kvm_host.h source code [linux/include/linux/kvm_host.h]

1	/ SPDX-License-Identifier: GPL-2.0-only /
2	#ifndef __KVM_HOST_H
3	#define __KVM_HOST_H
4
5
6	#include <linux/types.h>
7	#include <linux/hardirq.h>
8	#include <linux/list.h>
9	#include <linux/mutex.h>
10	#include <linux/spinlock.h>
11	#include <linux/signal.h>
12	#include <linux/sched.h>
13	#include <linux/sched/stat.h>
14	#include <linux/bug.h>
15	#include <linux/minmax.h>
16	#include <linux/mm.h>
17	#include <linux/mmu_notifier.h>
18	#include <linux/preempt.h>
19	#include <linux/msi.h>
20	#include <linux/slab.h>
21	#include <linux/vmalloc.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/rcuwait.h>
29	#include <linux/refcount.h>
30	#include <linux/nospec.h>
31	#include <linux/notifier.h>
32	#include <linux/ftrace.h>
33	#include <linux/hashtable.h>
34	#include <linux/instrumentation.h>
35	#include <linux/interval_tree.h>
36	#include <linux/rbtree.h>
37	#include <linux/xarray.h>
38	#include <asm/signal.h>
39
40	#include <linux/kvm.h>
41	#include <linux/kvm_para.h>
42
43	#include <linux/kvm_types.h>
44
45	#include <asm/kvm_host.h>
46	#include <linux/kvm_dirty_ring.h>
47
48	#ifndef KVM_MAX_VCPU_IDS
49	#define KVM_MAX_VCPU_IDS KVM_MAX_VCPUS
50	#endif
51
52	/*
53	* The bit 16 ~ bit 31 of kvm_userspace_memory_region::flags are internally
54	* used in kvm, other bits are visible for userspace which are defined in
55	* include/linux/kvm_h.
56	*/
57	#define KVM_MEMSLOT_INVALID (1UL << 16)
58
59	/*
60	* Bit 63 of the memslot generation number is an "update in-progress flag",
61	* e.g. is temporarily set for the duration of kvm_swap_active_memslots().
62	* This flag effectively creates a unique generation number that is used to
63	* mark cached memslot data, e.g. MMIO accesses, as potentially being stale,
64	* i.e. may (or may not) have come from the previous memslots generation.
65	*
66	* This is necessary because the actual memslots update is not atomic with
67	* respect to the generation number update. Updating the generation number
68	* first would allow a vCPU to cache a spte from the old memslots using the
69	* new generation number, and updating the generation number after switching
70	* to the new memslots would allow cache hits using the old generation number
71	* to reference the defunct memslots.
72	*
73	* This mechanism is used to prevent getting hits in KVM's caches while a
74	* memslot update is in-progress, and to prevent cache hits after updating
75	* the actual generation number against accesses that were inserted into the
76	* cache before the memslots were updated.
77	*/
78	#define KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS BIT_ULL(63)
79
80	/ Two fragments for cross MMIO pages. /
81	#define KVM_MAX_MMIO_FRAGMENTS 2
82
83	#ifndef KVM_MAX_NR_ADDRESS_SPACES
84	#define KVM_MAX_NR_ADDRESS_SPACES 1
85	#endif
86
87	/*
88	* For the normal pfn, the highest 12 bits should be zero,
89	* so we can mask bit 62 ~ bit 52 to indicate the error pfn,
90	* mask bit 63 to indicate the noslot pfn.
91	*/
92	#define KVM_PFN_ERR_MASK (0x7ffULL << 52)
93	#define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
94	#define KVM_PFN_NOSLOT (0x1ULL << 63)
95
96	#define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
97	#define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
98	#define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
99	#define KVM_PFN_ERR_SIGPENDING (KVM_PFN_ERR_MASK + 3)
100	#define KVM_PFN_ERR_NEEDS_IO (KVM_PFN_ERR_MASK + 4)
101
102	/*
103	* error pfns indicate that the gfn is in slot but faild to
104	* translate it to pfn on host.
105	*/
106	static inline bool is_error_pfn(kvm_pfn_t pfn)
107	{
108	return !!(pfn & KVM_PFN_ERR_MASK);
109	}
110
111	/*
112	* KVM_PFN_ERR_SIGPENDING indicates that fetching the PFN was interrupted
113	* by a pending signal. Note, the signal may or may not be fatal.
114	*/
115	static inline bool is_sigpending_pfn(kvm_pfn_t pfn)
116	{
117	return pfn == KVM_PFN_ERR_SIGPENDING;
118	}
119
120	/*
121	* error_noslot pfns indicate that the gfn can not be
122	* translated to pfn - it is not in slot or failed to
123	* translate it to pfn.
124	*/
125	static inline bool is_error_noslot_pfn(kvm_pfn_t pfn)
126	{
127	return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
128	}
129
130	/ noslot pfn indicates that the gfn is not in slot. /
131	static inline bool is_noslot_pfn(kvm_pfn_t pfn)
132	{
133	return pfn == KVM_PFN_NOSLOT;
134	}
135
136	/*
137	* architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
138	* provide own defines and kvm_is_error_hva
139	*/
140	#ifndef KVM_HVA_ERR_BAD
141
142	#define KVM_HVA_ERR_BAD (PAGE_OFFSET)
143	#define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
144
145	static inline bool kvm_is_error_hva(unsigned long addr)
146	{
147	return addr >= PAGE_OFFSET;
148	}
149
150	#endif
151
152	static inline bool kvm_is_error_gpa(gpa_t gpa)
153	{
154	return gpa == INVALID_GPA;
155	}
156
157	#define KVM_REQUEST_MASK GENMASK(7,0)
158	#define KVM_REQUEST_NO_WAKEUP BIT(8)
159	#define KVM_REQUEST_WAIT BIT(9)
160	#define KVM_REQUEST_NO_ACTION BIT(10)
161	/*
162	* Architecture-independent vcpu->requests bit members
163	* Bits 3-7 are reserved for more arch-independent bits.
164	*/
165	#define KVM_REQ_TLB_FLUSH (0 \| KVM_REQUEST_WAIT \| KVM_REQUEST_NO_WAKEUP)
166	#define KVM_REQ_VM_DEAD (1 \| KVM_REQUEST_WAIT \| KVM_REQUEST_NO_WAKEUP)
167	#define KVM_REQ_UNBLOCK 2
168	#define KVM_REQ_DIRTY_RING_SOFT_FULL 3
169	#define KVM_REQUEST_ARCH_BASE 8
170
171	/*
172	* KVM_REQ_OUTSIDE_GUEST_MODE exists is purely as way to force the vCPU to
173	* OUTSIDE_GUEST_MODE. KVM_REQ_OUTSIDE_GUEST_MODE differs from a vCPU "kick"
174	* in that it ensures the vCPU has reached OUTSIDE_GUEST_MODE before continuing
175	* on. A kick only guarantees that the vCPU is on its way out, e.g. a previous
176	* kick may have set vcpu->mode to EXITING_GUEST_MODE, and so there's no
177	* guarantee the vCPU received an IPI and has actually exited guest mode.
178	*/
179	#define KVM_REQ_OUTSIDE_GUEST_MODE (KVM_REQUEST_NO_ACTION \| KVM_REQUEST_WAIT \| KVM_REQUEST_NO_WAKEUP)
180
181	#define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
182	BUILD_BUG_ON((unsigned)(nr) >= (sizeof_field(struct kvm_vcpu, requests) * 8) - KVM_REQUEST_ARCH_BASE); \
183	(unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) \| (flags)); \
184	})
185	#define KVM_ARCH_REQ(nr) KVM_ARCH_REQ_FLAGS(nr, 0)
186
187	bool kvm_make_vcpus_request_mask(struct kvm kvm, unsigned* int req,
188	unsigned long *vcpu_bitmap);
189	bool kvm_make_all_cpus_request(struct kvm kvm, unsigned* int req);
190
191	#define KVM_USERSPACE_IRQ_SOURCE_ID 0
192	#define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
193
194	extern struct mutex kvm_lock;
195	extern struct list_head vm_list;
196
197	struct kvm_io_range {
198	gpa_t addr;
199	int len;
200	struct kvm_io_device *dev;
201	};
202
203	#define NR_IOBUS_DEVS 1000
204
205	struct kvm_io_bus {
206	int dev_count;
207	int ioeventfd_count;
208	struct kvm_io_range range[];
209	};
210
211	enum kvm_bus {
212	KVM_MMIO_BUS,
213	KVM_PIO_BUS,
214	KVM_VIRTIO_CCW_NOTIFY_BUS,
215	KVM_FAST_MMIO_BUS,
216	KVM_IOCSR_BUS,
217	KVM_NR_BUSES
218	};
219
220	int kvm_io_bus_write(struct kvm_vcpu vcpu, enum* kvm_bus bus_idx, gpa_t addr,
221	int len, const void *val);
222	int kvm_io_bus_write_cookie(struct kvm_vcpu vcpu, enum* kvm_bus bus_idx,
223	gpa_t addr, int len, const void val, long* cookie);
224	int kvm_io_bus_read(struct kvm_vcpu vcpu, enum* kvm_bus bus_idx, gpa_t addr,
225	int len, void *val);
226	int kvm_io_bus_register_dev(struct kvm kvm, enum* kvm_bus bus_idx, gpa_t addr,
227	int len, struct kvm_io_device *dev);
228	int kvm_io_bus_unregister_dev(struct kvm kvm, enum* kvm_bus bus_idx,
229	struct kvm_io_device *dev);
230	struct kvm_io_device kvm_io_bus_get_dev(struct* kvm kvm, enum* kvm_bus bus_idx,
231	gpa_t addr);
232
233	#ifdef CONFIG_KVM_ASYNC_PF
234	struct kvm_async_pf {
235	struct work_struct work;
236	struct list_head link;
237	struct list_head queue;
238	struct kvm_vcpu *vcpu;
239	gpa_t cr2_or_gpa;
240	unsigned long addr;
241	struct kvm_arch_async_pf arch;
242	bool wakeup_all;
243	bool notpresent_injected;
244	};
245
246	void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
247	void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
248	bool kvm_setup_async_pf(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
249	unsigned long hva, struct kvm_arch_async_pf *arch);
250	int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
251	#endif
252
253	#ifdef CONFIG_KVM_GENERIC_MMU_NOTIFIER
254	union kvm_mmu_notifier_arg {
255	unsigned long attributes;
256	};
257
258	enum kvm_gfn_range_filter {
259	KVM_FILTER_SHARED = BIT(`0`),
260	KVM_FILTER_PRIVATE = BIT(`1`),
261	};
262
263	struct kvm_gfn_range {
264	struct kvm_memory_slot *slot;
265	gfn_t start;
266	gfn_t end;
267	union kvm_mmu_notifier_arg arg;
268	enum kvm_gfn_range_filter attr_filter;
269	bool may_block;
270	bool lockless;
271	};
272	bool kvm_unmap_gfn_range(struct kvm kvm, struct* kvm_gfn_range *range);
273	bool kvm_age_gfn(struct kvm kvm, struct* kvm_gfn_range *range);
274	bool kvm_test_age_gfn(struct kvm kvm, struct* kvm_gfn_range *range);
275	#endif
276
277	enum {
278	OUTSIDE_GUEST_MODE,
279	IN_GUEST_MODE,
280	EXITING_GUEST_MODE,
281	READING_SHADOW_PAGE_TABLES,
282	};
283
284	struct kvm_host_map {
285	/*
286	* Only valid if the 'pfn' is managed by the host kernel (i.e. There is
287	* a 'struct page' for it. When using mem= kernel parameter some memory
288	* can be used as guest memory but they are not managed by host
289	* kernel).
290	*/
291	struct page *pinned_page;
292	struct page *page;
293	void *hva;
294	kvm_pfn_t pfn;
295	kvm_pfn_t gfn;
296	bool writable;
297	};
298
299	/*
300	* Used to check if the mapping is valid or not. Never use 'kvm_host_map'
301	* directly to check for that.
302	*/
303	static inline bool kvm_vcpu_mapped(struct kvm_host_map *map)
304	{
305	return !!map->hva;
306	}
307
308	static inline bool kvm_vcpu_can_poll(ktime_t cur, ktime_t stop)
309	{
310	return single_task_running() && !need_resched() && ktime_before(cmp1: cur, cmp2: stop);
311	}
312
313	/*
314	* Sometimes a large or cross-page mmio needs to be broken up into separate
315	* exits for userspace servicing.
