xref: /linux-6.15/include/linux/kvm_host.h (revision e8fa600e)
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/bug.h>
17 #include <linux/mm.h>
18 #include <linux/mmu_notifier.h>
19 #include <linux/preempt.h>
20 #include <linux/msi.h>
21 #include <linux/slab.h>
22 #include <linux/rcupdate.h>
23 #include <linux/ratelimit.h>
24 #include <asm/signal.h>
25 
26 #include <linux/kvm.h>
27 #include <linux/kvm_para.h>
28 
29 #include <linux/kvm_types.h>
30 
31 #include <asm/kvm_host.h>
32 
33 #ifndef KVM_MMIO_SIZE
34 #define KVM_MMIO_SIZE 8
35 #endif
36 
37 /*
38  * If we support unaligned MMIO, at most one fragment will be split into two:
39  */
40 #ifdef KVM_UNALIGNED_MMIO
41 #  define KVM_EXTRA_MMIO_FRAGMENTS 1
42 #else
43 #  define KVM_EXTRA_MMIO_FRAGMENTS 0
44 #endif
45 
46 #define KVM_USER_MMIO_SIZE 8
47 
48 #define KVM_MAX_MMIO_FRAGMENTS \
49 	(KVM_MMIO_SIZE / KVM_USER_MMIO_SIZE + KVM_EXTRA_MMIO_FRAGMENTS)
50 
51 /*
52  * vcpu->requests bit members
53  */
54 #define KVM_REQ_TLB_FLUSH          0
55 #define KVM_REQ_MIGRATE_TIMER      1
56 #define KVM_REQ_REPORT_TPR_ACCESS  2
57 #define KVM_REQ_MMU_RELOAD         3
58 #define KVM_REQ_TRIPLE_FAULT       4
59 #define KVM_REQ_PENDING_TIMER      5
60 #define KVM_REQ_UNHALT             6
61 #define KVM_REQ_MMU_SYNC           7
62 #define KVM_REQ_CLOCK_UPDATE       8
63 #define KVM_REQ_KICK               9
64 #define KVM_REQ_DEACTIVATE_FPU    10
65 #define KVM_REQ_EVENT             11
66 #define KVM_REQ_APF_HALT          12
67 #define KVM_REQ_STEAL_UPDATE      13
68 #define KVM_REQ_NMI               14
69 #define KVM_REQ_IMMEDIATE_EXIT    15
70 #define KVM_REQ_PMU               16
71 #define KVM_REQ_PMI               17
72 
73 #define KVM_USERSPACE_IRQ_SOURCE_ID	0
74 
75 struct kvm;
76 struct kvm_vcpu;
77 extern struct kmem_cache *kvm_vcpu_cache;
78 
79 struct kvm_io_range {
80 	gpa_t addr;
81 	int len;
82 	struct kvm_io_device *dev;
83 };
84 
85 #define NR_IOBUS_DEVS 1000
86 
87 struct kvm_io_bus {
88 	int                   dev_count;
89 	struct kvm_io_range range[];
90 };
91 
92 enum kvm_bus {
93 	KVM_MMIO_BUS,
94 	KVM_PIO_BUS,
95 	KVM_NR_BUSES
96 };
97 
98 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
99 		     int len, const void *val);
100 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
101 		    void *val);
102 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
103 			    int len, struct kvm_io_device *dev);
104 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
105 			      struct kvm_io_device *dev);
106 
107 #ifdef CONFIG_KVM_ASYNC_PF
108 struct kvm_async_pf {
109 	struct work_struct work;
110 	struct list_head link;
111 	struct list_head queue;
112 	struct kvm_vcpu *vcpu;
113 	struct mm_struct *mm;
114 	gva_t gva;
115 	unsigned long addr;
116 	struct kvm_arch_async_pf arch;
117 	struct page *page;
118 	bool done;
119 };
120 
121 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
122 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
123 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
124 		       struct kvm_arch_async_pf *arch);
125 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
126 #endif
127 
128 enum {
129 	OUTSIDE_GUEST_MODE,
130 	IN_GUEST_MODE,
131 	EXITING_GUEST_MODE,
132 	READING_SHADOW_PAGE_TABLES,
133 };
134 
135 /*
136  * Sometimes a large or cross-page mmio needs to be broken up into separate
137  * exits for userspace servicing.
