kvm_main.c 96.3 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
117 118 119
static long kvm_vcpu_compat_ioctl(struct file *file, unsigned int ioctl,
				  unsigned long arg);
#endif
120 121
static int hardware_enable_all(void);
static void hardware_disable_all(void);
A
Avi Kivity 已提交
122

M
Marcelo Tosatti 已提交
123
static void kvm_io_bus_destroy(struct kvm_io_bus *bus);
124

D
Dan Williams 已提交
125
static void kvm_release_pfn_dirty(kvm_pfn_t pfn);
126
static void mark_page_dirty_in_slot(struct kvm_memory_slot *memslot, gfn_t gfn);
M
Marcelo Tosatti 已提交
127

128
__visible bool kvm_rebooting;
129
EXPORT_SYMBOL_GPL(kvm_rebooting);
130

131 132
static bool largepages_enabled = true;

133 134 135 136 137 138
#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;

D
Dan Williams 已提交
139
bool kvm_is_reserved_pfn(kvm_pfn_t pfn)
B
Ben-Ami Yassour 已提交
140
{
141
	if (pfn_valid(pfn))
142
		return PageReserved(pfn_to_page(pfn));
B
Ben-Ami Yassour 已提交
143 144 145 146

	return true;
}

A
Avi Kivity 已提交
147 148 149
/*
 * Switches to specified vcpu, until a matching vcpu_put()
 */
150
int vcpu_load(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
151
{
152 153
	int cpu;

154 155
	if (mutex_lock_killable(&vcpu->mutex))
		return -EINTR;
156 157
	cpu = get_cpu();
	preempt_notifier_register(&vcpu->preempt_notifier);
158
	kvm_arch_vcpu_load(vcpu, cpu);
159
	put_cpu();
160
	return 0;
A
Avi Kivity 已提交
161
}
162
EXPORT_SYMBOL_GPL(vcpu_load);
A
Avi Kivity 已提交
163

164
void vcpu_put(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
165
{
166
	preempt_disable();
167
	kvm_arch_vcpu_put(vcpu);
168 169
	preempt_notifier_unregister(&vcpu->preempt_notifier);
	preempt_enable();
A
Avi Kivity 已提交
170 171
	mutex_unlock(&vcpu->mutex);
}
172
EXPORT_SYMBOL_GPL(vcpu_put);
A
Avi Kivity 已提交
173

174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191
/* 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;
}

192 193 194 195
static void ack_flush(void *_completed)
{
}

196 197 198 199 200 201 202 203 204 205 206 207
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;
}

208
bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req)
209
{
210
	int i, cpu, me;
211
	cpumask_var_t cpus;
212
	bool called;
213 214
	struct kvm_vcpu *vcpu;

215
	zalloc_cpumask_var(&cpus, GFP_ATOMIC);
216

217
	me = get_cpu();
218
	kvm_for_each_vcpu(i, vcpu, kvm) {
219
		kvm_make_request(req, vcpu);
220
		cpu = vcpu->cpu;
221

222 223
		if (!(req & KVM_REQUEST_NO_WAKEUP) && kvm_vcpu_wake_up(vcpu))
			continue;
224

225
		if (cpus != NULL && cpu != -1 && cpu != me &&
226
		    kvm_request_needs_ipi(vcpu, req))
227
			__cpumask_set_cpu(cpu, cpus);
228
	}
229
	called = kvm_kick_many_cpus(cpus, !!(req & KVM_REQUEST_WAIT));
230
	put_cpu();
231
	free_cpumask_var(cpus);
232
	return called;
233 234
}

235
#ifndef CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALL
236
void kvm_flush_remote_tlbs(struct kvm *kvm)
237
{
238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254
	/*
	 * 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.
	 */
255
	if (kvm_make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH))
256
		++kvm->stat.remote_tlb_flush;
257
	cmpxchg(&kvm->tlbs_dirty, dirty_count, 0);
258
}
259
EXPORT_SYMBOL_GPL(kvm_flush_remote_tlbs);
260
#endif
261

262 263
void kvm_reload_remote_mmus(struct kvm *kvm)
{
264
	kvm_make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD);
265
}
266

R
Rusty Russell 已提交
267 268 269 270 271 272 273 274 275
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;
276
	vcpu->pid = NULL;
277
	init_swait_queue_head(&vcpu->wq);
278
	kvm_async_pf_vcpu_init(vcpu);
R
Rusty Russell 已提交
279

280 281 282
	vcpu->pre_pcpu = -1;
	INIT_LIST_HEAD(&vcpu->blocked_vcpu_list);

R
Rusty Russell 已提交
283 284 285 286 287 288 289
	page = alloc_page(GFP_KERNEL | __GFP_ZERO);
	if (!page) {
		r = -ENOMEM;
		goto fail;
	}
	vcpu->run = page_address(page);

290 291
	kvm_vcpu_set_in_spin_loop(vcpu, false);
	kvm_vcpu_set_dy_eligible(vcpu, false);
292
	vcpu->preempted = false;
293

294
	r = kvm_arch_vcpu_init(vcpu);
R
Rusty Russell 已提交
295
	if (r < 0)
296
		goto fail_free_run;
R
Rusty Russell 已提交
297 298 299 300 301
	return 0;

fail_free_run:
	free_page((unsigned long)vcpu->run);
fail:
302
	return r;
R
Rusty Russell 已提交
303 304 305 306 307
}
EXPORT_SYMBOL_GPL(kvm_vcpu_init);

void kvm_vcpu_uninit(struct kvm_vcpu *vcpu)
{
308 309 310 311 312 313
	/*
	 * 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));
314
	kvm_arch_vcpu_uninit(vcpu);
R
Rusty Russell 已提交
315 316 317 318
	free_page((unsigned long)vcpu->run);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_uninit);

319 320 321 322 323 324 325 326 327 328 329
#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);
}

static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn,
					     struct mm_struct *mm,
					     unsigned long address)
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
330
	int need_tlb_flush, idx;
331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349

	/*
	 * When ->invalidate_page runs, the linux pte has been zapped
	 * already but the page is still allocated until
	 * ->invalidate_page returns. So if we increase the sequence
	 * here the kvm page fault will notice if the spte can't be
	 * established because the page is going to be freed. If
	 * instead the kvm page fault establishes the spte before
	 * ->invalidate_page runs, kvm_unmap_hva will release it
	 * before returning.
	 *
	 * The sequence increase only need to be seen at spin_unlock
	 * time, and not at spin_lock time.
	 *
	 * Increasing the sequence after the spin_unlock would be
	 * unsafe because the kvm page fault could then establish the
	 * pte after kvm_unmap_hva returned, without noticing the page
	 * is going to be freed.
	 */
350
	idx = srcu_read_lock(&kvm->srcu);
351
	spin_lock(&kvm->mmu_lock);
352

353
	kvm->mmu_notifier_seq++;
354
	need_tlb_flush = kvm_unmap_hva(kvm, address) | kvm->tlbs_dirty;
355 356 357 358
	/* we've to flush the tlb before the pages can be freed */
	if (need_tlb_flush)
		kvm_flush_remote_tlbs(kvm);

359
	spin_unlock(&kvm->mmu_lock);
360 361 362

	kvm_arch_mmu_notifier_invalidate_page(kvm, address);

363
	srcu_read_unlock(&kvm->srcu, idx);
364 365
}

366 367 368 369 370 371
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);
372
	int idx;
373

374
	idx = srcu_read_lock(&kvm->srcu);
375 376 377 378
	spin_lock(&kvm->mmu_lock);
	kvm->mmu_notifier_seq++;
	kvm_set_spte_hva(kvm, address, pte);
	spin_unlock(&kvm->mmu_lock);
379
	srcu_read_unlock(&kvm->srcu, idx);
380 381
}

382 383 384 385 386 387
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);
388
	int need_tlb_flush = 0, idx;
389

390
	idx = srcu_read_lock(&kvm->srcu);
391 392 393 394 395 396 397
	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++;
398
	need_tlb_flush = kvm_unmap_hva_range(kvm, start, end);
399
	need_tlb_flush |= kvm->tlbs_dirty;
400 401 402
	/* we've to flush the tlb before the pages can be freed */
	if (need_tlb_flush)
		kvm_flush_remote_tlbs(kvm);
403 404 405

	spin_unlock(&kvm->mmu_lock);
	srcu_read_unlock(&kvm->srcu, idx);
406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421
}

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++;
422
	smp_wmb();
423 424
	/*
	 * The above sequence increase must be visible before the
425 426
	 * below count decrease, which is ensured by the smp_wmb above
	 * in conjunction with the smp_rmb in mmu_notifier_retry().
427 428 429 430 431 432 433 434 435
	 */
	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
436 437
					      unsigned long start,
					      unsigned long end)
438 439
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
440
	int young, idx;
441

442
	idx = srcu_read_lock(&kvm->srcu);
443 444
	spin_lock(&kvm->mmu_lock);

A
Andres Lagar-Cavilla 已提交
445
	young = kvm_age_hva(kvm, start, end);
446 447 448
	if (young)
		kvm_flush_remote_tlbs(kvm);

449 450 451
	spin_unlock(&kvm->mmu_lock);
	srcu_read_unlock(&kvm->srcu, idx);

452 453 454
	return young;
}

455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484
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
485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500
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;
}

501 502 503 504
static void kvm_mmu_notifier_release(struct mmu_notifier *mn,
				     struct mm_struct *mm)
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
505 506 507
	int idx;

	idx = srcu_read_lock(&kvm->srcu);
508
	kvm_arch_flush_shadow_all(kvm);
509
	srcu_read_unlock(&kvm->srcu, idx);
510 511
}

512 513 514 515 516
static const struct mmu_notifier_ops kvm_mmu_notifier_ops = {
	.invalidate_page	= kvm_mmu_notifier_invalidate_page,
	.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,
517
	.clear_young		= kvm_mmu_notifier_clear_young,
A
Andrea Arcangeli 已提交
518
	.test_young		= kvm_mmu_notifier_test_young,
519
	.change_pte		= kvm_mmu_notifier_change_pte,
520
	.release		= kvm_mmu_notifier_release,
521
};
522 523 524 525 526 527 528 529 530 531 532 533 534 535

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;
}

536 537
#endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */

538
static struct kvm_memslots *kvm_alloc_memslots(void)
539 540
{
	int i;
541
	struct kvm_memslots *slots;
542

M
Michal Hocko 已提交
543
	slots = kvzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
544 545 546
	if (!slots)
		return NULL;

547
	for (i = 0; i < KVM_MEM_SLOTS_NUM; i++)
548
		slots->id_to_index[i] = slots->memslots[i].id = i;
549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586

	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);
587 588
}

589 590 591 592 593 594 595 596 597
static void kvm_destroy_vm_debugfs(struct kvm *kvm)
{
	int i;

	if (!kvm->debugfs_dentry)
		return;

	debugfs_remove_recursive(kvm->debugfs_dentry);

598 599 600 601 602
	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);
	}
603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633
}

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);
	kvm->debugfs_dentry = debugfs_create_dir(dir_name,
						 kvm_debugfs_dir);
	if (!kvm->debugfs_dentry)
		return -ENOMEM;

	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;
634
		if (!debugfs_create_file(p->name, 0644,
635 636 637 638 639 640 641 642
					 kvm->debugfs_dentry,
					 stat_data,
					 stat_fops_per_vm[p->kind]))
			return -ENOMEM;
	}
	return 0;
}

643
static struct kvm *kvm_create_vm(unsigned long type)
A
Avi Kivity 已提交
644
{
645 646
	int r, i;
	struct kvm *kvm = kvm_arch_alloc_vm();
A
Avi Kivity 已提交
647

648 649 650
	if (!kvm)
		return ERR_PTR(-ENOMEM);

651
	spin_lock_init(&kvm->mmu_lock);
V
Vegard Nossum 已提交
652
	mmgrab(current->mm);
653 654 655 656 657
	kvm->mm = current->mm;
	kvm_eventfd_init(kvm);
	mutex_init(&kvm->lock);
	mutex_init(&kvm->irq_lock);
	mutex_init(&kvm->slots_lock);
658
	refcount_set(&kvm->users_count, 1);
659 660
	INIT_LIST_HEAD(&kvm->devices);

661
	r = kvm_arch_init_vm(kvm, type);
662
	if (r)
663
		goto out_err_no_disable;
664 665 666

	r = hardware_enable_all();
	if (r)
667
		goto out_err_no_disable;
668

669
#ifdef CONFIG_HAVE_KVM_IRQFD
670
	INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list);
671
#endif
A
Avi Kivity 已提交
672

673 674
	BUILD_BUG_ON(KVM_MEM_SLOTS_NUM > SHRT_MAX);

675
	r = -ENOMEM;
676
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
677 678
		struct kvm_memslots *slots = kvm_alloc_memslots();
		if (!slots)
679
			goto out_err_no_srcu;
680 681 682 683 684 685 686
		/*
		 * 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);
687
	}
688

689
	if (init_srcu_struct(&kvm->srcu))
690 691 692
		goto out_err_no_srcu;
	if (init_srcu_struct(&kvm->irq_srcu))
		goto out_err_no_irq_srcu;
M
Marcelo Tosatti 已提交
693
	for (i = 0; i < KVM_NR_BUSES; i++) {
694 695
		rcu_assign_pointer(kvm->buses[i],
			kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL));
696
		if (!kvm->buses[i])
M
Marcelo Tosatti 已提交
697 698
			goto out_err;
	}
699

