kvm_main.c 99.5 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 85 86 87 88
/* The start value to grow halt_poll_ns from */
unsigned int halt_poll_ns_grow_start = 10000; /* 10us */
module_param(halt_poll_ns_grow_start, uint, 0644);
EXPORT_SYMBOL_GPL(halt_poll_ns_grow_start);

W
Wanpeng Li 已提交
89
/* Default resets per-vcpu halt_poll_ns . */
90
unsigned int halt_poll_ns_shrink;
91
module_param(halt_poll_ns_shrink, uint, 0644);
92
EXPORT_SYMBOL_GPL(halt_poll_ns_shrink);
W
Wanpeng Li 已提交
93

94 95 96
/*
 * Ordering of locks:
 *
97
 *	kvm->lock --> kvm->slots_lock --> kvm->irq_lock
98 99
 */

100
DEFINE_SPINLOCK(kvm_lock);
101
static DEFINE_RAW_SPINLOCK(kvm_count_lock);
102
LIST_HEAD(vm_list);
103

104
static cpumask_var_t cpus_hardware_enabled;
105
static int kvm_usage_count;
106
static atomic_t hardware_enable_failed;
107

108 109
struct kmem_cache *kvm_vcpu_cache;
EXPORT_SYMBOL_GPL(kvm_vcpu_cache);
A
Avi Kivity 已提交
110

111 112
static __read_mostly struct preempt_ops kvm_preempt_ops;

113
struct dentry *kvm_debugfs_dir;
114
EXPORT_SYMBOL_GPL(kvm_debugfs_dir);
A
Avi Kivity 已提交
115

116 117 118
static int kvm_debugfs_num_entries;
static const struct file_operations *stat_fops_per_vm[];

A
Avi Kivity 已提交
119 120
static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl,
			   unsigned long arg);
121
#ifdef CONFIG_KVM_COMPAT
122 123
static long kvm_vcpu_compat_ioctl(struct file *file, unsigned int ioctl,
				  unsigned long arg);
124 125 126 127 128
#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
129
#endif
130 131
static int hardware_enable_all(void);
static void hardware_disable_all(void);
A
Avi Kivity 已提交
132

M
Marcelo Tosatti 已提交
133
static void kvm_io_bus_destroy(struct kvm_io_bus *bus);
134

135
static void mark_page_dirty_in_slot(struct kvm_memory_slot *memslot, gfn_t gfn);
M
Marcelo Tosatti 已提交
136

137
__visible bool kvm_rebooting;
138
EXPORT_SYMBOL_GPL(kvm_rebooting);
139

140 141
static bool largepages_enabled = true;

142 143 144 145 146 147
#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;

148 149
__weak int kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
		unsigned long start, unsigned long end, bool blockable)
150
{
151
	return 0;
152 153
}

D
Dan Williams 已提交
154
bool kvm_is_reserved_pfn(kvm_pfn_t pfn)
B
Ben-Ami Yassour 已提交
155
{
156
	if (pfn_valid(pfn))
157
		return PageReserved(pfn_to_page(pfn));
B
Ben-Ami Yassour 已提交
158 159 160 161

	return true;
}

A
Avi Kivity 已提交
162 163 164
/*
 * Switches to specified vcpu, until a matching vcpu_put()
 */
165
void vcpu_load(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
166
{
167
	int cpu = get_cpu();
168
	preempt_notifier_register(&vcpu->preempt_notifier);
169
	kvm_arch_vcpu_load(vcpu, cpu);
170
	put_cpu();
A
Avi Kivity 已提交
171
}
172
EXPORT_SYMBOL_GPL(vcpu_load);
A
Avi Kivity 已提交
173

174
void vcpu_put(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
175
{
176
	preempt_disable();
177
	kvm_arch_vcpu_put(vcpu);
178 179
	preempt_notifier_unregister(&vcpu->preempt_notifier);
	preempt_enable();
A
Avi Kivity 已提交
180
}
181
EXPORT_SYMBOL_GPL(vcpu_put);
A
Avi Kivity 已提交
182

183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200
/* 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;
}

201 202 203 204
static void ack_flush(void *_completed)
{
}

205 206 207 208 209 210 211 212 213 214 215 216
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;
}

217 218
bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
				 unsigned long *vcpu_bitmap, cpumask_var_t tmp)
219
{
220
	int i, cpu, me;
221
	struct kvm_vcpu *vcpu;
222
	bool called;
223

224
	me = get_cpu();
225

226
	kvm_for_each_vcpu(i, vcpu, kvm) {
227
		if (vcpu_bitmap && !test_bit(i, vcpu_bitmap))
228 229
			continue;

230
		kvm_make_request(req, vcpu);
231
		cpu = vcpu->cpu;
232

233 234
		if (!(req & KVM_REQUEST_NO_WAKEUP) && kvm_vcpu_wake_up(vcpu))
			continue;
235

236
		if (tmp != NULL && cpu != -1 && cpu != me &&
237
		    kvm_request_needs_ipi(vcpu, req))
238
			__cpumask_set_cpu(cpu, tmp);
239
	}
240 241

	called = kvm_kick_many_cpus(tmp, !!(req & KVM_REQUEST_WAIT));
242
	put_cpu();
243 244 245 246 247 248 249 250 251 252 253

	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);

254
	called = kvm_make_vcpus_request_mask(kvm, req, NULL, cpus);
255

256
	free_cpumask_var(cpus);
257
	return called;
258 259
}

260
#ifndef CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALL
261
void kvm_flush_remote_tlbs(struct kvm *kvm)
262
{
263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279
	/*
	 * 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.
	 */
280 281
	if (!kvm_arch_flush_remote_tlb(kvm)
	    || kvm_make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH))
282
		++kvm->stat.remote_tlb_flush;
283
	cmpxchg(&kvm->tlbs_dirty, dirty_count, 0);
284
}
285
EXPORT_SYMBOL_GPL(kvm_flush_remote_tlbs);
286
#endif
287

288 289
void kvm_reload_remote_mmus(struct kvm *kvm)
{
290
	kvm_make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD);
291
}
292

R
Rusty Russell 已提交
293 294 295 296 297 298 299 300 301
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;
302
	vcpu->pid = NULL;
303
	init_swait_queue_head(&vcpu->wq);
304
	kvm_async_pf_vcpu_init(vcpu);
R
Rusty Russell 已提交
305

306 307 308
	vcpu->pre_pcpu = -1;
	INIT_LIST_HEAD(&vcpu->blocked_vcpu_list);

R
Rusty Russell 已提交
309 310 311 312 313 314 315
	page = alloc_page(GFP_KERNEL | __GFP_ZERO);
	if (!page) {
		r = -ENOMEM;
		goto fail;
	}
	vcpu->run = page_address(page);

316 317
	kvm_vcpu_set_in_spin_loop(vcpu, false);
	kvm_vcpu_set_dy_eligible(vcpu, false);
318
	vcpu->preempted = false;
319

320
	r = kvm_arch_vcpu_init(vcpu);
R
Rusty Russell 已提交
321
	if (r < 0)
322
		goto fail_free_run;
R
Rusty Russell 已提交
323 324 325 326 327
	return 0;

fail_free_run:
	free_page((unsigned long)vcpu->run);
fail:
328
	return r;
R
Rusty Russell 已提交
329 330 331 332 333
}
EXPORT_SYMBOL_GPL(kvm_vcpu_init);

void kvm_vcpu_uninit(struct kvm_vcpu *vcpu)
{
334 335 336 337 338 339
	/*
	 * 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));
340
	kvm_arch_vcpu_uninit(vcpu);
R
Rusty Russell 已提交
341 342 343 344
	free_page((unsigned long)vcpu->run);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_uninit);

345 346 347 348 349 350
#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);
}

351 352 353 354 355 356
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);
357
	int idx;
358

359
	idx = srcu_read_lock(&kvm->srcu);
360 361
	spin_lock(&kvm->mmu_lock);
	kvm->mmu_notifier_seq++;
362 363 364 365

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

366
	spin_unlock(&kvm->mmu_lock);
367
	srcu_read_unlock(&kvm->srcu, idx);
368 369
}

370
static int kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn,
371
					const struct mmu_notifier_range *range)
372 373
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
374
	int need_tlb_flush = 0, idx;
375
	int ret;
376

377
	idx = srcu_read_lock(&kvm->srcu);
378 379 380 381 382 383 384
	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++;
385
	need_tlb_flush = kvm_unmap_hva_range(kvm, range->start, range->end);
386
	need_tlb_flush |= kvm->tlbs_dirty;
387 388 389
	/* we've to flush the tlb before the pages can be freed */
	if (need_tlb_flush)
		kvm_flush_remote_tlbs(kvm);
390 391

	spin_unlock(&kvm->mmu_lock);
392

393 394
	ret = kvm_arch_mmu_notifier_invalidate_range(kvm, range->start,
					range->end, range->blockable);
395

396
	srcu_read_unlock(&kvm->srcu, idx);
397 398

	return ret;
399 400 401
}

static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
402
					const struct mmu_notifier_range *range)
403 404 405 406 407 408 409 410 411 412
{
	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++;
413
	smp_wmb();
414 415
	/*
	 * The above sequence increase must be visible before the
416 417
	 * below count decrease, which is ensured by the smp_wmb above
	 * in conjunction with the smp_rmb in mmu_notifier_retry().
418 419 420 421 422 423 424 425 426
	 */
	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
427 428
					      unsigned long start,
					      unsigned long end)
429 430
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
431
	int young, idx;
432

433
	idx = srcu_read_lock(&kvm->srcu);
434 435
	spin_lock(&kvm->mmu_lock);

A
Andres Lagar-Cavilla 已提交
436
	young = kvm_age_hva(kvm, start, end);
437 438 439
	if (young)
		kvm_flush_remote_tlbs(kvm);

440 441 442
	spin_unlock(&kvm->mmu_lock);
	srcu_read_unlock(&kvm->srcu, idx);

443 444 445
	return young;
}

446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475
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
476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491
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;
}

492 493 494 495
static void kvm_mmu_notifier_release(struct mmu_notifier *mn,
				     struct mm_struct *mm)
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
496 497 498
	int idx;

	idx = srcu_read_lock(&kvm->srcu);
499
	kvm_arch_flush_shadow_all(kvm);
500
	srcu_read_unlock(&kvm->srcu, idx);
501 502
}

503 504 505 506
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,
507
	.clear_young		= kvm_mmu_notifier_clear_young,
A
Andrea Arcangeli 已提交
508
	.test_young		= kvm_mmu_notifier_test_young,
509
	.change_pte		= kvm_mmu_notifier_change_pte,
510
	.release		= kvm_mmu_notifier_release,
511
};
512 513 514 515 516 517 518 519 520 521 522 523 524 525

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;
}

526 527
#endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */

528
static struct kvm_memslots *kvm_alloc_memslots(void)
529 530
{
	int i;
531
	struct kvm_memslots *slots;
532

533
	slots = kvzalloc(sizeof(struct kvm_memslots), GFP_KERNEL_ACCOUNT);
534 535 536
	if (!slots)
		return NULL;

537
	for (i = 0; i < KVM_MEM_SLOTS_NUM; i++)
538
		slots->id_to_index[i] = slots->memslots[i].id = i;
539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576

	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);
577 578
}

579 580 581 582 583 584 585 586 587
static void kvm_destroy_vm_debugfs(struct kvm *kvm)
{
	int i;

	if (!kvm->debugfs_dentry)
		return;

	debugfs_remove_recursive(kvm->debugfs_dentry);

588 589 590 591 592
	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);
	}
593 594 595 596 597 598 599 600 601 602 603 604
}

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);
605
	kvm->debugfs_dentry = debugfs_create_dir(dir_name, kvm_debugfs_dir);
606 607 608

	kvm->debugfs_stat_data = kcalloc(kvm_debugfs_num_entries,
					 sizeof(*kvm->debugfs_stat_data),
609
					 GFP_KERNEL_ACCOUNT);
610 611 612 613
	if (!kvm->debugfs_stat_data)
		return -ENOMEM;

	for (p = debugfs_entries; p->name; p++) {
614
		stat_data = kzalloc(sizeof(*stat_data), GFP_KERNEL_ACCOUNT);
615 616 617 618 619 620
		if (!stat_data)
			return -ENOMEM;

		stat_data->kvm = kvm;
		stat_data->offset = p->offset;
		kvm->debugfs_stat_data[p - debugfs_entries] = stat_data;
621 622
		debugfs_create_file(p->name, 0644, kvm->debugfs_dentry,
				    stat_data, stat_fops_per_vm[p->kind]);
623 624 625 626
	}
	return 0;
}

627
static struct kvm *kvm_create_vm(unsigned long type)
A
Avi Kivity 已提交
628
{
629 630
	int r, i;
	struct kvm *kvm = kvm_arch_alloc_vm();
A
Avi Kivity 已提交
631

632 633 634
	if (!kvm)
		return ERR_PTR(-ENOMEM);

635
	spin_lock_init(&kvm->mmu_lock);
V
Vegard Nossum 已提交
636
	mmgrab(current->mm);
637 638 639 640 641
	kvm->mm = current->mm;
	kvm_eventfd_init(kvm);
	mutex_init(&kvm->lock);
	mutex_init(&kvm->irq_lock);
	mutex_init(&kvm->slots_lock);
642
	refcount_set(&kvm->users_count, 1);
643 644
	INIT_LIST_HEAD(&kvm->devices);

645
	r = kvm_arch_init_vm(kvm, type);
646
	if (r)
647
		goto out_err_no_disable;
648 649 650

	r = hardware_enable_all();
	if (r)
651
		goto out_err_no_disable;
652

653
#ifdef CONFIG_HAVE_KVM_IRQFD
654
	INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list);
655
#endif
A
Avi Kivity 已提交
656

