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 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233
			if (!dirty_bitmap[i])
				continue;

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

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

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

/**
 * kvm_clear_dirty_log_protect - clear dirty bits in the bitmap
 *	and reenable dirty page tracking for the corresponding pages.
 * @kvm:	pointer to kvm instance
 * @log:	slot id and address from which to fetch the bitmap of dirty pages
 */
int kvm_clear_dirty_log_protect(struct kvm *kvm,
				struct kvm_clear_dirty_log *log, bool *flush)
{
	struct kvm_memslots *slots;
	struct kvm_memory_slot *memslot;
1234
	int as_id, id;
1235
	gfn_t offset;
1236
	unsigned long i, n;
1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255
	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);
1256 1257 1258 1259 1260

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

1261
	*flush = false;
1262 1263 1264
	dirty_bitmap_buffer = kvm_second_dirty_bitmap(memslot);
	if (copy_from_user(dirty_bitmap_buffer, log->dirty_bitmap, n))
		return -EFAULT;
1265

1266 1267 1268 1269 1270 1271 1272
	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)
1273 1274
			continue;

1275
		mask &= atomic_long_fetch_andnot(mask, p);
1276

1277 1278 1279 1280 1281 1282
		/*
		 * 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.
		*/
1283
		if (mask) {
1284
			*flush = true;
1285 1286 1287
			kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot,
								offset, mask);
		}
1288 1289
	}
	spin_unlock(&kvm->mmu_lock);
1290

1291
	return 0;
1292
}
1293
EXPORT_SYMBOL_GPL(kvm_clear_dirty_log_protect);
1294 1295
#endif

1296 1297 1298 1299 1300
bool kvm_largepages_enabled(void)
{
	return largepages_enabled;
}

1301 1302 1303 1304 1305 1306
void kvm_disable_largepages(void)
{
	largepages_enabled = false;
}
EXPORT_SYMBOL_GPL(kvm_disable_largepages);

1307 1308 1309 1310
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
1311
EXPORT_SYMBOL_GPL(gfn_to_memslot);
A
Avi Kivity 已提交
1312

1313 1314 1315 1316 1317
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);
}

1318
bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
1319
{
1320
	struct kvm_memory_slot *memslot = gfn_to_memslot(kvm, gfn);
1321

1322
	if (!memslot || memslot->id >= KVM_USER_MEM_SLOTS ||
1323
	      memslot->flags & KVM_MEMSLOT_INVALID)
1324
		return false;
1325

1326
	return true;
1327 1328 1329
}
EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);

J
Joerg Roedel 已提交
1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353
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
1354 1355 1356 1357 1358 1359 1360
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
1361
{
1362
	if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
X
Xiao Guangrong 已提交
1363
		return KVM_HVA_ERR_BAD;
1364

X
Xiao Guangrong 已提交
1365 1366
	if (memslot_is_readonly(slot) && write)
		return KVM_HVA_ERR_RO_BAD;
1367 1368 1369 1370

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

X
Xiao Guangrong 已提交
1371
	return __gfn_to_hva_memslot(slot, gfn);
I
Izik Eidus 已提交
1372
}
1373

X
Xiao Guangrong 已提交
1374 1375 1376 1377
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
1378
}
1379

X
Xiao Guangrong 已提交
1380
unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot,
1381
					gfn_t gfn)
X
Xiao Guangrong 已提交
1382 1383 1384 1385 1386
{
	return gfn_to_hva_many(slot, gfn, NULL);
}
EXPORT_SYMBOL_GPL(gfn_to_hva_memslot);

1387 1388
unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
{
1389
	return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL);
1390
}
1391
EXPORT_SYMBOL_GPL(gfn_to_hva);
I
Izik Eidus 已提交
1392

1393 1394 1395 1396 1397 1398
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);

1399
/*
1400 1401 1402 1403 1404 1405
 * 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
1406
 */
1407 1408
unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot,
				      gfn_t gfn, bool *writable)
1409
{
1410 1411 1412
	unsigned long hva = __gfn_to_hva_many(slot, gfn, NULL, false);

	if (!kvm_is_error_hva(hva) && writable)
1413 1414
		*writable = !memslot_is_readonly(slot);

1415
	return hva;
1416 1417
}

1418 1419 1420 1421 1422 1423 1424
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);
}

1425 1426 1427 1428 1429 1430 1431
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);
}

1432 1433
static inline int check_user_page_hwpoison(unsigned long addr)
{
L
Lorenzo Stoakes 已提交
1434
	int rc, flags = FOLL_HWPOISON | FOLL_WRITE;
1435

L
Lorenzo Stoakes 已提交
1436
	rc = get_user_pages(addr, 1, flags, NULL, NULL);
1437 1438 1439
	return rc == -EHWPOISON;
}

X
Xiao Guangrong 已提交
1440
/*
1441 1442 1443
 * 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
1444
 */
1445 1446
static bool hva_to_pfn_fast(unsigned long addr, bool write_fault,
			    bool *writable, kvm_pfn_t *pfn)
A
Avi Kivity 已提交
1447
{
1448
	struct page *page[1];
X
Xiao Guangrong 已提交
1449
	int npages;
A
Avi Kivity 已提交
1450

1451 1452 1453 1454 1455 1456 1457
	/*
	 * 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;
1458

X
Xiao Guangrong 已提交
1459 1460 1461
	npages = __get_user_pages_fast(addr, 1, 1, page);
	if (npages == 1) {
		*pfn = page_to_pfn(page[0]);
1462

X
Xiao Guangrong 已提交
1463 1464 1465 1466
		if (writable)
			*writable = true;
		return true;
	}
1467

X
Xiao Guangrong 已提交
1468 1469
	return false;
}
1470

X
Xiao Guangrong 已提交
1471 1472 1473 1474 1475
/*
 * 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
1476
			   bool *writable, kvm_pfn_t *pfn)
X
Xiao Guangrong 已提交
1477
{
1478 1479
	unsigned int flags = FOLL_HWPOISON;
	struct page *page;
X
Xiao Guangrong 已提交
1480
	int npages = 0;
1481

X
Xiao Guangrong 已提交
1482 1483 1484 1485 1486
	might_sleep();

	if (writable)
		*writable = write_fault;

1487 1488 1489 1490
	if (write_fault)
		flags |= FOLL_WRITE;
	if (async)
		flags |= FOLL_NOWAIT;
1491

1492
	npages = get_user_pages_unlocked(addr, 1, &page, flags);
X
Xiao Guangrong 已提交
1493 1494 1495 1496
	if (npages != 1)
		return npages;

	/* map read fault as writable if possible */
1497
	if (unlikely(!write_fault) && writable) {
1498
		struct page *wpage;
X
Xiao Guangrong 已提交
1499

1500
		if (__get_user_pages_fast(addr, 1, 1, &wpage) == 1) {
X
Xiao Guangrong 已提交
1501
			*writable = true;
1502 1503
			put_page(page);
			page = wpage;
1504
		}
1505
	}
1506
	*pfn = page_to_pfn(page);
X
Xiao Guangrong 已提交
1507 1508
	return npages;
}
I
Izik Eidus 已提交
1509

X
Xiao Guangrong 已提交
1510 1511 1512 1513
static bool vma_is_valid(struct vm_area_struct *vma, bool write_fault)
{
	if (unlikely(!(vma->vm_flags & VM_READ)))
		return false;
1514

X
Xiao Guangrong 已提交
1515 1516
	if (write_fault && (unlikely(!(vma->vm_flags & VM_WRITE))))
		return false;
1517

X
Xiao Guangrong 已提交
1518 1519
	return true;
}
1520

1521 1522
static int hva_to_pfn_remapped(struct vm_area_struct *vma,
			       unsigned long addr, bool *async,
1523 1524
			       bool write_fault, bool *writable,
			       kvm_pfn_t *p_pfn)
1525
{
1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549
	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;

	}

1550 1551
	if (writable)
		*writable = true;
1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566

	/*
	 * 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;
1567 1568 1569
	return 0;
}

1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583
/*
 * 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
1584
static kvm_pfn_t hva_to_pfn(unsigned long addr, bool atomic, bool *async,
X
Xiao Guangrong 已提交
1585 1586 1587
			bool write_fault, bool *writable)
{
	struct vm_area_struct *vma;
D
Dan Williams 已提交
1588
	kvm_pfn_t pfn = 0;
1589
	int npages, r;
1590

X
Xiao Guangrong 已提交
1591 1592
	/* we can do it either atomically or asynchronously, not both */
	BUG_ON(atomic && async);
1593

1594
	if (hva_to_pfn_fast(addr, write_fault, writable, &pfn))
X
Xiao Guangrong 已提交
1595 1596 1597 1598 1599 1600 1601 1602
		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;
1603

X
Xiao Guangrong 已提交
1604 1605 1606 1607 1608 1609 1610
	down_read(&current->mm->mmap_sem);
	if (npages == -EHWPOISON ||
	      (!async && check_user_page_hwpoison(addr))) {
		pfn = KVM_PFN_ERR_HWPOISON;
		goto exit;
	}

1611
retry:
X
Xiao Guangrong 已提交
1612 1613 1614 1615
	vma = find_vma_intersection(current->mm, addr, addr + 1);

	if (vma == NULL)
		pfn = KVM_PFN_ERR_FAULT;
1616
	else if (vma->vm_flags & (VM_IO | VM_PFNMAP)) {
1617
		r = hva_to_pfn_remapped(vma, addr, async, write_fault, writable, &pfn);
1618 1619
		if (r == -EAGAIN)
			goto retry;
1620 1621
		if (r < 0)
			pfn = KVM_PFN_ERR_FAULT;
X
Xiao Guangrong 已提交
1622
	} else {
X
Xiao Guangrong 已提交
1623
		if (async && vma_is_valid(vma, write_fault))
X
Xiao Guangrong 已提交
1624 1625 1626 1627 1628
			*async = true;
		pfn = KVM_PFN_ERR_FAULT;
	}
exit:
	up_read(&current->mm->mmap_sem);
1629
	return pfn;
1630 1631
}

