• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 <linux/irqbypass.h>
28 #include <linux/swait.h>
29 #include <linux/refcount.h>
30 #include <linux/nospec.h>
31 #include <asm/signal.h>
32 
33 #include <linux/kvm.h>
34 #include <linux/kvm_para.h>
35 
36 #include <linux/kvm_types.h>
37 
38 #include <asm/kvm_host.h>
39 
40 #ifndef KVM_MAX_VCPU_ID
41 #define KVM_MAX_VCPU_ID KVM_MAX_VCPUS
42 #endif
43 
44 /*
45  * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
46  * in kvm, other bits are visible for userspace which are defined in
47  * include/linux/kvm_h.
48  */
49 #define KVM_MEMSLOT_INVALID	(1UL << 16)
50 
51 /* Two fragments for cross MMIO pages. */
52 #define KVM_MAX_MMIO_FRAGMENTS	2
53 
54 #ifndef KVM_ADDRESS_SPACE_NUM
55 #define KVM_ADDRESS_SPACE_NUM	1
56 #endif
57 
58 /*
59  * For the normal pfn, the highest 12 bits should be zero,
60  * so we can mask bit 62 ~ bit 52  to indicate the error pfn,
61  * mask bit 63 to indicate the noslot pfn.
62  */
63 #define KVM_PFN_ERR_MASK	(0x7ffULL << 52)
64 #define KVM_PFN_ERR_NOSLOT_MASK	(0xfffULL << 52)
65 #define KVM_PFN_NOSLOT		(0x1ULL << 63)
66 
67 #define KVM_PFN_ERR_FAULT	(KVM_PFN_ERR_MASK)
68 #define KVM_PFN_ERR_HWPOISON	(KVM_PFN_ERR_MASK + 1)
69 #define KVM_PFN_ERR_RO_FAULT	(KVM_PFN_ERR_MASK + 2)
70 
71 /*
72  * error pfns indicate that the gfn is in slot but faild to
73  * translate it to pfn on host.
74  */
is_error_pfn(kvm_pfn_t pfn)75 static inline bool is_error_pfn(kvm_pfn_t pfn)
76 {
77 	return !!(pfn & KVM_PFN_ERR_MASK);
78 }
79 
80 /*
81  * error_noslot pfns indicate that the gfn can not be
82  * translated to pfn - it is not in slot or failed to
83  * translate it to pfn.
84  */
is_error_noslot_pfn(kvm_pfn_t pfn)85 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn)
86 {
87 	return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
88 }
89 
90 /* noslot pfn indicates that the gfn is not in slot. */
is_noslot_pfn(kvm_pfn_t pfn)91 static inline bool is_noslot_pfn(kvm_pfn_t pfn)
92 {
93 	return pfn == KVM_PFN_NOSLOT;
94 }
95 
96 /*
97  * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
98  * provide own defines and kvm_is_error_hva
99  */
100 #ifndef KVM_HVA_ERR_BAD
101 
102 #define KVM_HVA_ERR_BAD		(PAGE_OFFSET)
103 #define KVM_HVA_ERR_RO_BAD	(PAGE_OFFSET + PAGE_SIZE)
104 
kvm_is_error_hva(unsigned long addr)105 static inline bool kvm_is_error_hva(unsigned long addr)
106 {
107 	return addr >= PAGE_OFFSET;
108 }
109 
110 #endif
111 
112 #define KVM_ERR_PTR_BAD_PAGE	(ERR_PTR(-ENOENT))
113 
is_error_page(struct page * page)114 static inline bool is_error_page(struct page *page)
115 {
116 	return IS_ERR(page);
117 }
118 
119 #define KVM_REQUEST_MASK           GENMASK(7,0)
120 #define KVM_REQUEST_NO_WAKEUP      BIT(8)
121 #define KVM_REQUEST_WAIT           BIT(9)
122 /*
123  * Architecture-independent vcpu->requests bit members
124  * Bits 4-7 are reserved for more arch-independent bits.