657 658
	BUILD_BUG_ON(KVM_MEM_SLOTS_NUM > SHRT_MAX);

659
	r = -ENOMEM;
660
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
661 662
		struct kvm_memslots *slots = kvm_alloc_memslots();
		if (!slots)
663
			goto out_err_no_srcu;
664
		/* Generations must be different for each address space. */
665
		slots->generation = i;
666
		rcu_assign_pointer(kvm->memslots[i], slots);
667
	}
668

669
	if (init_srcu_struct(&kvm->srcu))
670 671 672
		goto out_err_no_srcu;
	if (init_srcu_struct(&kvm->irq_srcu))
		goto out_err_no_irq_srcu;
M
Marcelo Tosatti 已提交
673
	for (i = 0; i < KVM_NR_BUSES; i++) {
674
		rcu_assign_pointer(kvm->buses[i],
675
			kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL_ACCOUNT));
676
		if (!kvm->buses[i])
M
Marcelo Tosatti 已提交
677 678
			goto out_err;
	}
679

680 681 682 683
	r = kvm_init_mmu_notifier(kvm);
	if (r)
		goto out_err;

684
	spin_lock(&kvm_lock);
685
	list_add(&kvm->vm_list, &vm_list);
686
	spin_unlock(&kvm_lock);
687

688 689
	preempt_notifier_inc();

690
	return kvm;
691 692

out_err:
693 694
	cleanup_srcu_struct(&kvm->irq_srcu);
out_err_no_irq_srcu:
695
	cleanup_srcu_struct(&kvm->srcu);
696
out_err_no_srcu:
697
	hardware_disable_all();
698
out_err_no_disable:
699
	refcount_set(&kvm->users_count, 0);
M
Marcelo Tosatti 已提交
700
	for (i = 0; i < KVM_NR_BUSES; i++)
701
		kfree(kvm_get_bus(kvm, i));
702
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
703
		kvm_free_memslots(kvm, __kvm_memslots(kvm, i));
704
	kvm_arch_free_vm(kvm);
705
	mmdrop(current->mm);
706
	return ERR_PTR(r);
707 708
}

709 710
static void kvm_destroy_devices(struct kvm *kvm)
{
G
Geliang Tang 已提交
711
	struct kvm_device *dev, *tmp;
712

713 714 715 716 717
	/*
	 * 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
718 719
	list_for_each_entry_safe(dev, tmp, &kvm->devices, vm_node) {
		list_del(&dev->vm_node);
720 721 722 723
		dev->ops->destroy(dev);
	}
}

724 725
static void kvm_destroy_vm(struct kvm *kvm)
{
M
Marcelo Tosatti 已提交
726
	int i;
727 728
	struct mm_struct *mm = kvm->mm;

729
	kvm_uevent_notify_change(KVM_EVENT_DESTROY_VM, kvm);
730
	kvm_destroy_vm_debugfs(kvm);
731
	kvm_arch_sync_events(kvm);
732
	spin_lock(&kvm_lock);
733
	list_del(&kvm->vm_list);
734
	spin_unlock(&kvm_lock);
735
	kvm_free_irq_routing(kvm);
736
	for (i = 0; i < KVM_NR_BUSES; i++) {
737
		struct kvm_io_bus *bus = kvm_get_bus(kvm, i);
738 739 740

		if (bus)
			kvm_io_bus_destroy(bus);
741 742
		kvm->buses[i] = NULL;
	}
743
	kvm_coalesced_mmio_free(kvm);
744 745
#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
	mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
746
#else
747
	kvm_arch_flush_shadow_all(kvm);
748
#endif
749
	kvm_arch_destroy_vm(kvm);
750
	kvm_destroy_devices(kvm);
751
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
752
		kvm_free_memslots(kvm, __kvm_memslots(kvm, i));
753
	cleanup_srcu_struct(&kvm->irq_srcu);
754 755
	cleanup_srcu_struct(&kvm->srcu);
	kvm_arch_free_vm(kvm);
756
	preempt_notifier_dec();
757
	hardware_disable_all();
758
	mmdrop(mm);
759 760
}

I
Izik Eidus 已提交
761 762
void kvm_get_kvm(struct kvm *kvm)
{
763
	refcount_inc(&kvm->users_count);
I
Izik Eidus 已提交
764 765 766 767 768
}
EXPORT_SYMBOL_GPL(kvm_get_kvm);

void kvm_put_kvm(struct kvm *kvm)
{
769
	if (refcount_dec_and_test(&kvm->users_count))
I
Izik Eidus 已提交
770 771 772 773 774
		kvm_destroy_vm(kvm);
}
EXPORT_SYMBOL_GPL(kvm_put_kvm);


775 776 777 778
static int kvm_vm_release(struct inode *inode, struct file *filp)
{
	struct kvm *kvm = filp->private_data;

G
Gregory Haskins 已提交
779 780
	kvm_irqfd_release(kvm);

I
Izik Eidus 已提交
781
	kvm_put_kvm(kvm);
A
Avi Kivity 已提交
782 783 784
	return 0;
}

785 786
/*
 * Allocation size is twice as large as the actual dirty bitmap size.
787
 * See x86's kvm_vm_ioctl_get_dirty_log() why this is needed.
788
 */
789 790
static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot)
{
791
	unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot);
792

793
	memslot->dirty_bitmap = kvzalloc(dirty_bytes, GFP_KERNEL_ACCOUNT);
794 795 796 797 798 799
	if (!memslot->dirty_bitmap)
		return -ENOMEM;

	return 0;
}

800
/*
801 802 803 804
 * 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.
805
 */
806
static void update_memslots(struct kvm_memslots *slots,
807 808
			    struct kvm_memory_slot *new,
			    enum kvm_mr_change change)
809
{
810 811
	int id = new->id;
	int i = slots->id_to_index[id];
812
	struct kvm_memory_slot *mslots = slots->memslots;
813

814
	WARN_ON(mslots[i].id != id);
815 816 817 818 819 820 821 822 823 824 825
	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;
826
	}
827

828
	while (i < KVM_MEM_SLOTS_NUM - 1 &&
829 830 831
	       new->base_gfn <= mslots[i + 1].base_gfn) {
		if (!mslots[i + 1].npages)
			break;
832 833 834 835
		mslots[i] = mslots[i + 1];
		slots->id_to_index[mslots[i].id] = i;
		i++;
	}
836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852

	/*
	 * 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--;
		}
853 854
	} else
		WARN_ON_ONCE(i != slots->used_slots);
855

856 857
	mslots[i] = *new;
	slots->id_to_index[mslots[i].id] = i;
858 859
}

860
static int check_memory_region_flags(const struct kvm_userspace_memory_region *mem)
861
{
X
Xiao Guangrong 已提交
862 863
	u32 valid_flags = KVM_MEM_LOG_DIRTY_PAGES;

864
#ifdef __KVM_HAVE_READONLY_MEM
X
Xiao Guangrong 已提交
865 866 867 868
	valid_flags |= KVM_MEM_READONLY;
#endif

	if (mem->flags & ~valid_flags)
869 870 871 872 873
		return -EINVAL;

	return 0;
}

874
static struct kvm_memslots *install_new_memslots(struct kvm *kvm,
875
		int as_id, struct kvm_memslots *slots)
876
{
877
	struct kvm_memslots *old_memslots = __kvm_memslots(kvm, as_id);
878
	u64 gen = old_memslots->generation;
879

880 881
	WARN_ON(gen & KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS);
	slots->generation = gen | KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS;
882

883
	rcu_assign_pointer(kvm->memslots[as_id], slots);
884
	synchronize_srcu_expedited(&kvm->srcu);
885

886
	/*
887 888 889 890 891 892 893 894
	 * Increment the new memslot generation a second time, dropping the
	 * update in-progress flag and incrementing then generation based on
	 * the number of address spaces.  This provides a unique and easily
	 * identifiable generation number while the memslots are in flux.
	 */
	gen = slots->generation & ~KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS;

	/*
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
898 899
	 * space 0 will use generations 0, 2, 4, ... while address space 1 will
	 * use generations 1, 3, 5, ...
900
	 */
901
	gen += KVM_ADDRESS_SPACE_NUM;
902

903 904 905
	kvm_arch_memslots_updated(kvm, gen);

	slots->generation = gen;
906 907

	return old_memslots;
908 909
}

A
Avi Kivity 已提交
910 911 912 913 914
/*
 * Allocate some memory and give it an address in the guest physical address
 * space.
 *
 * Discontiguous memory is allowed, mostly for framebuffers.
915
 *
916
 * Must be called holding kvm->slots_lock for write.
A
Avi Kivity 已提交
917
 */
918
int __kvm_set_memory_region(struct kvm *kvm,
919
			    const struct kvm_userspace_memory_region *mem)
A
Avi Kivity 已提交
920
{
921
	int r;
A
Avi Kivity 已提交
922
	gfn_t base_gfn;
923
	unsigned long npages;
924
	struct kvm_memory_slot *slot;
A
Avi Kivity 已提交
925
	struct kvm_memory_slot old, new;
926
	struct kvm_memslots *slots = NULL, *old_memslots;
927
	int as_id, id;
928
	enum kvm_mr_change change;
A
Avi Kivity 已提交
929

930 931 932 933
	r = check_memory_region_flags(mem);
	if (r)
		goto out;

A
Avi Kivity 已提交
934
	r = -EINVAL;
935 936 937
	as_id = mem->slot >> 16;
	id = (u16)mem->slot;

A
Avi Kivity 已提交
938 939 940 941 942
	/* General sanity checks */
	if (mem->memory_size & (PAGE_SIZE - 1))
		goto out;
	if (mem->guest_phys_addr & (PAGE_SIZE - 1))
		goto out;
943
	/* We can read the guest memory with __xxx_user() later on. */
944
	if ((id < KVM_USER_MEM_SLOTS) &&
945
	    ((mem->userspace_addr & (PAGE_SIZE - 1)) ||
946
	     !access_ok((void __user *)(unsigned long)mem->userspace_addr,
947
			mem->memory_size)))
948
		goto out;
949
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_MEM_SLOTS_NUM)
A
Avi Kivity 已提交
950 951 952 953
		goto out;
	if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
		goto out;

954
	slot = id_to_memslot(__kvm_memslots(kvm, as_id), id);
A
Avi Kivity 已提交
955 956 957
	base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
	npages = mem->memory_size >> PAGE_SHIFT;

958 959 960
	if (npages > KVM_MEM_MAX_NR_PAGES)
		goto out;

961
	new = old = *slot;
A
Avi Kivity 已提交
962

963
	new.id = id;
A
Avi Kivity 已提交
964 965 966 967
	new.base_gfn = base_gfn;
	new.npages = npages;
	new.flags = mem->flags;

968 969 970 971 972
	if (npages) {
		if (!old.npages)
			change = KVM_MR_CREATE;
		else { /* Modify an existing slot. */
			if ((mem->userspace_addr != old.userspace_addr) ||
973 974
			    (npages != old.npages) ||
			    ((new.flags ^ old.flags) & KVM_MEM_READONLY))
975 976 977 978 979 980 981 982 983 984 985
				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;
			}
		}
986 987 988 989
	} else {
		if (!old.npages)
			goto out;

990
		change = KVM_MR_DELETE;
991 992 993
		new.base_gfn = 0;
		new.flags = 0;
	}
A
Avi Kivity 已提交
994

995
	if ((change == KVM_MR_CREATE) || (change == KVM_MR_MOVE)) {
996 997
		/* Check for overlaps */
		r = -EEXIST;
998
		kvm_for_each_memslot(slot, __kvm_memslots(kvm, as_id)) {
999
			if (slot->id == id)
1000 1001 1002 1003 1004
				continue;
			if (!((base_gfn + npages <= slot->base_gfn) ||
			      (base_gfn >= slot->base_gfn + slot->npages)))
				goto out;
		}
A
Avi Kivity 已提交
1005 1006 1007 1008
	}

	/* Free page dirty bitmap if unneeded */
	if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
A
Al Viro 已提交
1009
		new.dirty_bitmap = NULL;
A
Avi Kivity 已提交
1010 1011

	r = -ENOMEM;
1012
	if (change == KVM_MR_CREATE) {
1013
		new.userspace_addr = mem->userspace_addr;
1014

1015
		if (kvm_arch_create_memslot(kvm, &new, npages))
1016
			goto out_free;
A
Avi Kivity 已提交
1017
	}
1018

A
Avi Kivity 已提交
1019 1020
	/* Allocate page dirty bitmap if needed */
	if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
1021
		if (kvm_create_dirty_bitmap(&new) < 0)
1022
			goto out_free;
A
Avi Kivity 已提交
1023 1024
	}

1025
	slots = kvzalloc(sizeof(struct kvm_memslots), GFP_KERNEL_ACCOUNT);
1026 1027
	if (!slots)
		goto out_free;
1028
	memcpy(slots, __kvm_memslots(kvm, as_id), sizeof(struct kvm_memslots));
1029

1030
	if ((change == KVM_MR_DELETE) || (change == KVM_MR_MOVE)) {
1031
		slot = id_to_memslot(slots, id);
1032 1033
		slot->flags |= KVM_MEMSLOT_INVALID;

1034
		old_memslots = install_new_memslots(kvm, as_id, slots);
1035

1036 1037
		/* From this point no new shadow pages pointing to a deleted,
		 * or moved, memslot will be created.
1038 1039
		 *
		 * validation of sp->gfn happens in:
1040 1041
		 *	- gfn_to_hva (kvm_read_guest, gfn_to_pfn)
		 *	- kvm_is_visible_gfn (mmu_check_roots)
1042
		 */
1043
		kvm_arch_flush_shadow_memslot(kvm, slot);
1044 1045 1046 1047 1048 1049

