xref: /linux-6.15/include/linux/kvm_host.h (revision 6fa79bca)
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 <linux/err.h>
25 #include <asm/signal.h>
26 
27 #include <linux/kvm.h>
28 #include <linux/kvm_para.h>
29 
30 #include <linux/kvm_types.h>
31 
32 #include <asm/kvm_host.h>
33 
34 #ifndef KVM_MMIO_SIZE
35 #define KVM_MMIO_SIZE 8
36 #endif
37 
38 /*
39  * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
40  * in kvm, other bits are visible for userspace which are defined in
41  * include/linux/kvm_h.
42  */
43 #define KVM_MEMSLOT_INVALID	(1UL << 16)
44 
45 /* Two fragments for cross MMIO pages. */
46 #define KVM_MAX_MMIO_FRAGMENTS	2
47 
48 /*
49  * For the normal pfn, the highest 12 bits should be zero,
50  * so we can mask these bits to indicate the error.
51  */
52 #define KVM_PFN_ERR_MASK	(0xfffULL << 52)
53 
54 #define KVM_PFN_ERR_FAULT	(KVM_PFN_ERR_MASK)
55 #define KVM_PFN_ERR_HWPOISON	(KVM_PFN_ERR_MASK + 1)
56 #define KVM_PFN_ERR_BAD		(KVM_PFN_ERR_MASK + 2)
57 #define KVM_PFN_ERR_RO_FAULT	(KVM_PFN_ERR_MASK + 3)
58 
59 static inline bool is_error_pfn(pfn_t pfn)
60 {
61 	return !!(pfn & KVM_PFN_ERR_MASK);
62 }
63 
64 static inline bool is_noslot_pfn(pfn_t pfn)
65 {
66 	return pfn == KVM_PFN_ERR_BAD;
67 }
68 
69 static inline bool is_invalid_pfn(pfn_t pfn)
70 {
71 	return !is_noslot_pfn(pfn) && is_error_pfn(pfn);
72 }
73 
74 #define KVM_HVA_ERR_BAD		(PAGE_OFFSET)
75 #define KVM_HVA_ERR_RO_BAD	(PAGE_OFFSET + PAGE_SIZE)
76 
77 static inline bool kvm_is_error_hva(unsigned long addr)
78 {
79 	return addr >= PAGE_OFFSET;
80 }
81 
82 #define KVM_ERR_PTR_BAD_PAGE	(ERR_PTR(-ENOENT))
83 
84 static inline bool is_error_page(struct page *page)
85 {
86 	return IS_ERR(page);
87 }
88 
89 /*
90  * vcpu->requests bit members
91  */
92 #define KVM_REQ_TLB_FLUSH          0
93 #define KVM_REQ_MIGRATE_TIMER      1
94 #define KVM_REQ_REPORT_TPR_ACCESS  2
95 #define KVM_REQ_MMU_RELOAD         3
96 #define KVM_REQ_TRIPLE_FAULT       4
97 #define KVM_REQ_PENDING_TIMER      5
98 #define KVM_REQ_UNHALT             6
99 #define KVM_REQ_MMU_SYNC           7
100 #define KVM_REQ_CLOCK_UPDATE       8
101 #define KVM_REQ_KICK               9
102 #define KVM_REQ_DEACTIVATE_FPU    10
103 #define KVM_REQ_EVENT             11
104 #define KVM_REQ_APF_HALT          12
105 #define KVM_REQ_STEAL_UPDATE      13
106 #define KVM_REQ_NMI               14
107 #define KVM_REQ_IMMEDIATE_EXIT    15
108 #define KVM_REQ_PMU               16
109 #define KVM_REQ_PMI               17
110 
111 #define KVM_USERSPACE_IRQ_SOURCE_ID		0
112 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID	1
113 
114 struct kvm;
115 struct kvm_vcpu;
116 extern struct kmem_cache *kvm_vcpu_cache;
117 
118 struct kvm_io_range {
119 	gpa_t addr;
120 	int len;
121 	struct kvm_io_device *dev;
122 };
123 
124 #define NR_IOBUS_DEVS 1000
125 
126 struct kvm_io_bus {
127 	int                   dev_count;
128 	struct kvm_io_range range[];
129 };
130 
131 enum kvm_bus {
132 	KVM_MMIO_BUS,
133 	KVM_PIO_BUS,
134 	KVM_NR_BUSES
135 };
136 
