kvm_main.c 99.2 KB
Newer Older
A
Avi Kivity 已提交
1 2 3 4 5 6 7
/*
 * Kernel-based Virtual Machine driver for Linux
 *
 * This module enables machines with Intel VT-x extensions to run virtual
 * machines without emulation or binary translation.
 *
 * Copyright (C) 2006 Qumranet, Inc.
N
Nicolas Kaiser 已提交
8
 * Copyright 2010 Red Hat, Inc. and/or its affiliates.
A
Avi Kivity 已提交
9 10 11 12 13 14 15 16 17 18
 *
 * Authors:
 *   Avi Kivity   <avi@qumranet.com>
 *   Yaniv Kamay  <yaniv@qumranet.com>
 *
 * This work is licensed under the terms of the GNU GPL, version 2.  See
 * the COPYING file in the top-level directory.
 *
 */

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

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

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

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

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

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

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

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

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

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

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

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

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

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

106 107
static __read_mostly struct preempt_ops kvm_preempt_ops;

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

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

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

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

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

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

135 136
static bool largepages_enabled = true;

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

143 144
__weak int kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
		unsigned long start, unsigned long end, bool blockable)
145
{
146
	return 0;
147 148
}

D
Dan Williams 已提交
149
bool kvm_is_reserved_pfn(kvm_pfn_t pfn)
B
Ben-Ami Yassour 已提交
150
{
151
	if (pfn_valid(pfn))
152
		return PageReserved(pfn_to_page(pfn));
B
Ben-Ami Yassour 已提交
153 154 155 156

	return true;
}

A
Avi Kivity 已提交
157 158 159
/*
 * Switches to specified vcpu, until a matching vcpu_put()
 */
160
void vcpu_load(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
161
{
162
	int cpu = get_cpu();
163
	preempt_notifier_register(&vcpu->preempt_notifier);
164
	kvm_arch_vcpu_load(vcpu, cpu);
165
	put_cpu();
A
Avi Kivity 已提交
166
}
167
EXPORT_SYMBOL_GPL(vcpu_load);
A
Avi Kivity 已提交
168

169
void vcpu_put(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
170
{
171
	preempt_disable();
172
	kvm_arch_vcpu_put(vcpu);
173 174
	preempt_notifier_unregister(&vcpu->preempt_notifier);
	preempt_enable();
A
Avi Kivity 已提交
175
}
176
EXPORT_SYMBOL_GPL(vcpu_put);
A
Avi Kivity 已提交
177

178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195
/* 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;
}

196 197 198 199
static void ack_flush(void *_completed)
{
}

200 201 202 203 204 205 206 207 208 209 210 211
static inline bool kvm_kick_many_cpus(const struct cpumask *cpus, bool wait)
{
	if (unlikely(!cpus))
		cpus = cpu_online_mask;

	if (cpumask_empty(cpus))
		return false;

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

212 213
bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
				 unsigned long *vcpu_bitmap, cpumask_var_t tmp)
214
{
215
	int i, cpu, me;
216
	struct kvm_vcpu *vcpu;
217
	bool called;
218

219
	me = get_cpu();
220

221
	kvm_for_each_vcpu(i, vcpu, kvm) {
222
		if (vcpu_bitmap && !test_bit(i, vcpu_bitmap))
223 224
			continue;

225
		kvm_make_request(req, vcpu);
226
		cpu = vcpu->cpu;
227

228 229
		if (!(req & KVM_REQUEST_NO_WAKEUP) && kvm_vcpu_wake_up(vcpu))
			continue;
230

231
		if (tmp != NULL && cpu != -1 && cpu != me &&
232
		    kvm_request_needs_ipi(vcpu, req))
233
			__cpumask_set_cpu(cpu, tmp);
234
	}
235 236

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

	return called;
}

bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req)
{
	cpumask_var_t cpus;
	bool called;

	zalloc_cpumask_var(&cpus, GFP_ATOMIC);

249
	called = kvm_make_vcpus_request_mask(kvm, req, NULL, cpus);
250

251
	free_cpumask_var(cpus);
252
	return called;
253 254
}

255
#ifndef CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALL
256
void kvm_flush_remote_tlbs(struct kvm *kvm)
257
{
258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274
	/*
	 * 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.
	 */
275 276
	if (!kvm_arch_flush_remote_tlb(kvm)
	    || kvm_make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH))
277
		++kvm->stat.remote_tlb_flush;
278
	cmpxchg(&kvm->tlbs_dirty, dirty_count, 0);
279
}
280
EXPORT_SYMBOL_GPL(kvm_flush_remote_tlbs);
281
#endif
282

283 284
void kvm_reload_remote_mmus(struct kvm *kvm)
{
285
	kvm_make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD);
286
}
287

R
Rusty Russell 已提交
288 289 290 291 292 293 294 295 296
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;
297
	vcpu->pid = NULL;
298
	init_swait_queue_head(&vcpu->wq);
299
	kvm_async_pf_vcpu_init(vcpu);
R
Rusty Russell 已提交
300

301 302 303
	vcpu->pre_pcpu = -1;
	INIT_LIST_HEAD(&vcpu->blocked_vcpu_list);

R
Rusty Russell 已提交
304 305 306 307 308 309 310
	page = alloc_page(GFP_KERNEL | __GFP_ZERO);
	if (!page) {
		r = -ENOMEM;
		goto fail;
	}
	vcpu->run = page_address(page);

311 312
	kvm_vcpu_set_in_spin_loop(vcpu, false);
	kvm_vcpu_set_dy_eligible(vcpu, false);
313
	vcpu->preempted = false;
314

315
	r = kvm_arch_vcpu_init(vcpu);
R
Rusty Russell 已提交
316
	if (r < 0)
317
		goto fail_free_run;
R
Rusty Russell 已提交
318 319 320 321 322
	return 0;

fail_free_run:
	free_page((unsigned long)vcpu->run);
fail:
323
	return r;
R
Rusty Russell 已提交
324 325 326 327 328
}
EXPORT_SYMBOL_GPL(kvm_vcpu_init);

void kvm_vcpu_uninit(struct kvm_vcpu *vcpu)
{
329 330 331 332 333 334
	/*
	 * no need for rcu_read_lock as VCPU_RUN is the only place that
	 * will change the vcpu->pid pointer and on uninit all file
	 * descriptors are already gone.
	 */
	put_pid(rcu_dereference_protected(vcpu->pid, 1));
335
	kvm_arch_vcpu_uninit(vcpu);
R
Rusty Russell 已提交
336 337 338 339
	free_page((unsigned long)vcpu->run);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_uninit);

340 341 342 343 344 345
#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);
}

346 347 348 349 350 351
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);
352
	int idx;
353

354
	idx = srcu_read_lock(&kvm->srcu);
355 356
	spin_lock(&kvm->mmu_lock);
	kvm->mmu_notifier_seq++;
357 358 359 360

	if (kvm_set_spte_hva(kvm, address, pte))
		kvm_flush_remote_tlbs(kvm);

361
	spin_unlock(&kvm->mmu_lock);
362
	srcu_read_unlock(&kvm->srcu, idx);
363 364
}

365
static int kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn,
366
					const struct mmu_notifier_range *range)
367 368
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
369
	int need_tlb_flush = 0, idx;
370
	int ret;
371

372
	idx = srcu_read_lock(&kvm->srcu);
373 374 375 376 377 378 379
	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++;
380
	need_tlb_flush = kvm_unmap_hva_range(kvm, range->start, range->end);
381
	need_tlb_flush |= kvm->tlbs_dirty;
382 383 384
	/* we've to flush the tlb before the pages can be freed */
	if (need_tlb_flush)
		kvm_flush_remote_tlbs(kvm);
385 386

	spin_unlock(&kvm->mmu_lock);
387

388 389
	ret = kvm_arch_mmu_notifier_invalidate_range(kvm, range->start,
					range->end, range->blockable);
390

391
	srcu_read_unlock(&kvm->srcu, idx);
392 393

	return ret;
394 395 396
}

static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
397
					const struct mmu_notifier_range *range)
398 399 400 401 402 403 404 405 406 407
{
	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++;
408
	smp_wmb();
409 410
	/*
	 * The above sequence increase must be visible before the
411 412
	 * below count decrease, which is ensured by the smp_wmb above
	 * in conjunction with the smp_rmb in mmu_notifier_retry().
413 414 415 416 417 418 419 420 421
	 */
	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
422 423
					      unsigned long start,
					      unsigned long end)
424 425
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
426
	int young, idx;
427

428
	idx = srcu_read_lock(&kvm->srcu);
429 430
	spin_lock(&kvm->mmu_lock);

A
Andres Lagar-Cavilla 已提交
431
	young = kvm_age_hva(kvm, start, end);
432 433 434
	if (young)
		kvm_flush_remote_tlbs(kvm);

435 436 437
	spin_unlock(&kvm->mmu_lock);
	srcu_read_unlock(&kvm->srcu, idx);

438 439 440
	return young;
}

441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470
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
471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486
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;
}

487 488 489 490
static void kvm_mmu_notifier_release(struct mmu_notifier *mn,
				     struct mm_struct *mm)
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
491 492 493
	int idx;

	idx = srcu_read_lock(&kvm->srcu);
494
	kvm_arch_flush_shadow_all(kvm);
495
	srcu_read_unlock(&kvm->srcu, idx);
496 497
}

498 499 500 501
static const struct mmu_notifier_ops kvm_mmu_notifier_ops = {
	.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,
502
	.clear_young		= kvm_mmu_notifier_clear_young,
A
Andrea Arcangeli 已提交
503
	.test_young		= kvm_mmu_notifier_test_young,
504
	.change_pte		= kvm_mmu_notifier_change_pte,
505
	.release		= kvm_mmu_notifier_release,
506
};
507 508 509 510 511 512 513 514 515 516 517 518 519 520

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;
}

521 522
#endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */

523
static struct kvm_memslots *kvm_alloc_memslots(void)
524 525
{
	int i;
526
	struct kvm_memslots *slots;
527

M
Michal Hocko 已提交
528
	slots = kvzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
529 530 531
	if (!slots)
		return NULL;

532
	for (i = 0; i < KVM_MEM_SLOTS_NUM; i++)
533
		slots->id_to_index[i] = slots->memslots[i].id = i;
534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571

	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);
572 573
}

574 575 576 577 578 579 580 581 582
static void kvm_destroy_vm_debugfs(struct kvm *kvm)
{
	int i;

	if (!kvm->debugfs_dentry)
		return;

	debugfs_remove_recursive(kvm->debugfs_dentry);

583 584 585 586 587
	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);
	}
588 589 590 591 592 593 594 595 596 597 598 599
}

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);
600
	kvm->debugfs_dentry = debugfs_create_dir(dir_name, kvm_debugfs_dir);
601 602 603 604 605 606 607 608 609 610 611 612 613 614 615

	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;
616 617
		debugfs_create_file(p->name, 0644, kvm->debugfs_dentry,
				    stat_data, stat_fops_per_vm[p->kind]);
618 619 620 621
	}
	return 0;
}

622
static struct kvm *kvm_create_vm(unsigned long type)
A
Avi Kivity 已提交
623
{
624 625
	int r, i;
	struct kvm *kvm = kvm_arch_alloc_vm();
A
Avi Kivity 已提交
626

627 628 629
	if (!kvm)
		return ERR_PTR(-ENOMEM);

630
	spin_lock_init(&kvm->mmu_lock);
V
Vegard Nossum 已提交
631
	mmgrab(current->mm);
632 633 634 635 636
	kvm->mm = current->mm;
	kvm_eventfd_init(kvm);
	mutex_init(&kvm->lock);
	mutex_init(&kvm->irq_lock);
	mutex_init(&kvm->slots_lock);
637
	refcount_set(&kvm->users_count, 1);
638 639
	INIT_LIST_HEAD(&kvm->devices);

640
	r = kvm_arch_init_vm(kvm, type);
641
	if (r)
642
		goto out_err_no_disable;
643 644 645

	r = hardware_enable_all();
	if (r)
646
		goto out_err_no_disable;
647

648
#ifdef CONFIG_HAVE_KVM_IRQFD
649
	INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list);
650
#endif
A
Avi Kivity 已提交
651

652 653
	BUILD_BUG_ON(KVM_MEM_SLOTS_NUM > SHRT_MAX);

654
	r = -ENOMEM;
655
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
656 657
		struct kvm_memslots *slots = kvm_alloc_memslots();
		if (!slots)
658
			goto out_err_no_srcu;
659 660 661 662 663 664 665
		/*
		 * 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);
666
	}
667

668
	if (init_srcu_struct(&kvm->srcu))
669 670 671
		goto out_err_no_srcu;
	if (init_srcu_struct(&kvm->irq_srcu))
		goto out_err_no_irq_srcu;
M
Marcelo Tosatti 已提交
672
	for (i = 0; i < KVM_NR_BUSES; i++) {
673 674
		rcu_assign_pointer(kvm->buses[i],
			kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL));
675
		if (!kvm->buses[i])
M
Marcelo Tosatti 已提交
676 677
			goto out_err;
	}
678

679 680 681 682
	r = kvm_init_mmu_notifier(kvm);
	if (r)
		goto out_err;

683
	spin_lock(&kvm_lock);
684
	list_add(&kvm->vm_list, &vm_list);
685
	spin_unlock(&kvm_lock);
686

687 688
	preempt_notifier_inc();

689
	return kvm;
690 691

out_err:
692 693
	cleanup_srcu_struct(&kvm->irq_srcu);
out_err_no_irq_srcu:
694
	cleanup_srcu_struct(&kvm->srcu);
695
out_err_no_srcu:
696
	hardware_disable_all();
697
out_err_no_disable:
698
	refcount_set(&kvm->users_count, 0);
M
Marcelo Tosatti 已提交
699
	for (i = 0; i < KVM_NR_BUSES; i++)
700
		kfree(kvm_get_bus(kvm, i));
701
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
702
		kvm_free_memslots(kvm, __kvm_memslots(kvm, i));
703
	kvm_arch_free_vm(kvm);
704
	mmdrop(current->mm);
705
	return ERR_PTR(r);
706 707
}

708 709
static void kvm_destroy_devices(struct kvm *kvm)
{
G
Geliang Tang 已提交
710
	struct kvm_device *dev, *tmp;
711