125  */
126 #define KVM_REQ_TLB_FLUSH         (0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
127 #define KVM_REQ_MMU_RELOAD        (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
128 #define KVM_REQ_PENDING_TIMER     2
129 #define KVM_REQ_UNHALT            3
130 #define KVM_REQUEST_ARCH_BASE     8
131 
132 #define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
133 	BUILD_BUG_ON((unsigned)(nr) >= 32 - KVM_REQUEST_ARCH_BASE); \
134 	(unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) | (flags)); \
135 })
136 #define KVM_ARCH_REQ(nr)           KVM_ARCH_REQ_FLAGS(nr, 0)
137 
138 #define KVM_USERSPACE_IRQ_SOURCE_ID		0
139 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID	1
140 
141 extern struct kmem_cache *kvm_vcpu_cache;
142 
143 extern struct mutex kvm_lock;
144 extern struct list_head vm_list;
145 
146 struct kvm_io_range {
147 	gpa_t addr;
148 	int len;
149 	struct kvm_io_device *dev;
150 };
151 
152 #define NR_IOBUS_DEVS 1000
153 
154 struct kvm_io_bus {
155 	int dev_count;
156 	int ioeventfd_count;
157 	struct kvm_io_range range[];
158 };
159 
160 enum kvm_bus {
161 	KVM_MMIO_BUS,
162 	KVM_PIO_BUS,
163 	KVM_VIRTIO_CCW_NOTIFY_BUS,
164 	KVM_FAST_MMIO_BUS,
165 	KVM_NR_BUSES
166 };
167 
168 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
169 		     int len, const void *val);
170 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
171 			    gpa_t addr, int len, const void *val, long cookie);
172 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
173 		    int len, void *val);
174 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
175 			    int len, struct kvm_io_device *dev);
176 void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
177 			       struct kvm_io_device *dev);
178 struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx,
179 					 gpa_t addr);
180 
181 #ifdef CONFIG_KVM_ASYNC_PF
182 struct kvm_async_pf {
183 	struct work_struct work;
184 	struct list_head link;
185 	struct list_head queue;
186 	struct kvm_vcpu *vcpu;
187 	struct mm_struct *mm;
188 	gva_t gva;
189 	unsigned long addr;
190 	struct kvm_arch_async_pf arch;
191 	bool   wakeup_all;
192 };
193 
194 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
195 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
196 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
197 		       struct kvm_arch_async_pf *arch);
198 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
199 #endif
200 
201 enum {
202 	OUTSIDE_GUEST_MODE,
203 	IN_GUEST_MODE,
204 	EXITING_GUEST_MODE,
205 	READING_SHADOW_PAGE_TABLES,
206 };
207 
208 /*
209  * Sometimes a large or cross-page mmio needs to be broken up into separate
210  * exits for userspace servicing.
211  */
212 struct kvm_mmio_fragment {
213 	gpa_t gpa;
214 	void *data;
215 	unsigned len;
216 };
217 
218 struct kvm_vcpu {
219 	struct kvm *kvm;
220 #ifdef CONFIG_PREEMPT_NOTIFIERS
221 	struct preempt_notifier preempt_notifier;
222 #endif
223 	int cpu;
224 	int vcpu_id;
225 	int srcu_idx;
226 	int mode;
227 	unsigned long requests;
228 	unsigned long guest_debug;
229 
230 	int pre_pcpu;
231 	struct list_head blocked_vcpu_list;
232 
233 	struct mutex mutex;
234 	struct kvm_run *run;
235 
236 	int guest_xcr0_loaded;
237 	struct swait_queue_head wq;
238 	struct pid __rcu *pid;
239 	int sigset_active;
240 	sigset_t sigset;
241 	struct kvm_vcpu_stat stat;
242 	unsigned int halt_poll_ns;
243 	bool valid_wakeup;
244 
245 #ifdef CONFIG_HAS_IOMEM
246 	int mmio_needed;
247 	int mmio_read_completed;
248 	int mmio_is_write;
249 	int mmio_cur_fragment;
250 	int mmio_nr_fragments;
251 	struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
252 #endif
253 
254 #ifdef CONFIG_KVM_ASYNC_PF
255 	struct {
256 		u32 queued;
257 		struct list_head queue;
258 		struct list_head done;
259 		spinlock_t lock;
260 	} async_pf;
261 #endif
262 
263 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
264 	/*
265 	 * Cpu relax intercept or pause loop exit optimization
266 	 * in_spin_loop: set when a vcpu does a pause loop exit
267 	 *  or cpu relax intercepted.
268 	 * dy_eligible: indicates whether vcpu is eligible for directed yield.
269 	 */
270 	struct {
271 		bool in_spin_loop;
272 		bool dy_eligible;
273 	} spin_loop;
274 #endif
275 	bool preempted;
276 	struct kvm_vcpu_arch arch;
277 	struct dentry *debugfs_dentry;
278 };
279 
kvm_vcpu_exiting_guest_mode(struct kvm_vcpu * vcpu)280 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
281 {
282 	/*
283 	 * The memory barrier ensures a previous write to vcpu->requests cannot
284 	 * be reordered with the read of vcpu->mode.  It pairs with the general
285 	 * memory barrier following the write of vcpu->mode in VCPU RUN.
286 	 */
287 	smp_mb__before_atomic();
288 	return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
289 }
290 
291 /*
292  * Some of the bitops functions do not support too long bitmaps.
293  * This number must be determined not to exceed such limits.