137 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
138 		     int len, const void *val);
139 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
140 		    void *val);
141 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
142 			    int len, struct kvm_io_device *dev);
143 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
144 			      struct kvm_io_device *dev);
145 
146 #ifdef CONFIG_KVM_ASYNC_PF
147 struct kvm_async_pf {
148 	struct work_struct work;
149 	struct list_head link;
150 	struct list_head queue;
151 	struct kvm_vcpu *vcpu;
152 	struct mm_struct *mm;
153 	gva_t gva;
154 	unsigned long addr;
155 	struct kvm_arch_async_pf arch;
156 	struct page *page;
157 	bool done;
158 };
159 
160 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
161 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
162 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
163 		       struct kvm_arch_async_pf *arch);
164 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
165 #endif
166 
167 enum {
168 	OUTSIDE_GUEST_MODE,
169 	IN_GUEST_MODE,
170 	EXITING_GUEST_MODE,
171 	READING_SHADOW_PAGE_TABLES,
172 };
173 
174 /*
175  * Sometimes a large or cross-page mmio needs to be broken up into separate
176  * exits for userspace servicing.
177  */
178 struct kvm_mmio_fragment {
179 	gpa_t gpa;
180 	void *data;
181 	unsigned len;
182 };
183 
184 struct kvm_vcpu {
185 	struct kvm *kvm;
186 #ifdef CONFIG_PREEMPT_NOTIFIERS
187 	struct preempt_notifier preempt_notifier;
188 #endif
189 	int cpu;
190 	int vcpu_id;
191 	int srcu_idx;
192 	int mode;
193 	unsigned long requests;
194 	unsigned long guest_debug;
195 
196 	struct mutex mutex;
197 	struct kvm_run *run;
198 
199 	int fpu_active;
200 	int guest_fpu_loaded, guest_xcr0_loaded;
201 	wait_queue_head_t wq;
202 	struct pid *pid;
203 	int sigset_active;
204 	sigset_t sigset;
205 	struct kvm_vcpu_stat stat;
206 
207 #ifdef CONFIG_HAS_IOMEM
208 	int mmio_needed;
209 	int mmio_read_completed;
210 	int mmio_is_write;
211 	int mmio_cur_fragment;
212 	int mmio_nr_fragments;
213 	struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
214 #endif
215 
216 #ifdef CONFIG_KVM_ASYNC_PF
217 	struct {
218 		u32 queued;
219 		struct list_head queue;
220 		struct list_head done;
221 		spinlock_t lock;
222 	} async_pf;
223 #endif
224 
225 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
226 	/*
227 	 * Cpu relax intercept or pause loop exit optimization
228 	 * in_spin_loop: set when a vcpu does a pause loop exit
229 	 *  or cpu relax intercepted.
230 	 * dy_eligible: indicates whether vcpu is eligible for directed yield.
231 	 */
232 	struct {
233 		bool in_spin_loop;
234 		bool dy_eligible;
235 	} spin_loop;
236 #endif
237 	struct kvm_vcpu_arch arch;
238 };
239 
240 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
241 {
242 	return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
243 }
244 
245 /*
246  * Some of the bitops functions do not support too long bitmaps.
247  * This number must be determined not to exceed such limits.