D
Dan Williams 已提交
1632 1633 1634
kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
			       bool atomic, bool *async, bool write_fault,
			       bool *writable)
1635
{
X
Xiao Guangrong 已提交
1636 1637
	unsigned long addr = __gfn_to_hva_many(slot, gfn, NULL, write_fault);

1638 1639 1640
	if (addr == KVM_HVA_ERR_RO_BAD) {
		if (writable)
			*writable = false;
X
Xiao Guangrong 已提交
1641
		return KVM_PFN_ERR_RO_FAULT;
1642
	}
X
Xiao Guangrong 已提交
1643

1644 1645 1646
	if (kvm_is_error_hva(addr)) {
		if (writable)
			*writable = false;
1647
		return KVM_PFN_NOSLOT;
1648
	}
X
Xiao Guangrong 已提交
1649 1650 1651 1652 1653 1654 1655 1656 1657

	/* 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);
1658
}
1659
EXPORT_SYMBOL_GPL(__gfn_to_pfn_memslot);
1660

D
Dan Williams 已提交
1661
kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1662 1663
		      bool *writable)
{
P
Paolo Bonzini 已提交
1664 1665
	return __gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn, false, NULL,
				    write_fault, writable);
1666 1667 1668
}
EXPORT_SYMBOL_GPL(gfn_to_pfn_prot);

D
Dan Williams 已提交
1669
kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1670
{
X
Xiao Guangrong 已提交
1671
	return __gfn_to_pfn_memslot(slot, gfn, false, NULL, true, NULL);
1672
}
P
Paolo Bonzini 已提交
1673
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot);
1674

D
Dan Williams 已提交
1675
kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn)
1676
{
X
Xiao Guangrong 已提交
1677
	return __gfn_to_pfn_memslot(slot, gfn, true, NULL, true, NULL);
1678
}
1679
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot_atomic);
1680

D
Dan Williams 已提交
1681
kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1682 1683 1684 1685 1686
{
	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
1687
kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn)
1688 1689 1690 1691 1692
{
	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
1693
kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1694 1695 1696 1697 1698
{
	return gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn);
}
EXPORT_SYMBOL_GPL(gfn_to_pfn);

D
Dan Williams 已提交
1699
kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn)
1700 1701 1702 1703 1704
{
	return gfn_to_pfn_memslot(kvm_vcpu_gfn_to_memslot(vcpu, gfn), gfn);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_gfn_to_pfn);

1705 1706
int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
			    struct page **pages, int nr_pages)
1707 1708
{
	unsigned long addr;
1709
	gfn_t entry = 0;
1710

1711
	addr = gfn_to_hva_many(slot, gfn, &entry);
1712 1713 1714 1715 1716 1717 1718 1719 1720 1721
	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
1722
static struct page *kvm_pfn_to_page(kvm_pfn_t pfn)
X
Xiao Guangrong 已提交
1723
{
1724
	if (is_error_noslot_pfn(pfn))
1725
		return KVM_ERR_PTR_BAD_PAGE;
X
Xiao Guangrong 已提交
1726

1727
	if (kvm_is_reserved_pfn(pfn)) {
1728
		WARN_ON(1);
1729
		return KVM_ERR_PTR_BAD_PAGE;
1730
	}
X
Xiao Guangrong 已提交
1731 1732 1733 1734

	return pfn_to_page(pfn);
}

1735 1736
struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
{
D
Dan Williams 已提交
1737
	kvm_pfn_t pfn;
1738 1739 1740

	pfn = gfn_to_pfn(kvm, gfn);

X
Xiao Guangrong 已提交
1741
	return kvm_pfn_to_page(pfn);
A
Avi Kivity 已提交
1742 1743 1744
}
EXPORT_SYMBOL_GPL(gfn_to_page);

1745 1746
struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn)
{
D
Dan Williams 已提交
1747
	kvm_pfn_t pfn;
1748 1749 1750 1751 1752 1753 1754

	pfn = kvm_vcpu_gfn_to_pfn(vcpu, gfn);

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

1755 1756
void kvm_release_page_clean(struct page *page)
{
1757 1758
	WARN_ON(is_error_page(page));

1759
	kvm_release_pfn_clean(page_to_pfn(page));
1760 1761 1762
}
EXPORT_SYMBOL_GPL(kvm_release_page_clean);

D
Dan Williams 已提交
1763
void kvm_release_pfn_clean(kvm_pfn_t pfn)
1764
{
1765
	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn))
1766
		put_page(pfn_to_page(pfn));
1767 1768 1769
}
EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);

1770
void kvm_release_page_dirty(struct page *page)
1771
{
X
Xiao Guangrong 已提交
1772 1773
	WARN_ON(is_error_page(page));

1774 1775 1776 1777
	kvm_release_pfn_dirty(page_to_pfn(page));
}
EXPORT_SYMBOL_GPL(kvm_release_page_dirty);

1778
void kvm_release_pfn_dirty(kvm_pfn_t pfn)
1779 1780 1781 1782
{
	kvm_set_pfn_dirty(pfn);
	kvm_release_pfn_clean(pfn);
}
1783
EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
1784

D
Dan Williams 已提交
1785
void kvm_set_pfn_dirty(kvm_pfn_t pfn)
1786
{
1787
	if (!kvm_is_reserved_pfn(pfn)) {
1788
		struct page *page = pfn_to_page(pfn);
1789

1790 1791 1792
		if (!PageReserved(page))
			SetPageDirty(page);
	}
1793
}
1794 1795
EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);

D
Dan Williams 已提交
1796
void kvm_set_pfn_accessed(kvm_pfn_t pfn)
1797
{
1798
	if (!kvm_is_reserved_pfn(pfn))
1799
		mark_page_accessed(pfn_to_page(pfn));
1800 1801 1802
}
EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);

D
Dan Williams 已提交
1803
void kvm_get_pfn(kvm_pfn_t pfn)
1804
{
1805
	if (!kvm_is_reserved_pfn(pfn))
1806
		get_page(pfn_to_page(pfn));
1807 1808
}
EXPORT_SYMBOL_GPL(kvm_get_pfn);
1809

1810 1811 1812 1813 1814 1815 1816 1817
static int next_segment(unsigned long len, int offset)
{
	if (len > PAGE_SIZE - offset)
		return PAGE_SIZE - offset;
	else
		return len;
}

1818 1819
static int __kvm_read_guest_page(struct kvm_memory_slot *slot, gfn_t gfn,
				 void *data, int offset, int len)
1820
{
1821 1822
	int r;
	unsigned long addr;
1823

1824
	addr = gfn_to_hva_memslot_prot(slot, gfn, NULL);
1825 1826
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1827
	r = __copy_from_user(data, (void __user *)addr + offset, len);
1828
	if (r)
1829 1830 1831
		return -EFAULT;
	return 0;
}
1832 1833 1834 1835 1836 1837 1838 1839

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);
}
1840 1841
EXPORT_SYMBOL_GPL(kvm_read_guest_page);

1842 1843 1844 1845 1846 1847 1848 1849 1850
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);

1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870
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);

1871
int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data, unsigned long len)
1872 1873
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
1874
	int seg;
1875
	int offset = offset_in_page(gpa);
1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889
	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);
1890

1891 1892 1893 1894 1895 1896 1897
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);
1898 1899
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1900
	pagefault_disable();
1901
	r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
1902
	pagefault_enable();
1903 1904 1905 1906 1907
	if (r)
		return -EFAULT;
	return 0;
}

1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931
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)
1932
{
1933 1934
	int r;
	unsigned long addr;
1935

1936
	addr = gfn_to_hva_memslot(memslot, gfn);
1937 1938
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1939
	r = __copy_to_user((void __user *)addr + offset, data, len);
1940
	if (r)
1941
		return -EFAULT;
1942
	mark_page_dirty_in_slot(memslot, gfn);
1943 1944
	return 0;
}
1945 1946 1947 1948 1949 1950 1951 1952

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);
}
1953 1954
EXPORT_SYMBOL_GPL(kvm_write_guest_page);

1955 1956 1957 1958 1959 1960 1961 1962 1963
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);

1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982
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;
}
1983
EXPORT_SYMBOL_GPL(kvm_write_guest);
1984

1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
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);

2006 2007 2008
static int __kvm_gfn_to_hva_cache_init(struct kvm_memslots *slots,
				       struct gfn_to_hva_cache *ghc,
				       gpa_t gpa, unsigned long len)
2009 2010
{
	int offset = offset_in_page(gpa);
2011 2012 2013 2014
	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;
2015
	int r = start_gfn <= end_gfn ? 0 : -EINVAL;
2016 2017 2018

	ghc->gpa = gpa;
	ghc->generation = slots->generation;
2019
	ghc->len = len;
2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036
	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)
2037
		ghc->hva += offset;
2038
	else
2039
		ghc->memslot = NULL;
2040 2041

	return r;
2042
}
2043

2044
int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
2045 2046
			      gpa_t gpa, unsigned long len)
{
2047
	struct kvm_memslots *slots = kvm_memslots(kvm);
2048 2049
	return __kvm_gfn_to_hva_cache_init(slots, ghc, gpa, len);
}
2050
EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init);
2051

2052
int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
2053 2054
				  void *data, unsigned int offset,
				  unsigned long len)
2055
{
2056
	struct kvm_memslots *slots = kvm_memslots(kvm);
2057
	int r;
2058
	gpa_t gpa = ghc->gpa + offset;
2059