294  */
295 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
296 
297 struct kvm_memory_slot {
298 	gfn_t base_gfn;
299 	unsigned long npages;
300 	unsigned long *dirty_bitmap;
301 	struct kvm_arch_memory_slot arch;
302 	unsigned long userspace_addr;
303 	u32 flags;
304 	short id;
305 };
306 
kvm_dirty_bitmap_bytes(struct kvm_memory_slot * memslot)307 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
308 {
309 	return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
310 }
311 
312 struct kvm_s390_adapter_int {
313 	u64 ind_addr;
314 	u64 summary_addr;
315 	u64 ind_offset;
316 	u32 summary_offset;
317 	u32 adapter_id;
318 };
319 
320 struct kvm_hv_sint {
321 	u32 vcpu;
322 	u32 sint;
323 };
324 
325 struct kvm_kernel_irq_routing_entry {
326 	u32 gsi;
327 	u32 type;
328 	int (*set)(struct kvm_kernel_irq_routing_entry *e,
329 		   struct kvm *kvm, int irq_source_id, int level,
330 		   bool line_status);
331 	union {
332 		struct {
333 			unsigned irqchip;
334 			unsigned pin;
335 		} irqchip;
336 		struct {
337 			u32 address_lo;
338 			u32 address_hi;
339 			u32 data;
340 			u32 flags;
341 			u32 devid;
342 		} msi;
343 		struct kvm_s390_adapter_int adapter;
344 		struct kvm_hv_sint hv_sint;
345 	};
346 	struct hlist_node link;
347 };
348 
349 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
350 struct kvm_irq_routing_table {
351 	int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
352 	u32 nr_rt_entries;
353 	/*
354 	 * Array indexed by gsi. Each entry contains list of irq chips
355 	 * the gsi is connected to.
356 	 */
357 	struct hlist_head map[0];
358 };
359 #endif
360 
361 #ifndef KVM_PRIVATE_MEM_SLOTS
362 #define KVM_PRIVATE_MEM_SLOTS 0
363 #endif
364 
365 #ifndef KVM_MEM_SLOTS_NUM
366 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
367 #endif
368 
369 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
kvm_arch_vcpu_memslots_id(struct kvm_vcpu * vcpu)370 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
371 {
372 	return 0;
373 }
374 #endif
375 
376 /*
377  * Note:
378  * memslots are not sorted by id anymore, please use id_to_memslot()
379  * to get the memslot by its id.
380  */
381 struct kvm_memslots {
382 	u64 generation;
383 	struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
384 	/* The mapping table from slot id to the index in memslots[]. */
385 	short id_to_index[KVM_MEM_SLOTS_NUM];
386 	atomic_t lru_slot;
387 	int used_slots;
388 };
389 
390 struct kvm {
391 	spinlock_t mmu_lock;
392 	struct mutex slots_lock;
393 	struct mm_struct *mm; /* userspace tied to this vm */
394 	struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
395 	struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
396 
397 	/*
398 	 * created_vcpus is protected by kvm->lock, and is incremented
399 	 * at the beginning of KVM_CREATE_VCPU.  online_vcpus is only
400 	 * incremented after storing the kvm_vcpu pointer in vcpus,
401 	 * and is accessed atomically.
402 	 */
403 	atomic_t online_vcpus;
404 	int created_vcpus;
405 	int last_boosted_vcpu;
406 	struct list_head vm_list;
407 	struct mutex lock;
408 	struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
409 #ifdef CONFIG_HAVE_KVM_EVENTFD
410 	struct {
411 		spinlock_t        lock;
412 		struct list_head  items;
413 		struct list_head  resampler_list;
414 		struct mutex      resampler_lock;
415 	} irqfds;
416 	struct list_head ioeventfds;
417 #endif
418 	struct kvm_vm_stat stat;
419 	struct kvm_arch arch;
420 	refcount_t users_count;
421 #ifdef CONFIG_KVM_MMIO
422 	struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
423 	spinlock_t ring_lock;
424 	struct list_head coalesced_zones;
425 #endif
426 
427 	struct mutex irq_lock;
428 #ifdef CONFIG_HAVE_KVM_IRQCHIP
429 	/*
430 	 * Update side is protected by irq_lock.
431 	 */
432 	struct kvm_irq_routing_table __rcu *irq_routing;
433 #endif
434 #ifdef CONFIG_HAVE_KVM_IRQFD
435 	struct hlist_head irq_ack_notifier_list;
436 #endif
437 
438 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
439 	struct mmu_notifier mmu_notifier;
440 	unsigned long mmu_notifier_seq;
441 	long mmu_notifier_count;
442 #endif
443 	long tlbs_dirty;
444 	struct list_head devices;
445 	struct dentry *debugfs_dentry;
446 	struct kvm_stat_data **debugfs_stat_data;
447 	struct srcu_struct srcu;
448 	struct srcu_struct irq_srcu;
449 	pid_t userspace_pid;
450 };
451 
452 #define kvm_err(fmt, ...) \
453 	pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
454 #define kvm_info(fmt, ...) \
455 	pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
456 #define kvm_debug(fmt, ...) \
457 	pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
458 #define kvm_debug_ratelimited(fmt, ...) \
459 	pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
460 			     ## __VA_ARGS__)
461 #define kvm_pr_unimpl(fmt, ...) \
462 	pr_err_ratelimited("kvm [%i]: " fmt, \
463 			   task_tgid_nr(current), ## __VA_ARGS__)
464 
465 /* The guest did something we don't support. */
466 #define vcpu_unimpl(vcpu, fmt, ...)					\
467 	kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt,			\
468 			(vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
469 
470 #define vcpu_debug(vcpu, fmt, ...)					\
471 	kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
472 #define vcpu_debug_ratelimited(vcpu, fmt, ...)				\
473 	kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id,           \
474 			      ## __VA_ARGS__)
475 #define vcpu_err(vcpu, fmt, ...)					\
476 	kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
477 
kvm_get_bus(struct kvm * kvm,enum kvm_bus idx)478 static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx)
479 {
480 	return srcu_dereference_check(kvm->buses[idx], &kvm->srcu,
481 				      lockdep_is_held(&kvm->slots_lock) ||
482 				      !refcount_read(&kvm->users_count));
483 }
484 
