kvm_main.c 111.3 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-only
A
Avi Kivity 已提交
2 3 4 5 6 7 8
/*
 * 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
9
 * Copyright 2010 Red Hat, Inc. and/or its affiliates.
A
Avi Kivity 已提交
10 11 12 13 14 15
 *
 * Authors:
 *   Avi Kivity   <avi@qumranet.com>
 *   Yaniv Kamay  <yaniv@qumranet.com>
 */

16
#include <kvm/iodev.h>
A
Avi Kivity 已提交
17

18
#include <linux/kvm_host.h>
A
Avi Kivity 已提交
19 20 21 22 23 24 25 26 27 28 29
#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>
30
#include <linux/syscore_ops.h>
A
Avi Kivity 已提交
31
#include <linux/cpu.h>
32
#include <linux/sched/signal.h>
33
#include <linux/sched/mm.h>
34
#include <linux/sched/stat.h>
35 36
#include <linux/cpumask.h>
#include <linux/smp.h>
37
#include <linux/anon_inodes.h>
38
#include <linux/profile.h>
39
#include <linux/kvm_para.h>
40
#include <linux/pagemap.h>
41
#include <linux/mman.h>
42
#include <linux/swap.h>
43
#include <linux/bitops.h>
44
#include <linux/spinlock.h>
45
#include <linux/compat.h>
46
#include <linux/srcu.h>
J
Joerg Roedel 已提交
47
#include <linux/hugetlb.h>
48
#include <linux/slab.h>
49 50
#include <linux/sort.h>
#include <linux/bsearch.h>
P
Paolo Bonzini 已提交
51
#include <linux/io.h>
52
#include <linux/lockdep.h>
53
#include <linux/kthread.h>
A
Avi Kivity 已提交
54

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

60
#include "coalesced_mmio.h"
61
#include "async_pf.h"
P
Paolo Bonzini 已提交
62
#include "vfio.h"
63

64 65 66
#define CREATE_TRACE_POINTS
#include <trace/events/kvm.h>

67 68 69
/* Worst case buffer size needed for holding an integer. */
#define ITOA_MAX_LEN 12

A
Avi Kivity 已提交
70 71 72
MODULE_AUTHOR("Qumranet");
MODULE_LICENSE("GPL");

73
/* Architectures should define their poll value according to the halt latency */
74
unsigned int halt_poll_ns = KVM_HALT_POLL_NS_DEFAULT;
75
module_param(halt_poll_ns, uint, 0644);
76
EXPORT_SYMBOL_GPL(halt_poll_ns);
77

W
Wanpeng Li 已提交
78
/* Default doubles per-vcpu halt_poll_ns. */
79
unsigned int halt_poll_ns_grow = 2;
80
module_param(halt_poll_ns_grow, uint, 0644);
81
EXPORT_SYMBOL_GPL(halt_poll_ns_grow);
W
Wanpeng Li 已提交
82

83 84 85 86 87
/* 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
88
/* Default resets per-vcpu halt_poll_ns . */
89
unsigned int halt_poll_ns_shrink;
90
module_param(halt_poll_ns_shrink, uint, 0644);
91
EXPORT_SYMBOL_GPL(halt_poll_ns_shrink);
W
Wanpeng Li 已提交
92

93 94 95
/*
 * Ordering of locks:
 *
96
 *	kvm->lock --> kvm->slots_lock --> kvm->irq_lock
97 98
 */

J
Junaid Shahid 已提交
99
DEFINE_MUTEX(kvm_lock);
100
static DEFINE_RAW_SPINLOCK(kvm_count_lock);
101
LIST_HEAD(vm_list);
102

103
static cpumask_var_t cpus_hardware_enabled;
104
static int kvm_usage_count;
105
static atomic_t hardware_enable_failed;
106

107
static struct kmem_cache *kvm_vcpu_cache;
A
Avi Kivity 已提交
108

109
static __read_mostly struct preempt_ops kvm_preempt_ops;
110
static DEFINE_PER_CPU(struct kvm_vcpu *, kvm_running_vcpu);
111

112
struct dentry *kvm_debugfs_dir;
113
EXPORT_SYMBOL_GPL(kvm_debugfs_dir);
A
Avi Kivity 已提交
114

115
static int kvm_debugfs_num_entries;
116
static const struct file_operations stat_fops_per_vm;
117

A
Avi Kivity 已提交
118 119
static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl,
			   unsigned long arg);
120
#ifdef CONFIG_KVM_COMPAT
121 122
static long kvm_vcpu_compat_ioctl(struct file *file, unsigned int ioctl,
				  unsigned long arg);
123 124
#define KVM_COMPAT(c)	.compat_ioctl	= (c)
#else
125 126 127 128 129 130 131
/*
 * For architectures that don't implement a compat infrastructure,
 * adopt a double line of defense:
 * - Prevent a compat task from opening /dev/kvm
 * - If the open has been done by a 64bit task, and the KVM fd
 *   passed to a compat task, let the ioctls fail.
 */
132 133
static long kvm_no_compat_ioctl(struct file *file, unsigned int ioctl,
				unsigned long arg) { return -EINVAL; }
134 135 136 137 138 139 140

static int kvm_no_compat_open(struct inode *inode, struct file *file)
{
	return is_compat_task() ? -ENODEV : 0;
}
#define KVM_COMPAT(c)	.compat_ioctl	= kvm_no_compat_ioctl,	\
			.open		= kvm_no_compat_open
141
#endif
142 143
static int hardware_enable_all(void);
static void hardware_disable_all(void);
A
Avi Kivity 已提交
144

M
Marcelo Tosatti 已提交
145
static void kvm_io_bus_destroy(struct kvm_io_bus *bus);
146

147
static void mark_page_dirty_in_slot(struct kvm_memory_slot *memslot, gfn_t gfn);
M
Marcelo Tosatti 已提交
148

149
__visible bool kvm_rebooting;
150
EXPORT_SYMBOL_GPL(kvm_rebooting);
151

152 153
static bool largepages_enabled = true;

154 155 156 157 158 159
#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;

160 161
__weak int kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
		unsigned long start, unsigned long end, bool blockable)
162
{
163
	return 0;
164 165
}

166 167 168 169 170 171 172 173 174 175 176 177 178 179
bool kvm_is_zone_device_pfn(kvm_pfn_t pfn)
{
	/*
	 * The metadata used by is_zone_device_page() to determine whether or
	 * not a page is ZONE_DEVICE is guaranteed to be valid if and only if
	 * the device has been pinned, e.g. by get_user_pages().  WARN if the
	 * page_count() is zero to help detect bad usage of this helper.
	 */
	if (!pfn_valid(pfn) || WARN_ON_ONCE(!page_count(pfn_to_page(pfn))))
		return false;

	return is_zone_device_page(pfn_to_page(pfn));
}

D
Dan Williams 已提交
180
bool kvm_is_reserved_pfn(kvm_pfn_t pfn)
B
Ben-Ami Yassour 已提交
181
{
182 183 184 185 186
	/*
	 * ZONE_DEVICE pages currently set PG_reserved, but from a refcounting
	 * perspective they are "normal" pages, albeit with slightly different
	 * usage rules.
	 */
187
	if (pfn_valid(pfn))
188
		return PageReserved(pfn_to_page(pfn)) &&
189
		       !is_zero_pfn(pfn) &&
190
		       !kvm_is_zone_device_pfn(pfn);
B
Ben-Ami Yassour 已提交
191 192 193 194

	return true;
}

195 196 197 198 199 200 201 202 203 204
bool kvm_is_transparent_hugepage(kvm_pfn_t pfn)
{
	struct page *page = pfn_to_page(pfn);

	if (!PageTransCompoundMap(page))
		return false;

	return is_transparent_hugepage(compound_head(page));
}

A
Avi Kivity 已提交
205 206 207
/*
 * Switches to specified vcpu, until a matching vcpu_put()
 */
208
void vcpu_load(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
209
{
210
	int cpu = get_cpu();
211 212

	__this_cpu_write(kvm_running_vcpu, vcpu);
213
	preempt_notifier_register(&vcpu->preempt_notifier);
214
	kvm_arch_vcpu_load(vcpu, cpu);
215
	put_cpu();
A
Avi Kivity 已提交
216
}
217
EXPORT_SYMBOL_GPL(vcpu_load);
A
Avi Kivity 已提交
218

219
void vcpu_put(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
220
{
221
	preempt_disable();
222
	kvm_arch_vcpu_put(vcpu);
223
	preempt_notifier_unregister(&vcpu->preempt_notifier);
224
	__this_cpu_write(kvm_running_vcpu, NULL);
225
	preempt_enable();
A
Avi Kivity 已提交
226
}
227
EXPORT_SYMBOL_GPL(vcpu_put);
A
Avi Kivity 已提交
228

229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246
/* 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;
}

247 248 249 250
static void ack_flush(void *_completed)
{
}

251 252 253 254 255 256 257 258 259 260 261 262
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;
}

263 264
bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
				 unsigned long *vcpu_bitmap, cpumask_var_t tmp)
265
{
266
	int i, cpu, me;
267
	struct kvm_vcpu *vcpu;
268
	bool called;
269

270
	me = get_cpu();
271

272
	kvm_for_each_vcpu(i, vcpu, kvm) {
273
		if (vcpu_bitmap && !test_bit(i, vcpu_bitmap))
274 275
			continue;

276
		kvm_make_request(req, vcpu);
277
		cpu = vcpu->cpu;
278

279 280
		if (!(req & KVM_REQUEST_NO_WAKEUP) && kvm_vcpu_wake_up(vcpu))
			continue;
281

282
		if (tmp != NULL && cpu != -1 && cpu != me &&
283
		    kvm_request_needs_ipi(vcpu, req))
284
			__cpumask_set_cpu(cpu, tmp);
285
	}
286 287

	called = kvm_kick_many_cpus(tmp, !!(req & KVM_REQUEST_WAIT));
288
	put_cpu();
289 290 291 292 293 294 295 296 297 298 299

	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);

300
	called = kvm_make_vcpus_request_mask(kvm, req, NULL, cpus);
301

302
	free_cpumask_var(cpus);
303
	return called;
304 305
}

306
#ifndef CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALL
307
void kvm_flush_remote_tlbs(struct kvm *kvm)
308
{
309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
	/*
	 * 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.
	 */
326 327
	if (!kvm_arch_flush_remote_tlb(kvm)
	    || kvm_make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH))
328
		++kvm->stat.remote_tlb_flush;
329
	cmpxchg(&kvm->tlbs_dirty, dirty_count, 0);
330
}
331
EXPORT_SYMBOL_GPL(kvm_flush_remote_tlbs);
332
#endif
333

334 335
void kvm_reload_remote_mmus(struct kvm *kvm)
{
336
	kvm_make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD);
337
}
338

339
static void kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id)
R
Rusty Russell 已提交
340 341 342 343 344
{
	mutex_init(&vcpu->mutex);
	vcpu->cpu = -1;
	vcpu->kvm = kvm;
	vcpu->vcpu_id = id;
345
	vcpu->pid = NULL;
346
	init_swait_queue_head(&vcpu->wq);
347
	kvm_async_pf_vcpu_init(vcpu);
R
Rusty Russell 已提交
348

349 350 351
	vcpu->pre_pcpu = -1;
	INIT_LIST_HEAD(&vcpu->blocked_vcpu_list);

352 353
	kvm_vcpu_set_in_spin_loop(vcpu, false);
	kvm_vcpu_set_dy_eligible(vcpu, false);
354
	vcpu->preempted = false;
355
	vcpu->ready = false;
356
	preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);
R
Rusty Russell 已提交
357 358
}

359 360 361
void kvm_vcpu_destroy(struct kvm_vcpu *vcpu)
{
	kvm_arch_vcpu_destroy(vcpu);
362

363 364 365 366 367 368 369
	/*
	 * No need for rcu_read_lock as VCPU_RUN is the only place that changes
	 * the vcpu->pid pointer, and at destruction time all file descriptors
	 * are already gone.
	 */
	put_pid(rcu_dereference_protected(vcpu->pid, 1));

370
	free_page((unsigned long)vcpu->run);
371
	kmem_cache_free(kvm_vcpu_cache, vcpu);
372 373 374
}
EXPORT_SYMBOL_GPL(kvm_vcpu_destroy);

375 376 377 378 379 380
#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);
}

381 382 383 384 385 386
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);
387
	int idx;
388

389
	idx = srcu_read_lock(&kvm->srcu);
390 391
	spin_lock(&kvm->mmu_lock);
	kvm->mmu_notifier_seq++;
392 393 394 395

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

396
	spin_unlock(&kvm->mmu_lock);
397
	srcu_read_unlock(&kvm->srcu, idx);
398 399
}

400
static int kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn,
401
					const struct mmu_notifier_range *range)
402 403
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
404
	int need_tlb_flush = 0, idx;
405
	int ret;
406

407
	idx = srcu_read_lock(&kvm->srcu);
408 409 410 411 412 413 414
	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++;
415
	need_tlb_flush = kvm_unmap_hva_range(kvm, range->start, range->end);
416
	need_tlb_flush |= kvm->tlbs_dirty;
417 418 419
	/* we've to flush the tlb before the pages can be freed */
	if (need_tlb_flush)
		kvm_flush_remote_tlbs(kvm);
420 421

	spin_unlock(&kvm->mmu_lock);
422

423
	ret = kvm_arch_mmu_notifier_invalidate_range(kvm, range->start,
424 425
					range->end,
					mmu_notifier_range_blockable(range));
426