700 701 702 703
	r = kvm_init_mmu_notifier(kvm);
	if (r)
		goto out_err;

704
	spin_lock(&kvm_lock);
705
	list_add(&kvm->vm_list, &vm_list);
706
	spin_unlock(&kvm_lock);
707

708 709
	preempt_notifier_inc();

710
	return kvm;
711 712

out_err:
713 714
	cleanup_srcu_struct(&kvm->irq_srcu);
out_err_no_irq_srcu:
715
	cleanup_srcu_struct(&kvm->srcu);
716
out_err_no_srcu:
717
	hardware_disable_all();
718
out_err_no_disable:
M
Marcelo Tosatti 已提交
719
	for (i = 0; i < KVM_NR_BUSES; i++)
720
		kfree(kvm_get_bus(kvm, i));
721
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
722
		kvm_free_memslots(kvm, __kvm_memslots(kvm, i));
723
	kvm_arch_free_vm(kvm);
724
	mmdrop(current->mm);
725
	return ERR_PTR(r);
726 727
}

728 729
static void kvm_destroy_devices(struct kvm *kvm)
{
G
Geliang Tang 已提交
730
	struct kvm_device *dev, *tmp;
731

732 733 734 735 736
	/*
	 * 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
737 738
	list_for_each_entry_safe(dev, tmp, &kvm->devices, vm_node) {
		list_del(&dev->vm_node);
739 740 741 742
		dev->ops->destroy(dev);
	}
}

743 744
static void kvm_destroy_vm(struct kvm *kvm)
{
M
Marcelo Tosatti 已提交
745
	int i;
746 747
	struct mm_struct *mm = kvm->mm;

748
	kvm_uevent_notify_change(KVM_EVENT_DESTROY_VM, kvm);
749
	kvm_destroy_vm_debugfs(kvm);
750
	kvm_arch_sync_events(kvm);
751
	spin_lock(&kvm_lock);
752
	list_del(&kvm->vm_list);
753
	spin_unlock(&kvm_lock);
754
	kvm_free_irq_routing(kvm);
755
	for (i = 0; i < KVM_NR_BUSES; i++) {
756
		struct kvm_io_bus *bus = kvm_get_bus(kvm, i);
757 758 759

		if (bus)
			kvm_io_bus_destroy(bus);
760 761
		kvm->buses[i] = NULL;
	}
762
	kvm_coalesced_mmio_free(kvm);
763 764
#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
	mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
765
#else
766
	kvm_arch_flush_shadow_all(kvm);
767
#endif
768
	kvm_arch_destroy_vm(kvm);
769
	kvm_destroy_devices(kvm);
770
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
771
		kvm_free_memslots(kvm, __kvm_memslots(kvm, i));
772
	cleanup_srcu_struct(&kvm->irq_srcu);
773 774
	cleanup_srcu_struct(&kvm->srcu);
	kvm_arch_free_vm(kvm);
775
	preempt_notifier_dec();
776
	hardware_disable_all();
777
	mmdrop(mm);
778 779
}

I
Izik Eidus 已提交
780 781
void kvm_get_kvm(struct kvm *kvm)
{
782
	refcount_inc(&kvm->users_count);
I
Izik Eidus 已提交
783 784 785 786 787
}
EXPORT_SYMBOL_GPL(kvm_get_kvm);

void kvm_put_kvm(struct kvm *kvm)
{
788
	if (refcount_dec_and_test(&kvm->users_count))
I
Izik Eidus 已提交
789 790 791 792 793
		kvm_destroy_vm(kvm);
}
EXPORT_SYMBOL_GPL(kvm_put_kvm);


794 795 796 797
static int kvm_vm_release(struct inode *inode, struct file *filp)
{
	struct kvm *kvm = filp->private_data;

G
Gregory Haskins 已提交
798 799
	kvm_irqfd_release(kvm);

I
Izik Eidus 已提交
800
	kvm_put_kvm(kvm);
A
Avi Kivity 已提交
801 802 803
	return 0;
}

804 805
/*
 * Allocation size is twice as large as the actual dirty bitmap size.
806
 * See x86's kvm_vm_ioctl_get_dirty_log() why this is needed.
807
 */
808 809
static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot)
{
810
	unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot);
811

M
Michal Hocko 已提交
812
	memslot->dirty_bitmap = kvzalloc(dirty_bytes, GFP_KERNEL);
813 814 815 816 817 818
	if (!memslot->dirty_bitmap)
		return -ENOMEM;

	return 0;
}

819
/*
820 821 822 823
 * 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.
824
 */
825 826
static void update_memslots(struct kvm_memslots *slots,
			    struct kvm_memory_slot *new)
827
{
828 829
	int id = new->id;
	int i = slots->id_to_index[id];
830
	struct kvm_memory_slot *mslots = slots->memslots;
831

832
	WARN_ON(mslots[i].id != id);
833
	if (!new->npages) {
834
		WARN_ON(!mslots[i].npages);
835 836 837 838 839 840
		if (mslots[i].npages)
			slots->used_slots--;
	} else {
		if (!mslots[i].npages)
			slots->used_slots++;
	}
841

842
	while (i < KVM_MEM_SLOTS_NUM - 1 &&
843 844 845
	       new->base_gfn <= mslots[i + 1].base_gfn) {
		if (!mslots[i + 1].npages)
			break;
846 847 848 849
		mslots[i] = mslots[i + 1];
		slots->id_to_index[mslots[i].id] = i;
		i++;
	}
850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866

	/*
	 * 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--;
		}
867 868
	} else
		WARN_ON_ONCE(i != slots->used_slots);
869

870 871
	mslots[i] = *new;
	slots->id_to_index[mslots[i].id] = i;
872 873
}

874
static int check_memory_region_flags(const struct kvm_userspace_memory_region *mem)
875
{
X
Xiao Guangrong 已提交
876 877
	u32 valid_flags = KVM_MEM_LOG_DIRTY_PAGES;

878
#ifdef __KVM_HAVE_READONLY_MEM
X
Xiao Guangrong 已提交
879 880 881 882
	valid_flags |= KVM_MEM_READONLY;
#endif

	if (mem->flags & ~valid_flags)
883 884 885 886 887
		return -EINVAL;

	return 0;
}

888
static struct kvm_memslots *install_new_memslots(struct kvm *kvm,
889
		int as_id, struct kvm_memslots *slots)
890
{
891
	struct kvm_memslots *old_memslots = __kvm_memslots(kvm, as_id);
892

893 894 895 896 897 898 899
	/*
	 * 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;

900
	rcu_assign_pointer(kvm->memslots[as_id], slots);
901
	synchronize_srcu_expedited(&kvm->srcu);
902

903 904 905 906
	/*
	 * 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.
907 908 909 910 911 912
	 *
	 * 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, ...
913
	 */
914
	slots->generation += KVM_ADDRESS_SPACE_NUM * 2 - 1;
915

916
	kvm_arch_memslots_updated(kvm, slots);
917 918

	return old_memslots;
919 920
}

A
Avi Kivity 已提交
921 922 923 924 925
/*
 * Allocate some memory and give it an address in the guest physical address
 * space.
 *
 * Discontiguous memory is allowed, mostly for framebuffers.
926
 *
927
 * Must be called holding kvm->slots_lock for write.
A
Avi Kivity 已提交
928
 */
929
int __kvm_set_memory_region(struct kvm *kvm,
930
			    const struct kvm_userspace_memory_region *mem)
A
Avi Kivity 已提交
931
{
932
	int r;
A
Avi Kivity 已提交
933
	gfn_t base_gfn;
934
	unsigned long npages;
935
	struct kvm_memory_slot *slot;
A
Avi Kivity 已提交
936
	struct kvm_memory_slot old, new;
937
	struct kvm_memslots *slots = NULL, *old_memslots;
938
	int as_id, id;
939
	enum kvm_mr_change change;
A
Avi Kivity 已提交
940

941 942 943 944
	r = check_memory_region_flags(mem);
	if (r)
		goto out;

A
Avi Kivity 已提交
945
	r = -EINVAL;
946 947 948
	as_id = mem->slot >> 16;
	id = (u16)mem->slot;

A
Avi Kivity 已提交
949 950 951 952 953
	/* General sanity checks */
	if (mem->memory_size & (PAGE_SIZE - 1))
		goto out;
	if (mem->guest_phys_addr & (PAGE_SIZE - 1))
		goto out;
954
	/* We can read the guest memory with __xxx_user() later on. */
955
	if ((id < KVM_USER_MEM_SLOTS) &&
956
	    ((mem->userspace_addr & (PAGE_SIZE - 1)) ||
957 958 959
	     !access_ok(VERIFY_WRITE,
			(void __user *)(unsigned long)mem->userspace_addr,
			mem->memory_size)))
960
		goto out;
961
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_MEM_SLOTS_NUM)
A
Avi Kivity 已提交
962 963 964 965
		goto out;
	if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
		goto out;

966
	slot = id_to_memslot(__kvm_memslots(kvm, as_id), id);
A
Avi Kivity 已提交
967 968 969
	base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
	npages = mem->memory_size >> PAGE_SHIFT;

970 971 972
	if (npages > KVM_MEM_MAX_NR_PAGES)
		goto out;

973
	new = old = *slot;
A
Avi Kivity 已提交
974

975
	new.id = id;
A
Avi Kivity 已提交
976 977 978 979
	new.base_gfn = base_gfn;
	new.npages = npages;
	new.flags = mem->flags;

980 981 982 983 984
	if (npages) {
		if (!old.npages)
			change = KVM_MR_CREATE;
		else { /* Modify an existing slot. */
			if ((mem->userspace_addr != old.userspace_addr) ||
985 986
			    (npages != old.npages) ||
			    ((new.flags ^ old.flags) & KVM_MEM_READONLY))
987 988 989 990 991 992 993 994 995 996 997
				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;
			}
		}
998 999 1000 1001
	} else {
		if (!old.npages)
			goto out;

1002
		change = KVM_MR_DELETE;
1003 1004 1005
		new.base_gfn = 0;
		new.flags = 0;
	}
A
Avi Kivity 已提交
1006

1007
	if ((change == KVM_MR_CREATE) || (change == KVM_MR_MOVE)) {
1008 1009
		/* Check for overlaps */
		r = -EEXIST;
1010
		kvm_for_each_memslot(slot, __kvm_memslots(kvm, as_id)) {
1011
			if ((slot->id >= KVM_USER_MEM_SLOTS) ||
1012
			    (slot->id == id))
1013 1014 1015 1016 1017
				continue;
			if (!((base_gfn + npages <= slot->base_gfn) ||
			      (base_gfn >= slot->base_gfn + slot->npages)))
				goto out;
		}
A
Avi Kivity 已提交
1018 1019 1020 1021
	}

	/* Free page dirty bitmap if unneeded */
	if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
A
Al Viro 已提交
1022
		new.dirty_bitmap = NULL;
A
Avi Kivity 已提交
1023 1024

	r = -ENOMEM;
1025
	if (change == KVM_MR_CREATE) {
1026
		new.userspace_addr = mem->userspace_addr;
1027

1028
		if (kvm_arch_create_memslot(kvm, &new, npages))
1029
			goto out_free;
A
Avi Kivity 已提交
1030
	}
1031

A
Avi Kivity 已提交
1032 1033
	/* Allocate page dirty bitmap if needed */
	if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
1034
		if (kvm_create_dirty_bitmap(&new) < 0)
1035
			goto out_free;
A
Avi Kivity 已提交
1036 1037
	}

M
Michal Hocko 已提交
1038
	slots = kvzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
1039 1040
	if (!slots)
		goto out_free;
1041
	memcpy(slots, __kvm_memslots(kvm, as_id), sizeof(struct kvm_memslots));
1042

1043
	if ((change == KVM_MR_DELETE) || (change == KVM_MR_MOVE)) {
1044
		slot = id_to_memslot(slots, id);
1045 1046
		slot->flags |= KVM_MEMSLOT_INVALID;

1047
		old_memslots = install_new_memslots(kvm, as_id, slots);
1048

1049 1050
		/* From this point no new shadow pages pointing to a deleted,
		 * or moved, memslot will be created.
1051 1052
		 *
		 * validation of sp->gfn happens in:
1053 1054
		 *	- gfn_to_hva (kvm_read_guest, gfn_to_pfn)
		 *	- kvm_is_visible_gfn (mmu_check_roots)
1055
		 */
1056
		kvm_arch_flush_shadow_memslot(kvm, slot);
1057 1058 1059 1060 1061 1062

		/*
		 * 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.
		 */
1063
		slots = old_memslots;
1064
	}
1065

1066
	r = kvm_arch_prepare_memory_region(kvm, &new, mem, change);
1067
	if (r)
1068
		goto out_slots;
1069

1070
	/* actual memory is freed via old in kvm_free_memslot below */
1071
	if (change == KVM_MR_DELETE) {
1072
		new.dirty_bitmap = NULL;
1073
		memset(&new.arch, 0, sizeof(new.arch));
1074 1075
	}

1076
	update_memslots(slots, &new);
1077
	old_memslots = install_new_memslots(kvm, as_id, slots);
1078

1079
	kvm_arch_commit_memory_region(kvm, mem, &old, &new, change);
1080

1081
	kvm_free_memslot(kvm, &old, &new);
1082
	kvfree(old_memslots);
A
Avi Kivity 已提交
1083 1084
	return 0;

