xref: /openbmc/linux/include/linux/kvm_host.h (revision 7051924f771722c6dd235e693742cda6488ac700)
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 <linux/irqflags.h>
26 #include <linux/context_tracking.h>
27 #include <asm/signal.h>
28 
29 #include <linux/kvm.h>
30 #include <linux/kvm_para.h>
31 
32 #include <linux/kvm_types.h>
33 
34 #include <asm/kvm_host.h>
35 
36 #ifndef KVM_MMIO_SIZE
37 #define KVM_MMIO_SIZE 8
38 #endif
39 
40 /*
41  * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
42  * in kvm, other bits are visible for userspace which are defined in
43  * include/linux/kvm_h.
44  */
45 #define KVM_MEMSLOT_INVALID	(1UL << 16)
46 
47 /* Two fragments for cross MMIO pages. */
48 #define KVM_MAX_MMIO_FRAGMENTS	2
49 
50 /*
51  * For the normal pfn, the highest 12 bits should be zero,
52  * so we can mask bit 62 ~ bit 52  to indicate the error pfn,
53  * mask bit 63 to indicate the noslot pfn.
54  */
55 #define KVM_PFN_ERR_MASK	(0x7ffULL << 52)
56 #define KVM_PFN_ERR_NOSLOT_MASK	(0xfffULL << 52)
57 #define KVM_PFN_NOSLOT		(0x1ULL << 63)
58 
59 #define KVM_PFN_ERR_FAULT	(KVM_PFN_ERR_MASK)
60 #define KVM_PFN_ERR_HWPOISON	(KVM_PFN_ERR_MASK + 1)
61 #define KVM_PFN_ERR_RO_FAULT	(KVM_PFN_ERR_MASK + 2)
62 
63 /*
64  * error pfns indicate that the gfn is in slot but faild to
65  * translate it to pfn on host.
66  */
67 static inline bool is_error_pfn(pfn_t pfn)
68 {
69 	return !!(pfn & KVM_PFN_ERR_MASK);
70 }
71 
72 /*
73  * error_noslot pfns indicate that the gfn can not be
74  * translated to pfn - it is not in slot or failed to
75  * translate it to pfn.
76  */
77 static inline bool is_error_noslot_pfn(pfn_t pfn)
78 {
79 	return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
80 }
81 
82 /* noslot pfn indicates that the gfn is not in slot. */
83 static inline bool is_noslot_pfn(pfn_t pfn)
84 {
85 	return pfn == KVM_PFN_NOSLOT;
86 }
87 
88 /*
89  * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
90  * provide own defines and kvm_is_error_hva
91  */
92 #ifndef KVM_HVA_ERR_BAD
93 
94 #define KVM_HVA_ERR_BAD		(PAGE_OFFSET)
95 #define KVM_HVA_ERR_RO_BAD	(PAGE_OFFSET + PAGE_SIZE)
96 
97 static inline bool kvm_is_error_hva(unsigned long addr)
98 {
99 	return addr >= PAGE_OFFSET;
100 }
101 
102 #endif
103 
104 #define KVM_ERR_PTR_BAD_PAGE	(ERR_PTR(-ENOENT))
105 
106 static inline bool is_error_page(struct page *page)
107 {
108 	return IS_ERR(page);
109 }
110 
111 /*
112  * vcpu->requests bit members
113  */
114 #define KVM_REQ_TLB_FLUSH          0
115 #define KVM_REQ_MIGRATE_TIMER      1
116 #define KVM_REQ_REPORT_TPR_ACCESS  2
117 #define KVM_REQ_MMU_RELOAD         3
118 #define KVM_REQ_TRIPLE_FAULT       4
119 #define KVM_REQ_PENDING_TIMER      5
120 #define KVM_REQ_UNHALT             6
121 #define KVM_REQ_MMU_SYNC           7
122 #define KVM_REQ_CLOCK_UPDATE       8
123 #define KVM_REQ_KICK               9
124 #define KVM_REQ_DEACTIVATE_FPU    10
125 #define KVM_REQ_EVENT             11
126 #define KVM_REQ_APF_HALT          12
127 #define KVM_REQ_STEAL_UPDATE      13
128 #define KVM_REQ_NMI               14
129 #define KVM_REQ_PMU               15
130 #define KVM_REQ_PMI               16
131 #define KVM_REQ_WATCHDOG          17
132 #define KVM_REQ_MASTERCLOCK_UPDATE 18
133 #define KVM_REQ_MCLOCK_INPROGRESS 19
134 #define KVM_REQ_EPR_EXIT          20
135 #define KVM_REQ_SCAN_IOAPIC       21
136 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 22