		/*
		 * 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.
		 */
1050
		slots = old_memslots;
1051
	}
1052

1053
	r = kvm_arch_prepare_memory_region(kvm, &new, mem, change);
1054
	if (r)
1055
		goto out_slots;
1056

1057
	/* actual memory is freed via old in kvm_free_memslot below */
1058
	if (change == KVM_MR_DELETE) {
1059
		new.dirty_bitmap = NULL;
1060
		memset(&new.arch, 0, sizeof(new.arch));
1061 1062
	}

1063
	update_memslots(slots, &new, change);
1064
	old_memslots = install_new_memslots(kvm, as_id, slots);
1065

1066
	kvm_arch_commit_memory_region(kvm, mem, &old, &new, change);
1067

1068
	kvm_free_memslot(kvm, &old, &new);
1069
	kvfree(old_memslots);
A
Avi Kivity 已提交
1070 1071
	return 0;

1072
out_slots:
1073
	kvfree(slots);
1074
out_free:
1075
	kvm_free_memslot(kvm, &new, &old);
A
Avi Kivity 已提交
1076 1077
out:
	return r;
1078
}
1079 1080 1081
EXPORT_SYMBOL_GPL(__kvm_set_memory_region);

int kvm_set_memory_region(struct kvm *kvm,
1082
			  const struct kvm_userspace_memory_region *mem)
1083 1084 1085
{
	int r;

1086
	mutex_lock(&kvm->slots_lock);
1087
	r = __kvm_set_memory_region(kvm, mem);
1088
	mutex_unlock(&kvm->slots_lock);
1089 1090
	return r;
}
1091 1092
EXPORT_SYMBOL_GPL(kvm_set_memory_region);

1093 1094
static int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
					  struct kvm_userspace_memory_region *mem)
1095
{
1096
	if ((u16)mem->slot >= KVM_USER_MEM_SLOTS)
1097
		return -EINVAL;
1098

1099
	return kvm_set_memory_region(kvm, mem);
A
Avi Kivity 已提交
1100 1101
}

1102 1103
int kvm_get_dirty_log(struct kvm *kvm,
			struct kvm_dirty_log *log, int *is_dirty)
A
Avi Kivity 已提交
1104
{
1105
	struct kvm_memslots *slots;
A
Avi Kivity 已提交
1106
	struct kvm_memory_slot *memslot;
1107
	int i, as_id, id;
1108
	unsigned long n;
A
Avi Kivity 已提交
1109 1110
	unsigned long any = 0;

1111 1112 1113
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1114
		return -EINVAL;
A
Avi Kivity 已提交
1115

1116 1117
	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
A
Avi Kivity 已提交
1118
	if (!memslot->dirty_bitmap)
1119
		return -ENOENT;
A
Avi Kivity 已提交
1120

1121
	n = kvm_dirty_bitmap_bytes(memslot);
A
Avi Kivity 已提交
1122

1123
	for (i = 0; !any && i < n/sizeof(long); ++i)
A
Avi Kivity 已提交
1124 1125 1126
		any = memslot->dirty_bitmap[i];

	if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
1127
		return -EFAULT;
A
Avi Kivity 已提交
1128

1129 1130
	if (any)
		*is_dirty = 1;
1131
	return 0;
A
Avi Kivity 已提交
1132
}
1133
EXPORT_SYMBOL_GPL(kvm_get_dirty_log);
A
Avi Kivity 已提交
1134

1135 1136 1137
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
/**
 * kvm_get_dirty_log_protect - get a snapshot of dirty pages, and if any pages
1138
 *	and reenable dirty page tracking for the corresponding pages.
1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158
 * @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,
1159
			struct kvm_dirty_log *log, bool *flush)
1160
{
1161
	struct kvm_memslots *slots;
1162
	struct kvm_memory_slot *memslot;
1163
	int i, as_id, id;
1164 1165 1166 1167
	unsigned long n;
	unsigned long *dirty_bitmap;
	unsigned long *dirty_bitmap_buffer;

1168 1169 1170
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1171
		return -EINVAL;
1172

1173 1174
	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
1175 1176 1177

	dirty_bitmap = memslot->dirty_bitmap;
	if (!dirty_bitmap)
1178
		return -ENOENT;
1179 1180

	n = kvm_dirty_bitmap_bytes(memslot);
1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194
	*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);
1195

1196 1197 1198 1199
		spin_lock(&kvm->mmu_lock);
		for (i = 0; i < n / sizeof(long); i++) {
			unsigned long mask;
			gfn_t offset;
1200

1201 1202 1203 1204 1205 1206 1207
			if (!dirty_bitmap[i])
				continue;

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

1208 1209 1210
			offset = i * BITS_PER_LONG;
			kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot,
								offset, mask);
1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231
		}
		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;
1232
	int as_id, id;
1233
	gfn_t offset;
1234
	unsigned long i, n;
1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253
	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);
1254 1255 1256 1257 1258

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

1259
	*flush = false;
1260 1261 1262
	dirty_bitmap_buffer = kvm_second_dirty_bitmap(memslot);
	if (copy_from_user(dirty_bitmap_buffer, log->dirty_bitmap, n))
		return -EFAULT;
1263

1264 1265 1266 1267 1268 1269 1270
	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)
1271 1272
			continue;

1273
		mask &= atomic_long_fetch_andnot(mask, p);
1274

1275 1276 1277 1278 1279 1280
		/*
		 * 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.
		*/
1281
		if (mask) {
1282
			*flush = true;
1283 1284 1285
			kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot,
								offset, mask);
		}
1286 1287
	}
	spin_unlock(&kvm->mmu_lock);
1288

1289
	return 0;
1290
}
1291
EXPORT_SYMBOL_GPL(kvm_clear_dirty_log_protect);
1292 1293
#endif

1294 1295 1296 1297 1298
bool kvm_largepages_enabled(void)
{
	return largepages_enabled;
}

1299 1300 1301 1302 1303 1304
void kvm_disable_largepages(void)
{
	largepages_enabled = false;
}
EXPORT_SYMBOL_GPL(kvm_disable_largepages);

1305 1306 1307 1308
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
1309
EXPORT_SYMBOL_GPL(gfn_to_memslot);
A
Avi Kivity 已提交
1310

1311 1312 1313 1314 1315
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);
}

1316
bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
1317
{
1318
	struct kvm_memory_slot *memslot = gfn_to_memslot(kvm, gfn);
1319

1320
	if (!memslot || memslot->id >= KVM_USER_MEM_SLOTS ||
1321
	      memslot->flags & KVM_MEMSLOT_INVALID)
1322
		return false;
1323

1324
	return true;
1325 1326 1327
}
EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);

J
Joerg Roedel 已提交
1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351
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
1352 1353 1354 1355 1356 1357 1358
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
1359
{
1360
	if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
X
Xiao Guangrong 已提交
1361
		return KVM_HVA_ERR_BAD;
1362

X
Xiao Guangrong 已提交
1363 1364
	if (memslot_is_readonly(slot) && write)
		return KVM_HVA_ERR_RO_BAD;
1365 1366 1367 1368

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

X
Xiao Guangrong 已提交
1369
	return __gfn_to_hva_memslot(slot, gfn);
I
Izik Eidus 已提交
1370
}
1371

X
Xiao Guangrong 已提交
1372 1373 1374 1375
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
1376
}
1377

X
Xiao Guangrong 已提交
1378
unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot,
1379
					gfn_t gfn)
X
Xiao Guangrong 已提交
1380 1381 1382 1383 1384
{
	return gfn_to_hva_many(slot, gfn, NULL);
}
EXPORT_SYMBOL_GPL(gfn_to_hva_memslot);

1385 1386
unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
{
1387
	return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL);
1388
}
1389
EXPORT_SYMBOL_GPL(gfn_to_hva);
I
Izik Eidus 已提交
1390

1391 1392 1393 1394 1395 1396
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);

1397
/*
1398 1399 1400 1401 1402 1403
 * 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
1404
 */
1405 1406
unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot,
				      gfn_t gfn, bool *writable)
1407
{
1408 1409 1410
	unsigned long hva = __gfn_to_hva_many(slot, gfn, NULL, false);

	if (!kvm_is_error_hva(hva) && writable)
1411 1412
		*writable = !memslot_is_readonly(slot);

1413
	return hva;
1414 1415
}

1416 1417 1418 1419 1420 1421 1422
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);
}

1423 1424 1425 1426 1427 1428 1429
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);
}

1430 1431
static inline int check_user_page_hwpoison(unsigned long addr)
{
L
Lorenzo Stoakes 已提交
1432
	int rc, flags = FOLL_HWPOISON | FOLL_WRITE;
1433

L
Lorenzo Stoakes 已提交
1434
	rc = get_user_pages(addr, 1, flags, NULL, NULL);
1435 1436 1437
	return rc == -EHWPOISON;
}

X
Xiao Guangrong 已提交
1438
/*
1439 1440 1441
 * 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
1442
 */
1443 1444
static bool hva_to_pfn_fast(unsigned long addr, bool write_fault,
			    bool *writable, kvm_pfn_t *pfn)
A
Avi Kivity 已提交
1445
{
1446
	struct page *page[1];
X
Xiao Guangrong 已提交
1447
	int npages;
A
Avi Kivity 已提交
1448

1449 1450 1451 1452 1453 1454 1455
	/*
	 * 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;
1456

X
Xiao Guangrong 已提交
1457 1458 1459
	npages = __get_user_pages_fast(addr, 1, 1, page);
	if (npages == 1) {
		*pfn = page_to_pfn(page[0]);
1460

X
Xiao Guangrong 已提交
1461 1462 1463 1464
		if (writable)
			*writable = true;
		return true;
	}
1465

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

X
Xiao Guangrong 已提交
1469 1470 1471 1472 1473
/*
 * 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
1474
			   bool *writable, kvm_pfn_t *pfn)
X
Xiao Guangrong 已提交
1475
{
1476 1477
	unsigned int flags = FOLL_HWPOISON;
	struct page *page;
X
Xiao Guangrong 已提交
1478
	int npages = 0;
1479

X
Xiao Guangrong 已提交
1480 1481 1482 1483 1484
	might_sleep();

	if (writable)
		*writable = write_fault;

1485 1486 1487 1488
	if (write_fault)
		flags |= FOLL_WRITE;
	if (async)
		flags |= FOLL_NOWAIT;
1489

1490
	npages = get_user_pages_unlocked(addr, 1, &page, flags);
X
Xiao Guangrong 已提交
1491 1492 1493 1494
	if (npages != 1)
		return npages;

	/* map read fault as writable if possible */
1495
	if (unlikely(!write_fault) && writable) {
1496
		struct page *wpage;
X
Xiao Guangrong 已提交
1497

1498
		if (__get_user_pages_fast(addr, 1, 1, &wpage) == 1) {
X
Xiao Guangrong 已提交
1499
			*writable = true;
1500 1501
			put_page(page);
			page = wpage;
1502
		}
1503
	}
1504
	*pfn = page_to_pfn(page);
X
Xiao Guangrong 已提交
1505 1506
	return npages;
}
I
Izik Eidus 已提交
1507

X
Xiao Guangrong 已提交
1508 1509 1510 1511
static bool vma_is_valid(struct vm_area_struct *vma, bool write_fault)
{
	if (unlikely(!(vma->vm_flags & VM_READ)))
		return false;
1512

X
Xiao Guangrong 已提交
1513 1514
	if (write_fault && (unlikely(!(vma->vm_flags & VM_WRITE))))
		return false;
1515

X
Xiao Guangrong 已提交
1516 1517
	return true;
}
1518

1519 1520
static int hva_to_pfn_remapped(struct vm_area_struct *vma,
			       unsigned long addr, bool *async,
1521 1522
			       bool write_fault, bool *writable,
			       kvm_pfn_t *p_pfn)
1523
{
1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547
	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;

	}

1548 1549
	if (writable)
		*writable = true;
1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564

	/*
	 * 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;
1565 1566 1567
	return 0;
}

1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581
/*
 * 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
1582
static kvm_pfn_t hva_to_pfn(unsigned long addr, bool atomic, bool *async,
X
Xiao Guangrong 已提交
1583 1584 1585
			bool write_fault, bool *writable)
{
	struct vm_area_struct *vma;
D
Dan Williams 已提交
1586
	kvm_pfn_t pfn = 0;
1587
	int npages, r;
1588

X
Xiao Guangrong 已提交
1589 1590
	/* we can do it either atomically or asynchronously, not both */
	BUG_ON(atomic && async);
1591

1592
	if (hva_to_pfn_fast(addr, write_fault, writable, &pfn))
X
Xiao Guangrong 已提交
1593 1594 1595 1596 1597 1598 1599 1600
		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;
1601

X
Xiao Guangrong 已提交
1602 1603 1604 1605 1606 1607 1608
	down_read(&current->mm->mmap_sem);
	if (npages == -EHWPOISON ||
	      (!async && check_user_page_hwpoison(addr))) {
		pfn = KVM_PFN_ERR_HWPOISON;
		goto exit;
	}