1085
out_slots:
1086
	kvfree(slots);
1087
out_free:
1088
	kvm_free_memslot(kvm, &new, &old);
A
Avi Kivity 已提交
1089 1090
out:
	return r;
1091
}
1092 1093 1094
EXPORT_SYMBOL_GPL(__kvm_set_memory_region);

int kvm_set_memory_region(struct kvm *kvm,
1095
			  const struct kvm_userspace_memory_region *mem)
1096 1097 1098
{
	int r;

1099
	mutex_lock(&kvm->slots_lock);
1100
	r = __kvm_set_memory_region(kvm, mem);
1101
	mutex_unlock(&kvm->slots_lock);
1102 1103
	return r;
}
1104 1105
EXPORT_SYMBOL_GPL(kvm_set_memory_region);

1106 1107
static int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
					  struct kvm_userspace_memory_region *mem)
1108
{
1109
	if ((u16)mem->slot >= KVM_USER_MEM_SLOTS)
1110
		return -EINVAL;
1111

1112
	return kvm_set_memory_region(kvm, mem);
A
Avi Kivity 已提交
1113 1114
}

1115 1116
int kvm_get_dirty_log(struct kvm *kvm,
			struct kvm_dirty_log *log, int *is_dirty)
A
Avi Kivity 已提交
1117
{
1118
	struct kvm_memslots *slots;
A
Avi Kivity 已提交
1119
	struct kvm_memory_slot *memslot;
1120
	int i, as_id, id;
1121
	unsigned long n;
A
Avi Kivity 已提交
1122 1123
	unsigned long any = 0;

1124 1125 1126
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1127
		return -EINVAL;
A
Avi Kivity 已提交
1128

1129 1130
	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
A
Avi Kivity 已提交
1131
	if (!memslot->dirty_bitmap)
1132
		return -ENOENT;
A
Avi Kivity 已提交
1133

1134
	n = kvm_dirty_bitmap_bytes(memslot);
A
Avi Kivity 已提交
1135

1136
	for (i = 0; !any && i < n/sizeof(long); ++i)
A
Avi Kivity 已提交
1137 1138 1139
		any = memslot->dirty_bitmap[i];

	if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
1140
		return -EFAULT;
A
Avi Kivity 已提交
1141

1142 1143
	if (any)
		*is_dirty = 1;
1144
	return 0;
A
Avi Kivity 已提交
1145
}
1146
EXPORT_SYMBOL_GPL(kvm_get_dirty_log);
A
Avi Kivity 已提交
1147

1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173
#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)
{
1174
	struct kvm_memslots *slots;
1175
	struct kvm_memory_slot *memslot;
1176
	int i, as_id, id;
1177 1178 1179 1180
	unsigned long n;
	unsigned long *dirty_bitmap;
	unsigned long *dirty_bitmap_buffer;

1181 1182 1183
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1184
		return -EINVAL;
1185

1186 1187
	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
1188 1189 1190

	dirty_bitmap = memslot->dirty_bitmap;
	if (!dirty_bitmap)
1191
		return -ENOENT;
1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211

	n = kvm_dirty_bitmap_bytes(memslot);

	dirty_bitmap_buffer = dirty_bitmap + n / sizeof(long);
	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;

1212 1213 1214 1215 1216
		if (mask) {
			offset = i * BITS_PER_LONG;
			kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot,
								offset, mask);
		}
1217 1218 1219 1220
	}

	spin_unlock(&kvm->mmu_lock);
	if (copy_to_user(log->dirty_bitmap, dirty_bitmap_buffer, n))
1221 1222
		return -EFAULT;
	return 0;
1223 1224 1225 1226
}
EXPORT_SYMBOL_GPL(kvm_get_dirty_log_protect);
#endif

1227 1228 1229 1230 1231
bool kvm_largepages_enabled(void)
{
	return largepages_enabled;
}

1232 1233 1234 1235 1236 1237
void kvm_disable_largepages(void)
{
	largepages_enabled = false;
}
EXPORT_SYMBOL_GPL(kvm_disable_largepages);

1238 1239 1240 1241
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
1242
EXPORT_SYMBOL_GPL(gfn_to_memslot);
A
Avi Kivity 已提交
1243

1244 1245 1246 1247 1248
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);
}

1249
bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
1250
{
1251
	struct kvm_memory_slot *memslot = gfn_to_memslot(kvm, gfn);
1252

1253
	if (!memslot || memslot->id >= KVM_USER_MEM_SLOTS ||
1254
	      memslot->flags & KVM_MEMSLOT_INVALID)
1255
		return false;
1256

1257
	return true;
1258 1259 1260
}
EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);

J
Joerg Roedel 已提交
1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284
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
1285 1286 1287 1288 1289 1290 1291
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
1292
{
1293
	if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
X
Xiao Guangrong 已提交
1294
		return KVM_HVA_ERR_BAD;
1295

X
Xiao Guangrong 已提交
1296 1297
	if (memslot_is_readonly(slot) && write)
		return KVM_HVA_ERR_RO_BAD;
1298 1299 1300 1301

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

X
Xiao Guangrong 已提交
1302
	return __gfn_to_hva_memslot(slot, gfn);
I
Izik Eidus 已提交
1303
}
1304

X
Xiao Guangrong 已提交
1305 1306 1307 1308
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
1309
}
1310

X
Xiao Guangrong 已提交
1311
unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot,
1312
					gfn_t gfn)
X
Xiao Guangrong 已提交
1313 1314 1315 1316 1317
{
	return gfn_to_hva_many(slot, gfn, NULL);
}
EXPORT_SYMBOL_GPL(gfn_to_hva_memslot);

1318 1319
unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
{
1320
	return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL);
1321
}
1322
EXPORT_SYMBOL_GPL(gfn_to_hva);
I
Izik Eidus 已提交
1323

1324 1325 1326 1327 1328 1329
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);

1330
/*
1331 1332
 * If writable is set to false, the hva returned by this function is only
 * allowed to be read.
1333
 */
1334 1335
unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot,
				      gfn_t gfn, bool *writable)
1336
{
1337 1338 1339
	unsigned long hva = __gfn_to_hva_many(slot, gfn, NULL, false);

	if (!kvm_is_error_hva(hva) && writable)
1340 1341
		*writable = !memslot_is_readonly(slot);

1342
	return hva;
1343 1344
}

1345 1346 1347 1348 1349 1350 1351
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);
}

1352 1353 1354 1355 1356 1357 1358
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);
}

1359 1360
static int get_user_page_nowait(unsigned long start, int write,
		struct page **page)
1361
{
L
Lorenzo Stoakes 已提交
1362
	int flags = FOLL_NOWAIT | FOLL_HWPOISON;
1363 1364 1365 1366

	if (write)
		flags |= FOLL_WRITE;

L
Lorenzo Stoakes 已提交
1367
	return get_user_pages(start, 1, flags, page, NULL);
1368 1369
}

1370 1371
static inline int check_user_page_hwpoison(unsigned long addr)
{
L
Lorenzo Stoakes 已提交
1372
	int rc, flags = FOLL_HWPOISON | FOLL_WRITE;
1373

L
Lorenzo Stoakes 已提交
1374
	rc = get_user_pages(addr, 1, flags, NULL, NULL);
1375 1376 1377
	return rc == -EHWPOISON;
}

X
Xiao Guangrong 已提交
1378 1379 1380 1381 1382
/*
 * 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
1383
			    bool write_fault, bool *writable, kvm_pfn_t *pfn)
A
Avi Kivity 已提交
1384
{
1385
	struct page *page[1];
X
Xiao Guangrong 已提交
1386
	int npages;
A
Avi Kivity 已提交
1387

X
Xiao Guangrong 已提交
1388 1389
	if (!(async || atomic))
		return false;
1390

1391 1392 1393 1394 1395 1396 1397
	/*
	 * 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;
1398

X
Xiao Guangrong 已提交
1399 1400 1401
	npages = __get_user_pages_fast(addr, 1, 1, page);
	if (npages == 1) {
		*pfn = page_to_pfn(page[0]);
1402

X
Xiao Guangrong 已提交
1403 1404 1405 1406
		if (writable)
			*writable = true;
		return true;
	}
1407

X
Xiao Guangrong 已提交
1408 1409
	return false;
}
1410

X
Xiao Guangrong 已提交
1411 1412 1413 1414 1415
/*
 * 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
1416
			   bool *writable, kvm_pfn_t *pfn)
X
Xiao Guangrong 已提交
1417 1418 1419
{
	struct page *page[1];
	int npages = 0;
1420

X
Xiao Guangrong 已提交
1421 1422 1423 1424 1425 1426 1427
	might_sleep();

	if (writable)
		*writable = write_fault;

	if (async) {
		down_read(&current->mm->mmap_sem);
1428
		npages = get_user_page_nowait(addr, write_fault, page);
X
Xiao Guangrong 已提交
1429
		up_read(&current->mm->mmap_sem);
1430
	} else {
1431
		unsigned int flags = FOLL_HWPOISON;
1432 1433 1434 1435

		if (write_fault)
			flags |= FOLL_WRITE;

1436
		npages = get_user_pages_unlocked(addr, 1, page, flags);
1437
	}
X
Xiao Guangrong 已提交
1438 1439 1440 1441
	if (npages != 1)
		return npages;

	/* map read fault as writable if possible */
1442
	if (unlikely(!write_fault) && writable) {
X
Xiao Guangrong 已提交
1443 1444 1445 1446 1447 1448 1449
		struct page *wpage[1];

		npages = __get_user_pages_fast(addr, 1, 1, wpage);
		if (npages == 1) {
			*writable = true;
			put_page(page[0]);
			page[0] = wpage[0];
1450
		}
X
Xiao Guangrong 已提交
1451 1452

		npages = 1;
1453
	}
X
Xiao Guangrong 已提交
1454 1455 1456
	*pfn = page_to_pfn(page[0]);
	return npages;
}
I
Izik Eidus 已提交
1457

X
Xiao Guangrong 已提交
1458 1459 1460 1461
static bool vma_is_valid(struct vm_area_struct *vma, bool write_fault)
{
	if (unlikely(!(vma->vm_flags & VM_READ)))
		return false;
1462

X
Xiao Guangrong 已提交
1463 1464
	if (write_fault && (unlikely(!(vma->vm_flags & VM_WRITE))))
		return false;
1465

X
Xiao Guangrong 已提交
1466 1467
	return true;
}
1468

1469 1470 1471 1472
static int hva_to_pfn_remapped(struct vm_area_struct *vma,
			       unsigned long addr, bool *async,
			       bool write_fault, kvm_pfn_t *p_pfn)
{
1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511
	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;

	}


	/*
	 * 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;
1512 1513 1514
	return 0;
}

1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528
/*
 * 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
1529
static kvm_pfn_t hva_to_pfn(unsigned long addr, bool atomic, bool *async,
X
Xiao Guangrong 已提交
1530 1531 1532
			bool write_fault, bool *writable)
{
	struct vm_area_struct *vma;
D
Dan Williams 已提交
1533
	kvm_pfn_t pfn = 0;
1534
	int npages, r;
1535

X
Xiao Guangrong 已提交
1536 1537
	/* we can do it either atomically or asynchronously, not both */
	BUG_ON(atomic && async);
1538

X
Xiao Guangrong 已提交
1539 1540 1541 1542 1543 1544 1545 1546 1547
	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;
1548

X
Xiao Guangrong 已提交
1549 1550 1551 1552 1553 1554 1555
	down_read(&current->mm->mmap_sem);
	if (npages == -EHWPOISON ||
	      (!async && check_user_page_hwpoison(addr))) {
		pfn = KVM_PFN_ERR_HWPOISON;
		goto exit;
	}

1556
retry:
X
Xiao Guangrong 已提交
1557 1558 1559 1560
	vma = find_vma_intersection(current->mm, addr, addr + 1);

	if (vma == NULL)
		pfn = KVM_PFN_ERR_FAULT;
1561 1562
	else if (vma->vm_flags & (VM_IO | VM_PFNMAP)) {
		r = hva_to_pfn_remapped(vma, addr, async, write_fault, &pfn);
1563 1564
		if (r == -EAGAIN)
			goto retry;
1565 1566
		if (r < 0)
			pfn = KVM_PFN_ERR_FAULT;
X
Xiao Guangrong 已提交
1567
	} else {
X
Xiao Guangrong 已提交
1568
		if (async && vma_is_valid(vma, write_fault))
X
Xiao Guangrong 已提交
1569 1570 1571 1572 1573
			*async = true;
		pfn = KVM_PFN_ERR_FAULT;
	}
exit:
	up_read(&current->mm->mmap_sem);
1574
	return pfn;
1575 1576
}

D
Dan Williams 已提交
1577 1578 1579
kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
			       bool atomic, bool *async, bool write_fault,
			       bool *writable)
1580
{
X
Xiao Guangrong 已提交
1581 1582
	unsigned long addr = __gfn_to_hva_many(slot, gfn, NULL, write_fault);

1583 1584 1585
	if (addr == KVM_HVA_ERR_RO_BAD) {
		if (writable)
			*writable = false;
X
Xiao Guangrong 已提交
1586
		return KVM_PFN_ERR_RO_FAULT;
1587
	}
X
Xiao Guangrong 已提交
1588