316	*/
317	struct kvm_mmio_fragment {
318	gpa_t gpa;
319	void *data;
320	unsigned len;
321	};
322
323	struct kvm_vcpu {
324	struct kvm *kvm;
325	#ifdef CONFIG_PREEMPT_NOTIFIERS
326	struct preempt_notifier preempt_notifier;
327	#endif
328	int cpu;
329	int vcpu_id; / id given by userspace at creation /
330	int vcpu_idx; / index into kvm->vcpu_array /
331	int ____srcu_idx; / Don't use this directly. You've been warned. /
332	#ifdef CONFIG_PROVE_RCU
333	int srcu_depth;
334	#endif
335	int mode;
336	u64 requests;
337	unsigned long guest_debug;
338
339	struct mutex mutex;
340	struct kvm_run *run;
341
342	#ifndef __KVM_HAVE_ARCH_WQP
343	struct rcuwait wait;
344	#endif
345	struct pid *pid;
346	rwlock_t pid_lock;
347	int sigset_active;
348	sigset_t sigset;
349	unsigned int halt_poll_ns;
350	bool valid_wakeup;
351
352	#ifdef CONFIG_HAS_IOMEM
353	int mmio_needed;
354	int mmio_read_completed;
355	int mmio_is_write;
356	int mmio_cur_fragment;
357	int mmio_nr_fragments;
358	struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
359	#endif
360
361	#ifdef CONFIG_KVM_ASYNC_PF
362	struct {
363	u32 queued;
364	struct list_head queue;
365	struct list_head done;
366	spinlock_t lock;
367	} async_pf;
368	#endif
369
370	#ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
371	/*
372	* Cpu relax intercept or pause loop exit optimization
373	* in_spin_loop: set when a vcpu does a pause loop exit
374	* or cpu relax intercepted.
375	* dy_eligible: indicates whether vcpu is eligible for directed yield.
376	*/
377	struct {
378	bool in_spin_loop;
379	bool dy_eligible;
380	} spin_loop;
381	#endif
382	bool wants_to_run;
383	bool preempted;
384	bool ready;
385	bool scheduled_out;
386	struct kvm_vcpu_arch arch;
387	struct kvm_vcpu_stat stat;
388	char stats_id[KVM_STATS_NAME_SIZE];
389	struct kvm_dirty_ring dirty_ring;
390
391	/*
392	* The most recently used memslot by this vCPU and the slots generation
393	* for which it is valid.
394	* No wraparound protection is needed since generations won't overflow in
395	* thousands of years, even assuming 1M memslot operations per second.
396	*/
397	struct kvm_memory_slot *last_used_slot;
398	u64 last_used_slot_gen;
399	};
400
401	/*
402	* Start accounting time towards a guest.
403	* Must be called before entering guest context.
404	*/
405	static __always_inline void guest_timing_enter_irqoff(void)
406	{
407	/*
408	* This is running in ioctl context so its safe to assume that it's the
409	* stime pending cputime to flush.
410	*/
411	instrumentation_begin();
412	vtime_account_guest_enter();
413	instrumentation_end();
414	}
415
416	/*
417	* Enter guest context and enter an RCU extended quiescent state.
418	*
419	* Between guest_context_enter_irqoff() and guest_context_exit_irqoff() it is
420	* unsafe to use any code which may directly or indirectly use RCU, tracing
421	* (including IRQ flag tracing), or lockdep. All code in this period must be
422	* non-instrumentable.
423	*/
424	static __always_inline void guest_context_enter_irqoff(void)
425	{
426	/*
427	* KVM does not hold any references to rcu protected data when it
428	* switches CPU into a guest mode. In fact switching to a guest mode
429	* is very similar to exiting to userspace from rcu point of view. In
430	* addition CPU may stay in a guest mode for quite a long time (up to
431	* one time slice). Lets treat guest mode as quiescent state, just like
432	* we do with user-mode execution.
433	*/
434	if (!context_tracking_guest_enter()) {
435	instrumentation_begin();
436	rcu_virt_note_context_switch();
437	instrumentation_end();
438	}
439	}
440
441	/*
442	* Deprecated. Architectures should move to guest_timing_enter_irqoff() and
443	* guest_state_enter_irqoff().
444	*/
445	static __always_inline void guest_enter_irqoff(void)
446	{
447	guest_timing_enter_irqoff();
448	guest_context_enter_irqoff();
449	}
450
451	/**
452	* guest_state_enter_irqoff - Fixup state when entering a guest
453	*
454	* Entry to a guest will enable interrupts, but the kernel state is interrupts
455	* disabled when this is invoked. Also tell RCU about it.
456	*
457	* 1) Trace interrupts on state
458	* 2) Invoke context tracking if enabled to adjust RCU state
459	* 3) Tell lockdep that interrupts are enabled
460	*
461	* Invoked from architecture specific code before entering a guest.
462	* Must be called with interrupts disabled and the caller must be
463	* non-instrumentable.
464	* The caller has to invoke guest_timing_enter_irqoff() before this.
465	*
466	* Note: this is analogous to exit_to_user_mode().
467	*/
468	static __always_inline void guest_state_enter_irqoff(void)
469	{
470	instrumentation_begin();
471	trace_hardirqs_on_prepare();
472	lockdep_hardirqs_on_prepare();
473	instrumentation_end();
474
475	guest_context_enter_irqoff();
476	lockdep_hardirqs_on(CALLER_ADDR0);
477	}
478
479	/*
480	* Exit guest context and exit an RCU extended quiescent state.
481	*
482	* Between guest_context_enter_irqoff() and guest_context_exit_irqoff() it is
483	* unsafe to use any code which may directly or indirectly use RCU, tracing
484	* (including IRQ flag tracing), or lockdep. All code in this period must be
485	* non-instrumentable.
486	*/
487	static __always_inline void guest_context_exit_irqoff(void)
488	{
489	/*
490	* Guest mode is treated as a quiescent state, see
491	* guest_context_enter_irqoff() for more details.
492	*/
493	if (!context_tracking_guest_exit()) {
494	instrumentation_begin();
495	rcu_virt_note_context_switch();
496	instrumentation_end();
497	}
498	}
499
500	/*
501	* Stop accounting time towards a guest.
502	* Must be called after exiting guest context.
503	*/
504	static __always_inline void guest_timing_exit_irqoff(void)
505	{
506	instrumentation_begin();
507	/ Flush the guest cputime we spent on the guest /
508	vtime_account_guest_exit();
509	instrumentation_end();
510	}
511
512	/*
513	* Deprecated. Architectures should move to guest_state_exit_irqoff() and
514	* guest_timing_exit_irqoff().
515	*/
516	static __always_inline void guest_exit_irqoff(void)
517	{
518	guest_context_exit_irqoff();
519	guest_timing_exit_irqoff();
520	}
521
522	static inline void guest_exit(void)
523	{
524	unsigned long flags;
525
526	local_irq_save(flags);
527	guest_exit_irqoff();
528	local_irq_restore(flags);
529	}
530
531	/**
532	* guest_state_exit_irqoff - Establish state when returning from guest mode
533	*
534	* Entry from a guest disables interrupts, but guest mode is traced as
535	* interrupts enabled. Also with NO_HZ_FULL RCU might be idle.
536	*
537	* 1) Tell lockdep that interrupts are disabled
538	* 2) Invoke context tracking if enabled to reactivate RCU
539	* 3) Trace interrupts off state
540	*
541	* Invoked from architecture specific code after exiting a guest.
542	* Must be invoked with interrupts disabled and the caller must be
543	* non-instrumentable.
544	* The caller has to invoke guest_timing_exit_irqoff() after this.
545	*
546	* Note: this is analogous to enter_from_user_mode().
547	*/
548	static __always_inline void guest_state_exit_irqoff(void)
549	{
550	lockdep_hardirqs_off(CALLER_ADDR0);
551	guest_context_exit_irqoff();
552
553	instrumentation_begin();
554	trace_hardirqs_off_finish();
555	instrumentation_end();
556	}
557
558	static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
559	{
560	/*
561	* The memory barrier ensures a previous write to vcpu->requests cannot
562	* be reordered with the read of vcpu->mode. It pairs with the general
563	* memory barrier following the write of vcpu->mode in VCPU RUN.
564	*/
565	smp_mb__before_atomic();
566	return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
567	}
568
569	/*
570	* Some of the bitops functions do not support too long bitmaps.
571	* This number must be determined not to exceed such limits.
572	*/
573	#define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
574
575	/*
576	* Since at idle each memslot belongs to two memslot sets it has to contain
577	* two embedded nodes for each data structure that it forms a part of.
578	*
579	* Two memslot sets (one active and one inactive) are necessary so the VM
580	* continues to run on one memslot set while the other is being modified.
581	*
582	* These two memslot sets normally point to the same set of memslots.
583	* They can, however, be desynchronized when performing a memslot management
584	* operation by replacing the memslot to be modified by its copy.
585	* After the operation is complete, both memslot sets once again point to
586	* the same, common set of memslot data.
587	*
588	* The memslots themselves are independent of each other so they can be
589	* individually added or deleted.
590	*/
591	struct kvm_memory_slot {
592	struct hlist_node id_node[`2`];
593	struct interval_tree_node hva_node[`2`];
594	struct rb_node gfn_node[`2`];
595	gfn_t base_gfn;
596	unsigned long npages;
597	unsigned long *dirty_bitmap;
598	struct kvm_arch_memory_slot arch;
599	unsigned long userspace_addr;
600	u32 flags;
601	short id;
602	u16 as_id;
603
604	#ifdef CONFIG_KVM_PRIVATE_MEM
605	struct {
606	/*
607	* Writes protected by kvm->slots_lock. Acquiring a
608	* reference via kvm_gmem_get_file() is protected by
609	* either kvm->slots_lock or kvm->srcu.
610	*/
611	struct file *file;
612	pgoff_t pgoff;
613	} gmem;
614	#endif
615	};
616
617	static inline bool kvm_slot_can_be_private(const struct kvm_memory_slot *slot)
618	{
619	return slot && (slot->flags & KVM_MEM_GUEST_MEMFD);
620	}
621
622	static inline bool kvm_slot_dirty_track_enabled(const struct kvm_memory_slot *slot)
623	{
624	return slot->flags & KVM_MEM_LOG_DIRTY_PAGES;
625	}
626
627	static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
628	{
629	return ALIGN(memslot->npages, BITS_PER_LONG) / `8`;
630	}
631
632	static inline unsigned long kvm_second_dirty_bitmap(struct* kvm_memory_slot *memslot)
633	{
634	unsigned long len = kvm_dirty_bitmap_bytes(memslot);
635
636	return memslot->dirty_bitmap + len / sizeof(*memslot->dirty_bitmap);
637	}
638
639	#ifndef KVM_DIRTY_LOG_MANUAL_CAPS
640	#define KVM_DIRTY_LOG_MANUAL_CAPS KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE
641	#endif
642
643	struct kvm_s390_adapter_int {
644	u64 ind_addr;
645	u64 summary_addr;
646	u64 ind_offset;
647	u32 summary_offset;
648	u32 adapter_id;
649	};
650
651	struct kvm_hv_sint {
652	u32 vcpu;
653	u32 sint;
654	};
655
656	struct kvm_xen_evtchn {
657	u32 port;
658	u32 vcpu_id;
659	int vcpu_idx;
660	u32 priority;
661	};
662
663	struct kvm_kernel_irq_routing_entry {
664	u32 gsi;
665	u32 type;
666	int (set)(struct* kvm_kernel_irq_routing_entry *e,
667	struct kvm kvm, int* irq_source_id, int level,
668	bool line_status);
669	union {
670	struct {
671	unsigned irqchip;
672	unsigned pin;
673	} irqchip;
674	struct {
675	u32 address_lo;
676	u32 address_hi;
677	u32 data;
678	u32 flags;
679	u32 devid;
680	} msi;
681	struct kvm_s390_adapter_int adapter;
682	struct kvm_hv_sint hv_sint;
683	struct kvm_xen_evtchn xen_evtchn;
684	};
685	struct hlist_node link;
686	};
687
688	#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
689	struct kvm_irq_routing_table {
690	int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
691	u32 nr_rt_entries;
692	/*
693	* Array indexed by gsi. Each entry contains list of irq chips
694	* the gsi is connected to.
695	*/
696	struct hlist_head map[] __counted_by(nr_rt_entries);
697	};
698	#endif
699
700	bool kvm_arch_irqchip_in_kernel(struct kvm *kvm);
701
702	#ifndef KVM_INTERNAL_MEM_SLOTS
703	#define KVM_INTERNAL_MEM_SLOTS 0
704	#endif
705
706	#define KVM_MEM_SLOTS_NUM SHRT_MAX
707	#define KVM_USER_MEM_SLOTS (KVM_MEM_SLOTS_NUM - KVM_INTERNAL_MEM_SLOTS)
708
709	#if KVM_MAX_NR_ADDRESS_SPACES == 1
710	static inline int kvm_arch_nr_memslot_as_ids(struct kvm *kvm)
711	{
712	return KVM_MAX_NR_ADDRESS_SPACES;
713	}
714
715	static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
716	{
717	return `0`;
718	}
719	#endif
720
721	/*
722	* Arch code must define kvm_arch_has_private_mem if support for private memory
723	* is enabled.