137 #define KVM_REQ_ENABLE_IBS        23
138 #define KVM_REQ_DISABLE_IBS       24
139 
140 #define KVM_USERSPACE_IRQ_SOURCE_ID		0
141 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID	1
142 
143 struct kvm;
144 struct kvm_vcpu;
145 extern struct kmem_cache *kvm_vcpu_cache;
146 
147 extern spinlock_t kvm_lock;
148 extern struct list_head vm_list;
149 
150 struct kvm_io_range {
151 	gpa_t addr;
152 	int len;
153 	struct kvm_io_device *dev;
154 };
155 
156 #define NR_IOBUS_DEVS 1000
157 
158 struct kvm_io_bus {
159 	int dev_count;
160 	int ioeventfd_count;
161 	struct kvm_io_range range[];
162 };
163 
164 enum kvm_bus {
165 	KVM_MMIO_BUS,
166 	KVM_PIO_BUS,
167 	KVM_VIRTIO_CCW_NOTIFY_BUS,
168 	KVM_FAST_MMIO_BUS,
169 	KVM_NR_BUSES
170 };
171 
172 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
173 		     int len, const void *val);
174 int kvm_io_bus_write_cookie(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
175 			    int len, const void *val, long cookie);
176 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
177 		    void *val);
178 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
179 			    int len, struct kvm_io_device *dev);
180 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
181 			      struct kvm_io_device *dev);
182 
183 #ifdef CONFIG_KVM_ASYNC_PF
184 struct kvm_async_pf {
185 	struct work_struct work;
186 	struct list_head link;
187 	struct list_head queue;
188 	struct kvm_vcpu *vcpu;
189 	struct mm_struct *mm;
190 	gva_t gva;
191 	unsigned long addr;
192 	struct kvm_arch_async_pf arch;
193 	bool   wakeup_all;
194 };
195 
196 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
197 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
198 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
199 		       struct kvm_arch_async_pf *arch);
200 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
201 #endif
202 
203 enum {
204 	OUTSIDE_GUEST_MODE,
205 	IN_GUEST_MODE,
206 	EXITING_GUEST_MODE,
207 	READING_SHADOW_PAGE_TABLES,
208 };
209 
210 /*
211  * Sometimes a large or cross-page mmio needs to be broken up into separate
212  * exits for userspace servicing.
213  */
214 struct kvm_mmio_fragment {
215 	gpa_t gpa;
216 	void *data;
217 	unsigned len;
218 };
219 
220 struct kvm_vcpu {
221 	struct kvm *kvm;
222 #ifdef CONFIG_PREEMPT_NOTIFIERS
223 	struct preempt_notifier preempt_notifier;
224 #endif
225 	int cpu;
226 	int vcpu_id;
227 	int srcu_idx;
228 	int mode;
229 	unsigned long requests;
230 	unsigned long guest_debug;
231 
232 	struct mutex mutex;
233 	struct kvm_run *run;
234 
235 	int fpu_active;
236 	int guest_fpu_loaded, guest_xcr0_loaded;
237 	wait_queue_head_t wq;
238 	struct pid *pid;
239 	int sigset_active;
240 	sigset_t sigset;
241 	struct kvm_vcpu_stat stat;
242 
243 #ifdef CONFIG_HAS_IOMEM
244 	int mmio_needed;
245 	int mmio_read_completed;
246 	int mmio_is_write;
247 	int mmio_cur_fragment;
248 	int mmio_nr_fragments;
249 	struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
250 #endif
251 
252 #ifdef CONFIG_KVM_ASYNC_PF
253 	struct {
254 		u32 queued;
255 		struct list_head queue;
256 		struct list_head done;
257 		spinlock_t lock;
258 	} async_pf;
259 #endif
260 
261 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
262 	/*
263 	 * Cpu relax intercept or pause loop exit optimization
264 	 * in_spin_loop: set when a vcpu does a pause loop exit
265 	 *  or cpu relax intercepted.