1609
retry:
X
Xiao Guangrong 已提交
1610 1611 1612 1613
	vma = find_vma_intersection(current->mm, addr, addr + 1);

	if (vma == NULL)
		pfn = KVM_PFN_ERR_FAULT;
1614
	else if (vma->vm_flags & (VM_IO | VM_PFNMAP)) {
1615
		r = hva_to_pfn_remapped(vma, addr, async, write_fault, writable, &pfn);
1616 1617
		if (r == -EAGAIN)
			goto retry;
1618 1619
		if (r < 0)
			pfn = KVM_PFN_ERR_FAULT;
X
Xiao Guangrong 已提交
1620
	} else {
X
Xiao Guangrong 已提交
1621
		if (async && vma_is_valid(vma, write_fault))
X
Xiao Guangrong 已提交
1622 1623 1624 1625 1626
			*async = true;
		pfn = KVM_PFN_ERR_FAULT;
	}
exit:
	up_read(&current->mm->mmap_sem);
1627
	return pfn;
1628 1629
}

D
Dan Williams 已提交
1630 1631 1632
kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
			       bool atomic, bool *async, bool write_fault,
			       bool *writable)
1633
{
X
Xiao Guangrong 已提交
1634 1635
	unsigned long addr = __gfn_to_hva_many(slot, gfn, NULL, write_fault);

1636 1637 1638
	if (addr == KVM_HVA_ERR_RO_BAD) {
		if (writable)
			*writable = false;
X
Xiao Guangrong 已提交
1639
		return KVM_PFN_ERR_RO_FAULT;
1640
	}
X
Xiao Guangrong 已提交
1641

1642 1643 1644
	if (kvm_is_error_hva(addr)) {
		if (writable)
			*writable = false;
1645
		return KVM_PFN_NOSLOT;
1646
	}
X
Xiao Guangrong 已提交
1647 1648 1649 1650 1651 1652 1653 1654 1655

	/* 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);
1656
}
1657
EXPORT_SYMBOL_GPL(__gfn_to_pfn_memslot);
1658

D
Dan Williams 已提交
1659
kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1660 1661
		      bool *writable)
{
P
Paolo Bonzini 已提交
1662 1663
	return __gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn, false, NULL,
				    write_fault, writable);
1664 1665 1666
}
EXPORT_SYMBOL_GPL(gfn_to_pfn_prot);

D
Dan Williams 已提交
1667
kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1668
{
X
Xiao Guangrong 已提交
1669
	return __gfn_to_pfn_memslot(slot, gfn, false, NULL, true, NULL);
1670
}
P
Paolo Bonzini 已提交
1671
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot);
1672

D
Dan Williams 已提交
1673
kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn)
1674
{
X
Xiao Guangrong 已提交
1675
	return __gfn_to_pfn_memslot(slot, gfn, true, NULL, true, NULL);
1676
}
1677
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot_atomic);
1678

D
Dan Williams 已提交
1679
kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1680 1681 1682 1683 1684
{
	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
1685
kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn)
1686 1687 1688 1689 1690
{
	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
1691
kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1692 1693 1694 1695 1696
{
	return gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn);
}
EXPORT_SYMBOL_GPL(gfn_to_pfn);

D
Dan Williams 已提交
1697
kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn)
1698 1699 1700 1701 1702
{
	return gfn_to_pfn_memslot(kvm_vcpu_gfn_to_memslot(vcpu, gfn), gfn);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_gfn_to_pfn);

1703 1704
int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
			    struct page **pages, int nr_pages)
1705 1706
{
	unsigned long addr;
1707
	gfn_t entry = 0;
1708

1709
	addr = gfn_to_hva_many(slot, gfn, &entry);
1710 1711 1712 1713 1714 1715 1716 1717 1718 1719
	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
1720
static struct page *kvm_pfn_to_page(kvm_pfn_t pfn)
X
Xiao Guangrong 已提交
1721
{
1722
	if (is_error_noslot_pfn(pfn))
1723
		return KVM_ERR_PTR_BAD_PAGE;
X
Xiao Guangrong 已提交
1724

1725
	if (kvm_is_reserved_pfn(pfn)) {
1726
		WARN_ON(1);
1727
		return KVM_ERR_PTR_BAD_PAGE;
1728
	}
X
Xiao Guangrong 已提交
1729 1730 1731 1732

	return pfn_to_page(pfn);
}

1733 1734
struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
{
D
Dan Williams 已提交
1735
	kvm_pfn_t pfn;
1736 1737 1738

	pfn = gfn_to_pfn(kvm, gfn);

X
Xiao Guangrong 已提交
1739
	return kvm_pfn_to_page(pfn);
A
Avi Kivity 已提交
1740 1741 1742
}
EXPORT_SYMBOL_GPL(gfn_to_page);

1743 1744
struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn)
{
D
Dan Williams 已提交
1745
	kvm_pfn_t pfn;
1746 1747 1748 1749 1750 1751 1752

	pfn = kvm_vcpu_gfn_to_pfn(vcpu, gfn);

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

1753 1754
void kvm_release_page_clean(struct page *page)
{
1755 1756
	WARN_ON(is_error_page(page));

1757
	kvm_release_pfn_clean(page_to_pfn(page));
1758 1759 1760
}
EXPORT_SYMBOL_GPL(kvm_release_page_clean);

D
Dan Williams 已提交
1761
void kvm_release_pfn_clean(kvm_pfn_t pfn)
1762
{
1763
	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn))
1764
		put_page(pfn_to_page(pfn));
1765 1766 1767
}
EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);

1768
void kvm_release_page_dirty(struct page *page)
1769
{
X
Xiao Guangrong 已提交
1770 1771
	WARN_ON(is_error_page(page));

1772 1773 1774 1775
	kvm_release_pfn_dirty(page_to_pfn(page));
}
EXPORT_SYMBOL_GPL(kvm_release_page_dirty);

1776
void kvm_release_pfn_dirty(kvm_pfn_t pfn)
1777 1778 1779 1780
{
	kvm_set_pfn_dirty(pfn);
	kvm_release_pfn_clean(pfn);
}
1781
EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
1782

D
Dan Williams 已提交
1783
void kvm_set_pfn_dirty(kvm_pfn_t pfn)
1784
{
1785
	if (!kvm_is_reserved_pfn(pfn)) {
1786
		struct page *page = pfn_to_page(pfn);
1787

1788 1789 1790
		if (!PageReserved(page))
			SetPageDirty(page);
	}
1791
}
1792 1793
EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);

D
Dan Williams 已提交
1794
void kvm_set_pfn_accessed(kvm_pfn_t pfn)
1795
{
1796
	if (!kvm_is_reserved_pfn(pfn))
1797
		mark_page_accessed(pfn_to_page(pfn));
1798 1799 1800
}
EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);

D
Dan Williams 已提交
1801
void kvm_get_pfn(kvm_pfn_t pfn)
1802
{
1803
	if (!kvm_is_reserved_pfn(pfn))
1804
		get_page(pfn_to_page(pfn));
1805 1806
}
EXPORT_SYMBOL_GPL(kvm_get_pfn);
1807

1808 1809 1810 1811 1812 1813 1814 1815
static int next_segment(unsigned long len, int offset)
{
	if (len > PAGE_SIZE - offset)
		return PAGE_SIZE - offset;
	else
		return len;
}

1816 1817
static int __kvm_read_guest_page(struct kvm_memory_slot *slot, gfn_t gfn,
				 void *data, int offset, int len)
1818
{
1819 1820
	int r;
	unsigned long addr;
1821

1822
	addr = gfn_to_hva_memslot_prot(slot, gfn, NULL);
1823 1824
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1825
	r = __copy_from_user(data, (void __user *)addr + offset, len);
1826
	if (r)
1827 1828 1829
		return -EFAULT;
	return 0;
}
1830 1831 1832 1833 1834 1835 1836 1837

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);
}
1838 1839
EXPORT_SYMBOL_GPL(kvm_read_guest_page);

1840 1841 1842 1843 1844 1845 1846 1847 1848
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);

1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868
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);

1869
int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data, unsigned long len)
1870 1871
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
1872
	int seg;
1873
	int offset = offset_in_page(gpa);
1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887
	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);
1888

1889 1890 1891 1892 1893 1894 1895
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);
1896 1897
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1898
	pagefault_disable();
1899
	r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
1900
	pagefault_enable();
1901 1902 1903 1904 1905
	if (r)
		return -EFAULT;
	return 0;
}

1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929
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)
1930
{
1931 1932
	int r;
	unsigned long addr;
1933

1934
	addr = gfn_to_hva_memslot(memslot, gfn);
1935 1936
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1937
	r = __copy_to_user((void __user *)addr + offset, data, len);
1938
	if (r)
1939
		return -EFAULT;
1940
	mark_page_dirty_in_slot(memslot, gfn);
1941 1942
	return 0;
}
1943 1944 1945 1946 1947 1948 1949 1950

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);
}
1951 1952
EXPORT_SYMBOL_GPL(kvm_write_guest_page);

1953 1954 1955 1956 1957 1958 1959 1960 1961
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);

1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980
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;
}
1981
EXPORT_SYMBOL_GPL(kvm_write_guest);
1982

1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
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);

2004 2005 2006
static int __kvm_gfn_to_hva_cache_init(struct kvm_memslots *slots,
				       struct gfn_to_hva_cache *ghc,
				       gpa_t gpa, unsigned long len)
2007 2008
{
	int offset = offset_in_page(gpa);
2009 2010 2011 2012
	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;
2013
	int r = start_gfn <= end_gfn ? 0 : -EINVAL;
2014 2015 2016

	ghc->gpa = gpa;
	ghc->generation = slots->generation;
2017
	ghc->len = len;
2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034
	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)
2035
		ghc->hva += offset;
2036
	else
2037
		ghc->memslot = NULL;
2038 2039

	return r;
2040
}
2041

2042
int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
2043 2044
			      gpa_t gpa, unsigned long len)
{
2045
	struct kvm_memslots *slots = kvm_memslots(kvm);
2046 2047
	return __kvm_gfn_to_hva_cache_init(slots, ghc, gpa, len);
}
2048
EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init);
2049

2050
int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
2051 2052
				  void *data, unsigned int offset,
				  unsigned long len)
2053
{
2054
	struct kvm_memslots *slots = kvm_memslots(kvm);
2055
	int r;
2056
	gpa_t gpa = ghc->gpa + offset;
2057

2058
	BUG_ON(len + offset > ghc->len);
2059

2060
	if (slots->generation != ghc->generation)
2061
		__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len);
2062 2063

	if (unlikely(!ghc->memslot))
2064
		return kvm_write_guest(kvm, gpa, data, len);
2065 2066 2067 2068

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

2069
	r = __copy_to_user((void __user *)ghc->hva + offset, data, len);
2070 2071
	if (r)
		return -EFAULT;
2072
	mark_page_dirty_in_slot(ghc->memslot, gpa >> PAGE_SHIFT);
2073 2074 2075

	return 0;
}
2076
EXPORT_SYMBOL_GPL(kvm_write_guest_offset_cached);
2077

2078 2079
int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
2080
{
2081
	return kvm_write_guest_offset_cached(kvm, ghc, data, 0, len);
2082
}
2083
EXPORT_SYMBOL_GPL(kvm_write_guest_cached);
2084

2085 2086
int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
2087
{
2088
	struct kvm_memslots *slots = kvm_memslots(kvm);
2089 2090
	int r;

2091 2092
	BUG_ON(len > ghc->len);

2093
	if (slots->generation != ghc->generation)
2094
		__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len);
2095 2096

	if (unlikely(!ghc->memslot))
2097
		return kvm_read_guest(kvm, ghc->gpa, data, len);
2098 2099 2100 2101 2102 2103 2104 2105 2106 2107

	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;
}
2108
EXPORT_SYMBOL_GPL(kvm_read_guest_cached);
2109

2110 2111
int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
{
2112 2113 2114
	const void *zero_page = (const void *) __va(page_to_phys(ZERO_PAGE(0)));

	return kvm_write_guest_page(kvm, gfn, zero_page, offset, len);
2115 2116 2117 2118 2119 2120 2121 2122 2123 2124
}
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;

2125
	while ((seg = next_segment(len, offset)) != 0) {
2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136
		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);

2137
static void mark_page_dirty_in_slot(struct kvm_memory_slot *memslot,
2138
				    gfn_t gfn)
A
Avi Kivity 已提交
2139
{
R
Rusty Russell 已提交
2140 2141
	if (memslot && memslot->dirty_bitmap) {
		unsigned long rel_gfn = gfn - memslot->base_gfn;
A
Avi Kivity 已提交
2142

2143
		set_bit_le(rel_gfn, memslot->dirty_bitmap);
A
Avi Kivity 已提交
2144 2145 2146
	}
}

2147 2148 2149 2150 2151
void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
{
	struct kvm_memory_slot *memslot;

	memslot = gfn_to_memslot(kvm, gfn);
2152
	mark_page_dirty_in_slot(memslot, gfn);
2153
}
2154
EXPORT_SYMBOL_GPL(mark_page_dirty);
2155

2156 2157 2158 2159 2160 2161 2162 2163 2164
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);

2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187
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
2188 2189
static void grow_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2190
	unsigned int old, val, grow, grow_start;
W
Wanpeng Li 已提交
2191

2192
	old = val = vcpu->halt_poll_ns;
2193
	grow_start = READ_ONCE(halt_poll_ns_grow_start);
2194
	grow = READ_ONCE(halt_poll_ns_grow);
2195 2196 2197
	if (!grow)
		goto out;

2198 2199 2200
	val *= grow;
	if (val < grow_start)
		val = grow_start;
W
Wanpeng Li 已提交
2201

2202 2203 2204
	if (val > halt_poll_ns)
		val = halt_poll_ns;

W
Wanpeng Li 已提交
2205
	vcpu->halt_poll_ns = val;
2206
out:
2207
	trace_kvm_halt_poll_ns_grow(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2208 2209 2210 2211
}

static void shrink_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2212
	unsigned int old, val, shrink;
W
Wanpeng Li 已提交
2213