138  */
139 struct kvm_mmio_fragment {
140 	gpa_t gpa;
141 	void *data;
142 	unsigned len;
143 };
144 
145 struct kvm_vcpu {
146 	struct kvm *kvm;
147 #ifdef CONFIG_PREEMPT_NOTIFIERS
148 	struct preempt_notifier preempt_notifier;
149 #endif
150 	int cpu;
151 	int vcpu_id;
152 	int srcu_idx;
153 	int mode;
154 	unsigned long requests;
155 	unsigned long guest_debug;
156 
157 	struct mutex mutex;
158 	struct kvm_run *run;
159 
160 	int fpu_active;
161 	int guest_fpu_loaded, guest_xcr0_loaded;
162 	wait_queue_head_t wq;
163 	struct pid *pid;
164 	int sigset_active;
165 	sigset_t sigset;
166 	struct kvm_vcpu_stat stat;
167 
168 #ifdef CONFIG_HAS_IOMEM
169 	int mmio_needed;
170 	int mmio_read_completed;
171 	int mmio_is_write;
172 	int mmio_cur_fragment;
173 	int mmio_nr_fragments;
174 	struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
175 #endif
176 
177 #ifdef CONFIG_KVM_ASYNC_PF
178 	struct {
179 		u32 queued;
180 		struct list_head queue;
181 		struct list_head done;
182 		spinlock_t lock;
183 	} async_pf;
184 #endif
185 
186 	struct kvm_vcpu_arch arch;
187 };
188 
189 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
190 {
191 	return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
192 }
193 
194 /*
195  * Some of the bitops functions do not support too long bitmaps.
196  * This number must be determined not to exceed such limits.
197  */
198 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
199 
200 struct kvm_memory_slot {
201 	gfn_t base_gfn;
202 	unsigned long npages;
203 	unsigned long flags;
204 	unsigned long *rmap;
205 	unsigned long *dirty_bitmap;
206 	struct kvm_arch_memory_slot arch;
207 	unsigned long userspace_addr;
208 	int user_alloc;
209 	int id;
210 };
211 
212 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
213 {
214 	return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
215 }
216 
217 struct kvm_kernel_irq_routing_entry {
218 	u32 gsi;
219 	u32 type;
220 	int (*set)(struct kvm_kernel_irq_routing_entry *e,
221 		   struct kvm *kvm, int irq_source_id, int level);
222 	union {
223 		struct {
224 			unsigned irqchip;
225 			unsigned pin;
226 		} irqchip;
227 		struct msi_msg msi;
228 	};
229 	struct hlist_node link;
230 };
231 
232 #ifdef __KVM_HAVE_IOAPIC
233 
234 struct kvm_irq_routing_table {
235 	int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
236 	struct kvm_kernel_irq_routing_entry *rt_entries;
237 	u32 nr_rt_entries;
238 	/*
239 	 * Array indexed by gsi. Each entry contains list of irq chips
240 	 * the gsi is connected to.
241 	 */
242 	struct hlist_head map[0];
243 };
244 
245 #else
246 
247 struct kvm_irq_routing_table {};
248 
249 #endif
250 
251 #ifndef KVM_MEM_SLOTS_NUM
252 #define KVM_MEM_SLOTS_NUM (KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
253 #endif
254 
255 /*
256  * Note:
257  * memslots are not sorted by id anymore, please use id_to_memslot()
258  * to get the memslot by its id.
