kvm_main.c 95.6 KB
Newer Older
A
Avi Kivity 已提交
1 2 3 4 5 6 7
/*
 * Kernel-based Virtual Machine driver for Linux
 *
 * This module enables machines with Intel VT-x extensions to run virtual
 * machines without emulation or binary translation.
 *
 * Copyright (C) 2006 Qumranet, Inc.
N
Nicolas Kaiser 已提交
8
 * Copyright 2010 Red Hat, Inc. and/or its affiliates.
A
Avi Kivity 已提交
9 10 11 12 13 14 15 16 17 18
 *
 * Authors:
 *   Avi Kivity   <avi@qumranet.com>
 *   Yaniv Kamay  <yaniv@qumranet.com>
 *
 * This work is licensed under the terms of the GNU GPL, version 2.  See
 * the COPYING file in the top-level directory.
 *
 */

19
#include <kvm/iodev.h>
A
Avi Kivity 已提交
20

21
#include <linux/kvm_host.h>
A
Avi Kivity 已提交
22 23 24 25 26 27 28 29 30 31 32
#include <linux/kvm.h>
#include <linux/module.h>
#include <linux/errno.h>
#include <linux/percpu.h>
#include <linux/mm.h>
#include <linux/miscdevice.h>
#include <linux/vmalloc.h>
#include <linux/reboot.h>
#include <linux/debugfs.h>
#include <linux/highmem.h>
#include <linux/file.h>
33
#include <linux/syscore_ops.h>
A
Avi Kivity 已提交
34
#include <linux/cpu.h>
35
#include <linux/sched/signal.h>
36
#include <linux/sched/mm.h>
37
#include <linux/sched/stat.h>
38 39
#include <linux/cpumask.h>
#include <linux/smp.h>
40
#include <linux/anon_inodes.h>
41
#include <linux/profile.h>
42
#include <linux/kvm_para.h>
43
#include <linux/pagemap.h>
44
#include <linux/mman.h>
45
#include <linux/swap.h>
46
#include <linux/bitops.h>
47
#include <linux/spinlock.h>
48
#include <linux/compat.h>
49
#include <linux/srcu.h>
J
Joerg Roedel 已提交
50
#include <linux/hugetlb.h>
51
#include <linux/slab.h>
52 53
#include <linux/sort.h>
#include <linux/bsearch.h>
A
Avi Kivity 已提交
54

A
Avi Kivity 已提交
55 56
#include <asm/processor.h>
#include <asm/io.h>
57
#include <asm/ioctl.h>
58
#include <linux/uaccess.h>
59
#include <asm/pgtable.h>
A
Avi Kivity 已提交
60

61
#include "coalesced_mmio.h"
62
#include "async_pf.h"
P
Paolo Bonzini 已提交
63
#include "vfio.h"
64

65 66 67
#define CREATE_TRACE_POINTS
#include <trace/events/kvm.h>

68 69 70
/* Worst case buffer size needed for holding an integer. */
#define ITOA_MAX_LEN 12

A
Avi Kivity 已提交
71 72 73
MODULE_AUTHOR("Qumranet");
MODULE_LICENSE("GPL");

74
/* Architectures should define their poll value according to the halt latency */
75
unsigned int halt_poll_ns = KVM_HALT_POLL_NS_DEFAULT;
76
module_param(halt_poll_ns, uint, 0644);
77
EXPORT_SYMBOL_GPL(halt_poll_ns);
78

W
Wanpeng Li 已提交
79
/* Default doubles per-vcpu halt_poll_ns. */
80
unsigned int halt_poll_ns_grow = 2;
81
module_param(halt_poll_ns_grow, uint, 0644);
82
EXPORT_SYMBOL_GPL(halt_poll_ns_grow);
W
Wanpeng Li 已提交
83 84

/* Default resets per-vcpu halt_poll_ns . */
85
unsigned int halt_poll_ns_shrink;
86
module_param(halt_poll_ns_shrink, uint, 0644);
87
EXPORT_SYMBOL_GPL(halt_poll_ns_shrink);
W
Wanpeng Li 已提交
88

89 90 91
/*
 * Ordering of locks:
 *
92
 *	kvm->lock --> kvm->slots_lock --> kvm->irq_lock
93 94
 */

95
DEFINE_SPINLOCK(kvm_lock);
96
static DEFINE_RAW_SPINLOCK(kvm_count_lock);
97
LIST_HEAD(vm_list);
98

99
static cpumask_var_t cpus_hardware_enabled;
100
static int kvm_usage_count;
101
static atomic_t hardware_enable_failed;
102

103 104
struct kmem_cache *kvm_vcpu_cache;
EXPORT_SYMBOL_GPL(kvm_vcpu_cache);
A
Avi Kivity 已提交
105

106 107
static __read_mostly struct preempt_ops kvm_preempt_ops;

108
struct dentry *kvm_debugfs_dir;
109
EXPORT_SYMBOL_GPL(kvm_debugfs_dir);
A
Avi Kivity 已提交
110

111 112 113
static int kvm_debugfs_num_entries;
static const struct file_operations *stat_fops_per_vm[];

A
Avi Kivity 已提交
114 115
static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl,
			   unsigned long arg);
116
#ifdef CONFIG_KVM_COMPAT
117 118
static long kvm_vcpu_compat_ioctl(struct file *file, unsigned int ioctl,
				  unsigned long arg);
119 120 121 122 123
#define KVM_COMPAT(c)	.compat_ioctl	= (c)
#else
static long kvm_no_compat_ioctl(struct file *file, unsigned int ioctl,
				unsigned long arg) { return -EINVAL; }
#define KVM_COMPAT(c)	.compat_ioctl	= kvm_no_compat_ioctl
124
#endif
125 126
static int hardware_enable_all(void);
static void hardware_disable_all(void);
A
Avi Kivity 已提交
127

M
Marcelo Tosatti 已提交
128
static void kvm_io_bus_destroy(struct kvm_io_bus *bus);
129

130
static void mark_page_dirty_in_slot(struct kvm_memory_slot *memslot, gfn_t gfn);
M
Marcelo Tosatti 已提交
131

132
__visible bool kvm_rebooting;
133
EXPORT_SYMBOL_GPL(kvm_rebooting);
134

135 136
static bool largepages_enabled = true;

137 138 139 140 141 142
#define KVM_EVENT_CREATE_VM 0
#define KVM_EVENT_DESTROY_VM 1
static void kvm_uevent_notify_change(unsigned int type, struct kvm *kvm);
static unsigned long long kvm_createvm_count;
static unsigned long long kvm_active_vms;

143 144 145 146 147
__weak void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
		unsigned long start, unsigned long end)
{
}

D
Dan Williams 已提交
148
bool kvm_is_reserved_pfn(kvm_pfn_t pfn)
B
Ben-Ami Yassour 已提交
149
{
150
	if (pfn_valid(pfn))
151
		return PageReserved(pfn_to_page(pfn));
B
Ben-Ami Yassour 已提交
152 153 154 155

	return true;
}

A
Avi Kivity 已提交
156 157 158
/*
 * Switches to specified vcpu, until a matching vcpu_put()
 */
159
void vcpu_load(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
160
{
161
	int cpu = get_cpu();
162
	preempt_notifier_register(&vcpu->preempt_notifier);
163
	kvm_arch_vcpu_load(vcpu, cpu);
164
	put_cpu();
A
Avi Kivity 已提交
165
}
166
EXPORT_SYMBOL_GPL(vcpu_load);
A
Avi Kivity 已提交
167

168
void vcpu_put(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
169
{
170
	preempt_disable();
171
	kvm_arch_vcpu_put(vcpu);
172 173
	preempt_notifier_unregister(&vcpu->preempt_notifier);
	preempt_enable();
A
Avi Kivity 已提交
174
}
175
EXPORT_SYMBOL_GPL(vcpu_put);
A
Avi Kivity 已提交
176

177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194
/* TODO: merge with kvm_arch_vcpu_should_kick */
static bool kvm_request_needs_ipi(struct kvm_vcpu *vcpu, unsigned req)
{
	int mode = kvm_vcpu_exiting_guest_mode(vcpu);

	/*
	 * We need to wait for the VCPU to reenable interrupts and get out of
	 * READING_SHADOW_PAGE_TABLES mode.
	 */
	if (req & KVM_REQUEST_WAIT)
		return mode != OUTSIDE_GUEST_MODE;

	/*
	 * Need to kick a running VCPU, but otherwise there is nothing to do.
	 */
	return mode == IN_GUEST_MODE;
}

195 196 197 198
static void ack_flush(void *_completed)
{
}

199 200 201 202 203 204 205 206 207 208 209 210
static inline bool kvm_kick_many_cpus(const struct cpumask *cpus, bool wait)
{
	if (unlikely(!cpus))
		cpus = cpu_online_mask;

	if (cpumask_empty(cpus))
		return false;

	smp_call_function_many(cpus, ack_flush, NULL, wait);
	return true;
}

211 212
bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
				 unsigned long *vcpu_bitmap, cpumask_var_t tmp)
213
{
214
	int i, cpu, me;
215
	struct kvm_vcpu *vcpu;
216
	bool called;
217

218
	me = get_cpu();
219

220
	kvm_for_each_vcpu(i, vcpu, kvm) {
221 222 223
		if (!test_bit(i, vcpu_bitmap))
			continue;

224
		kvm_make_request(req, vcpu);
225
		cpu = vcpu->cpu;
226

227 228
		if (!(req & KVM_REQUEST_NO_WAKEUP) && kvm_vcpu_wake_up(vcpu))
			continue;
229

230
		if (tmp != NULL && cpu != -1 && cpu != me &&
231
		    kvm_request_needs_ipi(vcpu, req))
232
			__cpumask_set_cpu(cpu, tmp);
233
	}
234 235

	called = kvm_kick_many_cpus(tmp, !!(req & KVM_REQUEST_WAIT));
236
	put_cpu();
237 238 239 240 241 242 243 244 245 246 247 248 249 250 251

	return called;
}

bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req)
{
	cpumask_var_t cpus;
	bool called;
	static unsigned long vcpu_bitmap[BITS_TO_LONGS(KVM_MAX_VCPUS)]
		= {[0 ... BITS_TO_LONGS(KVM_MAX_VCPUS)-1] = ULONG_MAX};

	zalloc_cpumask_var(&cpus, GFP_ATOMIC);

	called = kvm_make_vcpus_request_mask(kvm, req, vcpu_bitmap, cpus);

252
	free_cpumask_var(cpus);
253
	return called;
254 255
}

256
#ifndef CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALL
257
void kvm_flush_remote_tlbs(struct kvm *kvm)
258
{
259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275
	/*
	 * Read tlbs_dirty before setting KVM_REQ_TLB_FLUSH in
	 * kvm_make_all_cpus_request.
	 */
	long dirty_count = smp_load_acquire(&kvm->tlbs_dirty);

	/*
	 * We want to publish modifications to the page tables before reading
	 * mode. Pairs with a memory barrier in arch-specific code.
	 * - x86: smp_mb__after_srcu_read_unlock in vcpu_enter_guest
	 * and smp_mb in walk_shadow_page_lockless_begin/end.
	 * - powerpc: smp_mb in kvmppc_prepare_to_enter.
	 *
	 * There is already an smp_mb__after_atomic() before
	 * kvm_make_all_cpus_request() reads vcpu->mode. We reuse that
	 * barrier here.
	 */
276 277
	if (!kvm_arch_flush_remote_tlb(kvm)
	    || kvm_make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH))
278
		++kvm->stat.remote_tlb_flush;
279
	cmpxchg(&kvm->tlbs_dirty, dirty_count, 0);
280
}
281
EXPORT_SYMBOL_GPL(kvm_flush_remote_tlbs);
282
#endif
283

284 285
void kvm_reload_remote_mmus(struct kvm *kvm)
{
286
	kvm_make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD);
287
}
288

R
Rusty Russell 已提交
289 290 291 292 293 294 295 296 297
int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id)
{
	struct page *page;
	int r;

	mutex_init(&vcpu->mutex);
	vcpu->cpu = -1;
	vcpu->kvm = kvm;
	vcpu->vcpu_id = id;
298
	vcpu->pid = NULL;
299
	init_swait_queue_head(&vcpu->wq);
300
	kvm_async_pf_vcpu_init(vcpu);
R
Rusty Russell 已提交
301

302 303 304
	vcpu->pre_pcpu = -1;
	INIT_LIST_HEAD(&vcpu->blocked_vcpu_list);

R
Rusty Russell 已提交
305 306 307 308 309 310 311
	page = alloc_page(GFP_KERNEL | __GFP_ZERO);
	if (!page) {
		r = -ENOMEM;
		goto fail;
	}
	vcpu->run = page_address(page);

312 313
	kvm_vcpu_set_in_spin_loop(vcpu, false);
	kvm_vcpu_set_dy_eligible(vcpu, false);
314
	vcpu->preempted = false;
315

316
	r = kvm_arch_vcpu_init(vcpu);
R
Rusty Russell 已提交
317
	if (r < 0)
318
		goto fail_free_run;
R
Rusty Russell 已提交
319 320 321 322 323
	return 0;

fail_free_run:
	free_page((unsigned long)vcpu->run);
fail:
324
	return r;
R
Rusty Russell 已提交
325 326 327 328 329
}
EXPORT_SYMBOL_GPL(kvm_vcpu_init);

void kvm_vcpu_uninit(struct kvm_vcpu *vcpu)
{
330 331 332 333 334 335
	/*
	 * no need for rcu_read_lock as VCPU_RUN is the only place that
	 * will change the vcpu->pid pointer and on uninit all file
	 * descriptors are already gone.
	 */
	put_pid(rcu_dereference_protected(vcpu->pid, 1));
336
	kvm_arch_vcpu_uninit(vcpu);
R
Rusty Russell 已提交
337 338 339 340
	free_page((unsigned long)vcpu->run);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_uninit);

341 342 343 344 345 346
#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn)
{
	return container_of(mn, struct kvm, mmu_notifier);
}

347 348 349 350 351 352
static void kvm_mmu_notifier_change_pte(struct mmu_notifier *mn,
					struct mm_struct *mm,
					unsigned long address,
					pte_t pte)
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
353
	int idx;
354

355
	idx = srcu_read_lock(&kvm->srcu);
356 357 358 359
	spin_lock(&kvm->mmu_lock);
	kvm->mmu_notifier_seq++;
	kvm_set_spte_hva(kvm, address, pte);
	spin_unlock(&kvm->mmu_lock);
360
	srcu_read_unlock(&kvm->srcu, idx);
361 362
}

363 364 365 366 367 368
static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn,
						    struct mm_struct *mm,
						    unsigned long start,
						    unsigned long end)
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
369
	int need_tlb_flush = 0, idx;
370

371
	idx = srcu_read_lock(&kvm->srcu);
372 373 374 375 376 377 378
	spin_lock(&kvm->mmu_lock);
	/*
	 * The count increase must become visible at unlock time as no
	 * spte can be established without taking the mmu_lock and
	 * count is also read inside the mmu_lock critical section.
	 */
	kvm->mmu_notifier_count++;
379
	need_tlb_flush = kvm_unmap_hva_range(kvm, start, end);
380
	need_tlb_flush |= kvm->tlbs_dirty;
381 382 383
	/* we've to flush the tlb before the pages can be freed */
	if (need_tlb_flush)
		kvm_flush_remote_tlbs(kvm);
384 385

	spin_unlock(&kvm->mmu_lock);
386 387 388

	kvm_arch_mmu_notifier_invalidate_range(kvm, start, end);

389
	srcu_read_unlock(&kvm->srcu, idx);
390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405
}

static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
						  struct mm_struct *mm,
						  unsigned long start,
						  unsigned long end)
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);

	spin_lock(&kvm->mmu_lock);
	/*
	 * This sequence increase will notify the kvm page fault that
	 * the page that is going to be mapped in the spte could have
	 * been freed.
	 */
	kvm->mmu_notifier_seq++;
406
	smp_wmb();
407 408
	/*
	 * The above sequence increase must be visible before the
409 410
	 * below count decrease, which is ensured by the smp_wmb above
	 * in conjunction with the smp_rmb in mmu_notifier_retry().
411 412 413 414 415 416 417 418 419
	 */
	kvm->mmu_notifier_count--;
	spin_unlock(&kvm->mmu_lock);

	BUG_ON(kvm->mmu_notifier_count < 0);
}

static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
					      struct mm_struct *mm,
A
Andres Lagar-Cavilla 已提交
420 421
					      unsigned long start,
					      unsigned long end)
422 423
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
424
	int young, idx;
425

426
	idx = srcu_read_lock(&kvm->srcu);
427 428
	spin_lock(&kvm->mmu_lock);

A
Andres Lagar-Cavilla 已提交
429
	young = kvm_age_hva(kvm, start, end);
430 431 432
	if (young)
		kvm_flush_remote_tlbs(kvm);

433 434 435
	spin_unlock(&kvm->mmu_lock);
	srcu_read_unlock(&kvm->srcu, idx);

436 437 438
	return young;
}

439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468
static int kvm_mmu_notifier_clear_young(struct mmu_notifier *mn,
					struct mm_struct *mm,
					unsigned long start,
					unsigned long end)
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
	int young, idx;

	idx = srcu_read_lock(&kvm->srcu);
	spin_lock(&kvm->mmu_lock);
	/*
	 * Even though we do not flush TLB, this will still adversely
	 * affect performance on pre-Haswell Intel EPT, where there is
	 * no EPT Access Bit to clear so that we have to tear down EPT
	 * tables instead. If we find this unacceptable, we can always
	 * add a parameter to kvm_age_hva so that it effectively doesn't
	 * do anything on clear_young.
	 *
	 * Also note that currently we never issue secondary TLB flushes
	 * from clear_young, leaving this job up to the regular system
	 * cadence. If we find this inaccurate, we might come up with a
	 * more sophisticated heuristic later.
	 */
	young = kvm_age_hva(kvm, start, end);
	spin_unlock(&kvm->mmu_lock);
	srcu_read_unlock(&kvm->srcu, idx);

	return young;
}

A
Andrea Arcangeli 已提交
469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484
static int kvm_mmu_notifier_test_young(struct mmu_notifier *mn,
				       struct mm_struct *mm,
				       unsigned long address)
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
	int young, idx;

	idx = srcu_read_lock(&kvm->srcu);
	spin_lock(&kvm->mmu_lock);
	young = kvm_test_age_hva(kvm, address);
	spin_unlock(&kvm->mmu_lock);
	srcu_read_unlock(&kvm->srcu, idx);

	return young;
}

485 486 487 488
static void kvm_mmu_notifier_release(struct mmu_notifier *mn,
				     struct mm_struct *mm)
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
489 490 491
	int idx;

	idx = srcu_read_lock(&kvm->srcu);
492
	kvm_arch_flush_shadow_all(kvm);
493
	srcu_read_unlock(&kvm->srcu, idx);
494 495
}

