kvm_main.c 94.4 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, S_IRUGO | S_IWUSR);
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, S_IRUGO | S_IWUSR);
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, S_IRUGO | S_IWUSR);
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;

D
Dan Williams 已提交
133
bool kvm_is_reserved_pfn(kvm_pfn_t pfn)
B
Ben-Ami Yassour 已提交
134
{
135
	if (pfn_valid(pfn))
136
		return PageReserved(pfn_to_page(pfn));
B
Ben-Ami Yassour 已提交
137 138 139 140

	return true;
}

A
Avi Kivity 已提交
141 142 143
/*
 * Switches to specified vcpu, until a matching vcpu_put()
 */
144
int vcpu_load(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
145
{
146 147
	int cpu;

148 149
	if (mutex_lock_killable(&vcpu->mutex))
		return -EINTR;
150 151
	cpu = get_cpu();
	preempt_notifier_register(&vcpu->preempt_notifier);
152
	kvm_arch_vcpu_load(vcpu, cpu);
153
	put_cpu();
154
	return 0;
A
Avi Kivity 已提交
155
}
156
EXPORT_SYMBOL_GPL(vcpu_load);
A
Avi Kivity 已提交
157

158
void vcpu_put(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
159
{
160
	preempt_disable();
161
	kvm_arch_vcpu_put(vcpu);
162 163
	preempt_notifier_unregister(&vcpu->preempt_notifier);
	preempt_enable();
A
Avi Kivity 已提交
164 165
	mutex_unlock(&vcpu->mutex);
}
166
EXPORT_SYMBOL_GPL(vcpu_put);
A
Avi Kivity 已提交
167

168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185
/* 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;
}

186 187 188 189
static void ack_flush(void *_completed)
{
}

190
bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req)
191
{
192
	int i, cpu, me;
193 194
	cpumask_var_t cpus;
	bool called = true;
195
	bool wait = req & KVM_REQUEST_WAIT;
196 197
	struct kvm_vcpu *vcpu;

198
	zalloc_cpumask_var(&cpus, GFP_ATOMIC);
199

200
	me = get_cpu();
201
	kvm_for_each_vcpu(i, vcpu, kvm) {
202
		kvm_make_request(req, vcpu);
203
		cpu = vcpu->cpu;
204

205 206
		if (!(req & KVM_REQUEST_NO_WAKEUP) && kvm_vcpu_wake_up(vcpu))
			continue;
207

208
		if (cpus != NULL && cpu != -1 && cpu != me &&
209
		    kvm_request_needs_ipi(vcpu, req))
210
			cpumask_set_cpu(cpu, cpus);
211
	}
212
	if (unlikely(cpus == NULL))
213
		smp_call_function_many(cpu_online_mask, ack_flush, NULL, wait);
214
	else if (!cpumask_empty(cpus))
215
		smp_call_function_many(cpus, ack_flush, NULL, wait);
216 217
	else
		called = false;
218
	put_cpu();
219
	free_cpumask_var(cpus);
220
	return called;
221 222
}

223
#ifndef CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALL
224
void kvm_flush_remote_tlbs(struct kvm *kvm)
225
{
226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242
	/*
	 * 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.
	 */
243
	if (kvm_make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH))
244
		++kvm->stat.remote_tlb_flush;
245
	cmpxchg(&kvm->tlbs_dirty, dirty_count, 0);
246
}
247
EXPORT_SYMBOL_GPL(kvm_flush_remote_tlbs);
248
#endif
249

250 251
void kvm_reload_remote_mmus(struct kvm *kvm)
{
252
	kvm_make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD);
253
}
254

R
Rusty Russell 已提交
255 256 257 258 259 260 261 262 263
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;
264
	vcpu->pid = NULL;
265
	init_swait_queue_head(&vcpu->wq);
266
	kvm_async_pf_vcpu_init(vcpu);
R
Rusty Russell 已提交
267

268 269 270
	vcpu->pre_pcpu = -1;
	INIT_LIST_HEAD(&vcpu->blocked_vcpu_list);

R
Rusty Russell 已提交
271 272 273 274 275 276 277
	page = alloc_page(GFP_KERNEL | __GFP_ZERO);
	if (!page) {
		r = -ENOMEM;
		goto fail;
	}
	vcpu->run = page_address(page);

278 279
	kvm_vcpu_set_in_spin_loop(vcpu, false);
	kvm_vcpu_set_dy_eligible(vcpu, false);
280
	vcpu->preempted = false;
281

282
	r = kvm_arch_vcpu_init(vcpu);
R
Rusty Russell 已提交
283
	if (r < 0)
284
		goto fail_free_run;
R
Rusty Russell 已提交
285 286 287 288 289
	return 0;

fail_free_run:
	free_page((unsigned long)vcpu->run);
fail:
290
	return r;
R
Rusty Russell 已提交
291 292 293 294 295
}
EXPORT_SYMBOL_GPL(kvm_vcpu_init);

void kvm_vcpu_uninit(struct kvm_vcpu *vcpu)
{
296
	put_pid(vcpu->pid);
297
	kvm_arch_vcpu_uninit(vcpu);
R
Rusty Russell 已提交
298 299 300 301
	free_page((unsigned long)vcpu->run);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_uninit);

302 303 304 305 306 307 308 309 310 311 312
#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);
313
	int need_tlb_flush, idx;
314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332

	/*
	 * 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.
	 */
333
	idx = srcu_read_lock(&kvm->srcu);
334
	spin_lock(&kvm->mmu_lock);
335

336
	kvm->mmu_notifier_seq++;
337
	need_tlb_flush = kvm_unmap_hva(kvm, address) | kvm->tlbs_dirty;
338 339 340 341
	/* we've to flush the tlb before the pages can be freed */
	if (need_tlb_flush)
		kvm_flush_remote_tlbs(kvm);

342
	spin_unlock(&kvm->mmu_lock);
343 344 345

	kvm_arch_mmu_notifier_invalidate_page(kvm, address);

346
	srcu_read_unlock(&kvm->srcu, idx);
347 348
}

349 350 351 352 353 354
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);
355
	int idx;
356

357
	idx = srcu_read_lock(&kvm->srcu);
358 359 360 361
	spin_lock(&kvm->mmu_lock);
	kvm->mmu_notifier_seq++;
	kvm_set_spte_hva(kvm, address, pte);
	spin_unlock(&kvm->mmu_lock);
362
	srcu_read_unlock(&kvm->srcu, idx);
363 364
}

365 366 367 368 369 370
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);
371
	int need_tlb_flush = 0, idx;
372

373
	idx = srcu_read_lock(&kvm->srcu);
374 375 376 377 378 379 380
	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++;
381
	need_tlb_flush = kvm_unmap_hva_range(kvm, start, end);
382
	need_tlb_flush |= kvm->tlbs_dirty;
383 384 385
	/* we've to flush the tlb before the pages can be freed */
	if (need_tlb_flush)
		kvm_flush_remote_tlbs(kvm);
386 387 388

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

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++;
405
	smp_wmb();
406 407
	/*
	 * The above sequence increase must be visible before the
408 409
	 * below count decrease, which is ensured by the smp_wmb above
	 * in conjunction with the smp_rmb in mmu_notifier_retry().
410 411 412 413 414 415 416 417 418
	 */
	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
419 420
					      unsigned long start,
					      unsigned long end)
421 422
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
423
	int young, idx;
424

425
	idx = srcu_read_lock(&kvm->srcu);
426 427
	spin_lock(&kvm->mmu_lock);

A
Andres Lagar-Cavilla 已提交
428
	young = kvm_age_hva(kvm, start, end);
429 430 431
	if (young)
		kvm_flush_remote_tlbs(kvm);

432 433 434
	spin_unlock(&kvm->mmu_lock);
	srcu_read_unlock(&kvm->srcu, idx);

435 436 437
	return young;
}

438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467
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
468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483
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;
}

484 485 486 487
static void kvm_mmu_notifier_release(struct mmu_notifier *mn,
				     struct mm_struct *mm)
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
488 489 490
	int idx;

	idx = srcu_read_lock(&kvm->srcu);
491
	kvm_arch_flush_shadow_all(kvm);
492
	srcu_read_unlock(&kvm->srcu, idx);
493 494
}

495 496 497 498 499
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,
500
	.clear_young		= kvm_mmu_notifier_clear_young,
A
Andrea Arcangeli 已提交
501
	.test_young		= kvm_mmu_notifier_test_young,
502
	.change_pte		= kvm_mmu_notifier_change_pte,
503
	.release		= kvm_mmu_notifier_release,
504
};
505 506 507 508 509 510 511 512 513 514 515 516 517 518

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;
}

519 520
#endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */

521
static struct kvm_memslots *kvm_alloc_memslots(void)
522 523
{
	int i;
524
	struct kvm_memslots *slots;
525

526 527 528 529
	slots = kvm_kvzalloc(sizeof(struct kvm_memslots));
	if (!slots)
		return NULL;

530
	for (i = 0; i < KVM_MEM_SLOTS_NUM; i++)
531
		slots->id_to_index[i] = slots->memslots[i].id = i;
532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569

	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);
570 571
}

572 573 574 575 576 577 578 579 580
static void kvm_destroy_vm_debugfs(struct kvm *kvm)
{
	int i;

	if (!kvm->debugfs_dentry)
		return;

	debugfs_remove_recursive(kvm->debugfs_dentry);

581 582 583 584 585
	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);
	}
586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616
}

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;
617
		if (!debugfs_create_file(p->name, 0644,
618 619 620 621 622 623 624 625
					 kvm->debugfs_dentry,
					 stat_data,
					 stat_fops_per_vm[p->kind]))
			return -ENOMEM;
	}
	return 0;
}

626
static struct kvm *kvm_create_vm(unsigned long type)
A
Avi Kivity 已提交
627
{
628 629
	int r, i;
	struct kvm *kvm = kvm_arch_alloc_vm();
A
Avi Kivity 已提交
630

631 632 633
	if (!kvm)
		return ERR_PTR(-ENOMEM);

634
	spin_lock_init(&kvm->mmu_lock);
V
Vegard Nossum 已提交
635
	mmgrab(current->mm);
636 637 638 639 640
	kvm->mm = current->mm;
	kvm_eventfd_init(kvm);
	mutex_init(&kvm->lock);
	mutex_init(&kvm->irq_lock);
	mutex_init(&kvm->slots_lock);
641
	refcount_set(&kvm->users_count, 1);
642 643
	INIT_LIST_HEAD(&kvm->devices);

644
	r = kvm_arch_init_vm(kvm, type);
645
	if (r)
646
		goto out_err_no_disable;
647 648 649

	r = hardware_enable_all();
	if (r)
650
		goto out_err_no_disable;
651

652
#ifdef CONFIG_HAVE_KVM_IRQFD
653
	INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list);
654
#endif
A
Avi Kivity 已提交
655

656 657
	BUILD_BUG_ON(KVM_MEM_SLOTS_NUM > SHRT_MAX);

658
	r = -ENOMEM;
659
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
660 661
		struct kvm_memslots *slots = kvm_alloc_memslots();
		if (!slots)
662
			goto out_err_no_srcu;
663 664 665 666 667 668 669
		/*
		 * 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);
670
	}
671

672
	if (init_srcu_struct(&kvm->srcu))
673 674 675
		goto out_err_no_srcu;
	if (init_srcu_struct(&kvm->irq_srcu))
		goto out_err_no_irq_srcu;
M
Marcelo Tosatti 已提交
676 677 678
	for (i = 0; i < KVM_NR_BUSES; i++) {
		kvm->buses[i] = kzalloc(sizeof(struct kvm_io_bus),
					GFP_KERNEL);
679
		if (!kvm->buses[i])
M
Marcelo Tosatti 已提交
680 681
			goto out_err;
	}
682

683 684 685 686
	r = kvm_init_mmu_notifier(kvm);
	if (r)
		goto out_err;

687
	spin_lock(&kvm_lock);
688
	list_add(&kvm->vm_list, &vm_list);
689
	spin_unlock(&kvm_lock);
690

691 692
	preempt_notifier_inc();

693
	return kvm;
694 695

out_err:
696 697
	cleanup_srcu_struct(&kvm->irq_srcu);
out_err_no_irq_srcu:
698
	cleanup_srcu_struct(&kvm->srcu);
699
out_err_no_srcu:
700
	hardware_disable_all();
701
out_err_no_disable:
M
Marcelo Tosatti 已提交
702 703
	for (i = 0; i < KVM_NR_BUSES; i++)
		kfree(kvm->buses[i]);
704 705
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
		kvm_free_memslots(kvm, kvm->memslots[i]);
706
	kvm_arch_free_vm(kvm);
707
	mmdrop(current->mm);
708
	return ERR_PTR(r);
709 710
}

711 712 713 714
/*
 * Avoid using vmalloc for a small buffer.
 * Should not be used when the size is statically known.
 */
715
void *kvm_kvzalloc(unsigned long size)
716 717 718 719 720 721 722
{
	if (size > PAGE_SIZE)
		return vzalloc(size);
	else
		return kzalloc(size, GFP_KERNEL);
}