248  */
249 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
250 
251 struct kvm_memory_slot {
252 	gfn_t base_gfn;
253 	unsigned long npages;
254 	unsigned long flags;
255 	unsigned long *dirty_bitmap;
256 	struct kvm_arch_memory_slot arch;
257 	unsigned long userspace_addr;
258 	int user_alloc;
259 	int id;
260 };
261 
262 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
263 {
264 	return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
265 }
266 
267 struct kvm_kernel_irq_routing_entry {
268 	u32 gsi;
269 	u32 type;
270 	int (*set)(struct kvm_kernel_irq_routing_entry *e,
271 		   struct kvm *kvm, int irq_source_id, int level);
272 	union {
273 		struct {
274 			unsigned irqchip;
275 			unsigned pin;
276 		} irqchip;
277 		struct msi_msg msi;
278 	};
279 	struct hlist_node link;
280 };
281 
282 #ifdef __KVM_HAVE_IOAPIC
283 
284 struct kvm_irq_routing_table {
285 	int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
286 	struct kvm_kernel_irq_routing_entry *rt_entries;
287 	u32 nr_rt_entries;
288 	/*
289 	 * Array indexed by gsi. Each entry contains list of irq chips
290 	 * the gsi is connected to.
291 	 */
292 	struct hlist_head map[0];
293 };
294 
295 #else
296 
297 struct kvm_irq_routing_table {};
298 
299 #endif
300 
301 #ifndef KVM_MEM_SLOTS_NUM
302 #define KVM_MEM_SLOTS_NUM (KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
303 #endif
304 
305 /*
306  * Note:
307  * memslots are not sorted by id anymore, please use id_to_memslot()
308  * to get the memslot by its id.
309  */
310 struct kvm_memslots {
311 	u64 generation;
312 	struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
313 	/* The mapping table from slot id to the index in memslots[]. */
314 	int id_to_index[KVM_MEM_SLOTS_NUM];
315 };
316 
317 struct kvm {
318 	spinlock_t mmu_lock;
319 	struct mutex slots_lock;
320 	struct mm_struct *mm; /* userspace tied to this vm */
321 	struct kvm_memslots *memslots;
322 	struct srcu_struct srcu;
323 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
324 	u32 bsp_vcpu_id;
325 #endif
326 	struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
327 	atomic_t online_vcpus;
328 	int last_boosted_vcpu;
329 	struct list_head vm_list;
330 	struct mutex lock;
331 	struct kvm_io_bus *buses[KVM_NR_BUSES];
332 #ifdef CONFIG_HAVE_KVM_EVENTFD
333 	struct {
334 		spinlock_t        lock;
335 		struct list_head  items;
336 		struct list_head  resampler_list;
337 		struct mutex      resampler_lock;
338 	} irqfds;
339 	struct list_head ioeventfds;
340 #endif
341 	struct kvm_vm_stat stat;
342 	struct kvm_arch arch;
343 	atomic_t users_count;
344 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
345 	struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
346 	spinlock_t ring_lock;
347 	struct list_head coalesced_zones;
348 #endif
349 
350 	struct mutex irq_lock;
351 #ifdef CONFIG_HAVE_KVM_IRQCHIP
352 	/*
353 	 * Update side is protected by irq_lock and,
354 	 * if configured, irqfds.lock.
355 	 */
356 	struct kvm_irq_routing_table __rcu *irq_routing;
357 	struct hlist_head mask_notifier_list;
358 	struct hlist_head irq_ack_notifier_list;
359 #endif
360 
361 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
362 	struct mmu_notifier mmu_notifier;
363 	unsigned long mmu_notifier_seq;
364 	long mmu_notifier_count;
365 #endif
366 	long tlbs_dirty;
367 };
368 
369 #define kvm_err(fmt, ...) \
370 	pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
371 #define kvm_info(fmt, ...) \
372 	pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
373 #define kvm_debug(fmt, ...) \
374 	pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
375 #define kvm_pr_unimpl(fmt, ...) \
376 	pr_err_ratelimited("kvm [%i]: " fmt, \
377 			   task_tgid_nr(current), ## __VA_ARGS__)
378 
379 /* The guest did something we don't support. */
380 #define vcpu_unimpl(vcpu, fmt, ...)					\
381 	kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
382 
383 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
384 {
385 	smp_rmb();
386 	return kvm->vcpus[i];
387 }
388 
389 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
390 	for (idx = 0; \
391 	     idx < atomic_read(&kvm->online_vcpus) && \
392 	     (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
393 	     idx++)
394 
395 #define kvm_for_each_memslot(memslot, slots)	\
396 	for (memslot = &slots->memslots[0];	\
397 	      memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
398 		memslot++)
399 
400 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
401 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
402 
403 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
404 void vcpu_put(struct kvm_vcpu *vcpu);
405 
406 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
407 		  struct module *module);
408 void kvm_exit(void);
409 
410 void kvm_get_kvm(struct kvm *kvm);
411 void kvm_put_kvm(struct kvm *kvm);
412 void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new);
413 