712 713 714 715 716
	/*
	 * 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
717 718
	list_for_each_entry_safe(dev, tmp, &kvm->devices, vm_node) {
		list_del(&dev->vm_node);
719 720 721 722
		dev->ops->destroy(dev);
	}
}

723 724
static void kvm_destroy_vm(struct kvm *kvm)
{
M
Marcelo Tosatti 已提交
725
	int i;
726 727
	struct mm_struct *mm = kvm->mm;

728
	kvm_uevent_notify_change(KVM_EVENT_DESTROY_VM, kvm);
729
	kvm_destroy_vm_debugfs(kvm);
730
	kvm_arch_sync_events(kvm);
731
	spin_lock(&kvm_lock);
732
	list_del(&kvm->vm_list);
733
	spin_unlock(&kvm_lock);
734
	kvm_free_irq_routing(kvm);
735
	for (i = 0; i < KVM_NR_BUSES; i++) {
736
		struct kvm_io_bus *bus = kvm_get_bus(kvm, i);
737 738 739

		if (bus)
			kvm_io_bus_destroy(bus);
740 741
		kvm->buses[i] = NULL;
	}
742
	kvm_coalesced_mmio_free(kvm);
743 744
#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
	mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
745
#else
746
	kvm_arch_flush_shadow_all(kvm);
747
#endif
748
	kvm_arch_destroy_vm(kvm);
749
	kvm_destroy_devices(kvm);
750
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
751
		kvm_free_memslots(kvm, __kvm_memslots(kvm, i));
752
	cleanup_srcu_struct(&kvm->irq_srcu);
753 754
	cleanup_srcu_struct(&kvm->srcu);
	kvm_arch_free_vm(kvm);
755
	preempt_notifier_dec();
756
	hardware_disable_all();
757
	mmdrop(mm);
758 759
}

I
Izik Eidus 已提交
760 761
void kvm_get_kvm(struct kvm *kvm)
{
762
	refcount_inc(&kvm->users_count);
I
Izik Eidus 已提交
763 764 765 766 767
}
EXPORT_SYMBOL_GPL(kvm_get_kvm);

void kvm_put_kvm(struct kvm *kvm)
{
768
	if (refcount_dec_and_test(&kvm->users_count))
I
Izik Eidus 已提交
769 770 771 772 773
		kvm_destroy_vm(kvm);
}
EXPORT_SYMBOL_GPL(kvm_put_kvm);


774 775 776 777
static int kvm_vm_release(struct inode *inode, struct file *filp)
{
	struct kvm *kvm = filp->private_data;

G
Gregory Haskins 已提交
778 779
	kvm_irqfd_release(kvm);

I
Izik Eidus 已提交
780
	kvm_put_kvm(kvm);
A
Avi Kivity 已提交
781 782 783
	return 0;
}

784 785
/*
 * Allocation size is twice as large as the actual dirty bitmap size.
786
 * See x86's kvm_vm_ioctl_get_dirty_log() why this is needed.
787
 */
788 789
static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot)
{
790
	unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot);
791

M
Michal Hocko 已提交
792
	memslot->dirty_bitmap = kvzalloc(dirty_bytes, GFP_KERNEL);
793 794 795 796 797 798
	if (!memslot->dirty_bitmap)
		return -ENOMEM;

	return 0;
}

799
/*
800 801 802 803
 * 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.
804
 */
805
static void update_memslots(struct kvm_memslots *slots,
806 807
			    struct kvm_memory_slot *new,
			    enum kvm_mr_change change)
808
{
809 810
	int id = new->id;
	int i = slots->id_to_index[id];
811
	struct kvm_memory_slot *mslots = slots->memslots;
812

813
	WARN_ON(mslots[i].id != id);
814 815 816 817 818 819 820 821 822 823 824
	switch (change) {
	case KVM_MR_CREATE:
		slots->used_slots++;
		WARN_ON(mslots[i].npages || !new->npages);
		break;
	case KVM_MR_DELETE:
		slots->used_slots--;
		WARN_ON(new->npages || !mslots[i].npages);
		break;
	default:
		break;
825
	}
826

827
	while (i < KVM_MEM_SLOTS_NUM - 1 &&
828 829 830
	       new->base_gfn <= mslots[i + 1].base_gfn) {
		if (!mslots[i + 1].npages)
			break;
831 832 833 834
		mslots[i] = mslots[i + 1];
		slots->id_to_index[mslots[i].id] = i;
		i++;
	}
835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851

	/*
	 * 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--;
		}
852 853
	} else
		WARN_ON_ONCE(i != slots->used_slots);
854

855 856
	mslots[i] = *new;
	slots->id_to_index[mslots[i].id] = i;
857 858
}

859
static int check_memory_region_flags(const struct kvm_userspace_memory_region *mem)
860
{
X
Xiao Guangrong 已提交
861 862
	u32 valid_flags = KVM_MEM_LOG_DIRTY_PAGES;

863
#ifdef __KVM_HAVE_READONLY_MEM
X
Xiao Guangrong 已提交
864 865 866 867
	valid_flags |= KVM_MEM_READONLY;
#endif

	if (mem->flags & ~valid_flags)
868 869 870 871 872
		return -EINVAL;

	return 0;
}

873
static struct kvm_memslots *install_new_memslots(struct kvm *kvm,
874
		int as_id, struct kvm_memslots *slots)
875
{
876
	struct kvm_memslots *old_memslots = __kvm_memslots(kvm, as_id);
877

878 879 880 881 882 883 884
	/*
	 * 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;

885
	rcu_assign_pointer(kvm->memslots[as_id], slots);
886
	synchronize_srcu_expedited(&kvm->srcu);
887

888 889 890 891
	/*
	 * 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.
892 893 894 895 896 897
	 *
	 * 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, ...
898
	 */
899
	slots->generation += KVM_ADDRESS_SPACE_NUM * 2 - 1;
900

901
	kvm_arch_memslots_updated(kvm, slots);
902 903

	return old_memslots;
904 905
}

A
Avi Kivity 已提交
906 907 908 909 910
/*
 * Allocate some memory and give it an address in the guest physical address
 * space.
 *
 * Discontiguous memory is allowed, mostly for framebuffers.
911
 *
912
 * Must be called holding kvm->slots_lock for write.
A
Avi Kivity 已提交
913
 */
914
int __kvm_set_memory_region(struct kvm *kvm,
915
			    const struct kvm_userspace_memory_region *mem)
A
Avi Kivity 已提交
916
{
917
	int r;
A
Avi Kivity 已提交
918
	gfn_t base_gfn;
919
	unsigned long npages;
920
	struct kvm_memory_slot *slot;
A
Avi Kivity 已提交
921
	struct kvm_memory_slot old, new;
922
	struct kvm_memslots *slots = NULL, *old_memslots;
923
	int as_id, id;
924
	enum kvm_mr_change change;
A
Avi Kivity 已提交
925

926 927 928 929
	r = check_memory_region_flags(mem);
	if (r)
		goto out;

A
Avi Kivity 已提交
930
	r = -EINVAL;
931 932 933
	as_id = mem->slot >> 16;
	id = (u16)mem->slot;

A
Avi Kivity 已提交
934 935 936 937 938
	/* General sanity checks */
	if (mem->memory_size & (PAGE_SIZE - 1))
		goto out;
	if (mem->guest_phys_addr & (PAGE_SIZE - 1))
		goto out;
939
	/* We can read the guest memory with __xxx_user() later on. */
940
	if ((id < KVM_USER_MEM_SLOTS) &&
941
	    ((mem->userspace_addr & (PAGE_SIZE - 1)) ||
942
	     !access_ok((void __user *)(unsigned long)mem->userspace_addr,
943
			mem->memory_size)))
944
		goto out;
945
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_MEM_SLOTS_NUM)
A
Avi Kivity 已提交
946 947 948 949
		goto out;
	if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
		goto out;

950
	slot = id_to_memslot(__kvm_memslots(kvm, as_id), id);
A
Avi Kivity 已提交
951 952 953
	base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
	npages = mem->memory_size >> PAGE_SHIFT;

954 955 956
	if (npages > KVM_MEM_MAX_NR_PAGES)
		goto out;

957
	new = old = *slot;
A
Avi Kivity 已提交
958

959
	new.id = id;
A
Avi Kivity 已提交
960 961 962 963
	new.base_gfn = base_gfn;
	new.npages = npages;
	new.flags = mem->flags;

964 965 966 967 968
	if (npages) {
		if (!old.npages)
			change = KVM_MR_CREATE;
		else { /* Modify an existing slot. */
			if ((mem->userspace_addr != old.userspace_addr) ||
969 970
			    (npages != old.npages) ||
			    ((new.flags ^ old.flags) & KVM_MEM_READONLY))
971 972 973 974 975 976 977 978 979 980 981
				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;
			}
		}
982 983 984 985
	} else {
		if (!old.npages)
			goto out;

986
		change = KVM_MR_DELETE;
987 988 989
		new.base_gfn = 0;
		new.flags = 0;
	}
A
Avi Kivity 已提交
990

991
	if ((change == KVM_MR_CREATE) || (change == KVM_MR_MOVE)) {
992 993
		/* Check for overlaps */
		r = -EEXIST;
994
		kvm_for_each_memslot(slot, __kvm_memslots(kvm, as_id)) {
995
			if (slot->id == id)
996 997 998 999 1000
				continue;
			if (!((base_gfn + npages <= slot->base_gfn) ||
			      (base_gfn >= slot->base_gfn + slot->npages)))
				goto out;
		}
A
Avi Kivity 已提交
1001 1002 1003 1004
	}

	/* Free page dirty bitmap if unneeded */
	if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
A
Al Viro 已提交
1005
		new.dirty_bitmap = NULL;
A
Avi Kivity 已提交
1006 1007

	r = -ENOMEM;
1008
	if (change == KVM_MR_CREATE) {
1009
		new.userspace_addr = mem->userspace_addr;
1010

1011
		if (kvm_arch_create_memslot(kvm, &new, npages))
1012
			goto out_free;
A
Avi Kivity 已提交
1013
	}
1014

A
Avi Kivity 已提交
1015 1016
	/* Allocate page dirty bitmap if needed */
	if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
1017
		if (kvm_create_dirty_bitmap(&new) < 0)
1018
			goto out_free;
A
Avi Kivity 已提交
1019 1020
	}

M
Michal Hocko 已提交
1021
	slots = kvzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
1022 1023
	if (!slots)
		goto out_free;
1024
	memcpy(slots, __kvm_memslots(kvm, as_id), sizeof(struct kvm_memslots));
1025

1026
	if ((change == KVM_MR_DELETE) || (change == KVM_MR_MOVE)) {
1027
		slot = id_to_memslot(slots, id);
1028 1029
		slot->flags |= KVM_MEMSLOT_INVALID;

1030
		old_memslots = install_new_memslots(kvm, as_id, slots);
1031

1032 1033
		/* From this point no new shadow pages pointing to a deleted,
		 * or moved, memslot will be created.
1034 1035
		 *
		 * validation of sp->gfn happens in:
1036 1037
		 *	- gfn_to_hva (kvm_read_guest, gfn_to_pfn)
		 *	- kvm_is_visible_gfn (mmu_check_roots)
1038
		 */
1039
		kvm_arch_flush_shadow_memslot(kvm, slot);
1040 1041 1042 1043 1044 1045

		/*
		 * 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.
		 */
1046
		slots = old_memslots;
1047
	}
1048

1049
	r = kvm_arch_prepare_memory_region(kvm, &new, mem, change);
1050
	if (r)
1051
		goto out_slots;
1052

1053
	/* actual memory is freed via old in kvm_free_memslot below */
1054
	if (change == KVM_MR_DELETE) {
1055
		new.dirty_bitmap = NULL;
1056
		memset(&new.arch, 0, sizeof(new.arch));
1057 1058
	}

1059
	update_memslots(slots, &new, change);
1060
	old_memslots = install_new_memslots(kvm, as_id, slots);
1061

1062
	kvm_arch_commit_memory_region(kvm, mem, &old, &new, change);
1063

1064
	kvm_free_memslot(kvm, &old, &new);
1065
	kvfree(old_memslots);
A
Avi Kivity 已提交
1066 1067
	return 0;

1068
out_slots:
1069
	kvfree(slots);
1070
out_free:
1071
	kvm_free_memslot(kvm, &new, &old);
A
Avi Kivity 已提交
1072 1073
out:
	return r;
1074
}
1075 1076 1077
EXPORT_SYMBOL_GPL(__kvm_set_memory_region);

int kvm_set_memory_region(struct kvm *kvm,
1078
			  const struct kvm_userspace_memory_region *mem)
1079 1080 1081
{
	int r;

1082
	mutex_lock(&kvm->slots_lock);
1083
	r = __kvm_set_memory_region(kvm, mem);
1084
	mutex_unlock(&kvm->slots_lock);
1085 1086
	return r;
}
1087 1088
EXPORT_SYMBOL_GPL(kvm_set_memory_region);

1089 1090
static int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
					  struct kvm_userspace_memory_region *mem)
1091
{
1092
	if ((u16)mem->slot >= KVM_USER_MEM_SLOTS)
1093
		return -EINVAL;
1094

1095
	return kvm_set_memory_region(kvm, mem);
A
Avi Kivity 已提交
1096 1097
}

1098 1099
int kvm_get_dirty_log(struct kvm *kvm,
			struct kvm_dirty_log *log, int *is_dirty)
A
Avi Kivity 已提交
1100
{
1101
	struct kvm_memslots *slots;
A
Avi Kivity 已提交
1102
	struct kvm_memory_slot *memslot;
1103
	int i, as_id, id;
1104
	unsigned long n;
A
Avi Kivity 已提交
1105 1106
	unsigned long any = 0;

1107 1108 1109
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1110
		return -EINVAL;
A
Avi Kivity 已提交
1111

1112 1113
	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
A
Avi Kivity 已提交
1114
	if (!memslot->dirty_bitmap)
1115
		return -ENOENT;
A
Avi Kivity 已提交
1116

1117
	n = kvm_dirty_bitmap_bytes(memslot);
A
Avi Kivity 已提交
1118

1119
	for (i = 0; !any && i < n/sizeof(long); ++i)
A
Avi Kivity 已提交
1120 1121 1122
		any = memslot->dirty_bitmap[i];

	if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
1123
		return -EFAULT;
A
Avi Kivity 已提交
1124