259  */
260 struct kvm_memslots {
261 	u64 generation;
262 	struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
263 	/* The mapping table from slot id to the index in memslots[]. */
264 	int id_to_index[KVM_MEM_SLOTS_NUM];
265 };
266 
267 struct kvm {
268 	spinlock_t mmu_lock;
269 	struct mutex slots_lock;
270 	struct mm_struct *mm; /* userspace tied to this vm */
271 	struct kvm_memslots *memslots;
272 	struct srcu_struct srcu;
273 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
274 	u32 bsp_vcpu_id;
275 #endif
276 	struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
277 	atomic_t online_vcpus;
278 	int last_boosted_vcpu;
279 	struct list_head vm_list;
280 	struct mutex lock;
281 	struct kvm_io_bus *buses[KVM_NR_BUSES];
282 #ifdef CONFIG_HAVE_KVM_EVENTFD
283 	struct {
284 		spinlock_t        lock;
285 		struct list_head  items;
286 	} irqfds;
287 	struct list_head ioeventfds;
288 #endif
289 	struct kvm_vm_stat stat;
290 	struct kvm_arch arch;
291 	atomic_t users_count;
292 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
293 	struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
294 	spinlock_t ring_lock;
295 	struct list_head coalesced_zones;
296 #endif
297 
298 	struct mutex irq_lock;
299 #ifdef CONFIG_HAVE_KVM_IRQCHIP
300 	/*
301 	 * Update side is protected by irq_lock and,
302 	 * if configured, irqfds.lock.
303 	 */
304 	struct kvm_irq_routing_table __rcu *irq_routing;
305 	struct hlist_head mask_notifier_list;
306 	struct hlist_head irq_ack_notifier_list;
307 #endif
308 
309 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
310 	struct mmu_notifier mmu_notifier;
311 	unsigned long mmu_notifier_seq;
312 	long mmu_notifier_count;
313 #endif
314 	long tlbs_dirty;
315 };
316 
317 #define kvm_err(fmt, ...) \
318 	pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
319 #define kvm_info(fmt, ...) \
320 	pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
321 #define kvm_debug(fmt, ...) \
322 	pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
323 #define kvm_pr_unimpl(fmt, ...) \
324 	pr_err_ratelimited("kvm [%i]: " fmt, \
325 			   task_tgid_nr(current), ## __VA_ARGS__)
326 
327 /* The guest did something we don't support. */
328 #define vcpu_unimpl(vcpu, fmt, ...)					\
329 	kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
330 
331 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
332 {
333 	smp_rmb();
334 	return kvm->vcpus[i];
335 }
336 
337 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
338 	for (idx = 0; \
339 	     idx < atomic_read(&kvm->online_vcpus) && \
340 	     (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
341 	     idx++)
342 
343 #define kvm_for_each_memslot(memslot, slots)	\
344 	for (memslot = &slots->memslots[0];	\
345 	      memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
346 		memslot++)
347 
348 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
349 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
350 
351 void vcpu_load(struct kvm_vcpu *vcpu);
352 void vcpu_put(struct kvm_vcpu *vcpu);
353 
354 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
355 		  struct module *module);
356 void kvm_exit(void);
357 
358 void kvm_get_kvm(struct kvm *kvm);
359 void kvm_put_kvm(struct kvm *kvm);
360 void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new);
361 
362 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
363 {
364 	return rcu_dereference_check(kvm->memslots,
365 			srcu_read_lock_held(&kvm->srcu)
366 			|| lockdep_is_held(&kvm->slots_lock));
367 }
368 
369 static inline struct kvm_memory_slot *
370 id_to_memslot(struct kvm_memslots *slots, int id)
371 {
372 	int index = slots->id_to_index[id];
373 	struct kvm_memory_slot *slot;
374 
375 	slot = &slots->memslots[index];
376 
377 	WARN_ON(slot->id != id);
378 	return slot;
379 }
380 
381 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
382 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
383 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
384 
385 extern struct page *bad_page;
386 extern struct page *fault_page;
387 
388 extern pfn_t bad_pfn;
389 extern pfn_t fault_pfn;
390 
391 int is_error_page(struct page *page);
392 int is_error_pfn(pfn_t pfn);