724	*/
725	#if !defined(kvm_arch_has_private_mem) && !IS_ENABLED(CONFIG_KVM_PRIVATE_MEM)
726	static inline bool kvm_arch_has_private_mem(struct kvm *kvm)
727	{
728	return false;
729	}
730	#endif
731
732	#ifndef kvm_arch_has_readonly_mem
733	static inline bool kvm_arch_has_readonly_mem(struct kvm *kvm)
734	{
735	return IS_ENABLED(CONFIG_HAVE_KVM_READONLY_MEM);
736	}
737	#endif
738
739	struct kvm_memslots {
740	u64 generation;
741	atomic_long_t last_used_slot;
742	struct rb_root_cached hva_tree;
743	struct rb_root gfn_tree;
744	/*
745	* The mapping table from slot id to memslot.
746	*
747	* 7-bit bucket count matches the size of the old id to index array for
748	* 512 slots, while giving good performance with this slot count.
749	* Higher bucket counts bring only small performance improvements but
750	* always result in higher memory usage (even for lower memslot counts).
751	*/
752	DECLARE_HASHTABLE(id_hash, `7`);
753	int node_idx;
754	};
755
756	struct kvm {
757	#ifdef KVM_HAVE_MMU_RWLOCK
758	rwlock_t mmu_lock;
759	#else
760	spinlock_t mmu_lock;
761	#endif /* KVM_HAVE_MMU_RWLOCK */
762
763	struct mutex slots_lock;
764
765	/*
766	* Protects the arch-specific fields of struct kvm_memory_slots in
767	* use by the VM. To be used under the slots_lock (above) or in a
768	* kvm->srcu critical section where acquiring the slots_lock would
769	* lead to deadlock with the synchronize_srcu in
770	* kvm_swap_active_memslots().
771	*/
772	struct mutex slots_arch_lock;
773	struct mm_struct mm; /* userspace tied to this vm /
774	unsigned long nr_memslot_pages;
775	/ The two memslot sets - active and inactive (per address space) /
776	struct kvm_memslots __memslots[KVM_MAX_NR_ADDRESS_SPACES][`2`];
777	/ The current active memslot set for each address space /
778	struct kvm_memslots __rcu *memslots[KVM_MAX_NR_ADDRESS_SPACES];
779	struct xarray vcpu_array;
780	/*
781	* Protected by slots_lock, but can be read outside if an
782	* incorrect answer is acceptable.
783	*/
784	atomic_t nr_memslots_dirty_logging;
785
786	/ Used to wait for completion of MMU notifiers. /
787	spinlock_t mn_invalidate_lock;
788	unsigned long mn_active_invalidate_count;
789	struct rcuwait mn_memslots_update_rcuwait;
790
791	/ For management / invalidation of gfn_to_pfn_caches /
792	spinlock_t gpc_lock;
793	struct list_head gpc_list;
794
795	/*
796	* created_vcpus is protected by kvm->lock, and is incremented
797	* at the beginning of KVM_CREATE_VCPU. online_vcpus is only
798	* incremented after storing the kvm_vcpu pointer in vcpus,
799	* and is accessed atomically.
800	*/
801	atomic_t online_vcpus;
802	int max_vcpus;
803	int created_vcpus;
804	int last_boosted_vcpu;
805	struct list_head vm_list;
806	struct mutex lock;
807	struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
808	#ifdef CONFIG_HAVE_KVM_IRQCHIP
809	struct {
810	spinlock_t lock;
811	struct list_head items;
812	/ resampler_list update side is protected by resampler_lock. /
813	struct list_head resampler_list;
814	struct mutex resampler_lock;
815	} irqfds;
816	#endif
817	struct list_head ioeventfds;
818	struct kvm_vm_stat stat;
819	struct kvm_arch arch;
820	refcount_t users_count;
821	#ifdef CONFIG_KVM_MMIO
822	struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
823	spinlock_t ring_lock;
824	struct list_head coalesced_zones;
825	#endif
826
827	struct mutex irq_lock;
828	#ifdef CONFIG_HAVE_KVM_IRQCHIP
829	/*
830	* Update side is protected by irq_lock.
831	*/
832	struct kvm_irq_routing_table __rcu *irq_routing;
833
834	struct hlist_head irq_ack_notifier_list;
835	#endif
836
837	#ifdef CONFIG_KVM_GENERIC_MMU_NOTIFIER
838	struct mmu_notifier mmu_notifier;
839	unsigned long mmu_invalidate_seq;
840	long mmu_invalidate_in_progress;
841	gfn_t mmu_invalidate_range_start;
842	gfn_t mmu_invalidate_range_end;
843	#endif
844	struct list_head devices;
845	u64 manual_dirty_log_protect;
846	struct dentry *debugfs_dentry;
847	struct kvm_stat_data **debugfs_stat_data;
848	struct srcu_struct srcu;
849	struct srcu_struct irq_srcu;
850	pid_t userspace_pid;
851	bool override_halt_poll_ns;
852	unsigned int max_halt_poll_ns;
853	u32 dirty_ring_size;
854	bool dirty_ring_with_bitmap;
855	bool vm_bugged;
856	bool vm_dead;
857
858	#ifdef CONFIG_HAVE_KVM_PM_NOTIFIER
859	struct notifier_block pm_notifier;
860	#endif
861	#ifdef CONFIG_KVM_GENERIC_MEMORY_ATTRIBUTES
862	/ Protected by slots_locks (for writes) and RCU (for reads) /
863	struct xarray mem_attr_array;
864	#endif
865	char stats_id[KVM_STATS_NAME_SIZE];
866	};
867
868	#define kvm_err(fmt, ...) \
869	pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
870	#define kvm_info(fmt, ...) \
871	pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
872	#define kvm_debug(fmt, ...) \
873	pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
874	#define kvm_debug_ratelimited(fmt, ...) \
875	pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
876	## __VA_ARGS__)
877	#define kvm_pr_unimpl(fmt, ...) \
878	pr_err_ratelimited("kvm [%i]: " fmt, \
879	task_tgid_nr(current), ## __VA_ARGS__)
880
881	/ The guest did something we don't support. /
882	#define vcpu_unimpl(vcpu, fmt, ...) \
883	kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
884	(vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
885
886	#define vcpu_debug(vcpu, fmt, ...) \
887	kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
888	#define vcpu_debug_ratelimited(vcpu, fmt, ...) \
889	kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
890	## __VA_ARGS__)
891	#define vcpu_err(vcpu, fmt, ...) \
892	kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
893
894	static inline void kvm_vm_dead(struct kvm *kvm)
895	{
896	kvm->vm_dead = true;
897	kvm_make_all_cpus_request(kvm, KVM_REQ_VM_DEAD);
898	}
899
900	static inline void kvm_vm_bugged(struct kvm *kvm)
901	{
902	kvm->vm_bugged = true;
903	kvm_vm_dead(kvm);
904	}
905
906
907	#define KVM_BUG(cond, kvm, fmt...) \
908	({ \
909	bool __ret = !!(cond); \
910	\
911	if (WARN_ONCE(__ret && !(kvm)->vm_bugged, fmt)) \
912	kvm_vm_bugged(kvm); \
913	unlikely(__ret); \
914	})
915
916	#define KVM_BUG_ON(cond, kvm) \
917	({ \
918	bool __ret = !!(cond); \
919	\
920	if (WARN_ON_ONCE(__ret && !(kvm)->vm_bugged)) \
921	kvm_vm_bugged(kvm); \
922	unlikely(__ret); \
923	})
924
925	/*
926	* Note, "data corruption" refers to corruption of host kernel data structures,
927	* not guest data. Guest data corruption, suspected or confirmed, that is tied
928	* and contained to a single VM should never BUG() and potentially panic the
929	* host, i.e. use this variant of KVM_BUG() if and only if a KVM data structure
930	* is corrupted and that corruption can have a cascading effect to other parts
931	* of the hosts and/or to other VMs.
932	*/
933	#define KVM_BUG_ON_DATA_CORRUPTION(cond, kvm) \
934	({ \
935	bool __ret = !!(cond); \
936	\
937	if (IS_ENABLED(CONFIG_BUG_ON_DATA_CORRUPTION)) \
938	BUG_ON(__ret); \
939	else if (WARN_ON_ONCE(__ret && !(kvm)->vm_bugged)) \
940	kvm_vm_bugged(kvm); \
941	unlikely(__ret); \
942	})
943
944	static inline void kvm_vcpu_srcu_read_lock(struct kvm_vcpu *vcpu)
945	{
946	#ifdef CONFIG_PROVE_RCU
947	WARN_ONCE(vcpu->srcu_depth++,
948	"KVM: Illegal vCPU srcu_idx LOCK, depth=%d", vcpu->srcu_depth - `1`);
949	#endif
950	vcpu->____srcu_idx = srcu_read_lock(ssp: &vcpu->kvm->srcu);
951	}
952
953	static inline void kvm_vcpu_srcu_read_unlock(struct kvm_vcpu *vcpu)
954	{
955	srcu_read_unlock(ssp: &vcpu->kvm->srcu, idx: vcpu->____srcu_idx);
956
957	#ifdef CONFIG_PROVE_RCU
958	WARN_ONCE(--vcpu->srcu_depth,
959	"KVM: Illegal vCPU srcu_idx UNLOCK, depth=%d", vcpu->srcu_depth);
960	#endif
961	}
962
963	static inline bool kvm_dirty_log_manual_protect_and_init_set(struct kvm *kvm)
964	{
965	return !!(kvm->manual_dirty_log_protect & KVM_DIRTY_LOG_INITIALLY_SET);
966	}
967
968	static inline struct kvm_io_bus kvm_get_bus(struct* kvm kvm, enum* kvm_bus idx)
969	{
970	return srcu_dereference_check(kvm->buses[idx], &kvm->srcu,
971	lockdep_is_held(&kvm->slots_lock) \|\|
972	!refcount_read(&kvm->users_count));
973	}
974
975	static inline struct kvm_vcpu kvm_get_vcpu(struct* kvm kvm, int* i)
976	{
977	int num_vcpus = atomic_read(v: &kvm->online_vcpus);
978
979	/*
980	* Explicitly verify the target vCPU is online, as the anti-speculation
981	* logic only limits the CPU's ability to speculate, e.g. given a "bad"
982	* index, clamping the index to 0 would return vCPU0, not NULL.
983	*/
984	if (i >= num_vcpus)
985	return NULL;
986
987	i = array_index_nospec(i, num_vcpus);
988
989	/ Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. /
990	smp_rmb();
991	return xa_load(&kvm->vcpu_array, index: i);
992	}
993
994	#define kvm_for_each_vcpu(idx, vcpup, kvm) \
995	if (atomic_read(&kvm->online_vcpus)) \
996	xa_for_each_range(&kvm->vcpu_array, idx, vcpup, 0, \
997	(atomic_read(&kvm->online_vcpus) - 1))
998
999	static inline struct kvm_vcpu kvm_get_vcpu_by_id(struct* kvm kvm, int* id)
1000	{
1001	struct kvm_vcpu *vcpu = NULL;
1002	unsigned long i;
1003
1004	if (id < `0`)
1005	return NULL;
1006	if (id < KVM_MAX_VCPUS)
1007	vcpu = kvm_get_vcpu(kvm, i: id);
1008	if (vcpu && vcpu->vcpu_id == id)
1009	return vcpu;
1010	kvm_for_each_vcpu(i, vcpu, kvm)
1011	if (vcpu->vcpu_id == id)
1012	return vcpu;
1013	return NULL;
1014	}
1015
1016	void kvm_destroy_vcpus(struct kvm *kvm);
1017
1018	int kvm_trylock_all_vcpus(struct kvm *kvm);
1019	int kvm_lock_all_vcpus(struct kvm *kvm);
1020	void kvm_unlock_all_vcpus(struct kvm *kvm);
1021
1022	void vcpu_load(struct kvm_vcpu *vcpu);
1023	void vcpu_put(struct kvm_vcpu *vcpu);
1024
1025	#ifdef __KVM_HAVE_IOAPIC
1026	void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
1027	void kvm_arch_post_irq_routing_update(struct kvm *kvm);
1028	#else
1029	static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
1030	{
1031	}
1032	static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
1033	{
1034	}
1035	#endif
1036
1037	#ifdef CONFIG_HAVE_KVM_IRQCHIP
1038	int kvm_irqfd_init(void);
1039	void kvm_irqfd_exit(void);
1040	#else
1041	static inline int kvm_irqfd_init(void)
1042	{
1043	return `0`;
1044	}
1045
1046	static inline void kvm_irqfd_exit(void)
1047	{
1048	}
1049	#endif
1050	int kvm_init(unsigned vcpu_size, unsigned vcpu_align, struct module *module);
1051	void kvm_exit(void);
1052
1053	void kvm_get_kvm(struct kvm *kvm);
1054	bool kvm_get_kvm_safe(struct kvm *kvm);
1055	void kvm_put_kvm(struct kvm *kvm);
1056	bool file_is_kvm(struct file *file);
1057	void kvm_put_kvm_no_destroy(struct kvm *kvm);
1058
1059	static inline struct kvm_memslots __kvm_memslots(struct* kvm kvm, int* as_id)
1060	{
1061	as_id = array_index_nospec(as_id, KVM_MAX_NR_ADDRESS_SPACES);
1062	return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu,
1063	lockdep_is_held(&kvm->slots_lock) \|\|
1064	!refcount_read(&kvm->users_count));
1065	}
1066
1067	static inline struct kvm_memslots kvm_memslots(struct* kvm *kvm)
1068	{
1069	return __kvm_memslots(kvm, as_id: `0`);
1070	}
1071
1072	static inline struct kvm_memslots kvm_vcpu_memslots(struct* kvm_vcpu *vcpu)
1073	{
1074	int as_id = kvm_arch_vcpu_memslots_id(vcpu);
1075
1076	return __kvm_memslots(kvm: vcpu->kvm, as_id);
1077	}
1078
1079	static inline bool kvm_memslots_empty(struct kvm_memslots *slots)
1080	{
1081	return RB_EMPTY_ROOT(&slots->gfn_tree);
1082	}
1083
1084	bool kvm_are_all_memslots_empty(struct kvm *kvm);
1085
1086	#define kvm_for_each_memslot(memslot, bkt, slots) \
1087	hash_for_each(slots->id_hash, bkt, memslot, id_node[slots->node_idx]) \
1088	if (WARN_ON_ONCE(!memslot->npages)) { \
1089	} else
1090
1091	static inline
1092	struct kvm_memory_slot id_to_memslot(struct* kvm_memslots slots, int* id)
1093	{
1094	struct kvm_memory_slot *slot;
1095	int idx = slots->node_idx;
1096
1097	hash_for_each_possible(slots->id_hash, slot, id_node[idx], id) {
1098	if (slot->id == id)
1099	return slot;
1100	}
1101
1102	return NULL;
1103	}
1104
1105	/ Iterator used for walking memslots that overlap a gfn range. /
1106	struct kvm_memslot_iter {
1107	struct kvm_memslots *slots;
1108	struct rb_node *node;
1109	struct kvm_memory_slot *slot;
1110	};
1111
1112	static inline void kvm_memslot_iter_next(struct kvm_memslot_iter *iter)
1113	{
1114	iter->node = rb_next(iter->node);
1115	if (!iter->node)
1116	return;
1117
1118	iter->slot = container_of(iter->node, struct kvm_memory_slot, gfn_node[iter->slots->node_idx]);
1119	}
1120
1121	static inline void kvm_memslot_iter_start(struct kvm_memslot_iter *iter,
1122	struct kvm_memslots *slots,
1123	gfn_t start)
1124	{
1125	int idx = slots->node_idx;
1126	struct rb_node *tmp;
1127	struct kvm_memory_slot *slot;
1128
1129	iter->slots = slots;
1130
1131	/*
1132	* Find the so called "upper bound" of a key - the first node that has
1133	* its key strictly greater than the searched one (the start gfn in our case).