1125 1126
	if (any)
		*is_dirty = 1;
1127
	return 0;
A
Avi Kivity 已提交
1128
}
1129
EXPORT_SYMBOL_GPL(kvm_get_dirty_log);
A
Avi Kivity 已提交
1130

1131 1132 1133
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
/**
 * kvm_get_dirty_log_protect - get a snapshot of dirty pages, and if any pages
1134
 *	and reenable dirty page tracking for the corresponding pages.
1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154
 * @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,
1155
			struct kvm_dirty_log *log, bool *flush)
1156
{
1157
	struct kvm_memslots *slots;
1158
	struct kvm_memory_slot *memslot;
1159
	int i, as_id, id;
1160 1161 1162 1163
	unsigned long n;
	unsigned long *dirty_bitmap;
	unsigned long *dirty_bitmap_buffer;

1164 1165 1166
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1167
		return -EINVAL;
1168

1169 1170
	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
1171 1172 1173

	dirty_bitmap = memslot->dirty_bitmap;
	if (!dirty_bitmap)
1174
		return -ENOENT;
1175 1176

	n = kvm_dirty_bitmap_bytes(memslot);
1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190
	*flush = false;
	if (kvm->manual_dirty_log_protect) {
		/*
		 * Unlike kvm_get_dirty_log, we always return false in *flush,
		 * because no flush is needed until KVM_CLEAR_DIRTY_LOG.  There
		 * is some code duplication between this function and
		 * kvm_get_dirty_log, but hopefully all architecture
		 * transition to kvm_get_dirty_log_protect and kvm_get_dirty_log
		 * can be eliminated.
		 */
		dirty_bitmap_buffer = dirty_bitmap;
	} else {
		dirty_bitmap_buffer = kvm_second_dirty_bitmap(memslot);
		memset(dirty_bitmap_buffer, 0, n);
1191

1192 1193 1194 1195
		spin_lock(&kvm->mmu_lock);
		for (i = 0; i < n / sizeof(long); i++) {
			unsigned long mask;
			gfn_t offset;
1196

1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229
			if (!dirty_bitmap[i])
				continue;

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

			if (mask) {
				offset = i * BITS_PER_LONG;
				kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot,
									offset, mask);
			}
		}
		spin_unlock(&kvm->mmu_lock);
	}

	if (copy_to_user(log->dirty_bitmap, dirty_bitmap_buffer, n))
		return -EFAULT;
	return 0;
}
EXPORT_SYMBOL_GPL(kvm_get_dirty_log_protect);

/**
 * kvm_clear_dirty_log_protect - clear dirty bits in the bitmap
 *	and reenable dirty page tracking for the corresponding pages.
 * @kvm:	pointer to kvm instance
 * @log:	slot id and address from which to fetch the bitmap of dirty pages
 */
int kvm_clear_dirty_log_protect(struct kvm *kvm,
				struct kvm_clear_dirty_log *log, bool *flush)
{
	struct kvm_memslots *slots;
	struct kvm_memory_slot *memslot;
1230
	int as_id, id;
1231
	gfn_t offset;
1232
	unsigned long i, n;
1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251
	unsigned long *dirty_bitmap;
	unsigned long *dirty_bitmap_buffer;

	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
		return -EINVAL;

	if ((log->first_page & 63) || (log->num_pages & 63))
		return -EINVAL;

	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);

	dirty_bitmap = memslot->dirty_bitmap;
	if (!dirty_bitmap)
		return -ENOENT;

	n = kvm_dirty_bitmap_bytes(memslot);
1252 1253 1254 1255 1256

	if (log->first_page > memslot->npages ||
	    log->num_pages > memslot->npages - log->first_page)
			return -EINVAL;

1257
	*flush = false;
1258 1259 1260
	dirty_bitmap_buffer = kvm_second_dirty_bitmap(memslot);
	if (copy_from_user(dirty_bitmap_buffer, log->dirty_bitmap, n))
		return -EFAULT;
1261

1262 1263 1264 1265 1266 1267 1268
	spin_lock(&kvm->mmu_lock);
	for (offset = log->first_page,
	     i = offset / BITS_PER_LONG, n = log->num_pages / BITS_PER_LONG; n--;
	     i++, offset += BITS_PER_LONG) {
		unsigned long mask = *dirty_bitmap_buffer++;
		atomic_long_t *p = (atomic_long_t *) &dirty_bitmap[i];
		if (!mask)
1269 1270
			continue;

1271
		mask &= atomic_long_fetch_andnot(mask, p);
1272

1273 1274 1275 1276 1277 1278
		/*
		 * mask contains the bits that really have been cleared.  This
		 * never includes any bits beyond the length of the memslot (if
		 * the length is not aligned to 64 pages), therefore it is not
		 * a problem if userspace sets them in log->dirty_bitmap.
		*/
1279
		if (mask) {
1280
			*flush = true;
1281 1282 1283
			kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot,
								offset, mask);
		}
1284 1285
	}
	spin_unlock(&kvm->mmu_lock);
1286

1287
	return 0;
1288
}
1289
EXPORT_SYMBOL_GPL(kvm_clear_dirty_log_protect);
1290 1291
#endif

1292 1293 1294 1295 1296
bool kvm_largepages_enabled(void)
{
	return largepages_enabled;
}

1297 1298 1299 1300 1301 1302
void kvm_disable_largepages(void)
{
	largepages_enabled = false;
}
EXPORT_SYMBOL_GPL(kvm_disable_largepages);

1303 1304 1305 1306
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
1307
EXPORT_SYMBOL_GPL(gfn_to_memslot);
A
Avi Kivity 已提交
1308

1309 1310 1311 1312 1313
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);
}

1314
bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
1315
{
1316
	struct kvm_memory_slot *memslot = gfn_to_memslot(kvm, gfn);
1317

1318
	if (!memslot || memslot->id >= KVM_USER_MEM_SLOTS ||
1319
	      memslot->flags & KVM_MEMSLOT_INVALID)
1320
		return false;
1321

1322
	return true;
1323 1324 1325
}
EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);

J
Joerg Roedel 已提交
1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349
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
1350 1351 1352 1353 1354 1355 1356
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
1357
{
1358
	if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
X
Xiao Guangrong 已提交
1359
		return KVM_HVA_ERR_BAD;
1360

X
Xiao Guangrong 已提交
1361 1362
	if (memslot_is_readonly(slot) && write)
		return KVM_HVA_ERR_RO_BAD;
1363 1364 1365 1366

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

X
Xiao Guangrong 已提交
1367
	return __gfn_to_hva_memslot(slot, gfn);
I
Izik Eidus 已提交
1368
}
1369

X
Xiao Guangrong 已提交
1370 1371 1372 1373
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
1374
}
1375

X
Xiao Guangrong 已提交
1376
unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot,
1377
					gfn_t gfn)
X
Xiao Guangrong 已提交
1378 1379 1380 1381 1382
{
	return gfn_to_hva_many(slot, gfn, NULL);
}
EXPORT_SYMBOL_GPL(gfn_to_hva_memslot);

1383 1384
unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
{
1385
	return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL);
1386
}
1387
EXPORT_SYMBOL_GPL(gfn_to_hva);
I
Izik Eidus 已提交
1388

1389 1390 1391 1392 1393 1394
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);

1395
/*
1396 1397 1398 1399 1400 1401
 * Return the hva of a @gfn and the R/W attribute if possible.
 *
 * @slot: the kvm_memory_slot which contains @gfn
 * @gfn: the gfn to be translated
 * @writable: used to return the read/write attribute of the @slot if the hva
 * is valid and @writable is not NULL
1402
 */
1403 1404
unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot,
				      gfn_t gfn, bool *writable)
1405
{
1406 1407 1408
	unsigned long hva = __gfn_to_hva_many(slot, gfn, NULL, false);

	if (!kvm_is_error_hva(hva) && writable)
1409 1410
		*writable = !memslot_is_readonly(slot);

1411
	return hva;
1412 1413
}

1414 1415 1416 1417 1418 1419 1420
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);
}

1421 1422 1423 1424 1425 1426 1427
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);
}

1428 1429
static inline int check_user_page_hwpoison(unsigned long addr)
{
L
Lorenzo Stoakes 已提交
1430
	int rc, flags = FOLL_HWPOISON | FOLL_WRITE;
1431

L
Lorenzo Stoakes 已提交
1432
	rc = get_user_pages(addr, 1, flags, NULL, NULL);
1433 1434 1435
	return rc == -EHWPOISON;
}

X
Xiao Guangrong 已提交
1436
/*
1437 1438 1439
 * The fast path to get the writable pfn which will be stored in @pfn,
 * true indicates success, otherwise false is returned.  It's also the
 * only part that runs if we can are in atomic context.
X
Xiao Guangrong 已提交
1440
 */
1441 1442
static bool hva_to_pfn_fast(unsigned long addr, bool write_fault,
			    bool *writable, kvm_pfn_t *pfn)
A
Avi Kivity 已提交
1443
{
1444
	struct page *page[1];
X
Xiao Guangrong 已提交
1445
	int npages;
A
Avi Kivity 已提交
1446

1447 1448 1449 1450 1451 1452 1453
	/*
	 * 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;
1454

X
Xiao Guangrong 已提交
1455 1456 1457
	npages = __get_user_pages_fast(addr, 1, 1, page);
	if (npages == 1) {
		*pfn = page_to_pfn(page[0]);
1458

X
Xiao Guangrong 已提交
1459 1460 1461 1462
		if (writable)
			*writable = true;
		return true;
	}
1463

X
Xiao Guangrong 已提交
1464 1465
	return false;
}
1466

X
Xiao Guangrong 已提交
1467 1468 1469 1470 1471
/*
 * 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
1472
			   bool *writable, kvm_pfn_t *pfn)
X
Xiao Guangrong 已提交
1473
{
1474 1475
	unsigned int flags = FOLL_HWPOISON;
	struct page *page;
X
Xiao Guangrong 已提交
1476
	int npages = 0;
1477

X
Xiao Guangrong 已提交
1478 1479 1480 1481 1482
	might_sleep();

	if (writable)
		*writable = write_fault;

1483 1484 1485 1486
	if (write_fault)
		flags |= FOLL_WRITE;
	if (async)
		flags |= FOLL_NOWAIT;
1487

1488
	npages = get_user_pages_unlocked(addr, 1, &page, flags);
X
Xiao Guangrong 已提交
1489 1490 1491 1492
	if (npages != 1)
		return npages;

	/* map read fault as writable if possible */
1493
	if (unlikely(!write_fault) && writable) {
1494
		struct page *wpage;
X
Xiao Guangrong 已提交
1495

1496
		if (__get_user_pages_fast(addr, 1, 1, &wpage) == 1) {
X
Xiao Guangrong 已提交
1497
			*writable = true;
1498 1499
			put_page(page);
			page = wpage;
1500
		}
1501
	}
1502
	*pfn = page_to_pfn(page);
X
Xiao Guangrong 已提交
1503 1504
	return npages;
}
I
Izik Eidus 已提交
1505

X
Xiao Guangrong 已提交
1506 1507 1508 1509
static bool vma_is_valid(struct vm_area_struct *vma, bool write_fault)
{
	if (unlikely(!(vma->vm_flags & VM_READ)))
		return false;
1510

X
Xiao Guangrong 已提交
1511 1512
	if (write_fault && (unlikely(!(vma->vm_flags & VM_WRITE))))
		return false;
1513

X
Xiao Guangrong 已提交
1514 1515
	return true;
}
1516

1517 1518
static int hva_to_pfn_remapped(struct vm_area_struct *vma,
			       unsigned long addr, bool *async,
1519 1520
			       bool write_fault, bool *writable,
			       kvm_pfn_t *p_pfn)
1521
{
1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545
	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;

	}

1546 1547
	if (writable)
		*writable = true;
1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562

	/*
	 * 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;
1563 1564 1565
	return 0;
}

1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579
/*
 * 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
1580
static kvm_pfn_t hva_to_pfn(unsigned long addr, bool atomic, bool *async,
X
Xiao Guangrong 已提交
1581 1582 1583
			bool write_fault, bool *writable)
{
	struct vm_area_struct *vma;
D
Dan Williams 已提交
1584
	kvm_pfn_t pfn = 0;
1585
	int npages, r;
1586

X
Xiao Guangrong 已提交
1587 1588
	/* we can do it either atomically or asynchronously, not both */
	BUG_ON(atomic && async);
1589

1590
	if (hva_to_pfn_fast(addr, write_fault, writable, &pfn))
X
Xiao Guangrong 已提交
1591 1592 1593 1594 1595 1596 1597 1598
		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;
1599

X
Xiao Guangrong 已提交
1600 1601 1602 1603 1604 1605 1606
	down_read(&current->mm->mmap_sem);
	if (npages == -EHWPOISON ||
	      (!async && check_user_page_hwpoison(addr))) {
		pfn = KVM_PFN_ERR_HWPOISON;
		goto exit;
	}

1607
retry:
X
Xiao Guangrong 已提交
1608 1609 1610 1611
	vma = find_vma_intersection(current->mm, addr, addr + 1);

	if (vma == NULL)
		pfn = KVM_PFN_ERR_FAULT;
1612
	else if (vma->vm_flags & (VM_IO | VM_PFNMAP)) {
1613
		r = hva_to_pfn_remapped(vma, addr, async, write_fault, writable, &pfn);
1614 1615
		if (r == -EAGAIN)
			goto retry;
1616 1617
		if (r < 0)
			pfn = KVM_PFN_ERR_FAULT;
X
Xiao Guangrong 已提交
1618
	} else {
X
Xiao Guangrong 已提交
1619
		if (async && vma_is_valid(vma, write_fault))
X
Xiao Guangrong 已提交
1620 1621 1622 1623 1624
			*async = true;
		pfn = KVM_PFN_ERR_FAULT;
	}
exit:
	up_read(&current->mm->mmap_sem);
1625
	return pfn;
1626 1627
}

D
Dan Williams 已提交
1628 1629 1630
kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
			       bool atomic, bool *async, bool write_fault,
			       bool *writable)
1631
{
X
Xiao Guangrong 已提交
1632 1633
	unsigned long addr = __gfn_to_hva_many(slot, gfn, NULL, write_fault);