427
	srcu_read_unlock(&kvm->srcu, idx);
428 429

	return ret;
430 431 432
}

static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
433
					const struct mmu_notifier_range *range)
434 435 436 437 438 439 440 441 442 443
{
	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++;
444
	smp_wmb();
445 446
	/*
	 * The above sequence increase must be visible before the
447 448
	 * below count decrease, which is ensured by the smp_wmb above
	 * in conjunction with the smp_rmb in mmu_notifier_retry().
449 450 451 452 453 454 455 456 457
	 */
	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
458 459
					      unsigned long start,
					      unsigned long end)
460 461
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
462
	int young, idx;
463

464
	idx = srcu_read_lock(&kvm->srcu);
465 466
	spin_lock(&kvm->mmu_lock);

A
Andres Lagar-Cavilla 已提交
467
	young = kvm_age_hva(kvm, start, end);
468 469 470
	if (young)
		kvm_flush_remote_tlbs(kvm);

471 472 473
	spin_unlock(&kvm->mmu_lock);
	srcu_read_unlock(&kvm->srcu, idx);

474 475 476
	return young;
}

477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506
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
507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522
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;
}

523 524 525 526
static void kvm_mmu_notifier_release(struct mmu_notifier *mn,
				     struct mm_struct *mm)
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
527 528 529
	int idx;

	idx = srcu_read_lock(&kvm->srcu);
530
	kvm_arch_flush_shadow_all(kvm);
531
	srcu_read_unlock(&kvm->srcu, idx);
532 533
}

534 535 536 537
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,
538
	.clear_young		= kvm_mmu_notifier_clear_young,
A
Andrea Arcangeli 已提交
539
	.test_young		= kvm_mmu_notifier_test_young,
540
	.change_pte		= kvm_mmu_notifier_change_pte,
541
	.release		= kvm_mmu_notifier_release,
542
};
543 544 545 546 547 548 549 550 551 552 553 554 555 556

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;
}

557 558
#endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */

559
static struct kvm_memslots *kvm_alloc_memslots(void)
560 561
{
	int i;
562
	struct kvm_memslots *slots;
563

564
	slots = kvzalloc(sizeof(struct kvm_memslots), GFP_KERNEL_ACCOUNT);
565 566 567
	if (!slots)
		return NULL;

568
	for (i = 0; i < KVM_MEM_SLOTS_NUM; i++)
569
		slots->id_to_index[i] = slots->memslots[i].id = i;
570 571 572 573 574 575 576 577 578 579 580 581 582

	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;
}

583
static void kvm_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot)
584
{
585
	kvm_destroy_dirty_bitmap(slot);
586

587
	kvm_arch_free_memslot(kvm, slot);
588

589 590
	slot->flags = 0;
	slot->npages = 0;
591 592 593 594 595 596 597 598 599 600
}

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)
601
		kvm_free_memslot(kvm, memslot);
602 603

	kvfree(slots);
604 605
}

606 607 608 609 610 611 612 613 614
static void kvm_destroy_vm_debugfs(struct kvm *kvm)
{
	int i;

	if (!kvm->debugfs_dentry)
		return;

	debugfs_remove_recursive(kvm->debugfs_dentry);

615 616 617 618 619
	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);
	}
620 621 622 623 624 625 626 627 628 629 630 631
}

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);
632
	kvm->debugfs_dentry = debugfs_create_dir(dir_name, kvm_debugfs_dir);
633 634 635

	kvm->debugfs_stat_data = kcalloc(kvm_debugfs_num_entries,
					 sizeof(*kvm->debugfs_stat_data),
636
					 GFP_KERNEL_ACCOUNT);
637 638 639 640
	if (!kvm->debugfs_stat_data)
		return -ENOMEM;

	for (p = debugfs_entries; p->name; p++) {
641
		stat_data = kzalloc(sizeof(*stat_data), GFP_KERNEL_ACCOUNT);
642 643 644 645
		if (!stat_data)
			return -ENOMEM;

		stat_data->kvm = kvm;
646
		stat_data->dbgfs_item = p;
647
		kvm->debugfs_stat_data[p - debugfs_entries] = stat_data;
648 649 650
		debugfs_create_file(p->name, KVM_DBGFS_GET_MODE(p),
				    kvm->debugfs_dentry, stat_data,
				    &stat_fops_per_vm);
651 652 653 654
	}
	return 0;
}

655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671
/*
 * Called after the VM is otherwise initialized, but just before adding it to
 * the vm_list.
 */
int __weak kvm_arch_post_init_vm(struct kvm *kvm)
{
	return 0;
}

/*
 * Called just after removing the VM from the vm_list, but before doing any
 * other destruction.
 */
void __weak kvm_arch_pre_destroy_vm(struct kvm *kvm)
{
}

672
static struct kvm *kvm_create_vm(unsigned long type)
A
Avi Kivity 已提交
673
{
674
	struct kvm *kvm = kvm_arch_alloc_vm();
675 676
	int r = -ENOMEM;
	int i;
A
Avi Kivity 已提交
677

678 679 680
	if (!kvm)
		return ERR_PTR(-ENOMEM);

681
	spin_lock_init(&kvm->mmu_lock);
V
Vegard Nossum 已提交
682
	mmgrab(current->mm);
683 684 685 686 687 688 689
	kvm->mm = current->mm;
	kvm_eventfd_init(kvm);
	mutex_init(&kvm->lock);
	mutex_init(&kvm->irq_lock);
	mutex_init(&kvm->slots_lock);
	INIT_LIST_HEAD(&kvm->devices);

690 691
	BUILD_BUG_ON(KVM_MEM_SLOTS_NUM > SHRT_MAX);

692 693 694 695 696
	if (init_srcu_struct(&kvm->srcu))
		goto out_err_no_srcu;
	if (init_srcu_struct(&kvm->irq_srcu))
		goto out_err_no_irq_srcu;

697
	refcount_set(&kvm->users_count, 1);
698
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
699
		struct kvm_memslots *slots = kvm_alloc_memslots();
700

701
		if (!slots)
702
			goto out_err_no_arch_destroy_vm;
703
		/* Generations must be different for each address space. */
704
		slots->generation = i;
705
		rcu_assign_pointer(kvm->memslots[i], slots);
706
	}
707

M
Marcelo Tosatti 已提交
708
	for (i = 0; i < KVM_NR_BUSES; i++) {
709
		rcu_assign_pointer(kvm->buses[i],
710
			kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL_ACCOUNT));
711
		if (!kvm->buses[i])
712
			goto out_err_no_arch_destroy_vm;
M
Marcelo Tosatti 已提交
713
	}
714

715
	r = kvm_arch_init_vm(kvm, type);
716
	if (r)
717
		goto out_err_no_arch_destroy_vm;
718 719 720

	r = hardware_enable_all();
	if (r)
721
		goto out_err_no_disable;
722

723
#ifdef CONFIG_HAVE_KVM_IRQFD
724
	INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list);
725
#endif
A
Avi Kivity 已提交
726

727
	r = kvm_init_mmu_notifier(kvm);
728 729 730 731
	if (r)
		goto out_err_no_mmu_notifier;

	r = kvm_arch_post_init_vm(kvm);
732 733 734
	if (r)
		goto out_err;

J
Junaid Shahid 已提交
735
	mutex_lock(&kvm_lock);
736
	list_add(&kvm->vm_list, &vm_list);
J
Junaid Shahid 已提交
737
	mutex_unlock(&kvm_lock);
738

739 740
	preempt_notifier_inc();

741
	return kvm;
742 743

out_err:
744 745 746 747 748
#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
	if (kvm->mmu_notifier.ops)
		mmu_notifier_unregister(&kvm->mmu_notifier, current->mm);
#endif
out_err_no_mmu_notifier:
749
	hardware_disable_all();
750
out_err_no_disable:
751 752
	kvm_arch_destroy_vm(kvm);
out_err_no_arch_destroy_vm:
753
	WARN_ON_ONCE(!refcount_dec_and_test(&kvm->users_count));
M
Marcelo Tosatti 已提交
754
	for (i = 0; i < KVM_NR_BUSES; i++)
755
		kfree(kvm_get_bus(kvm, i));
756
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
757
		kvm_free_memslots(kvm, __kvm_memslots(kvm, i));
758 759 760 761
	cleanup_srcu_struct(&kvm->irq_srcu);
out_err_no_irq_srcu:
	cleanup_srcu_struct(&kvm->srcu);
out_err_no_srcu:
762
	kvm_arch_free_vm(kvm);
763
	mmdrop(current->mm);
764
	return ERR_PTR(r);
765 766
}

767 768
static void kvm_destroy_devices(struct kvm *kvm)
{
G
Geliang Tang 已提交
769
	struct kvm_device *dev, *tmp;
770

771 772 773 774 775
	/*
	 * 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
776 777
	list_for_each_entry_safe(dev, tmp, &kvm->devices, vm_node) {
		list_del(&dev->vm_node);
778 779 780 781
		dev->ops->destroy(dev);
	}
}

782 783
static void kvm_destroy_vm(struct kvm *kvm)
{
M
Marcelo Tosatti 已提交
784
	int i;
785 786
	struct mm_struct *mm = kvm->mm;

787
	kvm_uevent_notify_change(KVM_EVENT_DESTROY_VM, kvm);
788
	kvm_destroy_vm_debugfs(kvm);
789
	kvm_arch_sync_events(kvm);
J
Junaid Shahid 已提交
790
	mutex_lock(&kvm_lock);
791
	list_del(&kvm->vm_list);
J
Junaid Shahid 已提交
792
	mutex_unlock(&kvm_lock);
793 794
	kvm_arch_pre_destroy_vm(kvm);

795
	kvm_free_irq_routing(kvm);
796
	for (i = 0; i < KVM_NR_BUSES; i++) {
797
		struct kvm_io_bus *bus = kvm_get_bus(kvm, i);
798 799 800

		if (bus)
			kvm_io_bus_destroy(bus);
801 802
		kvm->buses[i] = NULL;
	}
803
	kvm_coalesced_mmio_free(kvm);
804 805
#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
	mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
806
#else
807
	kvm_arch_flush_shadow_all(kvm);
808
#endif
809
	kvm_arch_destroy_vm(kvm);
810
	kvm_destroy_devices(kvm);
811
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
812
		kvm_free_memslots(kvm, __kvm_memslots(kvm, i));
813
	cleanup_srcu_struct(&kvm->irq_srcu);
814 815
	cleanup_srcu_struct(&kvm->srcu);
	kvm_arch_free_vm(kvm);
816
	preempt_notifier_dec();
817
	hardware_disable_all();
818
	mmdrop(mm);
819 820
}

I
Izik Eidus 已提交
821 822
void kvm_get_kvm(struct kvm *kvm)
{
823
	refcount_inc(&kvm->users_count);
I
Izik Eidus 已提交
824 825 826 827 828
}
EXPORT_SYMBOL_GPL(kvm_get_kvm);

void kvm_put_kvm(struct kvm *kvm)
{
829
	if (refcount_dec_and_test(&kvm->users_count))
I
Izik Eidus 已提交
830 831 832 833
		kvm_destroy_vm(kvm);
}
EXPORT_SYMBOL_GPL(kvm_put_kvm);

834 835 836 837 838 839 840 841 842 843 844 845
/*
 * Used to put a reference that was taken on behalf of an object associated
 * with a user-visible file descriptor, e.g. a vcpu or device, if installation
 * of the new file descriptor fails and the reference cannot be transferred to
 * its final owner.  In such cases, the caller is still actively using @kvm and
 * will fail miserably if the refcount unexpectedly hits zero.
 */
void kvm_put_kvm_no_destroy(struct kvm *kvm)
{
	WARN_ON(refcount_dec_and_test(&kvm->users_count));
}
EXPORT_SYMBOL_GPL(kvm_put_kvm_no_destroy);
I
Izik Eidus 已提交
846

847 848 849 850
static int kvm_vm_release(struct inode *inode, struct file *filp)
{
	struct kvm *kvm = filp->private_data;

G
Gregory Haskins 已提交
851 852
	kvm_irqfd_release(kvm);

I
Izik Eidus 已提交
853
	kvm_put_kvm(kvm);
A
Avi Kivity 已提交
854 855 856
	return 0;
}

857 858
/*
 * Allocation size is twice as large as the actual dirty bitmap size.
859
 * See kvm_vm_ioctl_get_dirty_log() why this is needed.
860
 */
861 862
static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot)
{
863
	unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot);
864

865
	memslot->dirty_bitmap = kvzalloc(dirty_bytes, GFP_KERNEL_ACCOUNT);
866 867 868 869 870 871
	if (!memslot->dirty_bitmap)
		return -ENOMEM;

	return 0;
}

872
/*
873 874 875 876
 * 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.
877
 */
878
static void update_memslots(struct kvm_memslots *slots,
879 880
			    struct kvm_memory_slot *new,
			    enum kvm_mr_change change)
881
{
882 883
	int id = new->id;
	int i = slots->id_to_index[id];
884
	struct kvm_memory_slot *mslots = slots->memslots;
885

886
	WARN_ON(mslots[i].id != id);
887 888 889 890 891 892 893 894 895 896 897
	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;
898
	}
899

900
	while (i < KVM_MEM_SLOTS_NUM - 1 &&
901 902 903
	       new->base_gfn <= mslots[i + 1].base_gfn) {
		if (!mslots[i + 1].npages)
			break;
904 905 906 907
		mslots[i] = mslots[i + 1];
		slots->id_to_index[mslots[i].id] = i;
		i++;
	}
908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924

	/*
	 * 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--;
		}
925 926
	} else
		WARN_ON_ONCE(i != slots->used_slots);
927