kvm_get_vcpu(struct kvm * kvm,int i)485 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
486 {
487 	int num_vcpus = atomic_read(&kvm->online_vcpus);
488 	i = array_index_nospec(i, num_vcpus);
489 
490 	/* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu.  */
491 	smp_rmb();
492 	return kvm->vcpus[i];
493 }
494 
495 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
496 	for (idx = 0; \
497 	     idx < atomic_read(&kvm->online_vcpus) && \
498 	     (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
499 	     idx++)
500 
kvm_get_vcpu_by_id(struct kvm * kvm,int id)501 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
502 {
503 	struct kvm_vcpu *vcpu = NULL;
504 	int i;
505 
506 	if (id < 0)
507 		return NULL;
508 	if (id < KVM_MAX_VCPUS)
509 		vcpu = kvm_get_vcpu(kvm, id);
510 	if (vcpu && vcpu->vcpu_id == id)
511 		return vcpu;
512 	kvm_for_each_vcpu(i, vcpu, kvm)
513 		if (vcpu->vcpu_id == id)
514 			return vcpu;
515 	return NULL;
516 }
517 
kvm_vcpu_get_idx(struct kvm_vcpu * vcpu)518 static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu)
519 {
520 	struct kvm_vcpu *tmp;
521 	int idx;
522 
523 	kvm_for_each_vcpu(idx, tmp, vcpu->kvm)
524 		if (tmp == vcpu)
525 			return idx;
526 	BUG();
527 }
528 
529 #define kvm_for_each_memslot(memslot, slots)	\
530 	for (memslot = &slots->memslots[0];	\
531 	      memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
532 		memslot++)
533 
534 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
535 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
536 
537 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
538 void vcpu_put(struct kvm_vcpu *vcpu);
539 
540 #ifdef __KVM_HAVE_IOAPIC
541 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
542 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
543 #else
kvm_arch_post_irq_ack_notifier_list_update(struct kvm * kvm)544 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
545 {
546 }
kvm_arch_post_irq_routing_update(struct kvm * kvm)547 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
548 {
549 }
550 #endif
551 
552 #ifdef CONFIG_HAVE_KVM_IRQFD
553 int kvm_irqfd_init(void);
554 void kvm_irqfd_exit(void);
555 #else
kvm_irqfd_init(void)556 static inline int kvm_irqfd_init(void)
557 {
558 	return 0;
559 }
560 
kvm_irqfd_exit(void)561 static inline void kvm_irqfd_exit(void)
562 {
563 }
564 #endif
565 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
566 		  struct module *module);
567 void kvm_exit(void);
568 
569 void kvm_get_kvm(struct kvm *kvm);
570 void kvm_put_kvm(struct kvm *kvm);
571 
__kvm_memslots(struct kvm * kvm,int as_id)572 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
573 {
574 	as_id = array_index_nospec(as_id, KVM_ADDRESS_SPACE_NUM);
575 	return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu,
576 			lockdep_is_held(&kvm->slots_lock) ||
577 			!refcount_read(&kvm->users_count));
578 }
579 
kvm_memslots(struct kvm * kvm)580 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
581 {
582 	return __kvm_memslots(kvm, 0);
583 }
584 
kvm_vcpu_memslots(struct kvm_vcpu * vcpu)585 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
586 {
587 	int as_id = kvm_arch_vcpu_memslots_id(vcpu);
588 
589 	return __kvm_memslots(vcpu->kvm, as_id);
590 }
591 
592 static inline struct kvm_memory_slot *
id_to_memslot(struct kvm_memslots * slots,int id)593 id_to_memslot(struct kvm_memslots *slots, int id)
594 {
595 	int index = slots->id_to_index[id];
596 	struct kvm_memory_slot *slot;
597 
598 	slot = &slots->memslots[index];
599 
600 	WARN_ON(slot->id != id);
601 	return slot;
602 }
603 
604 /*
605  * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
606  * - create a new memory slot
607  * - delete an existing memory slot
608  * - modify an existing memory slot
609  *   -- move it in the guest physical memory space
610  *   -- just change its flags
611  *
612  * Since flags can be changed by some of these operations, the following
613  * differentiation is the best we can do for __kvm_set_memory_region():
614  */
615 enum kvm_mr_change {
616 	KVM_MR_CREATE,
617 	KVM_MR_DELETE,
618 	KVM_MR_MOVE,
619 	KVM_MR_FLAGS_ONLY,
620 };
621 
622 int kvm_set_memory_region(struct kvm *kvm,
623 			  const struct kvm_userspace_memory_region *mem);
624 int __kvm_set_memory_region(struct kvm *kvm,
625 			    const struct kvm_userspace_memory_region *mem);
626 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
627 			   struct kvm_memory_slot *dont);
628 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
629 			    unsigned long npages);
630 void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen);
631 int kvm_arch_prepare_memory_region(struct kvm *kvm,
632 				struct kvm_memory_slot *memslot,
633 				const struct kvm_userspace_memory_region *mem,
634 				enum kvm_mr_change change);
635 void kvm_arch_commit_memory_region(struct kvm *kvm,
636 				const struct kvm_userspace_memory_region *mem,
637 				const struct kvm_memory_slot *old,
638 				const struct kvm_memory_slot *new,
639 				enum kvm_mr_change change);
640 bool kvm_largepages_enabled(void);
641 void kvm_disable_largepages(void);
642 /* flush all memory translations */
643 void kvm_arch_flush_shadow_all(struct kvm *kvm);
644 /* flush memory translations pointing to 'slot' */
645 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
646 				   struct kvm_memory_slot *slot);
647 
648 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
649 			    struct page **pages, int nr_pages);
650 
651 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
652 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
653 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
654 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
655 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
656 				      bool *writable);