496
static const struct mmu_notifier_ops kvm_mmu_notifier_ops = {
497
	.flags			= MMU_INVALIDATE_DOES_NOT_BLOCK,
498 499 500
	.invalidate_range_start	= kvm_mmu_notifier_invalidate_range_start,
	.invalidate_range_end	= kvm_mmu_notifier_invalidate_range_end,
	.clear_flush_young	= kvm_mmu_notifier_clear_flush_young,
501
	.clear_young		= kvm_mmu_notifier_clear_young,
A
Andrea Arcangeli 已提交
502
	.test_young		= kvm_mmu_notifier_test_young,
503
	.change_pte		= kvm_mmu_notifier_change_pte,
504
	.release		= kvm_mmu_notifier_release,
505
};
506 507 508 509 510 511 512 513 514 515 516 517 518 519

static int kvm_init_mmu_notifier(struct kvm *kvm)
{
	kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops;
	return mmu_notifier_register(&kvm->mmu_notifier, current->mm);
}

#else  /* !(CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER) */

static int kvm_init_mmu_notifier(struct kvm *kvm)
{
	return 0;
}

520 521
#endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */

522
static struct kvm_memslots *kvm_alloc_memslots(void)
523 524
{
	int i;
525
	struct kvm_memslots *slots;
526

M
Michal Hocko 已提交
527
	slots = kvzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
528 529 530
	if (!slots)
		return NULL;

531
	for (i = 0; i < KVM_MEM_SLOTS_NUM; i++)
532
		slots->id_to_index[i] = slots->memslots[i].id = i;
533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570

	return slots;
}

static void kvm_destroy_dirty_bitmap(struct kvm_memory_slot *memslot)
{
	if (!memslot->dirty_bitmap)
		return;

	kvfree(memslot->dirty_bitmap);
	memslot->dirty_bitmap = NULL;
}

/*
 * Free any memory in @free but not in @dont.
 */
static void kvm_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
			      struct kvm_memory_slot *dont)
{
	if (!dont || free->dirty_bitmap != dont->dirty_bitmap)
		kvm_destroy_dirty_bitmap(free);

	kvm_arch_free_memslot(kvm, free, dont);

	free->npages = 0;
}

static void kvm_free_memslots(struct kvm *kvm, struct kvm_memslots *slots)
{
	struct kvm_memory_slot *memslot;

	if (!slots)
		return;

	kvm_for_each_memslot(memslot, slots)
		kvm_free_memslot(kvm, memslot, NULL);

	kvfree(slots);
571 572
}

573 574 575 576 577 578 579 580 581
static void kvm_destroy_vm_debugfs(struct kvm *kvm)
{
	int i;

	if (!kvm->debugfs_dentry)
		return;

	debugfs_remove_recursive(kvm->debugfs_dentry);

582 583 584 585 586
	if (kvm->debugfs_stat_data) {
		for (i = 0; i < kvm_debugfs_num_entries; i++)
			kfree(kvm->debugfs_stat_data[i]);
		kfree(kvm->debugfs_stat_data);
	}
587 588 589 590 591 592 593 594 595 596 597 598
}

static int kvm_create_vm_debugfs(struct kvm *kvm, int fd)
{
	char dir_name[ITOA_MAX_LEN * 2];
	struct kvm_stat_data *stat_data;
	struct kvm_stats_debugfs_item *p;

	if (!debugfs_initialized())
		return 0;

	snprintf(dir_name, sizeof(dir_name), "%d-%d", task_pid_nr(current), fd);
599
	kvm->debugfs_dentry = debugfs_create_dir(dir_name, kvm_debugfs_dir);
600 601 602 603 604 605 606 607 608 609 610 611 612 613 614

	kvm->debugfs_stat_data = kcalloc(kvm_debugfs_num_entries,
					 sizeof(*kvm->debugfs_stat_data),
					 GFP_KERNEL);
	if (!kvm->debugfs_stat_data)
		return -ENOMEM;

	for (p = debugfs_entries; p->name; p++) {
		stat_data = kzalloc(sizeof(*stat_data), GFP_KERNEL);
		if (!stat_data)
			return -ENOMEM;

		stat_data->kvm = kvm;
		stat_data->offset = p->offset;
		kvm->debugfs_stat_data[p - debugfs_entries] = stat_data;
615 616
		debugfs_create_file(p->name, 0644, kvm->debugfs_dentry,
				    stat_data, stat_fops_per_vm[p->kind]);
617 618 619 620
	}
	return 0;
}

621
static struct kvm *kvm_create_vm(unsigned long type)
A
Avi Kivity 已提交
622
{
623 624
	int r, i;
	struct kvm *kvm = kvm_arch_alloc_vm();
A
Avi Kivity 已提交
625

626 627 628
	if (!kvm)
		return ERR_PTR(-ENOMEM);

629
	spin_lock_init(&kvm->mmu_lock);
V
Vegard Nossum 已提交
630
	mmgrab(current->mm);
631 632 633 634 635
	kvm->mm = current->mm;
	kvm_eventfd_init(kvm);
	mutex_init(&kvm->lock);
	mutex_init(&kvm->irq_lock);
	mutex_init(&kvm->slots_lock);
636
	refcount_set(&kvm->users_count, 1);
637 638
	INIT_LIST_HEAD(&kvm->devices);

639
	r = kvm_arch_init_vm(kvm, type);
640
	if (r)
641
		goto out_err_no_disable;
642 643 644

	r = hardware_enable_all();
	if (r)
645
		goto out_err_no_disable;
646

647
#ifdef CONFIG_HAVE_KVM_IRQFD
648
	INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list);
649
#endif
A
Avi Kivity 已提交
650

651 652
	BUILD_BUG_ON(KVM_MEM_SLOTS_NUM > SHRT_MAX);

653
	r = -ENOMEM;
654
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
655 656
		struct kvm_memslots *slots = kvm_alloc_memslots();
		if (!slots)
657
			goto out_err_no_srcu;
658 659 660 661 662 663 664
		/*
		 * Generations must be different for each address space.
		 * Init kvm generation close to the maximum to easily test the
		 * code of handling generation number wrap-around.
		 */
		slots->generation = i * 2 - 150;
		rcu_assign_pointer(kvm->memslots[i], slots);
665
	}
666

667
	if (init_srcu_struct(&kvm->srcu))
668 669 670
		goto out_err_no_srcu;
	if (init_srcu_struct(&kvm->irq_srcu))
		goto out_err_no_irq_srcu;
M
Marcelo Tosatti 已提交
671
	for (i = 0; i < KVM_NR_BUSES; i++) {
672 673
		rcu_assign_pointer(kvm->buses[i],
			kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL));
674
		if (!kvm->buses[i])
M
Marcelo Tosatti 已提交
675 676
			goto out_err;
	}
677

678 679 680 681
	r = kvm_init_mmu_notifier(kvm);
	if (r)
		goto out_err;

682
	spin_lock(&kvm_lock);
683
	list_add(&kvm->vm_list, &vm_list);
684
	spin_unlock(&kvm_lock);
685

686 687
	preempt_notifier_inc();

688
	return kvm;
689 690

out_err:
691 692
	cleanup_srcu_struct(&kvm->irq_srcu);
out_err_no_irq_srcu:
693
	cleanup_srcu_struct(&kvm->srcu);
694
out_err_no_srcu:
695
	hardware_disable_all();
696
out_err_no_disable:
697
	refcount_set(&kvm->users_count, 0);
M
Marcelo Tosatti 已提交
698
	for (i = 0; i < KVM_NR_BUSES; i++)
699
		kfree(kvm_get_bus(kvm, i));
700
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
701
		kvm_free_memslots(kvm, __kvm_memslots(kvm, i));
702
	kvm_arch_free_vm(kvm);
703
	mmdrop(current->mm);
704
	return ERR_PTR(r);
705 706
}

707 708
static void kvm_destroy_devices(struct kvm *kvm)
{
G
Geliang Tang 已提交
709
	struct kvm_device *dev, *tmp;
710

711 712 713 714 715
	/*
	 * We do not need to take the kvm->lock here, because nobody else
	 * has a reference to the struct kvm at this point and therefore
	 * cannot access the devices list anyhow.
	 */
G
Geliang Tang 已提交
716 717
	list_for_each_entry_safe(dev, tmp, &kvm->devices, vm_node) {
		list_del(&dev->vm_node);
718 719 720 721
		dev->ops->destroy(dev);
	}
}

722 723
static void kvm_destroy_vm(struct kvm *kvm)
{
M
Marcelo Tosatti 已提交
724
	int i;
725 726
	struct mm_struct *mm = kvm->mm;

727
	kvm_uevent_notify_change(KVM_EVENT_DESTROY_VM, kvm);
728
	kvm_destroy_vm_debugfs(kvm);
729
	kvm_arch_sync_events(kvm);
730
	spin_lock(&kvm_lock);
731
	list_del(&kvm->vm_list);
732
	spin_unlock(&kvm_lock);
733
	kvm_free_irq_routing(kvm);
734
	for (i = 0; i < KVM_NR_BUSES; i++) {
735
		struct kvm_io_bus *bus = kvm_get_bus(kvm, i);
736 737 738

		if (bus)
			kvm_io_bus_destroy(bus);
739 740
		kvm->buses[i] = NULL;
	}
741
	kvm_coalesced_mmio_free(kvm);
742 743
#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
	mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
744
#else
745
	kvm_arch_flush_shadow_all(kvm);
746
#endif
747
	kvm_arch_destroy_vm(kvm);
748
	kvm_destroy_devices(kvm);
749
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
750
		kvm_free_memslots(kvm, __kvm_memslots(kvm, i));
751
	cleanup_srcu_struct(&kvm->irq_srcu);
752 753
	cleanup_srcu_struct(&kvm->srcu);
	kvm_arch_free_vm(kvm);
754
	preempt_notifier_dec();
755
	hardware_disable_all();
756
	mmdrop(mm);
757 758
}

I
Izik Eidus 已提交
759 760
void kvm_get_kvm(struct kvm *kvm)
{
761
	refcount_inc(&kvm->users_count);
I
Izik Eidus 已提交
762 763 764 765 766
}
EXPORT_SYMBOL_GPL(kvm_get_kvm);

void kvm_put_kvm(struct kvm *kvm)
{
767
	if (refcount_dec_and_test(&kvm->users_count))
I
Izik Eidus 已提交
768 769 770 771 772
		kvm_destroy_vm(kvm);
}
EXPORT_SYMBOL_GPL(kvm_put_kvm);


773 774 775 776
static int kvm_vm_release(struct inode *inode, struct file *filp)
{
	struct kvm *kvm = filp->private_data;

G
Gregory Haskins 已提交
777 778
	kvm_irqfd_release(kvm);

I
Izik Eidus 已提交
779
	kvm_put_kvm(kvm);
A
Avi Kivity 已提交
780 781 782
	return 0;
}

783 784
/*
 * Allocation size is twice as large as the actual dirty bitmap size.
785
 * See x86's kvm_vm_ioctl_get_dirty_log() why this is needed.
786
 */
787 788
static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot)
{
789
	unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot);
790

M
Michal Hocko 已提交
791
	memslot->dirty_bitmap = kvzalloc(dirty_bytes, GFP_KERNEL);
792 793 794 795 796 797
	if (!memslot->dirty_bitmap)
		return -ENOMEM;

	return 0;
}

798
/*
799 800 801 802
 * Insert memslot and re-sort memslots based on their GFN,
 * so binary search could be used to lookup GFN.
 * Sorting algorithm takes advantage of having initially
 * sorted array and known changed memslot position.
803
 */
804 805
static void update_memslots(struct kvm_memslots *slots,
			    struct kvm_memory_slot *new)
806
{
807 808
	int id = new->id;
	int i = slots->id_to_index[id];
809
	struct kvm_memory_slot *mslots = slots->memslots;
810

811
	WARN_ON(mslots[i].id != id);
812
	if (!new->npages) {
813
		WARN_ON(!mslots[i].npages);
814 815 816 817 818 819
		if (mslots[i].npages)
			slots->used_slots--;
	} else {
		if (!mslots[i].npages)
			slots->used_slots++;
	}
820

821
	while (i < KVM_MEM_SLOTS_NUM - 1 &&
822 823 824
	       new->base_gfn <= mslots[i + 1].base_gfn) {
		if (!mslots[i + 1].npages)
			break;
825 826 827 828
		mslots[i] = mslots[i + 1];
		slots->id_to_index[mslots[i].id] = i;
		i++;
	}
829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845

	/*
	 * The ">=" is needed when creating a slot with base_gfn == 0,
	 * so that it moves before all those with base_gfn == npages == 0.
	 *
	 * On the other hand, if new->npages is zero, the above loop has
	 * already left i pointing to the beginning of the empty part of
	 * mslots, and the ">=" would move the hole backwards in this
	 * case---which is wrong.  So skip the loop when deleting a slot.
	 */
	if (new->npages) {
		while (i > 0 &&
		       new->base_gfn >= mslots[i - 1].base_gfn) {
			mslots[i] = mslots[i - 1];
			slots->id_to_index[mslots[i].id] = i;
			i--;
		}
846 847
	} else
		WARN_ON_ONCE(i != slots->used_slots);
848

849 850
	mslots[i] = *new;
	slots->id_to_index[mslots[i].id] = i;
851 852
}

853
static int check_memory_region_flags(const struct kvm_userspace_memory_region *mem)
854
{
X
Xiao Guangrong 已提交
855 856
	u32 valid_flags = KVM_MEM_LOG_DIRTY_PAGES;

857
#ifdef __KVM_HAVE_READONLY_MEM
X
Xiao Guangrong 已提交
858 859 860 861
	valid_flags |= KVM_MEM_READONLY;
#endif

	if (mem->flags & ~valid_flags)
862 863 864 865 866
		return -EINVAL;

	return 0;
}

867
static struct kvm_memslots *install_new_memslots(struct kvm *kvm,
868
		int as_id, struct kvm_memslots *slots)
869
{
870
	struct kvm_memslots *old_memslots = __kvm_memslots(kvm, as_id);
871

872 873 874 875 876 877 878
	/*
	 * Set the low bit in the generation, which disables SPTE caching
	 * until the end of synchronize_srcu_expedited.
	 */
	WARN_ON(old_memslots->generation & 1);
	slots->generation = old_memslots->generation + 1;

879
	rcu_assign_pointer(kvm->memslots[as_id], slots);
880
	synchronize_srcu_expedited(&kvm->srcu);
881

882 883 884 885
	/*
	 * Increment the new memslot generation a second time. This prevents
	 * vm exits that race with memslot updates from caching a memslot
	 * generation that will (potentially) be valid forever.
886 887 888 889 890 891
	 *
	 * Generations must be unique even across address spaces.  We do not need
	 * a global counter for that, instead the generation space is evenly split
	 * across address spaces.  For example, with two address spaces, address
	 * space 0 will use generations 0, 4, 8, ... while * address space 1 will
	 * use generations 2, 6, 10, 14, ...
892
	 */
893
	slots->generation += KVM_ADDRESS_SPACE_NUM * 2 - 1;
894

895
	kvm_arch_memslots_updated(kvm, slots);
896 897

	return old_memslots;
898 899
}

A
Avi Kivity 已提交
900 901 902 903 904
/*
 * Allocate some memory and give it an address in the guest physical address
 * space.
 *
 * Discontiguous memory is allowed, mostly for framebuffers.
905
 *
906
 * Must be called holding kvm->slots_lock for write.
A
Avi Kivity 已提交
907
 */
908
int __kvm_set_memory_region(struct kvm *kvm,
909
			    const struct kvm_userspace_memory_region *mem)
A
Avi Kivity 已提交
910
{
911
	int r;
A
Avi Kivity 已提交
912
	gfn_t base_gfn;
913
	unsigned long npages;
914
	struct kvm_memory_slot *slot;
A
Avi Kivity 已提交
915
	struct kvm_memory_slot old, new;
916
	struct kvm_memslots *slots = NULL, *old_memslots;
917
	int as_id, id;
918
	enum kvm_mr_change change;
A
Avi Kivity 已提交
919

920 921 922 923
	r = check_memory_region_flags(mem);
	if (r)
		goto out;

A
Avi Kivity 已提交
924
	r = -EINVAL;
925 926 927
	as_id = mem->slot >> 16;
	id = (u16)mem->slot;

A
Avi Kivity 已提交
928 929 930 931 932
	/* General sanity checks */
	if (mem->memory_size & (PAGE_SIZE - 1))
		goto out;
	if (mem->guest_phys_addr & (PAGE_SIZE - 1))
		goto out;
933
	/* We can read the guest memory with __xxx_user() later on. */
934
	if ((id < KVM_USER_MEM_SLOTS) &&
935
	    ((mem->userspace_addr & (PAGE_SIZE - 1)) ||
936 937 938
	     !access_ok(VERIFY_WRITE,
			(void __user *)(unsigned long)mem->userspace_addr,
			mem->memory_size)))
939
		goto out;
940
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_MEM_SLOTS_NUM)
A
Avi Kivity 已提交
941 942 943 944
		goto out;
	if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
		goto out;

945
	slot = id_to_memslot(__kvm_memslots(kvm, as_id), id);
A
Avi Kivity 已提交
946 947 948
	base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
	npages = mem->memory_size >> PAGE_SHIFT;

949 950 951
	if (npages > KVM_MEM_MAX_NR_PAGES)
		goto out;

952
	new = old = *slot;
A
Avi Kivity 已提交
953

954
	new.id = id;
A
Avi Kivity 已提交
955 956 957 958
	new.base_gfn = base_gfn;
	new.npages = npages;
	new.flags = mem->flags;

959 960 961 962 963
	if (npages) {
		if (!old.npages)
			change = KVM_MR_CREATE;
		else { /* Modify an existing slot. */
			if ((mem->userspace_addr != old.userspace_addr) ||
964 965
			    (npages != old.npages) ||
			    ((new.flags ^ old.flags) & KVM_MEM_READONLY))
966 967 968 969 970 971 972 973 974 975 976
				goto out;

			if (base_gfn != old.base_gfn)
				change = KVM_MR_MOVE;
			else if (new.flags != old.flags)
				change = KVM_MR_FLAGS_ONLY;
			else { /* Nothing to change. */
				r = 0;
				goto out;
			}
		}
977 978 979 980
	} else {
		if (!old.npages)
			goto out;

981
		change = KVM_MR_DELETE;
982 983 984
		new.base_gfn = 0;
		new.flags = 0;
	}
A
Avi Kivity 已提交
985

986
	if ((change == KVM_MR_CREATE) || (change == KVM_MR_MOVE)) {
987 988
		/* Check for overlaps */
		r = -EEXIST;
989
		kvm_for_each_memslot(slot, __kvm_memslots(kvm, as_id)) {
990
			if (slot->id == id)
991 992 993 994 995
				continue;
			if (!((base_gfn + npages <= slot->base_gfn) ||
			      (base_gfn >= slot->base_gfn + slot->npages)))
				goto out;
		}
A
Avi Kivity 已提交
996 997 998 999
	}