393 int is_hwpoison_pfn(pfn_t pfn);
394 int is_fault_pfn(pfn_t pfn);
395 int is_noslot_pfn(pfn_t pfn);
396 int is_invalid_pfn(pfn_t pfn);
397 int kvm_is_error_hva(unsigned long addr);
398 int kvm_set_memory_region(struct kvm *kvm,
399 			  struct kvm_userspace_memory_region *mem,
400 			  int user_alloc);
401 int __kvm_set_memory_region(struct kvm *kvm,
402 			    struct kvm_userspace_memory_region *mem,
403 			    int user_alloc);
404 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
405 			   struct kvm_memory_slot *dont);
406 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages);
407 int kvm_arch_prepare_memory_region(struct kvm *kvm,
408 				struct kvm_memory_slot *memslot,
409 				struct kvm_memory_slot old,
410 				struct kvm_userspace_memory_region *mem,
411 				int user_alloc);
412 void kvm_arch_commit_memory_region(struct kvm *kvm,
413 				struct kvm_userspace_memory_region *mem,
414 				struct kvm_memory_slot old,
415 				int user_alloc);
416 bool kvm_largepages_enabled(void);
417 void kvm_disable_largepages(void);
418 void kvm_arch_flush_shadow(struct kvm *kvm);
419 
420 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
421 			    int nr_pages);
422 
423 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
424 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
425 void kvm_release_page_clean(struct page *page);
426 void kvm_release_page_dirty(struct page *page);
427 void kvm_set_page_dirty(struct page *page);
428 void kvm_set_page_accessed(struct page *page);
429 
430 pfn_t hva_to_pfn_atomic(struct kvm *kvm, unsigned long addr);
431 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
432 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
433 		       bool write_fault, bool *writable);
434 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
435 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
436 		      bool *writable);
437 pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
438 			 struct kvm_memory_slot *slot, gfn_t gfn);
439 void kvm_release_pfn_dirty(pfn_t);
440 void kvm_release_pfn_clean(pfn_t pfn);
441 void kvm_set_pfn_dirty(pfn_t pfn);
442 void kvm_set_pfn_accessed(pfn_t pfn);
443 void kvm_get_pfn(pfn_t pfn);
444 
445 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
446 			int len);
447 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
448 			  unsigned long len);
449 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
450 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
451 			   void *data, unsigned long len);
452 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
453 			 int offset, int len);
454 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
455 		    unsigned long len);
456 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
457 			   void *data, unsigned long len);
458 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
459 			      gpa_t gpa);
460 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
461 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
462 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
463 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
464 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
465 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
466 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
467 			     gfn_t gfn);
468 
469 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
470 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
471 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
472 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
473 void kvm_resched(struct kvm_vcpu *vcpu);
474 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
475 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
476 
477 void kvm_flush_remote_tlbs(struct kvm *kvm);
478 void kvm_reload_remote_mmus(struct kvm *kvm);
479 
480 long kvm_arch_dev_ioctl(struct file *filp,
481 			unsigned int ioctl, unsigned long arg);
482 long kvm_arch_vcpu_ioctl(struct file *filp,
483 			 unsigned int ioctl, unsigned long arg);
484 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
485 
486 int kvm_dev_ioctl_check_extension(long ext);
487 
488 int kvm_get_dirty_log(struct kvm *kvm,
489 			struct kvm_dirty_log *log, int *is_dirty);