928 929
	mslots[i] = *new;
	slots->id_to_index[mslots[i].id] = i;
930 931
}

932
static int check_memory_region_flags(const struct kvm_userspace_memory_region *mem)
933
{
X
Xiao Guangrong 已提交
934 935
	u32 valid_flags = KVM_MEM_LOG_DIRTY_PAGES;

936
#ifdef __KVM_HAVE_READONLY_MEM
X
Xiao Guangrong 已提交
937 938 939 940
	valid_flags |= KVM_MEM_READONLY;
#endif

	if (mem->flags & ~valid_flags)
941 942 943 944 945
		return -EINVAL;

	return 0;
}

946
static struct kvm_memslots *install_new_memslots(struct kvm *kvm,
947
		int as_id, struct kvm_memslots *slots)
948
{
949
	struct kvm_memslots *old_memslots = __kvm_memslots(kvm, as_id);
950
	u64 gen = old_memslots->generation;
951

952 953
	WARN_ON(gen & KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS);
	slots->generation = gen | KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS;
954

955
	rcu_assign_pointer(kvm->memslots[as_id], slots);
956
	synchronize_srcu_expedited(&kvm->srcu);
957

958
	/*
959
	 * Increment the new memslot generation a second time, dropping the
M
Miaohe Lin 已提交
960
	 * update in-progress flag and incrementing the generation based on
961 962 963 964 965 966
	 * 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;

	/*
967 968 969
	 * 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
970 971
	 * space 0 will use generations 0, 2, 4, ... while address space 1 will
	 * use generations 1, 3, 5, ...
972
	 */
973
	gen += KVM_ADDRESS_SPACE_NUM;
974

975
	kvm_arch_memslots_updated(kvm, gen);
976

977
	slots->generation = gen;
978 979

	return old_memslots;
980 981
}

982 983
static int kvm_set_memslot(struct kvm *kvm,
			   const struct kvm_userspace_memory_region *mem,
984
			   struct kvm_memory_slot *old,
985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041
			   struct kvm_memory_slot *new, int as_id,
			   enum kvm_mr_change change)
{
	struct kvm_memory_slot *slot;
	struct kvm_memslots *slots;
	int r;

	slots = kvzalloc(sizeof(struct kvm_memslots), GFP_KERNEL_ACCOUNT);
	if (!slots)
		return -ENOMEM;
	memcpy(slots, __kvm_memslots(kvm, as_id), sizeof(struct kvm_memslots));

	if (change == KVM_MR_DELETE || change == KVM_MR_MOVE) {
		/*
		 * Note, the INVALID flag needs to be in the appropriate entry
		 * in the freshly allocated memslots, not in @old or @new.
		 */
		slot = id_to_memslot(slots, old->id);
		slot->flags |= KVM_MEMSLOT_INVALID;

		/*
		 * We can re-use the old memslots, the only difference from the
		 * newly installed memslots is the invalid flag, which will get
		 * dropped by update_memslots anyway.  We'll also revert to the
		 * old memslots if preparing the new memory region fails.
		 */
		slots = install_new_memslots(kvm, as_id, slots);

		/* From this point no new shadow pages pointing to a deleted,
		 * or moved, memslot will be created.
		 *
		 * validation of sp->gfn happens in:
		 *	- gfn_to_hva (kvm_read_guest, gfn_to_pfn)
		 *	- kvm_is_visible_gfn (mmu_check_root)
		 */
		kvm_arch_flush_shadow_memslot(kvm, slot);
	}

	r = kvm_arch_prepare_memory_region(kvm, new, mem, change);
	if (r)
		goto out_slots;

	update_memslots(slots, new, change);
	slots = install_new_memslots(kvm, as_id, slots);

	kvm_arch_commit_memory_region(kvm, mem, old, new, change);

	kvfree(slots);
	return 0;

out_slots:
	if (change == KVM_MR_DELETE || change == KVM_MR_MOVE)
		slots = install_new_memslots(kvm, as_id, slots);
	kvfree(slots);
	return r;
}

1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058
static int kvm_delete_memslot(struct kvm *kvm,
			      const struct kvm_userspace_memory_region *mem,
			      struct kvm_memory_slot *old, int as_id)
{
	struct kvm_memory_slot new;
	int r;

	if (!old->npages)
		return -EINVAL;

	memset(&new, 0, sizeof(new));
	new.id = old->id;

	r = kvm_set_memslot(kvm, mem, old, &new, as_id, KVM_MR_DELETE);
	if (r)
		return r;

1059
	kvm_free_memslot(kvm, old);
1060 1061 1062
	return 0;
}

A
Avi Kivity 已提交
1063 1064 1065 1066 1067
/*
 * Allocate some memory and give it an address in the guest physical address
 * space.
 *
 * Discontiguous memory is allowed, mostly for framebuffers.
1068
 *
1069
 * Must be called holding kvm->slots_lock for write.
A
Avi Kivity 已提交
1070
 */
1071
int __kvm_set_memory_region(struct kvm *kvm,
1072
			    const struct kvm_userspace_memory_region *mem)
A
Avi Kivity 已提交
1073 1074
{
	struct kvm_memory_slot old, new;
1075
	struct kvm_memory_slot *tmp;
1076
	enum kvm_mr_change change;
1077 1078
	int as_id, id;
	int r;
A
Avi Kivity 已提交
1079

1080 1081
	r = check_memory_region_flags(mem);
	if (r)
1082
		return r;
1083

1084 1085 1086
	as_id = mem->slot >> 16;
	id = (u16)mem->slot;

A
Avi Kivity 已提交
1087 1088
	/* General sanity checks */
	if (mem->memory_size & (PAGE_SIZE - 1))
1089
		return -EINVAL;
A
Avi Kivity 已提交
1090
	if (mem->guest_phys_addr & (PAGE_SIZE - 1))
1091
		return -EINVAL;
1092
	/* We can read the guest memory with __xxx_user() later on. */
1093
	if ((id < KVM_USER_MEM_SLOTS) &&
1094
	    ((mem->userspace_addr & (PAGE_SIZE - 1)) ||
1095
	     !access_ok((void __user *)(unsigned long)mem->userspace_addr,
1096
			mem->memory_size)))
1097
		return -EINVAL;
1098
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_MEM_SLOTS_NUM)
1099
		return -EINVAL;
A
Avi Kivity 已提交
1100
	if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
1101
		return -EINVAL;
A
Avi Kivity 已提交
1102

1103 1104 1105 1106
	/*
	 * Make a full copy of the old memslot, the pointer will become stale
	 * when the memslots are re-sorted by update_memslots(), and the old
	 * memslot needs to be referenced after calling update_memslots(), e.g.
1107
	 * to free its resources and for arch specific behavior.
1108
	 */
1109
	old = *id_to_memslot(__kvm_memslots(kvm, as_id), id);
1110

1111 1112 1113
	if (!mem->memory_size)
		return kvm_delete_memslot(kvm, mem, &old, as_id);

1114
	new.id = id;
1115 1116
	new.base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
	new.npages = mem->memory_size >> PAGE_SHIFT;
A
Avi Kivity 已提交
1117
	new.flags = mem->flags;
1118
	new.userspace_addr = mem->userspace_addr;
A
Avi Kivity 已提交
1119

1120 1121 1122
	if (new.npages > KVM_MEM_MAX_NR_PAGES)
		return -EINVAL;

1123 1124
	if (!old.npages) {
		change = KVM_MR_CREATE;
1125 1126
		new.dirty_bitmap = NULL;
		memset(&new.arch, 0, sizeof(new.arch));
1127 1128
	} else { /* Modify an existing slot. */
		if ((new.userspace_addr != old.userspace_addr) ||
1129
		    (new.npages != old.npages) ||
1130
		    ((new.flags ^ old.flags) & KVM_MEM_READONLY))
1131
			return -EINVAL;
1132

1133
		if (new.base_gfn != old.base_gfn)
1134 1135 1136 1137 1138
			change = KVM_MR_MOVE;
		else if (new.flags != old.flags)
			change = KVM_MR_FLAGS_ONLY;
		else /* Nothing to change. */
			return 0;
1139 1140 1141 1142

		/* Copy dirty_bitmap and arch from the current memslot. */
		new.dirty_bitmap = old.dirty_bitmap;
		memcpy(&new.arch, &old.arch, sizeof(new.arch));
1143
	}
A
Avi Kivity 已提交
1144

1145
	if ((change == KVM_MR_CREATE) || (change == KVM_MR_MOVE)) {
1146
		/* Check for overlaps */
1147 1148
		kvm_for_each_memslot(tmp, __kvm_memslots(kvm, as_id)) {
			if (tmp->id == id)
1149
				continue;
1150 1151
			if (!((new.base_gfn + new.npages <= tmp->base_gfn) ||
			      (new.base_gfn >= tmp->base_gfn + tmp->npages)))
1152
				return -EEXIST;
1153
		}
A
Avi Kivity 已提交
1154 1155
	}

1156 1157 1158 1159
	/* Allocate/free page dirty bitmap as needed */
	if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
		new.dirty_bitmap = NULL;
	else if (!new.dirty_bitmap) {
1160 1161 1162
		r = kvm_create_dirty_bitmap(&new);
		if (r)
			return r;
A
Avi Kivity 已提交
1163 1164
	}

1165 1166 1167
	r = kvm_set_memslot(kvm, mem, &old, &new, as_id, change);
	if (r)
		goto out_bitmap;
1168

1169 1170
	if (old.dirty_bitmap && !new.dirty_bitmap)
		kvm_destroy_dirty_bitmap(&old);
A
Avi Kivity 已提交
1171 1172
	return 0;

1173 1174 1175
out_bitmap:
	if (new.dirty_bitmap && !old.dirty_bitmap)
		kvm_destroy_dirty_bitmap(&new);
A
Avi Kivity 已提交
1176
	return r;
1177
}
1178 1179 1180
EXPORT_SYMBOL_GPL(__kvm_set_memory_region);

int kvm_set_memory_region(struct kvm *kvm,
1181
			  const struct kvm_userspace_memory_region *mem)
1182 1183 1184
{
	int r;

1185
	mutex_lock(&kvm->slots_lock);
1186
	r = __kvm_set_memory_region(kvm, mem);
1187
	mutex_unlock(&kvm->slots_lock);
1188 1189
	return r;
}
1190 1191
EXPORT_SYMBOL_GPL(kvm_set_memory_region);

1192 1193
static int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
					  struct kvm_userspace_memory_region *mem)
1194
{
1195
	if ((u16)mem->slot >= KVM_USER_MEM_SLOTS)
1196
		return -EINVAL;
1197

1198
	return kvm_set_memory_region(kvm, mem);
A
Avi Kivity 已提交
1199 1200
}

1201
#ifndef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
1202 1203
int kvm_get_dirty_log(struct kvm *kvm,
			struct kvm_dirty_log *log, int *is_dirty)
A
Avi Kivity 已提交
1204
{
1205
	struct kvm_memslots *slots;
A
Avi Kivity 已提交
1206
	struct kvm_memory_slot *memslot;
1207
	int i, as_id, id;
1208
	unsigned long n;
A
Avi Kivity 已提交
1209 1210
	unsigned long any = 0;

1211 1212 1213
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1214
		return -EINVAL;
A
Avi Kivity 已提交
1215

1216 1217
	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
A
Avi Kivity 已提交
1218
	if (!memslot->dirty_bitmap)
1219
		return -ENOENT;
A
Avi Kivity 已提交
1220

1221
	n = kvm_dirty_bitmap_bytes(memslot);
A
Avi Kivity 已提交
1222

1223
	for (i = 0; !any && i < n/sizeof(long); ++i)
A
Avi Kivity 已提交
1224 1225 1226
		any = memslot->dirty_bitmap[i];

	if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
1227
		return -EFAULT;
A
Avi Kivity 已提交
1228

1229 1230
	if (any)
		*is_dirty = 1;
1231
	return 0;
A
Avi Kivity 已提交
1232
}
1233
EXPORT_SYMBOL_GPL(kvm_get_dirty_log);
A
Avi Kivity 已提交
1234

1235
#else /* CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
1236
/**
J
Jiang Biao 已提交
1237
 * kvm_get_dirty_log_protect - get a snapshot of dirty pages
1238
 *	and reenable dirty page tracking for the corresponding pages.
1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256
 * @kvm:	pointer to kvm instance
 * @log:	slot id and address to which we copy the log
 *
 * 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.
 *
 */
1257
static int kvm_get_dirty_log_protect(struct kvm *kvm, struct kvm_dirty_log *log)
1258
{
1259
	struct kvm_memslots *slots;
1260
	struct kvm_memory_slot *memslot;
1261
	int i, as_id, id;
1262 1263 1264
	unsigned long n;
	unsigned long *dirty_bitmap;
	unsigned long *dirty_bitmap_buffer;
1265
	bool flush;
1266

1267 1268 1269
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1270
		return -EINVAL;
1271

1272 1273
	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
1274 1275 1276

	dirty_bitmap = memslot->dirty_bitmap;
	if (!dirty_bitmap)
1277
		return -ENOENT;
1278

1279 1280
	kvm_arch_sync_dirty_log(kvm, memslot);

1281
	n = kvm_dirty_bitmap_bytes(memslot);
1282
	flush = false;
1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295
	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);
1296

1297 1298 1299 1300
		spin_lock(&kvm->mmu_lock);
		for (i = 0; i < n / sizeof(long); i++) {
			unsigned long mask;
			gfn_t offset;
1301

1302 1303 1304
			if (!dirty_bitmap[i])
				continue;

1305
			flush = true;
1306 1307 1308
			mask = xchg(&dirty_bitmap[i], 0);
			dirty_bitmap_buffer[i] = mask;