2060
	BUG_ON(len + offset > ghc->len);
2061

2062
	if (slots->generation != ghc->generation)
2063
		__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len);
2064 2065

	if (unlikely(!ghc->memslot))
2066
		return kvm_write_guest(kvm, gpa, data, len);
2067 2068 2069 2070

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

2071
	r = __copy_to_user((void __user *)ghc->hva + offset, data, len);
2072 2073
	if (r)
		return -EFAULT;
2074
	mark_page_dirty_in_slot(ghc->memslot, gpa >> PAGE_SHIFT);
2075 2076 2077

	return 0;
}
2078
EXPORT_SYMBOL_GPL(kvm_write_guest_offset_cached);
2079

2080 2081
int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
2082
{
2083
	return kvm_write_guest_offset_cached(kvm, ghc, data, 0, len);
2084
}
2085
EXPORT_SYMBOL_GPL(kvm_write_guest_cached);
2086

2087 2088
int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
2089
{
2090
	struct kvm_memslots *slots = kvm_memslots(kvm);
2091 2092
	int r;

2093 2094
	BUG_ON(len > ghc->len);

2095
	if (slots->generation != ghc->generation)
2096
		__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len);
2097 2098

	if (unlikely(!ghc->memslot))
2099
		return kvm_read_guest(kvm, ghc->gpa, data, len);
2100 2101 2102 2103 2104 2105 2106 2107 2108 2109

	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;
}
2110
EXPORT_SYMBOL_GPL(kvm_read_guest_cached);
2111

2112 2113
int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
{
2114 2115 2116
	const void *zero_page = (const void *) __va(page_to_phys(ZERO_PAGE(0)));

	return kvm_write_guest_page(kvm, gfn, zero_page, offset, len);
2117 2118 2119 2120 2121 2122 2123 2124 2125 2126
}
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;

2127
	while ((seg = next_segment(len, offset)) != 0) {
2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138
		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);

2139
static void mark_page_dirty_in_slot(struct kvm_memory_slot *memslot,
2140
				    gfn_t gfn)
A
Avi Kivity 已提交
2141
{
R
Rusty Russell 已提交
2142 2143
	if (memslot && memslot->dirty_bitmap) {
		unsigned long rel_gfn = gfn - memslot->base_gfn;
A
Avi Kivity 已提交
2144

2145
		set_bit_le(rel_gfn, memslot->dirty_bitmap);
A
Avi Kivity 已提交
2146 2147 2148
	}
}

2149 2150 2151 2152 2153
void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
{
	struct kvm_memory_slot *memslot;

	memslot = gfn_to_memslot(kvm, gfn);
2154
	mark_page_dirty_in_slot(memslot, gfn);
2155
}
2156
EXPORT_SYMBOL_GPL(mark_page_dirty);
2157

2158 2159 2160 2161 2162 2163 2164 2165 2166
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);

2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189
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
2190 2191
static void grow_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2192
	unsigned int old, val, grow, grow_start;
W
Wanpeng Li 已提交
2193

2194
	old = val = vcpu->halt_poll_ns;
2195
	grow_start = READ_ONCE(halt_poll_ns_grow_start);
2196
	grow = READ_ONCE(halt_poll_ns_grow);
2197 2198 2199
	if (!grow)
		goto out;

2200 2201 2202
	val *= grow;
	if (val < grow_start)
		val = grow_start;
W
Wanpeng Li 已提交
2203

2204 2205 2206
	if (val > halt_poll_ns)
		val = halt_poll_ns;

W
Wanpeng Li 已提交
2207
	vcpu->halt_poll_ns = val;
2208
out:
2209
	trace_kvm_halt_poll_ns_grow(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2210 2211 2212 2213
}

static void shrink_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2214
	unsigned int old, val, shrink;
W
Wanpeng Li 已提交
2215

2216
	old = val = vcpu->halt_poll_ns;
2217 2218
	shrink = READ_ONCE(halt_poll_ns_shrink);
	if (shrink == 0)
W
Wanpeng Li 已提交
2219 2220
		val = 0;
	else
2221
		val /= shrink;
W
Wanpeng Li 已提交
2222 2223

	vcpu->halt_poll_ns = val;
2224
	trace_kvm_halt_poll_ns_shrink(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2225 2226
}

2227 2228
static int kvm_vcpu_check_block(struct kvm_vcpu *vcpu)
{
2229 2230 2231
	int ret = -EINTR;
	int idx = srcu_read_lock(&vcpu->kvm->srcu);

2232 2233
	if (kvm_arch_vcpu_runnable(vcpu)) {
		kvm_make_request(KVM_REQ_UNHALT, vcpu);
2234
		goto out;
2235 2236
	}
	if (kvm_cpu_has_pending_timer(vcpu))
2237
		goto out;
2238
	if (signal_pending(current))
2239
		goto out;
2240

2241 2242 2243 2244
	ret = 0;
out:
	srcu_read_unlock(&vcpu->kvm->srcu, idx);
	return ret;
2245 2246
}

E
Eddie Dong 已提交
2247 2248 2249
/*
 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
 */
2250
void kvm_vcpu_block(struct kvm_vcpu *vcpu)
2251
{
2252
	ktime_t start, cur;
2253
	DECLARE_SWAITQUEUE(wait);
2254
	bool waited = false;
W
Wanpeng Li 已提交
2255
	u64 block_ns;
2256 2257

	start = cur = ktime_get();
W
Wanpeng Li 已提交
2258 2259
	if (vcpu->halt_poll_ns) {
		ktime_t stop = ktime_add_ns(ktime_get(), vcpu->halt_poll_ns);
2260

2261
		++vcpu->stat.halt_attempted_poll;
2262 2263 2264 2265 2266 2267 2268
		do {
			/*
			 * This sets KVM_REQ_UNHALT if an interrupt
			 * arrives.
			 */
			if (kvm_vcpu_check_block(vcpu) < 0) {
				++vcpu->stat.halt_successful_poll;
2269 2270
				if (!vcpu_valid_wakeup(vcpu))
					++vcpu->stat.halt_poll_invalid;
2271 2272 2273 2274 2275
				goto out;
			}
			cur = ktime_get();
		} while (single_task_running() && ktime_before(cur, stop));
	}
2276

2277 2278
	kvm_arch_vcpu_blocking(vcpu);

2279
	for (;;) {
2280
		prepare_to_swait_exclusive(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
2281

2282
		if (kvm_vcpu_check_block(vcpu) < 0)
2283 2284
			break;

2285
		waited = true;
E
Eddie Dong 已提交
2286 2287
		schedule();
	}
2288

2289
	finish_swait(&vcpu->wq, &wait);
2290 2291
	cur = ktime_get();

2292
	kvm_arch_vcpu_unblocking(vcpu);
2293
out:
W
Wanpeng Li 已提交
2294 2295
	block_ns = ktime_to_ns(cur) - ktime_to_ns(start);

2296 2297 2298
	if (!vcpu_valid_wakeup(vcpu))
		shrink_halt_poll_ns(vcpu);
	else if (halt_poll_ns) {
W
Wanpeng Li 已提交
2299 2300 2301
		if (block_ns <= vcpu->halt_poll_ns)
			;
		/* we had a long block, shrink polling */
2302
		else if (vcpu->halt_poll_ns && block_ns > halt_poll_ns)
W
Wanpeng Li 已提交
2303 2304 2305 2306 2307
			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);
2308 2309
	} else
		vcpu->halt_poll_ns = 0;
W
Wanpeng Li 已提交
2310

2311 2312
	trace_kvm_vcpu_wakeup(block_ns, waited, vcpu_valid_wakeup(vcpu));
	kvm_arch_vcpu_block_finish(vcpu);
E
Eddie Dong 已提交
2313
}
2314
EXPORT_SYMBOL_GPL(kvm_vcpu_block);
E
Eddie Dong 已提交
2315

2316
bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu)
2317
{
2318
	struct swait_queue_head *wqp;
2319 2320

	wqp = kvm_arch_vcpu_wq(vcpu);
2321
	if (swq_has_sleeper(wqp)) {
2322
		swake_up_one(wqp);
2323
		++vcpu->stat.halt_wakeup;
2324
		return true;
2325 2326
	}

2327
	return false;
2328 2329 2330
}
EXPORT_SYMBOL_GPL(kvm_vcpu_wake_up);

2331
#ifndef CONFIG_S390
2332 2333 2334 2335 2336 2337 2338 2339
/*
 * 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;

2340 2341 2342
	if (kvm_vcpu_wake_up(vcpu))
		return;

2343 2344 2345 2346 2347 2348
	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();
}
2349
EXPORT_SYMBOL_GPL(kvm_vcpu_kick);
2350
#endif /* !CONFIG_S390 */
2351

2352
int kvm_vcpu_yield_to(struct kvm_vcpu *target)
2353 2354 2355
{
	struct pid *pid;
	struct task_struct *task = NULL;
2356
	int ret = 0;
2357 2358 2359 2360

	rcu_read_lock();
	pid = rcu_dereference(target->pid);
	if (pid)
2361
		task = get_pid_task(pid, PIDTYPE_PID);
2362 2363
	rcu_read_unlock();
	if (!task)
2364 2365
		return ret;
	ret = yield_to(task, 1);
2366
	put_task_struct(task);
2367 2368

	return ret;
2369 2370 2371
}
EXPORT_SYMBOL_GPL(kvm_vcpu_yield_to);

2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393
/*
 * 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.
 */
2394
static bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
2395
{
2396
#ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
2397 2398 2399
	bool eligible;

	eligible = !vcpu->spin_loop.in_spin_loop ||
2400
		    vcpu->spin_loop.dy_eligible;
2401 2402 2403 2404 2405