657 void kvm_release_page_clean(struct page *page);
658 void kvm_release_page_dirty(struct page *page);
659 void kvm_set_page_accessed(struct page *page);
660 
661 kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
662 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
663 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
664 		      bool *writable);
665 kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
666 kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
667 kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
668 			       bool atomic, bool *async, bool write_fault,
669 			       bool *writable);
670 
671 void kvm_release_pfn_clean(kvm_pfn_t pfn);
672 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
673 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
674 void kvm_get_pfn(kvm_pfn_t pfn);
675 
676 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
677 			int len);
678 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
679 			  unsigned long len);
680 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
681 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
682 			   void *data, unsigned long len);
683 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
684 			 int offset, int len);
685 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
686 		    unsigned long len);
687 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
688 			   void *data, unsigned long len);
689 int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
690 				  void *data, unsigned int offset,
691 				  unsigned long len);
692 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
693 			      gpa_t gpa, unsigned long len);
694 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
695 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
696 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
697 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
698 unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn);
699 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
700 
701 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
702 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
703 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
704 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
705 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
706 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
707 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
708 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
709 			     int len);
710 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
711 			       unsigned long len);
712 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
713 			unsigned long len);
714 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
715 			      int offset, int len);
716 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
717 			 unsigned long len);
718 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
719 
720 void kvm_sigset_activate(struct kvm_vcpu *vcpu);
721 void kvm_sigset_deactivate(struct kvm_vcpu *vcpu);
722 
723 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
724 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
725 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
726 bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
727 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
728 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
729 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool usermode_vcpu_not_eligible);
730 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
731 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
732 
733 void kvm_flush_remote_tlbs(struct kvm *kvm);
734 void kvm_reload_remote_mmus(struct kvm *kvm);
735 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
736 
737 long kvm_arch_dev_ioctl(struct file *filp,
738 			unsigned int ioctl, unsigned long arg);
739 long kvm_arch_vcpu_ioctl(struct file *filp,
740 			 unsigned int ioctl, unsigned long arg);
741 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
742 
743 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
744 
745 int kvm_get_dirty_log(struct kvm *kvm,
746 			struct kvm_dirty_log *log, int *is_dirty);
747 
748 int kvm_get_dirty_log_protect(struct kvm *kvm,
749 			struct kvm_dirty_log *log, bool *is_dirty);
750 
751 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
752 					struct kvm_memory_slot *slot,
753 					gfn_t gfn_offset,
754 					unsigned long mask);
755 
756 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
757 				struct kvm_dirty_log *log);
758 
759 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
760 			bool line_status);
761 long kvm_arch_vm_ioctl(struct file *filp,
762 		       unsigned int ioctl, unsigned long arg);
763 
764 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
765 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
766 
767 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
768 				    struct kvm_translation *tr);
769 
770 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
771 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
772 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
773 				  struct kvm_sregs *sregs);
774 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
775 				  struct kvm_sregs *sregs);
776 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
777 				    struct kvm_mp_state *mp_state);
778 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
779 				    struct kvm_mp_state *mp_state);
780 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
781 					struct kvm_guest_debug *dbg);
782 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
783 
784 int kvm_arch_init(void *opaque);
785 void kvm_arch_exit(void);
786 
787 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
788 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