414 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
415 {
416 	return rcu_dereference_check(kvm->memslots,
417 			srcu_read_lock_held(&kvm->srcu)
418 			|| lockdep_is_held(&kvm->slots_lock));
419 }
420 
421 static inline struct kvm_memory_slot *
422 id_to_memslot(struct kvm_memslots *slots, int id)
423 {
424 	int index = slots->id_to_index[id];
425 	struct kvm_memory_slot *slot;
426 
427 	slot = &slots->memslots[index];
428 
429 	WARN_ON(slot->id != id);
430 	return slot;
431 }
432 
433 int kvm_set_memory_region(struct kvm *kvm,
434 			  struct kvm_userspace_memory_region *mem,
435 			  int user_alloc);
436 int __kvm_set_memory_region(struct kvm *kvm,
437 			    struct kvm_userspace_memory_region *mem,
438 			    int user_alloc);
439 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
440 			   struct kvm_memory_slot *dont);
441 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages);
442 int kvm_arch_prepare_memory_region(struct kvm *kvm,
443 				struct kvm_memory_slot *memslot,
444 				struct kvm_memory_slot old,
445 				struct kvm_userspace_memory_region *mem,
446 				int user_alloc);
447 void kvm_arch_commit_memory_region(struct kvm *kvm,
448 				struct kvm_userspace_memory_region *mem,
449 				struct kvm_memory_slot old,
450 				int user_alloc);
451 bool kvm_largepages_enabled(void);
452 void kvm_disable_largepages(void);
453 /* flush all memory translations */
454 void kvm_arch_flush_shadow_all(struct kvm *kvm);
455 /* flush memory translations pointing to 'slot' */
456 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
457 				   struct kvm_memory_slot *slot);
458 
459 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
460 			    int nr_pages);
461 
462 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
463 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
464 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
465 void kvm_release_page_clean(struct page *page);
466 void kvm_release_page_dirty(struct page *page);
467 void kvm_set_page_dirty(struct page *page);
468 void kvm_set_page_accessed(struct page *page);
469 
470 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
471 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
472 		       bool write_fault, bool *writable);
473 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
474 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
475 		      bool *writable);
476 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
477 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
478 
479 void kvm_release_pfn_dirty(pfn_t pfn);
480 void kvm_release_pfn_clean(pfn_t pfn);
481 void kvm_set_pfn_dirty(pfn_t pfn);
482 void kvm_set_pfn_accessed(pfn_t pfn);
483 void kvm_get_pfn(pfn_t pfn);
484 
485 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
486 			int len);
487 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
488 			  unsigned long len);
489 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
490 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
491 			   void *data, unsigned long len);
492 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
493 			 int offset, int len);
494 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
495 		    unsigned long len);
496 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
497 			   void *data, unsigned long len);
498 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
499 			      gpa_t gpa);
500 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
501 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
502 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
503 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
504 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
505 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
506 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
507 			     gfn_t gfn);
508 
509 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
510 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
511 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
512 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
513 void kvm_resched(struct kvm_vcpu *vcpu);
514 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
515 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
516 
517 void kvm_flush_remote_tlbs(struct kvm *kvm);
518 void kvm_reload_remote_mmus(struct kvm *kvm);
519 
520 long kvm_arch_dev_ioctl(struct file *filp,
521 			unsigned int ioctl, unsigned long arg);
522 long kvm_arch_vcpu_ioctl(struct file *filp,
523 			 unsigned int ioctl, unsigned long arg);
524 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
525 
526 int kvm_dev_ioctl_check_extension(long ext);
527 
528 int kvm_get_dirty_log(struct kvm *kvm,
529 			struct kvm_dirty_log *log, int *is_dirty);
530 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
531 				struct kvm_dirty_log *log);
532 
533 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
534 				   struct
535 				   kvm_userspace_memory_region *mem,
536 				   int user_alloc);