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

	return eligible;
2406 2407
#else
	return true;
2408
#endif
2409
}
2410

2411
void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
Z
Zhai, Edwin 已提交
2412
{
2413 2414 2415 2416
	struct kvm *kvm = me->kvm;
	struct kvm_vcpu *vcpu;
	int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
	int yielded = 0;
2417
	int try = 3;
2418 2419
	int pass;
	int i;
Z
Zhai, Edwin 已提交
2420

2421
	kvm_vcpu_set_in_spin_loop(me, true);
2422 2423 2424 2425 2426 2427 2428
	/*
	 * 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.
	 */
2429
	for (pass = 0; pass < 2 && !yielded && try; pass++) {
2430
		kvm_for_each_vcpu(i, vcpu, kvm) {
2431
			if (!pass && i <= last_boosted_vcpu) {
2432 2433 2434 2435
				i = last_boosted_vcpu;
				continue;
			} else if (pass && i > last_boosted_vcpu)
				break;
2436
			if (!READ_ONCE(vcpu->preempted))
2437
				continue;
2438 2439
			if (vcpu == me)
				continue;
2440
			if (swait_active(&vcpu->wq) && !kvm_arch_vcpu_runnable(vcpu))
2441
				continue;
2442 2443
			if (yield_to_kernel_mode && !kvm_arch_vcpu_in_kernel(vcpu))
				continue;
2444 2445
			if (!kvm_vcpu_eligible_for_directed_yield(vcpu))
				continue;
2446 2447 2448

			yielded = kvm_vcpu_yield_to(vcpu);
			if (yielded > 0) {
2449 2450
				kvm->last_boosted_vcpu = i;
				break;
2451 2452 2453 2454
			} else if (yielded < 0) {
				try--;
				if (!try)
					break;
2455 2456 2457
			}
		}
	}
2458
	kvm_vcpu_set_in_spin_loop(me, false);
2459 2460 2461

	/* Ensure vcpu is not eligible during next spinloop */
	kvm_vcpu_set_dy_eligible(me, false);
Z
Zhai, Edwin 已提交
2462 2463 2464
}
EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin);

2465
static vm_fault_t kvm_vcpu_fault(struct vm_fault *vmf)
2466
{
2467
	struct kvm_vcpu *vcpu = vmf->vma->vm_file->private_data;
2468 2469
	struct page *page;

2470
	if (vmf->pgoff == 0)
2471
		page = virt_to_page(vcpu->run);
A
Avi Kivity 已提交
2472
#ifdef CONFIG_X86
2473
	else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
2474
		page = virt_to_page(vcpu->arch.pio_data);
2475
#endif
2476
#ifdef CONFIG_KVM_MMIO
2477 2478
	else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
		page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
A
Avi Kivity 已提交
2479
#endif
2480
	else
2481
		return kvm_arch_vcpu_fault(vcpu, vmf);
2482
	get_page(page);
2483 2484
	vmf->page = page;
	return 0;
2485 2486
}

2487
static const struct vm_operations_struct kvm_vcpu_vm_ops = {
2488
	.fault = kvm_vcpu_fault,
2489 2490 2491 2492 2493 2494 2495 2496
};

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
2497 2498 2499 2500
static int kvm_vcpu_release(struct inode *inode, struct file *filp)
{
	struct kvm_vcpu *vcpu = filp->private_data;

2501
	debugfs_remove_recursive(vcpu->debugfs_dentry);
A
Al Viro 已提交
2502
	kvm_put_kvm(vcpu->kvm);
A
Avi Kivity 已提交
2503 2504 2505
	return 0;
}

2506
static struct file_operations kvm_vcpu_fops = {
A
Avi Kivity 已提交
2507 2508
	.release        = kvm_vcpu_release,
	.unlocked_ioctl = kvm_vcpu_ioctl,
2509
	.mmap           = kvm_vcpu_mmap,
2510
	.llseek		= noop_llseek,
2511
	KVM_COMPAT(kvm_vcpu_compat_ioctl),
A
Avi Kivity 已提交
2512 2513 2514 2515 2516 2517 2518
};

/*
 * Allocates an inode for the vcpu.
 */
static int create_vcpu_fd(struct kvm_vcpu *vcpu)
{
2519 2520 2521 2522
	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
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 2549 2550
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;
}

2551 2552 2553
/*
 * Creates some virtual cpus.  Good luck creating more than one.
 */
2554
static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
2555 2556
{
	int r;
2557
	struct kvm_vcpu *vcpu;
2558

G
Greg Kurz 已提交
2559
	if (id >= KVM_MAX_VCPU_ID)
2560 2561
		return -EINVAL;

2562 2563 2564 2565 2566 2567 2568 2569 2570
	mutex_lock(&kvm->lock);
	if (kvm->created_vcpus == KVM_MAX_VCPUS) {
		mutex_unlock(&kvm->lock);
		return -EINVAL;
	}

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

2571
	vcpu = kvm_arch_vcpu_create(kvm, id);
2572 2573 2574 2575
	if (IS_ERR(vcpu)) {
		r = PTR_ERR(vcpu);
		goto vcpu_decrement;
	}
2576

2577 2578
	preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);

2579 2580
	r = kvm_arch_vcpu_setup(vcpu);
	if (r)
2581
		goto vcpu_destroy;
2582

2583 2584 2585 2586
	r = kvm_create_vcpu_debugfs(vcpu);
	if (r)
		goto vcpu_destroy;

S
Shaohua Li 已提交
2587
	mutex_lock(&kvm->lock);
2588 2589 2590 2591
	if (kvm_get_vcpu_by_id(kvm, id)) {
		r = -EEXIST;
		goto unlock_vcpu_destroy;
	}
2592 2593

	BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
2594

R
Rusty Russell 已提交
2595
	/* Now it's all set up, let userspace reach it */
A
Al Viro 已提交
2596
	kvm_get_kvm(kvm);
A
Avi Kivity 已提交
2597
	r = create_vcpu_fd(vcpu);
2598 2599
	if (r < 0) {
		kvm_put_kvm(kvm);
2600
		goto unlock_vcpu_destroy;
2601 2602 2603
	}

	kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu;
2604 2605 2606 2607 2608

	/*
	 * Pairs with smp_rmb() in kvm_get_vcpu.  Write kvm->vcpus
	 * before kvm->online_vcpu's incremented value.
	 */
2609 2610 2611 2612
	smp_wmb();
	atomic_inc(&kvm->online_vcpus);

	mutex_unlock(&kvm->lock);
2613
	kvm_arch_vcpu_postcreate(vcpu);
R
Rusty Russell 已提交
2614
	return r;
2615

2616
unlock_vcpu_destroy:
2617
	mutex_unlock(&kvm->lock);
2618
	debugfs_remove_recursive(vcpu->debugfs_dentry);
2619
vcpu_destroy:
2620
	kvm_arch_vcpu_destroy(vcpu);
2621 2622 2623 2624
vcpu_decrement:
	mutex_lock(&kvm->lock);
	kvm->created_vcpus--;
	mutex_unlock(&kvm->lock);
2625 2626 2627
	return r;
}

A
Avi Kivity 已提交
2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638
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
2639 2640
static long kvm_vcpu_ioctl(struct file *filp,
			   unsigned int ioctl, unsigned long arg)
A
Avi Kivity 已提交
2641
{
A
Avi Kivity 已提交
2642
	struct kvm_vcpu *vcpu = filp->private_data;
A
Al Viro 已提交
2643
	void __user *argp = (void __user *)arg;
2644
	int r;
2645 2646
	struct kvm_fpu *fpu = NULL;
	struct kvm_sregs *kvm_sregs = NULL;
A
Avi Kivity 已提交
2647

2648 2649
	if (vcpu->kvm->mm != current->mm)
		return -EIO;
2650

2651 2652 2653
	if (unlikely(_IOC_TYPE(ioctl) != KVMIO))
		return -EINVAL;

2654
	/*
2655 2656
	 * Some architectures have vcpu ioctls that are asynchronous to vcpu
	 * execution; mutex_lock() would break them.
2657
	 */
2658 2659
	r = kvm_arch_vcpu_async_ioctl(filp, ioctl, arg);
	if (r != -ENOIOCTLCMD)
2660
		return r;
2661

2662 2663
	if (mutex_lock_killable(&vcpu->mutex))
		return -EINTR;
A
Avi Kivity 已提交
2664
	switch (ioctl) {
2665 2666
	case KVM_RUN: {
		struct pid *oldpid;
2667 2668 2669
		r = -EINVAL;
		if (arg)
			goto out;
2670
		oldpid = rcu_access_pointer(vcpu->pid);
2671
		if (unlikely(oldpid != task_pid(current))) {
2672
			/* The thread running this VCPU changed. */
2673
			struct pid *newpid;
2674

2675 2676 2677 2678 2679
			r = kvm_arch_vcpu_run_pid_change(vcpu);
			if (r)
				break;

			newpid = get_task_pid(current, PIDTYPE_PID);
2680 2681 2682 2683 2684
			rcu_assign_pointer(vcpu->pid, newpid);
			if (oldpid)
				synchronize_rcu();
			put_pid(oldpid);
		}
2685
		r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
2686
		trace_kvm_userspace_exit(vcpu->run->exit_reason, r);
A
Avi Kivity 已提交
2687
		break;
2688
	}
A
Avi Kivity 已提交
2689
	case KVM_GET_REGS: {
2690
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
2691

2692
		r = -ENOMEM;
2693
		kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL_ACCOUNT);
2694
		if (!kvm_regs)
A
Avi Kivity 已提交
2695
			goto out;
2696 2697 2698
		r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
		if (r)
			goto out_free1;
A
Avi Kivity 已提交
2699
		r = -EFAULT;
2700 2701
		if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
			goto out_free1;
A
Avi Kivity 已提交
2702
		r = 0;
2703 2704
out_free1:
		kfree(kvm_regs);
A
Avi Kivity 已提交
2705 2706 2707
		break;
	}
	case KVM_SET_REGS: {
2708
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
2709