	/* Free page dirty bitmap if unneeded */
	if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
A
Al Viro 已提交
1000
		new.dirty_bitmap = NULL;
A
Avi Kivity 已提交
1001 1002

	r = -ENOMEM;
1003
	if (change == KVM_MR_CREATE) {
1004
		new.userspace_addr = mem->userspace_addr;
1005

1006
		if (kvm_arch_create_memslot(kvm, &new, npages))
1007
			goto out_free;
A
Avi Kivity 已提交
1008
	}
1009

A
Avi Kivity 已提交
1010 1011
	/* Allocate page dirty bitmap if needed */
	if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
1012
		if (kvm_create_dirty_bitmap(&new) < 0)
1013
			goto out_free;
A
Avi Kivity 已提交
1014 1015
	}

M
Michal Hocko 已提交
1016
	slots = kvzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
1017 1018
	if (!slots)
		goto out_free;
1019
	memcpy(slots, __kvm_memslots(kvm, as_id), sizeof(struct kvm_memslots));
1020

1021
	if ((change == KVM_MR_DELETE) || (change == KVM_MR_MOVE)) {
1022
		slot = id_to_memslot(slots, id);
1023 1024
		slot->flags |= KVM_MEMSLOT_INVALID;

1025
		old_memslots = install_new_memslots(kvm, as_id, slots);
1026

1027 1028
		/* From this point no new shadow pages pointing to a deleted,
		 * or moved, memslot will be created.
1029 1030
		 *
		 * validation of sp->gfn happens in:
1031 1032
		 *	- gfn_to_hva (kvm_read_guest, gfn_to_pfn)
		 *	- kvm_is_visible_gfn (mmu_check_roots)
1033
		 */
1034
		kvm_arch_flush_shadow_memslot(kvm, slot);
1035 1036 1037 1038 1039 1040

		/*
		 * We can re-use the old_memslots from above, the only difference
		 * from the currently installed memslots is the invalid flag.  This
		 * will get overwritten by update_memslots anyway.
		 */
1041
		slots = old_memslots;
1042
	}
1043

1044
	r = kvm_arch_prepare_memory_region(kvm, &new, mem, change);
1045
	if (r)
1046
		goto out_slots;
1047

1048
	/* actual memory is freed via old in kvm_free_memslot below */
1049
	if (change == KVM_MR_DELETE) {
1050
		new.dirty_bitmap = NULL;
1051
		memset(&new.arch, 0, sizeof(new.arch));
1052 1053
	}

1054
	update_memslots(slots, &new);
1055
	old_memslots = install_new_memslots(kvm, as_id, slots);
1056

1057
	kvm_arch_commit_memory_region(kvm, mem, &old, &new, change);
1058

1059
	kvm_free_memslot(kvm, &old, &new);
1060
	kvfree(old_memslots);
A
Avi Kivity 已提交
1061 1062
	return 0;

1063
out_slots:
1064
	kvfree(slots);
1065
out_free:
1066
	kvm_free_memslot(kvm, &new, &old);
A
Avi Kivity 已提交
1067 1068
out:
	return r;
1069
}
1070 1071 1072
EXPORT_SYMBOL_GPL(__kvm_set_memory_region);

int kvm_set_memory_region(struct kvm *kvm,
1073
			  const struct kvm_userspace_memory_region *mem)
1074 1075 1076
{
	int r;

1077
	mutex_lock(&kvm->slots_lock);
1078
	r = __kvm_set_memory_region(kvm, mem);
1079
	mutex_unlock(&kvm->slots_lock);
1080 1081
	return r;
}
1082 1083
EXPORT_SYMBOL_GPL(kvm_set_memory_region);

1084 1085
static int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
					  struct kvm_userspace_memory_region *mem)
1086
{
1087
	if ((u16)mem->slot >= KVM_USER_MEM_SLOTS)
1088
		return -EINVAL;
1089

1090
	return kvm_set_memory_region(kvm, mem);
A
Avi Kivity 已提交
1091 1092
}

1093 1094
int kvm_get_dirty_log(struct kvm *kvm,
			struct kvm_dirty_log *log, int *is_dirty)
A
Avi Kivity 已提交
1095
{
1096
	struct kvm_memslots *slots;
A
Avi Kivity 已提交
1097
	struct kvm_memory_slot *memslot;
1098
	int i, as_id, id;
1099
	unsigned long n;
A
Avi Kivity 已提交
1100 1101
	unsigned long any = 0;

1102 1103 1104
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1105
		return -EINVAL;
A
Avi Kivity 已提交
1106

1107 1108
	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
A
Avi Kivity 已提交
1109
	if (!memslot->dirty_bitmap)
1110
		return -ENOENT;
A
Avi Kivity 已提交
1111

1112
	n = kvm_dirty_bitmap_bytes(memslot);
A
Avi Kivity 已提交
1113

1114
	for (i = 0; !any && i < n/sizeof(long); ++i)
A
Avi Kivity 已提交
1115 1116 1117
		any = memslot->dirty_bitmap[i];

	if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
1118
		return -EFAULT;
A
Avi Kivity 已提交
1119

1120 1121
	if (any)
		*is_dirty = 1;
1122
	return 0;
A
Avi Kivity 已提交
1123
}
1124
EXPORT_SYMBOL_GPL(kvm_get_dirty_log);
A
Avi Kivity 已提交
1125

1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
/**
 * kvm_get_dirty_log_protect - get a snapshot of dirty pages, and if any pages
 *	are dirty write protect them for next write.
 * @kvm:	pointer to kvm instance
 * @log:	slot id and address to which we copy the log
 * @is_dirty:	flag set if any page is dirty
 *
 * We need to keep it in mind that VCPU threads can write to the bitmap
 * concurrently. So, to avoid losing track of dirty pages we keep the
 * following order:
 *
 *    1. Take a snapshot of the bit and clear it if needed.
 *    2. Write protect the corresponding page.
 *    3. Copy the snapshot to the userspace.
 *    4. Upon return caller flushes TLB's if needed.
 *
 * Between 2 and 4, the guest may write to the page using the remaining TLB
 * entry.  This is not a problem because the page is reported dirty using
 * the snapshot taken before and step 4 ensures that writes done after
 * exiting to userspace will be logged for the next call.
 *
 */
int kvm_get_dirty_log_protect(struct kvm *kvm,
			struct kvm_dirty_log *log, bool *is_dirty)
{
1152
	struct kvm_memslots *slots;
1153
	struct kvm_memory_slot *memslot;
1154
	int i, as_id, id;
1155 1156 1157 1158
	unsigned long n;
	unsigned long *dirty_bitmap;
	unsigned long *dirty_bitmap_buffer;

1159 1160 1161
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1162
		return -EINVAL;
1163

1164 1165
	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
1166 1167 1168

	dirty_bitmap = memslot->dirty_bitmap;
	if (!dirty_bitmap)
1169
		return -ENOENT;
1170 1171 1172

	n = kvm_dirty_bitmap_bytes(memslot);

1173
	dirty_bitmap_buffer = kvm_second_dirty_bitmap(memslot);
1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189
	memset(dirty_bitmap_buffer, 0, n);

	spin_lock(&kvm->mmu_lock);
	*is_dirty = false;
	for (i = 0; i < n / sizeof(long); i++) {
		unsigned long mask;
		gfn_t offset;

		if (!dirty_bitmap[i])
			continue;

		*is_dirty = true;

		mask = xchg(&dirty_bitmap[i], 0);
		dirty_bitmap_buffer[i] = mask;

1190 1191 1192 1193 1194
		if (mask) {
			offset = i * BITS_PER_LONG;
			kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot,
								offset, mask);
		}
1195 1196 1197 1198
	}

	spin_unlock(&kvm->mmu_lock);
	if (copy_to_user(log->dirty_bitmap, dirty_bitmap_buffer, n))
1199 1200
		return -EFAULT;
	return 0;
1201 1202 1203 1204
}
EXPORT_SYMBOL_GPL(kvm_get_dirty_log_protect);
#endif

1205 1206 1207 1208 1209
bool kvm_largepages_enabled(void)
{
	return largepages_enabled;
}

1210 1211 1212 1213 1214 1215
void kvm_disable_largepages(void)
{
	largepages_enabled = false;
}
EXPORT_SYMBOL_GPL(kvm_disable_largepages);

1216 1217 1218 1219
struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn)
{
	return __gfn_to_memslot(kvm_memslots(kvm), gfn);
}
A
Avi Kivity 已提交
1220
EXPORT_SYMBOL_GPL(gfn_to_memslot);
A
Avi Kivity 已提交
1221

1222 1223 1224 1225 1226
struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn)
{
	return __gfn_to_memslot(kvm_vcpu_memslots(vcpu), gfn);
}

1227
bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
1228
{
1229
	struct kvm_memory_slot *memslot = gfn_to_memslot(kvm, gfn);
1230

1231
	if (!memslot || memslot->id >= KVM_USER_MEM_SLOTS ||
1232
	      memslot->flags & KVM_MEMSLOT_INVALID)
1233
		return false;
1234

1235
	return true;
1236 1237 1238
}
EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);

J
Joerg Roedel 已提交
1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262
unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn)
{
	struct vm_area_struct *vma;
	unsigned long addr, size;

	size = PAGE_SIZE;

	addr = gfn_to_hva(kvm, gfn);
	if (kvm_is_error_hva(addr))
		return PAGE_SIZE;

	down_read(&current->mm->mmap_sem);
	vma = find_vma(current->mm, addr);
	if (!vma)
		goto out;

	size = vma_kernel_pagesize(vma);

out:
	up_read(&current->mm->mmap_sem);

	return size;
}

X
Xiao Guangrong 已提交
1263 1264 1265 1266 1267 1268 1269
static bool memslot_is_readonly(struct kvm_memory_slot *slot)
{
	return slot->flags & KVM_MEM_READONLY;
}

static unsigned long __gfn_to_hva_many(struct kvm_memory_slot *slot, gfn_t gfn,
				       gfn_t *nr_pages, bool write)
I
Izik Eidus 已提交
1270
{
1271
	if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
X
Xiao Guangrong 已提交
1272
		return KVM_HVA_ERR_BAD;
1273

X
Xiao Guangrong 已提交
1274 1275
	if (memslot_is_readonly(slot) && write)
		return KVM_HVA_ERR_RO_BAD;
1276 1277 1278 1279

	if (nr_pages)
		*nr_pages = slot->npages - (gfn - slot->base_gfn);

X
Xiao Guangrong 已提交
1280
	return __gfn_to_hva_memslot(slot, gfn);
I
Izik Eidus 已提交
1281
}
1282

X
Xiao Guangrong 已提交
1283 1284 1285 1286
static unsigned long gfn_to_hva_many(struct kvm_memory_slot *slot, gfn_t gfn,
				     gfn_t *nr_pages)
{
	return __gfn_to_hva_many(slot, gfn, nr_pages, true);
I
Izik Eidus 已提交
1287
}
1288

X
Xiao Guangrong 已提交
1289
unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot,
1290
					gfn_t gfn)
X
Xiao Guangrong 已提交
1291 1292 1293 1294 1295
{
	return gfn_to_hva_many(slot, gfn, NULL);
}
EXPORT_SYMBOL_GPL(gfn_to_hva_memslot);

1296 1297
unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
{
1298
	return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL);
1299
}
1300
EXPORT_SYMBOL_GPL(gfn_to_hva);
I
Izik Eidus 已提交
1301

1302 1303 1304 1305 1306 1307
unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn)
{
	return gfn_to_hva_many(kvm_vcpu_gfn_to_memslot(vcpu, gfn), gfn, NULL);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_gfn_to_hva);

1308
/*
1309 1310
 * If writable is set to false, the hva returned by this function is only
 * allowed to be read.
1311
 */
1312 1313
unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot,
				      gfn_t gfn, bool *writable)
1314
{
1315 1316 1317
	unsigned long hva = __gfn_to_hva_many(slot, gfn, NULL, false);

	if (!kvm_is_error_hva(hva) && writable)
1318 1319
		*writable = !memslot_is_readonly(slot);

1320
	return hva;
1321 1322
}

1323 1324 1325 1326 1327 1328 1329
unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable)
{
	struct kvm_memory_slot *slot = gfn_to_memslot(kvm, gfn);

	return gfn_to_hva_memslot_prot(slot, gfn, writable);
}

1330 1331 1332 1333 1334 1335 1336
unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable)
{
	struct kvm_memory_slot *slot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);

	return gfn_to_hva_memslot_prot(slot, gfn, writable);
}

1337 1338
static inline int check_user_page_hwpoison(unsigned long addr)
{
L
Lorenzo Stoakes 已提交
1339
	int rc, flags = FOLL_HWPOISON | FOLL_WRITE;
1340

L
Lorenzo Stoakes 已提交
1341
	rc = get_user_pages(addr, 1, flags, NULL, NULL);
1342 1343 1344
	return rc == -EHWPOISON;
}

X
Xiao Guangrong 已提交
1345 1346 1347 1348 1349
/*
 * The atomic path to get the writable pfn which will be stored in @pfn,
 * true indicates success, otherwise false is returned.
 */
static bool hva_to_pfn_fast(unsigned long addr, bool atomic, bool *async,
D
Dan Williams 已提交
1350
			    bool write_fault, bool *writable, kvm_pfn_t *pfn)
A
Avi Kivity 已提交
1351
{
1352
	struct page *page[1];
X
Xiao Guangrong 已提交
1353
	int npages;
A
Avi Kivity 已提交
1354

X
Xiao Guangrong 已提交
1355 1356
	if (!(async || atomic))
		return false;
1357

1358 1359 1360 1361 1362 1363 1364
	/*
	 * Fast pin a writable pfn only if it is a write fault request
	 * or the caller allows to map a writable pfn for a read fault
	 * request.
	 */
	if (!(write_fault || writable))
		return false;
1365

X
Xiao Guangrong 已提交
1366 1367 1368
	npages = __get_user_pages_fast(addr, 1, 1, page);
	if (npages == 1) {
		*pfn = page_to_pfn(page[0]);
1369

X
Xiao Guangrong 已提交
1370 1371 1372 1373
		if (writable)
			*writable = true;
		return true;
	}
1374

X
Xiao Guangrong 已提交
1375 1376
	return false;
}
1377

X
Xiao Guangrong 已提交
1378 1379 1380 1381 1382
/*
 * The slow path to get the pfn of the specified host virtual address,
 * 1 indicates success, -errno is returned if error is detected.
 */
static int hva_to_pfn_slow(unsigned long addr, bool *async, bool write_fault,
D
Dan Williams 已提交
1383
			   bool *writable, kvm_pfn_t *pfn)
X
Xiao Guangrong 已提交
1384
{
1385 1386
	unsigned int flags = FOLL_HWPOISON;
	struct page *page;
X
Xiao Guangrong 已提交
1387
	int npages = 0;
1388

X
Xiao Guangrong 已提交
1389 1390 1391 1392 1393
	might_sleep();

	if (writable)
		*writable = write_fault;

1394 1395 1396 1397
	if (write_fault)
		flags |= FOLL_WRITE;
	if (async)
		flags |= FOLL_NOWAIT;
1398

1399
	npages = get_user_pages_unlocked(addr, 1, &page, flags);
X
Xiao Guangrong 已提交
1400 1401 1402 1403
	if (npages != 1)
		return npages;

	/* map read fault as writable if possible */
1404
	if (unlikely(!write_fault) && writable) {
1405
		struct page *wpage;
X
Xiao Guangrong 已提交
1406

1407
		if (__get_user_pages_fast(addr, 1, 1, &wpage) == 1) {
X
Xiao Guangrong 已提交
1408
			*writable = true;
1409 1410
			put_page(page);
			page = wpage;
1411
		}
1412
	}
1413
	*pfn = page_to_pfn(page);
X
Xiao Guangrong 已提交
1414 1415
	return npages;
}
I
Izik Eidus 已提交
1416

X
Xiao Guangrong 已提交
1417 1418 1419 1420
static bool vma_is_valid(struct vm_area_struct *vma, bool write_fault)
{
	if (unlikely(!(vma->vm_flags & VM_READ)))
		return false;
1421

X
Xiao Guangrong 已提交
1422 1423
	if (write_fault && (unlikely(!(vma->vm_flags & VM_WRITE))))
		return false;
1424

X
Xiao Guangrong 已提交
1425 1426
	return true;
}
1427

1428 1429
static int hva_to_pfn_remapped(struct vm_area_struct *vma,
			       unsigned long addr, bool *async,
1430 1431
			       bool write_fault, bool *writable,
			       kvm_pfn_t *p_pfn)
1432
{
1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456
	unsigned long pfn;
	int r;

	r = follow_pfn(vma, addr, &pfn);
	if (r) {
		/*
		 * get_user_pages fails for VM_IO and VM_PFNMAP vmas and does
		 * not call the fault handler, so do it here.
		 */
		bool unlocked = false;
		r = fixup_user_fault(current, current->mm, addr,
				     (write_fault ? FAULT_FLAG_WRITE : 0),
				     &unlocked);
		if (unlocked)
			return -EAGAIN;
		if (r)
			return r;

		r = follow_pfn(vma, addr, &pfn);
		if (r)
			return r;

	}

1457 1458
	if (writable)
		*writable = true;
1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473

	/*
	 * Get a reference here because callers of *hva_to_pfn* and
	 * *gfn_to_pfn* ultimately call kvm_release_pfn_clean on the
	 * returned pfn.  This is only needed if the VMA has VM_MIXEDMAP
	 * set, but the kvm_get_pfn/kvm_release_pfn_clean pair will
	 * simply do nothing for reserved pfns.
	 *
	 * Whoever called remap_pfn_range is also going to call e.g.
	 * unmap_mapping_range before the underlying pages are freed,
	 * causing a call to our MMU notifier.
	 */ 
	kvm_get_pfn(pfn);

	*p_pfn = pfn;
1474 1475 1476
	return 0;
}

1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490
/*
 * Pin guest page in memory and return its pfn.
 * @addr: host virtual address which maps memory to the guest
 * @atomic: whether this function can sleep
 * @async: whether this function need to wait IO complete if the
 *         host page is not in the memory
 * @write_fault: whether we should get a writable host page
 * @writable: whether it allows to map a writable host page for !@write_fault
 *
 * The function will map a writable host page for these two cases:
 * 1): @write_fault = true
 * 2): @write_fault = false && @writable, @writable will tell the caller
 *     whether the mapping is writable.
 */
D
Dan Williams 已提交
1491
static kvm_pfn_t hva_to_pfn(unsigned long addr, bool atomic, bool *async,
X
Xiao Guangrong 已提交
1492 1493 1494
			bool write_fault, bool *writable)
{
	struct vm_area_struct *vma;
D
Dan Williams 已提交
1495
	kvm_pfn_t pfn = 0;
1496
	int npages, r;
1497