789 
790 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
791 
792 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
793 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
794 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
795 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
796 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
797 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
798 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
799 
800 bool kvm_arch_has_vcpu_debugfs(void);
801 int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu);
802 
803 int kvm_arch_hardware_enable(void);
804 void kvm_arch_hardware_disable(void);
805 int kvm_arch_hardware_setup(void);
806 void kvm_arch_hardware_unsetup(void);
807 void kvm_arch_check_processor_compat(void *rtn);
808 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
809 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
810 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
811 bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu);
812 
813 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
kvm_arch_alloc_vm(void)814 static inline struct kvm *kvm_arch_alloc_vm(void)
815 {
816 	return kzalloc(sizeof(struct kvm), GFP_KERNEL);
817 }
818 
kvm_arch_free_vm(struct kvm * kvm)819 static inline void kvm_arch_free_vm(struct kvm *kvm)
820 {
821 	kfree(kvm);
822 }
823 #endif
824 
825 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
826 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
827 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
828 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
829 #else
kvm_arch_register_noncoherent_dma(struct kvm * kvm)830 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
831 {
832 }
833 
kvm_arch_unregister_noncoherent_dma(struct kvm * kvm)834 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
835 {
836 }
837 
kvm_arch_has_noncoherent_dma(struct kvm * kvm)838 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
839 {
840 	return false;
841 }
842 #endif
843 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
844 void kvm_arch_start_assignment(struct kvm *kvm);
845 void kvm_arch_end_assignment(struct kvm *kvm);
846 bool kvm_arch_has_assigned_device(struct kvm *kvm);
847 #else
kvm_arch_start_assignment(struct kvm * kvm)848 static inline void kvm_arch_start_assignment(struct kvm *kvm)
849 {
850 }
851 
kvm_arch_end_assignment(struct kvm * kvm)852 static inline void kvm_arch_end_assignment(struct kvm *kvm)
853 {
854 }
855 
kvm_arch_has_assigned_device(struct kvm * kvm)856 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
857 {
858 	return false;
859 }
860 #endif
861 
kvm_arch_vcpu_wq(struct kvm_vcpu * vcpu)862 static inline struct swait_queue_head *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
863 {
864 #ifdef __KVM_HAVE_ARCH_WQP
865 	return vcpu->arch.wqp;
866 #else
867 	return &vcpu->wq;
868 #endif
869 }
870 
871 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
872 /*
873  * returns true if the virtual interrupt controller is initialized and
874  * ready to accept virtual IRQ. On some architectures the virtual interrupt
875  * controller is dynamically instantiated and this is not always true.
876  */
877 bool kvm_arch_intc_initialized(struct kvm *kvm);
878 #else
kvm_arch_intc_initialized(struct kvm * kvm)879 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
880 {
881 	return true;
882 }
883 #endif
884 
885 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
886 void kvm_arch_destroy_vm(struct kvm *kvm);
887 void kvm_arch_sync_events(struct kvm *kvm);
888 
889 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
890 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
891 
892 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
893 bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
894 
895 struct kvm_irq_ack_notifier {
896 	struct hlist_node link;
897 	unsigned gsi;
898 	void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
899 };
900 
901 int kvm_irq_map_gsi(struct kvm *kvm,
902 		    struct kvm_kernel_irq_routing_entry *entries, int gsi);
903 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
904 
905 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
906 		bool line_status);
907 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
908 		int irq_source_id, int level, bool line_status);
909 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
910 			       struct kvm *kvm, int irq_source_id,
911 			       int level, bool line_status);
912 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
913 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
914 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
915 void kvm_register_irq_ack_notifier(struct kvm *kvm,
916 				   struct kvm_irq_ack_notifier *kian);
917 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
918 				   struct kvm_irq_ack_notifier *kian);
919 int kvm_request_irq_source_id(struct kvm *kvm);
920 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
921 
922 /*
923  * search_memslots() and __gfn_to_memslot() are here because they are
924  * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
925  * gfn_to_memslot() itself isn't here as an inline because that would
926  * bloat other code too much.
927  */
928 static inline struct kvm_memory_slot *
search_memslots(struct kvm_memslots * slots,gfn_t gfn)929 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
930 {
931 	int start = 0, end = slots->used_slots;
932 	int slot = atomic_read(&slots->lru_slot);
933 	struct kvm_memory_slot *memslots = slots->memslots;
934 
935 	if (gfn >= memslots[slot].base_gfn &&
936 	    gfn < memslots[slot].base_gfn + memslots[slot].npages)
937 		return &memslots[slot];
938 
939 	while (start < end) {
940 		slot = start + (end - start) / 2;
941 