537 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level);
538 long kvm_arch_vm_ioctl(struct file *filp,
539 		       unsigned int ioctl, unsigned long arg);
540 
541 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
542 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
543 
544 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
545 				    struct kvm_translation *tr);
546 
547 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
548 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
549 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
550 				  struct kvm_sregs *sregs);
551 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
552 				  struct kvm_sregs *sregs);
553 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
554 				    struct kvm_mp_state *mp_state);
555 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
556 				    struct kvm_mp_state *mp_state);
557 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
558 					struct kvm_guest_debug *dbg);
559 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
560 
561 int kvm_arch_init(void *opaque);
562 void kvm_arch_exit(void);
563 
564 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
565 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
566 
567 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
568 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
569 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
570 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
571 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
572 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
573 
574 int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu);
575 int kvm_arch_hardware_enable(void *garbage);
576 void kvm_arch_hardware_disable(void *garbage);
577 int kvm_arch_hardware_setup(void);
578 void kvm_arch_hardware_unsetup(void);
579 void kvm_arch_check_processor_compat(void *rtn);
580 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
581 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
582 
583 void kvm_free_physmem(struct kvm *kvm);
584 
585 void *kvm_kvzalloc(unsigned long size);
586 void kvm_kvfree(const void *addr);
587 
588 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
589 static inline struct kvm *kvm_arch_alloc_vm(void)
590 {
591 	return kzalloc(sizeof(struct kvm), GFP_KERNEL);
592 }
593 
594 static inline void kvm_arch_free_vm(struct kvm *kvm)
595 {
596 	kfree(kvm);
597 }
598 #endif
599 
600 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
601 {
602 #ifdef __KVM_HAVE_ARCH_WQP
603 	return vcpu->arch.wqp;
604 #else
605 	return &vcpu->wq;
606 #endif
607 }
608 
609 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
610 void kvm_arch_destroy_vm(struct kvm *kvm);
611 void kvm_free_all_assigned_devices(struct kvm *kvm);
612 void kvm_arch_sync_events(struct kvm *kvm);
613 
614 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
615 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
616 
617 bool kvm_is_mmio_pfn(pfn_t pfn);
618 
619 struct kvm_irq_ack_notifier {
620 	struct hlist_node link;
621 	unsigned gsi;
622 	void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
623 };
624 
625 struct kvm_assigned_dev_kernel {
626 	struct kvm_irq_ack_notifier ack_notifier;
627 	struct list_head list;
628 	int assigned_dev_id;
629 	int host_segnr;
630 	int host_busnr;
631 	int host_devfn;
632 	unsigned int entries_nr;
633 	int host_irq;
634 	bool host_irq_disabled;
635 	bool pci_2_3;
636 	struct msix_entry *host_msix_entries;
637 	int guest_irq;
638 	struct msix_entry *guest_msix_entries;
639 	unsigned long irq_requested_type;
640 	int irq_source_id;
641 	int flags;
642 	struct pci_dev *dev;
643 	struct kvm *kvm;
644 	spinlock_t intx_lock;
645 	spinlock_t intx_mask_lock;
646 	char irq_name[32];
647 	struct pci_saved_state *pci_saved_state;
648 };
649 
650 struct kvm_irq_mask_notifier {
651 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
652 	int irq;
653 	struct hlist_node link;
654 };
655 
656 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
657 				    struct kvm_irq_mask_notifier *kimn);
658 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
659 				      struct kvm_irq_mask_notifier *kimn);
660 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
661 			     bool mask);
662 
663 #ifdef __KVM_HAVE_IOAPIC
664 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
665 				   union kvm_ioapic_redirect_entry *entry,
666 				   unsigned long *deliver_bitmask);
667 #endif
668 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
669 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
670 		int irq_source_id, int level);
671 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
672 void kvm_register_irq_ack_notifier(struct kvm *kvm,
673 				   struct kvm_irq_ack_notifier *kian);
674 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
675 				   struct kvm_irq_ack_notifier *kian);
676 int kvm_request_irq_source_id(struct kvm *kvm);