2214
	old = val = vcpu->halt_poll_ns;
2215 2216
	shrink = READ_ONCE(halt_poll_ns_shrink);
	if (shrink == 0)
W
Wanpeng Li 已提交
2217 2218
		val = 0;
	else
2219
		val /= shrink;
W
Wanpeng Li 已提交
2220 2221

	vcpu->halt_poll_ns = val;
2222
	trace_kvm_halt_poll_ns_shrink(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2223 2224
}

2225 2226
static int kvm_vcpu_check_block(struct kvm_vcpu *vcpu)
{
2227 2228 2229
	int ret = -EINTR;
	int idx = srcu_read_lock(&vcpu->kvm->srcu);

2230 2231
	if (kvm_arch_vcpu_runnable(vcpu)) {
		kvm_make_request(KVM_REQ_UNHALT, vcpu);
2232
		goto out;
2233 2234
	}
	if (kvm_cpu_has_pending_timer(vcpu))
2235
		goto out;
2236
	if (signal_pending(current))
2237
		goto out;
2238

2239 2240 2241 2242
	ret = 0;
out:
	srcu_read_unlock(&vcpu->kvm->srcu, idx);
	return ret;
2243 2244
}

E
Eddie Dong 已提交
2245 2246 2247
/*
 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
 */
2248
void kvm_vcpu_block(struct kvm_vcpu *vcpu)
2249
{
2250
	ktime_t start, cur;
2251
	DECLARE_SWAITQUEUE(wait);
2252
	bool waited = false;
W
Wanpeng Li 已提交
2253
	u64 block_ns;
2254 2255

	start = cur = ktime_get();
W
Wanpeng Li 已提交
2256 2257
	if (vcpu->halt_poll_ns) {
		ktime_t stop = ktime_add_ns(ktime_get(), vcpu->halt_poll_ns);
2258

2259
		++vcpu->stat.halt_attempted_poll;
2260 2261 2262 2263 2264 2265 2266
		do {
			/*
			 * This sets KVM_REQ_UNHALT if an interrupt
			 * arrives.
			 */
			if (kvm_vcpu_check_block(vcpu) < 0) {
				++vcpu->stat.halt_successful_poll;
2267 2268
				if (!vcpu_valid_wakeup(vcpu))
					++vcpu->stat.halt_poll_invalid;
2269 2270 2271 2272 2273
				goto out;
			}
			cur = ktime_get();
		} while (single_task_running() && ktime_before(cur, stop));
	}
2274

2275 2276
	kvm_arch_vcpu_blocking(vcpu);

2277
	for (;;) {
2278
		prepare_to_swait_exclusive(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
2279

2280
		if (kvm_vcpu_check_block(vcpu) < 0)
2281 2282
			break;

2283
		waited = true;
E
Eddie Dong 已提交
2284 2285
		schedule();
	}
2286

2287
	finish_swait(&vcpu->wq, &wait);
2288 2289
	cur = ktime_get();

2290
	kvm_arch_vcpu_unblocking(vcpu);
2291
out:
W
Wanpeng Li 已提交
2292 2293
	block_ns = ktime_to_ns(cur) - ktime_to_ns(start);

2294 2295 2296
	if (!vcpu_valid_wakeup(vcpu))
		shrink_halt_poll_ns(vcpu);
	else if (halt_poll_ns) {
W
Wanpeng Li 已提交
2297 2298 2299
		if (block_ns <= vcpu->halt_poll_ns)
			;
		/* we had a long block, shrink polling */
2300
		else if (vcpu->halt_poll_ns && block_ns > halt_poll_ns)
W
Wanpeng Li 已提交
2301 2302 2303 2304 2305
			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);
2306 2307
	} else
		vcpu->halt_poll_ns = 0;
W
Wanpeng Li 已提交
2308

2309 2310
	trace_kvm_vcpu_wakeup(block_ns, waited, vcpu_valid_wakeup(vcpu));
	kvm_arch_vcpu_block_finish(vcpu);
E
Eddie Dong 已提交
2311
}
2312
EXPORT_SYMBOL_GPL(kvm_vcpu_block);
E
Eddie Dong 已提交
2313

2314
bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu)
2315
{
2316
	struct swait_queue_head *wqp;
2317 2318

	wqp = kvm_arch_vcpu_wq(vcpu);
2319
	if (swq_has_sleeper(wqp)) {
2320
		swake_up_one(wqp);
2321
		++vcpu->stat.halt_wakeup;
2322
		return true;
2323 2324
	}

2325
	return false;
2326 2327 2328
}
EXPORT_SYMBOL_GPL(kvm_vcpu_wake_up);

2329
#ifndef CONFIG_S390
2330 2331 2332 2333 2334 2335 2336 2337
/*
 * 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;

2338 2339 2340
	if (kvm_vcpu_wake_up(vcpu))
		return;

2341 2342 2343 2344 2345 2346
	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();
}
2347
EXPORT_SYMBOL_GPL(kvm_vcpu_kick);
2348
#endif /* !CONFIG_S390 */
2349

2350
int kvm_vcpu_yield_to(struct kvm_vcpu *target)
2351 2352 2353
{
	struct pid *pid;
	struct task_struct *task = NULL;
2354
	int ret = 0;
2355 2356 2357 2358

	rcu_read_lock();
	pid = rcu_dereference(target->pid);
	if (pid)
2359
		task = get_pid_task(pid, PIDTYPE_PID);
2360 2361
	rcu_read_unlock();
	if (!task)
2362 2363
		return ret;
	ret = yield_to(task, 1);
2364
	put_task_struct(task);
2365 2366

	return ret;
2367 2368 2369
}
EXPORT_SYMBOL_GPL(kvm_vcpu_yield_to);

2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391
/*
 * 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.
 */
2392
static bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
2393
{
2394
#ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
2395 2396 2397
	bool eligible;

	eligible = !vcpu->spin_loop.in_spin_loop ||
2398
		    vcpu->spin_loop.dy_eligible;
2399 2400 2401 2402 2403

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

	return eligible;
2404 2405
#else
	return true;
2406
#endif
2407
}
2408

2409
void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
Z
Zhai, Edwin 已提交
2410
{
2411 2412 2413 2414
	struct kvm *kvm = me->kvm;
	struct kvm_vcpu *vcpu;
	int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
	int yielded = 0;
2415
	int try = 3;
2416 2417
	int pass;
	int i;
Z
Zhai, Edwin 已提交
2418

2419
	kvm_vcpu_set_in_spin_loop(me, true);
2420 2421 2422 2423 2424 2425 2426
	/*
	 * 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.
	 */
2427
	for (pass = 0; pass < 2 && !yielded && try; pass++) {
2428
		kvm_for_each_vcpu(i, vcpu, kvm) {
2429
			if (!pass && i <= last_boosted_vcpu) {
2430 2431 2432 2433
				i = last_boosted_vcpu;
				continue;
			} else if (pass && i > last_boosted_vcpu)
				break;
2434
			if (!READ_ONCE(vcpu->preempted))
2435
				continue;
2436 2437
			if (vcpu == me)
				continue;
2438
			if (swait_active(&vcpu->wq) && !kvm_arch_vcpu_runnable(vcpu))
2439
				continue;
2440 2441
			if (yield_to_kernel_mode && !kvm_arch_vcpu_in_kernel(vcpu))
				continue;
2442 2443
			if (!kvm_vcpu_eligible_for_directed_yield(vcpu))
				continue;
2444 2445 2446

			yielded = kvm_vcpu_yield_to(vcpu);
			if (yielded > 0) {
2447 2448
				kvm->last_boosted_vcpu = i;
				break;
2449 2450 2451 2452
			} else if (yielded < 0) {
				try--;
				if (!try)
					break;
2453 2454 2455
			}
		}
	}
2456
	kvm_vcpu_set_in_spin_loop(me, false);
2457 2458 2459

	/* Ensure vcpu is not eligible during next spinloop */
	kvm_vcpu_set_dy_eligible(me, false);
Z
Zhai, Edwin 已提交
2460 2461 2462
}
EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin);

2463
static vm_fault_t kvm_vcpu_fault(struct vm_fault *vmf)
2464
{
2465
	struct kvm_vcpu *vcpu = vmf->vma->vm_file->private_data;
2466 2467
	struct page *page;

2468
	if (vmf->pgoff == 0)
2469
		page = virt_to_page(vcpu->run);
A
Avi Kivity 已提交
2470
#ifdef CONFIG_X86
2471
	else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
2472
		page = virt_to_page(vcpu->arch.pio_data);
2473
#endif
2474
#ifdef CONFIG_KVM_MMIO
2475 2476
	else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
		page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
A
Avi Kivity 已提交
2477
#endif
2478
	else
2479
		return kvm_arch_vcpu_fault(vcpu, vmf);
2480
	get_page(page);
2481 2482
	vmf->page = page;
	return 0;
2483 2484
}

2485
static const struct vm_operations_struct kvm_vcpu_vm_ops = {
2486
	.fault = kvm_vcpu_fault,
2487 2488 2489 2490 2491 2492 2493 2494
};

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
2495 2496 2497 2498
static int kvm_vcpu_release(struct inode *inode, struct file *filp)
{
	struct kvm_vcpu *vcpu = filp->private_data;

2499
	debugfs_remove_recursive(vcpu->debugfs_dentry);
A
Al Viro 已提交
2500
	kvm_put_kvm(vcpu->kvm);
A
Avi Kivity 已提交
2501 2502 2503
	return 0;
}

2504
static struct file_operations kvm_vcpu_fops = {
A
Avi Kivity 已提交
2505 2506
	.release        = kvm_vcpu_release,
	.unlocked_ioctl = kvm_vcpu_ioctl,
2507
	.mmap           = kvm_vcpu_mmap,
2508
	.llseek		= noop_llseek,
2509
	KVM_COMPAT(kvm_vcpu_compat_ioctl),
A
Avi Kivity 已提交
2510 2511 2512 2513 2514 2515 2516
};

/*
 * Allocates an inode for the vcpu.
 */
static int create_vcpu_fd(struct kvm_vcpu *vcpu)
{
2517 2518 2519 2520
	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
2521 2522
}

2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548
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;
}

2549 2550 2551
/*
 * Creates some virtual cpus.  Good luck creating more than one.
 */
2552
static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
2553 2554
{
	int r;
2555
	struct kvm_vcpu *vcpu;
2556

G
Greg Kurz 已提交
2557
	if (id >= KVM_MAX_VCPU_ID)
2558 2559
		return -EINVAL;

2560 2561 2562 2563 2564 2565 2566 2567 2568
	mutex_lock(&kvm->lock);
	if (kvm->created_vcpus == KVM_MAX_VCPUS) {
		mutex_unlock(&kvm->lock);
		return -EINVAL;
	}

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

2569
	vcpu = kvm_arch_vcpu_create(kvm, id);
2570 2571 2572 2573
	if (IS_ERR(vcpu)) {
		r = PTR_ERR(vcpu);
		goto vcpu_decrement;
	}
2574

2575 2576
	preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);

2577 2578
	r = kvm_arch_vcpu_setup(vcpu);
	if (r)
2579
		goto vcpu_destroy;
2580

2581 2582 2583 2584
	r = kvm_create_vcpu_debugfs(vcpu);
	if (r)
		goto vcpu_destroy;

S
Shaohua Li 已提交
2585
	mutex_lock(&kvm->lock);
2586 2587 2588 2589
	if (kvm_get_vcpu_by_id(kvm, id)) {
		r = -EEXIST;
		goto unlock_vcpu_destroy;
	}
2590 2591

	BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
2592

R
Rusty Russell 已提交
2593
	/* Now it's all set up, let userspace reach it */
A
Al Viro 已提交
2594
	kvm_get_kvm(kvm);
A
Avi Kivity 已提交
2595
	r = create_vcpu_fd(vcpu);
2596 2597
	if (r < 0) {
		kvm_put_kvm(kvm);
2598
		goto unlock_vcpu_destroy;
2599 2600 2601
	}

	kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu;
2602 2603 2604 2605 2606

	/*
	 * Pairs with smp_rmb() in kvm_get_vcpu.  Write kvm->vcpus
	 * before kvm->online_vcpu's incremented value.
	 */
2607 2608 2609 2610
	smp_wmb();
	atomic_inc(&kvm->online_vcpus);

	mutex_unlock(&kvm->lock);
2611
	kvm_arch_vcpu_postcreate(vcpu);
R
Rusty Russell 已提交
2612
	return r;
2613

2614
unlock_vcpu_destroy:
2615
	mutex_unlock(&kvm->lock);
2616
	debugfs_remove_recursive(vcpu->debugfs_dentry);
2617
vcpu_destroy:
2618
	kvm_arch_vcpu_destroy(vcpu);
2619 2620 2621 2622
vcpu_decrement:
	mutex_lock(&kvm->lock);
	kvm->created_vcpus--;
	mutex_unlock(&kvm->lock);
2623 2624 2625
	return r;
}

A
Avi Kivity 已提交
2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636
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
2637 2638
static long kvm_vcpu_ioctl(struct file *filp,
			   unsigned int ioctl, unsigned long arg)
A
Avi Kivity 已提交
2639
{
A
Avi Kivity 已提交
2640
	struct kvm_vcpu *vcpu = filp->private_data;
A
Al Viro 已提交
2641
	void __user *argp = (void __user *)arg;
2642
	int r;
2643 2644
	struct kvm_fpu *fpu = NULL;
	struct kvm_sregs *kvm_sregs = NULL;
A
Avi Kivity 已提交
2645

2646 2647
	if (vcpu->kvm->mm != current->mm)
		return -EIO;
2648

2649 2650 2651
	if (unlikely(_IOC_TYPE(ioctl) != KVMIO))
		return -EINVAL;