266 	 * dy_eligible: indicates whether vcpu is eligible for directed yield.
267 	 */
268 	struct {
269 		bool in_spin_loop;
270 		bool dy_eligible;
271 	} spin_loop;
272 #endif
273 	bool preempted;
274 	struct kvm_vcpu_arch arch;
275 };
276 
277 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
278 {
279 	return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
280 }
281 
282 /*
283  * Some of the bitops functions do not support too long bitmaps.
284  * This number must be determined not to exceed such limits.
285  */
286 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
287 
288 struct kvm_memory_slot {
289 	gfn_t base_gfn;
290 	unsigned long npages;
291 	unsigned long *dirty_bitmap;
292 	struct kvm_arch_memory_slot arch;
293 	unsigned long userspace_addr;
294 	u32 flags;
295 	short id;
296 };
297 
298 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
299 {
300 	return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
301 }
302 
303 struct kvm_s390_adapter_int {
304 	u64 ind_addr;
305 	u64 summary_addr;
306 	u64 ind_offset;
307 	u32 summary_offset;
308 	u32 adapter_id;
309 };
310 
311 struct kvm_kernel_irq_routing_entry {
312 	u32 gsi;
313 	u32 type;
314 	int (*set)(struct kvm_kernel_irq_routing_entry *e,
315 		   struct kvm *kvm, int irq_source_id, int level,
316 		   bool line_status);
317 	union {
318 		struct {
319 			unsigned irqchip;
320 			unsigned pin;
321 		} irqchip;
322 		struct msi_msg msi;
323 		struct kvm_s390_adapter_int adapter;
324 	};
325 	struct hlist_node link;
326 };
327 
328 struct kvm_irq_routing_table;
329 
330 #ifndef KVM_PRIVATE_MEM_SLOTS
331 #define KVM_PRIVATE_MEM_SLOTS 0
332 #endif
333 
334 #ifndef KVM_MEM_SLOTS_NUM
335 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
336 #endif
337 
338 /*
339  * Note:
340  * memslots are not sorted by id anymore, please use id_to_memslot()
341  * to get the memslot by its id.
342  */
343 struct kvm_memslots {
344 	u64 generation;
345 	struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
346 	/* The mapping table from slot id to the index in memslots[]. */
347 	short id_to_index[KVM_MEM_SLOTS_NUM];
348 };
349 
350 struct kvm {
351 	spinlock_t mmu_lock;
352 	struct mutex slots_lock;
353 	struct mm_struct *mm; /* userspace tied to this vm */
354 	struct kvm_memslots *memslots;
355 	struct srcu_struct srcu;
356 	struct srcu_struct irq_srcu;
357 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
358 	u32 bsp_vcpu_id;
359 #endif
360 	struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
361 	atomic_t online_vcpus;
362 	int last_boosted_vcpu;
363 	struct list_head vm_list;
364 	struct mutex lock;
365 	struct kvm_io_bus *buses[KVM_NR_BUSES];
366 #ifdef CONFIG_HAVE_KVM_EVENTFD
367 	struct {
368 		spinlock_t        lock;
369 		struct list_head  items;
370 		struct list_head  resampler_list;
371 		struct mutex      resampler_lock;
372 	} irqfds;
373 	struct list_head ioeventfds;
374 #endif
375 	struct kvm_vm_stat stat;
376 	struct kvm_arch arch;
377 	atomic_t users_count;
378 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
379 	struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
380 	spinlock_t ring_lock;
381 	struct list_head coalesced_zones;
382 #endif
383 
384 	struct mutex irq_lock;
385 #ifdef CONFIG_HAVE_KVM_IRQCHIP
386 	/*
387 	 * Update side is protected by irq_lock.