1589 1590 1591
	if (kvm_is_error_hva(addr)) {
		if (writable)
			*writable = false;
1592
		return KVM_PFN_NOSLOT;
1593
	}
X
Xiao Guangrong 已提交
1594 1595 1596 1597 1598 1599 1600 1601 1602

	/* 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);
1603
}
1604
EXPORT_SYMBOL_GPL(__gfn_to_pfn_memslot);
1605

D
Dan Williams 已提交
1606
kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1607 1608
		      bool *writable)
{
P
Paolo Bonzini 已提交
1609 1610
	return __gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn, false, NULL,
				    write_fault, writable);
1611 1612 1613
}
EXPORT_SYMBOL_GPL(gfn_to_pfn_prot);

D
Dan Williams 已提交
1614
kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1615
{
X
Xiao Guangrong 已提交
1616
	return __gfn_to_pfn_memslot(slot, gfn, false, NULL, true, NULL);
1617
}
P
Paolo Bonzini 已提交
1618
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot);
1619

D
Dan Williams 已提交
1620
kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn)
1621
{
X
Xiao Guangrong 已提交
1622
	return __gfn_to_pfn_memslot(slot, gfn, true, NULL, true, NULL);
1623
}
1624
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot_atomic);
1625

D
Dan Williams 已提交
1626
kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1627 1628 1629 1630 1631
{
	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
1632
kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn)
1633 1634 1635 1636 1637
{
	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
1638
kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1639 1640 1641 1642 1643
{
	return gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn);
}
EXPORT_SYMBOL_GPL(gfn_to_pfn);

D
Dan Williams 已提交
1644
kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn)
1645 1646 1647 1648 1649
{
	return gfn_to_pfn_memslot(kvm_vcpu_gfn_to_memslot(vcpu, gfn), gfn);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_gfn_to_pfn);

1650 1651
int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
			    struct page **pages, int nr_pages)
1652 1653 1654 1655
{
	unsigned long addr;
	gfn_t entry;

1656
	addr = gfn_to_hva_many(slot, gfn, &entry);
1657 1658 1659 1660 1661 1662 1663 1664 1665 1666
	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
1667
static struct page *kvm_pfn_to_page(kvm_pfn_t pfn)
X
Xiao Guangrong 已提交
1668
{
1669
	if (is_error_noslot_pfn(pfn))
1670
		return KVM_ERR_PTR_BAD_PAGE;
X
Xiao Guangrong 已提交
1671

1672
	if (kvm_is_reserved_pfn(pfn)) {
1673
		WARN_ON(1);
1674
		return KVM_ERR_PTR_BAD_PAGE;
1675
	}
X
Xiao Guangrong 已提交
1676 1677 1678 1679

	return pfn_to_page(pfn);
}

1680 1681
struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
{
D
Dan Williams 已提交
1682
	kvm_pfn_t pfn;
1683 1684 1685

	pfn = gfn_to_pfn(kvm, gfn);

X
Xiao Guangrong 已提交
1686
	return kvm_pfn_to_page(pfn);
A
Avi Kivity 已提交
1687 1688 1689
}
EXPORT_SYMBOL_GPL(gfn_to_page);

1690 1691
struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn)
{
D
Dan Williams 已提交
1692
	kvm_pfn_t pfn;
1693 1694 1695 1696 1697 1698 1699

	pfn = kvm_vcpu_gfn_to_pfn(vcpu, gfn);

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

1700 1701
void kvm_release_page_clean(struct page *page)
{
1702 1703
	WARN_ON(is_error_page(page));

1704
	kvm_release_pfn_clean(page_to_pfn(page));
1705 1706 1707
}
EXPORT_SYMBOL_GPL(kvm_release_page_clean);

D
Dan Williams 已提交
1708
void kvm_release_pfn_clean(kvm_pfn_t pfn)
1709
{
1710
	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn))
1711
		put_page(pfn_to_page(pfn));
1712 1713 1714
}
EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);

1715
void kvm_release_page_dirty(struct page *page)
1716
{
X
Xiao Guangrong 已提交
1717 1718
	WARN_ON(is_error_page(page));

1719 1720 1721 1722
	kvm_release_pfn_dirty(page_to_pfn(page));
}
EXPORT_SYMBOL_GPL(kvm_release_page_dirty);

D
Dan Williams 已提交
1723
static void kvm_release_pfn_dirty(kvm_pfn_t pfn)
1724 1725 1726 1727 1728
{
	kvm_set_pfn_dirty(pfn);
	kvm_release_pfn_clean(pfn);
}

D
Dan Williams 已提交
1729
void kvm_set_pfn_dirty(kvm_pfn_t pfn)
1730
{
1731
	if (!kvm_is_reserved_pfn(pfn)) {
1732
		struct page *page = pfn_to_page(pfn);
1733

1734 1735 1736
		if (!PageReserved(page))
			SetPageDirty(page);
	}
1737
}
1738 1739
EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);

D
Dan Williams 已提交
1740
void kvm_set_pfn_accessed(kvm_pfn_t pfn)
1741
{
1742
	if (!kvm_is_reserved_pfn(pfn))
1743
		mark_page_accessed(pfn_to_page(pfn));
1744 1745 1746
}
EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);

D
Dan Williams 已提交
1747
void kvm_get_pfn(kvm_pfn_t pfn)
1748
{
1749
	if (!kvm_is_reserved_pfn(pfn))
1750
		get_page(pfn_to_page(pfn));
1751 1752
}
EXPORT_SYMBOL_GPL(kvm_get_pfn);
1753

1754 1755 1756 1757 1758 1759 1760 1761
static int next_segment(unsigned long len, int offset)
{
	if (len > PAGE_SIZE - offset)
		return PAGE_SIZE - offset;
	else
		return len;
}

1762 1763
static int __kvm_read_guest_page(struct kvm_memory_slot *slot, gfn_t gfn,
				 void *data, int offset, int len)
1764
{
1765 1766
	int r;
	unsigned long addr;
1767

1768
	addr = gfn_to_hva_memslot_prot(slot, gfn, NULL);
1769 1770
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1771
	r = __copy_from_user(data, (void __user *)addr + offset, len);
1772
	if (r)
1773 1774 1775
		return -EFAULT;
	return 0;
}
1776 1777 1778 1779 1780 1781 1782 1783

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);
}
1784 1785
EXPORT_SYMBOL_GPL(kvm_read_guest_page);

1786 1787 1788 1789 1790 1791 1792 1793 1794
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);

1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814
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);

1815
int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data, unsigned long len)
1816 1817
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
1818
	int seg;
1819
	int offset = offset_in_page(gpa);
1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833
	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);
1834

1835 1836 1837 1838 1839 1840 1841
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);
1842 1843
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1844
	pagefault_disable();
1845
	r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
1846
	pagefault_enable();
1847 1848 1849 1850 1851
	if (r)
		return -EFAULT;
	return 0;
}

1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875
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)
1876
{
1877 1878
	int r;
	unsigned long addr;
1879

1880
	addr = gfn_to_hva_memslot(memslot, gfn);
1881 1882
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1883
	r = __copy_to_user((void __user *)addr + offset, data, len);
1884
	if (r)
1885
		return -EFAULT;
1886
	mark_page_dirty_in_slot(memslot, gfn);
1887 1888
	return 0;
}
1889 1890 1891 1892 1893 1894 1895 1896

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);
}
1897 1898
EXPORT_SYMBOL_GPL(kvm_write_guest_page);

1899 1900 1901 1902 1903 1904 1905 1906 1907
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);

1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926
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;
}
1927
EXPORT_SYMBOL_GPL(kvm_write_guest);
1928

1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949
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);

1950 1951 1952
static int __kvm_gfn_to_hva_cache_init(struct kvm_memslots *slots,
				       struct gfn_to_hva_cache *ghc,
				       gpa_t gpa, unsigned long len)
1953 1954
{
	int offset = offset_in_page(gpa);
1955 1956 1957 1958
	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;
1959 1960 1961

	ghc->gpa = gpa;
	ghc->generation = slots->generation;
1962
	ghc->len = len;
1963
	ghc->memslot = __gfn_to_memslot(slots, start_gfn);
1964 1965
	ghc->hva = gfn_to_hva_many(ghc->memslot, start_gfn, NULL);
	if (!kvm_is_error_hva(ghc->hva) && nr_pages_needed <= 1) {
1966
		ghc->hva += offset;
1967 1968 1969 1970 1971 1972
	} else {
		/*
		 * If the requested region crosses two memslots, we still
		 * verify that the entire region is valid here.
		 */
		while (start_gfn <= end_gfn) {
1973
			ghc->memslot = __gfn_to_memslot(slots, start_gfn);
1974 1975 1976 1977 1978 1979 1980 1981 1982
			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;
	}
1983 1984
	return 0;
}
1985

1986
int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1987 1988
			      gpa_t gpa, unsigned long len)
{
1989
	struct kvm_memslots *slots = kvm_memslots(kvm);
1990 1991
	return __kvm_gfn_to_hva_cache_init(slots, ghc, gpa, len);
}
1992
EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init);
1993

1994 1995
int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, int offset, unsigned long len)
1996
{
1997
	struct kvm_memslots *slots = kvm_memslots(kvm);
1998
	int r;
1999
	gpa_t gpa = ghc->gpa + offset;
2000

2001
	BUG_ON(len + offset > ghc->len);
2002

2003
	if (slots->generation != ghc->generation)
2004
		__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len);
2005 2006

	if (unlikely(!ghc->memslot))
2007
		return kvm_write_guest(kvm, gpa, data, len);
2008 2009 2010 2011

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

2012
	r = __copy_to_user((void __user *)ghc->hva + offset, data, len);
2013 2014
	if (r)
		return -EFAULT;
2015
	mark_page_dirty_in_slot(ghc->memslot, gpa >> PAGE_SHIFT);
2016 2017 2018

	return 0;
}
2019
EXPORT_SYMBOL_GPL(kvm_write_guest_offset_cached);
2020

2021 2022
int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
2023
{
2024
	return kvm_write_guest_offset_cached(kvm, ghc, data, 0, len);
2025
}
2026
EXPORT_SYMBOL_GPL(kvm_write_guest_cached);
2027

2028 2029
int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
2030
{
2031
	struct kvm_memslots *slots = kvm_memslots(kvm);
2032 2033
	int r;

2034 2035
	BUG_ON(len > ghc->len);

2036
	if (slots->generation != ghc->generation)
2037
		__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len);
2038 2039

	if (unlikely(!ghc->memslot))
2040
		return kvm_read_guest(kvm, ghc->gpa, data, len);
2041 2042 2043 2044 2045 2046 2047 2048 2049 2050

	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;
}
2051
EXPORT_SYMBOL_GPL(kvm_read_guest_cached);
2052

2053 2054
int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
{
2055 2056 2057
	const void *zero_page = (const void *) __va(page_to_phys(ZERO_PAGE(0)));

	return kvm_write_guest_page(kvm, gfn, zero_page, offset, len);
2058 2059 2060 2061 2062 2063 2064 2065 2066 2067
}
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;

2068
	while ((seg = next_segment(len, offset)) != 0) {
2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079
		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);

2080
static void mark_page_dirty_in_slot(struct kvm_memory_slot *memslot,
2081
				    gfn_t gfn)
A
Avi Kivity 已提交
2082
{
R
Rusty Russell 已提交
2083 2084
	if (memslot && memslot->dirty_bitmap) {
		unsigned long rel_gfn = gfn - memslot->base_gfn;
A
Avi Kivity 已提交
2085

2086
		set_bit_le(rel_gfn, memslot->dirty_bitmap);
A
Avi Kivity 已提交
2087 2088 2089
	}
}

2090 2091 2092 2093 2094
void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
{
	struct kvm_memory_slot *memslot;

	memslot = gfn_to_memslot(kvm, gfn);
2095
	mark_page_dirty_in_slot(memslot, gfn);
2096
}
2097
EXPORT_SYMBOL_GPL(mark_page_dirty);
2098

2099 2100 2101 2102 2103 2104 2105 2106 2107
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);

W
Wanpeng Li 已提交
2108 2109
static void grow_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2110
	unsigned int old, val, grow;
W
Wanpeng Li 已提交
2111

2112
	old = val = vcpu->halt_poll_ns;
2113
	grow = READ_ONCE(halt_poll_ns_grow);
W
Wanpeng Li 已提交
2114
	/* 10us base */
2115
	if (val == 0 && grow)
W
Wanpeng Li 已提交
2116 2117
		val = 10000;
	else
2118
		val *= grow;
W
Wanpeng Li 已提交
2119

2120 2121 2122
	if (val > halt_poll_ns)
		val = halt_poll_ns;

W
Wanpeng Li 已提交
2123
	vcpu->halt_poll_ns = val;
2124
	trace_kvm_halt_poll_ns_grow(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2125 2126 2127 2128
}

static void shrink_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2129
	unsigned int old, val, shrink;
W
Wanpeng Li 已提交
2130

2131
	old = val = vcpu->halt_poll_ns;
2132 2133
	shrink = READ_ONCE(halt_poll_ns_shrink);
	if (shrink == 0)
W
Wanpeng Li 已提交
2134 2135
		val = 0;
	else
2136
		val /= shrink;
W
Wanpeng Li 已提交
2137 2138

	vcpu->halt_poll_ns = val;
2139
	trace_kvm_halt_poll_ns_shrink(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2140 2141
}