2710
		r = -ENOMEM;
2711 2712 2713
		kvm_regs = memdup_user(argp, sizeof(*kvm_regs));
		if (IS_ERR(kvm_regs)) {
			r = PTR_ERR(kvm_regs);
A
Avi Kivity 已提交
2714
			goto out;
2715
		}
2716 2717
		r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
		kfree(kvm_regs);
A
Avi Kivity 已提交
2718 2719 2720
		break;
	}
	case KVM_GET_SREGS: {
2721 2722
		kvm_sregs = kzalloc(sizeof(struct kvm_sregs),
				    GFP_KERNEL_ACCOUNT);
2723 2724 2725 2726
		r = -ENOMEM;
		if (!kvm_sregs)
			goto out;
		r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
2727 2728 2729
		if (r)
			goto out;
		r = -EFAULT;
2730
		if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
A
Avi Kivity 已提交
2731 2732 2733 2734 2735
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_SREGS: {
2736 2737 2738
		kvm_sregs = memdup_user(argp, sizeof(*kvm_sregs));
		if (IS_ERR(kvm_sregs)) {
			r = PTR_ERR(kvm_sregs);
G
Guo Chao 已提交
2739
			kvm_sregs = NULL;
A
Avi Kivity 已提交
2740
			goto out;
2741
		}
2742
		r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
2743 2744
		break;
	}
2745 2746 2747 2748 2749 2750 2751
	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;
2752
		if (copy_to_user(argp, &mp_state, sizeof(mp_state)))
2753 2754 2755 2756 2757 2758 2759 2760
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_MP_STATE: {
		struct kvm_mp_state mp_state;

		r = -EFAULT;
2761
		if (copy_from_user(&mp_state, argp, sizeof(mp_state)))
2762 2763 2764 2765
			goto out;
		r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
		break;
	}
A
Avi Kivity 已提交
2766 2767 2768 2769
	case KVM_TRANSLATE: {
		struct kvm_translation tr;

		r = -EFAULT;
2770
		if (copy_from_user(&tr, argp, sizeof(tr)))
A
Avi Kivity 已提交
2771
			goto out;
2772
		r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
A
Avi Kivity 已提交
2773 2774 2775
		if (r)
			goto out;
		r = -EFAULT;
2776
		if (copy_to_user(argp, &tr, sizeof(tr)))
A
Avi Kivity 已提交
2777 2778 2779 2780
			goto out;
		r = 0;
		break;
	}
J
Jan Kiszka 已提交
2781 2782
	case KVM_SET_GUEST_DEBUG: {
		struct kvm_guest_debug dbg;
A
Avi Kivity 已提交
2783 2784

		r = -EFAULT;
2785
		if (copy_from_user(&dbg, argp, sizeof(dbg)))
A
Avi Kivity 已提交
2786
			goto out;
J
Jan Kiszka 已提交
2787
		r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
A
Avi Kivity 已提交
2788 2789
		break;
	}
A
Avi Kivity 已提交
2790 2791 2792 2793 2794 2795 2796 2797 2798
	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,
2799
					   sizeof(kvm_sigmask)))
A
Avi Kivity 已提交
2800 2801
				goto out;
			r = -EINVAL;
2802
			if (kvm_sigmask.len != sizeof(sigset))
A
Avi Kivity 已提交
2803 2804 2805
				goto out;
			r = -EFAULT;
			if (copy_from_user(&sigset, sigmask_arg->sigset,
2806
					   sizeof(sigset)))
A
Avi Kivity 已提交
2807 2808 2809
				goto out;
			p = &sigset;
		}
2810
		r = kvm_vcpu_ioctl_set_sigmask(vcpu, p);
A
Avi Kivity 已提交
2811 2812
		break;
	}
A
Avi Kivity 已提交
2813
	case KVM_GET_FPU: {
2814
		fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL_ACCOUNT);
2815 2816 2817 2818
		r = -ENOMEM;
		if (!fpu)
			goto out;
		r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu);
A
Avi Kivity 已提交
2819 2820 2821
		if (r)
			goto out;
		r = -EFAULT;
2822
		if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
A
Avi Kivity 已提交
2823 2824 2825 2826 2827
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_FPU: {
2828 2829 2830
		fpu = memdup_user(argp, sizeof(*fpu));
		if (IS_ERR(fpu)) {
			r = PTR_ERR(fpu);
G
Guo Chao 已提交
2831
			fpu = NULL;
A
Avi Kivity 已提交
2832
			goto out;
2833
		}
2834
		r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
A
Avi Kivity 已提交
2835 2836
		break;
	}
A
Avi Kivity 已提交
2837
	default:
2838
		r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
2839 2840
	}
out:
2841
	mutex_unlock(&vcpu->mutex);
2842 2843
	kfree(fpu);
	kfree(kvm_sregs);
A
Avi Kivity 已提交
2844 2845 2846
	return r;
}

2847
#ifdef CONFIG_KVM_COMPAT
2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866
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,
2867
					   sizeof(kvm_sigmask)))
2868 2869
				goto out;
			r = -EINVAL;
A
Al Viro 已提交
2870
			if (kvm_sigmask.len != sizeof(compat_sigset_t))
2871 2872
				goto out;
			r = -EFAULT;
A
Al Viro 已提交
2873
			if (get_compat_sigset(&sigset, (void *)sigmask_arg->sigset))
2874
				goto out;
2875 2876 2877
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
		} else
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, NULL);
2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888
		break;
	}
	default:
		r = kvm_vcpu_ioctl(filp, ioctl, arg);
	}

out:
	return r;
}
#endif

S
Scott Wood 已提交
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 2935 2936
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,
2937
	KVM_COMPAT(kvm_device_ioctl),
S
Scott Wood 已提交
2938 2939 2940 2941 2942 2943 2944 2945 2946 2947
};

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

	return filp->private_data;
}

2948
static struct kvm_device_ops *kvm_device_ops_table[KVM_DEV_TYPE_MAX] = {
2949
#ifdef CONFIG_KVM_MPIC
2950 2951
	[KVM_DEV_TYPE_FSL_MPIC_20]	= &kvm_mpic_ops,
	[KVM_DEV_TYPE_FSL_MPIC_42]	= &kvm_mpic_ops,
2952
#endif
2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966
};

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;
}

2967 2968 2969 2970 2971 2972
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
2973 2974 2975 2976 2977 2978 2979 2980
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;

2981 2982 2983 2984 2985
	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
2986 2987 2988 2989 2990
		return -ENODEV;

	if (test)
		return 0;

2991
	dev = kzalloc(sizeof(*dev), GFP_KERNEL_ACCOUNT);
S
Scott Wood 已提交
2992 2993 2994 2995 2996 2997
	if (!dev)
		return -ENOMEM;

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

2998
	mutex_lock(&kvm->lock);
S
Scott Wood 已提交
2999 3000
	ret = ops->create(dev, cd->type);
	if (ret < 0) {
3001
		mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
3002 3003 3004
		kfree(dev);
		return ret;
	}
3005 3006
	list_add(&dev->vm_node, &kvm->devices);
	mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
3007

3008 3009 3010
	if (ops->init)
		ops->init(dev);

3011
	kvm_get_kvm(kvm);
3012
	ret = anon_inode_getfd(ops->name, &kvm_device_fops, dev, O_RDWR | O_CLOEXEC);
S
Scott Wood 已提交
3013
	if (ret < 0) {
3014
		kvm_put_kvm(kvm);
3015 3016 3017
		mutex_lock(&kvm->lock);
		list_del(&dev->vm_node);
		mutex_unlock(&kvm->lock);
3018
		ops->destroy(dev);
S
Scott Wood 已提交
3019 3020 3021 3022 3023 3024 3025
		return ret;
	}

	cd->fd = ret;
	return 0;
}

3026 3027 3028 3029 3030 3031 3032 3033 3034 3035
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
3036
#ifdef CONFIG_HAVE_KVM_IRQFD
3037
	case KVM_CAP_IRQFD:
3038 3039
	case KVM_CAP_IRQFD_RESAMPLE:
#endif
3040
	case KVM_CAP_IOEVENTFD_ANY_LENGTH:
3041
	case KVM_CAP_CHECK_EXTENSION_VM:
3042
	case KVM_CAP_ENABLE_CAP_VM:
3043 3044 3045
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
	case KVM_CAP_MANUAL_DIRTY_LOG_PROTECT:
#endif
3046
		return 1;
3047
#ifdef CONFIG_KVM_MMIO
3048 3049
	case KVM_CAP_COALESCED_MMIO:
		return KVM_COALESCED_MMIO_PAGE_OFFSET;
P
Peng Hao 已提交
3050 3051
	case KVM_CAP_COALESCED_PIO:
		return 1;
3052
#endif
3053 3054 3055
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_CAP_IRQ_ROUTING:
		return KVM_MAX_IRQ_ROUTES;
3056 3057 3058 3059
#endif
#if KVM_ADDRESS_SPACE_NUM > 1
	case KVM_CAP_MULTI_ADDRESS_SPACE:
		return KVM_ADDRESS_SPACE_NUM;
3060
#endif
G
Greg Kurz 已提交
3061 3062
	case KVM_CAP_MAX_VCPU_ID:
		return KVM_MAX_VCPU_ID;
3063 3064 3065 3066 3067 3068
	default:
		break;
	}
	return kvm_vm_ioctl_check_extension(kvm, arg);
}

3069 3070 3071 3072 3073 3074 3075 3076 3077 3078
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) {
3079 3080 3081 3082 3083 3084 3085
#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
3086 3087 3088 3089 3090
	default:
		return kvm_vm_ioctl_enable_cap(kvm, cap);
	}
}

A
Avi Kivity 已提交
3091 3092 3093 3094 3095
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;
3096
	int r;
A
Avi Kivity 已提交
3097