388 	 */
389 	struct kvm_irq_routing_table __rcu *irq_routing;
390 	struct hlist_head mask_notifier_list;
391 #endif
392 #ifdef CONFIG_HAVE_KVM_IRQFD
393 	struct hlist_head irq_ack_notifier_list;
394 #endif
395 
396 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
397 	struct mmu_notifier mmu_notifier;
398 	unsigned long mmu_notifier_seq;
399 	long mmu_notifier_count;
400 #endif
401 	long tlbs_dirty;
402 	struct list_head devices;
403 };
404 
405 #define kvm_err(fmt, ...) \
406 	pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
407 #define kvm_info(fmt, ...) \
408 	pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
409 #define kvm_debug(fmt, ...) \
410 	pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
411 #define kvm_pr_unimpl(fmt, ...) \
412 	pr_err_ratelimited("kvm [%i]: " fmt, \
413 			   task_tgid_nr(current), ## __VA_ARGS__)
414 
415 /* The guest did something we don't support. */
416 #define vcpu_unimpl(vcpu, fmt, ...)					\
417 	kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
418 
419 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
420 {
421 	smp_rmb();
422 	return kvm->vcpus[i];
423 }
424 
425 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
426 	for (idx = 0; \
427 	     idx < atomic_read(&kvm->online_vcpus) && \
428 	     (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
429 	     idx++)
430 
431 #define kvm_for_each_memslot(memslot, slots)	\
432 	for (memslot = &slots->memslots[0];	\
433 	      memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
434 		memslot++)
435 
436 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
437 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
438 
439 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
440 void vcpu_put(struct kvm_vcpu *vcpu);
441 
442 #ifdef CONFIG_HAVE_KVM_IRQFD
443 int kvm_irqfd_init(void);
444 void kvm_irqfd_exit(void);
445 #else
446 static inline int kvm_irqfd_init(void)
447 {
448 	return 0;
449 }
450 
451 static inline void kvm_irqfd_exit(void)
452 {
453 }
454 #endif
455 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
456 		  struct module *module);
457 void kvm_exit(void);
458 
459 void kvm_get_kvm(struct kvm *kvm);
460 void kvm_put_kvm(struct kvm *kvm);
461 
462 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
463 {
464 	return rcu_dereference_check(kvm->memslots,
465 			srcu_read_lock_held(&kvm->srcu)
466 			|| lockdep_is_held(&kvm->slots_lock));
467 }
468 
469 static inline struct kvm_memory_slot *
470 id_to_memslot(struct kvm_memslots *slots, int id)
471 {
472 	int index = slots->id_to_index[id];
473 	struct kvm_memory_slot *slot;
474 
475 	slot = &slots->memslots[index];
476 
477 	WARN_ON(slot->id != id);
478 	return slot;
479 }
480 
481 /*
482  * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
483  * - create a new memory slot
484  * - delete an existing memory slot
485  * - modify an existing memory slot
486  *   -- move it in the guest physical memory space
487  *   -- just change its flags
488  *
489  * Since flags can be changed by some of these operations, the following
490  * differentiation is the best we can do for __kvm_set_memory_region():
491  */
492 enum kvm_mr_change {
493 	KVM_MR_CREATE,
494 	KVM_MR_DELETE,
495 	KVM_MR_MOVE,
496 	KVM_MR_FLAGS_ONLY,
497 };
498 
499 int kvm_set_memory_region(struct kvm *kvm,
500 			  struct kvm_userspace_memory_region *mem);
501 int __kvm_set_memory_region(struct kvm *kvm,
502 			    struct kvm_userspace_memory_region *mem);
503 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
504 			   struct kvm_memory_slot *dont);
505 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
506 			    unsigned long npages);
507 void kvm_arch_memslots_updated(struct kvm *kvm);
508 int kvm_arch_prepare_memory_region(struct kvm *kvm,
509 				struct kvm_memory_slot *memslot,
510 				struct kvm_userspace_memory_region *mem,
511 				enum kvm_mr_change change);
512 void kvm_arch_commit_memory_region(struct kvm *kvm,
513 				struct kvm_userspace_memory_region *mem,
514 				const struct kvm_memory_slot *old,
515 				enum kvm_mr_change change);
516 bool kvm_largepages_enabled(void);
517 void kvm_disable_largepages(void);
518 /* flush all memory translations */
519 void kvm_arch_flush_shadow_all(struct kvm *kvm);
520 /* flush memory translations pointing to 'slot' */
521 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
522 				   struct kvm_memory_slot *slot);
523 
524 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
525 			    int nr_pages);
526 
527 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
528 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
529 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
530 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
531 void kvm_release_page_clean(struct page *page);
532 void kvm_release_page_dirty(struct page *page);
533 void kvm_set_page_accessed(struct page *page);