1634 1635 1636
	if (addr == KVM_HVA_ERR_RO_BAD) {
		if (writable)
			*writable = false;
X
Xiao Guangrong 已提交
1637
		return KVM_PFN_ERR_RO_FAULT;
1638
	}
X
Xiao Guangrong 已提交
1639

1640 1641 1642
	if (kvm_is_error_hva(addr)) {
		if (writable)
			*writable = false;
1643
		return KVM_PFN_NOSLOT;
1644
	}
X
Xiao Guangrong 已提交
1645 1646 1647 1648 1649 1650 1651 1652 1653

	/* 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);
1654
}
1655
EXPORT_SYMBOL_GPL(__gfn_to_pfn_memslot);
1656

D
Dan Williams 已提交
1657
kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1658 1659
		      bool *writable)
{
P
Paolo Bonzini 已提交
1660 1661
	return __gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn, false, NULL,
				    write_fault, writable);
1662 1663 1664
}
EXPORT_SYMBOL_GPL(gfn_to_pfn_prot);

D
Dan Williams 已提交
1665
kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1666
{
X
Xiao Guangrong 已提交
1667
	return __gfn_to_pfn_memslot(slot, gfn, false, NULL, true, NULL);
1668
}
P
Paolo Bonzini 已提交
1669
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot);
1670

D
Dan Williams 已提交
1671
kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn)
1672
{
X
Xiao Guangrong 已提交
1673
	return __gfn_to_pfn_memslot(slot, gfn, true, NULL, true, NULL);
1674
}
1675
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot_atomic);
1676

D
Dan Williams 已提交
1677
kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1678 1679 1680 1681 1682
{
	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
1683
kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn)
1684 1685 1686 1687 1688
{
	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
1689
kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1690 1691 1692 1693 1694
{
	return gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn);
}
EXPORT_SYMBOL_GPL(gfn_to_pfn);

D
Dan Williams 已提交
1695
kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn)
1696 1697 1698 1699 1700
{
	return gfn_to_pfn_memslot(kvm_vcpu_gfn_to_memslot(vcpu, gfn), gfn);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_gfn_to_pfn);

1701 1702
int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
			    struct page **pages, int nr_pages)
1703 1704
{
	unsigned long addr;
1705
	gfn_t entry = 0;
1706

1707
	addr = gfn_to_hva_many(slot, gfn, &entry);
1708 1709 1710 1711 1712 1713 1714 1715 1716 1717
	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
1718
static struct page *kvm_pfn_to_page(kvm_pfn_t pfn)
X
Xiao Guangrong 已提交
1719
{
1720
	if (is_error_noslot_pfn(pfn))
1721
		return KVM_ERR_PTR_BAD_PAGE;
X
Xiao Guangrong 已提交
1722

1723
	if (kvm_is_reserved_pfn(pfn)) {
1724
		WARN_ON(1);
1725
		return KVM_ERR_PTR_BAD_PAGE;
1726
	}
X
Xiao Guangrong 已提交
1727 1728 1729 1730

	return pfn_to_page(pfn);
}

1731 1732
struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
{
D
Dan Williams 已提交
1733
	kvm_pfn_t pfn;
1734 1735 1736

	pfn = gfn_to_pfn(kvm, gfn);

X
Xiao Guangrong 已提交
1737
	return kvm_pfn_to_page(pfn);
A
Avi Kivity 已提交
1738 1739 1740
}
EXPORT_SYMBOL_GPL(gfn_to_page);

1741 1742
struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn)
{
D
Dan Williams 已提交
1743
	kvm_pfn_t pfn;
1744 1745 1746 1747 1748 1749 1750

	pfn = kvm_vcpu_gfn_to_pfn(vcpu, gfn);

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

1751 1752
void kvm_release_page_clean(struct page *page)
{
1753 1754
	WARN_ON(is_error_page(page));

1755
	kvm_release_pfn_clean(page_to_pfn(page));
1756 1757 1758
}
EXPORT_SYMBOL_GPL(kvm_release_page_clean);

D
Dan Williams 已提交
1759
void kvm_release_pfn_clean(kvm_pfn_t pfn)
1760
{
1761
	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn))
1762
		put_page(pfn_to_page(pfn));
1763 1764 1765
}
EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);

1766
void kvm_release_page_dirty(struct page *page)
1767
{
X
Xiao Guangrong 已提交
1768 1769
	WARN_ON(is_error_page(page));

1770 1771 1772 1773
	kvm_release_pfn_dirty(page_to_pfn(page));
}
EXPORT_SYMBOL_GPL(kvm_release_page_dirty);

1774
void kvm_release_pfn_dirty(kvm_pfn_t pfn)
1775 1776 1777 1778
{
	kvm_set_pfn_dirty(pfn);
	kvm_release_pfn_clean(pfn);
}
1779
EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
1780

D
Dan Williams 已提交
1781
void kvm_set_pfn_dirty(kvm_pfn_t pfn)
1782
{
1783
	if (!kvm_is_reserved_pfn(pfn)) {
1784
		struct page *page = pfn_to_page(pfn);
1785

1786 1787 1788
		if (!PageReserved(page))
			SetPageDirty(page);
	}
1789
}
1790 1791
EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);

D
Dan Williams 已提交
1792
void kvm_set_pfn_accessed(kvm_pfn_t pfn)
1793
{
1794
	if (!kvm_is_reserved_pfn(pfn))
1795
		mark_page_accessed(pfn_to_page(pfn));
1796 1797 1798
}
EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);

D
Dan Williams 已提交
1799
void kvm_get_pfn(kvm_pfn_t pfn)
1800
{
1801
	if (!kvm_is_reserved_pfn(pfn))
1802
		get_page(pfn_to_page(pfn));
1803 1804
}
EXPORT_SYMBOL_GPL(kvm_get_pfn);
1805

1806 1807 1808 1809 1810 1811 1812 1813
static int next_segment(unsigned long len, int offset)
{
	if (len > PAGE_SIZE - offset)
		return PAGE_SIZE - offset;
	else
		return len;
}

1814 1815
static int __kvm_read_guest_page(struct kvm_memory_slot *slot, gfn_t gfn,
				 void *data, int offset, int len)
1816
{
1817 1818
	int r;
	unsigned long addr;
1819

1820
	addr = gfn_to_hva_memslot_prot(slot, gfn, NULL);
1821 1822
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1823
	r = __copy_from_user(data, (void __user *)addr + offset, len);
1824
	if (r)
1825 1826 1827
		return -EFAULT;
	return 0;
}
1828 1829 1830 1831 1832 1833 1834 1835

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);
}
1836 1837
EXPORT_SYMBOL_GPL(kvm_read_guest_page);

1838 1839 1840 1841 1842 1843 1844 1845 1846
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);

1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866
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);

1867
int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data, unsigned long len)
1868 1869
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
1870
	int seg;
1871
	int offset = offset_in_page(gpa);
1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885
	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);
1886

1887 1888 1889 1890 1891 1892 1893
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);
1894 1895
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1896
	pagefault_disable();
1897
	r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
1898
	pagefault_enable();
1899 1900 1901 1902 1903
	if (r)
		return -EFAULT;
	return 0;
}

1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927
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)
1928
{
1929 1930
	int r;
	unsigned long addr;
1931

1932
	addr = gfn_to_hva_memslot(memslot, gfn);
1933 1934
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1935
	r = __copy_to_user((void __user *)addr + offset, data, len);
1936
	if (r)
1937
		return -EFAULT;
1938
	mark_page_dirty_in_slot(memslot, gfn);
1939 1940
	return 0;
}
1941 1942 1943 1944 1945 1946 1947 1948

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);
}
1949 1950
EXPORT_SYMBOL_GPL(kvm_write_guest_page);

1951 1952 1953 1954 1955 1956 1957 1958 1959
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);

1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978
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;
}
1979
EXPORT_SYMBOL_GPL(kvm_write_guest);
1980

1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
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);

2002 2003 2004
static int __kvm_gfn_to_hva_cache_init(struct kvm_memslots *slots,
				       struct gfn_to_hva_cache *ghc,
				       gpa_t gpa, unsigned long len)
2005 2006
{
	int offset = offset_in_page(gpa);
2007 2008 2009 2010
	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;
2011
	int r = start_gfn <= end_gfn ? 0 : -EINVAL;
2012 2013 2014

	ghc->gpa = gpa;
	ghc->generation = slots->generation;
2015
	ghc->len = len;
2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032
	ghc->hva = KVM_HVA_ERR_BAD;

	/*
	 * If the requested region crosses two memslots, we still
	 * verify that the entire region is valid here.
	 */
	while (!r && start_gfn <= end_gfn) {
		ghc->memslot = __gfn_to_memslot(slots, start_gfn);
		ghc->hva = gfn_to_hva_many(ghc->memslot, start_gfn,
					   &nr_pages_avail);
		if (kvm_is_error_hva(ghc->hva))
			r = -EFAULT;
		start_gfn += nr_pages_avail;
	}

	/* Use the slow path for cross page reads and writes. */
	if (!r && nr_pages_needed == 1)
2033
		ghc->hva += offset;
2034
	else
2035
		ghc->memslot = NULL;
2036 2037

	return r;
2038
}
2039

2040
int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
2041 2042
			      gpa_t gpa, unsigned long len)
{
2043
	struct kvm_memslots *slots = kvm_memslots(kvm);
2044 2045
	return __kvm_gfn_to_hva_cache_init(slots, ghc, gpa, len);
}
2046
EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init);
2047

2048
int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
2049 2050
				  void *data, unsigned int offset,
				  unsigned long len)
2051
{
2052
	struct kvm_memslots *slots = kvm_memslots(kvm);
2053
	int r;
2054
	gpa_t gpa = ghc->gpa + offset;
2055

2056
	BUG_ON(len + offset > ghc->len);
2057

2058
	if (slots->generation != ghc->generation)
2059
		__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len);
2060 2061

	if (unlikely(!ghc->memslot))
2062
		return kvm_write_guest(kvm, gpa, data, len);
2063 2064 2065 2066

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

2067
	r = __copy_to_user((void __user *)ghc->hva + offset, data, len);
2068 2069
	if (r)
		return -EFAULT;
2070
	mark_page_dirty_in_slot(ghc->memslot, gpa >> PAGE_SHIFT);
2071 2072 2073

	return 0;
}
2074
EXPORT_SYMBOL_GPL(kvm_write_guest_offset_cached);
2075

2076 2077
int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
2078
{
2079
	return kvm_write_guest_offset_cached(kvm, ghc, data, 0, len);
2080
}
2081
EXPORT_SYMBOL_GPL(kvm_write_guest_cached);
2082

2083 2084
int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
2085
{
2086
	struct kvm_memslots *slots = kvm_memslots(kvm);
2087 2088
	int r;

2089 2090
	BUG_ON(len > ghc->len);

2091
	if (slots->generation != ghc->generation)
2092
		__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len);
2093 2094

	if (unlikely(!ghc->memslot))
2095
		return kvm_read_guest(kvm, ghc->gpa, data, len);
2096 2097 2098 2099 2100 2101 2102 2103 2104 2105

	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;
}
2106
EXPORT_SYMBOL_GPL(kvm_read_guest_cached);
2107

2108 2109
int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
{
2110 2111 2112
	const void *zero_page = (const void *) __va(page_to_phys(ZERO_PAGE(0)));

	return kvm_write_guest_page(kvm, gfn, zero_page, offset, len);
2113 2114 2115 2116 2117 2118 2119 2120 2121 2122
}
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;

2123
	while ((seg = next_segment(len, offset)) != 0) {
2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134
		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);

2135
static void mark_page_dirty_in_slot(struct kvm_memory_slot *memslot,
2136
				    gfn_t gfn)
A
Avi Kivity 已提交
2137
{
R
Rusty Russell 已提交
2138 2139
	if (memslot && memslot->dirty_bitmap) {
		unsigned long rel_gfn = gfn - memslot->base_gfn;
A
Avi Kivity 已提交
2140

2141
		set_bit_le(rel_gfn, memslot->dirty_bitmap);
A
Avi Kivity 已提交
2142 2143 2144
	}
}

2145 2146 2147 2148 2149
void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
{
	struct kvm_memory_slot *memslot;

	memslot = gfn_to_memslot(kvm, gfn);
2150
	mark_page_dirty_in_slot(memslot, gfn);
2151
}
2152
EXPORT_SYMBOL_GPL(mark_page_dirty);
2153

2154 2155 2156 2157 2158 2159 2160 2161 2162
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);

2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185
void kvm_sigset_activate(struct kvm_vcpu *vcpu)
{
	if (!vcpu->sigset_active)
		return;

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

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

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

W
Wanpeng Li 已提交
2186 2187
static void grow_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2188
	unsigned int old, val, grow;
W
Wanpeng Li 已提交
2189

2190
	old = val = vcpu->halt_poll_ns;
2191
	grow = READ_ONCE(halt_poll_ns_grow);
W
Wanpeng Li 已提交
2192
	/* 10us base */
2193
	if (val == 0 && grow)
W
Wanpeng Li 已提交
2194 2195
		val = 10000;
	else
2196
		val *= grow;
W
Wanpeng Li 已提交
2197

2198 2199 2200
	if (val > halt_poll_ns)
		val = halt_poll_ns;

W
Wanpeng Li 已提交
2201
	vcpu->halt_poll_ns = val;
2202
	trace_kvm_halt_poll_ns_grow(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2203 2204 2205 2206
}

static void shrink_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2207
	unsigned int old, val, shrink;
W
Wanpeng Li 已提交
2208

2209
	old = val = vcpu->halt_poll_ns;
2210 2211
	shrink = READ_ONCE(halt_poll_ns_shrink);
	if (shrink == 0)
W
Wanpeng Li 已提交
2212 2213
		val = 0;
	else
2214
		val /= shrink;
W
Wanpeng Li 已提交
2215 2216

	vcpu->halt_poll_ns = val;
2217
	trace_kvm_halt_poll_ns_shrink(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2218 2219
}

2220 2221
static int kvm_vcpu_check_block(struct kvm_vcpu *vcpu)
{
2222 2223 2224
	int ret = -EINTR;
	int idx = srcu_read_lock(&vcpu->kvm->srcu);