2652
	/*
2653 2654
	 * Some architectures have vcpu ioctls that are asynchronous to vcpu
	 * execution; mutex_lock() would break them.
2655
	 */
2656 2657
	r = kvm_arch_vcpu_async_ioctl(filp, ioctl, arg);
	if (r != -ENOIOCTLCMD)
2658
		return r;
2659

2660 2661
	if (mutex_lock_killable(&vcpu->mutex))
		return -EINTR;
A
Avi Kivity 已提交
2662
	switch (ioctl) {
2663 2664
	case KVM_RUN: {
		struct pid *oldpid;
2665 2666 2667
		r = -EINVAL;
		if (arg)
			goto out;
2668
		oldpid = rcu_access_pointer(vcpu->pid);
2669
		if (unlikely(oldpid != task_pid(current))) {
2670
			/* The thread running this VCPU changed. */
2671
			struct pid *newpid;
2672

2673 2674 2675 2676 2677
			r = kvm_arch_vcpu_run_pid_change(vcpu);
			if (r)
				break;

			newpid = get_task_pid(current, PIDTYPE_PID);
2678 2679 2680 2681 2682
			rcu_assign_pointer(vcpu->pid, newpid);
			if (oldpid)
				synchronize_rcu();
			put_pid(oldpid);
		}
2683
		r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
2684
		trace_kvm_userspace_exit(vcpu->run->exit_reason, r);
A
Avi Kivity 已提交
2685
		break;
2686
	}
A
Avi Kivity 已提交
2687
	case KVM_GET_REGS: {
2688
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
2689

2690
		r = -ENOMEM;
2691
		kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL_ACCOUNT);
2692
		if (!kvm_regs)
A
Avi Kivity 已提交
2693
			goto out;
2694 2695 2696
		r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
		if (r)
			goto out_free1;
A
Avi Kivity 已提交
2697
		r = -EFAULT;
2698 2699
		if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
			goto out_free1;
A
Avi Kivity 已提交
2700
		r = 0;
2701 2702
out_free1:
		kfree(kvm_regs);
A
Avi Kivity 已提交
2703 2704 2705
		break;
	}
	case KVM_SET_REGS: {
2706
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
2707

2708
		r = -ENOMEM;
2709 2710 2711
		kvm_regs = memdup_user(argp, sizeof(*kvm_regs));
		if (IS_ERR(kvm_regs)) {
			r = PTR_ERR(kvm_regs);
A
Avi Kivity 已提交
2712
			goto out;
2713
		}
2714 2715
		r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
		kfree(kvm_regs);
A
Avi Kivity 已提交
2716 2717 2718
		break;
	}
	case KVM_GET_SREGS: {
2719 2720
		kvm_sregs = kzalloc(sizeof(struct kvm_sregs),
				    GFP_KERNEL_ACCOUNT);
2721 2722 2723 2724
		r = -ENOMEM;
		if (!kvm_sregs)
			goto out;
		r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
2725 2726 2727
		if (r)
			goto out;
		r = -EFAULT;
2728
		if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
A
Avi Kivity 已提交
2729 2730 2731 2732 2733
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_SREGS: {
2734 2735 2736
		kvm_sregs = memdup_user(argp, sizeof(*kvm_sregs));
		if (IS_ERR(kvm_sregs)) {
			r = PTR_ERR(kvm_sregs);
G
Guo Chao 已提交
2737
			kvm_sregs = NULL;
A
Avi Kivity 已提交
2738
			goto out;
2739
		}
2740
		r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
2741 2742
		break;
	}
2743 2744 2745 2746 2747 2748 2749
	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;
2750
		if (copy_to_user(argp, &mp_state, sizeof(mp_state)))
2751 2752 2753 2754 2755 2756 2757 2758
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_MP_STATE: {
		struct kvm_mp_state mp_state;

		r = -EFAULT;
2759
		if (copy_from_user(&mp_state, argp, sizeof(mp_state)))
2760 2761 2762 2763
			goto out;
		r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
		break;
	}
A
Avi Kivity 已提交
2764 2765 2766 2767
	case KVM_TRANSLATE: {
		struct kvm_translation tr;

		r = -EFAULT;
2768
		if (copy_from_user(&tr, argp, sizeof(tr)))
A
Avi Kivity 已提交
2769
			goto out;
2770
		r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
A
Avi Kivity 已提交
2771 2772 2773
		if (r)
			goto out;
		r = -EFAULT;
2774
		if (copy_to_user(argp, &tr, sizeof(tr)))
A
Avi Kivity 已提交
2775 2776 2777 2778
			goto out;
		r = 0;
		break;
	}
J
Jan Kiszka 已提交
2779 2780
	case KVM_SET_GUEST_DEBUG: {
		struct kvm_guest_debug dbg;
A
Avi Kivity 已提交
2781 2782

		r = -EFAULT;
2783
		if (copy_from_user(&dbg, argp, sizeof(dbg)))
A
Avi Kivity 已提交
2784
			goto out;
J
Jan Kiszka 已提交
2785
		r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
A
Avi Kivity 已提交
2786 2787
		break;
	}
A
Avi Kivity 已提交
2788 2789 2790 2791 2792 2793 2794 2795 2796
	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,
2797
					   sizeof(kvm_sigmask)))
A
Avi Kivity 已提交
2798 2799
				goto out;
			r = -EINVAL;
2800
			if (kvm_sigmask.len != sizeof(sigset))
A
Avi Kivity 已提交
2801 2802 2803
				goto out;
			r = -EFAULT;
			if (copy_from_user(&sigset, sigmask_arg->sigset,
2804
					   sizeof(sigset)))
A
Avi Kivity 已提交
2805 2806 2807
				goto out;
			p = &sigset;
		}
2808
		r = kvm_vcpu_ioctl_set_sigmask(vcpu, p);
A
Avi Kivity 已提交
2809 2810
		break;
	}
A
Avi Kivity 已提交
2811
	case KVM_GET_FPU: {
2812
		fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL_ACCOUNT);
2813 2814 2815 2816
		r = -ENOMEM;
		if (!fpu)
			goto out;
		r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu);
A
Avi Kivity 已提交
2817 2818 2819
		if (r)
			goto out;
		r = -EFAULT;
2820
		if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
A
Avi Kivity 已提交
2821 2822 2823 2824 2825
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_FPU: {
2826 2827 2828
		fpu = memdup_user(argp, sizeof(*fpu));
		if (IS_ERR(fpu)) {
			r = PTR_ERR(fpu);
G
Guo Chao 已提交
2829
			fpu = NULL;
A
Avi Kivity 已提交
2830
			goto out;
2831
		}
2832
		r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
A
Avi Kivity 已提交
2833 2834
		break;
	}
A
Avi Kivity 已提交
2835
	default:
2836
		r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
2837 2838
	}
out:
2839
	mutex_unlock(&vcpu->mutex);
2840 2841
	kfree(fpu);
	kfree(kvm_sregs);
A
Avi Kivity 已提交
2842 2843 2844
	return r;
}

2845
#ifdef CONFIG_KVM_COMPAT
2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864
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,
2865
					   sizeof(kvm_sigmask)))
2866 2867
				goto out;
			r = -EINVAL;
A
Al Viro 已提交
2868
			if (kvm_sigmask.len != sizeof(compat_sigset_t))
2869 2870
				goto out;
			r = -EFAULT;
A
Al Viro 已提交
2871
			if (get_compat_sigset(&sigset, (void *)sigmask_arg->sigset))
2872
				goto out;
2873 2874 2875
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
		} else
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, NULL);
2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886
		break;
	}
	default:
		r = kvm_vcpu_ioctl(filp, ioctl, arg);
	}

out:
	return r;
}
#endif

S
Scott Wood 已提交
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 2929 2930 2931 2932 2933 2934
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,
2935
	KVM_COMPAT(kvm_device_ioctl),
S
Scott Wood 已提交
2936 2937 2938 2939 2940 2941 2942 2943 2944 2945
};

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

	return filp->private_data;
}

2946
static struct kvm_device_ops *kvm_device_ops_table[KVM_DEV_TYPE_MAX] = {
2947
#ifdef CONFIG_KVM_MPIC
2948 2949
	[KVM_DEV_TYPE_FSL_MPIC_20]	= &kvm_mpic_ops,
	[KVM_DEV_TYPE_FSL_MPIC_42]	= &kvm_mpic_ops,
2950
#endif
2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964
};

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;
}

2965 2966 2967 2968 2969 2970
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
2971 2972 2973 2974 2975 2976 2977 2978
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;

2979 2980 2981 2982 2983
	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
2984 2985 2986 2987 2988
		return -ENODEV;

	if (test)
		return 0;

2989
	dev = kzalloc(sizeof(*dev), GFP_KERNEL_ACCOUNT);
S
Scott Wood 已提交
2990 2991 2992 2993 2994 2995
	if (!dev)
		return -ENOMEM;

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

2996
	mutex_lock(&kvm->lock);
S
Scott Wood 已提交
2997 2998
	ret = ops->create(dev, cd->type);
	if (ret < 0) {
2999
		mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
3000 3001 3002
		kfree(dev);
		return ret;
	}
3003 3004
	list_add(&dev->vm_node, &kvm->devices);
	mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
3005

3006 3007 3008
	if (ops->init)
		ops->init(dev);

3009
	kvm_get_kvm(kvm);
3010
	ret = anon_inode_getfd(ops->name, &kvm_device_fops, dev, O_RDWR | O_CLOEXEC);
S
Scott Wood 已提交
3011
	if (ret < 0) {
3012
		kvm_put_kvm(kvm);
3013 3014 3015
		mutex_lock(&kvm->lock);
		list_del(&dev->vm_node);
		mutex_unlock(&kvm->lock);
3016
		ops->destroy(dev);
S
Scott Wood 已提交
3017 3018 3019 3020 3021 3022 3023
		return ret;
	}

	cd->fd = ret;
	return 0;
}

3024 3025 3026 3027 3028 3029 3030 3031 3032 3033
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
3034
#ifdef CONFIG_HAVE_KVM_IRQFD
3035
	case KVM_CAP_IRQFD:
3036 3037
	case KVM_CAP_IRQFD_RESAMPLE:
#endif
3038
	case KVM_CAP_IOEVENTFD_ANY_LENGTH:
3039
	case KVM_CAP_CHECK_EXTENSION_VM:
3040
	case KVM_CAP_ENABLE_CAP_VM:
3041 3042 3043
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
	case KVM_CAP_MANUAL_DIRTY_LOG_PROTECT:
#endif
3044
		return 1;
3045
#ifdef CONFIG_KVM_MMIO
3046 3047
	case KVM_CAP_COALESCED_MMIO:
		return KVM_COALESCED_MMIO_PAGE_OFFSET;
P
Peng Hao 已提交
3048 3049
	case KVM_CAP_COALESCED_PIO:
		return 1;
3050
#endif
3051 3052 3053
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_CAP_IRQ_ROUTING:
		return KVM_MAX_IRQ_ROUTES;
3054 3055 3056 3057
#endif
#if KVM_ADDRESS_SPACE_NUM > 1
	case KVM_CAP_MULTI_ADDRESS_SPACE:
		return KVM_ADDRESS_SPACE_NUM;
3058
#endif
G
Greg Kurz 已提交
3059 3060
	case KVM_CAP_MAX_VCPU_ID:
		return KVM_MAX_VCPU_ID;
3061 3062 3063 3064 3065 3066
	default:
		break;
	}
	return kvm_vm_ioctl_check_extension(kvm, arg);
}

3067 3068 3069 3070 3071 3072 3073 3074 3075 3076
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) {
3077 3078 3079 3080 3081 3082 3083
#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
3084 3085 3086 3087 3088
	default:
		return kvm_vm_ioctl_enable_cap(kvm, cap);
	}
}

A
Avi Kivity 已提交
3089 3090 3091 3092 3093
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;
3094
	int r;
A
Avi Kivity 已提交
3095

3096 3097
	if (kvm->mm != current->mm)
		return -EIO;
A
Avi Kivity 已提交
3098 3099 3100 3101
	switch (ioctl) {
	case KVM_CREATE_VCPU:
		r = kvm_vm_ioctl_create_vcpu(kvm, arg);
		break;
3102 3103 3104 3105 3106 3107 3108 3109 3110
	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;
	}
3111 3112 3113 3114 3115
	case KVM_SET_USER_MEMORY_REGION: {
		struct kvm_userspace_memory_region kvm_userspace_mem;

		r = -EFAULT;
		if (copy_from_user(&kvm_userspace_mem, argp,
3116
						sizeof(kvm_userspace_mem)))
3117 3118
			goto out;

3119
		r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem);
A
Avi Kivity 已提交
3120 3121 3122 3123 3124 3125
		break;
	}
	case KVM_GET_DIRTY_LOG: {
		struct kvm_dirty_log log;

		r = -EFAULT;
3126
		if (copy_from_user(&log, argp, sizeof(log)))
A
Avi Kivity 已提交
3127
			goto out;
3128
		r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
A
Avi Kivity 已提交
3129 3130
		break;
	}
3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141
#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
3142
#ifdef CONFIG_KVM_MMIO
3143 3144
	case KVM_REGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3145

3146
		r = -EFAULT;
3147
		if (copy_from_user(&zone, argp, sizeof(zone)))
3148 3149 3150 3151 3152 3153
			goto out;
		r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
		break;
	}
	case KVM_UNREGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3154