1134	*/
1135	iter->node = NULL;
1136	for (tmp = slots->gfn_tree.rb_node; tmp; ) {
1137	slot = container_of(tmp, struct kvm_memory_slot, gfn_node[idx]);
1138	if (start < slot->base_gfn) {
1139	iter->node = tmp;
1140	tmp = tmp->rb_left;
1141	} else {
1142	tmp = tmp->rb_right;
1143	}
1144	}
1145
1146	/*
1147	* Find the slot with the lowest gfn that can possibly intersect with
1148	* the range, so we'll ideally have slot start <= range start
1149	*/
1150	if (iter->node) {
1151	/*
1152	* A NULL previous node means that the very first slot
1153	* already has a higher start gfn.
1154	* In this case slot start > range start.
1155	*/
1156	tmp = rb_prev(iter->node);
1157	if (tmp)
1158	iter->node = tmp;
1159	} else {
1160	/ a NULL node below means no slots /
1161	iter->node = rb_last(&slots->gfn_tree);
1162	}
1163
1164	if (iter->node) {
1165	iter->slot = container_of(iter->node, struct kvm_memory_slot, gfn_node[idx]);
1166
1167	/*
1168	* It is possible in the slot start < range start case that the
1169	* found slot ends before or at range start (slot end <= range start)
1170	* and so it does not overlap the requested range.
1171	*
1172	* In such non-overlapping case the next slot (if it exists) will
1173	* already have slot start > range start, otherwise the logic above
1174	* would have found it instead of the current slot.
1175	*/
1176	if (iter->slot->base_gfn + iter->slot->npages <= start)
1177	kvm_memslot_iter_next(iter);
1178	}
1179	}
1180
1181	static inline bool kvm_memslot_iter_is_valid(struct kvm_memslot_iter *iter, gfn_t end)
1182	{
1183	if (!iter->node)
1184	return false;
1185
1186	/*
1187	* If this slot starts beyond or at the end of the range so does
1188	* every next one
1189	*/
1190	return iter->slot->base_gfn < end;
1191	}
1192
1193	/ Iterate over each memslot at least partially intersecting [start, end) range /
1194	#define kvm_for_each_memslot_in_gfn_range(iter, slots, start, end) \
1195	for (kvm_memslot_iter_start(iter, slots, start); \
1196	kvm_memslot_iter_is_valid(iter, end); \
1197	kvm_memslot_iter_next(iter))
1198
1199	struct kvm_memory_slot gfn_to_memslot(struct* kvm *kvm, gfn_t gfn);
1200	struct kvm_memslots kvm_vcpu_memslots(struct* kvm_vcpu *vcpu);
1201	struct kvm_memory_slot kvm_vcpu_gfn_to_memslot(struct* kvm_vcpu *vcpu, gfn_t gfn);
1202
1203	/*
1204	* KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
1205	* - create a new memory slot
1206	* - delete an existing memory slot
1207	* - modify an existing memory slot
1208	* -- move it in the guest physical memory space
1209	* -- just change its flags
1210	*
1211	* Since flags can be changed by some of these operations, the following
1212	* differentiation is the best we can do for kvm_set_memory_region():
1213	*/
1214	enum kvm_mr_change {
1215	KVM_MR_CREATE,
1216	KVM_MR_DELETE,
1217	KVM_MR_MOVE,
1218	KVM_MR_FLAGS_ONLY,
1219	};
1220
1221	int kvm_set_internal_memslot(struct kvm *kvm,
1222	const struct kvm_userspace_memory_region2 *mem);
1223	void kvm_arch_free_memslot(struct kvm kvm, struct* kvm_memory_slot *slot);
1224	void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen);
1225	int kvm_arch_prepare_memory_region(struct kvm *kvm,
1226	const struct kvm_memory_slot *old,
1227	struct kvm_memory_slot *new,
1228	enum kvm_mr_change change);
1229	void kvm_arch_commit_memory_region(struct kvm *kvm,
1230	struct kvm_memory_slot *old,
1231	const struct kvm_memory_slot *new,
1232	enum kvm_mr_change change);
1233	/ flush all memory translations /
1234	void kvm_arch_flush_shadow_all(struct kvm *kvm);
1235	/ flush memory translations pointing to 'slot' /
1236	void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
1237	struct kvm_memory_slot *slot);
1238
1239	int kvm_prefetch_pages(struct kvm_memory_slot *slot, gfn_t gfn,
1240	struct page *pages, int* nr_pages);
1241
1242	struct page __gfn_to_page(struct* kvm *kvm, gfn_t gfn, bool write);
1243	static inline struct page gfn_to_page(struct* kvm *kvm, gfn_t gfn)
1244	{
1245	return __gfn_to_page(kvm, gfn, write: true);
1246	}
1247
1248	unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
1249	unsigned long gfn_to_hva_prot(struct kvm kvm, gfn_t gfn, bool writable);
1250	unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
1251	unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
1252	bool *writable);
1253
1254	static inline void kvm_release_page_unused(struct page *page)
1255	{
1256	if (!page)
1257	return;
1258
1259	put_page(page);
1260	}
1261
1262	void kvm_release_page_clean(struct page *page);
1263	void kvm_release_page_dirty(struct page *page);
1264
1265	static inline void kvm_release_faultin_page(struct kvm kvm, struct* page *page,
1266	bool unused, bool dirty)
1267	{
1268	lockdep_assert_once(lockdep_is_held(&kvm->mmu_lock) \|\| unused);
1269
1270	if (!page)
1271	return;
1272
1273	/*
1274	* If the page that KVM got from the primary MMU is writable, and KVM
1275	* installed or reused a SPTE, mark the page/folio dirty. Note, this
1276	* may mark a folio dirty even if KVM created a read-only SPTE, e.g. if
1277	* the GFN is write-protected. Folios can't be safely marked dirty
1278	* outside of mmu_lock as doing so could race with writeback on the
1279	* folio. As a result, KVM can't mark folios dirty in the fast page
1280	* fault handler, and so KVM must (somewhat) speculatively mark the
1281	* folio dirty if KVM could locklessly make the SPTE writable.
1282	*/
1283	if (unused)
1284	kvm_release_page_unused(page);
1285	else if (dirty)
1286	kvm_release_page_dirty(page);
1287	else
1288	kvm_release_page_clean(page);
1289	}
1290
1291	kvm_pfn_t __kvm_faultin_pfn(const struct kvm_memory_slot *slot, gfn_t gfn,
1292	unsigned int foll, bool *writable,
1293	struct page **refcounted_page);
1294
1295	static inline kvm_pfn_t kvm_faultin_pfn(struct kvm_vcpu *vcpu, gfn_t gfn,
1296	bool write, bool *writable,
1297	struct page **refcounted_page)
1298	{
1299	return __kvm_faultin_pfn(slot: kvm_vcpu_gfn_to_memslot(vcpu, gfn), gfn,
1300	foll: write ? FOLL_WRITE : `0`, writable, refcounted_page);
1301	}
1302
1303	int kvm_read_guest_page(struct kvm kvm, gfn_t gfn, void* data, int* offset,
1304	int len);
1305	int kvm_read_guest(struct kvm kvm, gpa_t gpa, void* data, unsigned* long len);
1306	int kvm_read_guest_cached(struct kvm kvm, struct* gfn_to_hva_cache *ghc,
1307	void data, unsigned* long len);
1308	int kvm_read_guest_offset_cached(struct kvm kvm, struct* gfn_to_hva_cache *ghc,
1309	void data, unsigned* int offset,
1310	unsigned long len);
1311	int kvm_write_guest_page(struct kvm kvm, gfn_t gfn, const* void *data,
1312	int offset, int len);
1313	int kvm_write_guest(struct kvm kvm, gpa_t gpa, const* void *data,
1314	unsigned long len);
1315	int kvm_write_guest_cached(struct kvm kvm, struct* gfn_to_hva_cache *ghc,
1316	void data, unsigned* long len);
1317	int kvm_write_guest_offset_cached(struct kvm kvm, struct* gfn_to_hva_cache *ghc,
1318	void data, unsigned* int offset,
1319	unsigned long len);
1320	int kvm_gfn_to_hva_cache_init(struct kvm kvm, struct* gfn_to_hva_cache *ghc,
1321	gpa_t gpa, unsigned long len);
1322
1323	#define __kvm_get_guest(kvm, gfn, offset, v) \
1324	({ \
1325	unsigned long __addr = gfn_to_hva(kvm, gfn); \
1326	typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
1327	int __ret = -EFAULT; \
1328	\
1329	if (!kvm_is_error_hva(__addr)) \
1330	__ret = get_user(v, __uaddr); \
1331	__ret; \
1332	})
1333
1334	#define kvm_get_guest(kvm, gpa, v) \
1335	({ \
1336	gpa_t __gpa = gpa; \
1337	struct kvm *__kvm = kvm; \
1338	\
1339	__kvm_get_guest(__kvm, __gpa >> PAGE_SHIFT, \
1340	offset_in_page(__gpa), v); \
1341	})
1342
1343	#define __kvm_put_guest(kvm, gfn, offset, v) \
1344	({ \
1345	unsigned long __addr = gfn_to_hva(kvm, gfn); \
1346	typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
1347	int __ret = -EFAULT; \
1348	\
1349	if (!kvm_is_error_hva(__addr)) \
1350	__ret = put_user(v, __uaddr); \
1351	if (!__ret) \
1352	mark_page_dirty(kvm, gfn); \
1353	__ret; \
1354	})
1355
1356	#define kvm_put_guest(kvm, gpa, v) \
1357	({ \
1358	gpa_t __gpa = gpa; \
1359	struct kvm *__kvm = kvm; \
1360	\
1361	__kvm_put_guest(__kvm, __gpa >> PAGE_SHIFT, \
1362	offset_in_page(__gpa), v); \
1363	})
1364
1365	int kvm_clear_guest(struct kvm kvm, gpa_t gpa, unsigned* long len);
1366	bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
1367	bool kvm_vcpu_is_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1368	unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn);
1369	void mark_page_dirty_in_slot(struct kvm kvm, const* struct kvm_memory_slot *memslot, gfn_t gfn);
1370	void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
1371
1372	int __kvm_vcpu_map(struct kvm_vcpu vcpu, gpa_t gpa, struct* kvm_host_map *map,
1373	bool writable);
1374	void kvm_vcpu_unmap(struct kvm_vcpu vcpu, struct* kvm_host_map *map);
1375
1376	static inline int kvm_vcpu_map(struct kvm_vcpu *vcpu, gpa_t gpa,
1377	struct kvm_host_map *map)
1378	{
1379	return __kvm_vcpu_map(vcpu, gpa, map, writable: true);
1380	}
1381
1382	static inline int kvm_vcpu_map_readonly(struct kvm_vcpu *vcpu, gpa_t gpa,
1383	struct kvm_host_map *map)
1384	{
1385	return __kvm_vcpu_map(vcpu, gpa, map, writable: false);
1386	}
1387
1388	unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
1389	unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu vcpu, gfn_t gfn, bool writable);
1390	int kvm_vcpu_read_guest_page(struct kvm_vcpu vcpu, gfn_t gfn, void* data, int* offset,
1391	int len);
1392	int kvm_vcpu_read_guest_atomic(struct kvm_vcpu vcpu, gpa_t gpa, void* *data,
1393	unsigned long len);
1394	int kvm_vcpu_read_guest(struct kvm_vcpu vcpu, gpa_t gpa, void* *data,
1395	unsigned long len);
1396	int kvm_vcpu_write_guest_page(struct kvm_vcpu vcpu, gfn_t gfn, const* void *data,
1397	int offset, int len);
1398	int kvm_vcpu_write_guest(struct kvm_vcpu vcpu, gpa_t gpa, const* void *data,
1399	unsigned long len);
1400	void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
1401
1402	/**
1403	* kvm_gpc_init - initialize gfn_to_pfn_cache.