723 724
static void kvm_destroy_devices(struct kvm *kvm)
{
G
Geliang Tang 已提交
725
	struct kvm_device *dev, *tmp;
726

727 728 729 730 731
	/*
	 * 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
732 733
	list_for_each_entry_safe(dev, tmp, &kvm->devices, vm_node) {
		list_del(&dev->vm_node);
734 735 736 737
		dev->ops->destroy(dev);
	}
}

738 739
static void kvm_destroy_vm(struct kvm *kvm)
{
M
Marcelo Tosatti 已提交
740
	int i;
741 742
	struct mm_struct *mm = kvm->mm;

743
	kvm_destroy_vm_debugfs(kvm);
744
	kvm_arch_sync_events(kvm);
745
	spin_lock(&kvm_lock);
746
	list_del(&kvm->vm_list);
747
	spin_unlock(&kvm_lock);
748
	kvm_free_irq_routing(kvm);
749
	for (i = 0; i < KVM_NR_BUSES; i++) {
750 751
		if (kvm->buses[i])
			kvm_io_bus_destroy(kvm->buses[i]);
752 753
		kvm->buses[i] = NULL;
	}
754
	kvm_coalesced_mmio_free(kvm);
755 756
#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
	mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
757
#else
758
	kvm_arch_flush_shadow_all(kvm);
759
#endif
760
	kvm_arch_destroy_vm(kvm);
761
	kvm_destroy_devices(kvm);
762 763
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
		kvm_free_memslots(kvm, kvm->memslots[i]);
764
	cleanup_srcu_struct(&kvm->irq_srcu);
765 766
	cleanup_srcu_struct(&kvm->srcu);
	kvm_arch_free_vm(kvm);
767
	preempt_notifier_dec();
768
	hardware_disable_all();
769
	mmdrop(mm);
770 771
}

I
Izik Eidus 已提交
772 773
void kvm_get_kvm(struct kvm *kvm)
{
774
	refcount_inc(&kvm->users_count);
I
Izik Eidus 已提交
775 776 777 778 779
}
EXPORT_SYMBOL_GPL(kvm_get_kvm);

void kvm_put_kvm(struct kvm *kvm)
{
780
	if (refcount_dec_and_test(&kvm->users_count))
I
Izik Eidus 已提交
781 782 783 784 785
		kvm_destroy_vm(kvm);
}
EXPORT_SYMBOL_GPL(kvm_put_kvm);


786 787 788 789
static int kvm_vm_release(struct inode *inode, struct file *filp)
{
	struct kvm *kvm = filp->private_data;

G
Gregory Haskins 已提交
790 791
	kvm_irqfd_release(kvm);

I
Izik Eidus 已提交
792
	kvm_put_kvm(kvm);
A
Avi Kivity 已提交
793 794 795
	return 0;
}

796 797
/*
 * Allocation size is twice as large as the actual dirty bitmap size.
798
 * See x86's kvm_vm_ioctl_get_dirty_log() why this is needed.
799
 */
800 801
static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot)
{
802
	unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot);
803

804
	memslot->dirty_bitmap = kvm_kvzalloc(dirty_bytes);
805 806 807 808 809 810
	if (!memslot->dirty_bitmap)
		return -ENOMEM;

	return 0;
}

811
/*
812 813 814 815
 * 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.
816
 */
817 818
static void update_memslots(struct kvm_memslots *slots,
			    struct kvm_memory_slot *new)
819
{
820 821
	int id = new->id;
	int i = slots->id_to_index[id];
822
	struct kvm_memory_slot *mslots = slots->memslots;
823

824
	WARN_ON(mslots[i].id != id);
825
	if (!new->npages) {
826
		WARN_ON(!mslots[i].npages);
827 828 829 830 831 832
		if (mslots[i].npages)
			slots->used_slots--;
	} else {
		if (!mslots[i].npages)
			slots->used_slots++;
	}
833

834
	while (i < KVM_MEM_SLOTS_NUM - 1 &&
835 836 837
	       new->base_gfn <= mslots[i + 1].base_gfn) {
		if (!mslots[i + 1].npages)
			break;
838 839 840 841
		mslots[i] = mslots[i + 1];
		slots->id_to_index[mslots[i].id] = i;
		i++;
	}
842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858

	/*
	 * 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--;
		}
859 860
	} else
		WARN_ON_ONCE(i != slots->used_slots);
861

862 863
	mslots[i] = *new;
	slots->id_to_index[mslots[i].id] = i;
864 865
}

866
static int check_memory_region_flags(const struct kvm_userspace_memory_region *mem)
867
{
X
Xiao Guangrong 已提交
868 869
	u32 valid_flags = KVM_MEM_LOG_DIRTY_PAGES;

870
#ifdef __KVM_HAVE_READONLY_MEM
X
Xiao Guangrong 已提交
871 872 873 874
	valid_flags |= KVM_MEM_READONLY;
#endif

	if (mem->flags & ~valid_flags)
875 876 877 878 879
		return -EINVAL;

	return 0;
}

880
static struct kvm_memslots *install_new_memslots(struct kvm *kvm,
881
		int as_id, struct kvm_memslots *slots)
882
{
883
	struct kvm_memslots *old_memslots = __kvm_memslots(kvm, as_id);
884

885 886 887 888 889 890 891
	/*
	 * 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;

892
	rcu_assign_pointer(kvm->memslots[as_id], slots);
893
	synchronize_srcu_expedited(&kvm->srcu);
894

895 896 897 898
	/*
	 * 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.
899 900 901 902 903 904
	 *
	 * 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, ...
905
	 */
906
	slots->generation += KVM_ADDRESS_SPACE_NUM * 2 - 1;
907

908
	kvm_arch_memslots_updated(kvm, slots);
909 910

	return old_memslots;
911 912
}

A
Avi Kivity 已提交
913 914 915 916 917
/*
 * Allocate some memory and give it an address in the guest physical address
 * space.
 *
 * Discontiguous memory is allowed, mostly for framebuffers.
918
 *
919
 * Must be called holding kvm->slots_lock for write.
A
Avi Kivity 已提交
920
 */
921
int __kvm_set_memory_region(struct kvm *kvm,
922
			    const struct kvm_userspace_memory_region *mem)
A
Avi Kivity 已提交
923
{
924
	int r;
A
Avi Kivity 已提交
925
	gfn_t base_gfn;
926
	unsigned long npages;
927
	struct kvm_memory_slot *slot;
A
Avi Kivity 已提交
928
	struct kvm_memory_slot old, new;
929
	struct kvm_memslots *slots = NULL, *old_memslots;
930
	int as_id, id;
931
	enum kvm_mr_change change;
A
Avi Kivity 已提交
932

933 934 935 936
	r = check_memory_region_flags(mem);
	if (r)
		goto out;

A
Avi Kivity 已提交
937
	r = -EINVAL;
938 939 940
	as_id = mem->slot >> 16;
	id = (u16)mem->slot;

A
Avi Kivity 已提交
941 942 943 944 945
	/* General sanity checks */
	if (mem->memory_size & (PAGE_SIZE - 1))
		goto out;
	if (mem->guest_phys_addr & (PAGE_SIZE - 1))
		goto out;
946
	/* We can read the guest memory with __xxx_user() later on. */
947
	if ((id < KVM_USER_MEM_SLOTS) &&
948
	    ((mem->userspace_addr & (PAGE_SIZE - 1)) ||
949 950 951
	     !access_ok(VERIFY_WRITE,
			(void __user *)(unsigned long)mem->userspace_addr,
			mem->memory_size)))
952
		goto out;
953
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_MEM_SLOTS_NUM)
A
Avi Kivity 已提交
954 955 956 957
		goto out;
	if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
		goto out;

958
	slot = id_to_memslot(__kvm_memslots(kvm, as_id), id);
A
Avi Kivity 已提交
959 960 961
	base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
	npages = mem->memory_size >> PAGE_SHIFT;

962 963 964
	if (npages > KVM_MEM_MAX_NR_PAGES)
		goto out;

965
	new = old = *slot;
A
Avi Kivity 已提交
966

967
	new.id = id;
A
Avi Kivity 已提交
968 969 970 971
	new.base_gfn = base_gfn;
	new.npages = npages;
	new.flags = mem->flags;

972 973 974 975 976
	if (npages) {
		if (!old.npages)
			change = KVM_MR_CREATE;
		else { /* Modify an existing slot. */
			if ((mem->userspace_addr != old.userspace_addr) ||
977 978
			    (npages != old.npages) ||
			    ((new.flags ^ old.flags) & KVM_MEM_READONLY))
979 980 981 982 983 984 985 986 987 988 989
				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;
			}
		}
990 991 992 993
	} else {
		if (!old.npages)
			goto out;

994
		change = KVM_MR_DELETE;
995 996 997
		new.base_gfn = 0;
		new.flags = 0;
	}
A
Avi Kivity 已提交
998

999
	if ((change == KVM_MR_CREATE) || (change == KVM_MR_MOVE)) {
1000 1001
		/* Check for overlaps */
		r = -EEXIST;
1002
		kvm_for_each_memslot(slot, __kvm_memslots(kvm, as_id)) {
1003
			if ((slot->id >= KVM_USER_MEM_SLOTS) ||
1004
			    (slot->id == id))
1005 1006 1007 1008 1009
				continue;
			if (!((base_gfn + npages <= slot->base_gfn) ||
			      (base_gfn >= slot->base_gfn + slot->npages)))
				goto out;
		}
A
Avi Kivity 已提交
1010 1011 1012 1013
	}

	/* Free page dirty bitmap if unneeded */
	if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
A
Al Viro 已提交
1014
		new.dirty_bitmap = NULL;
A
Avi Kivity 已提交
1015 1016

	r = -ENOMEM;
1017
	if (change == KVM_MR_CREATE) {
1018
		new.userspace_addr = mem->userspace_addr;
1019

1020
		if (kvm_arch_create_memslot(kvm, &new, npages))
1021
			goto out_free;
A
Avi Kivity 已提交
1022
	}
1023

A
Avi Kivity 已提交
1024 1025
	/* Allocate page dirty bitmap if needed */
	if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
1026
		if (kvm_create_dirty_bitmap(&new) < 0)
1027
			goto out_free;
A
Avi Kivity 已提交
1028 1029
	}

1030
	slots = kvm_kvzalloc(sizeof(struct kvm_memslots));
1031 1032
	if (!slots)
		goto out_free;
1033
	memcpy(slots, __kvm_memslots(kvm, as_id), sizeof(struct kvm_memslots));
1034

1035
	if ((change == KVM_MR_DELETE) || (change == KVM_MR_MOVE)) {
1036
		slot = id_to_memslot(slots, id);
1037 1038
		slot->flags |= KVM_MEMSLOT_INVALID;

1039
		old_memslots = install_new_memslots(kvm, as_id, slots);
1040

1041 1042
		/* From this point no new shadow pages pointing to a deleted,
		 * or moved, memslot will be created.
1043 1044
		 *
		 * validation of sp->gfn happens in:
1045 1046
		 *	- gfn_to_hva (kvm_read_guest, gfn_to_pfn)
		 *	- kvm_is_visible_gfn (mmu_check_roots)
1047
		 */
1048
		kvm_arch_flush_shadow_memslot(kvm, slot);
1049 1050 1051 1052 1053 1054

		/*
		 * 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.
		 */
1055
		slots = old_memslots;
1056
	}
1057

1058
	r = kvm_arch_prepare_memory_region(kvm, &new, mem, change);
1059
	if (r)
1060
		goto out_slots;
1061

1062
	/* actual memory is freed via old in kvm_free_memslot below */
1063
	if (change == KVM_MR_DELETE) {
1064
		new.dirty_bitmap = NULL;
1065
		memset(&new.arch, 0, sizeof(new.arch));
1066 1067
	}

1068
	update_memslots(slots, &new);
1069
	old_memslots = install_new_memslots(kvm, as_id, slots);
1070

1071
	kvm_arch_commit_memory_region(kvm, mem, &old, &new, change);
1072

1073
	kvm_free_memslot(kvm, &old, &new);
1074
	kvfree(old_memslots);
A
Avi Kivity 已提交
1075 1076
	return 0;

1077
out_slots:
1078
	kvfree(slots);
1079
out_free:
1080
	kvm_free_memslot(kvm, &new, &old);
A
Avi Kivity 已提交
1081 1082
out:
	return r;
1083
}
1084 1085 1086
EXPORT_SYMBOL_GPL(__kvm_set_memory_region);

int kvm_set_memory_region(struct kvm *kvm,
1087
			  const struct kvm_userspace_memory_region *mem)
1088 1089 1090
{
	int r;