3155
		r = -EFAULT;
3156
		if (copy_from_user(&zone, argp, sizeof(zone)))
3157 3158 3159 3160 3161
			goto out;
		r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
		break;
	}
#endif
G
Gregory Haskins 已提交
3162 3163 3164 3165
	case KVM_IRQFD: {
		struct kvm_irqfd data;

		r = -EFAULT;
3166
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3167
			goto out;
3168
		r = kvm_irqfd(kvm, &data);
G
Gregory Haskins 已提交
3169 3170
		break;
	}
G
Gregory Haskins 已提交
3171 3172 3173 3174
	case KVM_IOEVENTFD: {
		struct kvm_ioeventfd data;

		r = -EFAULT;
3175
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3176 3177 3178 3179
			goto out;
		r = kvm_ioeventfd(kvm, &data);
		break;
	}
3180 3181 3182 3183 3184
#ifdef CONFIG_HAVE_KVM_MSI
	case KVM_SIGNAL_MSI: {
		struct kvm_msi msi;

		r = -EFAULT;
3185
		if (copy_from_user(&msi, argp, sizeof(msi)))
3186 3187 3188 3189
			goto out;
		r = kvm_send_userspace_msi(kvm, &msi);
		break;
	}
3190 3191 3192 3193 3194 3195 3196
#endif
#ifdef __KVM_HAVE_IRQ_LINE
	case KVM_IRQ_LINE_STATUS:
	case KVM_IRQ_LINE: {
		struct kvm_irq_level irq_event;

		r = -EFAULT;
3197
		if (copy_from_user(&irq_event, argp, sizeof(irq_event)))
3198 3199
			goto out;

3200 3201
		r = kvm_vm_ioctl_irq_line(kvm, &irq_event,
					ioctl == KVM_IRQ_LINE_STATUS);
3202 3203 3204 3205 3206
		if (r)
			goto out;

		r = -EFAULT;
		if (ioctl == KVM_IRQ_LINE_STATUS) {
3207
			if (copy_to_user(argp, &irq_event, sizeof(irq_event)))
3208 3209 3210 3211 3212 3213
				goto out;
		}

		r = 0;
		break;
	}
3214
#endif
3215 3216 3217 3218
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_SET_GSI_ROUTING: {
		struct kvm_irq_routing routing;
		struct kvm_irq_routing __user *urouting;
3219
		struct kvm_irq_routing_entry *entries = NULL;
3220 3221 3222 3223 3224

		r = -EFAULT;
		if (copy_from_user(&routing, argp, sizeof(routing)))
			goto out;
		r = -EINVAL;
3225 3226
		if (!kvm_arch_can_set_irq_routing(kvm))
			goto out;
3227
		if (routing.nr > KVM_MAX_IRQ_ROUTES)
3228 3229 3230
			goto out;
		if (routing.flags)
			goto out;
3231 3232
		if (routing.nr) {
			r = -ENOMEM;
3233 3234
			entries = vmalloc(array_size(sizeof(*entries),
						     routing.nr));
3235 3236 3237 3238 3239 3240 3241 3242
			if (!entries)
				goto out;
			r = -EFAULT;
			urouting = argp;
			if (copy_from_user(entries, urouting->entries,
					   routing.nr * sizeof(*entries)))
				goto out_free_irq_routing;
		}
3243 3244
		r = kvm_set_irq_routing(kvm, entries, routing.nr,
					routing.flags);
3245
out_free_irq_routing:
3246 3247 3248 3249
		vfree(entries);
		break;
	}
#endif /* CONFIG_HAVE_KVM_IRQ_ROUTING */
S
Scott Wood 已提交
3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267
	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;
	}
3268 3269 3270
	case KVM_CHECK_EXTENSION:
		r = kvm_vm_ioctl_check_extension_generic(kvm, arg);
		break;
3271
	default:
3272
		r = kvm_arch_vm_ioctl(filp, ioctl, arg);
3273 3274 3275 3276 3277
	}
out:
	return r;
}

3278
#ifdef CONFIG_KVM_COMPAT
3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302
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)))
3303
			return -EFAULT;
3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318
		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

3319
static struct file_operations kvm_vm_fops = {
3320 3321
	.release        = kvm_vm_release,
	.unlocked_ioctl = kvm_vm_ioctl,
3322
	.llseek		= noop_llseek,
3323
	KVM_COMPAT(kvm_vm_compat_ioctl),
3324 3325
};

3326
static int kvm_dev_ioctl_create_vm(unsigned long type)
3327
{
3328
	int r;
3329
	struct kvm *kvm;
3330
	struct file *file;
3331

3332
	kvm = kvm_create_vm(type);
3333 3334
	if (IS_ERR(kvm))
		return PTR_ERR(kvm);
3335
#ifdef CONFIG_KVM_MMIO
3336
	r = kvm_coalesced_mmio_init(kvm);
3337 3338
	if (r < 0)
		goto put_kvm;
3339
#endif
3340
	r = get_unused_fd_flags(O_CLOEXEC);
3341 3342 3343
	if (r < 0)
		goto put_kvm;

3344 3345 3346
	file = anon_inode_getfile("kvm-vm", &kvm_vm_fops, kvm, O_RDWR);
	if (IS_ERR(file)) {
		put_unused_fd(r);
3347 3348
		r = PTR_ERR(file);
		goto put_kvm;
3349
	}
3350

3351 3352 3353 3354 3355 3356
	/*
	 * 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).
	 */
3357
	if (kvm_create_vm_debugfs(kvm, r) < 0) {
3358 3359
		put_unused_fd(r);
		fput(file);
3360 3361
		return -ENOMEM;
	}
3362
	kvm_uevent_notify_change(KVM_EVENT_CREATE_VM, kvm);
3363

3364
	fd_install(r, file);
3365
	return r;
3366 3367 3368 3369

put_kvm:
	kvm_put_kvm(kvm);
	return r;
3370 3371 3372 3373 3374
}

static long kvm_dev_ioctl(struct file *filp,
			  unsigned int ioctl, unsigned long arg)
{
3375
	long r = -EINVAL;
3376 3377 3378

	switch (ioctl) {
	case KVM_GET_API_VERSION:
3379 3380
		if (arg)
			goto out;
3381 3382 3383
		r = KVM_API_VERSION;
		break;
	case KVM_CREATE_VM:
3384
		r = kvm_dev_ioctl_create_vm(arg);
3385
		break;
3386
	case KVM_CHECK_EXTENSION:
3387
		r = kvm_vm_ioctl_check_extension_generic(NULL, arg);
3388
		break;
3389 3390 3391
	case KVM_GET_VCPU_MMAP_SIZE:
		if (arg)
			goto out;
3392 3393 3394
		r = PAGE_SIZE;     /* struct kvm_run */
#ifdef CONFIG_X86
		r += PAGE_SIZE;    /* pio data page */
3395
#endif
3396
#ifdef CONFIG_KVM_MMIO
3397
		r += PAGE_SIZE;    /* coalesced mmio ring page */
3398
#endif
3399
		break;
3400 3401 3402
	case KVM_TRACE_ENABLE:
	case KVM_TRACE_PAUSE:
	case KVM_TRACE_DISABLE:
3403
		r = -EOPNOTSUPP;
3404
		break;
A
Avi Kivity 已提交
3405
	default:
3406
		return kvm_arch_dev_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
3407 3408 3409 3410 3411 3412 3413
	}
out:
	return r;
}

static struct file_operations kvm_chardev_ops = {
	.unlocked_ioctl = kvm_dev_ioctl,
3414
	.llseek		= noop_llseek,
3415
	KVM_COMPAT(kvm_dev_ioctl),
A
Avi Kivity 已提交
3416 3417 3418
};

static struct miscdevice kvm_dev = {
A
Avi Kivity 已提交
3419
	KVM_MINOR,
A
Avi Kivity 已提交
3420 3421 3422 3423
	"kvm",
	&kvm_chardev_ops,
};

3424
static void hardware_enable_nolock(void *junk)
3425 3426
{
	int cpu = raw_smp_processor_id();
3427
	int r;
3428

3429
	if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
3430
		return;
3431

3432
	cpumask_set_cpu(cpu, cpus_hardware_enabled);
3433

3434
	r = kvm_arch_hardware_enable();
3435 3436 3437 3438

	if (r) {
		cpumask_clear_cpu(cpu, cpus_hardware_enabled);
		atomic_inc(&hardware_enable_failed);
X
Xiubo Li 已提交
3439
		pr_info("kvm: enabling virtualization on CPU%d failed\n", cpu);
3440
	}
3441 3442
}

3443
static int kvm_starting_cpu(unsigned int cpu)
3444
{
3445
	raw_spin_lock(&kvm_count_lock);
3446 3447
	if (kvm_usage_count)
		hardware_enable_nolock(NULL);
3448
	raw_spin_unlock(&kvm_count_lock);
3449
	return 0;
3450 3451 3452
}

static void hardware_disable_nolock(void *junk)
3453 3454 3455
{
	int cpu = raw_smp_processor_id();

3456
	if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
3457
		return;
3458
	cpumask_clear_cpu(cpu, cpus_hardware_enabled);
3459
	kvm_arch_hardware_disable();
3460 3461
}

3462
static int kvm_dying_cpu(unsigned int cpu)
3463
{
3464
	raw_spin_lock(&kvm_count_lock);
3465 3466
	if (kvm_usage_count)
		hardware_disable_nolock(NULL);
3467
	raw_spin_unlock(&kvm_count_lock);
3468
	return 0;
3469 3470
}

3471 3472 3473 3474 3475 3476
static void hardware_disable_all_nolock(void)
{
	BUG_ON(!kvm_usage_count);

	kvm_usage_count--;
	if (!kvm_usage_count)
3477
		on_each_cpu(hardware_disable_nolock, NULL, 1);
3478 3479 3480 3481
}

static void hardware_disable_all(void)
{
3482
	raw_spin_lock(&kvm_count_lock);
3483
	hardware_disable_all_nolock();
3484
	raw_spin_unlock(&kvm_count_lock);
3485 3486 3487 3488 3489 3490
}

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

3491
	raw_spin_lock(&kvm_count_lock);
3492 3493 3494 3495

	kvm_usage_count++;
	if (kvm_usage_count == 1) {
		atomic_set(&hardware_enable_failed, 0);
3496
		on_each_cpu(hardware_enable_nolock, NULL, 1);
3497 3498 3499 3500 3501 3502 3503

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

3504
	raw_spin_unlock(&kvm_count_lock);
3505 3506 3507 3508

	return r;
}

3509
static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
M
Mike Day 已提交
3510
		      void *v)
3511
{
3512 3513 3514 3515 3516 3517
	/*
	 * 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
3518
	pr_info("kvm: exiting hardware virtualization\n");
3519
	kvm_rebooting = true;
3520
	on_each_cpu(hardware_disable_nolock, NULL, 1);
3521 3522 3523 3524 3525 3526 3527 3528
	return NOTIFY_OK;
}

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

M
Marcelo Tosatti 已提交
3529
static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
3530 3531 3532 3533
{
	int i;

	for (i = 0; i < bus->dev_count; i++) {
3534
		struct kvm_io_device *pos = bus->range[i].dev;
3535 3536 3537

		kvm_iodevice_destructor(pos);
	}
M
Marcelo Tosatti 已提交
3538
	kfree(bus);
3539 3540
}

3541
static inline int kvm_io_bus_cmp(const struct kvm_io_range *r1,
X
Xiubo Li 已提交
3542
				 const struct kvm_io_range *r2)
3543
{
J
Jason Wang 已提交
3544 3545 3546 3547
	gpa_t addr1 = r1->addr;
	gpa_t addr2 = r2->addr;

	if (addr1 < addr2)
3548
		return -1;
J
Jason Wang 已提交
3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560

	/* 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)
3561
		return 1;
J
Jason Wang 已提交
3562

3563 3564 3565
	return 0;
}

3566 3567
static int kvm_io_bus_sort_cmp(const void *p1, const void *p2)
{
3568
	return kvm_io_bus_cmp(p1, p2);
3569 3570
}

G
Geoff Levand 已提交
3571
static int kvm_io_bus_get_first_dev(struct kvm_io_bus *bus,
3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588
			     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;

3589
	while (off > 0 && kvm_io_bus_cmp(&key, &bus->range[off-1]) == 0)
3590 3591 3592 3593 3594
		off--;

	return off;
}

3595
static int __kvm_io_bus_write(struct kvm_vcpu *vcpu, struct kvm_io_bus *bus,
C
Cornelia Huck 已提交
3596 3597 3598 3599 3600 3601 3602 3603 3604
			      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 &&
3605
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
3606
		if (!kvm_iodevice_write(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
3607 3608 3609 3610 3611 3612 3613 3614
					range->len, val))
			return idx;
		idx++;
	}

	return -EOPNOTSUPP;
}

3615
/* kvm_io_bus_write - called under kvm->slots_lock */
3616
int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
3617
		     int len, const void *val)
3618
{
3619
	struct kvm_io_bus *bus;
3620
	struct kvm_io_range range;
C
Cornelia Huck 已提交
3621
	int r;
3622 3623 3624 3625 3626

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
3627

3628
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3629 3630
	if (!bus)
		return -ENOMEM;
3631
	r = __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3632 3633 3634 3635
	return r < 0 ? r : 0;
}

/* kvm_io_bus_write_cookie - called under kvm->slots_lock */
3636 3637
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
3638 3639 3640 3641 3642 3643 3644 3645 3646
{
	struct kvm_io_bus *bus;
	struct kvm_io_range range;

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

3647
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3648 3649
	if (!bus)
		return -ENOMEM;
C
Cornelia Huck 已提交
3650 3651 3652

	/* First try the device referenced by cookie. */
	if ((cookie >= 0) && (cookie < bus->dev_count) &&
3653
	    (kvm_io_bus_cmp(&range, &bus->range[cookie]) == 0))
3654
		if (!kvm_iodevice_write(vcpu, bus->range[cookie].dev, addr, len,
C
Cornelia Huck 已提交
3655 3656 3657 3658 3659 3660 3661
					val))
			return cookie;