1404	*
1405	* @gpc: struct gfn_to_pfn_cache object.
1406	* @kvm: pointer to kvm instance.
1407	*
1408	* This sets up a gfn_to_pfn_cache by initializing locks and assigning the
1409	* immutable attributes. Note, the cache must be zero-allocated (or zeroed by
1410	* the caller before init).
1411	*/
1412	void kvm_gpc_init(struct gfn_to_pfn_cache gpc, struct* kvm *kvm);
1413
1414	/**
1415	* kvm_gpc_activate - prepare a cached kernel mapping and HPA for a given guest
1416	* physical address.
1417	*
1418	* @gpc: struct gfn_to_pfn_cache object.
1419	* @gpa: guest physical address to map.
1420	* @len: sanity check; the range being access must fit a single page.
1421	*
1422	* @return: 0 for success.
1423	* -EINVAL for a mapping which would cross a page boundary.
1424	* -EFAULT for an untranslatable guest physical address.
1425	*
1426	* This primes a gfn_to_pfn_cache and links it into the @gpc->kvm's list for
1427	* invalidations to be processed. Callers are required to use kvm_gpc_check()
1428	* to ensure that the cache is valid before accessing the target page.
1429	*/
1430	int kvm_gpc_activate(struct gfn_to_pfn_cache gpc, gpa_t gpa, unsigned* long len);
1431
1432	/**
1433	* kvm_gpc_activate_hva - prepare a cached kernel mapping and HPA for a given HVA.
1434	*
1435	* @gpc: struct gfn_to_pfn_cache object.
1436	* @hva: userspace virtual address to map.
1437	* @len: sanity check; the range being access must fit a single page.
1438	*
1439	* @return: 0 for success.
1440	* -EINVAL for a mapping which would cross a page boundary.
1441	* -EFAULT for an untranslatable guest physical address.
1442	*
1443	* The semantics of this function are the same as those of kvm_gpc_activate(). It
1444	* merely bypasses a layer of address translation.
1445	*/
1446	int kvm_gpc_activate_hva(struct gfn_to_pfn_cache gpc, unsigned* long hva, unsigned long len);
1447
1448	/**
1449	* kvm_gpc_check - check validity of a gfn_to_pfn_cache.
1450	*
1451	* @gpc: struct gfn_to_pfn_cache object.
1452	* @len: sanity check; the range being access must fit a single page.
1453	*
1454	* @return: %true if the cache is still valid and the address matches.
1455	* %false if the cache is not valid.
1456	*
1457	* Callers outside IN_GUEST_MODE context should hold a read lock on @gpc->lock
1458	* while calling this function, and then continue to hold the lock until the
1459	* access is complete.
1460	*
1461	* Callers in IN_GUEST_MODE may do so without locking, although they should
1462	* still hold a read lock on kvm->scru for the memslot checks.
1463	*/
1464	bool kvm_gpc_check(struct gfn_to_pfn_cache gpc, unsigned* long len);
1465
1466	/**
1467	* kvm_gpc_refresh - update a previously initialized cache.
1468	*
1469	* @gpc: struct gfn_to_pfn_cache object.
1470	* @len: sanity check; the range being access must fit a single page.
1471	*
1472	* @return: 0 for success.
1473	* -EINVAL for a mapping which would cross a page boundary.
1474	* -EFAULT for an untranslatable guest physical address.
1475	*
1476	* This will attempt to refresh a gfn_to_pfn_cache. Note that a successful
1477	* return from this function does not mean the page can be immediately
1478	* accessed because it may have raced with an invalidation. Callers must
1479	* still lock and check the cache status, as this function does not return
1480	* with the lock still held to permit access.
1481	*/
1482	int kvm_gpc_refresh(struct gfn_to_pfn_cache gpc, unsigned* long len);
1483
1484	/**
1485	* kvm_gpc_deactivate - deactivate and unlink a gfn_to_pfn_cache.
1486	*
1487	* @gpc: struct gfn_to_pfn_cache object.
1488	*
1489	* This removes a cache from the VM's list to be processed on MMU notifier
1490	* invocation.
1491	*/
1492	void kvm_gpc_deactivate(struct gfn_to_pfn_cache *gpc);
1493
1494	static inline bool kvm_gpc_is_gpa_active(struct gfn_to_pfn_cache *gpc)
1495	{
1496	return gpc->active && !kvm_is_error_gpa(gpa: gpc->gpa);
1497	}
1498
1499	static inline bool kvm_gpc_is_hva_active(struct gfn_to_pfn_cache *gpc)
1500	{
1501	return gpc->active && kvm_is_error_gpa(gpa: gpc->gpa);
1502	}
1503
1504	void kvm_sigset_activate(struct kvm_vcpu *vcpu);
1505	void kvm_sigset_deactivate(struct kvm_vcpu *vcpu);
1506
1507	void kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1508	bool kvm_vcpu_block(struct kvm_vcpu *vcpu);
1509	void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
1510	void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
1511	bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
1512
1513	#ifndef CONFIG_S390
1514	void __kvm_vcpu_kick(struct kvm_vcpu *vcpu, bool wait);
1515
1516	static inline void kvm_vcpu_kick(struct kvm_vcpu *vcpu)
1517	{
1518	__kvm_vcpu_kick(vcpu, wait: false);
1519	}
1520	#endif
1521
1522	int kvm_vcpu_yield_to(struct kvm_vcpu *target);
1523	void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool yield_to_kernel_mode);
1524
1525	void kvm_flush_remote_tlbs(struct kvm *kvm);
1526	void kvm_flush_remote_tlbs_range(struct kvm *kvm, gfn_t gfn, u64 nr_pages);
1527	void kvm_flush_remote_tlbs_memslot(struct kvm *kvm,
1528	const struct kvm_memory_slot *memslot);
1529
1530	#ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE
1531	int kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache mc, int* min);
1532	int __kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache mc, int* capacity, int min);
1533	int kvm_mmu_memory_cache_nr_free_objects(struct kvm_mmu_memory_cache *mc);
1534	void kvm_mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc);
1535	void kvm_mmu_memory_cache_alloc(struct* kvm_mmu_memory_cache *mc);
1536	#endif
1537
1538	void kvm_mmu_invalidate_begin(struct kvm *kvm);
1539	void kvm_mmu_invalidate_range_add(struct kvm *kvm, gfn_t start, gfn_t end);
1540	void kvm_mmu_invalidate_end(struct kvm *kvm);
1541	bool kvm_mmu_unmap_gfn_range(struct kvm kvm, struct* kvm_gfn_range *range);
1542
1543	long kvm_arch_dev_ioctl(struct file *filp,
1544	unsigned int ioctl, unsigned long arg);
1545	long kvm_arch_vcpu_ioctl(struct file *filp,
1546	unsigned int ioctl, unsigned long arg);
1547	vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu vcpu, struct* vm_fault *vmf);
1548
1549	int kvm_vm_ioctl_check_extension(struct kvm kvm, long* ext);
1550
1551	void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
1552	struct kvm_memory_slot *slot,
1553	gfn_t gfn_offset,
1554	unsigned long mask);
1555	void kvm_arch_sync_dirty_log(struct kvm kvm, struct* kvm_memory_slot *memslot);
1556
1557	#ifndef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
1558	int kvm_vm_ioctl_get_dirty_log(struct kvm kvm, struct* kvm_dirty_log *log);
1559	int kvm_get_dirty_log(struct kvm kvm, struct* kvm_dirty_log *log,
1560	int is_dirty, struct* kvm_memory_slot **memslot);
1561	#endif
1562
1563	int kvm_vm_ioctl_irq_line(struct kvm kvm, struct* kvm_irq_level *irq_level,
1564	bool line_status);
1565	int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
1566	struct kvm_enable_cap *cap);
1567	int kvm_arch_vm_ioctl(struct file filp, unsigned* int ioctl, unsigned long arg);
1568	long kvm_arch_vm_compat_ioctl(struct file filp, unsigned* int ioctl,
1569	unsigned long arg);
1570
1571	int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu vcpu, struct* kvm_fpu *fpu);
1572	int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu vcpu, struct* kvm_fpu *fpu);
1573
1574	int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1575	struct kvm_translation *tr);
1576
1577	int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu vcpu, struct* kvm_regs *regs);
1578	int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu vcpu, struct* kvm_regs *regs);
1579	int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1580	struct kvm_sregs *sregs);
1581	int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1582	struct kvm_sregs *sregs);
1583	int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
1584	struct kvm_mp_state *mp_state);
1585	int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
1586	struct kvm_mp_state *mp_state);
1587	int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
1588	struct kvm_guest_debug *dbg);
1589	int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu);
1590
1591	void kvm_arch_vcpu_load(struct kvm_vcpu vcpu, int* cpu);
1592	void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
1593	int kvm_arch_vcpu_precreate(struct kvm kvm, unsigned* int id);
1594	int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu);
1595	void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
1596	void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
1597
1598	#ifdef CONFIG_HAVE_KVM_PM_NOTIFIER
1599	int kvm_arch_pm_notifier(struct kvm kvm, unsigned* long state);
1600	#endif
1601
1602	#ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
1603	void kvm_arch_create_vcpu_debugfs(struct kvm_vcpu vcpu, struct* dentry *debugfs_dentry);
1604	#else
1605	static inline void kvm_create_vcpu_debugfs(struct kvm_vcpu *vcpu) {}
1606	#endif
1607
1608	#ifdef CONFIG_KVM_GENERIC_HARDWARE_ENABLING
1609	/*
1610	* kvm_arch_{enable,disable}_virtualization() are called on one CPU, under
1611	* kvm_usage_lock, immediately after/before 0=>1 and 1=>0 transitions of
1612	* kvm_usage_count, i.e. at the beginning of the generic hardware enabling
1613	* sequence, and at the end of the generic hardware disabling sequence.
1614	*/
1615	void kvm_arch_enable_virtualization(void);
1616	void kvm_arch_disable_virtualization(void);
1617	/*
1618	* kvm_arch_{enable,disable}_virtualization_cpu() are called on "every" CPU to
1619	* do the actual twiddling of hardware bits. The hooks are called on all
1620	* online CPUs when KVM enables/disabled virtualization, and on a single CPU
1621	* when that CPU is onlined/offlined (including for Resume/Suspend).
1622	*/
1623	int kvm_arch_enable_virtualization_cpu(void);
1624	void kvm_arch_disable_virtualization_cpu(void);
1625	#endif
1626	bool kvm_vcpu_has_events(struct kvm_vcpu *vcpu);
1627	int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
1628	bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
1629	int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
1630	bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu);
1631	bool kvm_arch_dy_has_pending_interrupt(struct kvm_vcpu *vcpu);
1632	bool kvm_arch_vcpu_preempted_in_kernel(struct kvm_vcpu *vcpu);
1633	void kvm_arch_pre_destroy_vm(struct kvm *kvm);
1634	void kvm_arch_create_vm_debugfs(struct kvm *kvm);
1635
1636	#ifndef __KVM_HAVE_ARCH_VM_ALLOC
1637	/*
1638	* All architectures that want to use vzalloc currently also
1639	* need their own kvm_arch_alloc_vm implementation.