534 
535 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
536 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
537 		       bool write_fault, bool *writable);
538 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
539 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
540 		      bool *writable);
541 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
542 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
543 
544 void kvm_release_pfn_clean(pfn_t pfn);
545 void kvm_set_pfn_dirty(pfn_t pfn);
546 void kvm_set_pfn_accessed(pfn_t pfn);
547 void kvm_get_pfn(pfn_t pfn);
548 
549 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
550 			int len);
551 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
552 			  unsigned long len);
553 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
554 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
555 			   void *data, unsigned long len);
556 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
557 			 int offset, int len);
558 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
559 		    unsigned long len);
560 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
561 			   void *data, unsigned long len);
562 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
563 			      gpa_t gpa, unsigned long len);
564 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
565 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
566 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
567 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
568 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
569 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
570 
571 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
572 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
573 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
574 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
575 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
576 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
577 
578 void kvm_flush_remote_tlbs(struct kvm *kvm);
579 void kvm_reload_remote_mmus(struct kvm *kvm);
580 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
581 void kvm_make_scan_ioapic_request(struct kvm *kvm);
582 
583 long kvm_arch_dev_ioctl(struct file *filp,
584 			unsigned int ioctl, unsigned long arg);
585 long kvm_arch_vcpu_ioctl(struct file *filp,
586 			 unsigned int ioctl, unsigned long arg);
587 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
588 
589 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
590 
591 int kvm_get_dirty_log(struct kvm *kvm,
592 			struct kvm_dirty_log *log, int *is_dirty);
593 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
594 				struct kvm_dirty_log *log);
595 
596 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
597 			bool line_status);
598 long kvm_arch_vm_ioctl(struct file *filp,
599 		       unsigned int ioctl, unsigned long arg);
600 
601 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
602 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
603 
604 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
605 				    struct kvm_translation *tr);
606 
607 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
608 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
609 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
610 				  struct kvm_sregs *sregs);
611 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
612 				  struct kvm_sregs *sregs);
613 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
614 				    struct kvm_mp_state *mp_state);
615 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
616 				    struct kvm_mp_state *mp_state);
617 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
618 					struct kvm_guest_debug *dbg);
619 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
620 
621 int kvm_arch_init(void *opaque);
622 void kvm_arch_exit(void);
623 
624 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
625 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
626 
627 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
628 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
629 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
630 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
631 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
632 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
633 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
634 
635 int kvm_arch_hardware_enable(void *garbage);
636 void kvm_arch_hardware_disable(void *garbage);
637 int kvm_arch_hardware_setup(void);
638 void kvm_arch_hardware_unsetup(void);
639 void kvm_arch_check_processor_compat(void *rtn);
640 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
641 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
642 
643 void *kvm_kvzalloc(unsigned long size);
644 void kvm_kvfree(const void *addr);