2225 2226
	if (kvm_arch_vcpu_runnable(vcpu)) {
		kvm_make_request(KVM_REQ_UNHALT, vcpu);
2227
		goto out;
2228 2229
	}
	if (kvm_cpu_has_pending_timer(vcpu))
2230
		goto out;
2231
	if (signal_pending(current))
2232
		goto out;
2233

2234 2235 2236 2237
	ret = 0;
out:
	srcu_read_unlock(&vcpu->kvm->srcu, idx);
	return ret;
2238 2239
}

E
Eddie Dong 已提交
2240 2241 2242
/*
 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
 */
2243
void kvm_vcpu_block(struct kvm_vcpu *vcpu)
2244
{
2245
	ktime_t start, cur;
2246
	DECLARE_SWAITQUEUE(wait);
2247
	bool waited = false;
W
Wanpeng Li 已提交
2248
	u64 block_ns;
2249 2250

	start = cur = ktime_get();
W
Wanpeng Li 已提交
2251 2252
	if (vcpu->halt_poll_ns) {
		ktime_t stop = ktime_add_ns(ktime_get(), vcpu->halt_poll_ns);
2253

2254
		++vcpu->stat.halt_attempted_poll;
2255 2256 2257 2258 2259 2260 2261
		do {
			/*
			 * This sets KVM_REQ_UNHALT if an interrupt
			 * arrives.
			 */
			if (kvm_vcpu_check_block(vcpu) < 0) {
				++vcpu->stat.halt_successful_poll;
2262 2263
				if (!vcpu_valid_wakeup(vcpu))
					++vcpu->stat.halt_poll_invalid;
2264 2265 2266 2267 2268
				goto out;
			}
			cur = ktime_get();
		} while (single_task_running() && ktime_before(cur, stop));
	}
2269

2270 2271
	kvm_arch_vcpu_blocking(vcpu);

2272
	for (;;) {
2273
		prepare_to_swait_exclusive(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
2274

2275
		if (kvm_vcpu_check_block(vcpu) < 0)
2276 2277
			break;

2278
		waited = true;
E
Eddie Dong 已提交
2279 2280
		schedule();
	}
2281

2282
	finish_swait(&vcpu->wq, &wait);
2283 2284
	cur = ktime_get();

2285
	kvm_arch_vcpu_unblocking(vcpu);
2286
out:
W
Wanpeng Li 已提交
2287 2288
	block_ns = ktime_to_ns(cur) - ktime_to_ns(start);

2289 2290 2291
	if (!vcpu_valid_wakeup(vcpu))
		shrink_halt_poll_ns(vcpu);
	else if (halt_poll_ns) {
W
Wanpeng Li 已提交
2292 2293 2294
		if (block_ns <= vcpu->halt_poll_ns)
			;
		/* we had a long block, shrink polling */
2295
		else if (vcpu->halt_poll_ns && block_ns > halt_poll_ns)
W
Wanpeng Li 已提交
2296 2297 2298 2299 2300
			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);
2301 2302
	} else
		vcpu->halt_poll_ns = 0;
W
Wanpeng Li 已提交
2303

2304 2305
	trace_kvm_vcpu_wakeup(block_ns, waited, vcpu_valid_wakeup(vcpu));
	kvm_arch_vcpu_block_finish(vcpu);
E
Eddie Dong 已提交
2306
}
2307
EXPORT_SYMBOL_GPL(kvm_vcpu_block);
E
Eddie Dong 已提交
2308

2309
bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu)
2310
{
2311
	struct swait_queue_head *wqp;
2312 2313

	wqp = kvm_arch_vcpu_wq(vcpu);
2314
	if (swq_has_sleeper(wqp)) {
2315
		swake_up_one(wqp);
2316
		++vcpu->stat.halt_wakeup;
2317
		return true;
2318 2319
	}

2320
	return false;
2321 2322 2323
}
EXPORT_SYMBOL_GPL(kvm_vcpu_wake_up);

2324
#ifndef CONFIG_S390
2325 2326 2327 2328 2329 2330 2331 2332
/*
 * 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;

2333 2334 2335
	if (kvm_vcpu_wake_up(vcpu))
		return;

2336 2337 2338 2339 2340 2341
	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();
}
2342
EXPORT_SYMBOL_GPL(kvm_vcpu_kick);
2343
#endif /* !CONFIG_S390 */
2344

2345
int kvm_vcpu_yield_to(struct kvm_vcpu *target)
2346 2347 2348
{
	struct pid *pid;
	struct task_struct *task = NULL;
2349
	int ret = 0;
2350 2351 2352 2353

	rcu_read_lock();
	pid = rcu_dereference(target->pid);
	if (pid)
2354
		task = get_pid_task(pid, PIDTYPE_PID);
2355 2356
	rcu_read_unlock();
	if (!task)
2357 2358
		return ret;
	ret = yield_to(task, 1);
2359
	put_task_struct(task);
2360 2361

	return ret;
2362 2363 2364
}
EXPORT_SYMBOL_GPL(kvm_vcpu_yield_to);

2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386
/*
 * 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.
 */
2387
static bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
2388
{
2389
#ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
2390 2391 2392
	bool eligible;

	eligible = !vcpu->spin_loop.in_spin_loop ||
2393
		    vcpu->spin_loop.dy_eligible;
2394 2395 2396 2397 2398

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

	return eligible;
2399 2400
#else
	return true;
2401
#endif
2402
}
2403

2404
void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
Z
Zhai, Edwin 已提交
2405
{
2406 2407 2408 2409
	struct kvm *kvm = me->kvm;
	struct kvm_vcpu *vcpu;
	int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
	int yielded = 0;
2410
	int try = 3;
2411 2412
	int pass;
	int i;
Z
Zhai, Edwin 已提交
2413

2414
	kvm_vcpu_set_in_spin_loop(me, true);
2415 2416 2417 2418 2419 2420 2421
	/*
	 * 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.
	 */
2422
	for (pass = 0; pass < 2 && !yielded && try; pass++) {
2423
		kvm_for_each_vcpu(i, vcpu, kvm) {
2424
			if (!pass && i <= last_boosted_vcpu) {
2425 2426 2427 2428
				i = last_boosted_vcpu;
				continue;
			} else if (pass && i > last_boosted_vcpu)
				break;
2429
			if (!READ_ONCE(vcpu->preempted))
2430
				continue;
2431 2432
			if (vcpu == me)
				continue;
2433
			if (swait_active(&vcpu->wq) && !kvm_arch_vcpu_runnable(vcpu))
2434
				continue;
2435 2436
			if (yield_to_kernel_mode && !kvm_arch_vcpu_in_kernel(vcpu))
				continue;
2437 2438
			if (!kvm_vcpu_eligible_for_directed_yield(vcpu))
				continue;
2439 2440 2441

			yielded = kvm_vcpu_yield_to(vcpu);
			if (yielded > 0) {
2442 2443
				kvm->last_boosted_vcpu = i;
				break;
2444 2445 2446 2447
			} else if (yielded < 0) {
				try--;
				if (!try)
					break;
2448 2449 2450
			}
		}
	}
2451
	kvm_vcpu_set_in_spin_loop(me, false);
2452 2453 2454

	/* Ensure vcpu is not eligible during next spinloop */
	kvm_vcpu_set_dy_eligible(me, false);
Z
Zhai, Edwin 已提交
2455 2456 2457
}
EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin);

2458
static vm_fault_t kvm_vcpu_fault(struct vm_fault *vmf)
2459
{
2460
	struct kvm_vcpu *vcpu = vmf->vma->vm_file->private_data;
2461 2462
	struct page *page;

2463
	if (vmf->pgoff == 0)
2464
		page = virt_to_page(vcpu->run);
A
Avi Kivity 已提交
2465
#ifdef CONFIG_X86
2466
	else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
2467
		page = virt_to_page(vcpu->arch.pio_data);
2468
#endif
2469
#ifdef CONFIG_KVM_MMIO
2470 2471
	else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
		page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
A
Avi Kivity 已提交
2472
#endif
2473
	else
2474
		return kvm_arch_vcpu_fault(vcpu, vmf);
2475
	get_page(page);
2476 2477
	vmf->page = page;
	return 0;
2478 2479
}

2480
static const struct vm_operations_struct kvm_vcpu_vm_ops = {
2481
	.fault = kvm_vcpu_fault,
2482 2483 2484 2485 2486 2487 2488 2489
};

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
2490 2491 2492 2493
static int kvm_vcpu_release(struct inode *inode, struct file *filp)
{
	struct kvm_vcpu *vcpu = filp->private_data;

2494
	debugfs_remove_recursive(vcpu->debugfs_dentry);
A
Al Viro 已提交
2495
	kvm_put_kvm(vcpu->kvm);
A
Avi Kivity 已提交
2496 2497 2498
	return 0;
}

2499
static struct file_operations kvm_vcpu_fops = {
A
Avi Kivity 已提交
2500 2501
	.release        = kvm_vcpu_release,
	.unlocked_ioctl = kvm_vcpu_ioctl,
2502
	.mmap           = kvm_vcpu_mmap,
2503
	.llseek		= noop_llseek,
2504
	KVM_COMPAT(kvm_vcpu_compat_ioctl),
A
Avi Kivity 已提交
2505 2506 2507 2508 2509 2510 2511
};

/*
 * Allocates an inode for the vcpu.
 */
static int create_vcpu_fd(struct kvm_vcpu *vcpu)
{
2512 2513 2514 2515
	char name[8 + 1 + ITOA_MAX_LEN + 1];

	snprintf(name, sizeof(name), "kvm-vcpu:%d", vcpu->vcpu_id);
	return anon_inode_getfd(name, &kvm_vcpu_fops, vcpu, O_RDWR | O_CLOEXEC);
A
Avi Kivity 已提交
2516 2517
}

2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543
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;
}

2544 2545 2546
/*
 * Creates some virtual cpus.  Good luck creating more than one.
 */
2547
static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
2548 2549
{
	int r;
2550
	struct kvm_vcpu *vcpu;
2551

G
Greg Kurz 已提交
2552
	if (id >= KVM_MAX_VCPU_ID)
2553 2554
		return -EINVAL;

2555 2556 2557 2558 2559 2560 2561 2562 2563
	mutex_lock(&kvm->lock);
	if (kvm->created_vcpus == KVM_MAX_VCPUS) {
		mutex_unlock(&kvm->lock);
		return -EINVAL;
	}

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

2564
	vcpu = kvm_arch_vcpu_create(kvm, id);
2565 2566 2567 2568
	if (IS_ERR(vcpu)) {
		r = PTR_ERR(vcpu);
		goto vcpu_decrement;
	}
2569

2570 2571
	preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);

2572 2573
	r = kvm_arch_vcpu_setup(vcpu);
	if (r)
2574
		goto vcpu_destroy;
2575

2576 2577 2578 2579
	r = kvm_create_vcpu_debugfs(vcpu);
	if (r)
		goto vcpu_destroy;

S
Shaohua Li 已提交
2580
	mutex_lock(&kvm->lock);
2581 2582 2583 2584
	if (kvm_get_vcpu_by_id(kvm, id)) {
		r = -EEXIST;
		goto unlock_vcpu_destroy;
	}
2585 2586

	BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
2587

R
Rusty Russell 已提交
2588
	/* Now it's all set up, let userspace reach it */
A
Al Viro 已提交
2589
	kvm_get_kvm(kvm);
A
Avi Kivity 已提交
2590
	r = create_vcpu_fd(vcpu);
2591 2592
	if (r < 0) {
		kvm_put_kvm(kvm);
2593
		goto unlock_vcpu_destroy;
2594 2595 2596
	}

	kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu;
2597 2598 2599 2600 2601

	/*
	 * Pairs with smp_rmb() in kvm_get_vcpu.  Write kvm->vcpus
	 * before kvm->online_vcpu's incremented value.
	 */
2602 2603 2604 2605
	smp_wmb();
	atomic_inc(&kvm->online_vcpus);

	mutex_unlock(&kvm->lock);
2606
	kvm_arch_vcpu_postcreate(vcpu);
R
Rusty Russell 已提交
2607
	return r;
2608

2609
unlock_vcpu_destroy:
2610
	mutex_unlock(&kvm->lock);
2611
	debugfs_remove_recursive(vcpu->debugfs_dentry);
2612
vcpu_destroy:
2613
	kvm_arch_vcpu_destroy(vcpu);
2614 2615 2616 2617
vcpu_decrement:
	mutex_lock(&kvm->lock);
	kvm->created_vcpus--;
	mutex_unlock(&kvm->lock);
2618 2619 2620
	return r;
}

A
Avi Kivity 已提交
2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631
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
2632 2633
static long kvm_vcpu_ioctl(struct file *filp,
			   unsigned int ioctl, unsigned long arg)
A
Avi Kivity 已提交
2634
{
A
Avi Kivity 已提交
2635
	struct kvm_vcpu *vcpu = filp->private_data;
A
Al Viro 已提交
2636
	void __user *argp = (void __user *)arg;
2637
	int r;
2638 2639
	struct kvm_fpu *fpu = NULL;
	struct kvm_sregs *kvm_sregs = NULL;
A
Avi Kivity 已提交
2640

2641 2642
	if (vcpu->kvm->mm != current->mm)
		return -EIO;
2643

2644 2645 2646
	if (unlikely(_IOC_TYPE(ioctl) != KVMIO))
		return -EINVAL;

2647
	/*
2648 2649
	 * Some architectures have vcpu ioctls that are asynchronous to vcpu
	 * execution; mutex_lock() would break them.
2650
	 */
2651 2652
	r = kvm_arch_vcpu_async_ioctl(filp, ioctl, arg);
	if (r != -ENOIOCTLCMD)
2653
		return r;
2654

2655 2656
	if (mutex_lock_killable(&vcpu->mutex))
		return -EINTR;
A
Avi Kivity 已提交
2657
	switch (ioctl) {
2658 2659
	case KVM_RUN: {
		struct pid *oldpid;
2660 2661 2662
		r = -EINVAL;
		if (arg)
			goto out;
2663
		oldpid = rcu_access_pointer(vcpu->pid);
2664
		if (unlikely(oldpid != task_pid(current))) {
2665
			/* The thread running this VCPU changed. */
2666
			struct pid *newpid;
2667