942 		if (gfn >= memslots[slot].base_gfn)
943 			end = slot;
944 		else
945 			start = slot + 1;
946 	}
947 
948 	if (gfn >= memslots[start].base_gfn &&
949 	    gfn < memslots[start].base_gfn + memslots[start].npages) {
950 		atomic_set(&slots->lru_slot, start);
951 		return &memslots[start];
952 	}
953 
954 	return NULL;
955 }
956 
957 static inline struct kvm_memory_slot *
__gfn_to_memslot(struct kvm_memslots * slots,gfn_t gfn)958 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
959 {
960 	return search_memslots(slots, gfn);
961 }
962 
963 static inline unsigned long
__gfn_to_hva_memslot(struct kvm_memory_slot * slot,gfn_t gfn)964 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
965 {
966 	return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
967 }
968 
memslot_id(struct kvm * kvm,gfn_t gfn)969 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
970 {
971 	return gfn_to_memslot(kvm, gfn)->id;
972 }
973 
974 static inline gfn_t
hva_to_gfn_memslot(unsigned long hva,struct kvm_memory_slot * slot)975 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
976 {
977 	gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
978 
979 	return slot->base_gfn + gfn_offset;
980 }
981 
gfn_to_gpa(gfn_t gfn)982 static inline gpa_t gfn_to_gpa(gfn_t gfn)
983 {
984 	return (gpa_t)gfn << PAGE_SHIFT;
985 }
986 
gpa_to_gfn(gpa_t gpa)987 static inline gfn_t gpa_to_gfn(gpa_t gpa)
988 {
989 	return (gfn_t)(gpa >> PAGE_SHIFT);
990 }
991 
pfn_to_hpa(kvm_pfn_t pfn)992 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
993 {
994 	return (hpa_t)pfn << PAGE_SHIFT;
995 }
996 
kvm_vcpu_gpa_to_page(struct kvm_vcpu * vcpu,gpa_t gpa)997 static inline struct page *kvm_vcpu_gpa_to_page(struct kvm_vcpu *vcpu,
998 						gpa_t gpa)
999 {
1000 	return kvm_vcpu_gfn_to_page(vcpu, gpa_to_gfn(gpa));
1001 }
1002 
kvm_is_error_gpa(struct kvm * kvm,gpa_t gpa)1003 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
1004 {
1005 	unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
1006 
1007 	return kvm_is_error_hva(hva);
1008 }
1009 
1010 enum kvm_stat_kind {
1011 	KVM_STAT_VM,
1012 	KVM_STAT_VCPU,
1013 };
1014 
1015 struct kvm_stat_data {
1016 	int offset;
1017 	int mode;
1018 	struct kvm *kvm;
1019 };
1020 
1021 struct kvm_stats_debugfs_item {
1022 	const char *name;
1023 	int offset;
1024 	enum kvm_stat_kind kind;
1025 	int mode;
1026 };
1027 extern struct kvm_stats_debugfs_item debugfs_entries[];
1028 extern struct dentry *kvm_debugfs_dir;
1029 
1030 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
mmu_notifier_retry(struct kvm * kvm,unsigned long mmu_seq)1031 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1032 {
1033 	if (unlikely(kvm->mmu_notifier_count))
1034 		return 1;
1035 	/*
1036 	 * Ensure the read of mmu_notifier_count happens before the read
1037 	 * of mmu_notifier_seq.  This interacts with the smp_wmb() in
1038 	 * mmu_notifier_invalidate_range_end to make sure that the caller
1039 	 * either sees the old (non-zero) value of mmu_notifier_count or
1040 	 * the new (incremented) value of mmu_notifier_seq.
1041 	 * PowerPC Book3s HV KVM calls this under a per-page lock
1042 	 * rather than under kvm->mmu_lock, for scalability, so
1043 	 * can't rely on kvm->mmu_lock to keep things ordered.
1044 	 */
1045 	smp_rmb();
1046 	if (kvm->mmu_notifier_seq != mmu_seq)
1047 		return 1;
1048 	return 0;
1049 }
1050 #endif
1051 
1052 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1053 
1054 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1055 
1056 bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
1057 int kvm_set_irq_routing(struct kvm *kvm,
1058 			const struct kvm_irq_routing_entry *entries,
1059 			unsigned nr,
1060 			unsigned flags);
1061 int kvm_set_routing_entry(struct kvm *kvm,
1062 			  struct kvm_kernel_irq_routing_entry *e,
1063 			  const struct kvm_irq_routing_entry *ue);
1064 void kvm_free_irq_routing(struct kvm *kvm);
1065 
1066 #else
1067 
kvm_free_irq_routing(struct kvm * kvm)1068 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1069 
1070 #endif
1071 
1072 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1073 
1074 #ifdef CONFIG_HAVE_KVM_EVENTFD
1075 
1076 void kvm_eventfd_init(struct kvm *kvm);
1077 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1078 
1079 #ifdef CONFIG_HAVE_KVM_IRQFD
1080 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1081 void kvm_irqfd_release(struct kvm *kvm);
1082 void kvm_irq_routing_update(struct kvm *);
1083 #else
kvm_irqfd(struct kvm * kvm,struct kvm_irqfd * args)1084 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1085 {
1086 	return -EINVAL;
1087 }
1088 
kvm_irqfd_release(struct kvm * kvm)1089 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1090 #endif
1091 
1092 #else
1093 
kvm_eventfd_init(struct kvm * kvm)1094 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1095 
kvm_irqfd(struct kvm * kvm,struct kvm_irqfd * args)1096 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1097 {
1098 	return -EINVAL;
1099 }
1100 
kvm_irqfd_release(struct kvm * kvm)1101 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1102 
1103 #ifdef CONFIG_HAVE_KVM_IRQCHIP
kvm_irq_routing_update(struct kvm * kvm)1104 static inline void kvm_irq_routing_update(struct kvm *kvm)
1105 {
1106 }
1107 #endif
1108 
kvm_ioeventfd(struct kvm * kvm,struct kvm_ioeventfd * args)1109 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1110 {
1111 	return -ENOSYS;
1112 }
1113 
1114 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1115 
1116 void kvm_arch_irq_routing_update(struct kvm *kvm);
1117 
kvm_make_request(int req,struct kvm_vcpu * vcpu)1118 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1119 {
1120 	/*
1121 	 * Ensure the rest of the request is published to kvm_check_request's
1122 	 * caller.  Paired with the smp_mb__after_atomic in kvm_check_request.