1309 1310 1311
			offset = i * BITS_PER_LONG;
			kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot,
								offset, mask);
1312 1313 1314 1315
		}
		spin_unlock(&kvm->mmu_lock);
	}

1316 1317 1318
	if (flush)
		kvm_arch_flush_remote_tlbs_memslot(kvm, memslot);

1319 1320 1321 1322
	if (copy_to_user(log->dirty_bitmap, dirty_bitmap_buffer, n))
		return -EFAULT;
	return 0;
}
1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355


/**
 * kvm_vm_ioctl_get_dirty_log - get and clear the log of dirty pages in a slot
 * @kvm: kvm instance
 * @log: slot id and address to which we copy the log
 *
 * Steps 1-4 below provide general overview of dirty page logging. See
 * kvm_get_dirty_log_protect() function description for additional details.
 *
 * We call kvm_get_dirty_log_protect() to handle steps 1-3, upon return we
 * always flush the TLB (step 4) even if previous step failed  and the dirty
 * bitmap may be corrupt. Regardless of previous outcome the KVM logging API
 * does not preclude user space subsequent dirty log read. Flushing TLB ensures
 * writes will be marked dirty for next log read.
 *
 *   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. Flush TLB's if needed.
 */
static int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
				      struct kvm_dirty_log *log)
{
	int r;

	mutex_lock(&kvm->slots_lock);

	r = kvm_get_dirty_log_protect(kvm, log);

	mutex_unlock(&kvm->slots_lock);
	return r;
}
1356 1357 1358 1359 1360 1361 1362

/**
 * 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
 */
1363 1364
static int kvm_clear_dirty_log_protect(struct kvm *kvm,
				       struct kvm_clear_dirty_log *log)
1365 1366 1367
{
	struct kvm_memslots *slots;
	struct kvm_memory_slot *memslot;
1368
	int as_id, id;
1369
	gfn_t offset;
1370
	unsigned long i, n;
1371 1372
	unsigned long *dirty_bitmap;
	unsigned long *dirty_bitmap_buffer;
1373
	bool flush;
1374 1375 1376 1377 1378 1379

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

1380
	if (log->first_page & 63)
1381 1382 1383 1384 1385 1386 1387 1388 1389
		return -EINVAL;

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

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

1390
	n = ALIGN(log->num_pages, BITS_PER_LONG) / 8;
1391 1392

	if (log->first_page > memslot->npages ||
1393 1394 1395
	    log->num_pages > memslot->npages - log->first_page ||
	    (log->num_pages < memslot->npages - log->first_page && (log->num_pages & 63)))
	    return -EINVAL;
1396

1397 1398 1399
	kvm_arch_sync_dirty_log(kvm, memslot);

	flush = false;
1400 1401 1402
	dirty_bitmap_buffer = kvm_second_dirty_bitmap(memslot);
	if (copy_from_user(dirty_bitmap_buffer, log->dirty_bitmap, n))
		return -EFAULT;
1403

1404
	spin_lock(&kvm->mmu_lock);
1405 1406
	for (offset = log->first_page, i = offset / BITS_PER_LONG,
		 n = DIV_ROUND_UP(log->num_pages, BITS_PER_LONG); n--;
1407 1408 1409 1410
	     i++, offset += BITS_PER_LONG) {
		unsigned long mask = *dirty_bitmap_buffer++;
		atomic_long_t *p = (atomic_long_t *) &dirty_bitmap[i];
		if (!mask)
1411 1412
			continue;

1413
		mask &= atomic_long_fetch_andnot(mask, p);
1414

1415 1416 1417 1418 1419 1420
		/*
		 * 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.
		*/
1421
		if (mask) {
1422
			flush = true;
1423 1424 1425
			kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot,
								offset, mask);
		}
1426 1427
	}
	spin_unlock(&kvm->mmu_lock);
1428

1429 1430 1431
	if (flush)
		kvm_arch_flush_remote_tlbs_memslot(kvm, memslot);

1432
	return 0;
1433
}
1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447

static int kvm_vm_ioctl_clear_dirty_log(struct kvm *kvm,
					struct kvm_clear_dirty_log *log)
{
	int r;

	mutex_lock(&kvm->slots_lock);

	r = kvm_clear_dirty_log_protect(kvm, log);

	mutex_unlock(&kvm->slots_lock);
	return r;
}
#endif /* CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
1448

1449 1450 1451 1452 1453
bool kvm_largepages_enabled(void)
{
	return largepages_enabled;
}

1454 1455 1456 1457 1458 1459
void kvm_disable_largepages(void)
{
	largepages_enabled = false;
}
EXPORT_SYMBOL_GPL(kvm_disable_largepages);

1460 1461 1462 1463
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
1464
EXPORT_SYMBOL_GPL(gfn_to_memslot);
A
Avi Kivity 已提交
1465

1466 1467 1468 1469 1470
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);
}

1471
bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
1472
{
1473
	struct kvm_memory_slot *memslot = gfn_to_memslot(kvm, gfn);
1474

1475
	if (!memslot || memslot->id >= KVM_USER_MEM_SLOTS ||
1476
	      memslot->flags & KVM_MEMSLOT_INVALID)
1477
		return false;
1478

1479
	return true;
1480 1481 1482
}
EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);

1483
unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn)
J
Joerg Roedel 已提交
1484 1485 1486 1487 1488 1489
{
	struct vm_area_struct *vma;
	unsigned long addr, size;

	size = PAGE_SIZE;

1490
	addr = kvm_vcpu_gfn_to_hva_prot(vcpu, gfn, NULL);
J
Joerg Roedel 已提交
1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506
	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
1507 1508 1509 1510 1511 1512 1513
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
1514
{
1515
	if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
X
Xiao Guangrong 已提交
1516
		return KVM_HVA_ERR_BAD;
1517

X
Xiao Guangrong 已提交
1518 1519
	if (memslot_is_readonly(slot) && write)
		return KVM_HVA_ERR_RO_BAD;
1520 1521 1522 1523

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

X
Xiao Guangrong 已提交
1524
	return __gfn_to_hva_memslot(slot, gfn);
I
Izik Eidus 已提交
1525
}
1526

X
Xiao Guangrong 已提交
1527 1528 1529 1530
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
1531
}
1532

X
Xiao Guangrong 已提交
1533
unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot,
1534
					gfn_t gfn)
X
Xiao Guangrong 已提交
1535 1536 1537 1538 1539
{
	return gfn_to_hva_many(slot, gfn, NULL);
}
EXPORT_SYMBOL_GPL(gfn_to_hva_memslot);

1540 1541
unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
{
1542
	return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL);
1543
}
1544
EXPORT_SYMBOL_GPL(gfn_to_hva);
I
Izik Eidus 已提交
1545

1546 1547 1548 1549 1550 1551
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);

1552
/*
1553 1554 1555 1556 1557 1558
 * 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
1559
 */
1560 1561
unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot,
				      gfn_t gfn, bool *writable)
1562
{
1563 1564 1565
	unsigned long hva = __gfn_to_hva_many(slot, gfn, NULL, false);

	if (!kvm_is_error_hva(hva) && writable)
1566 1567
		*writable = !memslot_is_readonly(slot);

1568
	return hva;
1569 1570
}

1571 1572 1573 1574 1575 1576 1577
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);
}

1578 1579 1580 1581 1582 1583 1584
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);
}

1585 1586
static inline int check_user_page_hwpoison(unsigned long addr)
{
L
Lorenzo Stoakes 已提交
1587
	int rc, flags = FOLL_HWPOISON | FOLL_WRITE;
1588

L
Lorenzo Stoakes 已提交
1589
	rc = get_user_pages(addr, 1, flags, NULL, NULL);
1590 1591 1592
	return rc == -EHWPOISON;
}

X
Xiao Guangrong 已提交
1593
/*
1594 1595
 * 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
M
Miaohe Lin 已提交
1596
 * only part that runs if we can in atomic context.
X
Xiao Guangrong 已提交
1597
 */
1598 1599
static bool hva_to_pfn_fast(unsigned long addr, bool write_fault,
			    bool *writable, kvm_pfn_t *pfn)
A
Avi Kivity 已提交
1600
{
1601
	struct page *page[1];
X
Xiao Guangrong 已提交
1602
	int npages;
A
Avi Kivity 已提交
1603

1604 1605 1606 1607 1608 1609 1610
	/*
	 * 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;
1611

X
Xiao Guangrong 已提交
1612 1613 1614
	npages = __get_user_pages_fast(addr, 1, 1, page);
	if (npages == 1) {
		*pfn = page_to_pfn(page[0]);
1615

X
Xiao Guangrong 已提交
1616 1617 1618 1619
		if (writable)
			*writable = true;
		return true;
	}
1620

X
Xiao Guangrong 已提交
1621 1622
	return false;
}
1623

X
Xiao Guangrong 已提交
1624 1625 1626 1627 1628
/*
 * 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
1629
			   bool *writable, kvm_pfn_t *pfn)
X
Xiao Guangrong 已提交
1630
{
1631 1632
	unsigned int flags = FOLL_HWPOISON;
	struct page *page;
X
Xiao Guangrong 已提交
1633
	int npages = 0;
1634

X
Xiao Guangrong 已提交
1635 1636 1637 1638 1639
	might_sleep();

	if (writable)
		*writable = write_fault;

1640 1641 1642 1643
	if (write_fault)
		flags |= FOLL_WRITE;
	if (async)
		flags |= FOLL_NOWAIT;
1644

1645
	npages = get_user_pages_unlocked(addr, 1, &page, flags);
X
Xiao Guangrong 已提交
1646 1647 1648 1649
	if (npages != 1)
		return npages;

	/* map read fault as writable if possible */
1650
	if (unlikely(!write_fault) && writable) {
1651
		struct page *wpage;
X
Xiao Guangrong 已提交
1652

1653
		if (__get_user_pages_fast(addr, 1, 1, &wpage) == 1) {
X
Xiao Guangrong 已提交
1654
			*writable = true;
1655 1656
			put_page(page);
			page = wpage;
1657
		}
1658
	}
1659
	*pfn = page_to_pfn(page);
X
Xiao Guangrong 已提交
1660 1661
	return npages;
}
I
Izik Eidus 已提交
1662

X
Xiao Guangrong 已提交
1663 1664 1665 1666
static bool vma_is_valid(struct vm_area_struct *vma, bool write_fault)
{
	if (unlikely(!(vma->vm_flags & VM_READ)))
		return false;
1667

X
Xiao Guangrong 已提交
1668 1669
	if (write_fault && (unlikely(!(vma->vm_flags & VM_WRITE))))
		return false;
1670

X
Xiao Guangrong 已提交
1671 1672
	return true;
}
1673

1674 1675
static int hva_to_pfn_remapped(struct vm_area_struct *vma,
			       unsigned long addr, bool *async,
1676 1677
			       bool write_fault, bool *writable,
			       kvm_pfn_t *p_pfn)
1678
{
1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702
	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;

	}

1703 1704
	if (writable)
		*writable = true;
1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719

	/*
	 * 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;
1720 1721 1722
	return 0;
}

1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736
/*
 * 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
1737
static kvm_pfn_t hva_to_pfn(unsigned long addr, bool atomic, bool *async,
X
Xiao Guangrong 已提交
1738 1739 1740
			bool write_fault, bool *writable)
{
	struct vm_area_struct *vma;
D
Dan Williams 已提交
1741
	kvm_pfn_t pfn = 0;
1742
	int npages, r;
1743

X
Xiao Guangrong 已提交
1744 1745
	/* we can do it either atomically or asynchronously, not both */
	BUG_ON(atomic && async);
1746

1747
	if (hva_to_pfn_fast(addr, write_fault, writable, &pfn))
X
Xiao Guangrong 已提交
1748 1749 1750 1751 1752 1753 1754 1755
		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;
1756

X
Xiao Guangrong 已提交
1757 1758 1759 1760 1761 1762 1763
	down_read(&current->mm->mmap_sem);
	if (npages == -EHWPOISON ||
	      (!async && check_user_page_hwpoison(addr))) {
		pfn = KVM_PFN_ERR_HWPOISON;
		goto exit;
	}

1764
retry:
X
Xiao Guangrong 已提交
1765 1766 1767 1768
	vma = find_vma_intersection(current->mm, addr, addr + 1);

	if (vma == NULL)
		pfn = KVM_PFN_ERR_FAULT;
1769
	else if (vma->vm_flags & (VM_IO | VM_PFNMAP)) {
1770
		r = hva_to_pfn_remapped(vma, addr, async, write_fault, writable, &pfn);
1771 1772
		if (r == -EAGAIN)
			goto retry;
1773 1774
		if (r < 0)
			pfn = KVM_PFN_ERR_FAULT;
X
Xiao Guangrong 已提交
1775
	} else {
X
Xiao Guangrong 已提交
1776
		if (async && vma_is_valid(vma, write_fault))
X
Xiao Guangrong 已提交
1777 1778 1779 1780 1781
			*async = true;
		pfn = KVM_PFN_ERR_FAULT;
	}
exit:
	up_read(&current->mm->mmap_sem);
1782
	return pfn;
1783 1784
}

D
Dan Williams 已提交
1785 1786 1787
kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
			       bool atomic, bool *async, bool write_fault,
			       bool *writable)
1788
{
X
Xiao Guangrong 已提交
1789 1790
	unsigned long addr = __gfn_to_hva_many(slot, gfn, NULL, write_fault);