1091
	mutex_lock(&kvm->slots_lock);
1092
	r = __kvm_set_memory_region(kvm, mem);
1093
	mutex_unlock(&kvm->slots_lock);
1094 1095
	return r;
}
1096 1097
EXPORT_SYMBOL_GPL(kvm_set_memory_region);

1098 1099
static int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
					  struct kvm_userspace_memory_region *mem)
1100
{
1101
	if ((u16)mem->slot >= KVM_USER_MEM_SLOTS)
1102
		return -EINVAL;
1103

1104
	return kvm_set_memory_region(kvm, mem);
A
Avi Kivity 已提交
1105 1106
}

1107 1108
int kvm_get_dirty_log(struct kvm *kvm,
			struct kvm_dirty_log *log, int *is_dirty)
A
Avi Kivity 已提交
1109
{
1110
	struct kvm_memslots *slots;
A
Avi Kivity 已提交
1111
	struct kvm_memory_slot *memslot;
1112
	int i, as_id, id;
1113
	unsigned long n;
A
Avi Kivity 已提交
1114 1115
	unsigned long any = 0;

1116 1117 1118
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1119
		return -EINVAL;
A
Avi Kivity 已提交
1120

1121 1122
	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
A
Avi Kivity 已提交
1123
	if (!memslot->dirty_bitmap)
1124
		return -ENOENT;
A
Avi Kivity 已提交
1125

1126
	n = kvm_dirty_bitmap_bytes(memslot);
A
Avi Kivity 已提交
1127

1128
	for (i = 0; !any && i < n/sizeof(long); ++i)
A
Avi Kivity 已提交
1129 1130 1131
		any = memslot->dirty_bitmap[i];

	if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
1132
		return -EFAULT;
A
Avi Kivity 已提交
1133

1134 1135
	if (any)
		*is_dirty = 1;
1136
	return 0;
A
Avi Kivity 已提交
1137
}
1138
EXPORT_SYMBOL_GPL(kvm_get_dirty_log);
A
Avi Kivity 已提交
1139

1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165
#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)
{
1166
	struct kvm_memslots *slots;
1167
	struct kvm_memory_slot *memslot;
1168
	int i, as_id, id;
1169 1170 1171 1172
	unsigned long n;
	unsigned long *dirty_bitmap;
	unsigned long *dirty_bitmap_buffer;

1173 1174 1175
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1176
		return -EINVAL;
1177

1178 1179
	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
1180 1181 1182

	dirty_bitmap = memslot->dirty_bitmap;
	if (!dirty_bitmap)
1183
		return -ENOENT;
1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203

	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;

1204 1205 1206 1207 1208
		if (mask) {
			offset = i * BITS_PER_LONG;
			kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot,
								offset, mask);
		}
1209 1210 1211 1212
	}

	spin_unlock(&kvm->mmu_lock);
	if (copy_to_user(log->dirty_bitmap, dirty_bitmap_buffer, n))
1213 1214
		return -EFAULT;
	return 0;
1215 1216 1217 1218
}
EXPORT_SYMBOL_GPL(kvm_get_dirty_log_protect);
#endif

1219 1220 1221 1222 1223
bool kvm_largepages_enabled(void)
{
	return largepages_enabled;
}

1224 1225 1226 1227 1228 1229
void kvm_disable_largepages(void)
{
	largepages_enabled = false;
}
EXPORT_SYMBOL_GPL(kvm_disable_largepages);

1230 1231 1232 1233
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
1234
EXPORT_SYMBOL_GPL(gfn_to_memslot);
A
Avi Kivity 已提交
1235

1236 1237 1238 1239 1240
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);
}

1241
bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
1242
{
1243
	struct kvm_memory_slot *memslot = gfn_to_memslot(kvm, gfn);
1244

1245
	if (!memslot || memslot->id >= KVM_USER_MEM_SLOTS ||
1246
	      memslot->flags & KVM_MEMSLOT_INVALID)
1247
		return false;
1248

1249
	return true;
1250 1251 1252
}
EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);

J
Joerg Roedel 已提交
1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276
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
1277 1278 1279 1280 1281 1282 1283
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
1284
{
1285
	if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
X
Xiao Guangrong 已提交
1286
		return KVM_HVA_ERR_BAD;
1287

X
Xiao Guangrong 已提交
1288 1289
	if (memslot_is_readonly(slot) && write)
		return KVM_HVA_ERR_RO_BAD;
1290 1291 1292 1293

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

X
Xiao Guangrong 已提交
1294
	return __gfn_to_hva_memslot(slot, gfn);
I
Izik Eidus 已提交
1295
}
1296

X
Xiao Guangrong 已提交
1297 1298 1299 1300
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
1301
}
1302

X
Xiao Guangrong 已提交
1303
unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot,
1304
					gfn_t gfn)
X
Xiao Guangrong 已提交
1305 1306 1307 1308 1309
{
	return gfn_to_hva_many(slot, gfn, NULL);
}
EXPORT_SYMBOL_GPL(gfn_to_hva_memslot);

1310 1311
unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
{
1312
	return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL);
1313
}
1314
EXPORT_SYMBOL_GPL(gfn_to_hva);
I
Izik Eidus 已提交
1315

1316 1317 1318 1319 1320 1321
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);

1322
/*
1323 1324
 * If writable is set to false, the hva returned by this function is only
 * allowed to be read.
1325
 */
1326 1327
unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot,
				      gfn_t gfn, bool *writable)
1328
{
1329 1330 1331
	unsigned long hva = __gfn_to_hva_many(slot, gfn, NULL, false);

	if (!kvm_is_error_hva(hva) && writable)
1332 1333
		*writable = !memslot_is_readonly(slot);

1334
	return hva;
1335 1336
}

1337 1338 1339 1340 1341 1342 1343
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);
}

1344 1345 1346 1347 1348 1349 1350
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);
}

1351 1352
static int get_user_page_nowait(unsigned long start, int write,
		struct page **page)
1353
{
L
Lorenzo Stoakes 已提交
1354
	int flags = FOLL_NOWAIT | FOLL_HWPOISON;
1355 1356 1357 1358

	if (write)
		flags |= FOLL_WRITE;

L
Lorenzo Stoakes 已提交
1359
	return get_user_pages(start, 1, flags, page, NULL);
1360 1361
}

1362 1363
static inline int check_user_page_hwpoison(unsigned long addr)
{
L
Lorenzo Stoakes 已提交
1364
	int rc, flags = FOLL_HWPOISON | FOLL_WRITE;
1365

L
Lorenzo Stoakes 已提交
1366
	rc = get_user_pages(addr, 1, flags, NULL, NULL);
1367 1368 1369
	return rc == -EHWPOISON;
}

X
Xiao Guangrong 已提交
1370 1371 1372 1373 1374
/*
 * 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
1375
			    bool write_fault, bool *writable, kvm_pfn_t *pfn)
A
Avi Kivity 已提交
1376
{
1377
	struct page *page[1];
X
Xiao Guangrong 已提交
1378
	int npages;
A
Avi Kivity 已提交
1379

X
Xiao Guangrong 已提交
1380 1381
	if (!(async || atomic))
		return false;
1382

1383 1384 1385 1386 1387 1388 1389
	/*
	 * 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;
1390

X
Xiao Guangrong 已提交
1391 1392 1393
	npages = __get_user_pages_fast(addr, 1, 1, page);
	if (npages == 1) {
		*pfn = page_to_pfn(page[0]);
1394

X
Xiao Guangrong 已提交
1395 1396 1397 1398
		if (writable)
			*writable = true;
		return true;
	}
1399

X
Xiao Guangrong 已提交
1400 1401
	return false;
}
1402

X
Xiao Guangrong 已提交
1403 1404 1405 1406 1407
/*
 * 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
1408
			   bool *writable, kvm_pfn_t *pfn)
X
Xiao Guangrong 已提交
1409 1410 1411
{
	struct page *page[1];
	int npages = 0;
1412

X
Xiao Guangrong 已提交
1413 1414 1415 1416 1417 1418 1419
	might_sleep();

	if (writable)
		*writable = write_fault;

	if (async) {
		down_read(&current->mm->mmap_sem);
1420
		npages = get_user_page_nowait(addr, write_fault, page);
X
Xiao Guangrong 已提交
1421
		up_read(&current->mm->mmap_sem);
1422
	} else {
1423
		unsigned int flags = FOLL_HWPOISON;
1424 1425 1426 1427

		if (write_fault)
			flags |= FOLL_WRITE;

1428
		npages = get_user_pages_unlocked(addr, 1, page, flags);
1429
	}
X
Xiao Guangrong 已提交
1430 1431 1432 1433
	if (npages != 1)
		return npages;

	/* map read fault as writable if possible */
1434
	if (unlikely(!write_fault) && writable) {
X
Xiao Guangrong 已提交
1435 1436 1437 1438 1439 1440 1441
		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];
1442
		}
X
Xiao Guangrong 已提交
1443 1444

		npages = 1;
1445
	}
X
Xiao Guangrong 已提交
1446 1447 1448
	*pfn = page_to_pfn(page[0]);
	return npages;
}
I
Izik Eidus 已提交
1449

X
Xiao Guangrong 已提交
1450 1451 1452 1453
static bool vma_is_valid(struct vm_area_struct *vma, bool write_fault)
{
	if (unlikely(!(vma->vm_flags & VM_READ)))
		return false;
1454

X
Xiao Guangrong 已提交
1455 1456
	if (write_fault && (unlikely(!(vma->vm_flags & VM_WRITE))))
		return false;
1457

X
Xiao Guangrong 已提交
1458 1459
	return true;
}
1460

1461 1462 1463 1464
static int hva_to_pfn_remapped(struct vm_area_struct *vma,
			       unsigned long addr, bool *async,
			       bool write_fault, kvm_pfn_t *p_pfn)
{
1465 1466 1467 1468 1469 1470 1471 1472 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
	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;
1504 1505 1506
	return 0;
}

1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520
/*
 * 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
1521
static kvm_pfn_t hva_to_pfn(unsigned long addr, bool atomic, bool *async,
X
Xiao Guangrong 已提交
1522 1523 1524
			bool write_fault, bool *writable)
{
	struct vm_area_struct *vma;
D
Dan Williams 已提交
1525
	kvm_pfn_t pfn = 0;
1526
	int npages, r;
1527

X
Xiao Guangrong 已提交
1528 1529
	/* we can do it either atomically or asynchronously, not both */
	BUG_ON(atomic && async);
1530

X
Xiao Guangrong 已提交
1531 1532 1533 1534 1535 1536 1537 1538 1539
	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;
1540

X
Xiao Guangrong 已提交
1541 1542 1543 1544 1545 1546 1547
	down_read(&current->mm->mmap_sem);
	if (npages == -EHWPOISON ||
	      (!async && check_user_page_hwpoison(addr))) {
		pfn = KVM_PFN_ERR_HWPOISON;
		goto exit;
	}

1548
retry:
X
Xiao Guangrong 已提交
1549 1550 1551 1552
	vma = find_vma_intersection(current->mm, addr, addr + 1);

	if (vma == NULL)
		pfn = KVM_PFN_ERR_FAULT;
1553 1554
	else if (vma->vm_flags & (VM_IO | VM_PFNMAP)) {
		r = hva_to_pfn_remapped(vma, addr, async, write_fault, &pfn);
1555 1556
		if (r == -EAGAIN)
			goto retry;
1557 1558
		if (r < 0)
			pfn = KVM_PFN_ERR_FAULT;
X
Xiao Guangrong 已提交
1559
	} else {
X
Xiao Guangrong 已提交
1560
		if (async && vma_is_valid(vma, write_fault))
X
Xiao Guangrong 已提交
1561 1562 1563 1564 1565
			*async = true;
		pfn = KVM_PFN_ERR_FAULT;
	}
exit:
	up_read(&current->mm->mmap_sem);
1566
	return pfn;
1567 1568
}

D
Dan Williams 已提交
1569 1570 1571
kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
			       bool atomic, bool *async, bool write_fault,
			       bool *writable)
1572
{
X
Xiao Guangrong 已提交
1573 1574
	unsigned long addr = __gfn_to_hva_many(slot, gfn, NULL, write_fault);

1575 1576 1577
	if (addr == KVM_HVA_ERR_RO_BAD) {
		if (writable)
			*writable = false;
X
Xiao Guangrong 已提交
1578
		return KVM_PFN_ERR_RO_FAULT;
1579
	}
X
Xiao Guangrong 已提交
1580

1581 1582 1583
	if (kvm_is_error_hva(addr)) {
		if (writable)
			*writable = false;
1584
		return KVM_PFN_NOSLOT;
1585
	}
X
Xiao Guangrong 已提交
1586 1587 1588 1589 1590 1591 1592 1593 1594

	/* 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);
1595
}
1596
EXPORT_SYMBOL_GPL(__gfn_to_pfn_memslot);
1597

D
Dan Williams 已提交
1598
kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1599 1600
		      bool *writable)
{
P
Paolo Bonzini 已提交
1601 1602
	return __gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn, false, NULL,
				    write_fault, writable);
1603 1604 1605
}
EXPORT_SYMBOL_GPL(gfn_to_pfn_prot);

