xref: /linux-6.15/include/linux/kvm_host.h (revision b8bb7671)
1 #ifndef __KVM_HOST_H
2 #define __KVM_HOST_H
3 
4 /*
5  * This work is licensed under the terms of the GNU GPL, version 2.  See
6  * the COPYING file in the top-level directory.
7  */
8 
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>
16 #include <linux/mm.h>
17 #include <linux/preempt.h>
18 #include <linux/marker.h>
19 #include <linux/msi.h>
20 #include <asm/signal.h>
21 
22 #include <linux/kvm.h>
23 #include <linux/kvm_para.h>
24 
25 #include <linux/kvm_types.h>
26 
27 #include <asm/kvm_host.h>
28 
29 /*
30  * vcpu->requests bit members
31  */
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 
42 #define KVM_USERSPACE_IRQ_SOURCE_ID	0
43 
44 struct kvm_vcpu;
45 extern struct kmem_cache *kvm_vcpu_cache;
46 
47 /*
48  * It would be nice to use something smarter than a linear search, TBD...
49  * Thankfully we dont expect many devices to register (famous last words :),
50  * so until then it will suffice.  At least its abstracted so we can change
51  * in one place.
52  */
53 struct kvm_io_bus {
54 	int                   dev_count;
55 #define NR_IOBUS_DEVS 6
56 	struct kvm_io_device *devs[NR_IOBUS_DEVS];
57 };
58 
59 void kvm_io_bus_init(struct kvm_io_bus *bus);
60 void kvm_io_bus_destroy(struct kvm_io_bus *bus);
61 struct kvm_io_device *kvm_io_bus_find_dev(struct kvm_io_bus *bus,
62 					  gpa_t addr, int len, int is_write);
63 void kvm_io_bus_register_dev(struct kvm_io_bus *bus,
64 			     struct kvm_io_device *dev);
65 
66 struct kvm_vcpu {
67 	struct kvm *kvm;
68 #ifdef CONFIG_PREEMPT_NOTIFIERS
69 	struct preempt_notifier preempt_notifier;
70 #endif
71 	int vcpu_id;
72 	struct mutex mutex;
73 	int   cpu;
74 	struct kvm_run *run;
75 	int guest_mode;
76 	unsigned long requests;
77 	unsigned long guest_debug;
78 	int fpu_active;
79 	int guest_fpu_loaded;
80 	wait_queue_head_t wq;
81 	int sigset_active;
82 	sigset_t sigset;
83 	struct kvm_vcpu_stat stat;
84 
85 #ifdef CONFIG_HAS_IOMEM
86 	int mmio_needed;
87 	int mmio_read_completed;
88 	int mmio_is_write;
89 	int mmio_size;
90 	unsigned char mmio_data[8];
91 	gpa_t mmio_phys_addr;
92 #endif
93 
94 	struct kvm_vcpu_arch arch;
95 };
96 
97 struct kvm_memory_slot {
98 	gfn_t base_gfn;
99 	unsigned long npages;
100 	unsigned long flags;
101 	unsigned long *rmap;
102 	unsigned long *dirty_bitmap;
103 	struct {
104 		unsigned long rmap_pde;
105 		int write_count;
106 	} *lpage_info;
107 	unsigned long userspace_addr;
108 	int user_alloc;
109 };
110 
111 struct kvm_kernel_irq_routing_entry {
112 	u32 gsi;
113 	int (*set)(struct kvm_kernel_irq_routing_entry *e,
114 		    struct kvm *kvm, int level);
115 	union {
116 		struct {
117 			unsigned irqchip;
118 			unsigned pin;
119 		} irqchip;
120 		struct msi_msg msi;
121 	};
122 	struct list_head link;
123 };
124 
125 struct kvm {
126 	struct mutex lock; /* protects the vcpus array and APIC accesses */
127 	spinlock_t mmu_lock;
128 	struct rw_semaphore slots_lock;
129 	struct mm_struct *mm; /* userspace tied to this vm */
130 	int nmemslots;
131 	struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS +
132 					KVM_PRIVATE_MEM_SLOTS];
133 	struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
134 	struct list_head vm_list;
135 	struct kvm_io_bus mmio_bus;
136 	struct kvm_io_bus pio_bus;
137 	struct kvm_vm_stat stat;
138 	struct kvm_arch arch;
139 	atomic_t users_count;
140 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
141 	struct kvm_coalesced_mmio_dev *coalesced_mmio_dev;
142 	struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
143 #endif
144 
145 #ifdef CONFIG_HAVE_KVM_IRQCHIP
146 	struct list_head irq_routing; /* of kvm_kernel_irq_routing_entry */
147 	struct hlist_head mask_notifier_list;
148 #endif
149 
150 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
151 	struct mmu_notifier mmu_notifier;
152 	unsigned long mmu_notifier_seq;
153 	long mmu_notifier_count;
154 #endif
155 };
156 
157 /* The guest did something we don't support. */
158 #define pr_unimpl(vcpu, fmt, ...)					\
159  do {									\
160 	if (printk_ratelimit())						\
161 		printk(KERN_ERR "kvm: %i: cpu%i " fmt,			\