X
Xiao Guangrong 已提交
1498 1499
	/* we can do it either atomically or asynchronously, not both */
	BUG_ON(atomic && async);
1500

X
Xiao Guangrong 已提交
1501 1502 1503 1504 1505 1506 1507 1508 1509
	if (hva_to_pfn_fast(addr, atomic, async, write_fault, writable, &pfn))
		return pfn;

	if (atomic)
		return KVM_PFN_ERR_FAULT;

	npages = hva_to_pfn_slow(addr, async, write_fault, writable, &pfn);
	if (npages == 1)
		return pfn;
1510

X
Xiao Guangrong 已提交
1511 1512 1513 1514 1515 1516 1517
	down_read(&current->mm->mmap_sem);
	if (npages == -EHWPOISON ||
	      (!async && check_user_page_hwpoison(addr))) {
		pfn = KVM_PFN_ERR_HWPOISON;
		goto exit;
	}

1518
retry:
X
Xiao Guangrong 已提交
1519 1520 1521 1522
	vma = find_vma_intersection(current->mm, addr, addr + 1);

	if (vma == NULL)
		pfn = KVM_PFN_ERR_FAULT;
1523
	else if (vma->vm_flags & (VM_IO | VM_PFNMAP)) {
1524
		r = hva_to_pfn_remapped(vma, addr, async, write_fault, writable, &pfn);
1525 1526
		if (r == -EAGAIN)
			goto retry;
1527 1528
		if (r < 0)
			pfn = KVM_PFN_ERR_FAULT;
X
Xiao Guangrong 已提交
1529
	} else {
X
Xiao Guangrong 已提交
1530
		if (async && vma_is_valid(vma, write_fault))
X
Xiao Guangrong 已提交
1531 1532 1533 1534 1535
			*async = true;
		pfn = KVM_PFN_ERR_FAULT;
	}
exit:
	up_read(&current->mm->mmap_sem);
1536
	return pfn;
1537 1538
}

D
Dan Williams 已提交
1539 1540 1541
kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
			       bool atomic, bool *async, bool write_fault,
			       bool *writable)
1542
{
X
Xiao Guangrong 已提交
1543 1544
	unsigned long addr = __gfn_to_hva_many(slot, gfn, NULL, write_fault);

1545 1546 1547
	if (addr == KVM_HVA_ERR_RO_BAD) {
		if (writable)
			*writable = false;
X
Xiao Guangrong 已提交
1548
		return KVM_PFN_ERR_RO_FAULT;
1549
	}
X
Xiao Guangrong 已提交
1550

1551 1552 1553
	if (kvm_is_error_hva(addr)) {
		if (writable)
			*writable = false;
1554
		return KVM_PFN_NOSLOT;
1555
	}
X
Xiao Guangrong 已提交
1556 1557 1558 1559 1560 1561 1562 1563 1564

	/* Do not map writable pfn in the readonly memslot. */
	if (writable && memslot_is_readonly(slot)) {
		*writable = false;
		writable = NULL;
	}

	return hva_to_pfn(addr, atomic, async, write_fault,
			  writable);
1565
}
1566
EXPORT_SYMBOL_GPL(__gfn_to_pfn_memslot);
1567

D
Dan Williams 已提交
1568
kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1569 1570
		      bool *writable)
{
P
Paolo Bonzini 已提交
1571 1572
	return __gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn, false, NULL,
				    write_fault, writable);
1573 1574 1575
}
EXPORT_SYMBOL_GPL(gfn_to_pfn_prot);

D
Dan Williams 已提交
1576
kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1577
{
X
Xiao Guangrong 已提交
1578
	return __gfn_to_pfn_memslot(slot, gfn, false, NULL, true, NULL);
1579
}
P
Paolo Bonzini 已提交
1580
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot);
1581

D
Dan Williams 已提交
1582
kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn)
1583
{
X
Xiao Guangrong 已提交
1584
	return __gfn_to_pfn_memslot(slot, gfn, true, NULL, true, NULL);
1585
}
1586
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot_atomic);
1587

D
Dan Williams 已提交
1588
kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1589 1590 1591 1592 1593
{
	return gfn_to_pfn_memslot_atomic(gfn_to_memslot(kvm, gfn), gfn);
}
EXPORT_SYMBOL_GPL(gfn_to_pfn_atomic);

D
Dan Williams 已提交
1594
kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn)
1595 1596 1597 1598 1599
{
	return gfn_to_pfn_memslot_atomic(kvm_vcpu_gfn_to_memslot(vcpu, gfn), gfn);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_gfn_to_pfn_atomic);

D
Dan Williams 已提交
1600
kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1601 1602 1603 1604 1605
{
	return gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn);
}
EXPORT_SYMBOL_GPL(gfn_to_pfn);

D
Dan Williams 已提交
1606
kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn)
1607 1608 1609 1610 1611
{
	return gfn_to_pfn_memslot(kvm_vcpu_gfn_to_memslot(vcpu, gfn), gfn);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_gfn_to_pfn);

1612 1613
int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
			    struct page **pages, int nr_pages)
1614 1615
{
	unsigned long addr;
1616
	gfn_t entry = 0;
1617

1618
	addr = gfn_to_hva_many(slot, gfn, &entry);
1619 1620 1621 1622 1623 1624 1625 1626 1627 1628
	if (kvm_is_error_hva(addr))
		return -1;

	if (entry < nr_pages)
		return 0;

	return __get_user_pages_fast(addr, nr_pages, 1, pages);
}
EXPORT_SYMBOL_GPL(gfn_to_page_many_atomic);

D
Dan Williams 已提交
1629
static struct page *kvm_pfn_to_page(kvm_pfn_t pfn)
X
Xiao Guangrong 已提交
1630
{
1631
	if (is_error_noslot_pfn(pfn))
1632
		return KVM_ERR_PTR_BAD_PAGE;
X
Xiao Guangrong 已提交
1633

1634
	if (kvm_is_reserved_pfn(pfn)) {
1635
		WARN_ON(1);
1636
		return KVM_ERR_PTR_BAD_PAGE;
1637
	}
X
Xiao Guangrong 已提交
1638 1639 1640 1641

	return pfn_to_page(pfn);
}

1642 1643
struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
{
D
Dan Williams 已提交
1644
	kvm_pfn_t pfn;
1645 1646 1647

	pfn = gfn_to_pfn(kvm, gfn);

X
Xiao Guangrong 已提交
1648
	return kvm_pfn_to_page(pfn);
A
Avi Kivity 已提交
1649 1650 1651
}
EXPORT_SYMBOL_GPL(gfn_to_page);

1652 1653
struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn)
{
D
Dan Williams 已提交
1654
	kvm_pfn_t pfn;
1655 1656 1657 1658 1659 1660 1661

	pfn = kvm_vcpu_gfn_to_pfn(vcpu, gfn);

	return kvm_pfn_to_page(pfn);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_gfn_to_page);

1662 1663
void kvm_release_page_clean(struct page *page)
{
1664 1665
	WARN_ON(is_error_page(page));

1666
	kvm_release_pfn_clean(page_to_pfn(page));
1667 1668 1669
}
EXPORT_SYMBOL_GPL(kvm_release_page_clean);

D
Dan Williams 已提交
1670
void kvm_release_pfn_clean(kvm_pfn_t pfn)
1671
{
1672
	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn))
1673
		put_page(pfn_to_page(pfn));
1674 1675 1676
}
EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);

1677
void kvm_release_page_dirty(struct page *page)
1678
{
X
Xiao Guangrong 已提交
1679 1680
	WARN_ON(is_error_page(page));

1681 1682 1683 1684
	kvm_release_pfn_dirty(page_to_pfn(page));
}
EXPORT_SYMBOL_GPL(kvm_release_page_dirty);

1685
void kvm_release_pfn_dirty(kvm_pfn_t pfn)
1686 1687 1688 1689
{
	kvm_set_pfn_dirty(pfn);
	kvm_release_pfn_clean(pfn);
}
1690
EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
1691

D
Dan Williams 已提交
1692
void kvm_set_pfn_dirty(kvm_pfn_t pfn)
1693
{
1694
	if (!kvm_is_reserved_pfn(pfn)) {
1695
		struct page *page = pfn_to_page(pfn);
1696

1697 1698 1699
		if (!PageReserved(page))
			SetPageDirty(page);
	}
1700
}
1701 1702
EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);

D
Dan Williams 已提交
1703
void kvm_set_pfn_accessed(kvm_pfn_t pfn)
1704
{
1705
	if (!kvm_is_reserved_pfn(pfn))
1706
		mark_page_accessed(pfn_to_page(pfn));
1707 1708 1709
}
EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);

D
Dan Williams 已提交
1710
void kvm_get_pfn(kvm_pfn_t pfn)
1711
{
1712
	if (!kvm_is_reserved_pfn(pfn))
1713
		get_page(pfn_to_page(pfn));
1714 1715
}
EXPORT_SYMBOL_GPL(kvm_get_pfn);
1716

1717 1718 1719 1720 1721 1722 1723 1724
static int next_segment(unsigned long len, int offset)
{
	if (len > PAGE_SIZE - offset)
		return PAGE_SIZE - offset;
	else
		return len;
}

1725 1726
static int __kvm_read_guest_page(struct kvm_memory_slot *slot, gfn_t gfn,
				 void *data, int offset, int len)
1727
{
1728 1729
	int r;
	unsigned long addr;
1730

1731
	addr = gfn_to_hva_memslot_prot(slot, gfn, NULL);
1732 1733
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1734
	r = __copy_from_user(data, (void __user *)addr + offset, len);
1735
	if (r)
1736 1737 1738
		return -EFAULT;
	return 0;
}
1739 1740 1741 1742 1743 1744 1745 1746

int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
			int len)
{
	struct kvm_memory_slot *slot = gfn_to_memslot(kvm, gfn);

	return __kvm_read_guest_page(slot, gfn, data, offset, len);
}
1747 1748
EXPORT_SYMBOL_GPL(kvm_read_guest_page);

1749 1750 1751 1752 1753 1754 1755 1756 1757
int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data,
			     int offset, int len)
{
	struct kvm_memory_slot *slot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);

	return __kvm_read_guest_page(slot, gfn, data, offset, len);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_read_guest_page);

1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777
int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len)
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
	int seg;
	int offset = offset_in_page(gpa);
	int ret;

	while ((seg = next_segment(len, offset)) != 0) {
		ret = kvm_read_guest_page(kvm, gfn, data, offset, seg);
		if (ret < 0)
			return ret;
		offset = 0;
		len -= seg;
		data += seg;
		++gfn;
	}
	return 0;
}
EXPORT_SYMBOL_GPL(kvm_read_guest);

1778
int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data, unsigned long len)
1779 1780
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
1781
	int seg;
1782
	int offset = offset_in_page(gpa);
1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796
	int ret;

	while ((seg = next_segment(len, offset)) != 0) {
		ret = kvm_vcpu_read_guest_page(vcpu, gfn, data, offset, seg);
		if (ret < 0)
			return ret;
		offset = 0;
		len -= seg;
		data += seg;
		++gfn;
	}
	return 0;
}
EXPORT_SYMBOL_GPL(kvm_vcpu_read_guest);
1797

1798 1799 1800 1801 1802 1803 1804
static int __kvm_read_guest_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
			           void *data, int offset, unsigned long len)
{
	int r;
	unsigned long addr;

	addr = gfn_to_hva_memslot_prot(slot, gfn, NULL);
1805 1806
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1807
	pagefault_disable();
1808
	r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
1809
	pagefault_enable();
1810 1811 1812 1813 1814
	if (r)
		return -EFAULT;
	return 0;
}

1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838
int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
			  unsigned long len)
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
	struct kvm_memory_slot *slot = gfn_to_memslot(kvm, gfn);
	int offset = offset_in_page(gpa);

	return __kvm_read_guest_atomic(slot, gfn, data, offset, len);
}
EXPORT_SYMBOL_GPL(kvm_read_guest_atomic);

int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa,
			       void *data, unsigned long len)
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
	struct kvm_memory_slot *slot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);
	int offset = offset_in_page(gpa);

	return __kvm_read_guest_atomic(slot, gfn, data, offset, len);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_read_guest_atomic);

static int __kvm_write_guest_page(struct kvm_memory_slot *memslot, gfn_t gfn,
			          const void *data, int offset, int len)
1839
{
1840 1841
	int r;
	unsigned long addr;
1842

1843
	addr = gfn_to_hva_memslot(memslot, gfn);
1844 1845
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1846
	r = __copy_to_user((void __user *)addr + offset, data, len);
1847
	if (r)
1848
		return -EFAULT;
1849
	mark_page_dirty_in_slot(memslot, gfn);
1850 1851
	return 0;
}
1852 1853 1854 1855 1856 1857 1858 1859

int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn,
			 const void *data, int offset, int len)
{
	struct kvm_memory_slot *slot = gfn_to_memslot(kvm, gfn);

	return __kvm_write_guest_page(slot, gfn, data, offset, len);
}
1860 1861
EXPORT_SYMBOL_GPL(kvm_write_guest_page);

1862 1863 1864 1865 1866 1867 1868 1869 1870
int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn,
			      const void *data, int offset, int len)
{
	struct kvm_memory_slot *slot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);

	return __kvm_write_guest_page(slot, gfn, data, offset, len);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_write_guest_page);

1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889
int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
		    unsigned long len)
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
	int seg;
	int offset = offset_in_page(gpa);
	int ret;

	while ((seg = next_segment(len, offset)) != 0) {
		ret = kvm_write_guest_page(kvm, gfn, data, offset, seg);
		if (ret < 0)
			return ret;
		offset = 0;
		len -= seg;
		data += seg;
		++gfn;
	}
	return 0;
}
1890
EXPORT_SYMBOL_GPL(kvm_write_guest);
1891

1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912
int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
		         unsigned long len)
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
	int seg;
	int offset = offset_in_page(gpa);
	int ret;

	while ((seg = next_segment(len, offset)) != 0) {
		ret = kvm_vcpu_write_guest_page(vcpu, gfn, data, offset, seg);
		if (ret < 0)
			return ret;
		offset = 0;
		len -= seg;
		data += seg;
		++gfn;
	}
	return 0;
}
EXPORT_SYMBOL_GPL(kvm_vcpu_write_guest);

1913 1914 1915
static int __kvm_gfn_to_hva_cache_init(struct kvm_memslots *slots,
				       struct gfn_to_hva_cache *ghc,
				       gpa_t gpa, unsigned long len)
1916 1917
{
	int offset = offset_in_page(gpa);
1918 1919 1920 1921
	gfn_t start_gfn = gpa >> PAGE_SHIFT;
	gfn_t end_gfn = (gpa + len - 1) >> PAGE_SHIFT;
	gfn_t nr_pages_needed = end_gfn - start_gfn + 1;
	gfn_t nr_pages_avail;
1922 1923 1924

	ghc->gpa = gpa;
	ghc->generation = slots->generation;
1925
	ghc->len = len;
1926
	ghc->memslot = __gfn_to_memslot(slots, start_gfn);
1927 1928
	ghc->hva = gfn_to_hva_many(ghc->memslot, start_gfn, NULL);
	if (!kvm_is_error_hva(ghc->hva) && nr_pages_needed <= 1) {
1929
		ghc->hva += offset;
1930 1931 1932 1933 1934 1935
	} else {
		/*
		 * If the requested region crosses two memslots, we still
		 * verify that the entire region is valid here.
		 */
		while (start_gfn <= end_gfn) {
1936
			nr_pages_avail = 0;
1937
			ghc->memslot = __gfn_to_memslot(slots, start_gfn);
1938 1939 1940 1941 1942 1943 1944 1945 1946
			ghc->hva = gfn_to_hva_many(ghc->memslot, start_gfn,
						   &nr_pages_avail);
			if (kvm_is_error_hva(ghc->hva))
				return -EFAULT;
			start_gfn += nr_pages_avail;
		}
		/* Use the slow path for cross page reads and writes. */
		ghc->memslot = NULL;
	}
1947 1948
	return 0;
}
1949

1950
int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1951 1952
			      gpa_t gpa, unsigned long len)
{
1953
	struct kvm_memslots *slots = kvm_memslots(kvm);
1954 1955
	return __kvm_gfn_to_hva_cache_init(slots, ghc, gpa, len);
}
1956
EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init);
1957

1958 1959
int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, int offset, unsigned long len)
1960
{
1961
	struct kvm_memslots *slots = kvm_memslots(kvm);
1962
	int r;
1963
	gpa_t gpa = ghc->gpa + offset;
1964

1965
	BUG_ON(len + offset > ghc->len);
1966

1967
	if (slots->generation != ghc->generation)
1968
		__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len);
1969 1970

	if (unlikely(!ghc->memslot))
1971
		return kvm_write_guest(kvm, gpa, data, len);
1972 1973 1974 1975

	if (kvm_is_error_hva(ghc->hva))
		return -EFAULT;

1976
	r = __copy_to_user((void __user *)ghc->hva + offset, data, len);
1977 1978
	if (r)
		return -EFAULT;
1979
	mark_page_dirty_in_slot(ghc->memslot, gpa >> PAGE_SHIFT);
1980 1981 1982

	return 0;
}
1983
EXPORT_SYMBOL_GPL(kvm_write_guest_offset_cached);
1984

1985 1986
int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
1987
{
1988
	return kvm_write_guest_offset_cached(kvm, ghc, data, 0, len);
1989
}
1990
EXPORT_SYMBOL_GPL(kvm_write_guest_cached);
1991

1992 1993
int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
1994
{
1995
	struct kvm_memslots *slots = kvm_memslots(kvm);
1996 1997
	int r;

1998 1999
	BUG_ON(len > ghc->len);

2000
	if (slots->generation != ghc->generation)
2001
		__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len);
2002 2003

	if (unlikely(!ghc->memslot))
2004
		return kvm_read_guest(kvm, ghc->gpa, data, len);
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

	if (kvm_is_error_hva(ghc->hva))
		return -EFAULT;

	r = __copy_from_user(data, (void __user *)ghc->hva, len);
	if (r)
		return -EFAULT;

	return 0;
}
2015
EXPORT_SYMBOL_GPL(kvm_read_guest_cached);
2016

2017 2018
int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
{
2019 2020 2021
	const void *zero_page = (const void *) __va(page_to_phys(ZERO_PAGE(0)));

	return kvm_write_guest_page(kvm, gfn, zero_page, offset, len);
2022 2023 2024 2025 2026 2027 2028 2029 2030 2031
}
EXPORT_SYMBOL_GPL(kvm_clear_guest_page);

int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len)
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
	int seg;
	int offset = offset_in_page(gpa);
	int ret;