2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155
static int kvm_vcpu_check_block(struct kvm_vcpu *vcpu)
{
	if (kvm_arch_vcpu_runnable(vcpu)) {
		kvm_make_request(KVM_REQ_UNHALT, vcpu);
		return -EINTR;
	}
	if (kvm_cpu_has_pending_timer(vcpu))
		return -EINTR;
	if (signal_pending(current))
		return -EINTR;

	return 0;
}

E
Eddie Dong 已提交
2156 2157 2158
/*
 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
 */
2159
void kvm_vcpu_block(struct kvm_vcpu *vcpu)
2160
{
2161
	ktime_t start, cur;
2162
	DECLARE_SWAITQUEUE(wait);
2163
	bool waited = false;
W
Wanpeng Li 已提交
2164
	u64 block_ns;
2165 2166

	start = cur = ktime_get();
W
Wanpeng Li 已提交
2167 2168
	if (vcpu->halt_poll_ns) {
		ktime_t stop = ktime_add_ns(ktime_get(), vcpu->halt_poll_ns);
2169

2170
		++vcpu->stat.halt_attempted_poll;
2171 2172 2173 2174 2175 2176 2177
		do {
			/*
			 * This sets KVM_REQ_UNHALT if an interrupt
			 * arrives.
			 */
			if (kvm_vcpu_check_block(vcpu) < 0) {
				++vcpu->stat.halt_successful_poll;
2178 2179
				if (!vcpu_valid_wakeup(vcpu))
					++vcpu->stat.halt_poll_invalid;
2180 2181 2182 2183 2184
				goto out;
			}
			cur = ktime_get();
		} while (single_task_running() && ktime_before(cur, stop));
	}
2185

2186 2187
	kvm_arch_vcpu_blocking(vcpu);

2188
	for (;;) {
2189
		prepare_to_swait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
2190

2191
		if (kvm_vcpu_check_block(vcpu) < 0)
2192 2193
			break;

2194
		waited = true;
E
Eddie Dong 已提交
2195 2196
		schedule();
	}
2197

2198
	finish_swait(&vcpu->wq, &wait);
2199 2200
	cur = ktime_get();

2201
	kvm_arch_vcpu_unblocking(vcpu);
2202
out:
W
Wanpeng Li 已提交
2203 2204
	block_ns = ktime_to_ns(cur) - ktime_to_ns(start);

2205 2206 2207
	if (!vcpu_valid_wakeup(vcpu))
		shrink_halt_poll_ns(vcpu);
	else if (halt_poll_ns) {
W
Wanpeng Li 已提交
2208 2209 2210
		if (block_ns <= vcpu->halt_poll_ns)
			;
		/* we had a long block, shrink polling */
2211
		else if (vcpu->halt_poll_ns && block_ns > halt_poll_ns)
W
Wanpeng Li 已提交
2212 2213 2214 2215 2216
			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);
2217 2218
	} else
		vcpu->halt_poll_ns = 0;
W
Wanpeng Li 已提交
2219

2220 2221
	trace_kvm_vcpu_wakeup(block_ns, waited, vcpu_valid_wakeup(vcpu));
	kvm_arch_vcpu_block_finish(vcpu);
E
Eddie Dong 已提交
2222
}
2223
EXPORT_SYMBOL_GPL(kvm_vcpu_block);
E
Eddie Dong 已提交
2224

2225
bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu)
2226
{
2227
	struct swait_queue_head *wqp;
2228 2229

	wqp = kvm_arch_vcpu_wq(vcpu);
2230 2231
	if (swait_active(wqp)) {
		swake_up(wqp);
2232
		++vcpu->stat.halt_wakeup;
2233
		return true;
2234 2235
	}

2236
	return false;
2237 2238 2239
}
EXPORT_SYMBOL_GPL(kvm_vcpu_wake_up);

2240
#ifndef CONFIG_S390
2241 2242 2243 2244 2245 2246 2247 2248
/*
 * 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;

2249 2250 2251
	if (kvm_vcpu_wake_up(vcpu))
		return;

2252 2253 2254 2255 2256 2257
	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();
}
2258
EXPORT_SYMBOL_GPL(kvm_vcpu_kick);
2259
#endif /* !CONFIG_S390 */
2260

2261
int kvm_vcpu_yield_to(struct kvm_vcpu *target)
2262 2263 2264
{
	struct pid *pid;
	struct task_struct *task = NULL;
2265
	int ret = 0;
2266 2267 2268 2269

	rcu_read_lock();
	pid = rcu_dereference(target->pid);
	if (pid)
2270
		task = get_pid_task(pid, PIDTYPE_PID);
2271 2272
	rcu_read_unlock();
	if (!task)
2273 2274
		return ret;
	ret = yield_to(task, 1);
2275
	put_task_struct(task);
2276 2277

	return ret;
2278 2279 2280
}
EXPORT_SYMBOL_GPL(kvm_vcpu_yield_to);

2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302
/*
 * 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.
 */
2303
static bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
2304
{
2305
#ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
2306 2307 2308
	bool eligible;

	eligible = !vcpu->spin_loop.in_spin_loop ||
2309
		    vcpu->spin_loop.dy_eligible;
2310 2311 2312 2313 2314

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

	return eligible;
2315 2316
#else
	return true;
2317
#endif
2318
}
2319

2320
void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
Z
Zhai, Edwin 已提交
2321
{
2322 2323 2324 2325
	struct kvm *kvm = me->kvm;
	struct kvm_vcpu *vcpu;
	int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
	int yielded = 0;
2326
	int try = 3;
2327 2328
	int pass;
	int i;
Z
Zhai, Edwin 已提交
2329

2330
	kvm_vcpu_set_in_spin_loop(me, true);
2331 2332 2333 2334 2335 2336 2337
	/*
	 * 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.
	 */
2338
	for (pass = 0; pass < 2 && !yielded && try; pass++) {
2339
		kvm_for_each_vcpu(i, vcpu, kvm) {
2340
			if (!pass && i <= last_boosted_vcpu) {
2341 2342 2343 2344
				i = last_boosted_vcpu;
				continue;
			} else if (pass && i > last_boosted_vcpu)
				break;
2345 2346
			if (!ACCESS_ONCE(vcpu->preempted))
				continue;
2347 2348
			if (vcpu == me)
				continue;
2349
			if (swait_active(&vcpu->wq) && !kvm_arch_vcpu_runnable(vcpu))
2350
				continue;
2351 2352
			if (yield_to_kernel_mode && !kvm_arch_vcpu_in_kernel(vcpu))
				continue;
2353 2354
			if (!kvm_vcpu_eligible_for_directed_yield(vcpu))
				continue;
2355 2356 2357

			yielded = kvm_vcpu_yield_to(vcpu);
			if (yielded > 0) {
2358 2359
				kvm->last_boosted_vcpu = i;
				break;
2360 2361 2362 2363
			} else if (yielded < 0) {
				try--;
				if (!try)
					break;
2364 2365 2366
			}
		}
	}
2367
	kvm_vcpu_set_in_spin_loop(me, false);
2368 2369 2370

	/* Ensure vcpu is not eligible during next spinloop */
	kvm_vcpu_set_dy_eligible(me, false);
Z
Zhai, Edwin 已提交
2371 2372 2373
}
EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin);

2374
static int kvm_vcpu_fault(struct vm_fault *vmf)
2375
{
2376
	struct kvm_vcpu *vcpu = vmf->vma->vm_file->private_data;
2377 2378
	struct page *page;

2379
	if (vmf->pgoff == 0)
2380
		page = virt_to_page(vcpu->run);
A
Avi Kivity 已提交
2381
#ifdef CONFIG_X86
2382
	else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
2383
		page = virt_to_page(vcpu->arch.pio_data);
2384
#endif
2385
#ifdef CONFIG_KVM_MMIO
2386 2387
	else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
		page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
A
Avi Kivity 已提交
2388
#endif
2389
	else
2390
		return kvm_arch_vcpu_fault(vcpu, vmf);
2391
	get_page(page);
2392 2393
	vmf->page = page;
	return 0;
2394 2395
}

2396
static const struct vm_operations_struct kvm_vcpu_vm_ops = {
2397
	.fault = kvm_vcpu_fault,
2398 2399 2400 2401 2402 2403 2404 2405
};

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
2406 2407 2408 2409
static int kvm_vcpu_release(struct inode *inode, struct file *filp)
{
	struct kvm_vcpu *vcpu = filp->private_data;

2410
	debugfs_remove_recursive(vcpu->debugfs_dentry);
A
Al Viro 已提交
2411
	kvm_put_kvm(vcpu->kvm);
A
Avi Kivity 已提交
2412 2413 2414
	return 0;
}

2415
static struct file_operations kvm_vcpu_fops = {
A
Avi Kivity 已提交
2416 2417
	.release        = kvm_vcpu_release,
	.unlocked_ioctl = kvm_vcpu_ioctl,
2418
#ifdef CONFIG_KVM_COMPAT
2419 2420
	.compat_ioctl   = kvm_vcpu_compat_ioctl,
#endif
2421
	.mmap           = kvm_vcpu_mmap,
2422
	.llseek		= noop_llseek,
A
Avi Kivity 已提交
2423 2424 2425 2426 2427 2428 2429
};

/*
 * Allocates an inode for the vcpu.
 */
static int create_vcpu_fd(struct kvm_vcpu *vcpu)
{
2430
	return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, O_RDWR | O_CLOEXEC);
A
Avi Kivity 已提交
2431 2432
}

2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458
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;
}

2459 2460 2461
/*
 * Creates some virtual cpus.  Good luck creating more than one.
 */
2462
static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
2463 2464
{
	int r;
2465
	struct kvm_vcpu *vcpu;
2466

G
Greg Kurz 已提交
2467
	if (id >= KVM_MAX_VCPU_ID)
2468 2469
		return -EINVAL;

2470 2471 2472 2473 2474 2475 2476 2477 2478
	mutex_lock(&kvm->lock);
	if (kvm->created_vcpus == KVM_MAX_VCPUS) {
		mutex_unlock(&kvm->lock);
		return -EINVAL;
	}

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

2479
	vcpu = kvm_arch_vcpu_create(kvm, id);
2480 2481 2482 2483
	if (IS_ERR(vcpu)) {
		r = PTR_ERR(vcpu);
		goto vcpu_decrement;
	}
2484

2485 2486
	preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);

2487 2488
	r = kvm_arch_vcpu_setup(vcpu);
	if (r)
2489
		goto vcpu_destroy;
2490

2491 2492 2493 2494
	r = kvm_create_vcpu_debugfs(vcpu);
	if (r)
		goto vcpu_destroy;

S
Shaohua Li 已提交
2495
	mutex_lock(&kvm->lock);
2496 2497 2498 2499
	if (kvm_get_vcpu_by_id(kvm, id)) {
		r = -EEXIST;
		goto unlock_vcpu_destroy;
	}
2500 2501

	BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
2502

R
Rusty Russell 已提交
2503
	/* Now it's all set up, let userspace reach it */
A
Al Viro 已提交
2504
	kvm_get_kvm(kvm);
A
Avi Kivity 已提交
2505
	r = create_vcpu_fd(vcpu);
2506 2507
	if (r < 0) {
		kvm_put_kvm(kvm);
2508
		goto unlock_vcpu_destroy;
2509 2510 2511
	}

	kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu;
2512 2513 2514 2515 2516

	/*
	 * Pairs with smp_rmb() in kvm_get_vcpu.  Write kvm->vcpus
	 * before kvm->online_vcpu's incremented value.
	 */
2517 2518 2519 2520
	smp_wmb();
	atomic_inc(&kvm->online_vcpus);

	mutex_unlock(&kvm->lock);
2521
	kvm_arch_vcpu_postcreate(vcpu);
R
Rusty Russell 已提交
2522
	return r;
2523

2524
unlock_vcpu_destroy:
2525
	mutex_unlock(&kvm->lock);
2526
	debugfs_remove_recursive(vcpu->debugfs_dentry);
2527
vcpu_destroy:
2528
	kvm_arch_vcpu_destroy(vcpu);
2529 2530 2531 2532
vcpu_decrement:
	mutex_lock(&kvm->lock);
	kvm->created_vcpus--;
	mutex_unlock(&kvm->lock);
2533 2534 2535
	return r;
}

A
Avi Kivity 已提交
2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546
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
2547 2548
static long kvm_vcpu_ioctl(struct file *filp,
			   unsigned int ioctl, unsigned long arg)
A
Avi Kivity 已提交
2549
{
A
Avi Kivity 已提交
2550
	struct kvm_vcpu *vcpu = filp->private_data;
A
Al Viro 已提交
2551
	void __user *argp = (void __user *)arg;
2552
	int r;
2553 2554
	struct kvm_fpu *fpu = NULL;
	struct kvm_sregs *kvm_sregs = NULL;
A
Avi Kivity 已提交
2555

2556 2557
	if (vcpu->kvm->mm != current->mm)
		return -EIO;
2558

2559 2560 2561
	if (unlikely(_IOC_TYPE(ioctl) != KVMIO))
		return -EINVAL;

2562
#if defined(CONFIG_S390) || defined(CONFIG_PPC) || defined(CONFIG_MIPS)
2563 2564 2565 2566
	/*
	 * Special cases: vcpu ioctls that are asynchronous to vcpu execution,
	 * so vcpu_load() would break it.
	 */
2567
	if (ioctl == KVM_S390_INTERRUPT || ioctl == KVM_S390_IRQ || ioctl == KVM_INTERRUPT)
2568 2569 2570 2571
		return kvm_arch_vcpu_ioctl(filp, ioctl, arg);
#endif