D
Dan Williams 已提交
1606
kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1607
{
X
Xiao Guangrong 已提交
1608
	return __gfn_to_pfn_memslot(slot, gfn, false, NULL, true, NULL);
1609
}
P
Paolo Bonzini 已提交
1610
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot);
1611

D
Dan Williams 已提交
1612
kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn)
1613
{
X
Xiao Guangrong 已提交
1614
	return __gfn_to_pfn_memslot(slot, gfn, true, NULL, true, NULL);
1615
}
1616
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot_atomic);
1617

D
Dan Williams 已提交
1618
kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1619 1620 1621 1622 1623
{
	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
1624
kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn)
1625 1626 1627 1628 1629
{
	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
1630
kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1631 1632 1633 1634 1635
{
	return gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn);
}
EXPORT_SYMBOL_GPL(gfn_to_pfn);

D
Dan Williams 已提交
1636
kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn)
1637 1638 1639 1640 1641
{
	return gfn_to_pfn_memslot(kvm_vcpu_gfn_to_memslot(vcpu, gfn), gfn);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_gfn_to_pfn);

1642 1643
int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
			    struct page **pages, int nr_pages)
1644 1645 1646 1647
{
	unsigned long addr;
	gfn_t entry;

1648
	addr = gfn_to_hva_many(slot, gfn, &entry);
1649 1650 1651 1652 1653 1654 1655 1656 1657 1658
	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
1659
static struct page *kvm_pfn_to_page(kvm_pfn_t pfn)
X
Xiao Guangrong 已提交
1660
{
1661
	if (is_error_noslot_pfn(pfn))
1662
		return KVM_ERR_PTR_BAD_PAGE;
X
Xiao Guangrong 已提交
1663

1664
	if (kvm_is_reserved_pfn(pfn)) {
1665
		WARN_ON(1);
1666
		return KVM_ERR_PTR_BAD_PAGE;
1667
	}
X
Xiao Guangrong 已提交
1668 1669 1670 1671

	return pfn_to_page(pfn);
}

1672 1673
struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
{
D
Dan Williams 已提交
1674
	kvm_pfn_t pfn;
1675 1676 1677

	pfn = gfn_to_pfn(kvm, gfn);

X
Xiao Guangrong 已提交
1678
	return kvm_pfn_to_page(pfn);
A
Avi Kivity 已提交
1679 1680 1681
}
EXPORT_SYMBOL_GPL(gfn_to_page);

1682 1683
struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn)
{
D
Dan Williams 已提交
1684
	kvm_pfn_t pfn;
1685 1686 1687 1688 1689 1690 1691

	pfn = kvm_vcpu_gfn_to_pfn(vcpu, gfn);

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

1692 1693
void kvm_release_page_clean(struct page *page)
{
1694 1695
	WARN_ON(is_error_page(page));

1696
	kvm_release_pfn_clean(page_to_pfn(page));
1697 1698 1699
}
EXPORT_SYMBOL_GPL(kvm_release_page_clean);

D
Dan Williams 已提交
1700
void kvm_release_pfn_clean(kvm_pfn_t pfn)
1701
{
1702
	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn))
1703
		put_page(pfn_to_page(pfn));
1704 1705 1706
}
EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);

1707
void kvm_release_page_dirty(struct page *page)
1708
{
X
Xiao Guangrong 已提交
1709 1710
	WARN_ON(is_error_page(page));

1711 1712 1713 1714
	kvm_release_pfn_dirty(page_to_pfn(page));
}
EXPORT_SYMBOL_GPL(kvm_release_page_dirty);

D
Dan Williams 已提交
1715
static void kvm_release_pfn_dirty(kvm_pfn_t pfn)
1716 1717 1718 1719 1720
{
	kvm_set_pfn_dirty(pfn);
	kvm_release_pfn_clean(pfn);
}

D
Dan Williams 已提交
1721
void kvm_set_pfn_dirty(kvm_pfn_t pfn)
1722
{
1723
	if (!kvm_is_reserved_pfn(pfn)) {
1724
		struct page *page = pfn_to_page(pfn);
1725

1726 1727 1728
		if (!PageReserved(page))
			SetPageDirty(page);
	}
1729
}
1730 1731
EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);

D
Dan Williams 已提交
1732
void kvm_set_pfn_accessed(kvm_pfn_t pfn)
1733
{
1734
	if (!kvm_is_reserved_pfn(pfn))
1735
		mark_page_accessed(pfn_to_page(pfn));
1736 1737 1738
}
EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);

D
Dan Williams 已提交
1739
void kvm_get_pfn(kvm_pfn_t pfn)
1740
{
1741
	if (!kvm_is_reserved_pfn(pfn))
1742
		get_page(pfn_to_page(pfn));
1743 1744
}
EXPORT_SYMBOL_GPL(kvm_get_pfn);
1745

1746 1747 1748 1749 1750 1751 1752 1753
static int next_segment(unsigned long len, int offset)
{
	if (len > PAGE_SIZE - offset)
		return PAGE_SIZE - offset;
	else
		return len;
}

1754 1755
static int __kvm_read_guest_page(struct kvm_memory_slot *slot, gfn_t gfn,
				 void *data, int offset, int len)
1756
{
1757 1758
	int r;
	unsigned long addr;
1759

1760
	addr = gfn_to_hva_memslot_prot(slot, gfn, NULL);
1761 1762
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1763
	r = __copy_from_user(data, (void __user *)addr + offset, len);
1764
	if (r)
1765 1766 1767
		return -EFAULT;
	return 0;
}
1768 1769 1770 1771 1772 1773 1774 1775

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);
}
1776 1777
EXPORT_SYMBOL_GPL(kvm_read_guest_page);

1778 1779 1780 1781 1782 1783 1784 1785 1786
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);

1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806
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);

1807
int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data, unsigned long len)
1808 1809
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
1810
	int seg;
1811
	int offset = offset_in_page(gpa);
1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825
	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);
1826

1827 1828 1829 1830 1831 1832 1833
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);
1834 1835
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1836
	pagefault_disable();
1837
	r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
1838
	pagefault_enable();
1839 1840 1841 1842 1843
	if (r)
		return -EFAULT;
	return 0;
}

1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867
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)
1868
{
1869 1870
	int r;
	unsigned long addr;
1871

1872
	addr = gfn_to_hva_memslot(memslot, gfn);
1873 1874
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1875
	r = __copy_to_user((void __user *)addr + offset, data, len);
1876
	if (r)
1877
		return -EFAULT;
1878
	mark_page_dirty_in_slot(memslot, gfn);
1879 1880
	return 0;
}
1881 1882 1883 1884 1885 1886 1887 1888

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);
}
1889 1890
EXPORT_SYMBOL_GPL(kvm_write_guest_page);

1891 1892 1893 1894 1895 1896 1897 1898 1899
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);

1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918
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;
}
1919
EXPORT_SYMBOL_GPL(kvm_write_guest);
1920

1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941
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);

1942 1943 1944
static int __kvm_gfn_to_hva_cache_init(struct kvm_memslots *slots,
				       struct gfn_to_hva_cache *ghc,
				       gpa_t gpa, unsigned long len)
1945 1946
{
	int offset = offset_in_page(gpa);
1947 1948 1949 1950
	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;
1951 1952 1953

	ghc->gpa = gpa;
	ghc->generation = slots->generation;
1954
	ghc->len = len;
1955
	ghc->memslot = __gfn_to_memslot(slots, start_gfn);
1956 1957
	ghc->hva = gfn_to_hva_many(ghc->memslot, start_gfn, NULL);
	if (!kvm_is_error_hva(ghc->hva) && nr_pages_needed <= 1) {
1958
		ghc->hva += offset;
1959 1960 1961 1962 1963 1964
	} else {
		/*
		 * If the requested region crosses two memslots, we still
		 * verify that the entire region is valid here.
		 */
		while (start_gfn <= end_gfn) {
1965
			ghc->memslot = __gfn_to_memslot(slots, start_gfn);
1966 1967 1968 1969 1970 1971 1972 1973 1974
			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;
	}
1975 1976
	return 0;
}
1977

1978
int kvm_vcpu_gfn_to_hva_cache_init(struct kvm_vcpu *vcpu, struct gfn_to_hva_cache *ghc,
1979 1980
			      gpa_t gpa, unsigned long len)
{
1981
	struct kvm_memslots *slots = kvm_vcpu_memslots(vcpu);
1982 1983
	return __kvm_gfn_to_hva_cache_init(slots, ghc, gpa, len);
}
1984
EXPORT_SYMBOL_GPL(kvm_vcpu_gfn_to_hva_cache_init);
1985

1986 1987
int kvm_vcpu_write_guest_offset_cached(struct kvm_vcpu *vcpu, struct gfn_to_hva_cache *ghc,
				       void *data, int offset, unsigned long len)
1988
{
1989
	struct kvm_memslots *slots = kvm_vcpu_memslots(vcpu);
1990
	int r;
1991
	gpa_t gpa = ghc->gpa + offset;
1992

1993
	BUG_ON(len + offset > ghc->len);
1994

1995
	if (slots->generation != ghc->generation)
1996
		__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len);
1997 1998

	if (unlikely(!ghc->memslot))
1999
		return kvm_vcpu_write_guest(vcpu, gpa, data, len);
2000 2001 2002 2003

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

2004
	r = __copy_to_user((void __user *)ghc->hva + offset, data, len);
2005 2006
	if (r)
		return -EFAULT;
2007
	mark_page_dirty_in_slot(ghc->memslot, gpa >> PAGE_SHIFT);
2008 2009 2010

	return 0;
}
2011
EXPORT_SYMBOL_GPL(kvm_vcpu_write_guest_offset_cached);
2012

2013 2014
int kvm_vcpu_write_guest_cached(struct kvm_vcpu *vcpu, struct gfn_to_hva_cache *ghc,
			       void *data, unsigned long len)
2015
{
2016
	return kvm_vcpu_write_guest_offset_cached(vcpu, ghc, data, 0, len);
2017
}
2018
EXPORT_SYMBOL_GPL(kvm_vcpu_write_guest_cached);
2019

2020 2021
int kvm_vcpu_read_guest_cached(struct kvm_vcpu *vcpu, struct gfn_to_hva_cache *ghc,
			       void *data, unsigned long len)
2022
{
2023
	struct kvm_memslots *slots = kvm_vcpu_memslots(vcpu);
2024 2025
	int r;

2026 2027
	BUG_ON(len > ghc->len);

2028
	if (slots->generation != ghc->generation)
2029
		__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len);
2030 2031

	if (unlikely(!ghc->memslot))
2032
		return kvm_vcpu_read_guest(vcpu, ghc->gpa, data, len);
2033 2034 2035 2036 2037 2038 2039 2040 2041 2042

	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;
}
2043
EXPORT_SYMBOL_GPL(kvm_vcpu_read_guest_cached);
2044

2045 2046
int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
{
2047 2048 2049
	const void *zero_page = (const void *) __va(page_to_phys(ZERO_PAGE(0)));

	return kvm_write_guest_page(kvm, gfn, zero_page, offset, len);
2050 2051 2052 2053 2054 2055 2056 2057 2058 2059
}
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;

2060
	while ((seg = next_segment(len, offset)) != 0) {
2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071
		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);

2072
static void mark_page_dirty_in_slot(struct kvm_memory_slot *memslot,
2073
				    gfn_t gfn)
A
Avi Kivity 已提交
2074
{
R
Rusty Russell 已提交
2075 2076
	if (memslot && memslot->dirty_bitmap) {
		unsigned long rel_gfn = gfn - memslot->base_gfn;
A
Avi Kivity 已提交
2077

2078
		set_bit_le(rel_gfn, memslot->dirty_bitmap);
A
Avi Kivity 已提交
2079 2080 2081
	}
}

2082 2083 2084 2085 2086
void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
{
	struct kvm_memory_slot *memslot;

	memslot = gfn_to_memslot(kvm, gfn);
2087
	mark_page_dirty_in_slot(memslot, gfn);
2088
}
2089
EXPORT_SYMBOL_GPL(mark_page_dirty);
2090

2091 2092 2093 2094 2095 2096 2097 2098 2099
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
2100 2101
static void grow_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2102
	unsigned int old, val, grow;
W
Wanpeng Li 已提交
2103

2104
	old = val = vcpu->halt_poll_ns;
2105
	grow = READ_ONCE(halt_poll_ns_grow);
W
Wanpeng Li 已提交
2106
	/* 10us base */
2107
	if (val == 0 && grow)
W
Wanpeng Li 已提交
2108 2109
		val = 10000;
	else
2110
		val *= grow;
W
Wanpeng Li 已提交
2111

2112 2113 2114
	if (val > halt_poll_ns)
		val = halt_poll_ns;