645 
646 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
647 static inline struct kvm *kvm_arch_alloc_vm(void)
648 {
649 	return kzalloc(sizeof(struct kvm), GFP_KERNEL);
650 }
651 
652 static inline void kvm_arch_free_vm(struct kvm *kvm)
653 {
654 	kfree(kvm);
655 }
656 #endif
657 
658 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
659 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
660 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
661 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
662 #else
663 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
664 {
665 }
666 
667 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
668 {
669 }
670 
671 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
672 {
673 	return false;
674 }
675 #endif
676 
677 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
678 {
679 #ifdef __KVM_HAVE_ARCH_WQP
680 	return vcpu->arch.wqp;
681 #else
682 	return &vcpu->wq;
683 #endif
684 }
685 
686 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
687 void kvm_arch_destroy_vm(struct kvm *kvm);
688 void kvm_arch_sync_events(struct kvm *kvm);
689 
690 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
691 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
692 
693 bool kvm_is_mmio_pfn(pfn_t pfn);
694 
695 struct kvm_irq_ack_notifier {
696 	struct hlist_node link;
697 	unsigned gsi;
698 	void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
699 };
700 
701 struct kvm_assigned_dev_kernel {
702 	struct kvm_irq_ack_notifier ack_notifier;
703 	struct list_head list;
704 	int assigned_dev_id;
705 	int host_segnr;
706 	int host_busnr;
707 	int host_devfn;
708 	unsigned int entries_nr;
709 	int host_irq;
710 	bool host_irq_disabled;
711 	bool pci_2_3;
712 	struct msix_entry *host_msix_entries;
713 	int guest_irq;
714 	struct msix_entry *guest_msix_entries;
715 	unsigned long irq_requested_type;
716 	int irq_source_id;
717 	int flags;
718 	struct pci_dev *dev;
719 	struct kvm *kvm;
720 	spinlock_t intx_lock;
721 	spinlock_t intx_mask_lock;
722 	char irq_name[32];
723 	struct pci_saved_state *pci_saved_state;
724 };
725 
726 struct kvm_irq_mask_notifier {
727 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
728 	int irq;
729 	struct hlist_node link;
730 };
731 
732 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
733 				    struct kvm_irq_mask_notifier *kimn);
734 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
735 				      struct kvm_irq_mask_notifier *kimn);
736 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
737 			     bool mask);
738 
739 int kvm_irq_map_gsi(struct kvm *kvm,
740 		    struct kvm_kernel_irq_routing_entry *entries, int gsi);
741 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
742 
743 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
744 		bool line_status);
745 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
746 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
747 		int irq_source_id, int level, bool line_status);
748 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
749 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
750 void kvm_register_irq_ack_notifier(struct kvm *kvm,
751 				   struct kvm_irq_ack_notifier *kian);
752 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
753 				   struct kvm_irq_ack_notifier *kian);
754 int kvm_request_irq_source_id(struct kvm *kvm);
755 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
756 
757 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
758 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
759 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
760 int kvm_iommu_map_guest(struct kvm *kvm);
761 int kvm_iommu_unmap_guest(struct kvm *kvm);
762 int kvm_assign_device(struct kvm *kvm,
763 		      struct kvm_assigned_dev_kernel *assigned_dev);
764 int kvm_deassign_device(struct kvm *kvm,
765 			struct kvm_assigned_dev_kernel *assigned_dev);
766 #else
767 static inline int kvm_iommu_map_pages(struct kvm *kvm,
768 				      struct kvm_memory_slot *slot)
769 {
770 	return 0;
771 }
772 
773 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
774 					 struct kvm_memory_slot *slot)
775 {
776 }
777 
778 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
779 {
780 	return 0;
781 }
782 #endif
783 
784 static inline void kvm_guest_enter(void)
785 {
786 	unsigned long flags;
787 
788 	BUG_ON(preemptible());
789 
790 	local_irq_save(flags);
791 	guest_enter();
792 	local_irq_restore(flags);
793 
794 	/* KVM does not hold any references to rcu protected data when it
795 	 * switches CPU into a guest mode. In fact switching to a guest mode
796 	 * is very similar to exiting to userspace from rcu point of view. In
797 	 * addition CPU may stay in a guest mode for quite a long time (up to
798 	 * one time slice). Lets treat guest mode as quiescent state, just like
799 	 * we do with user-mode execution.