2668 2669 2670 2671 2672
			r = kvm_arch_vcpu_run_pid_change(vcpu);
			if (r)
				break;

			newpid = get_task_pid(current, PIDTYPE_PID);
2673 2674 2675 2676 2677
			rcu_assign_pointer(vcpu->pid, newpid);
			if (oldpid)
				synchronize_rcu();
			put_pid(oldpid);
		}
2678
		r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
2679
		trace_kvm_userspace_exit(vcpu->run->exit_reason, r);
A
Avi Kivity 已提交
2680
		break;
2681
	}
A
Avi Kivity 已提交
2682
	case KVM_GET_REGS: {
2683
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
2684

2685 2686 2687
		r = -ENOMEM;
		kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
		if (!kvm_regs)
A
Avi Kivity 已提交
2688
			goto out;
2689 2690 2691
		r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
		if (r)
			goto out_free1;
A
Avi Kivity 已提交
2692
		r = -EFAULT;
2693 2694
		if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
			goto out_free1;
A
Avi Kivity 已提交
2695
		r = 0;
2696 2697
out_free1:
		kfree(kvm_regs);
A
Avi Kivity 已提交
2698 2699 2700
		break;
	}
	case KVM_SET_REGS: {
2701
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
2702

2703
		r = -ENOMEM;
2704 2705 2706
		kvm_regs = memdup_user(argp, sizeof(*kvm_regs));
		if (IS_ERR(kvm_regs)) {
			r = PTR_ERR(kvm_regs);
A
Avi Kivity 已提交
2707
			goto out;
2708
		}
2709 2710
		r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
		kfree(kvm_regs);
A
Avi Kivity 已提交
2711 2712 2713
		break;
	}
	case KVM_GET_SREGS: {
2714 2715 2716 2717 2718
		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
2719 2720 2721
		if (r)
			goto out;
		r = -EFAULT;
2722
		if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
A
Avi Kivity 已提交
2723 2724 2725 2726 2727
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_SREGS: {
2728 2729 2730
		kvm_sregs = memdup_user(argp, sizeof(*kvm_sregs));
		if (IS_ERR(kvm_sregs)) {
			r = PTR_ERR(kvm_sregs);
G
Guo Chao 已提交
2731
			kvm_sregs = NULL;
A
Avi Kivity 已提交
2732
			goto out;
2733
		}
2734
		r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
2735 2736
		break;
	}
2737 2738 2739 2740 2741 2742 2743
	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;
2744
		if (copy_to_user(argp, &mp_state, sizeof(mp_state)))
2745 2746 2747 2748 2749 2750 2751 2752
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_MP_STATE: {
		struct kvm_mp_state mp_state;

		r = -EFAULT;
2753
		if (copy_from_user(&mp_state, argp, sizeof(mp_state)))
2754 2755 2756 2757
			goto out;
		r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
		break;
	}
A
Avi Kivity 已提交
2758 2759 2760 2761
	case KVM_TRANSLATE: {
		struct kvm_translation tr;

		r = -EFAULT;
2762
		if (copy_from_user(&tr, argp, sizeof(tr)))
A
Avi Kivity 已提交
2763
			goto out;
2764
		r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
A
Avi Kivity 已提交
2765 2766 2767
		if (r)
			goto out;
		r = -EFAULT;
2768
		if (copy_to_user(argp, &tr, sizeof(tr)))
A
Avi Kivity 已提交
2769 2770 2771 2772
			goto out;
		r = 0;
		break;
	}
J
Jan Kiszka 已提交
2773 2774
	case KVM_SET_GUEST_DEBUG: {
		struct kvm_guest_debug dbg;
A
Avi Kivity 已提交
2775 2776

		r = -EFAULT;
2777
		if (copy_from_user(&dbg, argp, sizeof(dbg)))
A
Avi Kivity 已提交
2778
			goto out;
J
Jan Kiszka 已提交
2779
		r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
A
Avi Kivity 已提交
2780 2781
		break;
	}
A
Avi Kivity 已提交
2782 2783 2784 2785 2786 2787 2788 2789 2790
	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,
2791
					   sizeof(kvm_sigmask)))
A
Avi Kivity 已提交
2792 2793
				goto out;
			r = -EINVAL;
2794
			if (kvm_sigmask.len != sizeof(sigset))
A
Avi Kivity 已提交
2795 2796 2797
				goto out;
			r = -EFAULT;
			if (copy_from_user(&sigset, sigmask_arg->sigset,
2798
					   sizeof(sigset)))
A
Avi Kivity 已提交
2799 2800 2801
				goto out;
			p = &sigset;
		}
2802
		r = kvm_vcpu_ioctl_set_sigmask(vcpu, p);
A
Avi Kivity 已提交
2803 2804
		break;
	}
A
Avi Kivity 已提交
2805
	case KVM_GET_FPU: {
2806 2807 2808 2809 2810
		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
2811 2812 2813
		if (r)
			goto out;
		r = -EFAULT;
2814
		if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
A
Avi Kivity 已提交
2815 2816 2817 2818 2819
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_FPU: {
2820 2821 2822
		fpu = memdup_user(argp, sizeof(*fpu));
		if (IS_ERR(fpu)) {
			r = PTR_ERR(fpu);
G
Guo Chao 已提交
2823
			fpu = NULL;
A
Avi Kivity 已提交
2824
			goto out;
2825
		}
2826
		r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
A
Avi Kivity 已提交
2827 2828
		break;
	}
A
Avi Kivity 已提交
2829
	default:
2830
		r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
2831 2832
	}
out:
2833
	mutex_unlock(&vcpu->mutex);
2834 2835
	kfree(fpu);
	kfree(kvm_sregs);
A
Avi Kivity 已提交
2836 2837 2838
	return r;
}

2839
#ifdef CONFIG_KVM_COMPAT
2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858
static long kvm_vcpu_compat_ioctl(struct file *filp,
				  unsigned int ioctl, unsigned long arg)
{
	struct kvm_vcpu *vcpu = filp->private_data;
	void __user *argp = compat_ptr(arg);
	int r;

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

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

		if (argp) {
			r = -EFAULT;
			if (copy_from_user(&kvm_sigmask, argp,
2859
					   sizeof(kvm_sigmask)))
2860 2861
				goto out;
			r = -EINVAL;
A
Al Viro 已提交
2862
			if (kvm_sigmask.len != sizeof(compat_sigset_t))
2863 2864
				goto out;
			r = -EFAULT;
A
Al Viro 已提交
2865
			if (get_compat_sigset(&sigset, (void *)sigmask_arg->sigset))
2866
				goto out;
2867 2868 2869
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
		} else
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, NULL);
2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880
		break;
	}
	default:
		r = kvm_vcpu_ioctl(filp, ioctl, arg);
	}

out:
	return r;
}
#endif

S
Scott Wood 已提交
2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928
static int kvm_device_ioctl_attr(struct kvm_device *dev,
				 int (*accessor)(struct kvm_device *dev,
						 struct kvm_device_attr *attr),
				 unsigned long arg)
{
	struct kvm_device_attr attr;

	if (!accessor)
		return -EPERM;

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

	return accessor(dev, &attr);
}

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

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

		return -ENOTTY;
	}
}

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

	kvm_put_kvm(kvm);
	return 0;
}

static const struct file_operations kvm_device_fops = {
	.unlocked_ioctl = kvm_device_ioctl,
	.release = kvm_device_release,
2929
	KVM_COMPAT(kvm_device_ioctl),
S
Scott Wood 已提交
2930 2931 2932 2933 2934 2935 2936 2937 2938 2939
};

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

	return filp->private_data;
}

2940
static struct kvm_device_ops *kvm_device_ops_table[KVM_DEV_TYPE_MAX] = {
2941
#ifdef CONFIG_KVM_MPIC
2942 2943
	[KVM_DEV_TYPE_FSL_MPIC_20]	= &kvm_mpic_ops,
	[KVM_DEV_TYPE_FSL_MPIC_42]	= &kvm_mpic_ops,
2944
#endif
2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958
};

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;
}

2959 2960 2961 2962 2963 2964
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
2965 2966 2967 2968 2969 2970 2971 2972
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;

2973 2974 2975 2976 2977
	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
2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989
		return -ENODEV;

	if (test)
		return 0;

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

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

2990
	mutex_lock(&kvm->lock);
S
Scott Wood 已提交
2991 2992
	ret = ops->create(dev, cd->type);
	if (ret < 0) {
2993
		mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
2994 2995 2996
		kfree(dev);
		return ret;
	}
2997 2998
	list_add(&dev->vm_node, &kvm->devices);
	mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
2999

3000 3001 3002
	if (ops->init)
		ops->init(dev);

3003
	kvm_get_kvm(kvm);
3004
	ret = anon_inode_getfd(ops->name, &kvm_device_fops, dev, O_RDWR | O_CLOEXEC);
S
Scott Wood 已提交
3005
	if (ret < 0) {
3006
		kvm_put_kvm(kvm);
3007 3008 3009
		mutex_lock(&kvm->lock);
		list_del(&dev->vm_node);
		mutex_unlock(&kvm->lock);
3010
		ops->destroy(dev);
S
Scott Wood 已提交
3011 3012 3013 3014 3015 3016 3017
		return ret;
	}

	cd->fd = ret;
	return 0;
}

3018 3019 3020 3021 3022 3023 3024 3025 3026 3027
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
3028
#ifdef CONFIG_HAVE_KVM_IRQFD
3029
	case KVM_CAP_IRQFD:
3030 3031
	case KVM_CAP_IRQFD_RESAMPLE:
#endif
3032
	case KVM_CAP_IOEVENTFD_ANY_LENGTH:
3033
	case KVM_CAP_CHECK_EXTENSION_VM:
3034
	case KVM_CAP_ENABLE_CAP_VM:
3035 3036 3037
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
	case KVM_CAP_MANUAL_DIRTY_LOG_PROTECT:
#endif
3038
		return 1;
3039
#ifdef CONFIG_KVM_MMIO
3040 3041
	case KVM_CAP_COALESCED_MMIO:
		return KVM_COALESCED_MMIO_PAGE_OFFSET;
P
Peng Hao 已提交
3042 3043
	case KVM_CAP_COALESCED_PIO:
		return 1;
3044
#endif
3045 3046 3047
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_CAP_IRQ_ROUTING:
		return KVM_MAX_IRQ_ROUTES;
3048 3049 3050 3051
#endif
#if KVM_ADDRESS_SPACE_NUM > 1
	case KVM_CAP_MULTI_ADDRESS_SPACE:
		return KVM_ADDRESS_SPACE_NUM;
3052
#endif
G
Greg Kurz 已提交
3053 3054
	case KVM_CAP_MAX_VCPU_ID:
		return KVM_MAX_VCPU_ID;
3055 3056 3057 3058 3059 3060
	default:
		break;
	}
	return kvm_vm_ioctl_check_extension(kvm, arg);
}

3061 3062 3063 3064 3065 3066 3067 3068 3069 3070
int __attribute__((weak)) kvm_vm_ioctl_enable_cap(struct kvm *kvm,
						  struct kvm_enable_cap *cap)
{
	return -EINVAL;
}

static int kvm_vm_ioctl_enable_cap_generic(struct kvm *kvm,
					   struct kvm_enable_cap *cap)
{
	switch (cap->cap) {
3071 3072 3073 3074 3075 3076 3077
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
	case KVM_CAP_MANUAL_DIRTY_LOG_PROTECT:
		if (cap->flags || (cap->args[0] & ~1))
			return -EINVAL;
		kvm->manual_dirty_log_protect = cap->args[0];
		return 0;
#endif
3078 3079 3080 3081 3082
	default:
		return kvm_vm_ioctl_enable_cap(kvm, cap);
	}
}

A
Avi Kivity 已提交
3083 3084 3085 3086 3087
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;
3088
	int r;
A
Avi Kivity 已提交
3089

3090 3091
	if (kvm->mm != current->mm)
		return -EIO;
A
Avi Kivity 已提交
3092 3093 3094 3095
	switch (ioctl) {
	case KVM_CREATE_VCPU:
		r = kvm_vm_ioctl_create_vcpu(kvm, arg);
		break;
3096 3097 3098 3099 3100 3101 3102 3103 3104
	case KVM_ENABLE_CAP: {
		struct kvm_enable_cap cap;

		r = -EFAULT;
		if (copy_from_user(&cap, argp, sizeof(cap)))
			goto out;
		r = kvm_vm_ioctl_enable_cap_generic(kvm, &cap);
		break;
	}
3105 3106 3107 3108 3109
	case KVM_SET_USER_MEMORY_REGION: {
		struct kvm_userspace_memory_region kvm_userspace_mem;

		r = -EFAULT;
		if (copy_from_user(&kvm_userspace_mem, argp,
3110
						sizeof(kvm_userspace_mem)))
3111 3112
			goto out;

3113
		r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem);
A
Avi Kivity 已提交
3114 3115 3116 3117 3118 3119
		break;
	}
	case KVM_GET_DIRTY_LOG: {
		struct kvm_dirty_log log;

		r = -EFAULT;
3120
		if (copy_from_user(&log, argp, sizeof(log)))
A
Avi Kivity 已提交
3121
			goto out;
3122
		r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
A
Avi Kivity 已提交
3123 3124
		break;
	}
3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
	case KVM_CLEAR_DIRTY_LOG: {
		struct kvm_clear_dirty_log log;

		r = -EFAULT;
		if (copy_from_user(&log, argp, sizeof(log)))
			goto out;
		r = kvm_vm_ioctl_clear_dirty_log(kvm, &log);
		break;
	}
#endif
3136
#ifdef CONFIG_KVM_MMIO
3137 3138
	case KVM_REGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3139

3140
		r = -EFAULT;
3141
		if (copy_from_user(&zone, argp, sizeof(zone)))
3142 3143 3144 3145 3146 3147
			goto out;
		r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
		break;
	}
	case KVM_UNREGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3148