2572 2573 2574
	r = vcpu_load(vcpu);
	if (r)
		return r;
A
Avi Kivity 已提交
2575
	switch (ioctl) {
2576 2577
	case KVM_RUN: {
		struct pid *oldpid;
2578 2579 2580
		r = -EINVAL;
		if (arg)
			goto out;
2581 2582
		oldpid = rcu_access_pointer(vcpu->pid);
		if (unlikely(oldpid != current->pids[PIDTYPE_PID].pid)) {
2583 2584
			/* The thread running this VCPU changed. */
			struct pid *newpid = get_task_pid(current, PIDTYPE_PID);
2585

2586 2587 2588 2589 2590
			rcu_assign_pointer(vcpu->pid, newpid);
			if (oldpid)
				synchronize_rcu();
			put_pid(oldpid);
		}
2591
		r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
2592
		trace_kvm_userspace_exit(vcpu->run->exit_reason, r);
A
Avi Kivity 已提交
2593
		break;
2594
	}
A
Avi Kivity 已提交
2595
	case KVM_GET_REGS: {
2596
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
2597

2598 2599 2600
		r = -ENOMEM;
		kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
		if (!kvm_regs)
A
Avi Kivity 已提交
2601
			goto out;
2602 2603 2604
		r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
		if (r)
			goto out_free1;
A
Avi Kivity 已提交
2605
		r = -EFAULT;
2606 2607
		if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
			goto out_free1;
A
Avi Kivity 已提交
2608
		r = 0;
2609 2610
out_free1:
		kfree(kvm_regs);
A
Avi Kivity 已提交
2611 2612 2613
		break;
	}
	case KVM_SET_REGS: {
2614
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
2615

2616
		r = -ENOMEM;
2617 2618 2619
		kvm_regs = memdup_user(argp, sizeof(*kvm_regs));
		if (IS_ERR(kvm_regs)) {
			r = PTR_ERR(kvm_regs);
A
Avi Kivity 已提交
2620
			goto out;
2621
		}
2622 2623
		r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
		kfree(kvm_regs);
A
Avi Kivity 已提交
2624 2625 2626
		break;
	}
	case KVM_GET_SREGS: {
2627 2628 2629 2630 2631
		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
2632 2633 2634
		if (r)
			goto out;
		r = -EFAULT;
2635
		if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
A
Avi Kivity 已提交
2636 2637 2638 2639 2640
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_SREGS: {
2641 2642 2643
		kvm_sregs = memdup_user(argp, sizeof(*kvm_sregs));
		if (IS_ERR(kvm_sregs)) {
			r = PTR_ERR(kvm_sregs);
G
Guo Chao 已提交
2644
			kvm_sregs = NULL;
A
Avi Kivity 已提交
2645
			goto out;
2646
		}
2647
		r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
2648 2649
		break;
	}
2650 2651 2652 2653 2654 2655 2656
	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;
2657
		if (copy_to_user(argp, &mp_state, sizeof(mp_state)))
2658 2659 2660 2661 2662 2663 2664 2665
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_MP_STATE: {
		struct kvm_mp_state mp_state;

		r = -EFAULT;
2666
		if (copy_from_user(&mp_state, argp, sizeof(mp_state)))
2667 2668 2669 2670
			goto out;
		r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
		break;
	}
A
Avi Kivity 已提交
2671 2672 2673 2674
	case KVM_TRANSLATE: {
		struct kvm_translation tr;

		r = -EFAULT;
2675
		if (copy_from_user(&tr, argp, sizeof(tr)))
A
Avi Kivity 已提交
2676
			goto out;
2677
		r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
A
Avi Kivity 已提交
2678 2679 2680
		if (r)
			goto out;
		r = -EFAULT;
2681
		if (copy_to_user(argp, &tr, sizeof(tr)))
A
Avi Kivity 已提交
2682 2683 2684 2685
			goto out;
		r = 0;
		break;
	}
J
Jan Kiszka 已提交
2686 2687
	case KVM_SET_GUEST_DEBUG: {
		struct kvm_guest_debug dbg;
A
Avi Kivity 已提交
2688 2689

		r = -EFAULT;
2690
		if (copy_from_user(&dbg, argp, sizeof(dbg)))
A
Avi Kivity 已提交
2691
			goto out;
J
Jan Kiszka 已提交
2692
		r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
A
Avi Kivity 已提交
2693 2694
		break;
	}
A
Avi Kivity 已提交
2695 2696 2697 2698 2699 2700 2701 2702 2703
	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,
2704
					   sizeof(kvm_sigmask)))
A
Avi Kivity 已提交
2705 2706
				goto out;
			r = -EINVAL;
2707
			if (kvm_sigmask.len != sizeof(sigset))
A
Avi Kivity 已提交
2708 2709 2710
				goto out;
			r = -EFAULT;
			if (copy_from_user(&sigset, sigmask_arg->sigset,
2711
					   sizeof(sigset)))
A
Avi Kivity 已提交
2712 2713 2714
				goto out;
			p = &sigset;
		}
2715
		r = kvm_vcpu_ioctl_set_sigmask(vcpu, p);
A
Avi Kivity 已提交
2716 2717
		break;
	}
A
Avi Kivity 已提交
2718
	case KVM_GET_FPU: {
2719 2720 2721 2722 2723
		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
2724 2725 2726
		if (r)
			goto out;
		r = -EFAULT;
2727
		if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
A
Avi Kivity 已提交
2728 2729 2730 2731 2732
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_FPU: {
2733 2734 2735
		fpu = memdup_user(argp, sizeof(*fpu));
		if (IS_ERR(fpu)) {
			r = PTR_ERR(fpu);
G
Guo Chao 已提交
2736
			fpu = NULL;
A
Avi Kivity 已提交
2737
			goto out;
2738
		}
2739
		r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
A
Avi Kivity 已提交
2740 2741
		break;
	}
A
Avi Kivity 已提交
2742
	default:
2743
		r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
2744 2745
	}
out:
2746
	vcpu_put(vcpu);
2747 2748
	kfree(fpu);
	kfree(kvm_sregs);
A
Avi Kivity 已提交
2749 2750 2751
	return r;
}

2752
#ifdef CONFIG_KVM_COMPAT
2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772
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;
		compat_sigset_t csigset;
		sigset_t sigset;

		if (argp) {
			r = -EFAULT;
			if (copy_from_user(&kvm_sigmask, argp,
2773
					   sizeof(kvm_sigmask)))
2774 2775
				goto out;
			r = -EINVAL;
2776
			if (kvm_sigmask.len != sizeof(csigset))
2777 2778 2779
				goto out;
			r = -EFAULT;
			if (copy_from_user(&csigset, sigmask_arg->sigset,
2780
					   sizeof(csigset)))
2781
				goto out;
2782 2783 2784 2785
			sigset_from_compat(&sigset, &csigset);
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
		} else
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, NULL);
2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796
		break;
	}
	default:
		r = kvm_vcpu_ioctl(filp, ioctl, arg);
	}

out:
	return r;
}
#endif

S
Scott Wood 已提交
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 2838 2839 2840 2841 2842 2843
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,
2844
#ifdef CONFIG_KVM_COMPAT
2845 2846
	.compat_ioctl = kvm_device_ioctl,
#endif
S
Scott Wood 已提交
2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857
	.release = kvm_device_release,
};

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

	return filp->private_data;
}

2858
static struct kvm_device_ops *kvm_device_ops_table[KVM_DEV_TYPE_MAX] = {
2859
#ifdef CONFIG_KVM_MPIC
2860 2861
	[KVM_DEV_TYPE_FSL_MPIC_20]	= &kvm_mpic_ops,
	[KVM_DEV_TYPE_FSL_MPIC_42]	= &kvm_mpic_ops,
2862
#endif
2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876
};

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;
}

2877 2878 2879 2880 2881 2882
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
2883 2884 2885 2886 2887 2888 2889 2890
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;

2891 2892 2893 2894 2895
	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
2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907
		return -ENODEV;

	if (test)
		return 0;

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

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

2908
	mutex_lock(&kvm->lock);
S
Scott Wood 已提交
2909 2910
	ret = ops->create(dev, cd->type);
	if (ret < 0) {
2911
		mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
2912 2913 2914
		kfree(dev);
		return ret;
	}
2915 2916
	list_add(&dev->vm_node, &kvm->devices);
	mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
2917

2918 2919 2920
	if (ops->init)
		ops->init(dev);

2921
	ret = anon_inode_getfd(ops->name, &kvm_device_fops, dev, O_RDWR | O_CLOEXEC);
S
Scott Wood 已提交
2922
	if (ret < 0) {
2923 2924 2925
		mutex_lock(&kvm->lock);
		list_del(&dev->vm_node);
		mutex_unlock(&kvm->lock);
2926
		ops->destroy(dev);
S
Scott Wood 已提交
2927 2928 2929 2930 2931 2932 2933 2934
		return ret;
	}

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

2935 2936 2937 2938 2939 2940 2941 2942 2943 2944
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
2945
#ifdef CONFIG_HAVE_KVM_IRQFD
2946
	case KVM_CAP_IRQFD:
2947 2948
	case KVM_CAP_IRQFD_RESAMPLE:
#endif
2949
	case KVM_CAP_IOEVENTFD_ANY_LENGTH:
2950 2951
	case KVM_CAP_CHECK_EXTENSION_VM:
		return 1;
2952
#ifdef CONFIG_KVM_MMIO
2953 2954 2955
	case KVM_CAP_COALESCED_MMIO:
		return KVM_COALESCED_MMIO_PAGE_OFFSET;
#endif
2956 2957 2958
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_CAP_IRQ_ROUTING:
		return KVM_MAX_IRQ_ROUTES;
2959 2960 2961 2962
#endif
#if KVM_ADDRESS_SPACE_NUM > 1
	case KVM_CAP_MULTI_ADDRESS_SPACE:
		return KVM_ADDRESS_SPACE_NUM;
2963
#endif
G
Greg Kurz 已提交
2964 2965
	case KVM_CAP_MAX_VCPU_ID:
		return KVM_MAX_VCPU_ID;
2966 2967 2968 2969 2970 2971
	default:
		break;
	}
	return kvm_vm_ioctl_check_extension(kvm, arg);
}

A
Avi Kivity 已提交
2972 2973 2974 2975 2976
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;
2977
	int r;
A
Avi Kivity 已提交
2978

2979 2980
	if (kvm->mm != current->mm)
		return -EIO;
A
Avi Kivity 已提交
2981 2982 2983 2984
	switch (ioctl) {
	case KVM_CREATE_VCPU:
		r = kvm_vm_ioctl_create_vcpu(kvm, arg);
		break;
2985 2986 2987 2988 2989
	case KVM_SET_USER_MEMORY_REGION: {
		struct kvm_userspace_memory_region kvm_userspace_mem;

		r = -EFAULT;
		if (copy_from_user(&kvm_userspace_mem, argp,
2990
						sizeof(kvm_userspace_mem)))
2991 2992
			goto out;

2993
		r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem);
A
Avi Kivity 已提交
2994 2995 2996 2997 2998 2999
		break;
	}
	case KVM_GET_DIRTY_LOG: {
		struct kvm_dirty_log log;

		r = -EFAULT;
3000
		if (copy_from_user(&log, argp, sizeof(log)))
A
Avi Kivity 已提交
3001
			goto out;
3002
		r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
A
Avi Kivity 已提交
3003 3004
		break;
	}
3005
#ifdef CONFIG_KVM_MMIO
3006 3007
	case KVM_REGISTER_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 3016
			goto out;
		r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
		break;
	}
	case KVM_UNREGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3017

3018
		r = -EFAULT;
3019
		if (copy_from_user(&zone, argp, sizeof(zone)))
3020 3021 3022 3023 3024
			goto out;
		r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
		break;
	}
#endif
G
Gregory Haskins 已提交
3025 3026 3027 3028
	case KVM_IRQFD: {
		struct kvm_irqfd data;

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

		r = -EFAULT;
3038
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3039 3040 3041 3042
			goto out;
		r = kvm_ioeventfd(kvm, &data);
		break;
	}
3043 3044 3045 3046 3047
#ifdef CONFIG_HAVE_KVM_MSI
	case KVM_SIGNAL_MSI: {
		struct kvm_msi msi;

		r = -EFAULT;
3048
		if (copy_from_user(&msi, argp, sizeof(msi)))
3049 3050 3051 3052
			goto out;
		r = kvm_send_userspace_msi(kvm, &msi);
		break;
	}
3053 3054 3055 3056 3057 3058 3059
#endif
#ifdef __KVM_HAVE_IRQ_LINE
	case KVM_IRQ_LINE_STATUS:
	case KVM_IRQ_LINE: {
		struct kvm_irq_level irq_event;

		r = -EFAULT;
3060
		if (copy_from_user(&irq_event, argp, sizeof(irq_event)))
3061 3062
			goto out;