W
Wanpeng Li 已提交
2115
	vcpu->halt_poll_ns = val;
2116
	trace_kvm_halt_poll_ns_grow(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2117 2118 2119 2120
}

static void shrink_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2121
	unsigned int old, val, shrink;
W
Wanpeng Li 已提交
2122

2123
	old = val = vcpu->halt_poll_ns;
2124 2125
	shrink = READ_ONCE(halt_poll_ns_shrink);
	if (shrink == 0)
W
Wanpeng Li 已提交
2126 2127
		val = 0;
	else
2128
		val /= shrink;
W
Wanpeng Li 已提交
2129 2130

	vcpu->halt_poll_ns = val;
2131
	trace_kvm_halt_poll_ns_shrink(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2132 2133
}

2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147
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
2148 2149 2150
/*
 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
 */
2151
void kvm_vcpu_block(struct kvm_vcpu *vcpu)
2152
{
2153
	ktime_t start, cur;
2154
	DECLARE_SWAITQUEUE(wait);
2155
	bool waited = false;
W
Wanpeng Li 已提交
2156
	u64 block_ns;
2157 2158

	start = cur = ktime_get();
W
Wanpeng Li 已提交
2159 2160
	if (vcpu->halt_poll_ns) {
		ktime_t stop = ktime_add_ns(ktime_get(), vcpu->halt_poll_ns);
2161

2162
		++vcpu->stat.halt_attempted_poll;
2163 2164 2165 2166 2167 2168 2169
		do {
			/*
			 * This sets KVM_REQ_UNHALT if an interrupt
			 * arrives.
			 */
			if (kvm_vcpu_check_block(vcpu) < 0) {
				++vcpu->stat.halt_successful_poll;
2170 2171
				if (!vcpu_valid_wakeup(vcpu))
					++vcpu->stat.halt_poll_invalid;
2172 2173 2174 2175 2176
				goto out;
			}
			cur = ktime_get();
		} while (single_task_running() && ktime_before(cur, stop));
	}
2177

2178 2179
	kvm_arch_vcpu_blocking(vcpu);

2180
	for (;;) {
2181
		prepare_to_swait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
2182

2183
		if (kvm_vcpu_check_block(vcpu) < 0)
2184 2185
			break;

2186
		waited = true;
E
Eddie Dong 已提交
2187 2188
		schedule();
	}
2189

2190
	finish_swait(&vcpu->wq, &wait);
2191 2192
	cur = ktime_get();

2193
	kvm_arch_vcpu_unblocking(vcpu);
2194
out:
W
Wanpeng Li 已提交
2195 2196
	block_ns = ktime_to_ns(cur) - ktime_to_ns(start);

2197 2198 2199
	if (!vcpu_valid_wakeup(vcpu))
		shrink_halt_poll_ns(vcpu);
	else if (halt_poll_ns) {
W
Wanpeng Li 已提交
2200 2201 2202
		if (block_ns <= vcpu->halt_poll_ns)
			;
		/* we had a long block, shrink polling */
2203
		else if (vcpu->halt_poll_ns && block_ns > halt_poll_ns)
W
Wanpeng Li 已提交
2204 2205 2206 2207 2208
			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);
2209 2210
	} else
		vcpu->halt_poll_ns = 0;
W
Wanpeng Li 已提交
2211

2212 2213
	trace_kvm_vcpu_wakeup(block_ns, waited, vcpu_valid_wakeup(vcpu));
	kvm_arch_vcpu_block_finish(vcpu);
E
Eddie Dong 已提交
2214
}
2215
EXPORT_SYMBOL_GPL(kvm_vcpu_block);
E
Eddie Dong 已提交
2216

2217
bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu)
2218
{
2219
	struct swait_queue_head *wqp;
2220 2221

	wqp = kvm_arch_vcpu_wq(vcpu);
2222 2223
	if (swait_active(wqp)) {
		swake_up(wqp);
2224
		++vcpu->stat.halt_wakeup;
2225
		return true;
2226 2227
	}

2228
	return false;
2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239
}
EXPORT_SYMBOL_GPL(kvm_vcpu_wake_up);

/*
 * 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;

2240 2241 2242
	if (kvm_vcpu_wake_up(vcpu))
		return;

2243 2244 2245 2246 2247 2248
	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();
}
2249
EXPORT_SYMBOL_GPL(kvm_vcpu_kick);
2250

2251
int kvm_vcpu_yield_to(struct kvm_vcpu *target)
2252 2253 2254
{
	struct pid *pid;
	struct task_struct *task = NULL;
2255
	int ret = 0;
2256 2257 2258 2259

	rcu_read_lock();
	pid = rcu_dereference(target->pid);
	if (pid)
2260
		task = get_pid_task(pid, PIDTYPE_PID);
2261 2262
	rcu_read_unlock();
	if (!task)
2263 2264
		return ret;
	ret = yield_to(task, 1);
2265
	put_task_struct(task);
2266 2267

	return ret;
2268 2269 2270
}
EXPORT_SYMBOL_GPL(kvm_vcpu_yield_to);

2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292
/*
 * 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.
 */
2293
static bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
2294
{
2295
#ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
2296 2297 2298
	bool eligible;

	eligible = !vcpu->spin_loop.in_spin_loop ||
2299
		    vcpu->spin_loop.dy_eligible;
2300 2301 2302 2303 2304

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

	return eligible;
2305 2306
#else
	return true;
2307
#endif
2308
}
2309

2310
void kvm_vcpu_on_spin(struct kvm_vcpu *me)
Z
Zhai, Edwin 已提交
2311
{
2312 2313 2314 2315
	struct kvm *kvm = me->kvm;
	struct kvm_vcpu *vcpu;
	int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
	int yielded = 0;
2316
	int try = 3;
2317 2318
	int pass;
	int i;
Z
Zhai, Edwin 已提交
2319

2320
	kvm_vcpu_set_in_spin_loop(me, true);
2321 2322 2323 2324 2325 2326 2327
	/*
	 * 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.
	 */
2328
	for (pass = 0; pass < 2 && !yielded && try; pass++) {
2329
		kvm_for_each_vcpu(i, vcpu, kvm) {
2330
			if (!pass && i <= last_boosted_vcpu) {
2331 2332 2333 2334
				i = last_boosted_vcpu;
				continue;
			} else if (pass && i > last_boosted_vcpu)
				break;
2335 2336
			if (!ACCESS_ONCE(vcpu->preempted))
				continue;
2337 2338
			if (vcpu == me)
				continue;
2339
			if (swait_active(&vcpu->wq) && !kvm_arch_vcpu_runnable(vcpu))
2340
				continue;
2341 2342
			if (!kvm_vcpu_eligible_for_directed_yield(vcpu))
				continue;
2343 2344 2345

			yielded = kvm_vcpu_yield_to(vcpu);
			if (yielded > 0) {
2346 2347
				kvm->last_boosted_vcpu = i;
				break;
2348 2349 2350 2351
			} else if (yielded < 0) {
				try--;
				if (!try)
					break;
2352 2353 2354
			}
		}
	}
2355
	kvm_vcpu_set_in_spin_loop(me, false);
2356 2357 2358

	/* Ensure vcpu is not eligible during next spinloop */
	kvm_vcpu_set_dy_eligible(me, false);
Z
Zhai, Edwin 已提交
2359 2360 2361
}
EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin);

2362
static int kvm_vcpu_fault(struct vm_fault *vmf)
2363
{
2364
	struct kvm_vcpu *vcpu = vmf->vma->vm_file->private_data;
2365 2366
	struct page *page;

2367
	if (vmf->pgoff == 0)
2368
		page = virt_to_page(vcpu->run);
A
Avi Kivity 已提交
2369
#ifdef CONFIG_X86
2370
	else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
2371
		page = virt_to_page(vcpu->arch.pio_data);
2372
#endif
2373
#ifdef CONFIG_KVM_MMIO
2374 2375
	else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
		page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
A
Avi Kivity 已提交
2376
#endif
2377
	else
2378
		return kvm_arch_vcpu_fault(vcpu, vmf);
2379
	get_page(page);
2380 2381
	vmf->page = page;
	return 0;
2382 2383
}

2384
static const struct vm_operations_struct kvm_vcpu_vm_ops = {
2385
	.fault = kvm_vcpu_fault,
2386 2387 2388 2389 2390 2391 2392 2393
};

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
2394 2395 2396 2397
static int kvm_vcpu_release(struct inode *inode, struct file *filp)
{
	struct kvm_vcpu *vcpu = filp->private_data;

2398
	debugfs_remove_recursive(vcpu->debugfs_dentry);
A
Al Viro 已提交
2399
	kvm_put_kvm(vcpu->kvm);
A
Avi Kivity 已提交
2400 2401 2402
	return 0;
}

2403
static struct file_operations kvm_vcpu_fops = {
A
Avi Kivity 已提交
2404 2405
	.release        = kvm_vcpu_release,
	.unlocked_ioctl = kvm_vcpu_ioctl,
2406
#ifdef CONFIG_KVM_COMPAT
2407 2408
	.compat_ioctl   = kvm_vcpu_compat_ioctl,
#endif
2409
	.mmap           = kvm_vcpu_mmap,
2410
	.llseek		= noop_llseek,
A
Avi Kivity 已提交
2411 2412 2413 2414 2415 2416 2417
};

/*
 * Allocates an inode for the vcpu.
 */
static int create_vcpu_fd(struct kvm_vcpu *vcpu)
{
2418
	return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, O_RDWR | O_CLOEXEC);
A
Avi Kivity 已提交
2419 2420
}

2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446
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;
}

2447 2448 2449
/*
 * Creates some virtual cpus.  Good luck creating more than one.
 */
2450
static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
2451 2452
{
	int r;
2453
	struct kvm_vcpu *vcpu;
2454

G
Greg Kurz 已提交
2455
	if (id >= KVM_MAX_VCPU_ID)
2456 2457
		return -EINVAL;

2458 2459 2460 2461 2462 2463 2464 2465 2466
	mutex_lock(&kvm->lock);
	if (kvm->created_vcpus == KVM_MAX_VCPUS) {
		mutex_unlock(&kvm->lock);
		return -EINVAL;
	}

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

2467
	vcpu = kvm_arch_vcpu_create(kvm, id);
2468 2469 2470 2471
	if (IS_ERR(vcpu)) {
		r = PTR_ERR(vcpu);
		goto vcpu_decrement;
	}
2472

2473 2474
	preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);

2475 2476
	r = kvm_arch_vcpu_setup(vcpu);
	if (r)
2477
		goto vcpu_destroy;
2478

2479 2480 2481 2482
	r = kvm_create_vcpu_debugfs(vcpu);
	if (r)
		goto vcpu_destroy;

S
Shaohua Li 已提交
2483
	mutex_lock(&kvm->lock);
2484 2485 2486 2487
	if (kvm_get_vcpu_by_id(kvm, id)) {
		r = -EEXIST;
		goto unlock_vcpu_destroy;
	}
2488 2489

	BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
2490

R
Rusty Russell 已提交
2491
	/* Now it's all set up, let userspace reach it */
A
Al Viro 已提交
2492
	kvm_get_kvm(kvm);
A
Avi Kivity 已提交
2493
	r = create_vcpu_fd(vcpu);
2494 2495
	if (r < 0) {
		kvm_put_kvm(kvm);
2496
		goto unlock_vcpu_destroy;
2497 2498 2499
	}

	kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu;
2500 2501 2502 2503 2504

	/*
	 * Pairs with smp_rmb() in kvm_get_vcpu.  Write kvm->vcpus
	 * before kvm->online_vcpu's incremented value.
	 */
2505 2506 2507 2508
	smp_wmb();
	atomic_inc(&kvm->online_vcpus);

	mutex_unlock(&kvm->lock);
2509
	kvm_arch_vcpu_postcreate(vcpu);
R
Rusty Russell 已提交
2510
	return r;
2511

2512
unlock_vcpu_destroy:
2513
	mutex_unlock(&kvm->lock);
2514
	debugfs_remove_recursive(vcpu->debugfs_dentry);
2515
vcpu_destroy:
2516
	kvm_arch_vcpu_destroy(vcpu);
2517 2518 2519 2520
vcpu_decrement:
	mutex_lock(&kvm->lock);
	kvm->created_vcpus--;
	mutex_unlock(&kvm->lock);
2521 2522 2523
	return r;
}

A
Avi Kivity 已提交
2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534
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
2535 2536
static long kvm_vcpu_ioctl(struct file *filp,
			   unsigned int ioctl, unsigned long arg)
A
Avi Kivity 已提交
2537
{
A
Avi Kivity 已提交
2538
	struct kvm_vcpu *vcpu = filp->private_data;
A
Al Viro 已提交
2539
	void __user *argp = (void __user *)arg;
2540
	int r;
2541 2542
	struct kvm_fpu *fpu = NULL;
	struct kvm_sregs *kvm_sregs = NULL;
A
Avi Kivity 已提交
2543