2032
	while ((seg = next_segment(len, offset)) != 0) {
2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043
		ret = kvm_clear_guest_page(kvm, gfn, offset, seg);
		if (ret < 0)
			return ret;
		offset = 0;
		len -= seg;
		++gfn;
	}
	return 0;
}
EXPORT_SYMBOL_GPL(kvm_clear_guest);

2044
static void mark_page_dirty_in_slot(struct kvm_memory_slot *memslot,
2045
				    gfn_t gfn)
A
Avi Kivity 已提交
2046
{
R
Rusty Russell 已提交
2047 2048
	if (memslot && memslot->dirty_bitmap) {
		unsigned long rel_gfn = gfn - memslot->base_gfn;
A
Avi Kivity 已提交
2049

2050
		set_bit_le(rel_gfn, memslot->dirty_bitmap);
A
Avi Kivity 已提交
2051 2052 2053
	}
}

2054 2055 2056 2057 2058
void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
{
	struct kvm_memory_slot *memslot;

	memslot = gfn_to_memslot(kvm, gfn);
2059
	mark_page_dirty_in_slot(memslot, gfn);
2060
}
2061
EXPORT_SYMBOL_GPL(mark_page_dirty);
2062

2063 2064 2065 2066 2067 2068 2069 2070 2071
void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn)
{
	struct kvm_memory_slot *memslot;

	memslot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);
	mark_page_dirty_in_slot(memslot, gfn);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_mark_page_dirty);

2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094
void kvm_sigset_activate(struct kvm_vcpu *vcpu)
{
	if (!vcpu->sigset_active)
		return;

	/*
	 * This does a lockless modification of ->real_blocked, which is fine
	 * because, only current can change ->real_blocked and all readers of
	 * ->real_blocked don't care as long ->real_blocked is always a subset
	 * of ->blocked.
	 */
	sigprocmask(SIG_SETMASK, &vcpu->sigset, &current->real_blocked);
}

void kvm_sigset_deactivate(struct kvm_vcpu *vcpu)
{
	if (!vcpu->sigset_active)
		return;

	sigprocmask(SIG_SETMASK, &current->real_blocked, NULL);
	sigemptyset(&current->real_blocked);
}

W
Wanpeng Li 已提交
2095 2096
static void grow_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2097
	unsigned int old, val, grow;
W
Wanpeng Li 已提交
2098

2099
	old = val = vcpu->halt_poll_ns;
2100
	grow = READ_ONCE(halt_poll_ns_grow);
W
Wanpeng Li 已提交
2101
	/* 10us base */
2102
	if (val == 0 && grow)
W
Wanpeng Li 已提交
2103 2104
		val = 10000;
	else
2105
		val *= grow;
W
Wanpeng Li 已提交
2106

2107 2108 2109
	if (val > halt_poll_ns)
		val = halt_poll_ns;

W
Wanpeng Li 已提交
2110
	vcpu->halt_poll_ns = val;
2111
	trace_kvm_halt_poll_ns_grow(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2112 2113 2114 2115
}

static void shrink_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2116
	unsigned int old, val, shrink;
W
Wanpeng Li 已提交
2117

2118
	old = val = vcpu->halt_poll_ns;
2119 2120
	shrink = READ_ONCE(halt_poll_ns_shrink);
	if (shrink == 0)
W
Wanpeng Li 已提交
2121 2122
		val = 0;
	else
2123
		val /= shrink;
W
Wanpeng Li 已提交
2124 2125

	vcpu->halt_poll_ns = val;
2126
	trace_kvm_halt_poll_ns_shrink(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2127 2128
}

2129 2130
static int kvm_vcpu_check_block(struct kvm_vcpu *vcpu)
{
2131 2132 2133
	int ret = -EINTR;
	int idx = srcu_read_lock(&vcpu->kvm->srcu);

2134 2135
	if (kvm_arch_vcpu_runnable(vcpu)) {
		kvm_make_request(KVM_REQ_UNHALT, vcpu);
2136
		goto out;
2137 2138
	}
	if (kvm_cpu_has_pending_timer(vcpu))
2139
		goto out;
2140
	if (signal_pending(current))
2141
		goto out;
2142

2143 2144 2145 2146
	ret = 0;
out:
	srcu_read_unlock(&vcpu->kvm->srcu, idx);
	return ret;
2147 2148
}

E
Eddie Dong 已提交
2149 2150 2151
/*
 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
 */
2152
void kvm_vcpu_block(struct kvm_vcpu *vcpu)
2153
{
2154
	ktime_t start, cur;
2155
	DECLARE_SWAITQUEUE(wait);
2156
	bool waited = false;
W
Wanpeng Li 已提交
2157
	u64 block_ns;
2158 2159

	start = cur = ktime_get();
W
Wanpeng Li 已提交
2160 2161
	if (vcpu->halt_poll_ns) {
		ktime_t stop = ktime_add_ns(ktime_get(), vcpu->halt_poll_ns);
2162

2163
		++vcpu->stat.halt_attempted_poll;
2164 2165 2166 2167 2168 2169 2170
		do {
			/*
			 * This sets KVM_REQ_UNHALT if an interrupt
			 * arrives.
			 */
			if (kvm_vcpu_check_block(vcpu) < 0) {
				++vcpu->stat.halt_successful_poll;
2171 2172
				if (!vcpu_valid_wakeup(vcpu))
					++vcpu->stat.halt_poll_invalid;
2173 2174 2175 2176 2177
				goto out;
			}
			cur = ktime_get();
		} while (single_task_running() && ktime_before(cur, stop));
	}
2178

2179 2180
	kvm_arch_vcpu_blocking(vcpu);

2181
	for (;;) {
2182
		prepare_to_swait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
2183

2184
		if (kvm_vcpu_check_block(vcpu) < 0)
2185 2186
			break;

2187
		waited = true;
E
Eddie Dong 已提交
2188 2189
		schedule();
	}
2190

2191
	finish_swait(&vcpu->wq, &wait);
2192 2193
	cur = ktime_get();

2194
	kvm_arch_vcpu_unblocking(vcpu);
2195
out:
W
Wanpeng Li 已提交
2196 2197
	block_ns = ktime_to_ns(cur) - ktime_to_ns(start);

2198 2199 2200
	if (!vcpu_valid_wakeup(vcpu))
		shrink_halt_poll_ns(vcpu);
	else if (halt_poll_ns) {
W
Wanpeng Li 已提交
2201 2202 2203
		if (block_ns <= vcpu->halt_poll_ns)
			;
		/* we had a long block, shrink polling */
2204
		else if (vcpu->halt_poll_ns && block_ns > halt_poll_ns)
W
Wanpeng Li 已提交
2205 2206 2207 2208 2209
			shrink_halt_poll_ns(vcpu);
		/* we had a short halt and our poll time is too small */
		else if (vcpu->halt_poll_ns < halt_poll_ns &&
			block_ns < halt_poll_ns)
			grow_halt_poll_ns(vcpu);
2210 2211
	} else
		vcpu->halt_poll_ns = 0;
W
Wanpeng Li 已提交
2212

2213 2214
	trace_kvm_vcpu_wakeup(block_ns, waited, vcpu_valid_wakeup(vcpu));
	kvm_arch_vcpu_block_finish(vcpu);
E
Eddie Dong 已提交
2215
}
2216
EXPORT_SYMBOL_GPL(kvm_vcpu_block);
E
Eddie Dong 已提交
2217

2218
bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu)
2219
{
2220
	struct swait_queue_head *wqp;
2221 2222

	wqp = kvm_arch_vcpu_wq(vcpu);
2223
	if (swq_has_sleeper(wqp)) {
2224
		swake_up(wqp);
2225
		++vcpu->stat.halt_wakeup;
2226
		return true;
2227 2228
	}

2229
	return false;
2230 2231 2232
}
EXPORT_SYMBOL_GPL(kvm_vcpu_wake_up);

2233
#ifndef CONFIG_S390
2234 2235 2236 2237 2238 2239 2240 2241
/*
 * Kick a sleeping VCPU, or a guest VCPU in guest mode, into host kernel mode.
 */
void kvm_vcpu_kick(struct kvm_vcpu *vcpu)
{
	int me;
	int cpu = vcpu->cpu;

2242 2243 2244
	if (kvm_vcpu_wake_up(vcpu))
		return;

2245 2246 2247 2248 2249 2250
	me = get_cpu();
	if (cpu != me && (unsigned)cpu < nr_cpu_ids && cpu_online(cpu))
		if (kvm_arch_vcpu_should_kick(vcpu))
			smp_send_reschedule(cpu);
	put_cpu();
}
2251
EXPORT_SYMBOL_GPL(kvm_vcpu_kick);
2252
#endif /* !CONFIG_S390 */
2253

2254
int kvm_vcpu_yield_to(struct kvm_vcpu *target)
2255 2256 2257
{
	struct pid *pid;
	struct task_struct *task = NULL;
2258
	int ret = 0;
2259 2260 2261 2262

	rcu_read_lock();
	pid = rcu_dereference(target->pid);
	if (pid)
2263
		task = get_pid_task(pid, PIDTYPE_PID);
2264 2265
	rcu_read_unlock();
	if (!task)
2266 2267
		return ret;
	ret = yield_to(task, 1);
2268
	put_task_struct(task);
2269 2270

	return ret;
2271 2272 2273
}
EXPORT_SYMBOL_GPL(kvm_vcpu_yield_to);

2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295
/*
 * Helper that checks whether a VCPU is eligible for directed yield.
 * Most eligible candidate to yield is decided by following heuristics:
 *
 *  (a) VCPU which has not done pl-exit or cpu relax intercepted recently
 *  (preempted lock holder), indicated by @in_spin_loop.
 *  Set at the beiginning and cleared at the end of interception/PLE handler.
 *
 *  (b) VCPU which has done pl-exit/ cpu relax intercepted but did not get
 *  chance last time (mostly it has become eligible now since we have probably
 *  yielded to lockholder in last iteration. This is done by toggling
 *  @dy_eligible each time a VCPU checked for eligibility.)
 *
 *  Yielding to a recently pl-exited/cpu relax intercepted VCPU before yielding
 *  to preempted lock-holder could result in wrong VCPU selection and CPU
 *  burning. Giving priority for a potential lock-holder increases lock
 *  progress.
 *
 *  Since algorithm is based on heuristics, accessing another VCPU data without
 *  locking does not harm. It may result in trying to yield to  same VCPU, fail
 *  and continue with next VCPU and so on.
 */
2296
static bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
2297
{
2298
#ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
2299 2300 2301
	bool eligible;

	eligible = !vcpu->spin_loop.in_spin_loop ||
2302
		    vcpu->spin_loop.dy_eligible;
2303 2304 2305 2306 2307

	if (vcpu->spin_loop.in_spin_loop)
		kvm_vcpu_set_dy_eligible(vcpu, !vcpu->spin_loop.dy_eligible);

	return eligible;
2308 2309
#else
	return true;
2310
#endif
2311
}
2312

2313
void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
Z
Zhai, Edwin 已提交
2314
{
2315 2316 2317 2318
	struct kvm *kvm = me->kvm;
	struct kvm_vcpu *vcpu;
	int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
	int yielded = 0;
2319
	int try = 3;
2320 2321
	int pass;
	int i;
Z
Zhai, Edwin 已提交
2322

2323
	kvm_vcpu_set_in_spin_loop(me, true);
2324 2325 2326 2327 2328 2329 2330
	/*
	 * We boost the priority of a VCPU that is runnable but not
	 * currently running, because it got preempted by something
	 * else and called schedule in __vcpu_run.  Hopefully that
	 * VCPU is holding the lock that we need and will release it.
	 * We approximate round-robin by starting at the last boosted VCPU.
	 */
2331
	for (pass = 0; pass < 2 && !yielded && try; pass++) {
2332
		kvm_for_each_vcpu(i, vcpu, kvm) {
2333
			if (!pass && i <= last_boosted_vcpu) {
2334 2335 2336 2337
				i = last_boosted_vcpu;
				continue;
			} else if (pass && i > last_boosted_vcpu)
				break;
2338
			if (!READ_ONCE(vcpu->preempted))
2339
				continue;
2340 2341
			if (vcpu == me)
				continue;
2342
			if (swait_active(&vcpu->wq) && !kvm_arch_vcpu_runnable(vcpu))
2343
				continue;
2344 2345
			if (yield_to_kernel_mode && !kvm_arch_vcpu_in_kernel(vcpu))
				continue;
2346 2347
			if (!kvm_vcpu_eligible_for_directed_yield(vcpu))
				continue;
2348 2349 2350

			yielded = kvm_vcpu_yield_to(vcpu);
			if (yielded > 0) {
2351 2352
				kvm->last_boosted_vcpu = i;
				break;
2353 2354 2355 2356
			} else if (yielded < 0) {
				try--;
				if (!try)
					break;
2357 2358 2359
			}
		}
	}
2360
	kvm_vcpu_set_in_spin_loop(me, false);
2361 2362 2363

	/* Ensure vcpu is not eligible during next spinloop */
	kvm_vcpu_set_dy_eligible(me, false);
Z
Zhai, Edwin 已提交
2364 2365 2366
}
EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin);

2367
static vm_fault_t kvm_vcpu_fault(struct vm_fault *vmf)
2368
{
2369
	struct kvm_vcpu *vcpu = vmf->vma->vm_file->private_data;
2370 2371
	struct page *page;

2372
	if (vmf->pgoff == 0)
2373
		page = virt_to_page(vcpu->run);
A
Avi Kivity 已提交
2374
#ifdef CONFIG_X86
2375
	else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
2376
		page = virt_to_page(vcpu->arch.pio_data);
2377
#endif
2378
#ifdef CONFIG_KVM_MMIO
2379 2380
	else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
		page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
A
Avi Kivity 已提交
2381
#endif
2382
	else
2383
		return kvm_arch_vcpu_fault(vcpu, vmf);
2384
	get_page(page);
2385 2386
	vmf->page = page;
	return 0;
2387 2388
}

2389
static const struct vm_operations_struct kvm_vcpu_vm_ops = {
2390
	.fault = kvm_vcpu_fault,
2391 2392 2393 2394 2395 2396 2397 2398
};

static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma)
{
	vma->vm_ops = &kvm_vcpu_vm_ops;
	return 0;
}

A
Avi Kivity 已提交
2399 2400 2401 2402
static int kvm_vcpu_release(struct inode *inode, struct file *filp)
{
	struct kvm_vcpu *vcpu = filp->private_data;

2403
	debugfs_remove_recursive(vcpu->debugfs_dentry);
A
Al Viro 已提交
2404
	kvm_put_kvm(vcpu->kvm);
A
Avi Kivity 已提交
2405 2406 2407
	return 0;
}

2408
static struct file_operations kvm_vcpu_fops = {
A
Avi Kivity 已提交
2409 2410
	.release        = kvm_vcpu_release,
	.unlocked_ioctl = kvm_vcpu_ioctl,
2411
	.mmap           = kvm_vcpu_mmap,
2412
	.llseek		= noop_llseek,
2413
	KVM_COMPAT(kvm_vcpu_compat_ioctl),
A
Avi Kivity 已提交
2414 2415 2416 2417 2418 2419 2420
};

/*
 * Allocates an inode for the vcpu.
 */
static int create_vcpu_fd(struct kvm_vcpu *vcpu)
{
2421 2422 2423 2424
	char name[8 + 1 + ITOA_MAX_LEN + 1];

	snprintf(name, sizeof(name), "kvm-vcpu:%d", vcpu->vcpu_id);
	return anon_inode_getfd(name, &kvm_vcpu_fops, vcpu, O_RDWR | O_CLOEXEC);
A
Avi Kivity 已提交
2425 2426
}

2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452
static int kvm_create_vcpu_debugfs(struct kvm_vcpu *vcpu)
{
	char dir_name[ITOA_MAX_LEN * 2];
	int ret;

	if (!kvm_arch_has_vcpu_debugfs())
		return 0;

	if (!debugfs_initialized())
		return 0;

	snprintf(dir_name, sizeof(dir_name), "vcpu%d", vcpu->vcpu_id);
	vcpu->debugfs_dentry = debugfs_create_dir(dir_name,
								vcpu->kvm->debugfs_dentry);
	if (!vcpu->debugfs_dentry)
		return -ENOMEM;

	ret = kvm_arch_create_vcpu_debugfs(vcpu);
	if (ret < 0) {
		debugfs_remove_recursive(vcpu->debugfs_dentry);
		return ret;
	}

	return 0;
}

2453 2454 2455
/*
 * Creates some virtual cpus.  Good luck creating more than one.
 */
2456
static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
2457 2458
{
	int r;
2459
	struct kvm_vcpu *vcpu;
2460

G
Greg Kurz 已提交
2461
	if (id >= KVM_MAX_VCPU_ID)
2462 2463
		return -EINVAL;

2464 2465 2466 2467 2468 2469 2470 2471 2472
	mutex_lock(&kvm->lock);
	if (kvm->created_vcpus == KVM_MAX_VCPUS) {
		mutex_unlock(&kvm->lock);
		return -EINVAL;
	}

	kvm->created_vcpus++;
	mutex_unlock(&kvm->lock);

2473
	vcpu = kvm_arch_vcpu_create(kvm, id);
2474 2475 2476 2477
	if (IS_ERR(vcpu)) {
		r = PTR_ERR(vcpu);
		goto vcpu_decrement;
	}
2478

2479 2480
	preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);

2481 2482
	r = kvm_arch_vcpu_setup(vcpu);
	if (r)
2483
		goto vcpu_destroy;
2484

2485 2486 2487 2488
	r = kvm_create_vcpu_debugfs(vcpu);
	if (r)
		goto vcpu_destroy;

S
Shaohua Li 已提交
2489
	mutex_lock(&kvm->lock);
2490 2491 2492 2493
	if (kvm_get_vcpu_by_id(kvm, id)) {
		r = -EEXIST;
		goto unlock_vcpu_destroy;
	}
2494 2495

	BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
2496

R
Rusty Russell 已提交
2497
	/* Now it's all set up, let userspace reach it */
A
Al Viro 已提交
2498
	kvm_get_kvm(kvm);
A
Avi Kivity 已提交
2499
	r = create_vcpu_fd(vcpu);
2500 2501
	if (r < 0) {
		kvm_put_kvm(kvm);
2502
		goto unlock_vcpu_destroy;
2503 2504 2505
	}

	kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu;
2506 2507 2508 2509 2510

	/*
	 * Pairs with smp_rmb() in kvm_get_vcpu.  Write kvm->vcpus
	 * before kvm->online_vcpu's incremented value.
	 */
2511 2512 2513 2514
	smp_wmb();
	atomic_inc(&kvm->online_vcpus);

	mutex_unlock(&kvm->lock);
2515
	kvm_arch_vcpu_postcreate(vcpu);
R
Rusty Russell 已提交
2516
	return r;
2517