3063 3064
		r = kvm_vm_ioctl_irq_line(kvm, &irq_event,
					ioctl == KVM_IRQ_LINE_STATUS);
3065 3066 3067 3068 3069
		if (r)
			goto out;

		r = -EFAULT;
		if (ioctl == KVM_IRQ_LINE_STATUS) {
3070
			if (copy_to_user(argp, &irq_event, sizeof(irq_event)))
3071 3072 3073 3074 3075 3076
				goto out;
		}

		r = 0;
		break;
	}
3077
#endif
3078 3079 3080 3081
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_SET_GSI_ROUTING: {
		struct kvm_irq_routing routing;
		struct kvm_irq_routing __user *urouting;
3082
		struct kvm_irq_routing_entry *entries = NULL;
3083 3084 3085 3086 3087

		r = -EFAULT;
		if (copy_from_user(&routing, argp, sizeof(routing)))
			goto out;
		r = -EINVAL;
3088 3089
		if (!kvm_arch_can_set_irq_routing(kvm))
			goto out;
3090
		if (routing.nr > KVM_MAX_IRQ_ROUTES)
3091 3092 3093
			goto out;
		if (routing.flags)
			goto out;
3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104
		if (routing.nr) {
			r = -ENOMEM;
			entries = vmalloc(routing.nr * sizeof(*entries));
			if (!entries)
				goto out;
			r = -EFAULT;
			urouting = argp;
			if (copy_from_user(entries, urouting->entries,
					   routing.nr * sizeof(*entries)))
				goto out_free_irq_routing;
		}
3105 3106
		r = kvm_set_irq_routing(kvm, entries, routing.nr,
					routing.flags);
3107
out_free_irq_routing:
3108 3109 3110 3111
		vfree(entries);
		break;
	}
#endif /* CONFIG_HAVE_KVM_IRQ_ROUTING */
S
Scott Wood 已提交
3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129
	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;
	}
3130 3131 3132
	case KVM_CHECK_EXTENSION:
		r = kvm_vm_ioctl_check_extension_generic(kvm, arg);
		break;
3133
	default:
3134
		r = kvm_arch_vm_ioctl(filp, ioctl, arg);
3135 3136 3137 3138 3139
	}
out:
	return r;
}

3140
#ifdef CONFIG_KVM_COMPAT
3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164
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)))
3165
			return -EFAULT;
3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180
		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

3181
static struct file_operations kvm_vm_fops = {
3182 3183
	.release        = kvm_vm_release,
	.unlocked_ioctl = kvm_vm_ioctl,
3184
#ifdef CONFIG_KVM_COMPAT
3185 3186
	.compat_ioctl   = kvm_vm_compat_ioctl,
#endif
3187
	.llseek		= noop_llseek,
3188 3189
};

3190
static int kvm_dev_ioctl_create_vm(unsigned long type)
3191
{
3192
	int r;
3193
	struct kvm *kvm;
3194
	struct file *file;
3195

3196
	kvm = kvm_create_vm(type);
3197 3198
	if (IS_ERR(kvm))
		return PTR_ERR(kvm);
3199
#ifdef CONFIG_KVM_MMIO
3200 3201 3202 3203 3204 3205
	r = kvm_coalesced_mmio_init(kvm);
	if (r < 0) {
		kvm_put_kvm(kvm);
		return r;
	}
#endif
3206
	r = get_unused_fd_flags(O_CLOEXEC);
3207
	if (r < 0) {
A
Al Viro 已提交
3208
		kvm_put_kvm(kvm);
3209 3210
		return r;
	}
3211 3212 3213 3214 3215 3216
	file = anon_inode_getfile("kvm-vm", &kvm_vm_fops, kvm, O_RDWR);
	if (IS_ERR(file)) {
		put_unused_fd(r);
		kvm_put_kvm(kvm);
		return PTR_ERR(file);
	}
3217

3218 3219 3220 3221 3222 3223
	/*
	 * 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).
	 */
3224
	if (kvm_create_vm_debugfs(kvm, r) < 0) {
3225 3226
		put_unused_fd(r);
		fput(file);
3227 3228
		return -ENOMEM;
	}
3229
	kvm_uevent_notify_change(KVM_EVENT_CREATE_VM, kvm);
3230

3231
	fd_install(r, file);
3232
	return r;
3233 3234 3235 3236 3237
}

static long kvm_dev_ioctl(struct file *filp,
			  unsigned int ioctl, unsigned long arg)
{
3238
	long r = -EINVAL;
3239 3240 3241

	switch (ioctl) {
	case KVM_GET_API_VERSION:
3242 3243
		if (arg)
			goto out;
3244 3245 3246
		r = KVM_API_VERSION;
		break;
	case KVM_CREATE_VM:
3247
		r = kvm_dev_ioctl_create_vm(arg);
3248
		break;
3249
	case KVM_CHECK_EXTENSION:
3250
		r = kvm_vm_ioctl_check_extension_generic(NULL, arg);
3251
		break;
3252 3253 3254
	case KVM_GET_VCPU_MMAP_SIZE:
		if (arg)
			goto out;
3255 3256 3257
		r = PAGE_SIZE;     /* struct kvm_run */
#ifdef CONFIG_X86
		r += PAGE_SIZE;    /* pio data page */
3258
#endif
3259
#ifdef CONFIG_KVM_MMIO
3260
		r += PAGE_SIZE;    /* coalesced mmio ring page */
3261
#endif
3262
		break;
3263 3264 3265
	case KVM_TRACE_ENABLE:
	case KVM_TRACE_PAUSE:
	case KVM_TRACE_DISABLE:
3266
		r = -EOPNOTSUPP;
3267
		break;
A
Avi Kivity 已提交
3268
	default:
3269
		return kvm_arch_dev_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
3270 3271 3272 3273 3274 3275 3276 3277
	}
out:
	return r;
}

static struct file_operations kvm_chardev_ops = {
	.unlocked_ioctl = kvm_dev_ioctl,
	.compat_ioctl   = kvm_dev_ioctl,
3278
	.llseek		= noop_llseek,
A
Avi Kivity 已提交
3279 3280 3281
};

static struct miscdevice kvm_dev = {
A
Avi Kivity 已提交
3282
	KVM_MINOR,
A
Avi Kivity 已提交
3283 3284 3285 3286
	"kvm",
	&kvm_chardev_ops,
};

3287
static void hardware_enable_nolock(void *junk)
3288 3289
{
	int cpu = raw_smp_processor_id();
3290
	int r;
3291

3292
	if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
3293
		return;
3294

3295
	cpumask_set_cpu(cpu, cpus_hardware_enabled);
3296

3297
	r = kvm_arch_hardware_enable();
3298 3299 3300 3301

	if (r) {
		cpumask_clear_cpu(cpu, cpus_hardware_enabled);
		atomic_inc(&hardware_enable_failed);
X
Xiubo Li 已提交
3302
		pr_info("kvm: enabling virtualization on CPU%d failed\n", cpu);
3303
	}
3304 3305
}

3306
static int kvm_starting_cpu(unsigned int cpu)
3307
{
3308
	raw_spin_lock(&kvm_count_lock);
3309 3310
	if (kvm_usage_count)
		hardware_enable_nolock(NULL);
3311
	raw_spin_unlock(&kvm_count_lock);
3312
	return 0;
3313 3314 3315
}

static void hardware_disable_nolock(void *junk)
3316 3317 3318
{
	int cpu = raw_smp_processor_id();

3319
	if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
3320
		return;
3321
	cpumask_clear_cpu(cpu, cpus_hardware_enabled);
3322
	kvm_arch_hardware_disable();
3323 3324
}

3325
static int kvm_dying_cpu(unsigned int cpu)
3326
{
3327
	raw_spin_lock(&kvm_count_lock);
3328 3329
	if (kvm_usage_count)
		hardware_disable_nolock(NULL);
3330
	raw_spin_unlock(&kvm_count_lock);
3331
	return 0;
3332 3333
}

3334 3335 3336 3337 3338 3339
static void hardware_disable_all_nolock(void)
{
	BUG_ON(!kvm_usage_count);

	kvm_usage_count--;
	if (!kvm_usage_count)
3340
		on_each_cpu(hardware_disable_nolock, NULL, 1);
3341 3342 3343 3344
}

static void hardware_disable_all(void)
{
3345
	raw_spin_lock(&kvm_count_lock);
3346
	hardware_disable_all_nolock();
3347
	raw_spin_unlock(&kvm_count_lock);
3348 3349 3350 3351 3352 3353
}

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

3354
	raw_spin_lock(&kvm_count_lock);
3355 3356 3357 3358

	kvm_usage_count++;
	if (kvm_usage_count == 1) {
		atomic_set(&hardware_enable_failed, 0);
3359
		on_each_cpu(hardware_enable_nolock, NULL, 1);
3360 3361 3362 3363 3364 3365 3366

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

3367
	raw_spin_unlock(&kvm_count_lock);
3368 3369 3370 3371

	return r;
}

3372
static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
M
Mike Day 已提交
3373
		      void *v)
3374
{
3375 3376 3377 3378 3379 3380
	/*
	 * 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
3381
	pr_info("kvm: exiting hardware virtualization\n");
3382
	kvm_rebooting = true;
3383
	on_each_cpu(hardware_disable_nolock, NULL, 1);
3384 3385 3386 3387 3388 3389 3390 3391
	return NOTIFY_OK;
}

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

M
Marcelo Tosatti 已提交
3392
static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
3393 3394 3395 3396
{
	int i;

	for (i = 0; i < bus->dev_count; i++) {
3397
		struct kvm_io_device *pos = bus->range[i].dev;
3398 3399 3400

		kvm_iodevice_destructor(pos);
	}
M
Marcelo Tosatti 已提交
3401
	kfree(bus);
3402 3403
}

3404
static inline int kvm_io_bus_cmp(const struct kvm_io_range *r1,
X
Xiubo Li 已提交
3405
				 const struct kvm_io_range *r2)
3406
{
J
Jason Wang 已提交
3407 3408 3409 3410
	gpa_t addr1 = r1->addr;
	gpa_t addr2 = r2->addr;

	if (addr1 < addr2)
3411
		return -1;
J
Jason Wang 已提交
3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423

	/* 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)
3424
		return 1;
J
Jason Wang 已提交
3425

3426 3427 3428
	return 0;
}

3429 3430
static int kvm_io_bus_sort_cmp(const void *p1, const void *p2)
{
3431
	return kvm_io_bus_cmp(p1, p2);
3432 3433
}

G
Geoff Levand 已提交
3434
static int kvm_io_bus_insert_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev,
3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448
			  gpa_t addr, int len)
{
	bus->range[bus->dev_count++] = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
		.dev = dev,
	};

	sort(bus->range, bus->dev_count, sizeof(struct kvm_io_range),
		kvm_io_bus_sort_cmp, NULL);

	return 0;
}

G
Geoff Levand 已提交
3449
static int kvm_io_bus_get_first_dev(struct kvm_io_bus *bus,
3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466
			     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;

3467
	while (off > 0 && kvm_io_bus_cmp(&key, &bus->range[off-1]) == 0)
3468 3469 3470 3471 3472
		off--;

	return off;
}

3473
static int __kvm_io_bus_write(struct kvm_vcpu *vcpu, struct kvm_io_bus *bus,
C
Cornelia Huck 已提交
3474 3475 3476 3477 3478 3479 3480 3481 3482
			      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 &&
3483
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
3484
		if (!kvm_iodevice_write(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
3485 3486 3487 3488 3489 3490 3491 3492
					range->len, val))
			return idx;
		idx++;
	}

	return -EOPNOTSUPP;
}

3493
/* kvm_io_bus_write - called under kvm->slots_lock */
3494
int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
3495
		     int len, const void *val)
3496
{
3497
	struct kvm_io_bus *bus;
3498
	struct kvm_io_range range;
C
Cornelia Huck 已提交
3499
	int r;
3500 3501 3502 3503 3504

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
3505

3506
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3507 3508
	if (!bus)
		return -ENOMEM;
3509
	r = __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3510 3511 3512 3513
	return r < 0 ? r : 0;
}

/* kvm_io_bus_write_cookie - called under kvm->slots_lock */
3514 3515
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
3516 3517 3518 3519 3520 3521 3522 3523 3524
{
	struct kvm_io_bus *bus;
	struct kvm_io_range range;

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

3525
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3526 3527
	if (!bus)
		return -ENOMEM;
C
Cornelia Huck 已提交
3528 3529 3530

	/* First try the device referenced by cookie. */
	if ((cookie >= 0) && (cookie < bus->dev_count) &&
3531
	    (kvm_io_bus_cmp(&range, &bus->range[cookie]) == 0))
3532
		if (!kvm_iodevice_write(vcpu, bus->range[cookie].dev, addr, len,
C
Cornelia Huck 已提交
3533 3534 3535 3536 3537 3538 3539
					val))
			return cookie;

	/*
	 * cookie contained garbage; fall back to search and return the
	 * correct cookie value.
	 */
3540
	return __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3541 3542
}

3543 3544
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
3545 3546 3547 3548
{
	int idx;

	idx = kvm_io_bus_get_first_dev(bus, range->addr, range->len);
3549 3550 3551 3552
	if (idx < 0)
		return -EOPNOTSUPP;

	while (idx < bus->dev_count &&
3553
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
3554
		if (!kvm_iodevice_read(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
3555 3556
				       range->len, val))
			return idx;
3557 3558 3559
		idx++;
	}

3560 3561
	return -EOPNOTSUPP;
}
3562
EXPORT_SYMBOL_GPL(kvm_io_bus_write);
3563

3564
/* kvm_io_bus_read - called under kvm->slots_lock */
3565
int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
M
Marcelo Tosatti 已提交
3566
		    int len, void *val)
3567
{
3568
	struct kvm_io_bus *bus;
3569
	struct kvm_io_range range;
C
Cornelia Huck 已提交
3570
	int r;
3571 3572 3573 3574 3575