1791 1792 1793
	if (addr == KVM_HVA_ERR_RO_BAD) {
		if (writable)
			*writable = false;
X
Xiao Guangrong 已提交
1794
		return KVM_PFN_ERR_RO_FAULT;
1795
	}
X
Xiao Guangrong 已提交
1796

1797 1798 1799
	if (kvm_is_error_hva(addr)) {
		if (writable)
			*writable = false;
1800
		return KVM_PFN_NOSLOT;
1801
	}
X
Xiao Guangrong 已提交
1802 1803 1804 1805 1806 1807 1808 1809 1810

	/* 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);
1811
}
1812
EXPORT_SYMBOL_GPL(__gfn_to_pfn_memslot);
1813

D
Dan Williams 已提交
1814
kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1815 1816
		      bool *writable)
{
P
Paolo Bonzini 已提交
1817 1818
	return __gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn, false, NULL,
				    write_fault, writable);
1819 1820 1821
}
EXPORT_SYMBOL_GPL(gfn_to_pfn_prot);

D
Dan Williams 已提交
1822
kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1823
{
X
Xiao Guangrong 已提交
1824
	return __gfn_to_pfn_memslot(slot, gfn, false, NULL, true, NULL);
1825
}
P
Paolo Bonzini 已提交
1826
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot);
1827

D
Dan Williams 已提交
1828
kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn)
1829
{
X
Xiao Guangrong 已提交
1830
	return __gfn_to_pfn_memslot(slot, gfn, true, NULL, true, NULL);
1831
}
1832
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot_atomic);
1833

D
Dan Williams 已提交
1834
kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1835 1836 1837 1838 1839
{
	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
1840
kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn)
1841 1842 1843 1844 1845
{
	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
1846
kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1847 1848 1849 1850 1851
{
	return gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn);
}
EXPORT_SYMBOL_GPL(gfn_to_pfn);

D
Dan Williams 已提交
1852
kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn)
1853 1854 1855 1856 1857
{
	return gfn_to_pfn_memslot(kvm_vcpu_gfn_to_memslot(vcpu, gfn), gfn);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_gfn_to_pfn);

1858 1859
int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
			    struct page **pages, int nr_pages)
1860 1861
{
	unsigned long addr;
1862
	gfn_t entry = 0;
1863

1864
	addr = gfn_to_hva_many(slot, gfn, &entry);
1865 1866 1867 1868 1869 1870 1871 1872 1873 1874
	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
1875
static struct page *kvm_pfn_to_page(kvm_pfn_t pfn)
X
Xiao Guangrong 已提交
1876
{
1877
	if (is_error_noslot_pfn(pfn))
1878
		return KVM_ERR_PTR_BAD_PAGE;
X
Xiao Guangrong 已提交
1879

1880
	if (kvm_is_reserved_pfn(pfn)) {
1881
		WARN_ON(1);
1882
		return KVM_ERR_PTR_BAD_PAGE;
1883
	}
X
Xiao Guangrong 已提交
1884 1885 1886 1887

	return pfn_to_page(pfn);
}

1888 1889
struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
{
D
Dan Williams 已提交
1890
	kvm_pfn_t pfn;
1891 1892 1893

	pfn = gfn_to_pfn(kvm, gfn);

X
Xiao Guangrong 已提交
1894
	return kvm_pfn_to_page(pfn);
A
Avi Kivity 已提交
1895 1896 1897
}
EXPORT_SYMBOL_GPL(gfn_to_page);

1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922
void kvm_release_pfn(kvm_pfn_t pfn, bool dirty, struct gfn_to_pfn_cache *cache)
{
	if (pfn == 0)
		return;

	if (cache)
		cache->pfn = cache->gfn = 0;

	if (dirty)
		kvm_release_pfn_dirty(pfn);
	else
		kvm_release_pfn_clean(pfn);
}

static void kvm_cache_gfn_to_pfn(struct kvm_memory_slot *slot, gfn_t gfn,
				 struct gfn_to_pfn_cache *cache, u64 gen)
{
	kvm_release_pfn(cache->pfn, cache->dirty, cache);

	cache->pfn = gfn_to_pfn_memslot(slot, gfn);
	cache->gfn = gfn;
	cache->dirty = false;
	cache->generation = gen;
}

1923
static int __kvm_map_gfn(struct kvm_memslots *slots, gfn_t gfn,
1924 1925 1926
			 struct kvm_host_map *map,
			 struct gfn_to_pfn_cache *cache,
			 bool atomic)
1927 1928 1929 1930
{
	kvm_pfn_t pfn;
	void *hva = NULL;
	struct page *page = KVM_UNMAPPED_PAGE;
1931
	struct kvm_memory_slot *slot = __gfn_to_memslot(slots, gfn);
1932
	u64 gen = slots->generation;
1933 1934 1935 1936

	if (!map)
		return -EINVAL;

1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949
	if (cache) {
		if (!cache->pfn || cache->gfn != gfn ||
			cache->generation != gen) {
			if (atomic)
				return -EAGAIN;
			kvm_cache_gfn_to_pfn(slot, gfn, cache, gen);
		}
		pfn = cache->pfn;
	} else {
		if (atomic)
			return -EAGAIN;
		pfn = gfn_to_pfn_memslot(slot, gfn);
	}
1950 1951 1952 1953 1954
	if (is_error_noslot_pfn(pfn))
		return -EINVAL;

	if (pfn_valid(pfn)) {
		page = pfn_to_page(pfn);
1955 1956 1957 1958
		if (atomic)
			hva = kmap_atomic(page);
		else
			hva = kmap(page);
P
Paolo Bonzini 已提交
1959
#ifdef CONFIG_HAS_IOMEM
1960
	} else if (!atomic) {
1961
		hva = memremap(pfn_to_hpa(pfn), PAGE_SIZE, MEMREMAP_WB);
1962 1963
	} else {
		return -EINVAL;
P
Paolo Bonzini 已提交
1964
#endif
1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977
	}

	if (!hva)
		return -EFAULT;

	map->page = page;
	map->hva = hva;
	map->pfn = pfn;
	map->gfn = gfn;

	return 0;
}

1978 1979
int kvm_map_gfn(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map,
		struct gfn_to_pfn_cache *cache, bool atomic)
1980
{
1981 1982
	return __kvm_map_gfn(kvm_memslots(vcpu->kvm), gfn, map,
			cache, atomic);
1983 1984 1985
}
EXPORT_SYMBOL_GPL(kvm_map_gfn);

1986 1987
int kvm_vcpu_map(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map)
{
1988 1989
	return __kvm_map_gfn(kvm_vcpu_memslots(vcpu), gfn, map,
		NULL, false);
1990 1991 1992
}
EXPORT_SYMBOL_GPL(kvm_vcpu_map);

1993
static void __kvm_unmap_gfn(struct kvm_memory_slot *memslot,
1994 1995 1996
			struct kvm_host_map *map,
			struct gfn_to_pfn_cache *cache,
			bool dirty, bool atomic)
1997 1998 1999 2000 2001 2002 2003
{
	if (!map)
		return;

	if (!map->hva)
		return;

2004 2005 2006 2007 2008 2009
	if (map->page != KVM_UNMAPPED_PAGE) {
		if (atomic)
			kunmap_atomic(map->hva);
		else
			kunmap(map->page);
	}
2010
#ifdef CONFIG_HAS_IOMEM
2011
	else if (!atomic)
2012
		memunmap(map->hva);
2013 2014
	else
		WARN_ONCE(1, "Unexpected unmapping in atomic context");
2015
#endif
2016

2017
	if (dirty)
2018
		mark_page_dirty_in_slot(memslot, map->gfn);
2019 2020 2021 2022 2023

	if (cache)
		cache->dirty |= dirty;
	else
		kvm_release_pfn(map->pfn, dirty, NULL);
2024 2025 2026 2027

	map->hva = NULL;
	map->page = NULL;
}
2028

2029 2030
int kvm_unmap_gfn(struct kvm_vcpu *vcpu, struct kvm_host_map *map, 
		  struct gfn_to_pfn_cache *cache, bool dirty, bool atomic)
2031
{
2032 2033
	__kvm_unmap_gfn(gfn_to_memslot(vcpu->kvm, map->gfn), map,
			cache, dirty, atomic);
2034 2035 2036 2037 2038 2039
	return 0;
}
EXPORT_SYMBOL_GPL(kvm_unmap_gfn);

void kvm_vcpu_unmap(struct kvm_vcpu *vcpu, struct kvm_host_map *map, bool dirty)
{
2040 2041
	__kvm_unmap_gfn(kvm_vcpu_gfn_to_memslot(vcpu, map->gfn), map, NULL,
			dirty, false);
2042
}
2043 2044
EXPORT_SYMBOL_GPL(kvm_vcpu_unmap);

2045 2046
struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn)
{
D
Dan Williams 已提交
2047
	kvm_pfn_t pfn;
2048 2049 2050 2051 2052 2053 2054

	pfn = kvm_vcpu_gfn_to_pfn(vcpu, gfn);

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

2055 2056
void kvm_release_page_clean(struct page *page)
{
2057 2058
	WARN_ON(is_error_page(page));

2059
	kvm_release_pfn_clean(page_to_pfn(page));
2060 2061 2062
}
EXPORT_SYMBOL_GPL(kvm_release_page_clean);

D
Dan Williams 已提交
2063
void kvm_release_pfn_clean(kvm_pfn_t pfn)
2064
{
2065
	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn))
2066
		put_page(pfn_to_page(pfn));
2067 2068 2069
}
EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);

2070
void kvm_release_page_dirty(struct page *page)
2071
{
X
Xiao Guangrong 已提交
2072 2073
	WARN_ON(is_error_page(page));

2074 2075 2076 2077
	kvm_release_pfn_dirty(page_to_pfn(page));
}
EXPORT_SYMBOL_GPL(kvm_release_page_dirty);

2078
void kvm_release_pfn_dirty(kvm_pfn_t pfn)
2079 2080 2081 2082
{
	kvm_set_pfn_dirty(pfn);
	kvm_release_pfn_clean(pfn);
}
2083
EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
2084

D
Dan Williams 已提交
2085
void kvm_set_pfn_dirty(kvm_pfn_t pfn)
2086
{
2087 2088
	if (!kvm_is_reserved_pfn(pfn) && !kvm_is_zone_device_pfn(pfn))
		SetPageDirty(pfn_to_page(pfn));
2089
}
2090 2091
EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);

D
Dan Williams 已提交
2092
void kvm_set_pfn_accessed(kvm_pfn_t pfn)
2093
{
2094
	if (!kvm_is_reserved_pfn(pfn) && !kvm_is_zone_device_pfn(pfn))
2095
		mark_page_accessed(pfn_to_page(pfn));
2096 2097 2098
}
EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);

D
Dan Williams 已提交
2099
void kvm_get_pfn(kvm_pfn_t pfn)
2100
{
2101
	if (!kvm_is_reserved_pfn(pfn))
2102
		get_page(pfn_to_page(pfn));
2103 2104
}
EXPORT_SYMBOL_GPL(kvm_get_pfn);
2105

2106 2107 2108 2109 2110 2111 2112 2113
static int next_segment(unsigned long len, int offset)
{
	if (len > PAGE_SIZE - offset)
		return PAGE_SIZE - offset;
	else
		return len;
}

2114 2115
static int __kvm_read_guest_page(struct kvm_memory_slot *slot, gfn_t gfn,
				 void *data, int offset, int len)
2116
{
2117 2118
	int r;
	unsigned long addr;
2119

2120
	addr = gfn_to_hva_memslot_prot(slot, gfn, NULL);
2121 2122
	if (kvm_is_error_hva(addr))
		return -EFAULT;
2123
	r = __copy_from_user(data, (void __user *)addr + offset, len);
2124
	if (r)
2125 2126 2127
		return -EFAULT;
	return 0;
}
2128 2129 2130 2131 2132 2133 2134 2135

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);
}
2136 2137
EXPORT_SYMBOL_GPL(kvm_read_guest_page);

2138 2139 2140 2141 2142 2143 2144 2145 2146
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);

2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166
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);

2167
int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data, unsigned long len)
2168 2169
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
2170
	int seg;
2171
	int offset = offset_in_page(gpa);
2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185
	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);
2186

2187 2188 2189 2190 2191 2192 2193
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);
2194 2195
	if (kvm_is_error_hva(addr))
		return -EFAULT;
2196
	pagefault_disable();
2197
	r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
2198
	pagefault_enable();
2199 2200 2201 2202 2203
	if (r)
		return -EFAULT;
	return 0;
}

2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216
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)
2217
{
2218 2219
	int r;
	unsigned long addr;
2220

2221
	addr = gfn_to_hva_memslot(memslot, gfn);
2222 2223
	if (kvm_is_error_hva(addr))
		return -EFAULT;
2224
	r = __copy_to_user((void __user *)addr + offset, data, len);
2225
	if (r)
2226
		return -EFAULT;
2227
	mark_page_dirty_in_slot(memslot, gfn);
2228 2229
	return 0;
}
2230 2231 2232 2233 2234 2235 2236 2237

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);
}
2238 2239
EXPORT_SYMBOL_GPL(kvm_write_guest_page);

2240 2241 2242 2243 2244 2245 2246 2247 2248
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);

2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267
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;
}
2268
EXPORT_SYMBOL_GPL(kvm_write_guest);
2269

2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290
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);

2291 2292 2293
static int __kvm_gfn_to_hva_cache_init(struct kvm_memslots *slots,
				       struct gfn_to_hva_cache *ghc,
				       gpa_t gpa, unsigned long len)
2294 2295
{
	int offset = offset_in_page(gpa);
2296 2297 2298 2299
	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;
2300