2544 2545
	if (vcpu->kvm->mm != current->mm)
		return -EIO;
2546

2547 2548 2549
	if (unlikely(_IOC_TYPE(ioctl) != KVMIO))
		return -EINVAL;

2550
#if defined(CONFIG_S390) || defined(CONFIG_PPC) || defined(CONFIG_MIPS)
2551 2552 2553 2554
	/*
	 * Special cases: vcpu ioctls that are asynchronous to vcpu execution,
	 * so vcpu_load() would break it.
	 */
2555
	if (ioctl == KVM_S390_INTERRUPT || ioctl == KVM_S390_IRQ || ioctl == KVM_INTERRUPT)
2556 2557 2558 2559
		return kvm_arch_vcpu_ioctl(filp, ioctl, arg);
#endif


2560 2561 2562
	r = vcpu_load(vcpu);
	if (r)
		return r;
A
Avi Kivity 已提交
2563
	switch (ioctl) {
2564
	case KVM_RUN:
2565 2566 2567
		r = -EINVAL;
		if (arg)
			goto out;
2568 2569 2570 2571
		if (unlikely(vcpu->pid != current->pids[PIDTYPE_PID].pid)) {
			/* The thread running this VCPU changed. */
			struct pid *oldpid = vcpu->pid;
			struct pid *newpid = get_task_pid(current, PIDTYPE_PID);
2572

2573 2574 2575 2576 2577
			rcu_assign_pointer(vcpu->pid, newpid);
			if (oldpid)
				synchronize_rcu();
			put_pid(oldpid);
		}
2578
		r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
2579
		trace_kvm_userspace_exit(vcpu->run->exit_reason, r);
A
Avi Kivity 已提交
2580 2581
		break;
	case KVM_GET_REGS: {
2582
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
2583

2584 2585 2586
		r = -ENOMEM;
		kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
		if (!kvm_regs)
A
Avi Kivity 已提交
2587
			goto out;
2588 2589 2590
		r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
		if (r)
			goto out_free1;
A
Avi Kivity 已提交
2591
		r = -EFAULT;
2592 2593
		if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
			goto out_free1;
A
Avi Kivity 已提交
2594
		r = 0;
2595 2596
out_free1:
		kfree(kvm_regs);
A
Avi Kivity 已提交
2597 2598 2599
		break;
	}
	case KVM_SET_REGS: {
2600
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
2601

2602
		r = -ENOMEM;
2603 2604 2605
		kvm_regs = memdup_user(argp, sizeof(*kvm_regs));
		if (IS_ERR(kvm_regs)) {
			r = PTR_ERR(kvm_regs);
A
Avi Kivity 已提交
2606
			goto out;
2607
		}
2608 2609
		r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
		kfree(kvm_regs);
A
Avi Kivity 已提交
2610 2611 2612
		break;
	}
	case KVM_GET_SREGS: {
2613 2614 2615 2616 2617
		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
2618 2619 2620
		if (r)
			goto out;
		r = -EFAULT;
2621
		if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
A
Avi Kivity 已提交
2622 2623 2624 2625 2626
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_SREGS: {
2627 2628 2629
		kvm_sregs = memdup_user(argp, sizeof(*kvm_sregs));
		if (IS_ERR(kvm_sregs)) {
			r = PTR_ERR(kvm_sregs);
G
Guo Chao 已提交
2630
			kvm_sregs = NULL;
A
Avi Kivity 已提交
2631
			goto out;
2632
		}
2633
		r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
2634 2635
		break;
	}
2636 2637 2638 2639 2640 2641 2642
	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;
2643
		if (copy_to_user(argp, &mp_state, sizeof(mp_state)))
2644 2645 2646 2647 2648 2649 2650 2651
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_MP_STATE: {
		struct kvm_mp_state mp_state;

		r = -EFAULT;
2652
		if (copy_from_user(&mp_state, argp, sizeof(mp_state)))
2653 2654 2655 2656
			goto out;
		r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
		break;
	}
A
Avi Kivity 已提交
2657 2658 2659 2660
	case KVM_TRANSLATE: {
		struct kvm_translation tr;

		r = -EFAULT;
2661
		if (copy_from_user(&tr, argp, sizeof(tr)))
A
Avi Kivity 已提交
2662
			goto out;
2663
		r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
A
Avi Kivity 已提交
2664 2665 2666
		if (r)
			goto out;
		r = -EFAULT;
2667
		if (copy_to_user(argp, &tr, sizeof(tr)))
A
Avi Kivity 已提交
2668 2669 2670 2671
			goto out;
		r = 0;
		break;
	}
J
Jan Kiszka 已提交
2672 2673
	case KVM_SET_GUEST_DEBUG: {
		struct kvm_guest_debug dbg;
A
Avi Kivity 已提交
2674 2675

		r = -EFAULT;
2676
		if (copy_from_user(&dbg, argp, sizeof(dbg)))
A
Avi Kivity 已提交
2677
			goto out;
J
Jan Kiszka 已提交
2678
		r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
A
Avi Kivity 已提交
2679 2680
		break;
	}
A
Avi Kivity 已提交
2681 2682 2683 2684 2685 2686 2687 2688 2689
	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,
2690
					   sizeof(kvm_sigmask)))
A
Avi Kivity 已提交
2691 2692
				goto out;
			r = -EINVAL;
2693
			if (kvm_sigmask.len != sizeof(sigset))
A
Avi Kivity 已提交
2694 2695 2696
				goto out;
			r = -EFAULT;
			if (copy_from_user(&sigset, sigmask_arg->sigset,
2697
					   sizeof(sigset)))
A
Avi Kivity 已提交
2698 2699 2700
				goto out;
			p = &sigset;
		}
2701
		r = kvm_vcpu_ioctl_set_sigmask(vcpu, p);
A
Avi Kivity 已提交
2702 2703
		break;
	}
A
Avi Kivity 已提交
2704
	case KVM_GET_FPU: {
2705 2706 2707 2708 2709
		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
2710 2711 2712
		if (r)
			goto out;
		r = -EFAULT;
2713
		if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
A
Avi Kivity 已提交
2714 2715 2716 2717 2718
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_FPU: {
2719 2720 2721
		fpu = memdup_user(argp, sizeof(*fpu));
		if (IS_ERR(fpu)) {
			r = PTR_ERR(fpu);
G
Guo Chao 已提交
2722
			fpu = NULL;
A
Avi Kivity 已提交
2723
			goto out;
2724
		}
2725
		r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
A
Avi Kivity 已提交
2726 2727
		break;
	}
A
Avi Kivity 已提交
2728
	default:
2729
		r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
2730 2731
	}
out:
2732
	vcpu_put(vcpu);
2733 2734
	kfree(fpu);
	kfree(kvm_sregs);
A
Avi Kivity 已提交
2735 2736 2737
	return r;
}

2738
#ifdef CONFIG_KVM_COMPAT
2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758
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,
2759
					   sizeof(kvm_sigmask)))
2760 2761
				goto out;
			r = -EINVAL;
2762
			if (kvm_sigmask.len != sizeof(csigset))
2763 2764 2765
				goto out;
			r = -EFAULT;
			if (copy_from_user(&csigset, sigmask_arg->sigset,
2766
					   sizeof(csigset)))
2767
				goto out;
2768 2769 2770 2771
			sigset_from_compat(&sigset, &csigset);
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
		} else
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, NULL);
2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782
		break;
	}
	default:
		r = kvm_vcpu_ioctl(filp, ioctl, arg);
	}

out:
	return r;
}
#endif

S
Scott Wood 已提交
2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829
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,
2830
#ifdef CONFIG_KVM_COMPAT
2831 2832
	.compat_ioctl = kvm_device_ioctl,
#endif
S
Scott Wood 已提交
2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843
	.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;
}

2844
static struct kvm_device_ops *kvm_device_ops_table[KVM_DEV_TYPE_MAX] = {
2845
#ifdef CONFIG_KVM_MPIC
2846 2847
	[KVM_DEV_TYPE_FSL_MPIC_20]	= &kvm_mpic_ops,
	[KVM_DEV_TYPE_FSL_MPIC_42]	= &kvm_mpic_ops,
2848
#endif
2849

2850
#ifdef CONFIG_KVM_XICS
2851
	[KVM_DEV_TYPE_XICS]		= &kvm_xics_ops,
A
Alex Williamson 已提交
2852
#endif
2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866
};

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;
}

2867 2868 2869 2870 2871 2872
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
2873 2874 2875 2876 2877 2878 2879 2880
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;

2881 2882 2883 2884 2885
	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
2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897
		return -ENODEV;

	if (test)
		return 0;

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

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

2898
	mutex_lock(&kvm->lock);
S
Scott Wood 已提交
2899 2900
	ret = ops->create(dev, cd->type);
	if (ret < 0) {
2901
		mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
2902 2903 2904
		kfree(dev);
		return ret;
	}
2905 2906
	list_add(&dev->vm_node, &kvm->devices);
	mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
2907

2908 2909 2910
	if (ops->init)
		ops->init(dev);

2911
	ret = anon_inode_getfd(ops->name, &kvm_device_fops, dev, O_RDWR | O_CLOEXEC);
S
Scott Wood 已提交
2912
	if (ret < 0) {
2913 2914 2915
		mutex_lock(&kvm->lock);
		list_del(&dev->vm_node);
		mutex_unlock(&kvm->lock);
2916
		ops->destroy(dev);
S
Scott Wood 已提交
2917 2918 2919 2920 2921 2922 2923 2924
		return ret;
	}

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

2925 2926 2927 2928 2929 2930 2931 2932 2933 2934
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
2935
#ifdef CONFIG_HAVE_KVM_IRQFD
2936
	case KVM_CAP_IRQFD:
2937 2938
	case KVM_CAP_IRQFD_RESAMPLE:
#endif
2939
	case KVM_CAP_IOEVENTFD_ANY_LENGTH:
2940 2941
	case KVM_CAP_CHECK_EXTENSION_VM:
		return 1;
2942
#ifdef CONFIG_KVM_MMIO
2943 2944 2945
	case KVM_CAP_COALESCED_MMIO:
		return KVM_COALESCED_MMIO_PAGE_OFFSET;
#endif
2946 2947 2948
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_CAP_IRQ_ROUTING:
		return KVM_MAX_IRQ_ROUTES;
2949 2950 2951 2952
#endif
#if KVM_ADDRESS_SPACE_NUM > 1
	case KVM_CAP_MULTI_ADDRESS_SPACE:
		return KVM_ADDRESS_SPACE_NUM;
2953
#endif
G
Greg Kurz 已提交
2954 2955
	case KVM_CAP_MAX_VCPU_ID:
		return KVM_MAX_VCPU_ID;
2956 2957 2958 2959 2960 2961
	default:
		break;
	}
	return kvm_vm_ioctl_check_extension(kvm, arg);
}

A
Avi Kivity 已提交
2962 2963 2964 2965 2966
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;
2967
	int r;
A
Avi Kivity 已提交
2968

2969 2970
	if (kvm->mm != current->mm)
		return -EIO;
A
Avi Kivity 已提交
2971 2972 2973 2974
	switch (ioctl) {
	case KVM_CREATE_VCPU:
		r = kvm_vm_ioctl_create_vcpu(kvm, arg);
		break;
2975 2976 2977 2978 2979
	case KVM_SET_USER_MEMORY_REGION: {
		struct kvm_userspace_memory_region kvm_userspace_mem;

		r = -EFAULT;
		if (copy_from_user(&kvm_userspace_mem, argp,
2980
						sizeof(kvm_userspace_mem)))
2981 2982
			goto out;

2983
		r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem);
A
Avi Kivity 已提交
2984 2985 2986 2987 2988 2989
		break;
	}
	case KVM_GET_DIRTY_LOG: {
		struct kvm_dirty_log log;

		r = -EFAULT;
2990
		if (copy_from_user(&log, argp, sizeof(log)))
A
Avi Kivity 已提交
2991
			goto out;
2992
		r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
A
Avi Kivity 已提交
2993 2994
		break;
	}
2995
#ifdef CONFIG_KVM_MMIO
2996 2997
	case KVM_REGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
2998

2999
		r = -EFAULT;
3000
		if (copy_from_user(&zone, argp, sizeof(zone)))
3001 3002 3003 3004 3005 3006
			goto out;
		r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
		break;
	}
	case KVM_UNREGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3007

3008
		r = -EFAULT;
3009
		if (copy_from_user(&zone, argp, sizeof(zone)))
3010 3011 3012 3013 3014
			goto out;
		r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
		break;
	}
#endif
G
Gregory Haskins 已提交
3015 3016 3017 3018
	case KVM_IRQFD: {
		struct kvm_irqfd data;