	/*
	 * cookie contained garbage; fall back to search and return the
	 * correct cookie value.
	 */
3662
	return __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3663 3664
}

3665 3666
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
3667 3668 3669 3670
{
	int idx;

	idx = kvm_io_bus_get_first_dev(bus, range->addr, range->len);
3671 3672 3673 3674
	if (idx < 0)
		return -EOPNOTSUPP;

	while (idx < bus->dev_count &&
3675
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
3676
		if (!kvm_iodevice_read(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
3677 3678
				       range->len, val))
			return idx;
3679 3680 3681
		idx++;
	}

3682 3683
	return -EOPNOTSUPP;
}
3684
EXPORT_SYMBOL_GPL(kvm_io_bus_write);
3685

3686
/* kvm_io_bus_read - called under kvm->slots_lock */
3687
int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
M
Marcelo Tosatti 已提交
3688
		    int len, void *val)
3689
{
3690
	struct kvm_io_bus *bus;
3691
	struct kvm_io_range range;
C
Cornelia Huck 已提交
3692
	int r;
3693 3694 3695 3696 3697

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
M
Marcelo Tosatti 已提交
3698

3699
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3700 3701
	if (!bus)
		return -ENOMEM;
3702
	r = __kvm_io_bus_read(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3703 3704
	return r < 0 ? r : 0;
}
3705

3706

3707
/* Caller must hold slots_lock. */
3708 3709
int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
			    int len, struct kvm_io_device *dev)
3710
{
3711
	int i;
M
Marcelo Tosatti 已提交
3712
	struct kvm_io_bus *new_bus, *bus;
3713
	struct kvm_io_range range;
3714

3715
	bus = kvm_get_bus(kvm, bus_idx);
3716 3717 3718
	if (!bus)
		return -ENOMEM;

3719 3720
	/* exclude ioeventfd which is limited by maximum fd */
	if (bus->dev_count - bus->ioeventfd_count > NR_IOBUS_DEVS - 1)
3721
		return -ENOSPC;
3722

3723
	new_bus = kmalloc(struct_size(bus, range, bus->dev_count + 1),
3724
			  GFP_KERNEL_ACCOUNT);
M
Marcelo Tosatti 已提交
3725 3726
	if (!new_bus)
		return -ENOMEM;
3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742

	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
3743 3744 3745
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
3746 3747 3748 3749

	return 0;
}

3750
/* Caller must hold slots_lock. */
3751 3752
void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
			       struct kvm_io_device *dev)
3753
{
3754
	int i;
M
Marcelo Tosatti 已提交
3755
	struct kvm_io_bus *new_bus, *bus;
3756

3757
	bus = kvm_get_bus(kvm, bus_idx);
3758
	if (!bus)
3759
		return;
3760

3761 3762
	for (i = 0; i < bus->dev_count; i++)
		if (bus->range[i].dev == dev) {
3763 3764
			break;
		}
M
Marcelo Tosatti 已提交
3765

3766 3767
	if (i == bus->dev_count)
		return;
3768

3769
	new_bus = kmalloc(struct_size(bus, range, bus->dev_count - 1),
3770
			  GFP_KERNEL_ACCOUNT);
3771 3772 3773 3774
	if (!new_bus)  {
		pr_err("kvm: failed to shrink bus, removing it completely\n");
		goto broken;
	}
3775 3776 3777 3778 3779

	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
3780

3781
broken:
M
Marcelo Tosatti 已提交
3782 3783 3784
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
3785
	return;
3786 3787
}

3788 3789 3790 3791 3792 3793 3794 3795 3796 3797
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);
3798 3799
	if (!bus)
		goto out_unlock;
3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813

	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);

3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825
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.
	 */
3826
	if (!refcount_inc_not_zero(&stat_data->kvm->users_count))
3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851
		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;

3852
	*val = *(ulong *)((void *)stat_data->kvm + stat_data->offset);
3853 3854 3855 3856

	return 0;
}

3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868
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;
}

3869 3870 3871 3872
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,
3873
				vm_stat_clear_per_vm, "%llu\n");
3874 3875 3876 3877 3878 3879 3880 3881
}

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,
3882
	.llseek  = no_llseek,
3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893
};

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)
3894
		*val += *(u64 *)((void *)vcpu + stat_data->offset);
3895 3896 3897 3898

	return 0;
}

3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913
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;
}

3914 3915 3916 3917
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,
3918
				 vcpu_stat_clear_per_vm, "%llu\n");
3919 3920 3921 3922 3923 3924 3925 3926
}

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,
3927
	.llseek  = no_llseek,
3928 3929 3930 3931 3932 3933 3934
};

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,
};

3935
static int vm_stat_get(void *_offset, u64 *val)
3936 3937 3938
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
3939 3940
	struct kvm_stat_data stat_tmp = {.offset = offset};
	u64 tmp_val;
3941

3942
	*val = 0;
3943
	spin_lock(&kvm_lock);
3944 3945 3946 3947 3948
	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;
	}
3949
	spin_unlock(&kvm_lock);
3950
	return 0;
3951 3952
}

3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972
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");
3973

3974
static int vcpu_stat_get(void *_offset, u64 *val)
A
Avi Kivity 已提交
3975 3976 3977
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
3978 3979
	struct kvm_stat_data stat_tmp = {.offset = offset};
	u64 tmp_val;
A
Avi Kivity 已提交
3980

3981
	*val = 0;
3982
	spin_lock(&kvm_lock);
3983 3984 3985 3986 3987
	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;
	}
3988
	spin_unlock(&kvm_lock);
3989
	return 0;
A
Avi Kivity 已提交
3990 3991
}

3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012
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");
4013

4014
static const struct file_operations *stat_fops[] = {
4015 4016 4017
	[KVM_STAT_VCPU] = &vcpu_stat_fops,
	[KVM_STAT_VM]   = &vm_stat_fops,
};
A
Avi Kivity 已提交
4018

4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037
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);

4038
	env = kzalloc(sizeof(*env), GFP_KERNEL_ACCOUNT);
4039 4040 4041 4042 4043 4044
	if (!env)
		return;

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

4045
	if (type == KVM_EVENT_CREATE_VM) {
4046
		add_uevent_var(env, "EVENT=create");
4047 4048
		kvm->userspace_pid = task_pid_nr(current);
	} else if (type == KVM_EVENT_DESTROY_VM) {
4049
		add_uevent_var(env, "EVENT=destroy");
4050 4051
	}
	add_uevent_var(env, "PID=%d", kvm->userspace_pid);
4052 4053

	if (kvm->debugfs_dentry) {
4054
		char *tmp, *p = kmalloc(PATH_MAX, GFP_KERNEL_ACCOUNT);
4055 4056 4057 4058 4059 4060

		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);
4061 4062 4063 4064 4065 4066 4067 4068
		}
	}
	/* 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);
}

4069
static void kvm_init_debug(void)
A
Avi Kivity 已提交
4070 4071 4072
{
	struct kvm_stats_debugfs_item *p;

4073
	kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
4074

4075 4076
	kvm_debugfs_num_entries = 0;
	for (p = debugfs_entries; p->name; ++p, kvm_debugfs_num_entries++) {
4077 4078 4079
		debugfs_create_file(p->name, 0644, kvm_debugfs_dir,
				    (void *)(long)p->offset,
				    stat_fops[p->kind]);
4080
	}
A
Avi Kivity 已提交
4081 4082
}

4083
static int kvm_suspend(void)
4084
{
4085
	if (kvm_usage_count)
4086
		hardware_disable_nolock(NULL);
4087 4088 4089
	return 0;
}

4090
static void kvm_resume(void)
4091
{
4092
	if (kvm_usage_count) {
4093
		WARN_ON(raw_spin_is_locked(&kvm_count_lock));
4094
		hardware_enable_nolock(NULL);
4095
	}
4096 4097
}

4098
static struct syscore_ops kvm_syscore_ops = {
4099 4100 4101 4102
	.suspend = kvm_suspend,
	.resume = kvm_resume,
};

4103 4104 4105 4106 4107 4108 4109 4110 4111
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);
4112

4113 4114
	if (vcpu->preempted)
		vcpu->preempted = false;
4115

R
Radim Krčmář 已提交
4116 4117
	kvm_arch_sched_in(vcpu, cpu);

4118
	kvm_arch_vcpu_load(vcpu, cpu);
4119 4120 4121 4122 4123 4124 4125
}

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

4126 4127
	if (current->state == TASK_RUNNING)
		vcpu->preempted = true;
4128
	kvm_arch_vcpu_put(vcpu);
4129 4130
}

4131
int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
4132
		  struct module *module)
A
Avi Kivity 已提交
4133 4134
{
	int r;
Y
Yang, Sheng 已提交
4135
	int cpu;
A
Avi Kivity 已提交
4136

4137 4138
	r = kvm_arch_init(opaque);
	if (r)
4139
		goto out_fail;
4140

4141 4142 4143 4144
	/*
	 * 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
4145 4146
	 * kvm_arch_init must be called before kvm_irqfd_init to avoid creating
	 * conflicts in case kvm is already setup for another implementation.
4147
	 */
P
Paolo Bonzini 已提交
4148 4149 4150
	r = kvm_irqfd_init();
	if (r)
		goto out_irqfd;
4151

4152
	if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
4153 4154 4155 4156
		r = -ENOMEM;
		goto out_free_0;
	}

4157
	r = kvm_arch_hardware_setup();
A
Avi Kivity 已提交
4158
	if (r < 0)
4159
		goto out_free_0a;
A
Avi Kivity 已提交
4160

Y
Yang, Sheng 已提交
4161 4162
	for_each_online_cpu(cpu) {
		smp_call_function_single(cpu,
4163
				kvm_arch_check_processor_compat,
4164
				&r, 1);
Y
Yang, Sheng 已提交
4165
		if (r < 0)
4166
			goto out_free_1;
Y
Yang, Sheng 已提交
4167 4168
	}

T
Thomas Gleixner 已提交
4169
	r = cpuhp_setup_state_nocalls(CPUHP_AP_KVM_STARTING, "kvm/cpu:starting",
4170
				      kvm_starting_cpu, kvm_dying_cpu);
A
Avi Kivity 已提交
4171
	if (r)
4172
		goto out_free_2;
A
Avi Kivity 已提交
4173 4174
	register_reboot_notifier(&kvm_reboot_notifier);

4175
	/* A kmem cache lets us meet the alignment requirements of fx_save. */
4176 4177
	if (!vcpu_align)
		vcpu_align = __alignof__(struct kvm_vcpu);
4178 4179 4180 4181 4182 4183
	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);
4184 4185
	if (!kvm_vcpu_cache) {
		r = -ENOMEM;
4186
		goto out_free_3;
4187 4188
	}

4189 4190 4191 4192
	r = kvm_async_pf_init();
	if (r)
		goto out_free;

A
Avi Kivity 已提交
4193
	kvm_chardev_ops.owner = module;
4194 4195
	kvm_vm_fops.owner = module;
	kvm_vcpu_fops.owner = module;
A
Avi Kivity 已提交
4196 4197 4198

	r = misc_register(&kvm_dev);
	if (r) {
X
Xiubo Li 已提交
4199
		pr_err("kvm: misc device register failed\n");
4200
		goto out_unreg;
A
Avi Kivity 已提交
4201 4202
	}

4203 4204
	register_syscore_ops(&kvm_syscore_ops);

4205 4206 4207
	kvm_preempt_ops.sched_in = kvm_sched_in;
	kvm_preempt_ops.sched_out = kvm_sched_out;

4208
	kvm_init_debug();
4209

P
Paolo Bonzini 已提交
4210 4211 4212
	r = kvm_vfio_ops_init();
	WARN_ON(r);

4213
	return 0;
A
Avi Kivity 已提交
4214

4215 4216
out_unreg:
	kvm_async_pf_deinit();
A
Avi Kivity 已提交
4217
out_free:
4218
	kmem_cache_destroy(kvm_vcpu_cache);
4219
out_free_3:
A
Avi Kivity 已提交
4220
	unregister_reboot_notifier(&kvm_reboot_notifier);
4221
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4222 4223
out_free_2:
out_free_1:
4224
	kvm_arch_hardware_unsetup();
4225 4226
out_free_0a:
	free_cpumask_var(cpus_hardware_enabled);
4227
out_free_0:
4228
	kvm_irqfd_exit();
P
Paolo Bonzini 已提交
4229
out_irqfd:
4230 4231
	kvm_arch_exit();
out_fail:
A
Avi Kivity 已提交
4232 4233
	return r;
}
4234
EXPORT_SYMBOL_GPL(kvm_init);
A
Avi Kivity 已提交
4235

4236
void kvm_exit(void)
A
Avi Kivity 已提交
4237
{
4238
	debugfs_remove_recursive(kvm_debugfs_dir);
A
Avi Kivity 已提交
4239
	misc_deregister(&kvm_dev);
4240
	kmem_cache_destroy(kvm_vcpu_cache);
4241
	kvm_async_pf_deinit();
4242
	unregister_syscore_ops(&kvm_syscore_ops);
A
Avi Kivity 已提交
4243
	unregister_reboot_notifier(&kvm_reboot_notifier);
4244
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4245
	on_each_cpu(hardware_disable_nolock, NULL, 1);
4246
	kvm_arch_hardware_unsetup();
4247
	kvm_arch_exit();
4248
	kvm_irqfd_exit();
4249
	free_cpumask_var(cpus_hardware_enabled);
4250
	kvm_vfio_ops_exit();
A
Avi Kivity 已提交
4251
}
4252
EXPORT_SYMBOL_GPL(kvm_exit);