2301
	/* Update ghc->generation before performing any error checks. */
2302
	ghc->generation = slots->generation;
2303 2304 2305 2306 2307

	if (start_gfn > end_gfn) {
		ghc->hva = KVM_HVA_ERR_BAD;
		return -EINVAL;
	}
2308 2309 2310 2311 2312

	/*
	 * If the requested region crosses two memslots, we still
	 * verify that the entire region is valid here.
	 */
2313
	for ( ; start_gfn <= end_gfn; start_gfn += nr_pages_avail) {
2314 2315 2316 2317
		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))
2318
			return -EFAULT;
2319 2320 2321
	}

	/* Use the slow path for cross page reads and writes. */
2322
	if (nr_pages_needed == 1)
2323
		ghc->hva += offset;
2324
	else
2325
		ghc->memslot = NULL;
2326

2327 2328 2329
	ghc->gpa = gpa;
	ghc->len = len;
	return 0;
2330
}
2331

2332
int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
2333 2334
			      gpa_t gpa, unsigned long len)
{
2335
	struct kvm_memslots *slots = kvm_memslots(kvm);
2336 2337
	return __kvm_gfn_to_hva_cache_init(slots, ghc, gpa, len);
}
2338
EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init);
2339

2340
int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
2341 2342
				  void *data, unsigned int offset,
				  unsigned long len)
2343
{
2344
	struct kvm_memslots *slots = kvm_memslots(kvm);
2345
	int r;
2346
	gpa_t gpa = ghc->gpa + offset;
2347

2348
	BUG_ON(len + offset > ghc->len);
2349

2350 2351 2352 2353
	if (slots->generation != ghc->generation) {
		if (__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len))
			return -EFAULT;
	}
2354

2355 2356 2357
	if (kvm_is_error_hva(ghc->hva))
		return -EFAULT;

2358 2359 2360
	if (unlikely(!ghc->memslot))
		return kvm_write_guest(kvm, gpa, data, len);

2361
	r = __copy_to_user((void __user *)ghc->hva + offset, data, len);
2362 2363
	if (r)
		return -EFAULT;
2364
	mark_page_dirty_in_slot(ghc->memslot, gpa >> PAGE_SHIFT);
2365 2366 2367

	return 0;
}
2368
EXPORT_SYMBOL_GPL(kvm_write_guest_offset_cached);
2369

2370 2371
int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
2372
{
2373
	return kvm_write_guest_offset_cached(kvm, ghc, data, 0, len);
2374
}
2375
EXPORT_SYMBOL_GPL(kvm_write_guest_cached);
2376

2377 2378
int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
2379
{
2380
	struct kvm_memslots *slots = kvm_memslots(kvm);
2381 2382
	int r;

2383 2384
	BUG_ON(len > ghc->len);

2385 2386 2387 2388
	if (slots->generation != ghc->generation) {
		if (__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len))
			return -EFAULT;
	}
2389

2390 2391 2392
	if (kvm_is_error_hva(ghc->hva))
		return -EFAULT;

2393 2394 2395
	if (unlikely(!ghc->memslot))
		return kvm_read_guest(kvm, ghc->gpa, data, len);

2396 2397 2398 2399 2400 2401
	r = __copy_from_user(data, (void __user *)ghc->hva, len);
	if (r)
		return -EFAULT;

	return 0;
}
2402
EXPORT_SYMBOL_GPL(kvm_read_guest_cached);
2403

2404 2405
int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
{
2406 2407 2408
	const void *zero_page = (const void *) __va(page_to_phys(ZERO_PAGE(0)));

	return kvm_write_guest_page(kvm, gfn, zero_page, offset, len);
2409 2410 2411 2412 2413 2414 2415 2416 2417 2418
}
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;

2419
	while ((seg = next_segment(len, offset)) != 0) {
2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430
		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);

2431
static void mark_page_dirty_in_slot(struct kvm_memory_slot *memslot,
2432
				    gfn_t gfn)
A
Avi Kivity 已提交
2433
{
R
Rusty Russell 已提交
2434 2435
	if (memslot && memslot->dirty_bitmap) {
		unsigned long rel_gfn = gfn - memslot->base_gfn;
A
Avi Kivity 已提交
2436

2437
		set_bit_le(rel_gfn, memslot->dirty_bitmap);
A
Avi Kivity 已提交
2438 2439 2440
	}
}

2441 2442 2443 2444 2445
void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
{
	struct kvm_memory_slot *memslot;

	memslot = gfn_to_memslot(kvm, gfn);
2446
	mark_page_dirty_in_slot(memslot, gfn);
2447
}
2448
EXPORT_SYMBOL_GPL(mark_page_dirty);
2449

2450 2451 2452 2453 2454 2455 2456 2457 2458
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);

2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481
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
2482 2483
static void grow_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2484
	unsigned int old, val, grow, grow_start;
W
Wanpeng Li 已提交
2485

2486
	old = val = vcpu->halt_poll_ns;
2487
	grow_start = READ_ONCE(halt_poll_ns_grow_start);
2488
	grow = READ_ONCE(halt_poll_ns_grow);
2489 2490 2491
	if (!grow)
		goto out;

2492 2493 2494
	val *= grow;
	if (val < grow_start)
		val = grow_start;
W
Wanpeng Li 已提交
2495

2496 2497 2498
	if (val > halt_poll_ns)
		val = halt_poll_ns;

W
Wanpeng Li 已提交
2499
	vcpu->halt_poll_ns = val;
2500
out:
2501
	trace_kvm_halt_poll_ns_grow(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2502 2503 2504 2505
}

static void shrink_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2506
	unsigned int old, val, shrink;
W
Wanpeng Li 已提交
2507

2508
	old = val = vcpu->halt_poll_ns;
2509 2510
	shrink = READ_ONCE(halt_poll_ns_shrink);
	if (shrink == 0)
W
Wanpeng Li 已提交
2511 2512
		val = 0;
	else
2513
		val /= shrink;
W
Wanpeng Li 已提交
2514 2515

	vcpu->halt_poll_ns = val;
2516
	trace_kvm_halt_poll_ns_shrink(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2517 2518
}

2519 2520
static int kvm_vcpu_check_block(struct kvm_vcpu *vcpu)
{
2521 2522 2523
	int ret = -EINTR;
	int idx = srcu_read_lock(&vcpu->kvm->srcu);

2524 2525
	if (kvm_arch_vcpu_runnable(vcpu)) {
		kvm_make_request(KVM_REQ_UNHALT, vcpu);
2526
		goto out;
2527 2528
	}
	if (kvm_cpu_has_pending_timer(vcpu))
2529
		goto out;
2530
	if (signal_pending(current))
2531
		goto out;
2532

2533 2534 2535 2536
	ret = 0;
out:
	srcu_read_unlock(&vcpu->kvm->srcu, idx);
	return ret;
2537 2538
}

E
Eddie Dong 已提交
2539 2540 2541
/*
 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
 */
2542
void kvm_vcpu_block(struct kvm_vcpu *vcpu)
2543
{
2544
	ktime_t start, cur;
2545
	DECLARE_SWAITQUEUE(wait);
2546
	bool waited = false;
W
Wanpeng Li 已提交
2547
	u64 block_ns;
2548

2549 2550
	kvm_arch_vcpu_blocking(vcpu);

2551
	start = cur = ktime_get();
2552
	if (vcpu->halt_poll_ns && !kvm_arch_no_poll(vcpu)) {
W
Wanpeng Li 已提交
2553
		ktime_t stop = ktime_add_ns(ktime_get(), vcpu->halt_poll_ns);
2554

2555
		++vcpu->stat.halt_attempted_poll;
2556 2557 2558 2559 2560 2561 2562
		do {
			/*
			 * This sets KVM_REQ_UNHALT if an interrupt
			 * arrives.
			 */
			if (kvm_vcpu_check_block(vcpu) < 0) {
				++vcpu->stat.halt_successful_poll;
2563 2564
				if (!vcpu_valid_wakeup(vcpu))
					++vcpu->stat.halt_poll_invalid;
2565 2566 2567 2568 2569
				goto out;
			}
			cur = ktime_get();
		} while (single_task_running() && ktime_before(cur, stop));
	}
2570 2571

	for (;;) {
2572
		prepare_to_swait_exclusive(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
2573

2574
		if (kvm_vcpu_check_block(vcpu) < 0)
2575 2576
			break;

2577
		waited = true;
E
Eddie Dong 已提交
2578 2579
		schedule();
	}
2580

2581
	finish_swait(&vcpu->wq, &wait);
2582 2583
	cur = ktime_get();
out:
2584
	kvm_arch_vcpu_unblocking(vcpu);
W
Wanpeng Li 已提交
2585 2586
	block_ns = ktime_to_ns(cur) - ktime_to_ns(start);

2587 2588
	if (!kvm_arch_no_poll(vcpu)) {
		if (!vcpu_valid_wakeup(vcpu)) {
W
Wanpeng Li 已提交
2589
			shrink_halt_poll_ns(vcpu);
2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603
		} else if (halt_poll_ns) {
			if (block_ns <= vcpu->halt_poll_ns)
				;
			/* we had a long block, shrink polling */
			else if (vcpu->halt_poll_ns && block_ns > halt_poll_ns)
				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);
		} else {
			vcpu->halt_poll_ns = 0;
		}
	}
W
Wanpeng Li 已提交
2604

2605 2606
	trace_kvm_vcpu_wakeup(block_ns, waited, vcpu_valid_wakeup(vcpu));
	kvm_arch_vcpu_block_finish(vcpu);
E
Eddie Dong 已提交
2607
}
2608
EXPORT_SYMBOL_GPL(kvm_vcpu_block);
E
Eddie Dong 已提交
2609

2610
bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu)
2611
{
2612
	struct swait_queue_head *wqp;
2613 2614

	wqp = kvm_arch_vcpu_wq(vcpu);
2615
	if (swq_has_sleeper(wqp)) {
2616
		swake_up_one(wqp);
2617
		WRITE_ONCE(vcpu->ready, true);
2618
		++vcpu->stat.halt_wakeup;
2619
		return true;
2620 2621
	}

2622
	return false;
2623 2624 2625
}
EXPORT_SYMBOL_GPL(kvm_vcpu_wake_up);

2626
#ifndef CONFIG_S390
2627 2628 2629 2630 2631 2632 2633 2634
/*
 * 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;

2635 2636 2637
	if (kvm_vcpu_wake_up(vcpu))
		return;

2638 2639 2640 2641 2642 2643
	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();
}
2644
EXPORT_SYMBOL_GPL(kvm_vcpu_kick);
2645
#endif /* !CONFIG_S390 */
2646

2647
int kvm_vcpu_yield_to(struct kvm_vcpu *target)
2648 2649 2650
{
	struct pid *pid;
	struct task_struct *task = NULL;
2651
	int ret = 0;
2652 2653 2654 2655

	rcu_read_lock();
	pid = rcu_dereference(target->pid);
	if (pid)
2656
		task = get_pid_task(pid, PIDTYPE_PID);
2657 2658
	rcu_read_unlock();
	if (!task)
2659 2660
		return ret;
	ret = yield_to(task, 1);
2661
	put_task_struct(task);
2662 2663

	return ret;
2664 2665 2666
}
EXPORT_SYMBOL_GPL(kvm_vcpu_yield_to);

2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688
/*
 * 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.
 */
2689
static bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
2690
{
2691
#ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
2692 2693 2694
	bool eligible;

	eligible = !vcpu->spin_loop.in_spin_loop ||
2695
		    vcpu->spin_loop.dy_eligible;
2696 2697 2698 2699 2700

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

	return eligible;
2701 2702
#else
	return true;
2703
#endif
2704
}
2705

2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728
/*
 * Unlike kvm_arch_vcpu_runnable, this function is called outside
 * a vcpu_load/vcpu_put pair.  However, for most architectures
 * kvm_arch_vcpu_runnable does not require vcpu_load.
 */
bool __weak kvm_arch_dy_runnable(struct kvm_vcpu *vcpu)
{
	return kvm_arch_vcpu_runnable(vcpu);
}

static bool vcpu_dy_runnable(struct kvm_vcpu *vcpu)
{
	if (kvm_arch_dy_runnable(vcpu))
		return true;

#ifdef CONFIG_KVM_ASYNC_PF
	if (!list_empty_careful(&vcpu->async_pf.done))
		return true;
#endif

	return false;
}

2729
void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
Z
Zhai, Edwin 已提交
2730
{
2731 2732 2733 2734
	struct kvm *kvm = me->kvm;
	struct kvm_vcpu *vcpu;
	int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
	int yielded = 0;
2735
	int try = 3;
2736 2737
	int pass;
	int i;
Z
Zhai, Edwin 已提交
2738

2739
	kvm_vcpu_set_in_spin_loop(me, true);
2740 2741 2742 2743 2744 2745 2746
	/*
	 * 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.
	 */
2747
	for (pass = 0; pass < 2 && !yielded && try; pass++) {
2748
		kvm_for_each_vcpu(i, vcpu, kvm) {
2749
			if (!pass && i <= last_boosted_vcpu) {
2750 2751 2752 2753
				i = last_boosted_vcpu;
				continue;
			} else if (pass && i > last_boosted_vcpu)
				break;
2754
			if (!READ_ONCE(vcpu->ready))
2755
				continue;
2756 2757
			if (vcpu == me)
				continue;
2758
			if (swait_active(&vcpu->wq) && !vcpu_dy_runnable(vcpu))
2759
				continue;
2760 2761
			if (READ_ONCE(vcpu->preempted) && yield_to_kernel_mode &&
				!kvm_arch_vcpu_in_kernel(vcpu))
2762
				continue;
2763 2764
			if (!kvm_vcpu_eligible_for_directed_yield(vcpu))
				continue;
2765 2766 2767

			yielded = kvm_vcpu_yield_to(vcpu);
			if (yielded > 0) {
2768 2769
				kvm->last_boosted_vcpu = i;
				break;
2770 2771 2772 2773
			} else if (yielded < 0) {
				try--;
				if (!try)
					break;
2774 2775 2776
			}
		}
	}
2777
	kvm_vcpu_set_in_spin_loop(me, false);
2778 2779 2780