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

		r = -EFAULT;
3028
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3029 3030 3031 3032
			goto out;
		r = kvm_ioeventfd(kvm, &data);
		break;
	}
3033 3034 3035 3036 3037
#ifdef CONFIG_HAVE_KVM_MSI
	case KVM_SIGNAL_MSI: {
		struct kvm_msi msi;

		r = -EFAULT;
3038
		if (copy_from_user(&msi, argp, sizeof(msi)))
3039 3040 3041 3042
			goto out;
		r = kvm_send_userspace_msi(kvm, &msi);
		break;
	}
3043 3044 3045 3046 3047 3048 3049
#endif
#ifdef __KVM_HAVE_IRQ_LINE
	case KVM_IRQ_LINE_STATUS:
	case KVM_IRQ_LINE: {
		struct kvm_irq_level irq_event;

		r = -EFAULT;
3050
		if (copy_from_user(&irq_event, argp, sizeof(irq_event)))
3051 3052
			goto out;

3053 3054
		r = kvm_vm_ioctl_irq_line(kvm, &irq_event,
					ioctl == KVM_IRQ_LINE_STATUS);
3055 3056 3057 3058 3059
		if (r)
			goto out;

		r = -EFAULT;
		if (ioctl == KVM_IRQ_LINE_STATUS) {
3060
			if (copy_to_user(argp, &irq_event, sizeof(irq_event)))
3061 3062 3063 3064 3065 3066
				goto out;
		}

		r = 0;
		break;
	}
3067
#endif
3068 3069 3070 3071
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_SET_GSI_ROUTING: {
		struct kvm_irq_routing routing;
		struct kvm_irq_routing __user *urouting;
3072
		struct kvm_irq_routing_entry *entries = NULL;
3073 3074 3075 3076 3077

		r = -EFAULT;
		if (copy_from_user(&routing, argp, sizeof(routing)))
			goto out;
		r = -EINVAL;
3078 3079
		if (!kvm_arch_can_set_irq_routing(kvm))
			goto out;
3080
		if (routing.nr > KVM_MAX_IRQ_ROUTES)
3081 3082 3083
			goto out;
		if (routing.flags)
			goto out;
3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094
		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;
		}
3095 3096
		r = kvm_set_irq_routing(kvm, entries, routing.nr,
					routing.flags);
3097
out_free_irq_routing:
3098 3099 3100 3101
		vfree(entries);
		break;
	}
#endif /* CONFIG_HAVE_KVM_IRQ_ROUTING */
S
Scott Wood 已提交
3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119
	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;
	}
3120 3121 3122
	case KVM_CHECK_EXTENSION:
		r = kvm_vm_ioctl_check_extension_generic(kvm, arg);
		break;
3123
	default:
3124
		r = kvm_arch_vm_ioctl(filp, ioctl, arg);
3125 3126 3127 3128 3129
	}
out:
	return r;
}

3130
#ifdef CONFIG_KVM_COMPAT
3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154
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)))
3155
			return -EFAULT;
3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170
		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

3171
static struct file_operations kvm_vm_fops = {
3172 3173
	.release        = kvm_vm_release,
	.unlocked_ioctl = kvm_vm_ioctl,
3174
#ifdef CONFIG_KVM_COMPAT
3175 3176
	.compat_ioctl   = kvm_vm_compat_ioctl,
#endif
3177
	.llseek		= noop_llseek,
3178 3179
};

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

3186
	kvm = kvm_create_vm(type);
3187 3188
	if (IS_ERR(kvm))
		return PTR_ERR(kvm);
3189
#ifdef CONFIG_KVM_MMIO
3190 3191 3192 3193 3194 3195
	r = kvm_coalesced_mmio_init(kvm);
	if (r < 0) {
		kvm_put_kvm(kvm);
		return r;
	}
#endif
3196
	r = get_unused_fd_flags(O_CLOEXEC);
3197
	if (r < 0) {
A
Al Viro 已提交
3198
		kvm_put_kvm(kvm);
3199 3200
		return r;
	}
3201 3202 3203 3204 3205 3206
	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);
	}
3207 3208

	if (kvm_create_vm_debugfs(kvm, r) < 0) {
3209 3210
		put_unused_fd(r);
		fput(file);
3211 3212
		return -ENOMEM;
	}
3213

3214
	fd_install(r, file);
3215
	return r;
3216 3217 3218 3219 3220
}

static long kvm_dev_ioctl(struct file *filp,
			  unsigned int ioctl, unsigned long arg)
{
3221
	long r = -EINVAL;
3222 3223 3224

	switch (ioctl) {
	case KVM_GET_API_VERSION:
3225 3226
		if (arg)
			goto out;
3227 3228 3229
		r = KVM_API_VERSION;
		break;
	case KVM_CREATE_VM:
3230
		r = kvm_dev_ioctl_create_vm(arg);
3231
		break;
3232
	case KVM_CHECK_EXTENSION:
3233
		r = kvm_vm_ioctl_check_extension_generic(NULL, arg);
3234
		break;
3235 3236 3237
	case KVM_GET_VCPU_MMAP_SIZE:
		if (arg)
			goto out;
3238 3239 3240
		r = PAGE_SIZE;     /* struct kvm_run */
#ifdef CONFIG_X86
		r += PAGE_SIZE;    /* pio data page */
3241
#endif
3242
#ifdef CONFIG_KVM_MMIO
3243
		r += PAGE_SIZE;    /* coalesced mmio ring page */
3244
#endif
3245
		break;
3246 3247 3248
	case KVM_TRACE_ENABLE:
	case KVM_TRACE_PAUSE:
	case KVM_TRACE_DISABLE:
3249
		r = -EOPNOTSUPP;
3250
		break;
A
Avi Kivity 已提交
3251
	default:
3252
		return kvm_arch_dev_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
3253 3254 3255 3256 3257 3258 3259 3260
	}
out:
	return r;
}

static struct file_operations kvm_chardev_ops = {
	.unlocked_ioctl = kvm_dev_ioctl,
	.compat_ioctl   = kvm_dev_ioctl,
3261
	.llseek		= noop_llseek,
A
Avi Kivity 已提交
3262 3263 3264
};

static struct miscdevice kvm_dev = {
A
Avi Kivity 已提交
3265
	KVM_MINOR,
A
Avi Kivity 已提交
3266 3267 3268 3269
	"kvm",
	&kvm_chardev_ops,
};

3270
static void hardware_enable_nolock(void *junk)
3271 3272
{
	int cpu = raw_smp_processor_id();
3273
	int r;
3274

3275
	if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
3276
		return;
3277

3278
	cpumask_set_cpu(cpu, cpus_hardware_enabled);
3279

3280
	r = kvm_arch_hardware_enable();
3281 3282 3283 3284

	if (r) {
		cpumask_clear_cpu(cpu, cpus_hardware_enabled);
		atomic_inc(&hardware_enable_failed);
X
Xiubo Li 已提交
3285
		pr_info("kvm: enabling virtualization on CPU%d failed\n", cpu);
3286
	}
3287 3288
}

3289
static int kvm_starting_cpu(unsigned int cpu)
3290
{
3291
	raw_spin_lock(&kvm_count_lock);
3292 3293
	if (kvm_usage_count)
		hardware_enable_nolock(NULL);
3294
	raw_spin_unlock(&kvm_count_lock);
3295
	return 0;
3296 3297 3298
}

static void hardware_disable_nolock(void *junk)
3299 3300 3301
{
	int cpu = raw_smp_processor_id();

3302
	if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
3303
		return;
3304
	cpumask_clear_cpu(cpu, cpus_hardware_enabled);
3305
	kvm_arch_hardware_disable();
3306 3307
}

3308
static int kvm_dying_cpu(unsigned int cpu)
3309
{
3310
	raw_spin_lock(&kvm_count_lock);
3311 3312
	if (kvm_usage_count)
		hardware_disable_nolock(NULL);
3313
	raw_spin_unlock(&kvm_count_lock);
3314
	return 0;
3315 3316
}

3317 3318 3319 3320 3321 3322
static void hardware_disable_all_nolock(void)
{
	BUG_ON(!kvm_usage_count);

	kvm_usage_count--;
	if (!kvm_usage_count)
3323
		on_each_cpu(hardware_disable_nolock, NULL, 1);
3324 3325 3326 3327
}

static void hardware_disable_all(void)
{
3328
	raw_spin_lock(&kvm_count_lock);
3329
	hardware_disable_all_nolock();
3330
	raw_spin_unlock(&kvm_count_lock);
3331 3332 3333 3334 3335 3336
}

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

3337
	raw_spin_lock(&kvm_count_lock);
3338 3339 3340 3341

	kvm_usage_count++;
	if (kvm_usage_count == 1) {
		atomic_set(&hardware_enable_failed, 0);
3342
		on_each_cpu(hardware_enable_nolock, NULL, 1);
3343 3344 3345 3346 3347 3348 3349

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

3350
	raw_spin_unlock(&kvm_count_lock);
3351 3352 3353 3354

	return r;
}

3355
static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
M
Mike Day 已提交
3356
		      void *v)
3357
{
3358 3359 3360 3361 3362 3363
	/*
	 * 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
3364
	pr_info("kvm: exiting hardware virtualization\n");
3365
	kvm_rebooting = true;
3366
	on_each_cpu(hardware_disable_nolock, NULL, 1);
3367 3368 3369 3370 3371 3372 3373 3374
	return NOTIFY_OK;
}

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

M
Marcelo Tosatti 已提交
3375
static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
3376 3377 3378 3379
{
	int i;

	for (i = 0; i < bus->dev_count; i++) {
3380
		struct kvm_io_device *pos = bus->range[i].dev;
3381 3382 3383

		kvm_iodevice_destructor(pos);
	}
M
Marcelo Tosatti 已提交
3384
	kfree(bus);
3385 3386
}

3387
static inline int kvm_io_bus_cmp(const struct kvm_io_range *r1,
X
Xiubo Li 已提交
3388
				 const struct kvm_io_range *r2)
3389
{
J
Jason Wang 已提交
3390 3391 3392 3393
	gpa_t addr1 = r1->addr;
	gpa_t addr2 = r2->addr;

	if (addr1 < addr2)
3394
		return -1;
J
Jason Wang 已提交
3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406

	/* 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)
3407
		return 1;
J
Jason Wang 已提交
3408

3409 3410 3411
	return 0;
}

3412 3413
static int kvm_io_bus_sort_cmp(const void *p1, const void *p2)
{
3414
	return kvm_io_bus_cmp(p1, p2);
3415 3416
}

G
Geoff Levand 已提交
3417
static int kvm_io_bus_insert_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev,
3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431
			  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
3432
static int kvm_io_bus_get_first_dev(struct kvm_io_bus *bus,
3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449
			     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;

3450
	while (off > 0 && kvm_io_bus_cmp(&key, &bus->range[off-1]) == 0)
3451 3452 3453 3454 3455
		off--;

	return off;
}

3456
static int __kvm_io_bus_write(struct kvm_vcpu *vcpu, struct kvm_io_bus *bus,
C
Cornelia Huck 已提交
3457 3458 3459 3460 3461 3462 3463 3464 3465
			      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 &&
3466
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
3467
		if (!kvm_iodevice_write(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
3468 3469 3470 3471 3472 3473 3474 3475
					range->len, val))
			return idx;
		idx++;
	}

	return -EOPNOTSUPP;
}

3476
/* kvm_io_bus_write - called under kvm->slots_lock */
3477
int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
3478
		     int len, const void *val)
3479
{
3480
	struct kvm_io_bus *bus;
3481
	struct kvm_io_range range;
C
Cornelia Huck 已提交
3482
	int r;
3483 3484 3485 3486 3487

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
3488

3489
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3490 3491
	if (!bus)
		return -ENOMEM;
3492
	r = __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3493 3494 3495 3496
	return r < 0 ? r : 0;
}

/* kvm_io_bus_write_cookie - called under kvm->slots_lock */
3497 3498
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
3499 3500 3501 3502 3503 3504 3505 3506 3507
{
	struct kvm_io_bus *bus;
	struct kvm_io_range range;

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

3508
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3509 3510
	if (!bus)
		return -ENOMEM;
C
Cornelia Huck 已提交
3511 3512 3513

	/* First try the device referenced by cookie. */
	if ((cookie >= 0) && (cookie < bus->dev_count) &&
3514
	    (kvm_io_bus_cmp(&range, &bus->range[cookie]) == 0))
3515
		if (!kvm_iodevice_write(vcpu, bus->range[cookie].dev, addr, len,
C
Cornelia Huck 已提交
3516 3517 3518 3519 3520 3521 3522
					val))
			return cookie;

	/*
	 * cookie contained garbage; fall back to search and return the
	 * correct cookie value.
	 */
3523
	return __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3524 3525
}

3526 3527
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
3528 3529 3530 3531
{
	int idx;

	idx = kvm_io_bus_get_first_dev(bus, range->addr, range->len);
3532 3533 3534 3535
	if (idx < 0)
		return -EOPNOTSUPP;

	while (idx < bus->dev_count &&
3536
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
3537
		if (!kvm_iodevice_read(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
3538 3539
				       range->len, val))
			return idx;
3540 3541 3542
		idx++;
	}