490 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
491 				struct kvm_dirty_log *log);
492 
493 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
494 				   struct
495 				   kvm_userspace_memory_region *mem,
496 				   int user_alloc);
497 long kvm_arch_vm_ioctl(struct file *filp,
498 		       unsigned int ioctl, unsigned long arg);
499 
500 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
501 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
502 
503 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
504 				    struct kvm_translation *tr);
505 
506 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
507 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
508 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
509 				  struct kvm_sregs *sregs);
510 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
511 				  struct kvm_sregs *sregs);
512 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
513 				    struct kvm_mp_state *mp_state);
514 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
515 				    struct kvm_mp_state *mp_state);
516 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
517 					struct kvm_guest_debug *dbg);
518 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
519 
520 int kvm_arch_init(void *opaque);
521 void kvm_arch_exit(void);
522 
523 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
524 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
525 
526 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
527 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
528 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
529 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
530 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
531 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
532 
533 int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu);
534 int kvm_arch_hardware_enable(void *garbage);
535 void kvm_arch_hardware_disable(void *garbage);
536 int kvm_arch_hardware_setup(void);
537 void kvm_arch_hardware_unsetup(void);
538 void kvm_arch_check_processor_compat(void *rtn);
539 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
540 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
541 
542 void kvm_free_physmem(struct kvm *kvm);
543 
544 void *kvm_kvzalloc(unsigned long size);
545 void kvm_kvfree(const void *addr);
546 
547 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
548 static inline struct kvm *kvm_arch_alloc_vm(void)
549 {
550 	return kzalloc(sizeof(struct kvm), GFP_KERNEL);
551 }
552 
553 static inline void kvm_arch_free_vm(struct kvm *kvm)
554 {
555 	kfree(kvm);
556 }
557 #endif
558 
559 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
560 {
561 #ifdef __KVM_HAVE_ARCH_WQP
562 	return vcpu->arch.wqp;
563 #else
564 	return &vcpu->wq;
565 #endif
566 }
567 
568 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
569 void kvm_arch_destroy_vm(struct kvm *kvm);
570 void kvm_free_all_assigned_devices(struct kvm *kvm);
571 void kvm_arch_sync_events(struct kvm *kvm);
572 
573 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
574 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
575 
576 int kvm_is_mmio_pfn(pfn_t pfn);
577 
578 struct kvm_irq_ack_notifier {
579 	struct hlist_node link;
580 	unsigned gsi;
581 	void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
582 };
583 
584 struct kvm_assigned_dev_kernel {
585 	struct kvm_irq_ack_notifier ack_notifier;
586 	struct list_head list;
587 	int assigned_dev_id;
588 	int host_segnr;
589 	int host_busnr;
590 	int host_devfn;
591 	unsigned int entries_nr;
592 	int host_irq;
593 	bool host_irq_disabled;
594 	bool pci_2_3;
595 	struct msix_entry *host_msix_entries;
596 	int guest_irq;
597 	struct msix_entry *guest_msix_entries;
598 	unsigned long irq_requested_type;
599 	int irq_source_id;
600 	int flags;
601 	struct pci_dev *dev;
602 	struct kvm *kvm;
603 	spinlock_t intx_lock;
604 	spinlock_t intx_mask_lock;
605 	char irq_name[32];
606 	struct pci_saved_state *pci_saved_state;
607 };
608 
609 struct kvm_irq_mask_notifier {
610 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
611 	int irq;
612 	struct hlist_node link;
613 };
614 
615 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