800 	 */
801 	rcu_virt_note_context_switch(smp_processor_id());
802 }
803 
804 static inline void kvm_guest_exit(void)
805 {
806 	unsigned long flags;
807 
808 	local_irq_save(flags);
809 	guest_exit();
810 	local_irq_restore(flags);
811 }
812 
813 /*
814  * search_memslots() and __gfn_to_memslot() are here because they are
815  * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
816  * gfn_to_memslot() itself isn't here as an inline because that would
817  * bloat other code too much.
818  */
819 static inline struct kvm_memory_slot *
820 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
821 {
822 	struct kvm_memory_slot *memslot;
823 
824 	kvm_for_each_memslot(memslot, slots)
825 		if (gfn >= memslot->base_gfn &&
826 		      gfn < memslot->base_gfn + memslot->npages)
827 			return memslot;
828 
829 	return NULL;
830 }
831 
832 static inline struct kvm_memory_slot *
833 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
834 {
835 	return search_memslots(slots, gfn);
836 }
837 
838 static inline unsigned long
839 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
840 {
841 	return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
842 }
843 
844 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
845 {
846 	return gfn_to_memslot(kvm, gfn)->id;
847 }
848 
849 static inline gfn_t
850 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
851 {
852 	gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
853 
854 	return slot->base_gfn + gfn_offset;
855 }
856 
857 static inline gpa_t gfn_to_gpa(gfn_t gfn)
858 {
859 	return (gpa_t)gfn << PAGE_SHIFT;
860 }
861 
862 static inline gfn_t gpa_to_gfn(gpa_t gpa)
863 {
864 	return (gfn_t)(gpa >> PAGE_SHIFT);
865 }
866 
867 static inline hpa_t pfn_to_hpa(pfn_t pfn)
868 {
869 	return (hpa_t)pfn << PAGE_SHIFT;
870 }
871 
872 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
873 {
874 	unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
875 
876 	return kvm_is_error_hva(hva);
877 }
878 
879 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
880 {
881 	set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
882 }
883 
884 enum kvm_stat_kind {
885 	KVM_STAT_VM,
886 	KVM_STAT_VCPU,
887 };
888 
889 struct kvm_stats_debugfs_item {
890 	const char *name;
891 	int offset;
892 	enum kvm_stat_kind kind;
893 	struct dentry *dentry;
894 };
895 extern struct kvm_stats_debugfs_item debugfs_entries[];
896 extern struct dentry *kvm_debugfs_dir;
897 
898 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
899 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
900 {
901 	if (unlikely(kvm->mmu_notifier_count))
902 		return 1;
903 	/*
904 	 * Ensure the read of mmu_notifier_count happens before the read
905 	 * of mmu_notifier_seq.  This interacts with the smp_wmb() in
906 	 * mmu_notifier_invalidate_range_end to make sure that the caller
907 	 * either sees the old (non-zero) value of mmu_notifier_count or
908 	 * the new (incremented) value of mmu_notifier_seq.
909 	 * PowerPC Book3s HV KVM calls this under a per-page lock
910 	 * rather than under kvm->mmu_lock, for scalability, so
911 	 * can't rely on kvm->mmu_lock to keep things ordered.