162 		       current->tgid, (vcpu)->vcpu_id , ## __VA_ARGS__); \
163  } while (0)
164 
165 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
166 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
167 
168 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
169 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
170 
171 void vcpu_load(struct kvm_vcpu *vcpu);
172 void vcpu_put(struct kvm_vcpu *vcpu);
173 
174 int kvm_init(void *opaque, unsigned int vcpu_size,
175 		  struct module *module);
176 void kvm_exit(void);
177 
178 void kvm_get_kvm(struct kvm *kvm);
179 void kvm_put_kvm(struct kvm *kvm);
180 
181 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
182 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
183 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
184 struct page *gva_to_page(struct kvm_vcpu *vcpu, gva_t gva);
185 
186 extern struct page *bad_page;
187 extern pfn_t bad_pfn;
188 
189 int is_error_page(struct page *page);
190 int is_error_pfn(pfn_t pfn);
191 int kvm_is_error_hva(unsigned long addr);
192 int kvm_set_memory_region(struct kvm *kvm,
193 			  struct kvm_userspace_memory_region *mem,
194 			  int user_alloc);
195 int __kvm_set_memory_region(struct kvm *kvm,
196 			    struct kvm_userspace_memory_region *mem,
197 			    int user_alloc);
198 int kvm_arch_set_memory_region(struct kvm *kvm,
199 				struct kvm_userspace_memory_region *mem,
200 				struct kvm_memory_slot old,
201 				int user_alloc);
202 void kvm_arch_flush_shadow(struct kvm *kvm);
203 gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn);
204 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
205 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
206 void kvm_release_page_clean(struct page *page);
207 void kvm_release_page_dirty(struct page *page);
208 void kvm_set_page_dirty(struct page *page);
209 void kvm_set_page_accessed(struct page *page);
210 
211 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
212 void kvm_release_pfn_dirty(pfn_t);
213 void kvm_release_pfn_clean(pfn_t pfn);
214 void kvm_set_pfn_dirty(pfn_t pfn);
215 void kvm_set_pfn_accessed(pfn_t pfn);
216 void kvm_get_pfn(pfn_t pfn);
217 
218 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
219 			int len);
220 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
221 			  unsigned long len);
222 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
223 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
224 			 int offset, int len);
225 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
226 		    unsigned long len);
227 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
228 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
229 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
230 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
231 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
232 
233 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
234 void kvm_resched(struct kvm_vcpu *vcpu);
235 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
236 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
237 void kvm_flush_remote_tlbs(struct kvm *kvm);
238 void kvm_reload_remote_mmus(struct kvm *kvm);
239 
240 long kvm_arch_dev_ioctl(struct file *filp,
241 			unsigned int ioctl, unsigned long arg);
242 long kvm_arch_vcpu_ioctl(struct file *filp,
243 			 unsigned int ioctl, unsigned long arg);
244 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
245 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
246 
247 int kvm_dev_ioctl_check_extension(long ext);
248 
249 int kvm_get_dirty_log(struct kvm *kvm,
250 			struct kvm_dirty_log *log, int *is_dirty);
251 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
252 				struct kvm_dirty_log *log);
253 
254 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
255 				   struct
256 				   kvm_userspace_memory_region *mem,
257 				   int user_alloc);
258 long kvm_arch_vm_ioctl(struct file *filp,
259 		       unsigned int ioctl, unsigned long arg);
260 
261 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
262 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
263 