616 				    struct kvm_irq_mask_notifier *kimn);
617 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
618 				      struct kvm_irq_mask_notifier *kimn);
619 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
620 			     bool mask);
621 
622 #ifdef __KVM_HAVE_IOAPIC
623 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
624 				   union kvm_ioapic_redirect_entry *entry,
625 				   unsigned long *deliver_bitmask);
626 #endif
627 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
628 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
629 		int irq_source_id, int level);
630 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
631 void kvm_register_irq_ack_notifier(struct kvm *kvm,
632 				   struct kvm_irq_ack_notifier *kian);
633 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
634 				   struct kvm_irq_ack_notifier *kian);
635 int kvm_request_irq_source_id(struct kvm *kvm);
636 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
637 
638 /* For vcpu->arch.iommu_flags */
639 #define KVM_IOMMU_CACHE_COHERENCY	0x1
640 
641 #ifdef CONFIG_IOMMU_API
642 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
643 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
644 int kvm_iommu_map_guest(struct kvm *kvm);
645 int kvm_iommu_unmap_guest(struct kvm *kvm);
646 int kvm_assign_device(struct kvm *kvm,
647 		      struct kvm_assigned_dev_kernel *assigned_dev);
648 int kvm_deassign_device(struct kvm *kvm,
649 			struct kvm_assigned_dev_kernel *assigned_dev);
650 #else /* CONFIG_IOMMU_API */
651 static inline int kvm_iommu_map_pages(struct kvm *kvm,
652 				      struct kvm_memory_slot *slot)
653 {
654 	return 0;
655 }
656 
657 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
658 					 struct kvm_memory_slot *slot)
659 {
660 }
661 
662 static inline int kvm_iommu_map_guest(struct kvm *kvm)
663 {
664 	return -ENODEV;
665 }
666 
667 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
668 {
669 	return 0;
670 }
671 
672 static inline int kvm_assign_device(struct kvm *kvm,
673 		struct kvm_assigned_dev_kernel *assigned_dev)
674 {
675 	return 0;
676 }
677 
678 static inline int kvm_deassign_device(struct kvm *kvm,
679 		struct kvm_assigned_dev_kernel *assigned_dev)
680 {
681 	return 0;
682 }
683 #endif /* CONFIG_IOMMU_API */
684 
685 static inline void kvm_guest_enter(void)
686 {
687 	BUG_ON(preemptible());
688 	account_system_vtime(current);
689 	current->flags |= PF_VCPU;
690 	/* KVM does not hold any references to rcu protected data when it
691 	 * switches CPU into a guest mode. In fact switching to a guest mode
692 	 * is very similar to exiting to userspase from rcu point of view. In
693 	 * addition CPU may stay in a guest mode for quite a long time (up to
694 	 * one time slice). Lets treat guest mode as quiescent state, just like
695 	 * we do with user-mode execution.
696 	 */
697 	rcu_virt_note_context_switch(smp_processor_id());
698 }
699 
700 static inline void kvm_guest_exit(void)
701 {
702 	account_system_vtime(current);
703 	current->flags &= ~PF_VCPU;
704 }
705 
706 /*
707  * search_memslots() and __gfn_to_memslot() are here because they are
708  * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
709  * gfn_to_memslot() itself isn't here as an inline because that would
710  * bloat other code too much.
711  */
712 static inline struct kvm_memory_slot *
713 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
714 {
715 	struct kvm_memory_slot *memslot;
716 
717 	kvm_for_each_memslot(memslot, slots)
718 		if (gfn >= memslot->base_gfn &&
719 		      gfn < memslot->base_gfn + memslot->npages)
720 			return memslot;
721 
722 	return NULL;
723 }
724 
725 static inline struct kvm_memory_slot *
726 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
727 {
728 	return search_memslots(slots, gfn);
729 }
730 
731 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
732 {
733 	return gfn_to_memslot(kvm, gfn)->id;
734 }
735 
736 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
737 {
738 	/* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
739 	return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
740 		(base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
741 }
742 