677 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
678 
679 /* For vcpu->arch.iommu_flags */
680 #define KVM_IOMMU_CACHE_COHERENCY	0x1
681 
682 #ifdef CONFIG_IOMMU_API
683 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
684 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
685 int kvm_iommu_map_guest(struct kvm *kvm);
686 int kvm_iommu_unmap_guest(struct kvm *kvm);
687 int kvm_assign_device(struct kvm *kvm,
688 		      struct kvm_assigned_dev_kernel *assigned_dev);
689 int kvm_deassign_device(struct kvm *kvm,
690 			struct kvm_assigned_dev_kernel *assigned_dev);
691 #else /* CONFIG_IOMMU_API */
692 static inline int kvm_iommu_map_pages(struct kvm *kvm,
693 				      struct kvm_memory_slot *slot)
694 {
695 	return 0;
696 }
697 
698 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
699 					 struct kvm_memory_slot *slot)
700 {
701 }
702 
703 static inline int kvm_iommu_map_guest(struct kvm *kvm)
704 {
705 	return -ENODEV;
706 }
707 
708 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
709 {
710 	return 0;
711 }
712 
713 static inline int kvm_assign_device(struct kvm *kvm,
714 		struct kvm_assigned_dev_kernel *assigned_dev)
715 {
716 	return 0;
717 }
718 
719 static inline int kvm_deassign_device(struct kvm *kvm,
720 		struct kvm_assigned_dev_kernel *assigned_dev)
721 {
722 	return 0;
723 }
724 #endif /* CONFIG_IOMMU_API */
725 
726 static inline void kvm_guest_enter(void)
727 {
728 	BUG_ON(preemptible());
729 	vtime_account(current);
730 	current->flags |= PF_VCPU;
731 	/* KVM does not hold any references to rcu protected data when it
732 	 * switches CPU into a guest mode. In fact switching to a guest mode
733 	 * is very similar to exiting to userspase from rcu point of view. In
734 	 * addition CPU may stay in a guest mode for quite a long time (up to
735 	 * one time slice). Lets treat guest mode as quiescent state, just like
736 	 * we do with user-mode execution.
737 	 */
738 	rcu_virt_note_context_switch(smp_processor_id());
739 }
740 
741 static inline void kvm_guest_exit(void)
742 {
743 	vtime_account(current);
744 	current->flags &= ~PF_VCPU;
745 }
746 
747 /*
748  * search_memslots() and __gfn_to_memslot() are here because they are
749  * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
750  * gfn_to_memslot() itself isn't here as an inline because that would
751  * bloat other code too much.
752  */
753 static inline struct kvm_memory_slot *
754 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
755 {
756 	struct kvm_memory_slot *memslot;
757 
758 	kvm_for_each_memslot(memslot, slots)
759 		if (gfn >= memslot->base_gfn &&
760 		      gfn < memslot->base_gfn + memslot->npages)
761 			return memslot;
762 
763 	return NULL;
764 }
765 
766 static inline struct kvm_memory_slot *
767 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
768 {
769 	return search_memslots(slots, gfn);
770 }
771 
772 static inline unsigned long
773 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
774 {
775 	return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
776 }
777 
778 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
779 {
780 	return gfn_to_memslot(kvm, gfn)->id;
781 }
782 
783 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
784 {
785 	/* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
786 	return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
787 		(base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
788 }
789 
790 static inline gfn_t
791 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
792 {
793 	gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
794 
795 	return slot->base_gfn + gfn_offset;
796 }
797 
798 static inline gpa_t gfn_to_gpa(gfn_t gfn)
799 {
800 	return (gpa_t)gfn << PAGE_SHIFT;
801 }
802 
803 static inline gfn_t gpa_to_gfn(gpa_t gpa)
804 {
805 	return (gfn_t)(gpa >> PAGE_SHIFT);
806 }
807 
808 static inline hpa_t pfn_to_hpa(pfn_t pfn)
809 {
810 	return (hpa_t)pfn << PAGE_SHIFT;
811 }
812 
813 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
814 {
815 	set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
816 }
817 
818 enum kvm_stat_kind {
819 	KVM_STAT_VM,
820 	KVM_STAT_VCPU,
821 };
822 
823 struct kvm_stats_debugfs_item {
824 	const char *name;
825 	int offset;
826 	enum kvm_stat_kind kind;
827 	struct dentry *dentry;
828 };
829 extern struct kvm_stats_debugfs_item debugfs_entries[];
830 extern struct dentry *kvm_debugfs_dir;
831 
832 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
833 static inline int mmu_notifier_retry(struct kvm_vcpu *vcpu, unsigned long mmu_seq)
834 {
835 	if (unlikely(vcpu->kvm->mmu_notifier_count))
836 		return 1;
837 	/*
838 	 * Ensure the read of mmu_notifier_count happens before the read
839 	 * of mmu_notifier_seq.  This interacts with the smp_wmb() in
840 	 * mmu_notifier_invalidate_range_end to make sure that the caller
841 	 * either sees the old (non-zero) value of mmu_notifier_count or
842 	 * the new (incremented) value of mmu_notifier_seq.