2518
unlock_vcpu_destroy:
2519
	mutex_unlock(&kvm->lock);
2520
	debugfs_remove_recursive(vcpu->debugfs_dentry);
2521
vcpu_destroy:
2522
	kvm_arch_vcpu_destroy(vcpu);
2523 2524 2525 2526
vcpu_decrement:
	mutex_lock(&kvm->lock);
	kvm->created_vcpus--;
	mutex_unlock(&kvm->lock);
2527 2528 2529
	return r;
}

A
Avi Kivity 已提交
2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540
static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset)
{
	if (sigset) {
		sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP));
		vcpu->sigset_active = 1;
		vcpu->sigset = *sigset;
	} else
		vcpu->sigset_active = 0;
	return 0;
}

A
Avi Kivity 已提交
2541 2542
static long kvm_vcpu_ioctl(struct file *filp,
			   unsigned int ioctl, unsigned long arg)
A
Avi Kivity 已提交
2543
{
A
Avi Kivity 已提交
2544
	struct kvm_vcpu *vcpu = filp->private_data;
A
Al Viro 已提交
2545
	void __user *argp = (void __user *)arg;
2546
	int r;
2547 2548
	struct kvm_fpu *fpu = NULL;
	struct kvm_sregs *kvm_sregs = NULL;
A
Avi Kivity 已提交
2549

2550 2551
	if (vcpu->kvm->mm != current->mm)
		return -EIO;
2552

2553 2554 2555
	if (unlikely(_IOC_TYPE(ioctl) != KVMIO))
		return -EINVAL;

2556
	/*
2557 2558
	 * Some architectures have vcpu ioctls that are asynchronous to vcpu
	 * execution; mutex_lock() would break them.
2559
	 */
2560 2561
	r = kvm_arch_vcpu_async_ioctl(filp, ioctl, arg);
	if (r != -ENOIOCTLCMD)
2562
		return r;
2563

2564 2565
	if (mutex_lock_killable(&vcpu->mutex))
		return -EINTR;
A
Avi Kivity 已提交
2566
	switch (ioctl) {
2567 2568
	case KVM_RUN: {
		struct pid *oldpid;
2569 2570 2571
		r = -EINVAL;
		if (arg)
			goto out;
2572 2573
		oldpid = rcu_access_pointer(vcpu->pid);
		if (unlikely(oldpid != current->pids[PIDTYPE_PID].pid)) {
2574
			/* The thread running this VCPU changed. */
2575
			struct pid *newpid;
2576

2577 2578 2579 2580 2581
			r = kvm_arch_vcpu_run_pid_change(vcpu);
			if (r)
				break;

			newpid = get_task_pid(current, PIDTYPE_PID);
2582 2583 2584 2585 2586
			rcu_assign_pointer(vcpu->pid, newpid);
			if (oldpid)
				synchronize_rcu();
			put_pid(oldpid);
		}
2587
		r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
2588
		trace_kvm_userspace_exit(vcpu->run->exit_reason, r);
A
Avi Kivity 已提交
2589
		break;
2590
	}
A
Avi Kivity 已提交
2591
	case KVM_GET_REGS: {
2592
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
2593

2594 2595 2596
		r = -ENOMEM;
		kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
		if (!kvm_regs)
A
Avi Kivity 已提交
2597
			goto out;
2598 2599 2600
		r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
		if (r)
			goto out_free1;
A
Avi Kivity 已提交
2601
		r = -EFAULT;
2602 2603
		if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
			goto out_free1;
A
Avi Kivity 已提交
2604
		r = 0;
2605 2606
out_free1:
		kfree(kvm_regs);
A
Avi Kivity 已提交
2607 2608 2609
		break;
	}
	case KVM_SET_REGS: {
2610
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
2611

2612
		r = -ENOMEM;
2613 2614 2615
		kvm_regs = memdup_user(argp, sizeof(*kvm_regs));
		if (IS_ERR(kvm_regs)) {
			r = PTR_ERR(kvm_regs);
A
Avi Kivity 已提交
2616
			goto out;
2617
		}
2618 2619
		r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
		kfree(kvm_regs);
A
Avi Kivity 已提交
2620 2621 2622
		break;
	}
	case KVM_GET_SREGS: {
2623 2624 2625 2626 2627
		kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL);
		r = -ENOMEM;
		if (!kvm_sregs)
			goto out;
		r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
2628 2629 2630
		if (r)
			goto out;
		r = -EFAULT;
2631
		if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
A
Avi Kivity 已提交
2632 2633 2634 2635 2636
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_SREGS: {
2637 2638 2639
		kvm_sregs = memdup_user(argp, sizeof(*kvm_sregs));
		if (IS_ERR(kvm_sregs)) {
			r = PTR_ERR(kvm_sregs);
G
Guo Chao 已提交
2640
			kvm_sregs = NULL;
A
Avi Kivity 已提交
2641
			goto out;
2642
		}
2643
		r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
2644 2645
		break;
	}
2646 2647 2648 2649 2650 2651 2652
	case KVM_GET_MP_STATE: {
		struct kvm_mp_state mp_state;

		r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state);
		if (r)
			goto out;
		r = -EFAULT;
2653
		if (copy_to_user(argp, &mp_state, sizeof(mp_state)))
2654 2655 2656 2657 2658 2659 2660 2661
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_MP_STATE: {
		struct kvm_mp_state mp_state;

		r = -EFAULT;
2662
		if (copy_from_user(&mp_state, argp, sizeof(mp_state)))
2663 2664 2665 2666
			goto out;
		r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
		break;
	}
A
Avi Kivity 已提交
2667 2668 2669 2670
	case KVM_TRANSLATE: {
		struct kvm_translation tr;

		r = -EFAULT;
2671
		if (copy_from_user(&tr, argp, sizeof(tr)))
A
Avi Kivity 已提交
2672
			goto out;
2673
		r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
A
Avi Kivity 已提交
2674 2675 2676
		if (r)
			goto out;
		r = -EFAULT;
2677
		if (copy_to_user(argp, &tr, sizeof(tr)))
A
Avi Kivity 已提交
2678 2679 2680 2681
			goto out;
		r = 0;
		break;
	}
J
Jan Kiszka 已提交
2682 2683
	case KVM_SET_GUEST_DEBUG: {
		struct kvm_guest_debug dbg;
A
Avi Kivity 已提交
2684 2685

		r = -EFAULT;
2686
		if (copy_from_user(&dbg, argp, sizeof(dbg)))
A
Avi Kivity 已提交
2687
			goto out;
J
Jan Kiszka 已提交
2688
		r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
A
Avi Kivity 已提交
2689 2690
		break;
	}
A
Avi Kivity 已提交
2691 2692 2693 2694 2695 2696 2697 2698 2699
	case KVM_SET_SIGNAL_MASK: {
		struct kvm_signal_mask __user *sigmask_arg = argp;
		struct kvm_signal_mask kvm_sigmask;
		sigset_t sigset, *p;

		p = NULL;
		if (argp) {
			r = -EFAULT;
			if (copy_from_user(&kvm_sigmask, argp,
2700
					   sizeof(kvm_sigmask)))
A
Avi Kivity 已提交
2701 2702
				goto out;
			r = -EINVAL;
2703
			if (kvm_sigmask.len != sizeof(sigset))
A
Avi Kivity 已提交
2704 2705 2706
				goto out;
			r = -EFAULT;
			if (copy_from_user(&sigset, sigmask_arg->sigset,
2707
					   sizeof(sigset)))
A
Avi Kivity 已提交
2708 2709 2710
				goto out;
			p = &sigset;
		}
2711
		r = kvm_vcpu_ioctl_set_sigmask(vcpu, p);
A
Avi Kivity 已提交
2712 2713
		break;
	}
A
Avi Kivity 已提交
2714
	case KVM_GET_FPU: {
2715 2716 2717 2718 2719
		fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL);
		r = -ENOMEM;
		if (!fpu)
			goto out;
		r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu);
A
Avi Kivity 已提交
2720 2721 2722
		if (r)
			goto out;
		r = -EFAULT;
2723
		if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
A
Avi Kivity 已提交
2724 2725 2726 2727 2728
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_FPU: {
2729 2730 2731
		fpu = memdup_user(argp, sizeof(*fpu));
		if (IS_ERR(fpu)) {
			r = PTR_ERR(fpu);
G
Guo Chao 已提交
2732
			fpu = NULL;
A
Avi Kivity 已提交
2733
			goto out;
2734
		}
2735
		r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
A
Avi Kivity 已提交
2736 2737
		break;
	}
A
Avi Kivity 已提交
2738
	default:
2739
		r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
2740 2741
	}
out:
2742
	mutex_unlock(&vcpu->mutex);
2743 2744
	kfree(fpu);
	kfree(kvm_sregs);
A
Avi Kivity 已提交
2745 2746 2747
	return r;
}

2748
#ifdef CONFIG_KVM_COMPAT
2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767
static long kvm_vcpu_compat_ioctl(struct file *filp,
				  unsigned int ioctl, unsigned long arg)
{
	struct kvm_vcpu *vcpu = filp->private_data;
	void __user *argp = compat_ptr(arg);
	int r;

	if (vcpu->kvm->mm != current->mm)
		return -EIO;

	switch (ioctl) {
	case KVM_SET_SIGNAL_MASK: {
		struct kvm_signal_mask __user *sigmask_arg = argp;
		struct kvm_signal_mask kvm_sigmask;
		sigset_t sigset;

		if (argp) {
			r = -EFAULT;
			if (copy_from_user(&kvm_sigmask, argp,
2768
					   sizeof(kvm_sigmask)))
2769 2770
				goto out;
			r = -EINVAL;
A
Al Viro 已提交
2771
			if (kvm_sigmask.len != sizeof(compat_sigset_t))
2772 2773
				goto out;
			r = -EFAULT;
A
Al Viro 已提交
2774
			if (get_compat_sigset(&sigset, (void *)sigmask_arg->sigset))
2775
				goto out;
2776 2777 2778
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
		} else
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, NULL);
2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789
		break;
	}
	default:
		r = kvm_vcpu_ioctl(filp, ioctl, arg);
	}

out:
	return r;
}
#endif

S
Scott Wood 已提交
2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837
static int kvm_device_ioctl_attr(struct kvm_device *dev,
				 int (*accessor)(struct kvm_device *dev,
						 struct kvm_device_attr *attr),
				 unsigned long arg)
{
	struct kvm_device_attr attr;

	if (!accessor)
		return -EPERM;

	if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
		return -EFAULT;

	return accessor(dev, &attr);
}

static long kvm_device_ioctl(struct file *filp, unsigned int ioctl,
			     unsigned long arg)
{
	struct kvm_device *dev = filp->private_data;

	switch (ioctl) {
	case KVM_SET_DEVICE_ATTR:
		return kvm_device_ioctl_attr(dev, dev->ops->set_attr, arg);
	case KVM_GET_DEVICE_ATTR:
		return kvm_device_ioctl_attr(dev, dev->ops->get_attr, arg);
	case KVM_HAS_DEVICE_ATTR:
		return kvm_device_ioctl_attr(dev, dev->ops->has_attr, arg);
	default:
		if (dev->ops->ioctl)
			return dev->ops->ioctl(dev, ioctl, arg);

		return -ENOTTY;
	}
}

static int kvm_device_release(struct inode *inode, struct file *filp)
{
	struct kvm_device *dev = filp->private_data;
	struct kvm *kvm = dev->kvm;

	kvm_put_kvm(kvm);
	return 0;
}

static const struct file_operations kvm_device_fops = {
	.unlocked_ioctl = kvm_device_ioctl,
	.release = kvm_device_release,
2838
	KVM_COMPAT(kvm_device_ioctl),
S
Scott Wood 已提交
2839 2840 2841 2842 2843 2844 2845 2846 2847 2848
};

struct kvm_device *kvm_device_from_filp(struct file *filp)
{
	if (filp->f_op != &kvm_device_fops)
		return NULL;

	return filp->private_data;
}

2849
static struct kvm_device_ops *kvm_device_ops_table[KVM_DEV_TYPE_MAX] = {
2850
#ifdef CONFIG_KVM_MPIC
2851 2852
	[KVM_DEV_TYPE_FSL_MPIC_20]	= &kvm_mpic_ops,
	[KVM_DEV_TYPE_FSL_MPIC_42]	= &kvm_mpic_ops,
2853
#endif
2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867
};

int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type)
{
	if (type >= ARRAY_SIZE(kvm_device_ops_table))
		return -ENOSPC;

	if (kvm_device_ops_table[type] != NULL)
		return -EEXIST;

	kvm_device_ops_table[type] = ops;
	return 0;
}

2868 2869 2870 2871 2872 2873
void kvm_unregister_device_ops(u32 type)
{
	if (kvm_device_ops_table[type] != NULL)
		kvm_device_ops_table[type] = NULL;
}

S
Scott Wood 已提交
2874 2875 2876 2877 2878 2879 2880 2881
static int kvm_ioctl_create_device(struct kvm *kvm,
				   struct kvm_create_device *cd)
{
	struct kvm_device_ops *ops = NULL;
	struct kvm_device *dev;
	bool test = cd->flags & KVM_CREATE_DEVICE_TEST;
	int ret;

2882 2883 2884 2885 2886
	if (cd->type >= ARRAY_SIZE(kvm_device_ops_table))
		return -ENODEV;

	ops = kvm_device_ops_table[cd->type];
	if (ops == NULL)
S
Scott Wood 已提交
2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898
		return -ENODEV;

	if (test)
		return 0;

	dev = kzalloc(sizeof(*dev), GFP_KERNEL);
	if (!dev)
		return -ENOMEM;

	dev->ops = ops;
	dev->kvm = kvm;

2899
	mutex_lock(&kvm->lock);
S
Scott Wood 已提交
2900 2901
	ret = ops->create(dev, cd->type);
	if (ret < 0) {
2902
		mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
2903 2904 2905
		kfree(dev);
		return ret;
	}
2906 2907
	list_add(&dev->vm_node, &kvm->devices);
	mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
2908

2909 2910 2911
	if (ops->init)
		ops->init(dev);

2912
	ret = anon_inode_getfd(ops->name, &kvm_device_fops, dev, O_RDWR | O_CLOEXEC);
S
Scott Wood 已提交
2913
	if (ret < 0) {
2914 2915 2916
		mutex_lock(&kvm->lock);
		list_del(&dev->vm_node);
		mutex_unlock(&kvm->lock);
2917
		ops->destroy(dev);
S
Scott Wood 已提交
2918 2919 2920 2921 2922 2923 2924 2925
		return ret;
	}

	kvm_get_kvm(kvm);
	cd->fd = ret;
	return 0;
}

2926 2927 2928 2929 2930 2931 2932 2933 2934 2935
static long kvm_vm_ioctl_check_extension_generic(struct kvm *kvm, long arg)
{
	switch (arg) {
	case KVM_CAP_USER_MEMORY:
	case KVM_CAP_DESTROY_MEMORY_REGION_WORKS:
	case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS:
	case KVM_CAP_INTERNAL_ERROR_DATA:
#ifdef CONFIG_HAVE_KVM_MSI
	case KVM_CAP_SIGNAL_MSI:
#endif
2936
#ifdef CONFIG_HAVE_KVM_IRQFD
2937
	case KVM_CAP_IRQFD:
2938 2939
	case KVM_CAP_IRQFD_RESAMPLE:
#endif
2940
	case KVM_CAP_IOEVENTFD_ANY_LENGTH:
2941 2942
	case KVM_CAP_CHECK_EXTENSION_VM:
		return 1;
2943
#ifdef CONFIG_KVM_MMIO
2944 2945 2946
	case KVM_CAP_COALESCED_MMIO:
		return KVM_COALESCED_MMIO_PAGE_OFFSET;
#endif
2947 2948 2949
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_CAP_IRQ_ROUTING:
		return KVM_MAX_IRQ_ROUTES;
2950 2951 2952 2953
#endif
#if KVM_ADDRESS_SPACE_NUM > 1
	case KVM_CAP_MULTI_ADDRESS_SPACE:
		return KVM_ADDRESS_SPACE_NUM;
2954
#endif
G
Greg Kurz 已提交
2955 2956
	case KVM_CAP_MAX_VCPU_ID:
		return KVM_MAX_VCPU_ID;
2957 2958 2959 2960 2961 2962
	default:
		break;
	}
	return kvm_vm_ioctl_check_extension(kvm, arg);
}

A
Avi Kivity 已提交
2963 2964 2965 2966 2967
static long kvm_vm_ioctl(struct file *filp,
			   unsigned int ioctl, unsigned long arg)
{
	struct kvm *kvm = filp->private_data;
	void __user *argp = (void __user *)arg;
2968
	int r;
A
Avi Kivity 已提交
2969

2970 2971
	if (kvm->mm != current->mm)
		return -EIO;
A
Avi Kivity 已提交
2972 2973 2974 2975
	switch (ioctl) {
	case KVM_CREATE_VCPU:
		r = kvm_vm_ioctl_create_vcpu(kvm, arg);
		break;
2976 2977 2978 2979 2980
	case KVM_SET_USER_MEMORY_REGION: {
		struct kvm_userspace_memory_region kvm_userspace_mem;

		r = -EFAULT;
		if (copy_from_user(&kvm_userspace_mem, argp,
2981
						sizeof(kvm_userspace_mem)))
2982 2983
			goto out;

2984
		r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem);
A
Avi Kivity 已提交
2985 2986 2987 2988 2989 2990
		break;
	}
	case KVM_GET_DIRTY_LOG: {
		struct kvm_dirty_log log;

		r = -EFAULT;
2991
		if (copy_from_user(&log, argp, sizeof(log)))
A
Avi Kivity 已提交
2992
			goto out;
2993
		r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
A
Avi Kivity 已提交
2994 2995
		break;
	}
2996
#ifdef CONFIG_KVM_MMIO
2997 2998
	case KVM_REGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
2999

3000
		r = -EFAULT;
3001
		if (copy_from_user(&zone, argp, sizeof(zone)))
3002 3003 3004 3005 3006 3007
			goto out;
		r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
		break;
	}
	case KVM_UNREGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3008