1640	*/
1641	static inline struct kvm kvm_arch_alloc_vm(void*)
1642	{
1643	return kzalloc(sizeof(struct kvm), GFP_KERNEL_ACCOUNT);
1644	}
1645	#endif
1646
1647	static inline void __kvm_arch_free_vm(struct kvm *kvm)
1648	{
1649	kvfree(addr: kvm);
1650	}
1651
1652	#ifndef __KVM_HAVE_ARCH_VM_FREE
1653	static inline void kvm_arch_free_vm(struct kvm *kvm)
1654	{
1655	__kvm_arch_free_vm(kvm);
1656	}
1657	#endif
1658
1659	#ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS
1660	static inline int kvm_arch_flush_remote_tlbs(struct kvm *kvm)
1661	{
1662	return -ENOTSUPP;
1663	}
1664	#else
1665	int kvm_arch_flush_remote_tlbs(struct kvm *kvm);
1666	#endif
1667
1668	#ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS_RANGE
1669	static inline int kvm_arch_flush_remote_tlbs_range(struct kvm *kvm,
1670	gfn_t gfn, u64 nr_pages)
1671	{
1672	return -EOPNOTSUPP;
1673	}
1674	#else
1675	int kvm_arch_flush_remote_tlbs_range(struct kvm *kvm, gfn_t gfn, u64 nr_pages);
1676	#endif
1677
1678	#ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
1679	void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
1680	void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
1681	bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
1682	#else
1683	static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
1684	{
1685	}
1686
1687	static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
1688	{
1689	}
1690
1691	static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
1692	{
1693	return false;
1694	}
1695	#endif
1696	#ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
1697	void kvm_arch_start_assignment(struct kvm *kvm);
1698	void kvm_arch_end_assignment(struct kvm *kvm);
1699	bool kvm_arch_has_assigned_device(struct kvm *kvm);
1700	#else
1701	static inline void kvm_arch_start_assignment(struct kvm *kvm)
1702	{
1703	}
1704
1705	static inline void kvm_arch_end_assignment(struct kvm *kvm)
1706	{
1707	}
1708
1709	static __always_inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
1710	{
1711	return false;
1712	}
1713	#endif
1714
1715	static inline struct rcuwait kvm_arch_vcpu_get_wait(struct* kvm_vcpu *vcpu)
1716	{
1717	#ifdef __KVM_HAVE_ARCH_WQP
1718	return vcpu->arch.waitp;
1719	#else
1720	return &vcpu->wait;
1721	#endif
1722	}
1723
1724	/*
1725	* Wake a vCPU if necessary, but don't do any stats/metadata updates. Returns
1726	* true if the vCPU was blocking and was awakened, false otherwise.
1727	*/
1728	static inline bool __kvm_vcpu_wake_up(struct kvm_vcpu *vcpu)
1729	{
1730	return !!rcuwait_wake_up(w: kvm_arch_vcpu_get_wait(vcpu));
1731	}
1732
1733	static inline bool kvm_vcpu_is_blocking(struct kvm_vcpu *vcpu)
1734	{
1735	return rcuwait_active(w: kvm_arch_vcpu_get_wait(vcpu));
1736	}
1737
1738	#ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
1739	/*
1740	* returns true if the virtual interrupt controller is initialized and
1741	* ready to accept virtual IRQ. On some architectures the virtual interrupt
1742	* controller is dynamically instantiated and this is not always true.
1743	*/
1744	bool kvm_arch_intc_initialized(struct kvm *kvm);
1745	#else
1746	static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
1747	{
1748	return true;
1749	}
1750	#endif
1751
1752	#ifdef CONFIG_GUEST_PERF_EVENTS
1753	unsigned long kvm_arch_vcpu_get_ip(struct kvm_vcpu *vcpu);
1754
1755	void kvm_register_perf_callbacks(unsigned int (pt_intr_handler)(void*));
1756	void kvm_unregister_perf_callbacks(void);
1757	#else
1758	static inline void kvm_register_perf_callbacks(void *ign) {}
1759	static inline void kvm_unregister_perf_callbacks(void) {}
1760	#endif /* CONFIG_GUEST_PERF_EVENTS */
1761
1762	int kvm_arch_init_vm(struct kvm kvm, unsigned* long type);
1763	void kvm_arch_destroy_vm(struct kvm *kvm);
1764
1765	int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
1766
1767	struct kvm_irq_ack_notifier {
1768	struct hlist_node link;
1769	unsigned gsi;
1770	void (irq_acked)(struct* kvm_irq_ack_notifier *kian);
1771	};
1772
1773	int kvm_irq_map_gsi(struct kvm *kvm,
1774	struct kvm_kernel_irq_routing_entry entries, int* gsi);
1775	int kvm_irq_map_chip_pin(struct kvm kvm, unsigned* irqchip, unsigned pin);
1776
1777	int kvm_set_irq(struct kvm kvm, int* irq_source_id, u32 irq, int level,
1778	bool line_status);
1779	int kvm_set_msi(struct kvm_kernel_irq_routing_entry irq_entry, struct* kvm *kvm,
1780	int irq_source_id, int level, bool line_status);
1781	int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
1782	struct kvm kvm, int* irq_source_id,
1783	int level, bool line_status);
1784	bool kvm_irq_has_notifier(struct kvm kvm, unsigned* irqchip, unsigned pin);
1785	void kvm_notify_acked_gsi(struct kvm kvm, int* gsi);
1786	void kvm_notify_acked_irq(struct kvm kvm, unsigned* irqchip, unsigned pin);
1787	void kvm_register_irq_ack_notifier(struct kvm *kvm,
1788	struct kvm_irq_ack_notifier *kian);
1789	void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
1790	struct kvm_irq_ack_notifier *kian);
1791	int kvm_request_irq_source_id(struct kvm *kvm);
1792	void kvm_free_irq_source_id(struct kvm kvm, int* irq_source_id);
1793	bool kvm_arch_irqfd_allowed(struct kvm kvm, struct* kvm_irqfd *args);
1794
1795	/*
1796	* Returns a pointer to the memslot if it contains gfn.
1797	* Otherwise returns NULL.
1798	*/
1799	static inline struct kvm_memory_slot *
1800	try_get_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1801	{
1802	if (!slot)
1803	return NULL;
1804
1805	if (gfn >= slot->base_gfn && gfn < slot->base_gfn + slot->npages)
1806	return slot;
1807	else
1808	return NULL;
1809	}
1810
1811	/*
1812	* Returns a pointer to the memslot that contains gfn. Otherwise returns NULL.
1813	*
1814	* With "approx" set returns the memslot also when the address falls
1815	* in a hole. In that case one of the memslots bordering the hole is
1816	* returned.
1817	*/
1818	static inline struct kvm_memory_slot *
1819	search_memslots(struct kvm_memslots *slots, gfn_t gfn, bool approx)
1820	{
1821	struct kvm_memory_slot *slot;
1822	struct rb_node *node;
1823	int idx = slots->node_idx;
1824
1825	slot = NULL;
1826	for (node = slots->gfn_tree.rb_node; node; ) {
1827	slot = container_of(node, struct kvm_memory_slot, gfn_node[idx]);
1828	if (gfn >= slot->base_gfn) {
1829	if (gfn < slot->base_gfn + slot->npages)
1830	return slot;
1831	node = node->rb_right;
1832	} else
1833	node = node->rb_left;
1834	}
1835
1836	return approx ? slot : NULL;
1837	}
1838
1839	static inline struct kvm_memory_slot *
1840	____gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn, bool approx)
1841	{
1842	struct kvm_memory_slot *slot;
1843
1844	slot = (struct kvm_memory_slot *)atomic_long_read(v: &slots->last_used_slot);
1845	slot = try_get_memslot(slot, gfn);
1846	if (slot)
1847	return slot;
1848
1849	slot = search_memslots(slots, gfn, approx);
1850	if (slot) {
1851	atomic_long_set(v: &slots->last_used_slot, i: (unsigned long)slot);
1852	return slot;
1853	}
1854
1855	return NULL;
1856	}
1857
1858	/*
1859	* __gfn_to_memslot() and its descendants are here to allow arch code to inline
1860	* the lookups in hot paths. gfn_to_memslot() itself isn't here as an inline
1861	* because that would bloat other code too much.
1862	*/
1863	static inline struct kvm_memory_slot *
1864	__gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
1865	{
1866	return ____gfn_to_memslot(slots, gfn, approx: false);
1867	}
1868
1869	static inline unsigned long
1870	__gfn_to_hva_memslot(const struct kvm_memory_slot *slot, gfn_t gfn)
1871	{
1872	/*
1873	* The index was checked originally in search_memslots. To avoid
1874	* that a malicious guest builds a Spectre gadget out of e.g. page
1875	* table walks, do not let the processor speculate loads outside
1876	* the guest's registered memslots.
1877	*/
1878	unsigned long offset = gfn - slot->base_gfn;
1879	offset = array_index_nospec(offset, slot->npages);
1880	return slot->userspace_addr + offset * PAGE_SIZE;
1881	}
1882
1883	static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
1884	{
1885	return gfn_to_memslot(kvm, gfn)->id;
1886	}
1887
1888	static inline gfn_t
1889	hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
1890	{
1891	gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
1892
1893	return slot->base_gfn + gfn_offset;
1894	}
1895
1896	static inline gpa_t gfn_to_gpa(gfn_t gfn)
1897	{
1898	return (gpa_t)gfn << PAGE_SHIFT;
1899	}
1900
1901	static inline gfn_t gpa_to_gfn(gpa_t gpa)
1902	{
1903	return (gfn_t)(gpa >> PAGE_SHIFT);
1904	}
1905
1906	static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
1907	{
1908	return (hpa_t)pfn << PAGE_SHIFT;
1909	}
1910
1911	static inline bool kvm_is_gpa_in_memslot(struct kvm *kvm, gpa_t gpa)
1912	{
1913	unsigned long hva = gfn_to_hva(kvm, gfn: gpa_to_gfn(gpa));
1914
1915	return !kvm_is_error_hva(addr: hva);
1916	}
1917
1918	static inline void kvm_gpc_mark_dirty_in_slot(struct gfn_to_pfn_cache *gpc)
1919	{
1920	lockdep_assert_held(&gpc->lock);
1921
1922	if (!gpc->memslot)
1923	return;
1924
1925	mark_page_dirty_in_slot(kvm: gpc->kvm, memslot: gpc->memslot, gfn: gpa_to_gfn(gpa: gpc->gpa));
1926	}
1927
1928	enum kvm_stat_kind {
1929	KVM_STAT_VM,
1930	KVM_STAT_VCPU,
1931	};
1932
1933	struct kvm_stat_data {
1934	struct kvm *kvm;
1935	const struct _kvm_stats_desc *desc;
1936	enum kvm_stat_kind kind;
1937	};
1938
1939	struct _kvm_stats_desc {
1940	struct kvm_stats_desc desc;
1941	char name[KVM_STATS_NAME_SIZE];
1942	};
1943
1944	#define STATS_DESC_COMMON(type, unit, base, exp, sz, bsz) \
1945	.flags = type \| unit \| base \| \
1946	BUILD_BUG_ON_ZERO(type & ~KVM_STATS_TYPE_MASK) \| \
1947	BUILD_BUG_ON_ZERO(unit & ~KVM_STATS_UNIT_MASK) \| \
1948	BUILD_BUG_ON_ZERO(base & ~KVM_STATS_BASE_MASK), \
1949	.exponent = exp, \
1950	.size = sz, \
1951	.bucket_size = bsz
1952
1953	#define VM_GENERIC_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1954	{ \
1955	{ \
1956	STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1957	.offset = offsetof(struct kvm_vm_stat, generic.stat) \
1958	}, \
1959	.name = #stat, \
1960	}
1961	#define VCPU_GENERIC_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1962	{ \
1963	{ \
1964	STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1965	.offset = offsetof(struct kvm_vcpu_stat, generic.stat) \
1966	}, \
1967	.name = #stat, \
1968	}
1969	#define VM_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1970	{ \
1971	{ \
1972	STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1973	.offset = offsetof(struct kvm_vm_stat, stat) \
1974	}, \
1975	.name = #stat, \
1976	}
1977	#define VCPU_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1978	{ \
1979	{ \
1980	STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1981	.offset = offsetof(struct kvm_vcpu_stat, stat) \
1982	}, \
1983	.name = #stat, \
1984	}
1985	/ SCOPE: VM, VM_GENERIC, VCPU, VCPU_GENERIC /
1986	#define STATS_DESC(SCOPE, stat, type, unit, base, exp, sz, bsz) \
1987	SCOPE##_STATS_DESC(stat, type, unit, base, exp, sz, bsz)
1988
1989	#define STATS_DESC_CUMULATIVE(SCOPE, name, unit, base, exponent) \
1990	STATS_DESC(SCOPE, name, KVM_STATS_TYPE_CUMULATIVE, \
1991	unit, base, exponent, 1, 0)
1992	#define STATS_DESC_INSTANT(SCOPE, name, unit, base, exponent) \
1993	STATS_DESC(SCOPE, name, KVM_STATS_TYPE_INSTANT, \
1994	unit, base, exponent, 1, 0)
1995	#define STATS_DESC_PEAK(SCOPE, name, unit, base, exponent) \
1996	STATS_DESC(SCOPE, name, KVM_STATS_TYPE_PEAK, \
1997	unit, base, exponent, 1, 0)
