0)
433 #define KVMTRACE_0D(evt, vcpu, name) \
434 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
435 vcpu, 0, 0, 0, 0, 0, 0)
437 #ifdef CONFIG_KVM_TRACE
438 int kvm_trace_ioctl(unsigned int ioctl, unsigned long arg);
439 void kvm_trace_cleanup(void);
442 int kvm_trace_ioctl(unsigned int ioctl, unsigned long arg)
446 #define kvm_trace_cleanup() ((void)0)
449 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
450 static inline int mmu_notifier_retry(struct kvm_vcpu *vcpu, unsigned long mmu_seq)
452 if (unlikely(vcpu->kvm->mmu_notifier_count))
455 * Both reads happen under the mmu_lock and both values are
456 * modified under mmu_lock, so there's no need of smb_rmb()
457 * here in between, otherwise mmu_notifier_count should be
458 * read before mmu_notifier_seq, see
459 * mmu_notifier_invalidate_range_end write side.
461 if (vcpu->kvm->mmu_notifier_seq != mmu_seq)