3149
		r = -EFAULT;
3150
		if (copy_from_user(&zone, argp, sizeof(zone)))
3151 3152 3153 3154 3155
			goto out;
		r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
		break;
	}
#endif
G
Gregory Haskins 已提交
3156 3157 3158 3159
	case KVM_IRQFD: {
		struct kvm_irqfd data;

		r = -EFAULT;
3160
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3161
			goto out;
3162
		r = kvm_irqfd(kvm, &data);
G
Gregory Haskins 已提交
3163 3164
		break;
	}
G
Gregory Haskins 已提交
3165 3166 3167 3168
	case KVM_IOEVENTFD: {
		struct kvm_ioeventfd data;

		r = -EFAULT;
3169
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3170 3171 3172 3173
			goto out;
		r = kvm_ioeventfd(kvm, &data);
		break;
	}
3174 3175 3176 3177 3178
#ifdef CONFIG_HAVE_KVM_MSI
	case KVM_SIGNAL_MSI: {
		struct kvm_msi msi;

		r = -EFAULT;
3179
		if (copy_from_user(&msi, argp, sizeof(msi)))
3180 3181 3182 3183
			goto out;
		r = kvm_send_userspace_msi(kvm, &msi);
		break;
	}
3184 3185 3186 3187 3188 3189 3190
#endif
#ifdef __KVM_HAVE_IRQ_LINE
	case KVM_IRQ_LINE_STATUS:
	case KVM_IRQ_LINE: {
		struct kvm_irq_level irq_event;

		r = -EFAULT;
3191
		if (copy_from_user(&irq_event, argp, sizeof(irq_event)))
3192 3193
			goto out;

3194 3195
		r = kvm_vm_ioctl_irq_line(kvm, &irq_event,
					ioctl == KVM_IRQ_LINE_STATUS);
3196 3197 3198 3199 3200
		if (r)
			goto out;

		r = -EFAULT;
		if (ioctl == KVM_IRQ_LINE_STATUS) {
3201
			if (copy_to_user(argp, &irq_event, sizeof(irq_event)))
3202 3203 3204 3205 3206 3207
				goto out;
		}

		r = 0;
		break;
	}
3208
#endif
3209 3210 3211 3212
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_SET_GSI_ROUTING: {
		struct kvm_irq_routing routing;
		struct kvm_irq_routing __user *urouting;
3213
		struct kvm_irq_routing_entry *entries = NULL;
3214 3215 3216 3217 3218

		r = -EFAULT;
		if (copy_from_user(&routing, argp, sizeof(routing)))
			goto out;
		r = -EINVAL;
3219 3220
		if (!kvm_arch_can_set_irq_routing(kvm))
			goto out;
3221
		if (routing.nr > KVM_MAX_IRQ_ROUTES)
3222 3223 3224
			goto out;
		if (routing.flags)
			goto out;
3225 3226
		if (routing.nr) {
			r = -ENOMEM;
3227 3228
			entries = vmalloc(array_size(sizeof(*entries),
						     routing.nr));
3229 3230 3231 3232 3233 3234 3235 3236
			if (!entries)
				goto out;
			r = -EFAULT;
			urouting = argp;
			if (copy_from_user(entries, urouting->entries,
					   routing.nr * sizeof(*entries)))
				goto out_free_irq_routing;
		}
3237 3238
		r = kvm_set_irq_routing(kvm, entries, routing.nr,
					routing.flags);
3239
out_free_irq_routing:
3240 3241 3242 3243
		vfree(entries);
		break;
	}
#endif /* CONFIG_HAVE_KVM_IRQ_ROUTING */
S
Scott Wood 已提交
3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261
	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;
	}
3262 3263 3264
	case KVM_CHECK_EXTENSION:
		r = kvm_vm_ioctl_check_extension_generic(kvm, arg);
		break;
3265
	default:
3266
		r = kvm_arch_vm_ioctl(filp, ioctl, arg);
3267 3268 3269 3270 3271
	}
out:
	return r;
}

3272
#ifdef CONFIG_KVM_COMPAT
3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296
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)))
3297
			return -EFAULT;
3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312
		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

3313
static struct file_operations kvm_vm_fops = {
3314 3315
	.release        = kvm_vm_release,
	.unlocked_ioctl = kvm_vm_ioctl,
3316
	.llseek		= noop_llseek,
3317
	KVM_COMPAT(kvm_vm_compat_ioctl),
3318 3319
};

3320
static int kvm_dev_ioctl_create_vm(unsigned long type)
3321
{
3322
	int r;
3323
	struct kvm *kvm;
3324
	struct file *file;
3325

3326
	kvm = kvm_create_vm(type);
3327 3328
	if (IS_ERR(kvm))
		return PTR_ERR(kvm);
3329
#ifdef CONFIG_KVM_MMIO
3330
	r = kvm_coalesced_mmio_init(kvm);
3331 3332
	if (r < 0)
		goto put_kvm;
3333
#endif
3334
	r = get_unused_fd_flags(O_CLOEXEC);
3335 3336 3337
	if (r < 0)
		goto put_kvm;

3338 3339 3340
	file = anon_inode_getfile("kvm-vm", &kvm_vm_fops, kvm, O_RDWR);
	if (IS_ERR(file)) {
		put_unused_fd(r);
3341 3342
		r = PTR_ERR(file);
		goto put_kvm;
3343
	}
3344

3345 3346 3347 3348 3349 3350
	/*
	 * Don't call kvm_put_kvm anymore at this point; file->f_op is
	 * already set, with ->release() being kvm_vm_release().  In error
	 * cases it will be called by the final fput(file) and will take
	 * care of doing kvm_put_kvm(kvm).
	 */
3351
	if (kvm_create_vm_debugfs(kvm, r) < 0) {
3352 3353
		put_unused_fd(r);
		fput(file);
3354 3355
		return -ENOMEM;
	}
3356
	kvm_uevent_notify_change(KVM_EVENT_CREATE_VM, kvm);
3357

3358
	fd_install(r, file);
3359
	return r;
3360 3361 3362 3363

put_kvm:
	kvm_put_kvm(kvm);
	return r;
3364 3365 3366 3367 3368
}

static long kvm_dev_ioctl(struct file *filp,
			  unsigned int ioctl, unsigned long arg)
{
3369
	long r = -EINVAL;
3370 3371 3372

	switch (ioctl) {
	case KVM_GET_API_VERSION:
3373 3374
		if (arg)
			goto out;
3375 3376 3377
		r = KVM_API_VERSION;
		break;
	case KVM_CREATE_VM:
3378
		r = kvm_dev_ioctl_create_vm(arg);
3379
		break;
3380
	case KVM_CHECK_EXTENSION:
3381
		r = kvm_vm_ioctl_check_extension_generic(NULL, arg);
3382
		break;
3383 3384 3385
	case KVM_GET_VCPU_MMAP_SIZE:
		if (arg)
			goto out;
3386 3387 3388
		r = PAGE_SIZE;     /* struct kvm_run */
#ifdef CONFIG_X86
		r += PAGE_SIZE;    /* pio data page */
3389
#endif
3390
#ifdef CONFIG_KVM_MMIO
3391
		r += PAGE_SIZE;    /* coalesced mmio ring page */
3392
#endif
3393
		break;
3394 3395 3396
	case KVM_TRACE_ENABLE:
	case KVM_TRACE_PAUSE:
	case KVM_TRACE_DISABLE:
3397
		r = -EOPNOTSUPP;
3398
		break;
A
Avi Kivity 已提交
3399
	default:
3400
		return kvm_arch_dev_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
3401 3402 3403 3404 3405 3406 3407
	}
out:
	return r;
}

static struct file_operations kvm_chardev_ops = {
	.unlocked_ioctl = kvm_dev_ioctl,
3408
	.llseek		= noop_llseek,
3409
	KVM_COMPAT(kvm_dev_ioctl),
A
Avi Kivity 已提交
3410 3411 3412
};

static struct miscdevice kvm_dev = {
A
Avi Kivity 已提交
3413
	KVM_MINOR,
A
Avi Kivity 已提交
3414 3415 3416 3417
	"kvm",
	&kvm_chardev_ops,
};

3418
static void hardware_enable_nolock(void *junk)
3419 3420
{
	int cpu = raw_smp_processor_id();
3421
	int r;
3422

3423
	if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
3424
		return;
3425

3426
	cpumask_set_cpu(cpu, cpus_hardware_enabled);
3427

3428
	r = kvm_arch_hardware_enable();
3429 3430 3431 3432

	if (r) {
		cpumask_clear_cpu(cpu, cpus_hardware_enabled);
		atomic_inc(&hardware_enable_failed);
X
Xiubo Li 已提交
3433
		pr_info("kvm: enabling virtualization on CPU%d failed\n", cpu);
3434
	}
3435 3436
}

3437
static int kvm_starting_cpu(unsigned int cpu)
3438
{
3439
	raw_spin_lock(&kvm_count_lock);
3440 3441
	if (kvm_usage_count)
		hardware_enable_nolock(NULL);
3442
	raw_spin_unlock(&kvm_count_lock);
3443
	return 0;
3444 3445 3446
}

static void hardware_disable_nolock(void *junk)
3447 3448 3449
{
	int cpu = raw_smp_processor_id();

3450
	if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
3451
		return;
3452
	cpumask_clear_cpu(cpu, cpus_hardware_enabled);
3453
	kvm_arch_hardware_disable();
3454 3455
}

3456
static int kvm_dying_cpu(unsigned int cpu)
3457
{
3458
	raw_spin_lock(&kvm_count_lock);
3459 3460
	if (kvm_usage_count)
		hardware_disable_nolock(NULL);
3461
	raw_spin_unlock(&kvm_count_lock);
3462
	return 0;
3463 3464
}

3465 3466 3467 3468 3469 3470
static void hardware_disable_all_nolock(void)
{
	BUG_ON(!kvm_usage_count);

	kvm_usage_count--;
	if (!kvm_usage_count)
3471
		on_each_cpu(hardware_disable_nolock, NULL, 1);
3472 3473 3474 3475
}

static void hardware_disable_all(void)
{
3476
	raw_spin_lock(&kvm_count_lock);
3477
	hardware_disable_all_nolock();
3478
	raw_spin_unlock(&kvm_count_lock);
3479 3480 3481 3482 3483 3484
}

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

3485
	raw_spin_lock(&kvm_count_lock);
3486 3487 3488 3489

	kvm_usage_count++;
	if (kvm_usage_count == 1) {
		atomic_set(&hardware_enable_failed, 0);
3490
		on_each_cpu(hardware_enable_nolock, NULL, 1);
3491 3492 3493 3494 3495 3496 3497

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

3498
	raw_spin_unlock(&kvm_count_lock);
3499 3500 3501 3502

	return r;
}

3503
static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
M
Mike Day 已提交
3504
		      void *v)
3505
{
3506 3507 3508 3509 3510 3511
	/*
	 * 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
3512
	pr_info("kvm: exiting hardware virtualization\n");
3513
	kvm_rebooting = true;
3514
	on_each_cpu(hardware_disable_nolock, NULL, 1);
3515 3516 3517 3518 3519 3520 3521 3522
	return NOTIFY_OK;
}

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

M
Marcelo Tosatti 已提交
3523
static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
3524 3525 3526 3527
{
	int i;

	for (i = 0; i < bus->dev_count; i++) {
3528
		struct kvm_io_device *pos = bus->range[i].dev;
3529 3530 3531

		kvm_iodevice_destructor(pos);
	}
M
Marcelo Tosatti 已提交
3532
	kfree(bus);
3533 3534
}

3535
static inline int kvm_io_bus_cmp(const struct kvm_io_range *r1,
X
Xiubo Li 已提交
3536
				 const struct kvm_io_range *r2)
3537
{
J
Jason Wang 已提交
3538 3539 3540 3541
	gpa_t addr1 = r1->addr;
	gpa_t addr2 = r2->addr;

	if (addr1 < addr2)
3542
		return -1;
J
Jason Wang 已提交
3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554

	/* 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)
3555
		return 1;
J
Jason Wang 已提交
3556

3557 3558 3559
	return 0;
}

3560 3561
static int kvm_io_bus_sort_cmp(const void *p1, const void *p2)
{
3562
	return kvm_io_bus_cmp(p1, p2);
3563 3564
}

G
Geoff Levand 已提交
3565
static int kvm_io_bus_get_first_dev(struct kvm_io_bus *bus,
3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582
			     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;

3583
	while (off > 0 && kvm_io_bus_cmp(&key, &bus->range[off-1]) == 0)
3584 3585 3586 3587 3588
		off--;

	return off;
}

3589
static int __kvm_io_bus_write(struct kvm_vcpu *vcpu, struct kvm_io_bus *bus,
C
Cornelia Huck 已提交
3590 3591 3592 3593 3594 3595 3596 3597 3598
			      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 &&
3599
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
3600
		if (!kvm_iodevice_write(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
3601 3602 3603 3604 3605 3606 3607 3608
					range->len, val))
			return idx;
		idx++;
	}

	return -EOPNOTSUPP;
}

3609
/* kvm_io_bus_write - called under kvm->slots_lock */
3610
int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
3611
		     int len, const void *val)
3612
{
3613
	struct kvm_io_bus *bus;
3614
	struct kvm_io_range range;
C
Cornelia Huck 已提交
3615
	int r;
3616 3617 3618 3619 3620

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
3621

3622
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3623 3624
	if (!bus)
		return -ENOMEM;
3625
	r = __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3626 3627 3628 3629
	return r < 0 ? r : 0;
}

/* kvm_io_bus_write_cookie - called under kvm->slots_lock */
3630 3631
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
3632 3633 3634 3635 3636 3637 3638 3639 3640
{
	struct kvm_io_bus *bus;
	struct kvm_io_range range;

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

3641
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3642 3643
	if (!bus)
		return -ENOMEM;
C
Cornelia Huck 已提交
3644 3645 3646

	/* First try the device referenced by cookie. */
	if ((cookie >= 0) && (cookie < bus->dev_count) &&
3647
	    (kvm_io_bus_cmp(&range, &bus->range[cookie]) == 0))
3648
		if (!kvm_iodevice_write(vcpu, bus->range[cookie].dev, addr, len,
C
Cornelia Huck 已提交
3649 3650 3651 3652 3653 3654 3655
					val))
			return cookie;

	/*
	 * cookie contained garbage; fall back to search and return the
	 * correct cookie value.
	 */
3656
	return __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3657 3658
}

3659 3660
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
3661 3662 3663 3664
{
	int idx;

	idx = kvm_io_bus_get_first_dev(bus, range->addr, range->len);
3665 3666 3667 3668
	if (idx < 0)
		return -EOPNOTSUPP;

	while (idx < bus->dev_count &&
3669
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
3670
		if (!kvm_iodevice_read(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
3671 3672
				       range->len, val))
			return idx;
3673 3674 3675
		idx++;
	}