3009
		r = -EFAULT;
3010
		if (copy_from_user(&zone, argp, sizeof(zone)))
3011 3012 3013 3014 3015
			goto out;
		r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
		break;
	}
#endif
G
Gregory Haskins 已提交
3016 3017 3018 3019
	case KVM_IRQFD: {
		struct kvm_irqfd data;

		r = -EFAULT;
3020
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3021
			goto out;
3022
		r = kvm_irqfd(kvm, &data);
G
Gregory Haskins 已提交
3023 3024
		break;
	}
G
Gregory Haskins 已提交
3025 3026 3027 3028
	case KVM_IOEVENTFD: {
		struct kvm_ioeventfd data;

		r = -EFAULT;
3029
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3030 3031 3032 3033
			goto out;
		r = kvm_ioeventfd(kvm, &data);
		break;
	}
3034 3035 3036 3037 3038
#ifdef CONFIG_HAVE_KVM_MSI
	case KVM_SIGNAL_MSI: {
		struct kvm_msi msi;

		r = -EFAULT;
3039
		if (copy_from_user(&msi, argp, sizeof(msi)))
3040 3041 3042 3043
			goto out;
		r = kvm_send_userspace_msi(kvm, &msi);
		break;
	}
3044 3045 3046 3047 3048 3049 3050
#endif
#ifdef __KVM_HAVE_IRQ_LINE
	case KVM_IRQ_LINE_STATUS:
	case KVM_IRQ_LINE: {
		struct kvm_irq_level irq_event;

		r = -EFAULT;
3051
		if (copy_from_user(&irq_event, argp, sizeof(irq_event)))
3052 3053
			goto out;

3054 3055
		r = kvm_vm_ioctl_irq_line(kvm, &irq_event,
					ioctl == KVM_IRQ_LINE_STATUS);
3056 3057 3058 3059 3060
		if (r)
			goto out;

		r = -EFAULT;
		if (ioctl == KVM_IRQ_LINE_STATUS) {
3061
			if (copy_to_user(argp, &irq_event, sizeof(irq_event)))
3062 3063 3064 3065 3066 3067
				goto out;
		}

		r = 0;
		break;
	}
3068
#endif
3069 3070 3071 3072
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_SET_GSI_ROUTING: {
		struct kvm_irq_routing routing;
		struct kvm_irq_routing __user *urouting;
3073
		struct kvm_irq_routing_entry *entries = NULL;
3074 3075 3076 3077 3078

		r = -EFAULT;
		if (copy_from_user(&routing, argp, sizeof(routing)))
			goto out;
		r = -EINVAL;
3079 3080
		if (!kvm_arch_can_set_irq_routing(kvm))
			goto out;
3081
		if (routing.nr > KVM_MAX_IRQ_ROUTES)
3082 3083 3084
			goto out;
		if (routing.flags)
			goto out;
3085 3086
		if (routing.nr) {
			r = -ENOMEM;
3087 3088
			entries = vmalloc(array_size(sizeof(*entries),
						     routing.nr));
3089 3090 3091 3092 3093 3094 3095 3096
			if (!entries)
				goto out;
			r = -EFAULT;
			urouting = argp;
			if (copy_from_user(entries, urouting->entries,
					   routing.nr * sizeof(*entries)))
				goto out_free_irq_routing;
		}
3097 3098
		r = kvm_set_irq_routing(kvm, entries, routing.nr,
					routing.flags);
3099
out_free_irq_routing:
3100 3101 3102 3103
		vfree(entries);
		break;
	}
#endif /* CONFIG_HAVE_KVM_IRQ_ROUTING */
S
Scott Wood 已提交
3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121
	case KVM_CREATE_DEVICE: {
		struct kvm_create_device cd;

		r = -EFAULT;
		if (copy_from_user(&cd, argp, sizeof(cd)))
			goto out;

		r = kvm_ioctl_create_device(kvm, &cd);
		if (r)
			goto out;

		r = -EFAULT;
		if (copy_to_user(argp, &cd, sizeof(cd)))
			goto out;

		r = 0;
		break;
	}
3122 3123 3124
	case KVM_CHECK_EXTENSION:
		r = kvm_vm_ioctl_check_extension_generic(kvm, arg);
		break;
3125
	default:
3126
		r = kvm_arch_vm_ioctl(filp, ioctl, arg);
3127 3128 3129 3130 3131
	}
out:
	return r;
}

3132
#ifdef CONFIG_KVM_COMPAT
3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156
struct compat_kvm_dirty_log {
	__u32 slot;
	__u32 padding1;
	union {
		compat_uptr_t dirty_bitmap; /* one bit per page */
		__u64 padding2;
	};
};

static long kvm_vm_compat_ioctl(struct file *filp,
			   unsigned int ioctl, unsigned long arg)
{
	struct kvm *kvm = filp->private_data;
	int r;

	if (kvm->mm != current->mm)
		return -EIO;
	switch (ioctl) {
	case KVM_GET_DIRTY_LOG: {
		struct compat_kvm_dirty_log compat_log;
		struct kvm_dirty_log log;

		if (copy_from_user(&compat_log, (void __user *)arg,
				   sizeof(compat_log)))
3157
			return -EFAULT;
3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172
		log.slot	 = compat_log.slot;
		log.padding1	 = compat_log.padding1;
		log.padding2	 = compat_log.padding2;
		log.dirty_bitmap = compat_ptr(compat_log.dirty_bitmap);

		r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
		break;
	}
	default:
		r = kvm_vm_ioctl(filp, ioctl, arg);
	}
	return r;
}
#endif

3173
static struct file_operations kvm_vm_fops = {
3174 3175
	.release        = kvm_vm_release,
	.unlocked_ioctl = kvm_vm_ioctl,
3176
	.llseek		= noop_llseek,
3177
	KVM_COMPAT(kvm_vm_compat_ioctl),
3178 3179
};

3180
static int kvm_dev_ioctl_create_vm(unsigned long type)
3181
{
3182
	int r;
3183
	struct kvm *kvm;
3184
	struct file *file;
3185

3186
	kvm = kvm_create_vm(type);
3187 3188
	if (IS_ERR(kvm))
		return PTR_ERR(kvm);
3189
#ifdef CONFIG_KVM_MMIO
3190
	r = kvm_coalesced_mmio_init(kvm);
3191 3192
	if (r < 0)
		goto put_kvm;
3193
#endif
3194
	r = get_unused_fd_flags(O_CLOEXEC);
3195 3196 3197
	if (r < 0)
		goto put_kvm;

3198 3199 3200
	file = anon_inode_getfile("kvm-vm", &kvm_vm_fops, kvm, O_RDWR);
	if (IS_ERR(file)) {
		put_unused_fd(r);
3201 3202
		r = PTR_ERR(file);
		goto put_kvm;
3203
	}
3204

3205 3206 3207 3208 3209 3210
	/*
	 * Don't call kvm_put_kvm anymore at this point; file->f_op is
	 * already set, with ->release() being kvm_vm_release().  In error
	 * cases it will be called by the final fput(file) and will take
	 * care of doing kvm_put_kvm(kvm).
	 */
3211
	if (kvm_create_vm_debugfs(kvm, r) < 0) {
3212 3213
		put_unused_fd(r);
		fput(file);
3214 3215
		return -ENOMEM;
	}
3216
	kvm_uevent_notify_change(KVM_EVENT_CREATE_VM, kvm);
3217

3218
	fd_install(r, file);
3219
	return r;
3220 3221 3222 3223

put_kvm:
	kvm_put_kvm(kvm);
	return r;
3224 3225 3226 3227 3228
}

static long kvm_dev_ioctl(struct file *filp,
			  unsigned int ioctl, unsigned long arg)
{
3229
	long r = -EINVAL;
3230 3231 3232

	switch (ioctl) {
	case KVM_GET_API_VERSION:
3233 3234
		if (arg)
			goto out;
3235 3236 3237
		r = KVM_API_VERSION;
		break;
	case KVM_CREATE_VM:
3238
		r = kvm_dev_ioctl_create_vm(arg);
3239
		break;
3240
	case KVM_CHECK_EXTENSION:
3241
		r = kvm_vm_ioctl_check_extension_generic(NULL, arg);
3242
		break;
3243 3244 3245
	case KVM_GET_VCPU_MMAP_SIZE:
		if (arg)
			goto out;
3246 3247 3248
		r = PAGE_SIZE;     /* struct kvm_run */
#ifdef CONFIG_X86
		r += PAGE_SIZE;    /* pio data page */
3249
#endif
3250
#ifdef CONFIG_KVM_MMIO
3251
		r += PAGE_SIZE;    /* coalesced mmio ring page */
3252
#endif
3253
		break;
3254 3255 3256
	case KVM_TRACE_ENABLE:
	case KVM_TRACE_PAUSE:
	case KVM_TRACE_DISABLE:
3257
		r = -EOPNOTSUPP;
3258
		break;
A
Avi Kivity 已提交
3259
	default:
3260
		return kvm_arch_dev_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
3261 3262 3263 3264 3265 3266 3267
	}
out:
	return r;
}

static struct file_operations kvm_chardev_ops = {
	.unlocked_ioctl = kvm_dev_ioctl,
3268
	.llseek		= noop_llseek,
3269
	KVM_COMPAT(kvm_dev_ioctl),
A
Avi Kivity 已提交
3270 3271 3272
};

static struct miscdevice kvm_dev = {
A
Avi Kivity 已提交
3273
	KVM_MINOR,
A
Avi Kivity 已提交
3274 3275 3276 3277
	"kvm",
	&kvm_chardev_ops,
};

3278
static void hardware_enable_nolock(void *junk)
3279 3280
{
	int cpu = raw_smp_processor_id();
3281
	int r;
3282

3283
	if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
3284
		return;
3285

3286
	cpumask_set_cpu(cpu, cpus_hardware_enabled);
3287

3288
	r = kvm_arch_hardware_enable();
3289 3290 3291 3292

	if (r) {
		cpumask_clear_cpu(cpu, cpus_hardware_enabled);
		atomic_inc(&hardware_enable_failed);
X
Xiubo Li 已提交
3293
		pr_info("kvm: enabling virtualization on CPU%d failed\n", cpu);
3294
	}
3295 3296
}

3297
static int kvm_starting_cpu(unsigned int cpu)
3298
{
3299
	raw_spin_lock(&kvm_count_lock);
3300 3301
	if (kvm_usage_count)
		hardware_enable_nolock(NULL);
3302
	raw_spin_unlock(&kvm_count_lock);
3303
	return 0;
3304 3305 3306
}

static void hardware_disable_nolock(void *junk)
3307 3308 3309
{
	int cpu = raw_smp_processor_id();

3310
	if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
3311
		return;
3312
	cpumask_clear_cpu(cpu, cpus_hardware_enabled);
3313
	kvm_arch_hardware_disable();
3314 3315
}

3316
static int kvm_dying_cpu(unsigned int cpu)
3317
{
3318
	raw_spin_lock(&kvm_count_lock);
3319 3320
	if (kvm_usage_count)
		hardware_disable_nolock(NULL);
3321
	raw_spin_unlock(&kvm_count_lock);
3322
	return 0;
3323 3324
}

3325 3326 3327 3328 3329 3330
static void hardware_disable_all_nolock(void)
{
	BUG_ON(!kvm_usage_count);

	kvm_usage_count--;
	if (!kvm_usage_count)
3331
		on_each_cpu(hardware_disable_nolock, NULL, 1);
3332 3333 3334 3335
}

static void hardware_disable_all(void)
{
3336
	raw_spin_lock(&kvm_count_lock);
3337
	hardware_disable_all_nolock();
3338
	raw_spin_unlock(&kvm_count_lock);
3339 3340 3341 3342 3343 3344
}

static int hardware_enable_all(void)
{
	int r = 0;

3345
	raw_spin_lock(&kvm_count_lock);
3346 3347 3348 3349

	kvm_usage_count++;
	if (kvm_usage_count == 1) {
		atomic_set(&hardware_enable_failed, 0);
3350
		on_each_cpu(hardware_enable_nolock, NULL, 1);
3351 3352 3353 3354 3355 3356 3357

		if (atomic_read(&hardware_enable_failed)) {
			hardware_disable_all_nolock();
			r = -EBUSY;
		}
	}

3358
	raw_spin_unlock(&kvm_count_lock);
3359 3360 3361 3362

	return r;
}

3363
static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
M
Mike Day 已提交
3364
		      void *v)
3365
{
3366 3367 3368 3369 3370 3371
	/*
	 * Some (well, at least mine) BIOSes hang on reboot if
	 * in vmx root mode.
	 *
	 * And Intel TXT required VMX off for all cpu when system shutdown.
	 */
X
Xiubo Li 已提交
3372
	pr_info("kvm: exiting hardware virtualization\n");
3373
	kvm_rebooting = true;
3374
	on_each_cpu(hardware_disable_nolock, NULL, 1);
3375 3376 3377 3378 3379 3380 3381 3382
	return NOTIFY_OK;
}

static struct notifier_block kvm_reboot_notifier = {
	.notifier_call = kvm_reboot,
	.priority = 0,
};

M
Marcelo Tosatti 已提交
3383
static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
3384 3385 3386 3387
{
	int i;

	for (i = 0; i < bus->dev_count; i++) {
3388
		struct kvm_io_device *pos = bus->range[i].dev;
3389 3390 3391

		kvm_iodevice_destructor(pos);
	}
M
Marcelo Tosatti 已提交
3392
	kfree(bus);
3393 3394
}

3395
static inline int kvm_io_bus_cmp(const struct kvm_io_range *r1,
X
Xiubo Li 已提交
3396
				 const struct kvm_io_range *r2)
3397
{
J
Jason Wang 已提交
3398 3399 3400 3401
	gpa_t addr1 = r1->addr;
	gpa_t addr2 = r2->addr;

	if (addr1 < addr2)
3402
		return -1;
J
Jason Wang 已提交
3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414

	/* If r2->len == 0, match the exact address.  If r2->len != 0,
	 * accept any overlapping write.  Any order is acceptable for
	 * overlapping ranges, because kvm_io_bus_get_first_dev ensures
	 * we process all of them.
	 */
	if (r2->len) {
		addr1 += r1->len;
		addr2 += r2->len;
	}

	if (addr1 > addr2)
3415
		return 1;
J
Jason Wang 已提交
3416

3417 3418 3419
	return 0;
}

3420 3421
static int kvm_io_bus_sort_cmp(const void *p1, const void *p2)
{
3422
	return kvm_io_bus_cmp(p1, p2);
3423 3424
}

G
Geoff Levand 已提交
3425
static int kvm_io_bus_get_first_dev(struct kvm_io_bus *bus,
3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442
			     gpa_t addr, int len)
{
	struct kvm_io_range *range, key;
	int off;

	key = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};

	range = bsearch(&key, bus->range, bus->dev_count,
			sizeof(struct kvm_io_range), kvm_io_bus_sort_cmp);
	if (range == NULL)
		return -ENOENT;

	off = range - bus->range;

3443
	while (off > 0 && kvm_io_bus_cmp(&key, &bus->range[off-1]) == 0)
3444 3445 3446 3447 3448
		off--;

	return off;
}

3449
static int __kvm_io_bus_write(struct kvm_vcpu *vcpu, struct kvm_io_bus *bus,
C
Cornelia Huck 已提交
3450 3451 3452 3453 3454 3455 3456 3457 3458
			      struct kvm_io_range *range, const void *val)
{
	int idx;

	idx = kvm_io_bus_get_first_dev(bus, range->addr, range->len);
	if (idx < 0)
		return -EOPNOTSUPP;

	while (idx < bus->dev_count &&
3459
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
3460
		if (!kvm_iodevice_write(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
3461 3462 3463 3464 3465 3466 3467 3468
					range->len, val))
			return idx;
		idx++;
	}

	return -EOPNOTSUPP;
}

3469
/* kvm_io_bus_write - called under kvm->slots_lock */
3470
int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
3471
		     int len, const void *val)
3472
{
3473
	struct kvm_io_bus *bus;
3474
	struct kvm_io_range range;
C
Cornelia Huck 已提交
3475
	int r;
3476 3477 3478 3479 3480

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
3481

3482
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3483 3484
	if (!bus)
		return -ENOMEM;
3485
	r = __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3486 3487 3488 3489
	return r < 0 ? r : 0;
}

/* kvm_io_bus_write_cookie - called under kvm->slots_lock */
3490 3491
int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
			    gpa_t addr, int len, const void *val, long cookie)
C
Cornelia Huck 已提交
3492 3493 3494 3495 3496 3497 3498 3499 3500
{
	struct kvm_io_bus *bus;
	struct kvm_io_range range;

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};

3501
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3502 3503
	if (!bus)
		return -ENOMEM;
C
Cornelia Huck 已提交
3504 3505 3506

	/* First try the device referenced by cookie. */
	if ((cookie >= 0) && (cookie < bus->dev_count) &&
3507
	    (kvm_io_bus_cmp(&range, &bus->range[cookie]) == 0))
3508
		if (!kvm_iodevice_write(vcpu, bus->range[cookie].dev, addr, len,
C
Cornelia Huck 已提交
3509 3510 3511 3512 3513 3514 3515
					val))
			return cookie;

	/*
	 * cookie contained garbage; fall back to search and return the
	 * correct cookie value.
	 */
3516
	return __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3517 3518
}

3519 3520
static int __kvm_io_bus_read(struct kvm_vcpu *vcpu, struct kvm_io_bus *bus,
			     struct kvm_io_range *range, void *val)
C
Cornelia Huck 已提交
3521 3522 3523 3524
{
	int idx;

	idx = kvm_io_bus_get_first_dev(bus, range->addr, range->len);
3525 3526 3527 3528
	if (idx < 0)
		return -EOPNOTSUPP;

	while (idx < bus->dev_count &&
3529
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
3530
		if (!kvm_iodevice_read(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
3531 3532
				       range->len, val))
			return idx;
3533 3534 3535
		idx++;
	}

3536 3537
	return -EOPNOTSUPP;
}
3538
EXPORT_SYMBOL_GPL(kvm_io_bus_write);
3539

3540
/* kvm_io_bus_read - called under kvm->slots_lock */
3541
int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
M
Marcelo Tosatti 已提交
3542
		    int len, void *val)
3543
{
3544
	struct kvm_io_bus *bus;
3545
	struct kvm_io_range range;
C
Cornelia Huck 已提交
3546
	int r;
3547 3548 3549 3550 3551

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
M
Marcelo Tosatti 已提交
3552

3553
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3554 3555
	if (!bus)
		return -ENOMEM;
3556
	r = __kvm_io_bus_read(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3557 3558
	return r < 0 ? r : 0;
}
3559

3560

3561
/* Caller must hold slots_lock. */
3562 3563
int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
			    int len, struct kvm_io_device *dev)
3564
{
3565
	int i;
M
Marcelo Tosatti 已提交
3566
	struct kvm_io_bus *new_bus, *bus;
3567
	struct kvm_io_range range;
3568

3569
	bus = kvm_get_bus(kvm, bus_idx);
3570 3571 3572
	if (!bus)
		return -ENOMEM;

3573 3574
	/* exclude ioeventfd which is limited by maximum fd */
	if (bus->dev_count - bus->ioeventfd_count > NR_IOBUS_DEVS - 1)
3575
		return -ENOSPC;
3576