264 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
265 				    struct kvm_translation *tr);
266 
267 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
268 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
269 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
270 				  struct kvm_sregs *sregs);
271 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
272 				  struct kvm_sregs *sregs);
273 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
274 				    struct kvm_mp_state *mp_state);
275 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
276 				    struct kvm_mp_state *mp_state);
277 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
278 					struct kvm_guest_debug *dbg);
279 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
280 
281 int kvm_arch_init(void *opaque);
282 void kvm_arch_exit(void);
283 
284 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
285 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
286 
287 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
288 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
289 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
290 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
291 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
292 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
293 
294 int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu);
295 void kvm_arch_hardware_enable(void *garbage);
296 void kvm_arch_hardware_disable(void *garbage);
297 int kvm_arch_hardware_setup(void);
298 void kvm_arch_hardware_unsetup(void);
299 void kvm_arch_check_processor_compat(void *rtn);
300 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
301 
302 void kvm_free_physmem(struct kvm *kvm);
303 
304 struct  kvm *kvm_arch_create_vm(void);
305 void kvm_arch_destroy_vm(struct kvm *kvm);
306 void kvm_free_all_assigned_devices(struct kvm *kvm);
307 void kvm_arch_sync_events(struct kvm *kvm);
308 
309 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
310 int kvm_cpu_has_interrupt(struct kvm_vcpu *v);
311 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
312 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
313 
314 int kvm_is_mmio_pfn(pfn_t pfn);
315 
316 struct kvm_irq_ack_notifier {
317 	struct hlist_node link;
318 	unsigned gsi;
319 	void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
320 };
321 
322 struct kvm_assigned_dev_kernel {
323 	struct kvm_irq_ack_notifier ack_notifier;
324 	struct work_struct interrupt_work;
325 	struct list_head list;
326 	int assigned_dev_id;
327 	int host_busnr;
328 	int host_devfn;
329 	int host_irq;
330 	bool host_irq_disabled;
331 	int guest_irq;
332 #define KVM_ASSIGNED_DEV_GUEST_INTX	(1 << 0)
333 #define KVM_ASSIGNED_DEV_GUEST_MSI	(1 << 1)
334 #define KVM_ASSIGNED_DEV_HOST_INTX	(1 << 8)
335 #define KVM_ASSIGNED_DEV_HOST_MSI	(1 << 9)
336 	unsigned long irq_requested_type;
337 	int irq_source_id;
338 	int flags;
339 	struct pci_dev *dev;
340 	struct kvm *kvm;
341 };
342 
343 struct kvm_irq_mask_notifier {
344 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
345 	int irq;
346 	struct hlist_node link;
347 };
348 
349 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
350 				    struct kvm_irq_mask_notifier *kimn);
351 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
352 				      struct kvm_irq_mask_notifier *kimn);
353 void kvm_fire_mask_notifiers(struct kvm *kvm, int irq, bool mask);
354 
355 int kvm_set_irq(struct kvm *kvm, int irq_source_id, int irq, int level);
356 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
357 void kvm_register_irq_ack_notifier(struct kvm *kvm,
358 				   struct kvm_irq_ack_notifier *kian);
359 void kvm_unregister_irq_ack_notifier(struct kvm_irq_ack_notifier *kian);
360 int kvm_request_irq_source_id(struct kvm *kvm);
361 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
362 
363 #ifdef CONFIG_IOMMU_API
364 int kvm_iommu_map_pages(struct kvm *kvm, gfn_t base_gfn,
365 			unsigned long npages);
366 int kvm_iommu_map_guest(struct kvm *kvm);
367 int kvm_iommu_unmap_guest(struct kvm *kvm);
368 int kvm_assign_device(struct kvm *kvm,
369 		      struct kvm_assigned_dev_kernel *assigned_dev);