	/* Ensure vcpu is not eligible during next spinloop */
	kvm_vcpu_set_dy_eligible(me, false);
Z
Zhai, Edwin 已提交
2781 2782 2783
}
EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin);

2784
static vm_fault_t kvm_vcpu_fault(struct vm_fault *vmf)
2785
{
2786
	struct kvm_vcpu *vcpu = vmf->vma->vm_file->private_data;
2787 2788
	struct page *page;

2789
	if (vmf->pgoff == 0)
2790
		page = virt_to_page(vcpu->run);
A
Avi Kivity 已提交
2791
#ifdef CONFIG_X86
2792
	else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
2793
		page = virt_to_page(vcpu->arch.pio_data);
2794
#endif
2795
#ifdef CONFIG_KVM_MMIO
2796 2797
	else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
		page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
A
Avi Kivity 已提交
2798
#endif
2799
	else
2800
		return kvm_arch_vcpu_fault(vcpu, vmf);
2801
	get_page(page);
2802 2803
	vmf->page = page;
	return 0;
2804 2805
}

2806
static const struct vm_operations_struct kvm_vcpu_vm_ops = {
2807
	.fault = kvm_vcpu_fault,
2808 2809 2810 2811 2812 2813 2814 2815
};

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
2816 2817 2818 2819
static int kvm_vcpu_release(struct inode *inode, struct file *filp)
{
	struct kvm_vcpu *vcpu = filp->private_data;

2820
	debugfs_remove_recursive(vcpu->debugfs_dentry);
A
Al Viro 已提交
2821
	kvm_put_kvm(vcpu->kvm);
A
Avi Kivity 已提交
2822 2823 2824
	return 0;
}

2825
static struct file_operations kvm_vcpu_fops = {
A
Avi Kivity 已提交
2826 2827
	.release        = kvm_vcpu_release,
	.unlocked_ioctl = kvm_vcpu_ioctl,
2828
	.mmap           = kvm_vcpu_mmap,
2829
	.llseek		= noop_llseek,
2830
	KVM_COMPAT(kvm_vcpu_compat_ioctl),
A
Avi Kivity 已提交
2831 2832 2833 2834 2835 2836 2837
};

/*
 * Allocates an inode for the vcpu.
 */
static int create_vcpu_fd(struct kvm_vcpu *vcpu)
{
2838 2839 2840 2841
	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
2842 2843
}

2844
static void kvm_create_vcpu_debugfs(struct kvm_vcpu *vcpu)
2845
{
2846
#ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
2847 2848 2849
	char dir_name[ITOA_MAX_LEN * 2];

	if (!debugfs_initialized())
2850
		return;
2851 2852 2853

	snprintf(dir_name, sizeof(dir_name), "vcpu%d", vcpu->vcpu_id);
	vcpu->debugfs_dentry = debugfs_create_dir(dir_name,
2854
						  vcpu->kvm->debugfs_dentry);
2855

2856
	kvm_arch_create_vcpu_debugfs(vcpu);
2857
#endif
2858 2859
}

2860 2861 2862
/*
 * Creates some virtual cpus.  Good luck creating more than one.
 */
2863
static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
2864 2865
{
	int r;
2866
	struct kvm_vcpu *vcpu;
2867
	struct page *page;
2868

G
Greg Kurz 已提交
2869
	if (id >= KVM_MAX_VCPU_ID)
2870 2871
		return -EINVAL;

2872 2873 2874 2875 2876 2877 2878 2879 2880
	mutex_lock(&kvm->lock);
	if (kvm->created_vcpus == KVM_MAX_VCPUS) {
		mutex_unlock(&kvm->lock);
		return -EINVAL;
	}

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

2881 2882 2883 2884
	r = kvm_arch_vcpu_precreate(kvm, id);
	if (r)
		goto vcpu_decrement;

2885 2886 2887
	vcpu = kmem_cache_zalloc(kvm_vcpu_cache, GFP_KERNEL);
	if (!vcpu) {
		r = -ENOMEM;
2888 2889
		goto vcpu_decrement;
	}
2890

2891
	BUILD_BUG_ON(sizeof(struct kvm_run) > PAGE_SIZE);
2892 2893 2894
	page = alloc_page(GFP_KERNEL | __GFP_ZERO);
	if (!page) {
		r = -ENOMEM;
2895
		goto vcpu_free;
2896 2897 2898 2899
	}
	vcpu->run = page_address(page);

	kvm_vcpu_init(vcpu, kvm, id);
2900 2901 2902

	r = kvm_arch_vcpu_create(vcpu);
	if (r)
2903
		goto vcpu_free_run_page;
2904

2905
	kvm_create_vcpu_debugfs(vcpu);
2906

S
Shaohua Li 已提交
2907
	mutex_lock(&kvm->lock);
2908 2909 2910 2911
	if (kvm_get_vcpu_by_id(kvm, id)) {
		r = -EEXIST;
		goto unlock_vcpu_destroy;
	}
2912

2913 2914
	vcpu->vcpu_idx = atomic_read(&kvm->online_vcpus);
	BUG_ON(kvm->vcpus[vcpu->vcpu_idx]);
2915

R
Rusty Russell 已提交
2916
	/* Now it's all set up, let userspace reach it */
A
Al Viro 已提交
2917
	kvm_get_kvm(kvm);
A
Avi Kivity 已提交
2918
	r = create_vcpu_fd(vcpu);
2919
	if (r < 0) {
2920
		kvm_put_kvm_no_destroy(kvm);
2921
		goto unlock_vcpu_destroy;
2922 2923
	}

2924
	kvm->vcpus[vcpu->vcpu_idx] = vcpu;
2925 2926 2927 2928 2929

	/*
	 * Pairs with smp_rmb() in kvm_get_vcpu.  Write kvm->vcpus
	 * before kvm->online_vcpu's incremented value.
	 */
2930 2931 2932 2933
	smp_wmb();
	atomic_inc(&kvm->online_vcpus);

	mutex_unlock(&kvm->lock);
2934
	kvm_arch_vcpu_postcreate(vcpu);
R
Rusty Russell 已提交
2935
	return r;
2936

2937
unlock_vcpu_destroy:
2938
	mutex_unlock(&kvm->lock);
2939
	debugfs_remove_recursive(vcpu->debugfs_dentry);
2940
	kvm_arch_vcpu_destroy(vcpu);
2941 2942
vcpu_free_run_page:
	free_page((unsigned long)vcpu->run);
2943 2944
vcpu_free:
	kmem_cache_free(kvm_vcpu_cache, vcpu);
2945 2946 2947 2948
vcpu_decrement:
	mutex_lock(&kvm->lock);
	kvm->created_vcpus--;
	mutex_unlock(&kvm->lock);
2949 2950 2951
	return r;
}

A
Avi Kivity 已提交
2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962
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
2963 2964
static long kvm_vcpu_ioctl(struct file *filp,
			   unsigned int ioctl, unsigned long arg)
A
Avi Kivity 已提交
2965
{
A
Avi Kivity 已提交
2966
	struct kvm_vcpu *vcpu = filp->private_data;
A
Al Viro 已提交
2967
	void __user *argp = (void __user *)arg;
2968
	int r;
2969 2970
	struct kvm_fpu *fpu = NULL;
	struct kvm_sregs *kvm_sregs = NULL;
A
Avi Kivity 已提交
2971

2972 2973
	if (vcpu->kvm->mm != current->mm)
		return -EIO;
2974

2975 2976 2977
	if (unlikely(_IOC_TYPE(ioctl) != KVMIO))
		return -EINVAL;

2978
	/*
2979 2980
	 * Some architectures have vcpu ioctls that are asynchronous to vcpu
	 * execution; mutex_lock() would break them.
2981
	 */
2982 2983
	r = kvm_arch_vcpu_async_ioctl(filp, ioctl, arg);
	if (r != -ENOIOCTLCMD)
2984
		return r;
2985

2986 2987
	if (mutex_lock_killable(&vcpu->mutex))
		return -EINTR;
A
Avi Kivity 已提交
2988
	switch (ioctl) {
2989 2990
	case KVM_RUN: {
		struct pid *oldpid;
2991 2992 2993
		r = -EINVAL;
		if (arg)
			goto out;
2994
		oldpid = rcu_access_pointer(vcpu->pid);
2995
		if (unlikely(oldpid != task_pid(current))) {
2996
			/* The thread running this VCPU changed. */
2997
			struct pid *newpid;
2998

2999 3000 3001 3002 3003
			r = kvm_arch_vcpu_run_pid_change(vcpu);
			if (r)
				break;

			newpid = get_task_pid(current, PIDTYPE_PID);
3004 3005 3006 3007 3008
			rcu_assign_pointer(vcpu->pid, newpid);
			if (oldpid)
				synchronize_rcu();
			put_pid(oldpid);
		}
3009
		r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
3010
		trace_kvm_userspace_exit(vcpu->run->exit_reason, r);
A
Avi Kivity 已提交
3011
		break;
3012
	}
A
Avi Kivity 已提交
3013
	case KVM_GET_REGS: {
3014
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
3015

3016
		r = -ENOMEM;
3017
		kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL_ACCOUNT);
3018
		if (!kvm_regs)
A
Avi Kivity 已提交
3019
			goto out;
3020 3021 3022
		r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
		if (r)
			goto out_free1;
A
Avi Kivity 已提交
3023
		r = -EFAULT;
3024 3025
		if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
			goto out_free1;
A
Avi Kivity 已提交
3026
		r = 0;
3027 3028
out_free1:
		kfree(kvm_regs);
A
Avi Kivity 已提交
3029 3030 3031
		break;
	}
	case KVM_SET_REGS: {
3032
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
3033

3034
		r = -ENOMEM;
3035 3036 3037
		kvm_regs = memdup_user(argp, sizeof(*kvm_regs));
		if (IS_ERR(kvm_regs)) {
			r = PTR_ERR(kvm_regs);
A
Avi Kivity 已提交
3038
			goto out;
3039
		}
3040 3041
		r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
		kfree(kvm_regs);
A
Avi Kivity 已提交
3042 3043 3044
		break;
	}
	case KVM_GET_SREGS: {
3045 3046
		kvm_sregs = kzalloc(sizeof(struct kvm_sregs),
				    GFP_KERNEL_ACCOUNT);
3047 3048 3049 3050
		r = -ENOMEM;
		if (!kvm_sregs)
			goto out;
		r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
3051 3052 3053
		if (r)
			goto out;
		r = -EFAULT;
3054
		if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
A
Avi Kivity 已提交
3055 3056 3057 3058 3059
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_SREGS: {
3060 3061 3062
		kvm_sregs = memdup_user(argp, sizeof(*kvm_sregs));
		if (IS_ERR(kvm_sregs)) {
			r = PTR_ERR(kvm_sregs);
G
Guo Chao 已提交
3063
			kvm_sregs = NULL;
A
Avi Kivity 已提交
3064
			goto out;
3065
		}
3066
		r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
3067 3068
		break;
	}
3069 3070 3071 3072 3073 3074 3075
	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;
3076
		if (copy_to_user(argp, &mp_state, sizeof(mp_state)))
3077 3078 3079 3080 3081 3082 3083 3084
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_MP_STATE: {
		struct kvm_mp_state mp_state;

		r = -EFAULT;
3085
		if (copy_from_user(&mp_state, argp, sizeof(mp_state)))
3086 3087 3088 3089
			goto out;
		r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
		break;
	}
A
Avi Kivity 已提交
3090 3091 3092 3093
	case KVM_TRANSLATE: {
		struct kvm_translation tr;

		r = -EFAULT;
3094
		if (copy_from_user(&tr, argp, sizeof(tr)))
A
Avi Kivity 已提交
3095
			goto out;
3096
		r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
A
Avi Kivity 已提交
3097 3098 3099
		if (r)
			goto out;
		r = -EFAULT;
3100
		if (copy_to_user(argp, &tr, sizeof(tr)))
A
Avi Kivity 已提交
3101 3102 3103 3104
			goto out;
		r = 0;
		break;
	}
J
Jan Kiszka 已提交
3105 3106
	case KVM_SET_GUEST_DEBUG: {
		struct kvm_guest_debug dbg;
A
Avi Kivity 已提交
3107 3108

		r = -EFAULT;
3109
		if (copy_from_user(&dbg, argp, sizeof(dbg)))
A
Avi Kivity 已提交
3110
			goto out;
J
Jan Kiszka 已提交
3111
		r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
A
Avi Kivity 已提交
3112 3113
		break;
	}
A
Avi Kivity 已提交
3114 3115 3116 3117 3118 3119 3120 3121 3122
	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,
3123
					   sizeof(kvm_sigmask)))
A
Avi Kivity 已提交
3124 3125
				goto out;
			r = -EINVAL;
3126
			if (kvm_sigmask.len != sizeof(sigset))
A
Avi Kivity 已提交
3127 3128 3129
				goto out;
			r = -EFAULT;
			if (copy_from_user(&sigset, sigmask_arg->sigset,
3130
					   sizeof(sigset)))
A
Avi Kivity 已提交
3131 3132 3133
				goto out;
			p = &sigset;
		}
3134
		r = kvm_vcpu_ioctl_set_sigmask(vcpu, p);
A
Avi Kivity 已提交
3135 3136
		break;
	}
A
Avi Kivity 已提交
3137
	case KVM_GET_FPU: {
3138
		fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL_ACCOUNT);
3139 3140 3141 3142
		r = -ENOMEM;
		if (!fpu)
			goto out;
		r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu);
A
Avi Kivity 已提交
3143 3144 3145
		if (r)
			goto out;
		r = -EFAULT;
3146
		if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
A
Avi Kivity 已提交
3147 3148 3149 3150 3151
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_FPU: {
3152 3153 3154
		fpu = memdup_user(argp, sizeof(*fpu));
		if (IS_ERR(fpu)) {
			r = PTR_ERR(fpu);
G
Guo Chao 已提交
3155
			fpu = NULL;
A
Avi Kivity 已提交
3156
			goto out;
3157
		}
3158
		r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
A
Avi Kivity 已提交
3159 3160
		break;
	}
A
Avi Kivity 已提交
3161
	default:
3162
		r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
3163 3164
	}
out:
3165
	mutex_unlock(&vcpu->mutex);
3166 3167
	kfree(fpu);
	kfree(kvm_sregs);
A
Avi Kivity 已提交
3168 3169 3170
	return r;
}