3098 3099
	if (kvm->mm != current->mm)
		return -EIO;
A
Avi Kivity 已提交
3100 3101 3102 3103
	switch (ioctl) {
	case KVM_CREATE_VCPU:
		r = kvm_vm_ioctl_create_vcpu(kvm, arg);
		break;
3104 3105 3106 3107 3108 3109 3110 3111 3112
	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;
	}
3113 3114 3115 3116 3117
	case KVM_SET_USER_MEMORY_REGION: {
		struct kvm_userspace_memory_region kvm_userspace_mem;

		r = -EFAULT;
		if (copy_from_user(&kvm_userspace_mem, argp,
3118
						sizeof(kvm_userspace_mem)))
3119 3120
			goto out;

3121
		r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem);
A
Avi Kivity 已提交
3122 3123 3124 3125 3126 3127
		break;
	}
	case KVM_GET_DIRTY_LOG: {
		struct kvm_dirty_log log;

		r = -EFAULT;
3128
		if (copy_from_user(&log, argp, sizeof(log)))
A
Avi Kivity 已提交
3129
			goto out;
3130
		r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
A
Avi Kivity 已提交
3131 3132
		break;
	}
3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143
#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
3144
#ifdef CONFIG_KVM_MMIO
3145 3146
	case KVM_REGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3147

3148
		r = -EFAULT;
3149
		if (copy_from_user(&zone, argp, sizeof(zone)))
3150 3151 3152 3153 3154 3155
			goto out;
		r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
		break;
	}
	case KVM_UNREGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3156

3157
		r = -EFAULT;
3158
		if (copy_from_user(&zone, argp, sizeof(zone)))
3159 3160 3161 3162 3163
			goto out;
		r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
		break;
	}
#endif
G
Gregory Haskins 已提交
3164 3165 3166 3167
	case KVM_IRQFD: {
		struct kvm_irqfd data;

		r = -EFAULT;
3168
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3169
			goto out;
3170
		r = kvm_irqfd(kvm, &data);
G
Gregory Haskins 已提交
3171 3172
		break;
	}
G
Gregory Haskins 已提交
3173 3174 3175 3176
	case KVM_IOEVENTFD: {
		struct kvm_ioeventfd data;

		r = -EFAULT;
3177
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3178 3179 3180 3181
			goto out;
		r = kvm_ioeventfd(kvm, &data);
		break;
	}
3182 3183 3184 3185 3186
#ifdef CONFIG_HAVE_KVM_MSI
	case KVM_SIGNAL_MSI: {
		struct kvm_msi msi;

		r = -EFAULT;
3187
		if (copy_from_user(&msi, argp, sizeof(msi)))
3188 3189 3190 3191
			goto out;
		r = kvm_send_userspace_msi(kvm, &msi);
		break;
	}
3192 3193 3194 3195 3196 3197 3198
#endif
#ifdef __KVM_HAVE_IRQ_LINE
	case KVM_IRQ_LINE_STATUS:
	case KVM_IRQ_LINE: {
		struct kvm_irq_level irq_event;

		r = -EFAULT;
3199
		if (copy_from_user(&irq_event, argp, sizeof(irq_event)))
3200 3201
			goto out;

3202 3203
		r = kvm_vm_ioctl_irq_line(kvm, &irq_event,
					ioctl == KVM_IRQ_LINE_STATUS);
3204 3205 3206 3207 3208
		if (r)
			goto out;

		r = -EFAULT;
		if (ioctl == KVM_IRQ_LINE_STATUS) {
3209
			if (copy_to_user(argp, &irq_event, sizeof(irq_event)))
3210 3211 3212 3213 3214 3215
				goto out;
		}

		r = 0;
		break;
	}
3216
#endif
3217 3218 3219 3220
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_SET_GSI_ROUTING: {
		struct kvm_irq_routing routing;
		struct kvm_irq_routing __user *urouting;
3221
		struct kvm_irq_routing_entry *entries = NULL;
3222 3223 3224 3225 3226

		r = -EFAULT;
		if (copy_from_user(&routing, argp, sizeof(routing)))
			goto out;
		r = -EINVAL;
3227 3228
		if (!kvm_arch_can_set_irq_routing(kvm))
			goto out;
3229
		if (routing.nr > KVM_MAX_IRQ_ROUTES)
3230 3231 3232
			goto out;
		if (routing.flags)
			goto out;
3233 3234
		if (routing.nr) {
			r = -ENOMEM;
3235 3236
			entries = vmalloc(array_size(sizeof(*entries),
						     routing.nr));
3237 3238 3239 3240 3241 3242 3243 3244
			if (!entries)
				goto out;
			r = -EFAULT;
			urouting = argp;
			if (copy_from_user(entries, urouting->entries,
					   routing.nr * sizeof(*entries)))
				goto out_free_irq_routing;
		}
3245 3246
		r = kvm_set_irq_routing(kvm, entries, routing.nr,
					routing.flags);
3247
out_free_irq_routing:
3248 3249 3250 3251
		vfree(entries);
		break;
	}
#endif /* CONFIG_HAVE_KVM_IRQ_ROUTING */
S
Scott Wood 已提交
3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269
	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;
	}
3270 3271 3272
	case KVM_CHECK_EXTENSION:
		r = kvm_vm_ioctl_check_extension_generic(kvm, arg);
		break;
3273
	default:
3274
		r = kvm_arch_vm_ioctl(filp, ioctl, arg);
3275 3276 3277 3278 3279
	}
out:
	return r;
}

3280
#ifdef CONFIG_KVM_COMPAT
3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304
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)))
3305
			return -EFAULT;
3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320
		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

3321
static struct file_operations kvm_vm_fops = {
3322 3323
	.release        = kvm_vm_release,
	.unlocked_ioctl = kvm_vm_ioctl,
3324
	.llseek		= noop_llseek,
3325
	KVM_COMPAT(kvm_vm_compat_ioctl),
3326 3327
};

3328
static int kvm_dev_ioctl_create_vm(unsigned long type)
3329
{
3330
	int r;
3331
	struct kvm *kvm;
3332
	struct file *file;
3333

3334
	kvm = kvm_create_vm(type);
3335 3336
	if (IS_ERR(kvm))
		return PTR_ERR(kvm);
3337
#ifdef CONFIG_KVM_MMIO
3338
	r = kvm_coalesced_mmio_init(kvm);
3339 3340
	if (r < 0)
		goto put_kvm;
3341
#endif
3342
	r = get_unused_fd_flags(O_CLOEXEC);
3343 3344 3345
	if (r < 0)
		goto put_kvm;

3346 3347 3348
	file = anon_inode_getfile("kvm-vm", &kvm_vm_fops, kvm, O_RDWR);
	if (IS_ERR(file)) {
		put_unused_fd(r);
3349 3350
		r = PTR_ERR(file);
		goto put_kvm;
3351
	}
3352

3353 3354 3355 3356 3357 3358
	/*
	 * 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).
	 */
3359
	if (kvm_create_vm_debugfs(kvm, r) < 0) {
3360 3361
		put_unused_fd(r);
		fput(file);
3362 3363
		return -ENOMEM;
	}
3364
	kvm_uevent_notify_change(KVM_EVENT_CREATE_VM, kvm);
3365

3366
	fd_install(r, file);
3367
	return r;
3368 3369 3370 3371

put_kvm:
	kvm_put_kvm(kvm);
	return r;
3372 3373 3374 3375 3376
}

static long kvm_dev_ioctl(struct file *filp,
			  unsigned int ioctl, unsigned long arg)
{
3377
	long r = -EINVAL;
3378 3379 3380

	switch (ioctl) {
	case KVM_GET_API_VERSION:
3381 3382
		if (arg)
			goto out;
3383 3384 3385
		r = KVM_API_VERSION;
		break;
	case KVM_CREATE_VM:
3386
		r = kvm_dev_ioctl_create_vm(arg);
3387
		break;
3388
	case KVM_CHECK_EXTENSION:
3389
		r = kvm_vm_ioctl_check_extension_generic(NULL, arg);
3390
		break;
3391 3392 3393
	case KVM_GET_VCPU_MMAP_SIZE:
		if (arg)
			goto out;
3394 3395 3396
		r = PAGE_SIZE;     /* struct kvm_run */
#ifdef CONFIG_X86
		r += PAGE_SIZE;    /* pio data page */
3397
#endif
3398
#ifdef CONFIG_KVM_MMIO
3399
		r += PAGE_SIZE;    /* coalesced mmio ring page */
3400
#endif
3401
		break;
3402 3403 3404
	case KVM_TRACE_ENABLE:
	case KVM_TRACE_PAUSE:
	case KVM_TRACE_DISABLE:
3405
		r = -EOPNOTSUPP;
3406
		break;
A
Avi Kivity 已提交
3407
	default:
3408
		return kvm_arch_dev_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
3409 3410 3411 3412 3413 3414 3415
	}
out:
	return r;
}

static struct file_operations kvm_chardev_ops = {
	.unlocked_ioctl = kvm_dev_ioctl,
3416
	.llseek		= noop_llseek,
3417
	KVM_COMPAT(kvm_dev_ioctl),
A
Avi Kivity 已提交
3418 3419 3420
};

static struct miscdevice kvm_dev = {
A
Avi Kivity 已提交
3421
	KVM_MINOR,
A
Avi Kivity 已提交
3422 3423 3424 3425
	"kvm",
	&kvm_chardev_ops,
};

3426
static void hardware_enable_nolock(void *junk)
3427 3428
{
	int cpu = raw_smp_processor_id();
3429
	int r;
3430

3431
	if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
3432
		return;
3433