3676 3677
	return -EOPNOTSUPP;
}
3678
EXPORT_SYMBOL_GPL(kvm_io_bus_write);
3679

3680
/* kvm_io_bus_read - called under kvm->slots_lock */
3681
int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
M
Marcelo Tosatti 已提交
3682
		    int len, void *val)
3683
{
3684
	struct kvm_io_bus *bus;
3685
	struct kvm_io_range range;
C
Cornelia Huck 已提交
3686
	int r;
3687 3688 3689 3690 3691

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
M
Marcelo Tosatti 已提交
3692

3693
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3694 3695
	if (!bus)
		return -ENOMEM;
3696
	r = __kvm_io_bus_read(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3697 3698
	return r < 0 ? r : 0;
}
3699

3700

3701
/* Caller must hold slots_lock. */
3702 3703
int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
			    int len, struct kvm_io_device *dev)
3704
{
3705
	int i;
M
Marcelo Tosatti 已提交
3706
	struct kvm_io_bus *new_bus, *bus;
3707
	struct kvm_io_range range;
3708

3709
	bus = kvm_get_bus(kvm, bus_idx);
3710 3711 3712
	if (!bus)
		return -ENOMEM;

3713 3714
	/* exclude ioeventfd which is limited by maximum fd */
	if (bus->dev_count - bus->ioeventfd_count > NR_IOBUS_DEVS - 1)
3715
		return -ENOSPC;
3716

3717
	new_bus = kmalloc(sizeof(*bus) + ((bus->dev_count + 1) *
3718
			  sizeof(struct kvm_io_range)), GFP_KERNEL);
M
Marcelo Tosatti 已提交
3719 3720
	if (!new_bus)
		return -ENOMEM;
3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736

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

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

	memcpy(new_bus, bus, sizeof(*bus) + i * sizeof(struct kvm_io_range));
	new_bus->dev_count++;
	new_bus->range[i] = range;
	memcpy(new_bus->range + i + 1, bus->range + i,
		(bus->dev_count - i) * sizeof(struct kvm_io_range));
M
Marcelo Tosatti 已提交
3737 3738 3739
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
3740 3741 3742 3743

	return 0;
}

3744
/* Caller must hold slots_lock. */
3745 3746
void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
			       struct kvm_io_device *dev)
3747
{
3748
	int i;
M
Marcelo Tosatti 已提交
3749
	struct kvm_io_bus *new_bus, *bus;
3750

3751
	bus = kvm_get_bus(kvm, bus_idx);
3752
	if (!bus)
3753
		return;
3754

3755 3756
	for (i = 0; i < bus->dev_count; i++)
		if (bus->range[i].dev == dev) {
3757 3758
			break;
		}
M
Marcelo Tosatti 已提交
3759

3760 3761
	if (i == bus->dev_count)
		return;
3762

3763
	new_bus = kmalloc(sizeof(*bus) + ((bus->dev_count - 1) *
3764
			  sizeof(struct kvm_io_range)), GFP_KERNEL);
3765 3766 3767 3768
	if (!new_bus)  {
		pr_err("kvm: failed to shrink bus, removing it completely\n");
		goto broken;
	}
3769 3770 3771 3772 3773

	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
3774

3775
broken:
M
Marcelo Tosatti 已提交
3776 3777 3778
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
3779
	return;
3780 3781
}

3782 3783 3784 3785 3786 3787 3788 3789 3790 3791
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);
3792 3793
	if (!bus)
		goto out_unlock;
3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807

	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);

3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819
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.
	 */
3820
	if (!refcount_inc_not_zero(&stat_data->kvm->users_count))
3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845
		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;

3846
	*val = *(ulong *)((void *)stat_data->kvm + stat_data->offset);
3847 3848 3849 3850

	return 0;
}

3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862
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;
}

3863 3864 3865 3866
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,
3867
				vm_stat_clear_per_vm, "%llu\n");
3868 3869 3870 3871 3872 3873 3874 3875
}

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,
3876
	.llseek  = no_llseek,
3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887
};

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)
3888
		*val += *(u64 *)((void *)vcpu + stat_data->offset);
3889 3890 3891 3892

	return 0;
}

3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907
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;
}

3908 3909 3910 3911
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,
3912
				 vcpu_stat_clear_per_vm, "%llu\n");
3913 3914 3915 3916 3917 3918 3919 3920
}

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,
3921
	.llseek  = no_llseek,
3922 3923 3924 3925 3926 3927 3928
};

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,
};

3929
static int vm_stat_get(void *_offset, u64 *val)
3930 3931 3932
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
3933 3934
	struct kvm_stat_data stat_tmp = {.offset = offset};
	u64 tmp_val;
3935

3936
	*val = 0;
3937
	spin_lock(&kvm_lock);
3938 3939 3940 3941 3942
	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;
	}
3943
	spin_unlock(&kvm_lock);
3944
	return 0;
3945 3946
}

3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966
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");
3967

3968
static int vcpu_stat_get(void *_offset, u64 *val)
A
Avi Kivity 已提交
3969 3970 3971
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
3972 3973
	struct kvm_stat_data stat_tmp = {.offset = offset};
	u64 tmp_val;
A
Avi Kivity 已提交
3974

3975
	*val = 0;
3976
	spin_lock(&kvm_lock);
3977 3978 3979 3980 3981
	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;
	}
3982
	spin_unlock(&kvm_lock);
3983
	return 0;
A
Avi Kivity 已提交
3984 3985
}

3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006
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");
4007

4008
static const struct file_operations *stat_fops[] = {
4009 4010 4011
	[KVM_STAT_VCPU] = &vcpu_stat_fops,
	[KVM_STAT_VM]   = &vm_stat_fops,
};
A
Avi Kivity 已提交
4012

4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038
static void kvm_uevent_notify_change(unsigned int type, struct kvm *kvm)
{
	struct kobj_uevent_env *env;
	unsigned long long created, active;

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

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

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

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

4039
	if (type == KVM_EVENT_CREATE_VM) {
4040
		add_uevent_var(env, "EVENT=create");
4041 4042
		kvm->userspace_pid = task_pid_nr(current);
	} else if (type == KVM_EVENT_DESTROY_VM) {
4043
		add_uevent_var(env, "EVENT=destroy");
4044 4045
	}
	add_uevent_var(env, "PID=%d", kvm->userspace_pid);
4046 4047

	if (kvm->debugfs_dentry) {
4048 4049 4050 4051 4052 4053 4054
		char *tmp, *p = kmalloc(PATH_MAX, GFP_KERNEL);

		if (p) {
			tmp = dentry_path_raw(kvm->debugfs_dentry, p, PATH_MAX);
			if (!IS_ERR(tmp))
				add_uevent_var(env, "STATS_PATH=%s", tmp);
			kfree(p);
4055 4056 4057 4058 4059 4060 4061 4062
		}
	}
	/* no need for checks, since we are adding at most only 5 keys */
	env->envp[env->envp_idx++] = NULL;
	kobject_uevent_env(&kvm_dev.this_device->kobj, KOBJ_CHANGE, env->envp);
	kfree(env);
}

4063
static void kvm_init_debug(void)
A
Avi Kivity 已提交
4064 4065 4066
{
	struct kvm_stats_debugfs_item *p;

4067
	kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
4068

4069 4070
	kvm_debugfs_num_entries = 0;
	for (p = debugfs_entries; p->name; ++p, kvm_debugfs_num_entries++) {
4071 4072 4073
		debugfs_create_file(p->name, 0644, kvm_debugfs_dir,
				    (void *)(long)p->offset,
				    stat_fops[p->kind]);
4074
	}
A
Avi Kivity 已提交
4075 4076
}

4077
static int kvm_suspend(void)
4078
{
4079
	if (kvm_usage_count)
4080
		hardware_disable_nolock(NULL);
4081 4082 4083
	return 0;
}

4084
static void kvm_resume(void)
4085
{
4086
	if (kvm_usage_count) {
4087
		WARN_ON(raw_spin_is_locked(&kvm_count_lock));
4088
		hardware_enable_nolock(NULL);
4089
	}
4090 4091
}

4092
static struct syscore_ops kvm_syscore_ops = {
4093 4094 4095 4096
	.suspend = kvm_suspend,
	.resume = kvm_resume,
};

4097 4098 4099 4100 4101 4102 4103 4104 4105
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);
4106

4107 4108
	if (vcpu->preempted)
		vcpu->preempted = false;
4109

R
Radim Krčmář 已提交
4110 4111
	kvm_arch_sched_in(vcpu, cpu);

4112
	kvm_arch_vcpu_load(vcpu, cpu);
4113 4114 4115 4116 4117 4118 4119
}

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

4120 4121
	if (current->state == TASK_RUNNING)
		vcpu->preempted = true;
4122
	kvm_arch_vcpu_put(vcpu);
4123 4124
}

4125
int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
4126
		  struct module *module)
A
Avi Kivity 已提交
4127 4128
{
	int r;
Y
Yang, Sheng 已提交
4129
	int cpu;
A
Avi Kivity 已提交
4130

4131 4132
	r = kvm_arch_init(opaque);
	if (r)
4133
		goto out_fail;
4134

4135 4136 4137 4138
	/*
	 * 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
4139 4140
	 * kvm_arch_init must be called before kvm_irqfd_init to avoid creating
	 * conflicts in case kvm is already setup for another implementation.
4141
	 */
P
Paolo Bonzini 已提交
4142 4143 4144
	r = kvm_irqfd_init();
	if (r)
		goto out_irqfd;
4145

4146
	if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
4147 4148 4149 4150
		r = -ENOMEM;
		goto out_free_0;
	}

4151
	r = kvm_arch_hardware_setup();
A
Avi Kivity 已提交
4152
	if (r < 0)
4153
		goto out_free_0a;
A
Avi Kivity 已提交
4154

Y
Yang, Sheng 已提交
4155 4156
	for_each_online_cpu(cpu) {
		smp_call_function_single(cpu,
4157
				kvm_arch_check_processor_compat,
4158
				&r, 1);
Y
Yang, Sheng 已提交
4159
		if (r < 0)
4160
			goto out_free_1;
Y
Yang, Sheng 已提交
4161 4162
	}

T
Thomas Gleixner 已提交
4163
	r = cpuhp_setup_state_nocalls(CPUHP_AP_KVM_STARTING, "kvm/cpu:starting",
4164
				      kvm_starting_cpu, kvm_dying_cpu);
A
Avi Kivity 已提交
4165
	if (r)
4166
		goto out_free_2;
A
Avi Kivity 已提交
4167 4168
	register_reboot_notifier(&kvm_reboot_notifier);

4169
	/* A kmem cache lets us meet the alignment requirements of fx_save. */
4170 4171
	if (!vcpu_align)
		vcpu_align = __alignof__(struct kvm_vcpu);
4172 4173 4174 4175 4176 4177
	kvm_vcpu_cache =
		kmem_cache_create_usercopy("kvm_vcpu", vcpu_size, vcpu_align,
					   SLAB_ACCOUNT,
					   offsetof(struct kvm_vcpu, arch),
					   sizeof_field(struct kvm_vcpu, arch),
					   NULL);
4178 4179
	if (!kvm_vcpu_cache) {
		r = -ENOMEM;
4180
		goto out_free_3;
4181 4182
	}

4183 4184 4185 4186
	r = kvm_async_pf_init();
	if (r)
		goto out_free;

A
Avi Kivity 已提交
4187
	kvm_chardev_ops.owner = module;
4188 4189
	kvm_vm_fops.owner = module;
	kvm_vcpu_fops.owner = module;
A
Avi Kivity 已提交
4190 4191 4192

	r = misc_register(&kvm_dev);
	if (r) {
X
Xiubo Li 已提交
4193
		pr_err("kvm: misc device register failed\n");
4194
		goto out_unreg;
A
Avi Kivity 已提交
4195 4196
	}

4197 4198
	register_syscore_ops(&kvm_syscore_ops);

4199 4200 4201
	kvm_preempt_ops.sched_in = kvm_sched_in;
	kvm_preempt_ops.sched_out = kvm_sched_out;

4202
	kvm_init_debug();
4203

P
Paolo Bonzini 已提交
4204 4205 4206
	r = kvm_vfio_ops_init();
	WARN_ON(r);

4207
	return 0;
A
Avi Kivity 已提交
4208

4209 4210
out_unreg:
	kvm_async_pf_deinit();
A
Avi Kivity 已提交
4211
out_free:
4212
	kmem_cache_destroy(kvm_vcpu_cache);
4213
out_free_3:
A
Avi Kivity 已提交
4214
	unregister_reboot_notifier(&kvm_reboot_notifier);
4215
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4216 4217
out_free_2:
out_free_1:
4218
	kvm_arch_hardware_unsetup();
4219 4220
out_free_0a:
	free_cpumask_var(cpus_hardware_enabled);
4221
out_free_0:
4222
	kvm_irqfd_exit();
P
Paolo Bonzini 已提交
4223
out_irqfd:
4224 4225
	kvm_arch_exit();
out_fail:
A
Avi Kivity 已提交
4226 4227
	return r;
}
4228
EXPORT_SYMBOL_GPL(kvm_init);
A
Avi Kivity 已提交
4229

4230
void kvm_exit(void)
A
Avi Kivity 已提交
4231
{
4232
	debugfs_remove_recursive(kvm_debugfs_dir);
A
Avi Kivity 已提交
4233
	misc_deregister(&kvm_dev);
4234
	kmem_cache_destroy(kvm_vcpu_cache);
4235
	kvm_async_pf_deinit();
4236
	unregister_syscore_ops(&kvm_syscore_ops);
A
Avi Kivity 已提交
4237
	unregister_reboot_notifier(&kvm_reboot_notifier);
4238
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4239
	on_each_cpu(hardware_disable_nolock, NULL, 1);
4240
	kvm_arch_hardware_unsetup();
4241
	kvm_arch_exit();
4242
	kvm_irqfd_exit();
4243
	free_cpumask_var(cpus_hardware_enabled);
4244
	kvm_vfio_ops_exit();
A
Avi Kivity 已提交
4245
}
4246
EXPORT_SYMBOL_GPL(kvm_exit);