743 static inline unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot,
744 					       gfn_t gfn)
745 {
746 	return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
747 }
748 
749 static inline gpa_t gfn_to_gpa(gfn_t gfn)
750 {
751 	return (gpa_t)gfn << PAGE_SHIFT;
752 }
753 
754 static inline gfn_t gpa_to_gfn(gpa_t gpa)
755 {
756 	return (gfn_t)(gpa >> PAGE_SHIFT);
757 }
758 
759 static inline hpa_t pfn_to_hpa(pfn_t pfn)
760 {
761 	return (hpa_t)pfn << PAGE_SHIFT;
762 }
763 
764 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
765 {
766 	set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
767 }
768 
769 enum kvm_stat_kind {
770 	KVM_STAT_VM,
771 	KVM_STAT_VCPU,
772 };
773 
774 struct kvm_stats_debugfs_item {
775 	const char *name;
776 	int offset;
777 	enum kvm_stat_kind kind;
778 	struct dentry *dentry;
779 };
780 extern struct kvm_stats_debugfs_item debugfs_entries[];
781 extern struct dentry *kvm_debugfs_dir;
782 
783 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
784 static inline int mmu_notifier_retry(struct kvm_vcpu *vcpu, unsigned long mmu_seq)
785 {
786 	if (unlikely(vcpu->kvm->mmu_notifier_count))
787 		return 1;
788 	/*
789 	 * Ensure the read of mmu_notifier_count happens before the read
790 	 * of mmu_notifier_seq.  This interacts with the smp_wmb() in
791 	 * mmu_notifier_invalidate_range_end to make sure that the caller
792 	 * either sees the old (non-zero) value of mmu_notifier_count or
793 	 * the new (incremented) value of mmu_notifier_seq.
794 	 * PowerPC Book3s HV KVM calls this under a per-page lock
795 	 * rather than under kvm->mmu_lock, for scalability, so
796 	 * can't rely on kvm->mmu_lock to keep things ordered.
797 	 */
798 	smp_rmb();
799 	if (vcpu->kvm->mmu_notifier_seq != mmu_seq)
800 		return 1;
801 	return 0;
802 }
803 #endif
804 
805 #ifdef KVM_CAP_IRQ_ROUTING
806 
807 #define KVM_MAX_IRQ_ROUTES 1024
808 
809 int kvm_setup_default_irq_routing(struct kvm *kvm);
810 int kvm_set_irq_routing(struct kvm *kvm,
811 			const struct kvm_irq_routing_entry *entries,
812 			unsigned nr,
813 			unsigned flags);
814 void kvm_free_irq_routing(struct kvm *kvm);
815 
816 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
817 
818 #else
819 
820 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
821 
822 #endif
823 
824 #ifdef CONFIG_HAVE_KVM_EVENTFD
825 
826 void kvm_eventfd_init(struct kvm *kvm);
827 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
828 void kvm_irqfd_release(struct kvm *kvm);
829 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
830 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
831 
832 #else
833 
834 static inline void kvm_eventfd_init(struct kvm *kvm) {}
835 
836 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
837 {
838 	return -EINVAL;
839 }
840 
841 static inline void kvm_irqfd_release(struct kvm *kvm) {}
842 
843 #ifdef CONFIG_HAVE_KVM_IRQCHIP
844 static inline void kvm_irq_routing_update(struct kvm *kvm,
845 					  struct kvm_irq_routing_table *irq_rt)
846 {
847 	rcu_assign_pointer(kvm->irq_routing, irq_rt);
848 }
849 #endif
850 
851 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
852 {
853 	return -ENOSYS;
854 }
855 
856 #endif /* CONFIG_HAVE_KVM_EVENTFD */
857 
858 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
859 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
860 {
861 	return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
862 }
863 
864 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
865 
866 #else
867 
868 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
869 
870 #endif
871 
872 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
873 
874 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
875 				  unsigned long arg);
876 
877 #else
878 
879 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
880 						unsigned long arg)
881 {
882 	return -ENOTTY;
883 }
884 
885 #endif
886 
887 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
888 {
889 	set_bit(req, &vcpu->requests);
890 }
891 
892 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
893 {
894 	if (test_bit(req, &vcpu->requests)) {
895 		clear_bit(req, &vcpu->requests);
896 		return true;
897 	} else {
898 		return false;
899 	}
900 }
901 
902 #endif
903 
904