1123 	 */
1124 	smp_wmb();
1125 	set_bit(req & KVM_REQUEST_MASK, &vcpu->requests);
1126 }
1127 
kvm_request_pending(struct kvm_vcpu * vcpu)1128 static inline bool kvm_request_pending(struct kvm_vcpu *vcpu)
1129 {
1130 	return READ_ONCE(vcpu->requests);
1131 }
1132 
kvm_test_request(int req,struct kvm_vcpu * vcpu)1133 static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)
1134 {
1135 	return test_bit(req & KVM_REQUEST_MASK, &vcpu->requests);
1136 }
1137 
kvm_clear_request(int req,struct kvm_vcpu * vcpu)1138 static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu)
1139 {
1140 	clear_bit(req & KVM_REQUEST_MASK, &vcpu->requests);
1141 }
1142 
kvm_check_request(int req,struct kvm_vcpu * vcpu)1143 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1144 {
1145 	if (kvm_test_request(req, vcpu)) {
1146 		kvm_clear_request(req, vcpu);
1147 
1148 		/*
1149 		 * Ensure the rest of the request is visible to kvm_check_request's
1150 		 * caller.  Paired with the smp_wmb in kvm_make_request.
1151 		 */
1152 		smp_mb__after_atomic();
1153 		return true;
1154 	} else {
1155 		return false;
1156 	}
1157 }
1158 
1159 extern bool kvm_rebooting;
1160 
1161 extern unsigned int halt_poll_ns;
1162 extern unsigned int halt_poll_ns_grow;
1163 extern unsigned int halt_poll_ns_shrink;
1164 
1165 struct kvm_device {
1166 	struct kvm_device_ops *ops;
1167 	struct kvm *kvm;
1168 	void *private;
1169 	struct list_head vm_node;
1170 };
1171 
1172 /* create, destroy, and name are mandatory */
1173 struct kvm_device_ops {
1174 	const char *name;
1175 
1176 	/*
1177 	 * create is called holding kvm->lock and any operations not suitable
1178 	 * to do while holding the lock should be deferred to init (see
1179 	 * below).
1180 	 */
1181 	int (*create)(struct kvm_device *dev, u32 type);
1182 
1183 	/*
1184 	 * init is called after create if create is successful and is called
1185 	 * outside of holding kvm->lock.
1186 	 */
1187 	void (*init)(struct kvm_device *dev);
1188 
1189 	/*
1190 	 * Destroy is responsible for freeing dev.
1191 	 *
1192 	 * Destroy may be called before or after destructors are called
1193 	 * on emulated I/O regions, depending on whether a reference is
1194 	 * held by a vcpu or other kvm component that gets destroyed
1195 	 * after the emulated I/O.
1196 	 */
1197 	void (*destroy)(struct kvm_device *dev);
1198 
1199 	int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1200 	int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1201 	int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1202 	long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1203 		      unsigned long arg);
1204 };
1205 
1206 void kvm_device_get(struct kvm_device *dev);
1207 void kvm_device_put(struct kvm_device *dev);
1208 struct kvm_device *kvm_device_from_filp(struct file *filp);
1209 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1210 void kvm_unregister_device_ops(u32 type);
1211 
1212 extern struct kvm_device_ops kvm_mpic_ops;
1213 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1214 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1215 
1216 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1217 
kvm_vcpu_set_in_spin_loop(struct kvm_vcpu * vcpu,bool val)1218 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1219 {
1220 	vcpu->spin_loop.in_spin_loop = val;
1221 }
kvm_vcpu_set_dy_eligible(struct kvm_vcpu * vcpu,bool val)1222 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1223 {
1224 	vcpu->spin_loop.dy_eligible = val;
1225 }
1226 
1227 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1228 
kvm_vcpu_set_in_spin_loop(struct kvm_vcpu * vcpu,bool val)1229 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1230 {
1231 }
1232 
kvm_vcpu_set_dy_eligible(struct kvm_vcpu * vcpu,bool val)1233 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1234 {
1235 }
1236 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1237 
1238 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1239 bool kvm_arch_has_irq_bypass(void);
1240 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1241 			   struct irq_bypass_producer *);
1242 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1243 			   struct irq_bypass_producer *);
1244 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1245 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1246 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1247 				  uint32_t guest_irq, bool set);
1248 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1249 
1250 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1251 /* If we wakeup during the poll time, was it a sucessful poll? */
vcpu_valid_wakeup(struct kvm_vcpu * vcpu)1252 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1253 {
1254 	return vcpu->valid_wakeup;
1255 }
1256 
1257 #else
vcpu_valid_wakeup(struct kvm_vcpu * vcpu)1258 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1259 {
1260 	return true;
1261 }
1262 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
1263 
1264 typedef int (*kvm_vm_thread_fn_t)(struct kvm *kvm, uintptr_t data);
1265 
1266 int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn,
1267 				uintptr_t data, const char *name,
1268 				struct task_struct **thread_ptr);
1269 
1270 #endif
1271