1998	#define STATS_DESC_LINEAR_HIST(SCOPE, name, unit, base, exponent, sz, bsz) \
1999	STATS_DESC(SCOPE, name, KVM_STATS_TYPE_LINEAR_HIST, \
2000	unit, base, exponent, sz, bsz)
2001	#define STATS_DESC_LOG_HIST(SCOPE, name, unit, base, exponent, sz) \
2002	STATS_DESC(SCOPE, name, KVM_STATS_TYPE_LOG_HIST, \
2003	unit, base, exponent, sz, 0)
2004
2005	/ Cumulative counter, read/write /
2006	#define STATS_DESC_COUNTER(SCOPE, name) \
2007	STATS_DESC_CUMULATIVE(SCOPE, name, KVM_STATS_UNIT_NONE, \
2008	KVM_STATS_BASE_POW10, 0)
2009	/ Instantaneous counter, read only /
2010	#define STATS_DESC_ICOUNTER(SCOPE, name) \
2011	STATS_DESC_INSTANT(SCOPE, name, KVM_STATS_UNIT_NONE, \
2012	KVM_STATS_BASE_POW10, 0)
2013	/ Peak counter, read/write /
2014	#define STATS_DESC_PCOUNTER(SCOPE, name) \
2015	STATS_DESC_PEAK(SCOPE, name, KVM_STATS_UNIT_NONE, \
2016	KVM_STATS_BASE_POW10, 0)
2017
2018	/ Instantaneous boolean value, read only /
2019	#define STATS_DESC_IBOOLEAN(SCOPE, name) \
2020	STATS_DESC_INSTANT(SCOPE, name, KVM_STATS_UNIT_BOOLEAN, \
2021	KVM_STATS_BASE_POW10, 0)
2022	/ Peak (sticky) boolean value, read/write /
2023	#define STATS_DESC_PBOOLEAN(SCOPE, name) \
2024	STATS_DESC_PEAK(SCOPE, name, KVM_STATS_UNIT_BOOLEAN, \
2025	KVM_STATS_BASE_POW10, 0)
2026
2027	/ Cumulative time in nanosecond /
2028	#define STATS_DESC_TIME_NSEC(SCOPE, name) \
2029	STATS_DESC_CUMULATIVE(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
2030	KVM_STATS_BASE_POW10, -9)
2031	/ Linear histogram for time in nanosecond /
2032	#define STATS_DESC_LINHIST_TIME_NSEC(SCOPE, name, sz, bsz) \
2033	STATS_DESC_LINEAR_HIST(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
2034	KVM_STATS_BASE_POW10, -9, sz, bsz)
2035	/ Logarithmic histogram for time in nanosecond /
2036	#define STATS_DESC_LOGHIST_TIME_NSEC(SCOPE, name, sz) \
2037	STATS_DESC_LOG_HIST(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
2038	KVM_STATS_BASE_POW10, -9, sz)
2039
2040	#define KVM_GENERIC_VM_STATS() \
2041	STATS_DESC_COUNTER(VM_GENERIC, remote_tlb_flush), \
2042	STATS_DESC_COUNTER(VM_GENERIC, remote_tlb_flush_requests)
2043
2044	#define KVM_GENERIC_VCPU_STATS() \
2045	STATS_DESC_COUNTER(VCPU_GENERIC, halt_successful_poll), \
2046	STATS_DESC_COUNTER(VCPU_GENERIC, halt_attempted_poll), \
2047	STATS_DESC_COUNTER(VCPU_GENERIC, halt_poll_invalid), \
2048	STATS_DESC_COUNTER(VCPU_GENERIC, halt_wakeup), \
2049	STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_poll_success_ns), \
2050	STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_poll_fail_ns), \
2051	STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_wait_ns), \
2052	STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_poll_success_hist, \
2053	HALT_POLL_HIST_COUNT), \
2054	STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_poll_fail_hist, \
2055	HALT_POLL_HIST_COUNT), \
2056	STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_wait_hist, \
2057	HALT_POLL_HIST_COUNT), \
2058	STATS_DESC_IBOOLEAN(VCPU_GENERIC, blocking)
2059
2060	ssize_t kvm_stats_read(char id, const* struct kvm_stats_header *header,
2061	const struct _kvm_stats_desc *desc,
2062	void *stats, size_t size_stats,
2063	char __user user_buffer, size_t size, loff_t offset);
2064
2065	/**
2066	* kvm_stats_linear_hist_update() - Update bucket value for linear histogram
2067	* statistics data.
2068	*
2069	* @data: start address of the stats data
2070	* @size: the number of bucket of the stats data
2071	* @value: the new value used to update the linear histogram's bucket
2072	* @bucket_size: the size (width) of a bucket
2073	*/
2074	static inline void kvm_stats_linear_hist_update(u64 *data, size_t size,
2075	u64 value, size_t bucket_size)
2076	{
2077	size_t index = div64_u64(dividend: value, divisor: bucket_size);
2078
2079	index = min(index, size - `1`);
2080	++data[index];
2081	}
2082
2083	/**
2084	* kvm_stats_log_hist_update() - Update bucket value for logarithmic histogram
2085	* statistics data.
2086	*
2087	* @data: start address of the stats data
2088	* @size: the number of bucket of the stats data
2089	* @value: the new value used to update the logarithmic histogram's bucket
2090	*/
2091	static inline void kvm_stats_log_hist_update(u64 *data, size_t size, u64 value)
2092	{
2093	size_t index = fls64(x: value);
2094
2095	index = min(index, size - `1`);
2096	++data[index];
2097	}
2098
2099	#define KVM_STATS_LINEAR_HIST_UPDATE(array, value, bsize) \
2100	kvm_stats_linear_hist_update(array, ARRAY_SIZE(array), value, bsize)
2101	#define KVM_STATS_LOG_HIST_UPDATE(array, value) \
2102	kvm_stats_log_hist_update(array, ARRAY_SIZE(array), value)
2103
2104
2105	extern const struct kvm_stats_header kvm_vm_stats_header;
2106	extern const struct _kvm_stats_desc kvm_vm_stats_desc[];
2107	extern const struct kvm_stats_header kvm_vcpu_stats_header;
2108	extern const struct _kvm_stats_desc kvm_vcpu_stats_desc[];
2109
2110	#ifdef CONFIG_KVM_GENERIC_MMU_NOTIFIER
2111	static inline int mmu_invalidate_retry(struct kvm kvm, unsigned* long mmu_seq)
2112	{
2113	if (unlikely(kvm->mmu_invalidate_in_progress))
2114	return `1`;
2115	/*
2116	* Ensure the read of mmu_invalidate_in_progress happens before
2117	* the read of mmu_invalidate_seq. This interacts with the
2118	* smp_wmb() in mmu_notifier_invalidate_range_end to make sure
2119	* that the caller either sees the old (non-zero) value of
2120	* mmu_invalidate_in_progress or the new (incremented) value of
2121	* mmu_invalidate_seq.
2122	*
2123	* PowerPC Book3s HV KVM calls this under a per-page lock rather
2124	* than under kvm->mmu_lock, for scalability, so can't rely on
2125	* kvm->mmu_lock to keep things ordered.
2126	*/
2127	smp_rmb();
2128	if (kvm->mmu_invalidate_seq != mmu_seq)
2129	return `1`;
2130	return `0`;
2131	}
2132
2133	static inline int mmu_invalidate_retry_gfn(struct kvm *kvm,
2134	unsigned long mmu_seq,
2135	gfn_t gfn)
2136	{
2137	lockdep_assert_held(&kvm->mmu_lock);
2138	/*
2139	* If mmu_invalidate_in_progress is non-zero, then the range maintained
2140	* by kvm_mmu_notifier_invalidate_range_start contains all addresses
2141	* that might be being invalidated. Note that it may include some false
2142	* positives, due to shortcuts when handing concurrent invalidations.
2143	*/
2144	if (unlikely(kvm->mmu_invalidate_in_progress)) {
2145	/*
2146	* Dropping mmu_lock after bumping mmu_invalidate_in_progress
2147	* but before updating the range is a KVM bug.
2148	*/
2149	if (WARN_ON_ONCE(kvm->mmu_invalidate_range_start == INVALID_GPA \|\|
2150	kvm->mmu_invalidate_range_end == INVALID_GPA))
2151	return `1`;
2152
2153	if (gfn >= kvm->mmu_invalidate_range_start &&
2154	gfn < kvm->mmu_invalidate_range_end)
2155	return `1`;
2156	}
2157
2158	if (kvm->mmu_invalidate_seq != mmu_seq)
2159	return `1`;
2160	return `0`;
2161	}
2162
2163	/*
2164	* This lockless version of the range-based retry check must be paired with a
2165	* call to the locked version after acquiring mmu_lock, i.e. this is safe to
2166	* use only as a pre-check to avoid contending mmu_lock. This version will
2167	* get false negatives and false positives.
2168	*/
2169	static inline bool mmu_invalidate_retry_gfn_unsafe(struct kvm *kvm,
2170	unsigned long mmu_seq,
2171	gfn_t gfn)
2172	{
2173	/*
2174	* Use READ_ONCE() to ensure the in-progress flag and sequence counter
2175	* are always read from memory, e.g. so that checking for retry in a
2176	* loop won't result in an infinite retry loop. Don't force loads for
2177	* start+end, as the key to avoiding infinite retry loops is observing
2178	* the 1=>0 transition of in-progress, i.e. getting false negatives
2179	* due to stale start+end values is acceptable.
2180	*/
2181	if (unlikely(READ_ONCE(kvm->mmu_invalidate_in_progress)) &&
2182	gfn >= kvm->mmu_invalidate_range_start &&
2183	gfn < kvm->mmu_invalidate_range_end)
2184	return true;
2185
2186	return READ_ONCE(kvm->mmu_invalidate_seq) != mmu_seq;
2187	}
2188	#endif
2189
2190	#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
2191
2192	#define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
2193
2194	bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
2195	int kvm_set_irq_routing(struct kvm *kvm,
2196	const struct kvm_irq_routing_entry *entries,
2197	unsigned nr,
2198	unsigned flags);
2199	int kvm_init_irq_routing(struct kvm *kvm);
2200	int kvm_set_routing_entry(struct kvm *kvm,
2201	struct kvm_kernel_irq_routing_entry *e,
2202	const struct kvm_irq_routing_entry *ue);
2203	void kvm_free_irq_routing(struct kvm *kvm);
2204
2205	#else
2206
2207	static inline void kvm_free_irq_routing(struct kvm *kvm) {}
2208
2209	static inline int kvm_init_irq_routing(struct kvm *kvm)
2210	{
2211	return `0`;
2212	}
2213
2214	#endif
2215
2216	int kvm_send_userspace_msi(struct kvm kvm, struct* kvm_msi *msi);
2217
2218	void kvm_eventfd_init(struct kvm *kvm);
2219	int kvm_ioeventfd(struct kvm kvm, struct* kvm_ioeventfd *args);
2220
2221	#ifdef CONFIG_HAVE_KVM_IRQCHIP
2222	int kvm_irqfd(struct kvm kvm, struct* kvm_irqfd *args);
2223	void kvm_irqfd_release(struct kvm *kvm);
2224	bool kvm_notify_irqfd_resampler(struct kvm *kvm,
2225	unsigned int irqchip,
2226	unsigned int pin);
2227	void kvm_irq_routing_update(struct kvm *);
2228	#else
2229	static inline int kvm_irqfd(struct kvm kvm, struct* kvm_irqfd *args)
2230	{
2231	return -EINVAL;
2232	}
2233
2234	static inline void kvm_irqfd_release(struct kvm *kvm) {}
2235
2236	static inline bool kvm_notify_irqfd_resampler(struct kvm *kvm,
2237	unsigned int irqchip,
2238	unsigned int pin)
2239	{
2240	return false;
2241	}
2242	#endif /* CONFIG_HAVE_KVM_IRQCHIP */
2243
2244	void kvm_arch_irq_routing_update(struct kvm *kvm);
2245
2246	static inline void __kvm_make_request(int req, struct kvm_vcpu *vcpu)
2247	{
2248	/*
2249	* Ensure the rest of the request is published to kvm_check_request's
2250	* caller. Paired with the smp_mb__after_atomic in kvm_check_request.
2251	*/
2252	smp_wmb();
2253	set_bit(nr: req & KVM_REQUEST_MASK, addr: (void *)&vcpu->requests);
2254	}
2255
2256	static __always_inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
2257	{
2258	/*
2259	* Request that don't require vCPU action should never be logged in
2260	* vcpu->requests. The vCPU won't clear the request, so it will stay
2261	* logged indefinitely and prevent the vCPU from entering the guest.
2262	*/
2263	BUILD_BUG_ON(!__builtin_constant_p(req) \|\|
2264	(req & KVM_REQUEST_NO_ACTION));
2265
2266	__kvm_make_request(req, vcpu);
2267	}
2268
2269	#ifndef CONFIG_S390
2270	static inline void kvm_make_request_and_kick(int req, struct kvm_vcpu *vcpu)
2271	{
2272	kvm_make_request(req, vcpu);
2273	__kvm_vcpu_kick(vcpu, wait: req & KVM_REQUEST_WAIT);
2274	}
2275	#endif
2276
2277	static inline bool kvm_request_pending(struct kvm_vcpu *vcpu)
2278	{
2279	return READ_ONCE(vcpu->requests);
2280	}
2281
2282	static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)
2283	{
2284	return test_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
2285	}
2286
2287	static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu)
2288	{
2289	clear_bit(nr: req & KVM_REQUEST_MASK, addr: (void *)&vcpu->requests);
2290	}
2291
2292	static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
2293	{
2294	if (kvm_test_request(req, vcpu)) {
2295	kvm_clear_request(req, vcpu);
2296
2297	/*
2298	* Ensure the rest of the request is visible to kvm_check_request's
2299	* caller. Paired with the smp_wmb in kvm_make_request.