912 	 */
913 	smp_rmb();
914 	if (kvm->mmu_notifier_seq != mmu_seq)
915 		return 1;
916 	return 0;
917 }
918 #endif
919 
920 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
921 
922 #ifdef CONFIG_S390
923 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
924 #else
925 #define KVM_MAX_IRQ_ROUTES 1024
926 #endif
927 
928 int kvm_setup_default_irq_routing(struct kvm *kvm);
929 int kvm_set_irq_routing(struct kvm *kvm,
930 			const struct kvm_irq_routing_entry *entries,
931 			unsigned nr,
932 			unsigned flags);
933 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
934 			  const struct kvm_irq_routing_entry *ue);
935 void kvm_free_irq_routing(struct kvm *kvm);
936 
937 #else
938 
939 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
940 
941 #endif
942 
943 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
944 
945 #ifdef CONFIG_HAVE_KVM_EVENTFD
946 
947 void kvm_eventfd_init(struct kvm *kvm);
948 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
949 
950 #ifdef CONFIG_HAVE_KVM_IRQFD
951 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
952 void kvm_irqfd_release(struct kvm *kvm);
953 void kvm_irq_routing_update(struct kvm *);
954 #else
955 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
956 {
957 	return -EINVAL;
958 }
959 
960 static inline void kvm_irqfd_release(struct kvm *kvm) {}
961 #endif
962 
963 #else
964 
965 static inline void kvm_eventfd_init(struct kvm *kvm) {}
966 
967 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
968 {
969 	return -EINVAL;
970 }
971 
972 static inline void kvm_irqfd_release(struct kvm *kvm) {}
973 
974 #ifdef CONFIG_HAVE_KVM_IRQCHIP
975 static inline void kvm_irq_routing_update(struct kvm *kvm)
976 {
977 }
978 #endif
979 
980 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
981 {
982 	return -ENOSYS;
983 }
984 
985 #endif /* CONFIG_HAVE_KVM_EVENTFD */
986 
987 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
988 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
989 {
990 	return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
991 }
992 
993 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
994 
995 #else
996 
997 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
998 
999 #endif
1000 
1001 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
1002 
1003 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
1004 				  unsigned long arg);
1005 
1006 void kvm_free_all_assigned_devices(struct kvm *kvm);
1007 
1008 #else
1009 
1010 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
1011 						unsigned long arg)
1012 {
1013 	return -ENOTTY;
1014 }
1015 
1016 static inline void kvm_free_all_assigned_devices(struct kvm *kvm) {}
1017 
1018 #endif
1019 
1020 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1021 {
1022 	set_bit(req, &vcpu->requests);
1023 }
1024 
1025 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1026 {
1027 	if (test_bit(req, &vcpu->requests)) {
1028 		clear_bit(req, &vcpu->requests);
1029 		return true;
1030 	} else {
1031 		return false;
1032 	}
1033 }
1034 
1035 extern bool kvm_rebooting;
1036 
1037 struct kvm_device_ops;
1038 
1039 struct kvm_device {
1040 	struct kvm_device_ops *ops;
1041 	struct kvm *kvm;
1042 	void *private;
1043 	struct list_head vm_node;
1044 };
1045 
1046 /* create, destroy, and name are mandatory */
1047 struct kvm_device_ops {
1048 	const char *name;
1049 	int (*create)(struct kvm_device *dev, u32 type);
1050 
1051 	/*
1052 	 * Destroy is responsible for freeing dev.
1053 	 *
1054 	 * Destroy may be called before or after destructors are called
1055 	 * on emulated I/O regions, depending on whether a reference is
1056 	 * held by a vcpu or other kvm component that gets destroyed
1057 	 * after the emulated I/O.
1058 	 */
1059 	void (*destroy)(struct kvm_device *dev);
1060 
1061 	int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1062 	int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1063 	int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1064 	long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1065 		      unsigned long arg);
1066 };
1067 
1068 void kvm_device_get(struct kvm_device *dev);
1069 void kvm_device_put(struct kvm_device *dev);
1070 struct kvm_device *kvm_device_from_filp(struct file *filp);
1071 
1072 extern struct kvm_device_ops kvm_mpic_ops;
1073 extern struct kvm_device_ops kvm_xics_ops;
1074 extern struct kvm_device_ops kvm_vfio_ops;
1075 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1076 extern struct kvm_device_ops kvm_flic_ops;
1077 
1078 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1079 
1080 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1081 {
1082 	vcpu->spin_loop.in_spin_loop = val;
1083 }
1084 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1085 {
1086 	vcpu->spin_loop.dy_eligible = val;
1087 }
1088 
1089 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1090 
1091 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1092 {
1093 }
1094 
1095 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1096 {
1097 }
1098 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1099 #endif
1100 
1101