3434
	cpumask_set_cpu(cpu, cpus_hardware_enabled);
3435

3436
	r = kvm_arch_hardware_enable();
3437 3438 3439 3440

	if (r) {
		cpumask_clear_cpu(cpu, cpus_hardware_enabled);
		atomic_inc(&hardware_enable_failed);
X
Xiubo Li 已提交
3441
		pr_info("kvm: enabling virtualization on CPU%d failed\n", cpu);
3442
	}
3443 3444
}

3445
static int kvm_starting_cpu(unsigned int cpu)
3446
{
3447
	raw_spin_lock(&kvm_count_lock);
3448 3449
	if (kvm_usage_count)
		hardware_enable_nolock(NULL);
3450
	raw_spin_unlock(&kvm_count_lock);
3451
	return 0;
3452 3453 3454
}

static void hardware_disable_nolock(void *junk)
3455 3456 3457
{
	int cpu = raw_smp_processor_id();

3458
	if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
3459
		return;
3460
	cpumask_clear_cpu(cpu, cpus_hardware_enabled);
3461
	kvm_arch_hardware_disable();
3462 3463
}

3464
static int kvm_dying_cpu(unsigned int cpu)
3465
{
3466
	raw_spin_lock(&kvm_count_lock);
3467 3468
	if (kvm_usage_count)
		hardware_disable_nolock(NULL);
3469
	raw_spin_unlock(&kvm_count_lock);
3470
	return 0;
3471 3472
}

3473 3474 3475 3476 3477 3478
static void hardware_disable_all_nolock(void)
{
	BUG_ON(!kvm_usage_count);

	kvm_usage_count--;
	if (!kvm_usage_count)
3479
		on_each_cpu(hardware_disable_nolock, NULL, 1);
3480 3481 3482 3483
}

static void hardware_disable_all(void)
{
3484
	raw_spin_lock(&kvm_count_lock);
3485
	hardware_disable_all_nolock();
3486
	raw_spin_unlock(&kvm_count_lock);
3487 3488 3489 3490 3491 3492
}

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

3493
	raw_spin_lock(&kvm_count_lock);
3494 3495 3496 3497

	kvm_usage_count++;
	if (kvm_usage_count == 1) {
		atomic_set(&hardware_enable_failed, 0);
3498
		on_each_cpu(hardware_enable_nolock, NULL, 1);
3499 3500 3501 3502 3503 3504 3505

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

3506
	raw_spin_unlock(&kvm_count_lock);
3507 3508 3509 3510

	return r;
}

3511
static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
M
Mike Day 已提交
3512
		      void *v)
3513
{
3514 3515 3516 3517 3518 3519
	/*
	 * 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
3520
	pr_info("kvm: exiting hardware virtualization\n");
3521
	kvm_rebooting = true;
3522
	on_each_cpu(hardware_disable_nolock, NULL, 1);
3523 3524 3525 3526 3527 3528 3529 3530
	return NOTIFY_OK;
}

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

M
Marcelo Tosatti 已提交
3531
static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
3532 3533 3534 3535
{
	int i;

	for (i = 0; i < bus->dev_count; i++) {
3536
		struct kvm_io_device *pos = bus->range[i].dev;
3537 3538 3539

		kvm_iodevice_destructor(pos);
	}
M
Marcelo Tosatti 已提交
3540
	kfree(bus);
3541 3542
}

3543
static inline int kvm_io_bus_cmp(const struct kvm_io_range *r1,
X
Xiubo Li 已提交
3544
				 const struct kvm_io_range *r2)
3545
{
J
Jason Wang 已提交
3546 3547 3548 3549
	gpa_t addr1 = r1->addr;
	gpa_t addr2 = r2->addr;

	if (addr1 < addr2)
3550
		return -1;
J
Jason Wang 已提交
3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562

	/* 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)
3563
		return 1;
J
Jason Wang 已提交
3564

3565 3566 3567
	return 0;
}

3568 3569
static int kvm_io_bus_sort_cmp(const void *p1, const void *p2)
{
3570
	return kvm_io_bus_cmp(p1, p2);
3571 3572
}

G
Geoff Levand 已提交
3573
static int kvm_io_bus_get_first_dev(struct kvm_io_bus *bus,
3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590
			     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;

3591
	while (off > 0 && kvm_io_bus_cmp(&key, &bus->range[off-1]) == 0)
3592 3593 3594 3595 3596
		off--;

	return off;
}

3597
static int __kvm_io_bus_write(struct kvm_vcpu *vcpu, struct kvm_io_bus *bus,
C
Cornelia Huck 已提交
3598 3599 3600 3601 3602 3603 3604 3605 3606
			      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 &&
3607
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
3608
		if (!kvm_iodevice_write(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
3609 3610 3611 3612 3613 3614 3615 3616
					range->len, val))
			return idx;
		idx++;
	}

	return -EOPNOTSUPP;
}

3617
/* kvm_io_bus_write - called under kvm->slots_lock */
3618
int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
3619
		     int len, const void *val)
3620
{
3621
	struct kvm_io_bus *bus;
3622
	struct kvm_io_range range;
C
Cornelia Huck 已提交
3623
	int r;
3624 3625 3626 3627 3628

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
3629

3630
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3631 3632
	if (!bus)
		return -ENOMEM;
3633
	r = __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3634 3635 3636 3637
	return r < 0 ? r : 0;
}

/* kvm_io_bus_write_cookie - called under kvm->slots_lock */
3638 3639
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
3640 3641 3642 3643 3644 3645 3646 3647 3648
{
	struct kvm_io_bus *bus;
	struct kvm_io_range range;

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

3649
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3650 3651
	if (!bus)
		return -ENOMEM;
C
Cornelia Huck 已提交
3652 3653 3654

	/* First try the device referenced by cookie. */
	if ((cookie >= 0) && (cookie < bus->dev_count) &&
3655
	    (kvm_io_bus_cmp(&range, &bus->range[cookie]) == 0))
3656
		if (!kvm_iodevice_write(vcpu, bus->range[cookie].dev, addr, len,
C
Cornelia Huck 已提交
3657 3658 3659 3660 3661 3662 3663
					val))
			return cookie;

	/*
	 * cookie contained garbage; fall back to search and return the
	 * correct cookie value.
	 */
3664
	return __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3665 3666
}

3667 3668
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
3669 3670 3671 3672
{
	int idx;

	idx = kvm_io_bus_get_first_dev(bus, range->addr, range->len);
3673 3674 3675 3676
	if (idx < 0)
		return -EOPNOTSUPP;

	while (idx < bus->dev_count &&
3677
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
3678
		if (!kvm_iodevice_read(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
3679 3680
				       range->len, val))
			return idx;
3681 3682 3683
		idx++;
	}

3684 3685
	return -EOPNOTSUPP;
}
3686
EXPORT_SYMBOL_GPL(kvm_io_bus_write);
3687

3688
/* kvm_io_bus_read - called under kvm->slots_lock */
3689
int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
M
Marcelo Tosatti 已提交
3690
		    int len, void *val)
3691
{
3692
	struct kvm_io_bus *bus;
3693
	struct kvm_io_range range;
C
Cornelia Huck 已提交
3694
	int r;
3695 3696 3697 3698 3699

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
M
Marcelo Tosatti 已提交
3700

3701
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3702 3703
	if (!bus)
		return -ENOMEM;
3704
	r = __kvm_io_bus_read(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3705 3706
	return r < 0 ? r : 0;
}
3707

3708

3709
/* Caller must hold slots_lock. */
3710 3711
int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
			    int len, struct kvm_io_device *dev)
3712
{
3713
	int i;
M
Marcelo Tosatti 已提交
3714
	struct kvm_io_bus *new_bus, *bus;
3715
	struct kvm_io_range range;
3716

3717
	bus = kvm_get_bus(kvm, bus_idx);
3718 3719 3720
	if (!bus)
		return -ENOMEM;

3721 3722
	/* exclude ioeventfd which is limited by maximum fd */
	if (bus->dev_count - bus->ioeventfd_count > NR_IOBUS_DEVS - 1)
3723
		return -ENOSPC;
3724

3725
	new_bus = kmalloc(struct_size(bus, range, bus->dev_count + 1),
3726
			  GFP_KERNEL_ACCOUNT);
M
Marcelo Tosatti 已提交
3727 3728
	if (!new_bus)
		return -ENOMEM;
3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744

	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
3745 3746 3747
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
3748 3749 3750 3751

	return 0;
}

3752
/* Caller must hold slots_lock. */
3753 3754
void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
			       struct kvm_io_device *dev)
3755
{
3756
	int i;
M
Marcelo Tosatti 已提交
3757
	struct kvm_io_bus *new_bus, *bus;
3758

3759
	bus = kvm_get_bus(kvm, bus_idx);
3760
	if (!bus)
3761
		return;
3762

3763 3764
	for (i = 0; i < bus->dev_count; i++)
		if (bus->range[i].dev == dev) {
3765 3766
			break;
		}
M
Marcelo Tosatti 已提交
3767

3768 3769
	if (i == bus->dev_count)
		return;
3770

3771
	new_bus = kmalloc(struct_size(bus, range, bus->dev_count - 1),
3772
			  GFP_KERNEL_ACCOUNT);
3773 3774 3775 3776
	if (!new_bus)  {
		pr_err("kvm: failed to shrink bus, removing it completely\n");
		goto broken;
	}
3777 3778 3779 3780 3781

	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
3782

3783
broken:
M
Marcelo Tosatti 已提交
3784 3785 3786
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
3787
	return;
3788 3789
}

3790 3791 3792 3793 3794 3795 3796 3797 3798 3799
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);
3800 3801
	if (!bus)
		goto out_unlock;
3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815

	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);

3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827
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.
	 */
3828
	if (!refcount_inc_not_zero(&stat_data->kvm->users_count))
3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853
		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;

3854
	*val = *(ulong *)((void *)stat_data->kvm + stat_data->offset);
3855 3856 3857 3858

	return 0;
}

3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870
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;
}