3543 3544
	return -EOPNOTSUPP;
}
3545
EXPORT_SYMBOL_GPL(kvm_io_bus_write);
3546

3547
/* kvm_io_bus_read - called under kvm->slots_lock */
3548
int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
M
Marcelo Tosatti 已提交
3549
		    int len, void *val)
3550
{
3551
	struct kvm_io_bus *bus;
3552
	struct kvm_io_range range;
C
Cornelia Huck 已提交
3553
	int r;
3554 3555 3556 3557 3558

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
M
Marcelo Tosatti 已提交
3559

3560
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3561 3562
	if (!bus)
		return -ENOMEM;
3563
	r = __kvm_io_bus_read(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3564 3565
	return r < 0 ? r : 0;
}
3566

3567

3568
/* Caller must hold slots_lock. */
3569 3570
int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
			    int len, struct kvm_io_device *dev)
3571
{
M
Marcelo Tosatti 已提交
3572
	struct kvm_io_bus *new_bus, *bus;
3573

M
Marcelo Tosatti 已提交
3574
	bus = kvm->buses[bus_idx];
3575 3576 3577
	if (!bus)
		return -ENOMEM;

3578 3579
	/* exclude ioeventfd which is limited by maximum fd */
	if (bus->dev_count - bus->ioeventfd_count > NR_IOBUS_DEVS - 1)
3580
		return -ENOSPC;
3581

3582
	new_bus = kmalloc(sizeof(*bus) + ((bus->dev_count + 1) *
3583
			  sizeof(struct kvm_io_range)), GFP_KERNEL);
M
Marcelo Tosatti 已提交
3584 3585
	if (!new_bus)
		return -ENOMEM;
3586 3587
	memcpy(new_bus, bus, sizeof(*bus) + (bus->dev_count *
	       sizeof(struct kvm_io_range)));
3588
	kvm_io_bus_insert_dev(new_bus, dev, addr, len);
M
Marcelo Tosatti 已提交
3589 3590 3591
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
3592 3593 3594 3595

	return 0;
}

3596
/* Caller must hold slots_lock. */
3597 3598
void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
			       struct kvm_io_device *dev)
3599
{
3600
	int i;
M
Marcelo Tosatti 已提交
3601
	struct kvm_io_bus *new_bus, *bus;
3602

3603
	bus = kvm->buses[bus_idx];
3604
	if (!bus)
3605
		return;
3606

3607 3608
	for (i = 0; i < bus->dev_count; i++)
		if (bus->range[i].dev == dev) {
3609 3610
			break;
		}
M
Marcelo Tosatti 已提交
3611

3612 3613
	if (i == bus->dev_count)
		return;
3614

3615
	new_bus = kmalloc(sizeof(*bus) + ((bus->dev_count - 1) *
3616
			  sizeof(struct kvm_io_range)), GFP_KERNEL);
3617 3618 3619 3620
	if (!new_bus)  {
		pr_err("kvm: failed to shrink bus, removing it completely\n");
		goto broken;
	}
3621 3622 3623 3624 3625

	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
3626

3627
broken:
M
Marcelo Tosatti 已提交
3628 3629 3630
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
3631
	return;
3632 3633
}

3634 3635 3636 3637 3638 3639 3640 3641 3642 3643
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);
3644 3645
	if (!bus)
		goto out_unlock;
3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659

	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);

3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671
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.
	 */
3672
	if (!refcount_inc_not_zero(&stat_data->kvm->users_count))
3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697
		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;

3698
	*val = *(ulong *)((void *)stat_data->kvm + stat_data->offset);
3699 3700 3701 3702

	return 0;
}

3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714
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;
}

3715 3716 3717 3718
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,
3719
				vm_stat_clear_per_vm, "%llu\n");
3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739
}

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,
	.llseek  = generic_file_llseek,
};

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)
3740
		*val += *(u64 *)((void *)vcpu + stat_data->offset);
3741 3742 3743 3744

	return 0;
}

3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759
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;
}

3760 3761 3762 3763
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,
3764
				 vcpu_stat_clear_per_vm, "%llu\n");
3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780
}

static const struct file_operations vcpu_stat_get_per_vm_fops = {
	.owner   = THIS_MODULE,
	.open    = vcpu_stat_get_per_vm_open,
	.release = kvm_debugfs_release,
	.read    = simple_attr_read,
	.write   = simple_attr_write,
	.llseek  = generic_file_llseek,
};

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,
};

3781
static int vm_stat_get(void *_offset, u64 *val)
3782 3783 3784
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
3785 3786
	struct kvm_stat_data stat_tmp = {.offset = offset};
	u64 tmp_val;
3787

3788
	*val = 0;
3789
	spin_lock(&kvm_lock);
3790 3791 3792 3793 3794
	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;
	}
3795
	spin_unlock(&kvm_lock);
3796
	return 0;
3797 3798
}

3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818
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");
3819

3820
static int vcpu_stat_get(void *_offset, u64 *val)
A
Avi Kivity 已提交
3821 3822 3823
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
3824 3825
	struct kvm_stat_data stat_tmp = {.offset = offset};
	u64 tmp_val;
A
Avi Kivity 已提交
3826

3827
	*val = 0;
3828
	spin_lock(&kvm_lock);
3829 3830 3831 3832 3833
	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;
	}
3834
	spin_unlock(&kvm_lock);
3835
	return 0;
A
Avi Kivity 已提交
3836 3837
}

3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858
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");
3859

3860
static const struct file_operations *stat_fops[] = {
3861 3862 3863
	[KVM_STAT_VCPU] = &vcpu_stat_fops,
	[KVM_STAT_VM]   = &vm_stat_fops,
};
A
Avi Kivity 已提交
3864

3865
static int kvm_init_debug(void)
A
Avi Kivity 已提交
3866
{
3867
	int r = -EEXIST;
A
Avi Kivity 已提交
3868 3869
	struct kvm_stats_debugfs_item *p;

3870
	kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
3871 3872 3873
	if (kvm_debugfs_dir == NULL)
		goto out;

3874 3875
	kvm_debugfs_num_entries = 0;
	for (p = debugfs_entries; p->name; ++p, kvm_debugfs_num_entries++) {
3876
		if (!debugfs_create_file(p->name, 0644, kvm_debugfs_dir,
3877 3878
					 (void *)(long)p->offset,
					 stat_fops[p->kind]))
3879 3880 3881 3882 3883 3884 3885 3886 3887
			goto out_dir;
	}

	return 0;

out_dir:
	debugfs_remove_recursive(kvm_debugfs_dir);
out:
	return r;
A
Avi Kivity 已提交
3888 3889
}

3890
static int kvm_suspend(void)
3891
{
3892
	if (kvm_usage_count)
3893
		hardware_disable_nolock(NULL);
3894 3895 3896
	return 0;
}

3897
static void kvm_resume(void)
3898
{
3899
	if (kvm_usage_count) {
3900
		WARN_ON(raw_spin_is_locked(&kvm_count_lock));
3901
		hardware_enable_nolock(NULL);
3902
	}
3903 3904
}

3905
static struct syscore_ops kvm_syscore_ops = {
3906 3907 3908 3909
	.suspend = kvm_suspend,
	.resume = kvm_resume,
};

3910 3911 3912 3913 3914 3915 3916 3917 3918
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);
3919

3920 3921
	if (vcpu->preempted)
		vcpu->preempted = false;
3922

R
Radim Krčmář 已提交
3923 3924
	kvm_arch_sched_in(vcpu, cpu);

3925
	kvm_arch_vcpu_load(vcpu, cpu);
3926 3927 3928 3929 3930 3931 3932
}

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

3933 3934
	if (current->state == TASK_RUNNING)
		vcpu->preempted = true;
3935
	kvm_arch_vcpu_put(vcpu);
3936 3937
}

3938
int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
3939
		  struct module *module)
A
Avi Kivity 已提交
3940 3941
{
	int r;
Y
Yang, Sheng 已提交
3942
	int cpu;
A
Avi Kivity 已提交
3943

3944 3945
	r = kvm_arch_init(opaque);
	if (r)
3946
		goto out_fail;
3947

3948 3949 3950 3951
	/*
	 * 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
3952 3953
	 * kvm_arch_init must be called before kvm_irqfd_init to avoid creating
	 * conflicts in case kvm is already setup for another implementation.
3954
	 */
P
Paolo Bonzini 已提交
3955 3956 3957
	r = kvm_irqfd_init();
	if (r)
		goto out_irqfd;
3958

3959
	if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
3960 3961 3962 3963
		r = -ENOMEM;
		goto out_free_0;
	}

3964
	r = kvm_arch_hardware_setup();
A
Avi Kivity 已提交
3965
	if (r < 0)
3966
		goto out_free_0a;
A
Avi Kivity 已提交
3967

Y
Yang, Sheng 已提交
3968 3969
	for_each_online_cpu(cpu) {
		smp_call_function_single(cpu,
3970
				kvm_arch_check_processor_compat,
3971
				&r, 1);
Y
Yang, Sheng 已提交
3972
		if (r < 0)
3973
			goto out_free_1;
Y
Yang, Sheng 已提交
3974 3975
	}

T
Thomas Gleixner 已提交
3976
	r = cpuhp_setup_state_nocalls(CPUHP_AP_KVM_STARTING, "kvm/cpu:starting",
3977
				      kvm_starting_cpu, kvm_dying_cpu);
A
Avi Kivity 已提交
3978
	if (r)
3979
		goto out_free_2;
A
Avi Kivity 已提交
3980 3981
	register_reboot_notifier(&kvm_reboot_notifier);

3982
	/* A kmem cache lets us meet the alignment requirements of fx_save. */
3983 3984 3985
	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
3986
					   0, NULL);
3987 3988
	if (!kvm_vcpu_cache) {
		r = -ENOMEM;
3989
		goto out_free_3;
3990 3991
	}

3992 3993 3994 3995
	r = kvm_async_pf_init();
	if (r)
		goto out_free;

A
Avi Kivity 已提交
3996
	kvm_chardev_ops.owner = module;
3997 3998
	kvm_vm_fops.owner = module;
	kvm_vcpu_fops.owner = module;
A
Avi Kivity 已提交
3999 4000 4001

	r = misc_register(&kvm_dev);
	if (r) {
X
Xiubo Li 已提交
4002
		pr_err("kvm: misc device register failed\n");
4003
		goto out_unreg;
A
Avi Kivity 已提交
4004 4005
	}

4006 4007
	register_syscore_ops(&kvm_syscore_ops);

4008 4009 4010
	kvm_preempt_ops.sched_in = kvm_sched_in;
	kvm_preempt_ops.sched_out = kvm_sched_out;

4011 4012
	r = kvm_init_debug();
	if (r) {
X
Xiubo Li 已提交
4013
		pr_err("kvm: create debugfs files failed\n");
4014 4015
		goto out_undebugfs;
	}
4016

P
Paolo Bonzini 已提交
4017 4018 4019
	r = kvm_vfio_ops_init();
	WARN_ON(r);

4020
	return 0;
A
Avi Kivity 已提交
4021

4022 4023
out_undebugfs:
	unregister_syscore_ops(&kvm_syscore_ops);
4024
	misc_deregister(&kvm_dev);
4025 4026
out_unreg:
	kvm_async_pf_deinit();
A
Avi Kivity 已提交
4027
out_free:
4028
	kmem_cache_destroy(kvm_vcpu_cache);
4029
out_free_3:
A
Avi Kivity 已提交
4030
	unregister_reboot_notifier(&kvm_reboot_notifier);
4031
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4032 4033
out_free_2:
out_free_1:
4034
	kvm_arch_hardware_unsetup();
4035 4036
out_free_0a:
	free_cpumask_var(cpus_hardware_enabled);
4037
out_free_0:
4038
	kvm_irqfd_exit();
P
Paolo Bonzini 已提交
4039
out_irqfd:
4040 4041
	kvm_arch_exit();
out_fail:
A
Avi Kivity 已提交
4042 4043
	return r;
}
4044
EXPORT_SYMBOL_GPL(kvm_init);
A
Avi Kivity 已提交
4045

4046
void kvm_exit(void)
A
Avi Kivity 已提交
4047
{
4048
	debugfs_remove_recursive(kvm_debugfs_dir);
A
Avi Kivity 已提交
4049
	misc_deregister(&kvm_dev);
4050
	kmem_cache_destroy(kvm_vcpu_cache);
4051
	kvm_async_pf_deinit();
4052
	unregister_syscore_ops(&kvm_syscore_ops);
A
Avi Kivity 已提交
4053
	unregister_reboot_notifier(&kvm_reboot_notifier);
4054
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4055
	on_each_cpu(hardware_disable_nolock, NULL, 1);
4056
	kvm_arch_hardware_unsetup();
4057
	kvm_arch_exit();
4058
	kvm_irqfd_exit();
4059
	free_cpumask_var(cpus_hardware_enabled);
4060
	kvm_vfio_ops_exit();
A
Avi Kivity 已提交
4061
}
4062
EXPORT_SYMBOL_GPL(kvm_exit);