843 	 * PowerPC Book3s HV KVM calls this under a per-page lock
844 	 * rather than under kvm->mmu_lock, for scalability, so
845 	 * can't rely on kvm->mmu_lock to keep things ordered.
846 	 */
847 	smp_rmb();
848 	if (vcpu->kvm->mmu_notifier_seq != mmu_seq)
849 		return 1;
850 	return 0;
851 }
852 #endif
853 
854 #ifdef KVM_CAP_IRQ_ROUTING
855 
856 #define KVM_MAX_IRQ_ROUTES 1024
857 
858 int kvm_setup_default_irq_routing(struct kvm *kvm);
859 int kvm_set_irq_routing(struct kvm *kvm,
860 			const struct kvm_irq_routing_entry *entries,
861 			unsigned nr,
862 			unsigned flags);
863 void kvm_free_irq_routing(struct kvm *kvm);
864 
865 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
866 
867 #else
868 
869 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
870 
871 #endif
872 
873 #ifdef CONFIG_HAVE_KVM_EVENTFD
874 
875 void kvm_eventfd_init(struct kvm *kvm);
876 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
877 void kvm_irqfd_release(struct kvm *kvm);
878 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
879 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
880 
881 #else
882 
883 static inline void kvm_eventfd_init(struct kvm *kvm) {}
884 
885 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
886 {
887 	return -EINVAL;
888 }
889 
890 static inline void kvm_irqfd_release(struct kvm *kvm) {}
891 
892 #ifdef CONFIG_HAVE_KVM_IRQCHIP
893 static inline void kvm_irq_routing_update(struct kvm *kvm,
894 					  struct kvm_irq_routing_table *irq_rt)
895 {
896 	rcu_assign_pointer(kvm->irq_routing, irq_rt);
897 }
898 #endif
899 
900 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
901 {
902 	return -ENOSYS;
903 }
904 
905 #endif /* CONFIG_HAVE_KVM_EVENTFD */
906 
907 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
908 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
909 {
910 	return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
911 }
912 
913 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
914 
915 #else
916 
917 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
918 
919 #endif
920 
921 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
922 
923 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
924 				  unsigned long arg);
925 
926 #else
927 
928 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
929 						unsigned long arg)
930 {
931 	return -ENOTTY;
932 }
933 
934 #endif
935 
936 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
937 {
938 	set_bit(req, &vcpu->requests);
939 }
940 
941 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
942 {
943 	if (test_bit(req, &vcpu->requests)) {
944 		clear_bit(req, &vcpu->requests);
945 		return true;
946 	} else {
947 		return false;
948 	}
949 }
950 
951 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
952 
953 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
954 {
955 	vcpu->spin_loop.in_spin_loop = val;
956 }
957 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
958 {
959 	vcpu->spin_loop.dy_eligible = val;
960 }
961 
962 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
963 
964 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
965 {
966 }
967 
968 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
969 {
970 }
971 
972 static inline bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
973 {
974 	return true;
975 }
976 
977 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
978 #endif
979 
980