370 int kvm_deassign_device(struct kvm *kvm,
371 			struct kvm_assigned_dev_kernel *assigned_dev);
372 #else /* CONFIG_IOMMU_API */
373 static inline int kvm_iommu_map_pages(struct kvm *kvm,
374 				      gfn_t base_gfn,
375 				      unsigned long npages)
376 {
377 	return 0;
378 }
379 
380 static inline int kvm_iommu_map_guest(struct kvm *kvm)
381 {
382 	return -ENODEV;
383 }
384 
385 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
386 {
387 	return 0;
388 }
389 
390 static inline int kvm_assign_device(struct kvm *kvm,
391 		struct kvm_assigned_dev_kernel *assigned_dev)
392 {
393 	return 0;
394 }
395 
396 static inline int kvm_deassign_device(struct kvm *kvm,
397 		struct kvm_assigned_dev_kernel *assigned_dev)
398 {
399 	return 0;
400 }
401 #endif /* CONFIG_IOMMU_API */
402 
403 static inline void kvm_guest_enter(void)
404 {
405 	account_system_vtime(current);
406 	current->flags |= PF_VCPU;
407 }
408 
409 static inline void kvm_guest_exit(void)
410 {
411 	account_system_vtime(current);
412 	current->flags &= ~PF_VCPU;
413 }
414 
415 static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
416 {
417 	return slot - kvm->memslots;
418 }
419 
420 static inline gpa_t gfn_to_gpa(gfn_t gfn)
421 {
422 	return (gpa_t)gfn << PAGE_SHIFT;
423 }
424 
425 static inline hpa_t pfn_to_hpa(pfn_t pfn)
426 {
427 	return (hpa_t)pfn << PAGE_SHIFT;
428 }
429 
430 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
431 {
432 	set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
433 }
434 
435 enum kvm_stat_kind {
436 	KVM_STAT_VM,
437 	KVM_STAT_VCPU,
438 };
439 
440 struct kvm_stats_debugfs_item {
441 	const char *name;
442 	int offset;
443 	enum kvm_stat_kind kind;
444 	struct dentry *dentry;
445 };
446 extern struct kvm_stats_debugfs_item debugfs_entries[];
447 extern struct dentry *kvm_debugfs_dir;
448 
449 #define KVMTRACE_5D(evt, vcpu, d1, d2, d3, d4, d5, name) \
450 	trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
451 						vcpu, 5, d1, d2, d3, d4, d5)
452 #define KVMTRACE_4D(evt, vcpu, d1, d2, d3, d4, name) \
453 	trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
454 						vcpu, 4, d1, d2, d3, d4, 0)
455 #define KVMTRACE_3D(evt, vcpu, d1, d2, d3, name) \
456 	trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
457 						vcpu, 3, d1, d2, d3, 0, 0)
458 #define KVMTRACE_2D(evt, vcpu, d1, d2, name) \
459 	trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
460 						vcpu, 2, d1, d2, 0, 0, 0)
461 #define KVMTRACE_1D(evt, vcpu, d1, name) \
462 	trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
463 						vcpu, 1, d1, 0, 0, 0, 0)
464 #define KVMTRACE_0D(evt, vcpu, name) \
465 	trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
466 						vcpu, 0, 0, 0, 0, 0, 0)
467 
468 #ifdef CONFIG_KVM_TRACE
469 int kvm_trace_ioctl(unsigned int ioctl, unsigned long arg);
470 void kvm_trace_cleanup(void);
471 #else
472 static inline
473 int kvm_trace_ioctl(unsigned int ioctl, unsigned long arg)
474 {
475 	return -EINVAL;
476 }
477 #define kvm_trace_cleanup() ((void)0)
478 #endif
479 
480 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
481 static inline int mmu_notifier_retry(struct kvm_vcpu *vcpu, unsigned long mmu_seq)
482 {
483 	if (unlikely(vcpu->kvm->mmu_notifier_count))
484 		return 1;
485 	/*
486 	 * Both reads happen under the mmu_lock and both values are
487 	 * modified under mmu_lock, so there's no need of smb_rmb()
488 	 * here in between, otherwise mmu_notifier_count should be
489 	 * read before mmu_notifier_seq, see
490 	 * mmu_notifier_invalidate_range_end write side.
491 	 */
492 	if (vcpu->kvm->mmu_notifier_seq != mmu_seq)
493 		return 1;
494 	return 0;
495 }
496 #endif
497 
498 #ifdef CONFIG_HAVE_KVM_IRQCHIP
499 
500 #define KVM_MAX_IRQ_ROUTES 1024
501 
502 int kvm_setup_default_irq_routing(struct kvm *kvm);
503 int kvm_set_irq_routing(struct kvm *kvm,
504 			const struct kvm_irq_routing_entry *entries,
505 			unsigned nr,
506 			unsigned flags);
507 void kvm_free_irq_routing(struct kvm *kvm);
508 
509 #else
510 
511 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
512 
513 #endif
514 
515 #endif
516