3577
	new_bus = kmalloc(sizeof(*bus) + ((bus->dev_count + 1) *
3578
			  sizeof(struct kvm_io_range)), GFP_KERNEL);
M
Marcelo Tosatti 已提交
3579 3580
	if (!new_bus)
		return -ENOMEM;
3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
		.dev = dev,
	};

	for (i = 0; i < bus->dev_count; i++)
		if (kvm_io_bus_cmp(&bus->range[i], &range) > 0)
			break;

	memcpy(new_bus, bus, sizeof(*bus) + i * sizeof(struct kvm_io_range));
	new_bus->dev_count++;
	new_bus->range[i] = range;
	memcpy(new_bus->range + i + 1, bus->range + i,
		(bus->dev_count - i) * sizeof(struct kvm_io_range));
M
Marcelo Tosatti 已提交
3597 3598 3599
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
3600 3601 3602 3603

	return 0;
}

3604
/* Caller must hold slots_lock. */
3605 3606
void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
			       struct kvm_io_device *dev)
3607
{
3608
	int i;
M
Marcelo Tosatti 已提交
3609
	struct kvm_io_bus *new_bus, *bus;
3610

3611
	bus = kvm_get_bus(kvm, bus_idx);
3612
	if (!bus)
3613
		return;
3614

3615 3616
	for (i = 0; i < bus->dev_count; i++)
		if (bus->range[i].dev == dev) {
3617 3618
			break;
		}
M
Marcelo Tosatti 已提交
3619

3620 3621
	if (i == bus->dev_count)
		return;
3622

3623
	new_bus = kmalloc(sizeof(*bus) + ((bus->dev_count - 1) *
3624
			  sizeof(struct kvm_io_range)), GFP_KERNEL);
3625 3626 3627 3628
	if (!new_bus)  {
		pr_err("kvm: failed to shrink bus, removing it completely\n");
		goto broken;
	}
3629 3630 3631 3632 3633

	memcpy(new_bus, bus, sizeof(*bus) + i * sizeof(struct kvm_io_range));
	new_bus->dev_count--;
	memcpy(new_bus->range + i, bus->range + i + 1,
	       (new_bus->dev_count - i) * sizeof(struct kvm_io_range));
M
Marcelo Tosatti 已提交
3634

3635
broken:
M
Marcelo Tosatti 已提交
3636 3637 3638
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
3639
	return;
3640 3641
}

3642 3643 3644 3645 3646 3647 3648 3649 3650 3651
struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx,
					 gpa_t addr)
{
	struct kvm_io_bus *bus;
	int dev_idx, srcu_idx;
	struct kvm_io_device *iodev = NULL;

	srcu_idx = srcu_read_lock(&kvm->srcu);

	bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu);
3652 3653
	if (!bus)
		goto out_unlock;
3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667

	dev_idx = kvm_io_bus_get_first_dev(bus, addr, 1);
	if (dev_idx < 0)
		goto out_unlock;

	iodev = bus->range[dev_idx].dev;

out_unlock:
	srcu_read_unlock(&kvm->srcu, srcu_idx);

	return iodev;
}
EXPORT_SYMBOL_GPL(kvm_io_bus_get_dev);

3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679
static int kvm_debugfs_open(struct inode *inode, struct file *file,
			   int (*get)(void *, u64 *), int (*set)(void *, u64),
			   const char *fmt)
{
	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)
					  inode->i_private;

	/* The debugfs files are a reference to the kvm struct which
	 * is still valid when kvm_destroy_vm is called.
	 * To avoid the race between open and the removal of the debugfs
	 * directory we test against the users count.
	 */
3680
	if (!refcount_inc_not_zero(&stat_data->kvm->users_count))
3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705
		return -ENOENT;

	if (simple_attr_open(inode, file, get, set, fmt)) {
		kvm_put_kvm(stat_data->kvm);
		return -ENOMEM;
	}

	return 0;
}

static int kvm_debugfs_release(struct inode *inode, struct file *file)
{
	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)
					  inode->i_private;

	simple_attr_release(inode, file);
	kvm_put_kvm(stat_data->kvm);

	return 0;
}

static int vm_stat_get_per_vm(void *data, u64 *val)
{
	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)data;

3706
	*val = *(ulong *)((void *)stat_data->kvm + stat_data->offset);
3707 3708 3709 3710

	return 0;
}

3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722
static int vm_stat_clear_per_vm(void *data, u64 val)
{
	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)data;

	if (val)
		return -EINVAL;

	*(ulong *)((void *)stat_data->kvm + stat_data->offset) = 0;

	return 0;
}

3723 3724 3725 3726
static int vm_stat_get_per_vm_open(struct inode *inode, struct file *file)
{
	__simple_attr_check_format("%llu\n", 0ull);
	return kvm_debugfs_open(inode, file, vm_stat_get_per_vm,
3727
				vm_stat_clear_per_vm, "%llu\n");
3728 3729 3730 3731 3732 3733 3734 3735
}

static const struct file_operations vm_stat_get_per_vm_fops = {
	.owner   = THIS_MODULE,
	.open    = vm_stat_get_per_vm_open,
	.release = kvm_debugfs_release,
	.read    = simple_attr_read,
	.write   = simple_attr_write,
3736
	.llseek  = no_llseek,
3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747
};

static int vcpu_stat_get_per_vm(void *data, u64 *val)
{
	int i;
	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)data;
	struct kvm_vcpu *vcpu;

	*val = 0;

	kvm_for_each_vcpu(i, vcpu, stat_data->kvm)
3748
		*val += *(u64 *)((void *)vcpu + stat_data->offset);
3749 3750 3751 3752

	return 0;
}

3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767
static int vcpu_stat_clear_per_vm(void *data, u64 val)
{
	int i;
	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)data;
	struct kvm_vcpu *vcpu;

	if (val)
		return -EINVAL;

	kvm_for_each_vcpu(i, vcpu, stat_data->kvm)
		*(u64 *)((void *)vcpu + stat_data->offset) = 0;

	return 0;
}

3768 3769 3770 3771
static int vcpu_stat_get_per_vm_open(struct inode *inode, struct file *file)
{
	__simple_attr_check_format("%llu\n", 0ull);
	return kvm_debugfs_open(inode, file, vcpu_stat_get_per_vm,
3772
				 vcpu_stat_clear_per_vm, "%llu\n");
3773 3774 3775 3776 3777 3778 3779 3780
}

static const struct file_operations vcpu_stat_get_per_vm_fops = {
	.owner   = THIS_MODULE,
	.open    = vcpu_stat_get_per_vm_open,
	.release = kvm_debugfs_release,
	.read    = simple_attr_read,
	.write   = simple_attr_write,
3781
	.llseek  = no_llseek,
3782 3783 3784 3785 3786 3787 3788
};

static const struct file_operations *stat_fops_per_vm[] = {
	[KVM_STAT_VCPU] = &vcpu_stat_get_per_vm_fops,
	[KVM_STAT_VM]   = &vm_stat_get_per_vm_fops,
};

3789
static int vm_stat_get(void *_offset, u64 *val)
3790 3791 3792
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
3793 3794
	struct kvm_stat_data stat_tmp = {.offset = offset};
	u64 tmp_val;
3795

3796
	*val = 0;
3797
	spin_lock(&kvm_lock);
3798 3799 3800 3801 3802
	list_for_each_entry(kvm, &vm_list, vm_list) {
		stat_tmp.kvm = kvm;
		vm_stat_get_per_vm((void *)&stat_tmp, &tmp_val);
		*val += tmp_val;
	}
3803
	spin_unlock(&kvm_lock);
3804
	return 0;
3805 3806
}

3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826
static int vm_stat_clear(void *_offset, u64 val)
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
	struct kvm_stat_data stat_tmp = {.offset = offset};

	if (val)
		return -EINVAL;

	spin_lock(&kvm_lock);
	list_for_each_entry(kvm, &vm_list, vm_list) {
		stat_tmp.kvm = kvm;
		vm_stat_clear_per_vm((void *)&stat_tmp, 0);
	}
	spin_unlock(&kvm_lock);

	return 0;
}

DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, vm_stat_clear, "%llu\n");
3827

3828
static int vcpu_stat_get(void *_offset, u64 *val)
A
Avi Kivity 已提交
3829 3830 3831
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
3832 3833
	struct kvm_stat_data stat_tmp = {.offset = offset};
	u64 tmp_val;
A
Avi Kivity 已提交
3834

3835
	*val = 0;
3836
	spin_lock(&kvm_lock);
3837 3838 3839 3840 3841
	list_for_each_entry(kvm, &vm_list, vm_list) {
		stat_tmp.kvm = kvm;
		vcpu_stat_get_per_vm((void *)&stat_tmp, &tmp_val);
		*val += tmp_val;
	}
3842
	spin_unlock(&kvm_lock);
3843
	return 0;
A
Avi Kivity 已提交
3844 3845
}

3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866
static int vcpu_stat_clear(void *_offset, u64 val)
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
	struct kvm_stat_data stat_tmp = {.offset = offset};

	if (val)
		return -EINVAL;

	spin_lock(&kvm_lock);
	list_for_each_entry(kvm, &vm_list, vm_list) {
		stat_tmp.kvm = kvm;
		vcpu_stat_clear_per_vm((void *)&stat_tmp, 0);
	}
	spin_unlock(&kvm_lock);

	return 0;
}

DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, vcpu_stat_clear,
			"%llu\n");
3867

3868
static const struct file_operations *stat_fops[] = {
3869 3870 3871
	[KVM_STAT_VCPU] = &vcpu_stat_fops,
	[KVM_STAT_VM]   = &vm_stat_fops,
};
A
Avi Kivity 已提交
3872

3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898
static void kvm_uevent_notify_change(unsigned int type, struct kvm *kvm)
{
	struct kobj_uevent_env *env;
	unsigned long long created, active;

	if (!kvm_dev.this_device || !kvm)
		return;

	spin_lock(&kvm_lock);
	if (type == KVM_EVENT_CREATE_VM) {
		kvm_createvm_count++;
		kvm_active_vms++;
	} else if (type == KVM_EVENT_DESTROY_VM) {
		kvm_active_vms--;
	}
	created = kvm_createvm_count;
	active = kvm_active_vms;
	spin_unlock(&kvm_lock);

	env = kzalloc(sizeof(*env), GFP_KERNEL);
	if (!env)
		return;

	add_uevent_var(env, "CREATED=%llu", created);
	add_uevent_var(env, "COUNT=%llu", active);

3899
	if (type == KVM_EVENT_CREATE_VM) {
3900
		add_uevent_var(env, "EVENT=create");
3901 3902
		kvm->userspace_pid = task_pid_nr(current);
	} else if (type == KVM_EVENT_DESTROY_VM) {
3903
		add_uevent_var(env, "EVENT=destroy");
3904 3905
	}
	add_uevent_var(env, "PID=%d", kvm->userspace_pid);
3906 3907

	if (kvm->debugfs_dentry) {
3908 3909 3910 3911 3912 3913 3914
		char *tmp, *p = kmalloc(PATH_MAX, GFP_KERNEL);

		if (p) {
			tmp = dentry_path_raw(kvm->debugfs_dentry, p, PATH_MAX);
			if (!IS_ERR(tmp))
				add_uevent_var(env, "STATS_PATH=%s", tmp);
			kfree(p);
3915 3916 3917 3918 3919 3920 3921 3922
		}
	}
	/* no need for checks, since we are adding at most only 5 keys */
	env->envp[env->envp_idx++] = NULL;
	kobject_uevent_env(&kvm_dev.this_device->kobj, KOBJ_CHANGE, env->envp);
	kfree(env);
}

3923
static void kvm_init_debug(void)
A
Avi Kivity 已提交
3924 3925 3926
{
	struct kvm_stats_debugfs_item *p;

3927
	kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
3928

3929 3930
	kvm_debugfs_num_entries = 0;
	for (p = debugfs_entries; p->name; ++p, kvm_debugfs_num_entries++) {
3931 3932 3933
		debugfs_create_file(p->name, 0644, kvm_debugfs_dir,
				    (void *)(long)p->offset,
				    stat_fops[p->kind]);
3934
	}
A
Avi Kivity 已提交
3935 3936
}

3937
static int kvm_suspend(void)
3938
{
3939
	if (kvm_usage_count)
3940
		hardware_disable_nolock(NULL);
3941 3942 3943
	return 0;
}

3944
static void kvm_resume(void)
3945
{
3946
	if (kvm_usage_count) {
3947
		WARN_ON(raw_spin_is_locked(&kvm_count_lock));
3948
		hardware_enable_nolock(NULL);
3949
	}
3950 3951
}

3952
static struct syscore_ops kvm_syscore_ops = {
3953 3954 3955 3956
	.suspend = kvm_suspend,
	.resume = kvm_resume,
};

3957 3958 3959 3960 3961 3962 3963 3964 3965
static inline
struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn)
{
	return container_of(pn, struct kvm_vcpu, preempt_notifier);
}

static void kvm_sched_in(struct preempt_notifier *pn, int cpu)
{
	struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
3966

3967 3968
	if (vcpu->preempted)
		vcpu->preempted = false;
3969

R
Radim Krčmář 已提交
3970 3971
	kvm_arch_sched_in(vcpu, cpu);

3972
	kvm_arch_vcpu_load(vcpu, cpu);
3973 3974 3975 3976 3977 3978 3979
}

static void kvm_sched_out(struct preempt_notifier *pn,
			  struct task_struct *next)
{
	struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);

3980 3981
	if (current->state == TASK_RUNNING)
		vcpu->preempted = true;
3982
	kvm_arch_vcpu_put(vcpu);
3983 3984
}

3985
int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
3986
		  struct module *module)
A
Avi Kivity 已提交
3987 3988
{
	int r;
Y
Yang, Sheng 已提交
3989
	int cpu;
A
Avi Kivity 已提交
3990

3991 3992
	r = kvm_arch_init(opaque);
	if (r)
3993
		goto out_fail;
3994

3995 3996 3997 3998
	/*
	 * kvm_arch_init makes sure there's at most one caller
	 * for architectures that support multiple implementations,
	 * like intel and amd on x86.
P
Paolo Bonzini 已提交
3999 4000
	 * kvm_arch_init must be called before kvm_irqfd_init to avoid creating
	 * conflicts in case kvm is already setup for another implementation.
4001
	 */
P
Paolo Bonzini 已提交
4002 4003 4004
	r = kvm_irqfd_init();
	if (r)
		goto out_irqfd;
4005

4006
	if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
4007 4008 4009 4010
		r = -ENOMEM;
		goto out_free_0;
	}

4011
	r = kvm_arch_hardware_setup();
A
Avi Kivity 已提交
4012
	if (r < 0)
4013
		goto out_free_0a;
A
Avi Kivity 已提交
4014

Y
Yang, Sheng 已提交
4015 4016
	for_each_online_cpu(cpu) {
		smp_call_function_single(cpu,
4017
				kvm_arch_check_processor_compat,
4018
				&r, 1);
Y
Yang, Sheng 已提交
4019
		if (r < 0)
4020
			goto out_free_1;
Y
Yang, Sheng 已提交
4021 4022
	}

T
Thomas Gleixner 已提交
4023
	r = cpuhp_setup_state_nocalls(CPUHP_AP_KVM_STARTING, "kvm/cpu:starting",
4024
				      kvm_starting_cpu, kvm_dying_cpu);
A
Avi Kivity 已提交
4025
	if (r)
4026
		goto out_free_2;
A
Avi Kivity 已提交
4027 4028
	register_reboot_notifier(&kvm_reboot_notifier);

4029
	/* A kmem cache lets us meet the alignment requirements of fx_save. */
4030 4031
	if (!vcpu_align)
		vcpu_align = __alignof__(struct kvm_vcpu);
4032 4033 4034 4035 4036 4037
	kvm_vcpu_cache =
		kmem_cache_create_usercopy("kvm_vcpu", vcpu_size, vcpu_align,
					   SLAB_ACCOUNT,
					   offsetof(struct kvm_vcpu, arch),
					   sizeof_field(struct kvm_vcpu, arch),
					   NULL);
4038 4039
	if (!kvm_vcpu_cache) {
		r = -ENOMEM;
4040
		goto out_free_3;
4041 4042
	}

4043 4044 4045 4046
	r = kvm_async_pf_init();
	if (r)
		goto out_free;

A
Avi Kivity 已提交
4047
	kvm_chardev_ops.owner = module;
4048 4049
	kvm_vm_fops.owner = module;
	kvm_vcpu_fops.owner = module;
A
Avi Kivity 已提交
4050 4051 4052

	r = misc_register(&kvm_dev);
	if (r) {
X
Xiubo Li 已提交
4053
		pr_err("kvm: misc device register failed\n");
4054
		goto out_unreg;
A
Avi Kivity 已提交
4055 4056
	}

4057 4058
	register_syscore_ops(&kvm_syscore_ops);

4059 4060 4061
	kvm_preempt_ops.sched_in = kvm_sched_in;
	kvm_preempt_ops.sched_out = kvm_sched_out;

4062
	kvm_init_debug();
4063

P
Paolo Bonzini 已提交
4064 4065 4066
	r = kvm_vfio_ops_init();
	WARN_ON(r);

4067
	return 0;
A
Avi Kivity 已提交
4068

4069 4070
out_unreg:
	kvm_async_pf_deinit();
A
Avi Kivity 已提交
4071
out_free:
4072
	kmem_cache_destroy(kvm_vcpu_cache);
4073
out_free_3:
A
Avi Kivity 已提交
4074
	unregister_reboot_notifier(&kvm_reboot_notifier);
4075
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4076 4077
out_free_2:
out_free_1:
4078
	kvm_arch_hardware_unsetup();
4079 4080
out_free_0a:
	free_cpumask_var(cpus_hardware_enabled);
4081
out_free_0:
4082
	kvm_irqfd_exit();
P
Paolo Bonzini 已提交
4083
out_irqfd:
4084 4085
	kvm_arch_exit();
out_fail:
A
Avi Kivity 已提交
4086 4087
	return r;
}
4088
EXPORT_SYMBOL_GPL(kvm_init);
A
Avi Kivity 已提交
4089

4090
void kvm_exit(void)
A
Avi Kivity 已提交
4091
{
4092
	debugfs_remove_recursive(kvm_debugfs_dir);
A
Avi Kivity 已提交
4093
	misc_deregister(&kvm_dev);
4094
	kmem_cache_destroy(kvm_vcpu_cache);
4095
	kvm_async_pf_deinit();
4096
	unregister_syscore_ops(&kvm_syscore_ops);
A
Avi Kivity 已提交
4097
	unregister_reboot_notifier(&kvm_reboot_notifier);
4098
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4099
	on_each_cpu(hardware_disable_nolock, NULL, 1);
4100
	kvm_arch_hardware_unsetup();
4101
	kvm_arch_exit();
4102
	kvm_irqfd_exit();
4103
	free_cpumask_var(cpus_hardware_enabled);
4104
	kvm_vfio_ops_exit();
A
Avi Kivity 已提交
4105
}
4106
EXPORT_SYMBOL_GPL(kvm_exit);