2300	*/
2301	smp_mb__after_atomic();
2302	return true;
2303	} else {
2304	return false;
2305	}
2306	}
2307
2308	#ifdef CONFIG_KVM_GENERIC_HARDWARE_ENABLING
2309	extern bool enable_virt_at_load;
2310	extern bool kvm_rebooting;
2311	#endif
2312
2313	extern unsigned int halt_poll_ns;
2314	extern unsigned int halt_poll_ns_grow;
2315	extern unsigned int halt_poll_ns_grow_start;
2316	extern unsigned int halt_poll_ns_shrink;
2317
2318	struct kvm_device {
2319	const struct kvm_device_ops *ops;
2320	struct kvm *kvm;
2321	void *private;
2322	struct list_head vm_node;
2323	};
2324
2325	/ create, destroy, and name are mandatory /
2326	struct kvm_device_ops {
2327	const char *name;
2328
2329	/*
2330	* create is called holding kvm->lock and any operations not suitable
2331	* to do while holding the lock should be deferred to init (see
2332	* below).
2333	*/
2334	int (create)(struct* kvm_device *dev, u32 type);
2335
2336	/*
2337	* init is called after create if create is successful and is called
2338	* outside of holding kvm->lock.
2339	*/
2340	void (init)(struct* kvm_device *dev);
2341
2342	/*
2343	* Destroy is responsible for freeing dev.
2344	*
2345	* Destroy may be called before or after destructors are called
2346	* on emulated I/O regions, depending on whether a reference is
2347	* held by a vcpu or other kvm component that gets destroyed
2348	* after the emulated I/O.
2349	*/
2350	void (destroy)(struct* kvm_device *dev);
2351
2352	/*
2353	* Release is an alternative method to free the device. It is
2354	* called when the device file descriptor is closed. Once
2355	* release is called, the destroy method will not be called
2356	* anymore as the device is removed from the device list of
2357	* the VM. kvm->lock is held.
2358	*/
2359	void (release)(struct* kvm_device *dev);
2360
2361	int (set_attr)(struct* kvm_device dev, struct* kvm_device_attr *attr);
2362	int (get_attr)(struct* kvm_device dev, struct* kvm_device_attr *attr);
2363	int (has_attr)(struct* kvm_device dev, struct* kvm_device_attr *attr);
2364	long (ioctl)(struct* kvm_device dev, unsigned* int ioctl,
2365	unsigned long arg);
2366	int (mmap)(struct* kvm_device dev, struct* vm_area_struct *vma);
2367	};
2368
2369	struct kvm_device kvm_device_from_filp(struct* file *filp);
2370	int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type);
2371	void kvm_unregister_device_ops(u32 type);
2372
2373	extern struct kvm_device_ops kvm_mpic_ops;
2374	extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
2375	extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
2376
2377	#ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
2378
2379	static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
2380	{
2381	vcpu->spin_loop.in_spin_loop = val;
2382	}
2383	static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
2384	{
2385	vcpu->spin_loop.dy_eligible = val;
2386	}
2387
2388	#else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
2389
2390	static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
2391	{
2392	}
2393
2394	static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
2395	{
2396	}
2397	#endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
2398
2399	static inline bool kvm_is_visible_memslot(struct kvm_memory_slot *memslot)
2400	{
2401	return (memslot && memslot->id < KVM_USER_MEM_SLOTS &&
2402	!(memslot->flags & KVM_MEMSLOT_INVALID));
2403	}
2404
2405	struct kvm_vcpu kvm_get_running_vcpu(void*);
2406	struct kvm_vcpu * __percpu kvm_get_running_vcpus(void*);
2407
2408	#if IS_ENABLED(CONFIG_HAVE_KVM_IRQ_BYPASS)
2409	bool kvm_arch_has_irq_bypass(void);
2410	int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
2411	struct irq_bypass_producer *);
2412	void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
2413	struct irq_bypass_producer *);
2414	void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
2415	void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
2416	int kvm_arch_update_irqfd_routing(struct kvm kvm, unsigned* int host_irq,
2417	uint32_t guest_irq, bool set);
2418	bool kvm_arch_irqfd_route_changed(struct kvm_kernel_irq_routing_entry *,
2419	struct kvm_kernel_irq_routing_entry *);
2420	#endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
2421
2422	#ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
2423	/ If we wakeup during the poll time, was it a sucessful poll? /
2424	static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
2425	{
2426	return vcpu->valid_wakeup;
2427	}
2428
2429	#else
2430	static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
2431	{
2432	return true;
2433	}
2434	#endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
2435
2436	#ifdef CONFIG_HAVE_KVM_NO_POLL
2437	/ Callback that tells if we must not poll /
2438	bool kvm_arch_no_poll(struct kvm_vcpu *vcpu);
2439	#else
2440	static inline bool kvm_arch_no_poll(struct kvm_vcpu *vcpu)
2441	{
2442	return false;
2443	}
2444	#endif /* CONFIG_HAVE_KVM_NO_POLL */
2445
2446	#ifdef CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL
2447	long kvm_arch_vcpu_async_ioctl(struct file *filp,
2448	unsigned int ioctl, unsigned long arg);
2449	#else
2450	static inline long kvm_arch_vcpu_async_ioctl(struct file *filp,
2451	unsigned int ioctl,
2452	unsigned long arg)
2453	{
2454	return -ENOIOCTLCMD;
2455	}
2456	#endif /* CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL */
2457
2458	void kvm_arch_guest_memory_reclaimed(struct kvm *kvm);
2459
2460	#ifdef CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE
2461	int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu);
2462	#else
2463	static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
2464	{
2465	return `0`;
2466	}
2467	#endif /* CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE */
2468
2469	#ifdef CONFIG_KVM_XFER_TO_GUEST_WORK
2470	static inline void kvm_handle_signal_exit(struct kvm_vcpu *vcpu)
2471	{
2472	vcpu->run->exit_reason = KVM_EXIT_INTR;
2473	vcpu->stat.signal_exits++;
2474	}
2475	#endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */
2476
2477	/*
2478	* If more than one page is being (un)accounted, @virt must be the address of
2479	* the first page of a block of pages what were allocated together (i.e
2480	* accounted together).
2481	*
2482	* kvm_account_pgtable_pages() is thread-safe because mod_lruvec_page_state()
2483	* is thread-safe.
2484	*/
2485	static inline void kvm_account_pgtable_pages(void virt, int* nr)
2486	{
2487	mod_lruvec_page_state(virt_to_page(virt), idx: NR_SECONDARY_PAGETABLE, val: nr);
2488	}
2489
2490	/*
2491	* This defines how many reserved entries we want to keep before we
2492	* kick the vcpu to the userspace to avoid dirty ring full. This
2493	* value can be tuned to higher if e.g. PML is enabled on the host.
2494	*/
2495	#define KVM_DIRTY_RING_RSVD_ENTRIES 64
2496
2497	/ Max number of entries allowed for each kvm dirty ring /
2498	#define KVM_DIRTY_RING_MAX_ENTRIES 65536
2499
2500	static inline void kvm_prepare_memory_fault_exit(struct kvm_vcpu *vcpu,
2501	gpa_t gpa, gpa_t size,
2502	bool is_write, bool is_exec,
2503	bool is_private)
2504	{
2505	vcpu->run->exit_reason = KVM_EXIT_MEMORY_FAULT;
2506	vcpu->run->memory_fault.gpa = gpa;
2507	vcpu->run->memory_fault.size = size;
2508
2509	/ RWX flags are not (yet) defined or communicated to userspace. /
2510	vcpu->run->memory_fault.flags = `0`;
2511	if (is_private)
2512	vcpu->run->memory_fault.flags \|= KVM_MEMORY_EXIT_FLAG_PRIVATE;
2513	}
2514
2515	#ifdef CONFIG_KVM_GENERIC_MEMORY_ATTRIBUTES
2516	static inline unsigned long kvm_get_memory_attributes(struct kvm *kvm, gfn_t gfn)
2517	{
2518	return xa_to_value(entry: xa_load(&kvm->mem_attr_array, index: gfn));
2519	}
2520
2521	bool kvm_range_has_memory_attributes(struct kvm *kvm, gfn_t start, gfn_t end,
2522	unsigned long mask, unsigned long attrs);
2523	bool kvm_arch_pre_set_memory_attributes(struct kvm *kvm,
2524	struct kvm_gfn_range *range);
2525	bool kvm_arch_post_set_memory_attributes(struct kvm *kvm,
2526	struct kvm_gfn_range *range);
2527
2528	static inline bool kvm_mem_is_private(struct kvm *kvm, gfn_t gfn)
2529	{
2530	return IS_ENABLED(CONFIG_KVM_PRIVATE_MEM) &&
2531	kvm_get_memory_attributes(kvm, gfn) & KVM_MEMORY_ATTRIBUTE_PRIVATE;
2532	}
2533	#else
2534	static inline bool kvm_mem_is_private(struct kvm *kvm, gfn_t gfn)
2535	{
2536	return false;
2537	}
2538	#endif /* CONFIG_KVM_GENERIC_MEMORY_ATTRIBUTES */
2539
2540	#ifdef CONFIG_KVM_PRIVATE_MEM
2541	int kvm_gmem_get_pfn(struct kvm kvm, struct* kvm_memory_slot *slot,
2542	gfn_t gfn, kvm_pfn_t pfn, struct* page **page,
2543	int *max_order);
2544	#else
2545	static inline int kvm_gmem_get_pfn(struct kvm *kvm,
2546	struct kvm_memory_slot *slot, gfn_t gfn,
2547	kvm_pfn_t pfn, struct* page **page,
2548	int *max_order)
2549	{
2550	KVM_BUG_ON(`1`, kvm);
2551	return -EIO;
2552	}
2553	#endif /* CONFIG_KVM_PRIVATE_MEM */
2554
2555	#ifdef CONFIG_HAVE_KVM_ARCH_GMEM_PREPARE
2556	int kvm_arch_gmem_prepare(struct kvm kvm, gfn_t gfn, kvm_pfn_t pfn, int* max_order);
2557	#endif
2558
2559	#ifdef CONFIG_KVM_GENERIC_PRIVATE_MEM
2560	/**
2561	* kvm_gmem_populate() - Populate/prepare a GPA range with guest data
2562	*
2563	* @kvm: KVM instance
2564	* @gfn: starting GFN to be populated
2565	* @src: userspace-provided buffer containing data to copy into GFN range
2566	* (passed to @post_populate, and incremented on each iteration
2567	* if not NULL)
2568	* @npages: number of pages to copy from userspace-buffer
2569	* @post_populate: callback to issue for each gmem page that backs the GPA
2570	* range
2571	* @opaque: opaque data to pass to @post_populate callback
2572	*
2573	* This is primarily intended for cases where a gmem-backed GPA range needs
2574	* to be initialized with userspace-provided data prior to being mapped into
2575	* the guest as a private page. This should be called with the slots->lock
2576	* held so that caller-enforced invariants regarding the expected memory
2577	* attributes of the GPA range do not race with KVM_SET_MEMORY_ATTRIBUTES.
2578	*
2579	* Returns the number of pages that were populated.
2580	*/
2581	typedef int (kvm_gmem_populate_cb)(struct* kvm *kvm, gfn_t gfn, kvm_pfn_t pfn,
2582	void __user src, int* order, void *opaque);
2583
2584	long kvm_gmem_populate(struct kvm kvm, gfn_t gfn, void* __user src, long* npages,
2585	kvm_gmem_populate_cb post_populate, void *opaque);
2586	#endif
2587
2588	#ifdef CONFIG_HAVE_KVM_ARCH_GMEM_INVALIDATE
2589	void kvm_arch_gmem_invalidate(kvm_pfn_t start, kvm_pfn_t end);
2590	#endif
2591
2592	#ifdef CONFIG_KVM_GENERIC_PRE_FAULT_MEMORY
2593	long kvm_arch_vcpu_pre_fault_memory(struct kvm_vcpu *vcpu,
2594	struct kvm_pre_fault_memory *range);
2595	#endif
2596
2597	#ifdef CONFIG_KVM_GENERIC_HARDWARE_ENABLING
2598	int kvm_enable_virtualization(void);
2599	void kvm_disable_virtualization(void);
2600	#else
2601	static inline int kvm_enable_virtualization(void) { return `0`; }
2602	static inline void kvm_disable_virtualization(void) { }
2603	#endif
2604
2605	#endif
2606

source code of linux/include/linux/kvm_host.h