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
M
Marcelo Tosatti 已提交
3576

3577
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3578 3579
	if (!bus)
		return -ENOMEM;
3580
	r = __kvm_io_bus_read(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3581 3582
	return r < 0 ? r : 0;
}
3583

3584

3585
/* Caller must hold slots_lock. */
3586 3587
int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
			    int len, struct kvm_io_device *dev)
3588
{
M
Marcelo Tosatti 已提交
3589
	struct kvm_io_bus *new_bus, *bus;
3590

3591
	bus = kvm_get_bus(kvm, bus_idx);
3592 3593 3594
	if (!bus)
		return -ENOMEM;

3595 3596
	/* exclude ioeventfd which is limited by maximum fd */
	if (bus->dev_count - bus->ioeventfd_count > NR_IOBUS_DEVS - 1)
3597
		return -ENOSPC;
3598

3599
	new_bus = kmalloc(sizeof(*bus) + ((bus->dev_count + 1) *
3600
			  sizeof(struct kvm_io_range)), GFP_KERNEL);
M
Marcelo Tosatti 已提交
3601 3602
	if (!new_bus)
		return -ENOMEM;
3603 3604
	memcpy(new_bus, bus, sizeof(*bus) + (bus->dev_count *
	       sizeof(struct kvm_io_range)));
3605
	kvm_io_bus_insert_dev(new_bus, dev, addr, len);
M
Marcelo Tosatti 已提交
3606 3607 3608
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
3609 3610 3611 3612

	return 0;
}

3613
/* Caller must hold slots_lock. */
3614 3615
void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
			       struct kvm_io_device *dev)
3616
{
3617
	int i;
M
Marcelo Tosatti 已提交
3618
	struct kvm_io_bus *new_bus, *bus;
3619

3620
	bus = kvm_get_bus(kvm, bus_idx);
3621
	if (!bus)
3622
		return;
3623

3624 3625
	for (i = 0; i < bus->dev_count; i++)
		if (bus->range[i].dev == dev) {
3626 3627
			break;
		}
M
Marcelo Tosatti 已提交
3628

3629 3630
	if (i == bus->dev_count)
		return;
3631

3632
	new_bus = kmalloc(sizeof(*bus) + ((bus->dev_count - 1) *
3633
			  sizeof(struct kvm_io_range)), GFP_KERNEL);
3634 3635 3636 3637
	if (!new_bus)  {
		pr_err("kvm: failed to shrink bus, removing it completely\n");
		goto broken;
	}
3638 3639 3640 3641 3642

	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
3643

3644
broken:
M
Marcelo Tosatti 已提交
3645 3646 3647
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
3648
	return;
3649 3650
}

3651 3652 3653 3654 3655 3656 3657 3658 3659 3660
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);
3661 3662
	if (!bus)
		goto out_unlock;
3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676

	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);

3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688
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.
	 */
3689
	if (!refcount_inc_not_zero(&stat_data->kvm->users_count))
3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714
		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;

3715
	*val = *(ulong *)((void *)stat_data->kvm + stat_data->offset);
3716 3717 3718 3719

	return 0;
}

3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731
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;
}

3732 3733 3734 3735
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,
3736
				vm_stat_clear_per_vm, "%llu\n");
3737 3738 3739 3740 3741 3742 3743 3744
}

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,
3745
	.llseek  = no_llseek,
3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756
};

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)
3757
		*val += *(u64 *)((void *)vcpu + stat_data->offset);
3758 3759 3760 3761

	return 0;
}

3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776
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;
}

3777 3778 3779 3780
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,
3781
				 vcpu_stat_clear_per_vm, "%llu\n");
3782 3783 3784 3785 3786 3787 3788 3789
}

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,
3790
	.llseek  = no_llseek,
3791 3792 3793 3794 3795 3796 3797
};

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,
};

3798
static int vm_stat_get(void *_offset, u64 *val)
3799 3800 3801
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
3802 3803
	struct kvm_stat_data stat_tmp = {.offset = offset};
	u64 tmp_val;
3804

3805
	*val = 0;
3806
	spin_lock(&kvm_lock);
3807 3808 3809 3810 3811
	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;
	}
3812
	spin_unlock(&kvm_lock);
3813
	return 0;
3814 3815
}

3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835
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");
3836

3837
static int vcpu_stat_get(void *_offset, u64 *val)
A
Avi Kivity 已提交
3838 3839 3840
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
3841 3842
	struct kvm_stat_data stat_tmp = {.offset = offset};
	u64 tmp_val;
A
Avi Kivity 已提交
3843

3844
	*val = 0;
3845
	spin_lock(&kvm_lock);
3846 3847 3848 3849 3850
	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;
	}
3851
	spin_unlock(&kvm_lock);
3852
	return 0;
A
Avi Kivity 已提交
3853 3854
}

3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875
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");
3876

3877
static const struct file_operations *stat_fops[] = {
3878 3879 3880
	[KVM_STAT_VCPU] = &vcpu_stat_fops,
	[KVM_STAT_VM]   = &vm_stat_fops,
};
A
Avi Kivity 已提交
3881

3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907
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);

3908
	if (type == KVM_EVENT_CREATE_VM) {
3909
		add_uevent_var(env, "EVENT=create");
3910 3911
		kvm->userspace_pid = task_pid_nr(current);
	} else if (type == KVM_EVENT_DESTROY_VM) {
3912
		add_uevent_var(env, "EVENT=destroy");
3913 3914
	}
	add_uevent_var(env, "PID=%d", kvm->userspace_pid);
3915 3916

	if (kvm->debugfs_dentry) {
3917 3918 3919 3920 3921 3922 3923
		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);
3924 3925 3926 3927 3928 3929 3930 3931
		}
	}
	/* 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);
}

3932
static int kvm_init_debug(void)
A
Avi Kivity 已提交
3933
{
3934
	int r = -EEXIST;
A
Avi Kivity 已提交
3935 3936
	struct kvm_stats_debugfs_item *p;

3937
	kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
3938 3939 3940
	if (kvm_debugfs_dir == NULL)
		goto out;

3941 3942
	kvm_debugfs_num_entries = 0;
	for (p = debugfs_entries; p->name; ++p, kvm_debugfs_num_entries++) {
3943
		if (!debugfs_create_file(p->name, 0644, kvm_debugfs_dir,
3944 3945
					 (void *)(long)p->offset,
					 stat_fops[p->kind]))
3946 3947 3948 3949 3950 3951 3952 3953 3954
			goto out_dir;
	}

	return 0;

out_dir:
	debugfs_remove_recursive(kvm_debugfs_dir);
out:
	return r;
A
Avi Kivity 已提交
3955 3956
}

3957
static int kvm_suspend(void)
3958
{
3959
	if (kvm_usage_count)
3960
		hardware_disable_nolock(NULL);
3961 3962 3963
	return 0;
}

3964
static void kvm_resume(void)
3965
{
3966
	if (kvm_usage_count) {
3967
		WARN_ON(raw_spin_is_locked(&kvm_count_lock));
3968
		hardware_enable_nolock(NULL);
3969
	}
3970 3971
}

3972
static struct syscore_ops kvm_syscore_ops = {
3973 3974 3975 3976
	.suspend = kvm_suspend,
	.resume = kvm_resume,
};

3977 3978 3979 3980 3981 3982 3983 3984 3985
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);
3986

3987 3988
	if (vcpu->preempted)
		vcpu->preempted = false;
3989

R
Radim Krčmář 已提交
3990 3991
	kvm_arch_sched_in(vcpu, cpu);

3992
	kvm_arch_vcpu_load(vcpu, cpu);
3993 3994 3995 3996 3997 3998 3999
}

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

4000 4001
	if (current->state == TASK_RUNNING)
		vcpu->preempted = true;
4002
	kvm_arch_vcpu_put(vcpu);
4003 4004
}

4005
int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
4006
		  struct module *module)
A
Avi Kivity 已提交
4007 4008
{
	int r;
Y
Yang, Sheng 已提交
4009
	int cpu;
A
Avi Kivity 已提交
4010

4011 4012
	r = kvm_arch_init(opaque);
	if (r)
4013
		goto out_fail;
4014

4015 4016 4017 4018
	/*
	 * 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
4019 4020
	 * kvm_arch_init must be called before kvm_irqfd_init to avoid creating
	 * conflicts in case kvm is already setup for another implementation.
4021
	 */
P
Paolo Bonzini 已提交
4022 4023 4024
	r = kvm_irqfd_init();
	if (r)
		goto out_irqfd;
4025

4026
	if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
4027 4028 4029 4030
		r = -ENOMEM;
		goto out_free_0;
	}

4031
	r = kvm_arch_hardware_setup();
A
Avi Kivity 已提交
4032
	if (r < 0)
4033
		goto out_free_0a;
A
Avi Kivity 已提交
4034

Y
Yang, Sheng 已提交
4035 4036
	for_each_online_cpu(cpu) {
		smp_call_function_single(cpu,
4037
				kvm_arch_check_processor_compat,
4038
				&r, 1);
Y
Yang, Sheng 已提交
4039
		if (r < 0)
4040
			goto out_free_1;
Y
Yang, Sheng 已提交
4041 4042
	}

T
Thomas Gleixner 已提交
4043
	r = cpuhp_setup_state_nocalls(CPUHP_AP_KVM_STARTING, "kvm/cpu:starting",
4044
				      kvm_starting_cpu, kvm_dying_cpu);
A
Avi Kivity 已提交
4045
	if (r)
4046
		goto out_free_2;
A
Avi Kivity 已提交
4047 4048
	register_reboot_notifier(&kvm_reboot_notifier);

4049
	/* A kmem cache lets us meet the alignment requirements of fx_save. */
4050 4051 4052
	if (!vcpu_align)
		vcpu_align = __alignof__(struct kvm_vcpu);
	kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, vcpu_align,
J
Joe Perches 已提交
4053
					   0, NULL);
4054 4055
	if (!kvm_vcpu_cache) {
		r = -ENOMEM;
4056
		goto out_free_3;
4057 4058
	}

4059 4060 4061 4062
	r = kvm_async_pf_init();
	if (r)
		goto out_free;

A
Avi Kivity 已提交
4063
	kvm_chardev_ops.owner = module;
4064 4065
	kvm_vm_fops.owner = module;
	kvm_vcpu_fops.owner = module;
A
Avi Kivity 已提交
4066 4067 4068

	r = misc_register(&kvm_dev);
	if (r) {
X
Xiubo Li 已提交
4069
		pr_err("kvm: misc device register failed\n");
4070
		goto out_unreg;
A
Avi Kivity 已提交
4071 4072
	}

4073 4074
	register_syscore_ops(&kvm_syscore_ops);

4075 4076 4077
	kvm_preempt_ops.sched_in = kvm_sched_in;
	kvm_preempt_ops.sched_out = kvm_sched_out;

4078 4079
	r = kvm_init_debug();
	if (r) {
X
Xiubo Li 已提交
4080
		pr_err("kvm: create debugfs files failed\n");
4081 4082
		goto out_undebugfs;
	}
4083

P
Paolo Bonzini 已提交
4084 4085 4086
	r = kvm_vfio_ops_init();
	WARN_ON(r);

4087
	return 0;
A
Avi Kivity 已提交
4088

4089 4090
out_undebugfs:
	unregister_syscore_ops(&kvm_syscore_ops);
4091
	misc_deregister(&kvm_dev);
4092 4093
out_unreg:
	kvm_async_pf_deinit();
A
Avi Kivity 已提交
4094
out_free:
4095
	kmem_cache_destroy(kvm_vcpu_cache);
4096
out_free_3:
A
Avi Kivity 已提交
4097
	unregister_reboot_notifier(&kvm_reboot_notifier);
4098
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4099 4100
out_free_2:
out_free_1:
4101
	kvm_arch_hardware_unsetup();
4102 4103
out_free_0a:
	free_cpumask_var(cpus_hardware_enabled);
4104
out_free_0:
4105
	kvm_irqfd_exit();
P
Paolo Bonzini 已提交
4106
out_irqfd:
4107 4108
	kvm_arch_exit();
out_fail:
A
Avi Kivity 已提交
4109 4110
	return r;
}
4111
EXPORT_SYMBOL_GPL(kvm_init);
A
Avi Kivity 已提交
4112

4113
void kvm_exit(void)
A
Avi Kivity 已提交
4114
{
4115
	debugfs_remove_recursive(kvm_debugfs_dir);
A
Avi Kivity 已提交
4116
	misc_deregister(&kvm_dev);
4117
	kmem_cache_destroy(kvm_vcpu_cache);
4118
	kvm_async_pf_deinit();
4119
	unregister_syscore_ops(&kvm_syscore_ops);
A
Avi Kivity 已提交
4120
	unregister_reboot_notifier(&kvm_reboot_notifier);
4121
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4122
	on_each_cpu(hardware_disable_nolock, NULL, 1);
4123
	kvm_arch_hardware_unsetup();
4124
	kvm_arch_exit();
4125
	kvm_irqfd_exit();
4126
	free_cpumask_var(cpus_hardware_enabled);
4127
	kvm_vfio_ops_exit();
A
Avi Kivity 已提交
4128
}
4129
EXPORT_SYMBOL_GPL(kvm_exit);