3871 3872 3873 3874
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,
3875
				vm_stat_clear_per_vm, "%llu\n");
3876 3877 3878 3879 3880 3881 3882 3883
}

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,
3884
	.llseek  = no_llseek,
3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895
};

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)
3896
		*val += *(u64 *)((void *)vcpu + stat_data->offset);
3897 3898 3899 3900

	return 0;
}

3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915
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;
}

3916 3917 3918 3919
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,
3920
				 vcpu_stat_clear_per_vm, "%llu\n");
3921 3922 3923 3924 3925 3926 3927 3928
}

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,
3929
	.llseek  = no_llseek,
3930 3931 3932 3933 3934 3935 3936
};

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,
};

3937
static int vm_stat_get(void *_offset, u64 *val)
3938 3939 3940
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
3941 3942
	struct kvm_stat_data stat_tmp = {.offset = offset};
	u64 tmp_val;
3943

3944
	*val = 0;
3945
	spin_lock(&kvm_lock);
3946 3947 3948 3949 3950
	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;
	}
3951
	spin_unlock(&kvm_lock);
3952
	return 0;
3953 3954
}

3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974
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");
3975

3976
static int vcpu_stat_get(void *_offset, u64 *val)
A
Avi Kivity 已提交
3977 3978 3979
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
3980 3981
	struct kvm_stat_data stat_tmp = {.offset = offset};
	u64 tmp_val;
A
Avi Kivity 已提交
3982

3983
	*val = 0;
3984
	spin_lock(&kvm_lock);
3985 3986 3987 3988 3989
	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;
	}
3990
	spin_unlock(&kvm_lock);
3991
	return 0;
A
Avi Kivity 已提交
3992 3993
}

3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014
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");
4015

4016
static const struct file_operations *stat_fops[] = {
4017 4018 4019
	[KVM_STAT_VCPU] = &vcpu_stat_fops,
	[KVM_STAT_VM]   = &vm_stat_fops,
};
A
Avi Kivity 已提交
4020

4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039
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);

4040
	env = kzalloc(sizeof(*env), GFP_KERNEL_ACCOUNT);
4041 4042 4043 4044 4045 4046
	if (!env)
		return;

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

4047
	if (type == KVM_EVENT_CREATE_VM) {
4048
		add_uevent_var(env, "EVENT=create");
4049 4050
		kvm->userspace_pid = task_pid_nr(current);
	} else if (type == KVM_EVENT_DESTROY_VM) {
4051
		add_uevent_var(env, "EVENT=destroy");
4052 4053
	}
	add_uevent_var(env, "PID=%d", kvm->userspace_pid);
4054 4055

	if (kvm->debugfs_dentry) {
4056
		char *tmp, *p = kmalloc(PATH_MAX, GFP_KERNEL_ACCOUNT);
4057 4058 4059 4060 4061 4062

		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);
4063 4064 4065 4066 4067 4068 4069 4070
		}
	}
	/* 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);
}

4071
static void kvm_init_debug(void)
A
Avi Kivity 已提交
4072 4073 4074
{
	struct kvm_stats_debugfs_item *p;

4075
	kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
4076

4077 4078
	kvm_debugfs_num_entries = 0;
	for (p = debugfs_entries; p->name; ++p, kvm_debugfs_num_entries++) {
4079 4080 4081
		debugfs_create_file(p->name, 0644, kvm_debugfs_dir,
				    (void *)(long)p->offset,
				    stat_fops[p->kind]);
4082
	}
A
Avi Kivity 已提交
4083 4084
}

4085
static int kvm_suspend(void)
4086
{
4087
	if (kvm_usage_count)
4088
		hardware_disable_nolock(NULL);
4089 4090 4091
	return 0;
}

4092
static void kvm_resume(void)
4093
{
4094
	if (kvm_usage_count) {
4095
		WARN_ON(raw_spin_is_locked(&kvm_count_lock));
4096
		hardware_enable_nolock(NULL);
4097
	}
4098 4099
}

4100
static struct syscore_ops kvm_syscore_ops = {
4101 4102 4103 4104
	.suspend = kvm_suspend,
	.resume = kvm_resume,
};

4105 4106 4107 4108 4109 4110 4111 4112 4113
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);
4114

4115 4116
	if (vcpu->preempted)
		vcpu->preempted = false;
4117

R
Radim Krčmář 已提交
4118 4119
	kvm_arch_sched_in(vcpu, cpu);

4120
	kvm_arch_vcpu_load(vcpu, cpu);
4121 4122 4123 4124 4125 4126 4127
}

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

4128 4129
	if (current->state == TASK_RUNNING)
		vcpu->preempted = true;
4130
	kvm_arch_vcpu_put(vcpu);
4131 4132
}

4133
int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
4134
		  struct module *module)
A
Avi Kivity 已提交
4135 4136
{
	int r;
Y
Yang, Sheng 已提交
4137
	int cpu;
A
Avi Kivity 已提交
4138

4139 4140
	r = kvm_arch_init(opaque);
	if (r)
4141
		goto out_fail;
4142

4143 4144 4145 4146
	/*
	 * 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
4147 4148
	 * kvm_arch_init must be called before kvm_irqfd_init to avoid creating
	 * conflicts in case kvm is already setup for another implementation.
4149
	 */
P
Paolo Bonzini 已提交
4150 4151 4152
	r = kvm_irqfd_init();
	if (r)
		goto out_irqfd;
4153

4154
	if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
4155 4156 4157 4158
		r = -ENOMEM;
		goto out_free_0;
	}

4159
	r = kvm_arch_hardware_setup();
A
Avi Kivity 已提交
4160
	if (r < 0)
4161
		goto out_free_0a;
A
Avi Kivity 已提交
4162

Y
Yang, Sheng 已提交
4163 4164
	for_each_online_cpu(cpu) {
		smp_call_function_single(cpu,
4165
				kvm_arch_check_processor_compat,
4166
				&r, 1);
Y
Yang, Sheng 已提交
4167
		if (r < 0)
4168
			goto out_free_1;
Y
Yang, Sheng 已提交
4169 4170
	}

T
Thomas Gleixner 已提交
4171
	r = cpuhp_setup_state_nocalls(CPUHP_AP_KVM_STARTING, "kvm/cpu:starting",
4172
				      kvm_starting_cpu, kvm_dying_cpu);
A
Avi Kivity 已提交
4173
	if (r)
4174
		goto out_free_2;
A
Avi Kivity 已提交
4175 4176
	register_reboot_notifier(&kvm_reboot_notifier);

4177
	/* A kmem cache lets us meet the alignment requirements of fx_save. */
4178 4179
	if (!vcpu_align)
		vcpu_align = __alignof__(struct kvm_vcpu);
4180 4181 4182 4183 4184 4185
	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);
4186 4187
	if (!kvm_vcpu_cache) {
		r = -ENOMEM;
4188
		goto out_free_3;
4189 4190
	}

4191 4192 4193 4194
	r = kvm_async_pf_init();
	if (r)
		goto out_free;

A
Avi Kivity 已提交
4195
	kvm_chardev_ops.owner = module;
4196 4197
	kvm_vm_fops.owner = module;
	kvm_vcpu_fops.owner = module;
A
Avi Kivity 已提交
4198 4199 4200

	r = misc_register(&kvm_dev);
	if (r) {
X
Xiubo Li 已提交
4201
		pr_err("kvm: misc device register failed\n");
4202
		goto out_unreg;
A
Avi Kivity 已提交
4203 4204
	}

4205 4206
	register_syscore_ops(&kvm_syscore_ops);

4207 4208 4209
	kvm_preempt_ops.sched_in = kvm_sched_in;
	kvm_preempt_ops.sched_out = kvm_sched_out;

4210
	kvm_init_debug();
4211

P
Paolo Bonzini 已提交
4212 4213 4214
	r = kvm_vfio_ops_init();
	WARN_ON(r);

4215
	return 0;
A
Avi Kivity 已提交
4216

4217 4218
out_unreg:
	kvm_async_pf_deinit();
A
Avi Kivity 已提交
4219
out_free:
4220
	kmem_cache_destroy(kvm_vcpu_cache);
4221
out_free_3:
A
Avi Kivity 已提交
4222
	unregister_reboot_notifier(&kvm_reboot_notifier);
4223
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4224 4225
out_free_2:
out_free_1:
4226
	kvm_arch_hardware_unsetup();
4227 4228
out_free_0a:
	free_cpumask_var(cpus_hardware_enabled);
4229
out_free_0:
4230
	kvm_irqfd_exit();
P
Paolo Bonzini 已提交
4231
out_irqfd:
4232 4233
	kvm_arch_exit();
out_fail:
A
Avi Kivity 已提交
4234 4235
	return r;
}
4236
EXPORT_SYMBOL_GPL(kvm_init);
A
Avi Kivity 已提交
4237

4238
void kvm_exit(void)
A
Avi Kivity 已提交
4239
{
4240
	debugfs_remove_recursive(kvm_debugfs_dir);
A
Avi Kivity 已提交
4241
	misc_deregister(&kvm_dev);
4242
	kmem_cache_destroy(kvm_vcpu_cache);
4243
	kvm_async_pf_deinit();
4244
	unregister_syscore_ops(&kvm_syscore_ops);
A
Avi Kivity 已提交
4245
	unregister_reboot_notifier(&kvm_reboot_notifier);
4246
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4247
	on_each_cpu(hardware_disable_nolock, NULL, 1);
4248
	kvm_arch_hardware_unsetup();
4249
	kvm_arch_exit();
4250
	kvm_irqfd_exit();
4251
	free_cpumask_var(cpus_hardware_enabled);
4252
	kvm_vfio_ops_exit();
A
Avi Kivity 已提交
4253
}
4254
EXPORT_SYMBOL_GPL(kvm_exit);