3171
#ifdef CONFIG_KVM_COMPAT
3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190
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,
3191
					   sizeof(kvm_sigmask)))
3192 3193
				goto out;
			r = -EINVAL;
A
Al Viro 已提交
3194
			if (kvm_sigmask.len != sizeof(compat_sigset_t))
3195 3196
				goto out;
			r = -EFAULT;
A
Al Viro 已提交
3197
			if (get_compat_sigset(&sigset, (void *)sigmask_arg->sigset))
3198
				goto out;
3199 3200 3201
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
		} else
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, NULL);
3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212
		break;
	}
	default:
		r = kvm_vcpu_ioctl(filp, ioctl, arg);
	}

out:
	return r;
}
#endif

3213 3214 3215 3216 3217 3218 3219 3220 3221 3222
static int kvm_device_mmap(struct file *filp, struct vm_area_struct *vma)
{
	struct kvm_device *dev = filp->private_data;

	if (dev->ops->mmap)
		return dev->ops->mmap(dev, vma);

	return -ENODEV;
}

S
Scott Wood 已提交
3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243
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;

3244 3245 3246
	if (dev->kvm->mm != current->mm)
		return -EIO;

S
Scott Wood 已提交
3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266
	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;

3267 3268 3269 3270 3271 3272 3273
	if (dev->ops->release) {
		mutex_lock(&kvm->lock);
		list_del(&dev->vm_node);
		dev->ops->release(dev);
		mutex_unlock(&kvm->lock);
	}

S
Scott Wood 已提交
3274 3275 3276 3277 3278 3279 3280
	kvm_put_kvm(kvm);
	return 0;
}

static const struct file_operations kvm_device_fops = {
	.unlocked_ioctl = kvm_device_ioctl,
	.release = kvm_device_release,
3281
	KVM_COMPAT(kvm_device_ioctl),
3282
	.mmap = kvm_device_mmap,
S
Scott Wood 已提交
3283 3284 3285 3286 3287 3288 3289 3290 3291 3292
};

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

	return filp->private_data;
}

3293
static const struct kvm_device_ops *kvm_device_ops_table[KVM_DEV_TYPE_MAX] = {
3294
#ifdef CONFIG_KVM_MPIC
3295 3296
	[KVM_DEV_TYPE_FSL_MPIC_20]	= &kvm_mpic_ops,
	[KVM_DEV_TYPE_FSL_MPIC_42]	= &kvm_mpic_ops,
3297
#endif
3298 3299
};

3300
int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type)
3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311
{
	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;
}

3312 3313 3314 3315 3316 3317
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
3318 3319 3320
static int kvm_ioctl_create_device(struct kvm *kvm,
				   struct kvm_create_device *cd)
{
3321
	const struct kvm_device_ops *ops = NULL;
S
Scott Wood 已提交
3322 3323
	struct kvm_device *dev;
	bool test = cd->flags & KVM_CREATE_DEVICE_TEST;
P
Paolo Bonzini 已提交
3324
	int type;
S
Scott Wood 已提交
3325 3326
	int ret;

3327 3328 3329
	if (cd->type >= ARRAY_SIZE(kvm_device_ops_table))
		return -ENODEV;

P
Paolo Bonzini 已提交
3330 3331
	type = array_index_nospec(cd->type, ARRAY_SIZE(kvm_device_ops_table));
	ops = kvm_device_ops_table[type];
3332
	if (ops == NULL)
S
Scott Wood 已提交
3333 3334 3335 3336 3337
		return -ENODEV;

	if (test)
		return 0;

3338
	dev = kzalloc(sizeof(*dev), GFP_KERNEL_ACCOUNT);
S
Scott Wood 已提交
3339 3340 3341 3342 3343 3344
	if (!dev)
		return -ENOMEM;

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

3345
	mutex_lock(&kvm->lock);
P
Paolo Bonzini 已提交
3346
	ret = ops->create(dev, type);
S
Scott Wood 已提交
3347
	if (ret < 0) {
3348
		mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
3349 3350 3351
		kfree(dev);
		return ret;
	}
3352 3353
	list_add(&dev->vm_node, &kvm->devices);
	mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
3354

3355 3356 3357
	if (ops->init)
		ops->init(dev);

3358
	kvm_get_kvm(kvm);
3359
	ret = anon_inode_getfd(ops->name, &kvm_device_fops, dev, O_RDWR | O_CLOEXEC);
S
Scott Wood 已提交
3360
	if (ret < 0) {
3361
		kvm_put_kvm_no_destroy(kvm);
3362 3363 3364
		mutex_lock(&kvm->lock);
		list_del(&dev->vm_node);
		mutex_unlock(&kvm->lock);
3365
		ops->destroy(dev);
S
Scott Wood 已提交
3366 3367 3368 3369 3370 3371 3372
		return ret;
	}

	cd->fd = ret;
	return 0;
}

3373 3374 3375 3376 3377 3378 3379 3380 3381 3382
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
3383
#ifdef CONFIG_HAVE_KVM_IRQFD
3384
	case KVM_CAP_IRQFD:
3385 3386
	case KVM_CAP_IRQFD_RESAMPLE:
#endif
3387
	case KVM_CAP_IOEVENTFD_ANY_LENGTH:
3388
	case KVM_CAP_CHECK_EXTENSION_VM:
3389
	case KVM_CAP_ENABLE_CAP_VM:
3390
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
3391
	case KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2:
3392
#endif
3393
		return 1;
3394
#ifdef CONFIG_KVM_MMIO
3395 3396
	case KVM_CAP_COALESCED_MMIO:
		return KVM_COALESCED_MMIO_PAGE_OFFSET;
P
Peng Hao 已提交
3397 3398
	case KVM_CAP_COALESCED_PIO:
		return 1;
3399
#endif
3400 3401 3402
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_CAP_IRQ_ROUTING:
		return KVM_MAX_IRQ_ROUTES;
3403 3404 3405 3406
#endif
#if KVM_ADDRESS_SPACE_NUM > 1
	case KVM_CAP_MULTI_ADDRESS_SPACE:
		return KVM_ADDRESS_SPACE_NUM;
3407
#endif
3408 3409
	case KVM_CAP_NR_MEMSLOTS:
		return KVM_USER_MEM_SLOTS;
3410 3411 3412 3413 3414 3415
	default:
		break;
	}
	return kvm_vm_ioctl_check_extension(kvm, arg);
}

3416 3417 3418 3419 3420 3421 3422 3423 3424 3425
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) {
3426
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
3427
	case KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2:
3428 3429 3430 3431 3432
		if (cap->flags || (cap->args[0] & ~1))
			return -EINVAL;
		kvm->manual_dirty_log_protect = cap->args[0];
		return 0;
#endif
3433 3434 3435 3436 3437
	default:
		return kvm_vm_ioctl_enable_cap(kvm, cap);
	}
}

A
Avi Kivity 已提交
3438 3439 3440 3441 3442
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;
3443
	int r;
A
Avi Kivity 已提交
3444

3445 3446
	if (kvm->mm != current->mm)
		return -EIO;
A
Avi Kivity 已提交
3447 3448 3449 3450
	switch (ioctl) {
	case KVM_CREATE_VCPU:
		r = kvm_vm_ioctl_create_vcpu(kvm, arg);
		break;
3451 3452 3453 3454 3455 3456 3457 3458 3459
	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;
	}
3460 3461 3462 3463 3464
	case KVM_SET_USER_MEMORY_REGION: {
		struct kvm_userspace_memory_region kvm_userspace_mem;

		r = -EFAULT;
		if (copy_from_user(&kvm_userspace_mem, argp,
3465
						sizeof(kvm_userspace_mem)))
3466 3467
			goto out;

3468
		r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem);
A
Avi Kivity 已提交
3469 3470 3471 3472 3473 3474
		break;
	}
	case KVM_GET_DIRTY_LOG: {
		struct kvm_dirty_log log;

		r = -EFAULT;
3475
		if (copy_from_user(&log, argp, sizeof(log)))
A
Avi Kivity 已提交
3476
			goto out;
3477
		r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
A
Avi Kivity 已提交
3478 3479
		break;
	}
3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490
#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
3491
#ifdef CONFIG_KVM_MMIO
3492 3493
	case KVM_REGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3494

3495
		r = -EFAULT;
3496
		if (copy_from_user(&zone, argp, sizeof(zone)))
3497 3498 3499 3500 3501 3502
			goto out;
		r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
		break;
	}
	case KVM_UNREGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3503

3504
		r = -EFAULT;
3505
		if (copy_from_user(&zone, argp, sizeof(zone)))
3506 3507 3508 3509 3510
			goto out;
		r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
		break;
	}
#endif
G
Gregory Haskins 已提交
3511 3512 3513 3514
	case KVM_IRQFD: {
		struct kvm_irqfd data;

		r = -EFAULT;
3515
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3516
			goto out;
3517
		r = kvm_irqfd(kvm, &data);
G
Gregory Haskins 已提交
3518 3519
		break;
	}
G
Gregory Haskins 已提交
3520 3521 3522 3523
	case KVM_IOEVENTFD: {
		struct kvm_ioeventfd data;

		r = -EFAULT;
3524
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3525 3526 3527 3528
			goto out;
		r = kvm_ioeventfd(kvm, &data);
		break;
	}
3529 3530 3531 3532 3533
#ifdef CONFIG_HAVE_KVM_MSI
	case KVM_SIGNAL_MSI: {
		struct kvm_msi msi;

		r = -EFAULT;
3534
		if (copy_from_user(&msi, argp, sizeof(msi)))
3535 3536 3537 3538
			goto out;
		r = kvm_send_userspace_msi(kvm, &msi);
		break;
	}
3539 3540 3541 3542 3543 3544 3545
#endif
#ifdef __KVM_HAVE_IRQ_LINE
	case KVM_IRQ_LINE_STATUS:
	case KVM_IRQ_LINE: {
		struct kvm_irq_level irq_event;

		r = -EFAULT;
3546
		if (copy_from_user(&irq_event, argp, sizeof(irq_event)))
3547 3548
			goto out;

3549 3550
		r = kvm_vm_ioctl_irq_line(kvm, &irq_event,
					ioctl == KVM_IRQ_LINE_STATUS);
3551 3552 3553 3554 3555
		if (r)
			goto out;

		r = -EFAULT;
		if (ioctl == KVM_IRQ_LINE_STATUS) {
3556
			if (copy_to_user(argp, &irq_event, sizeof(irq_event)))
3557 3558 3559 3560 3561 3562
				goto out;
		}

		r = 0;
		break;
	}
3563
#endif
3564 3565 3566 3567
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_SET_GSI_ROUTING: {
		struct kvm_irq_routing routing;
		struct kvm_irq_routing __user *urouting;
3568
		struct kvm_irq_routing_entry *entries = NULL;
3569 3570 3571 3572 3573

		r = -EFAULT;
		if (copy_from_user(&routing, argp, sizeof(routing)))
			goto out;
		r = -EINVAL;
3574 3575
		if (!kvm_arch_can_set_irq_routing(kvm))
			goto out;
3576
		if (routing.nr > KVM_MAX_IRQ_ROUTES)
3577 3578 3579
			goto out;
		if (routing.flags)
			goto out;
3580 3581
		if (routing.nr) {
			r = -ENOMEM;
3582 3583
			entries = vmalloc(array_size(sizeof(*entries),
						     routing.nr));
3584 3585 3586 3587 3588 3589 3590 3591
			if (!entries)
				goto out;
			r = -EFAULT;
			urouting = argp;
			if (copy_from_user(entries, urouting->entries,
					   routing.nr * sizeof(*entries)))
				goto out_free_irq_routing;
		}
3592 3593
		r = kvm_set_irq_routing(kvm, entries, routing.nr,
					routing.flags);
3594
out_free_irq_routing:
3595 3596 3597 3598
		vfree(entries);
		break;
	}
#endif /* CONFIG_HAVE_KVM_IRQ_ROUTING */
S
Scott Wood 已提交
3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616
	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;
	}
3617 3618 3619
	case KVM_CHECK_EXTENSION:
		r = kvm_vm_ioctl_check_extension_generic(kvm, arg);
		break;
3620
	default:
3621
		r = kvm_arch_vm_ioctl(filp, ioctl, arg);
3622 3623 3624 3625 3626
	}
out:
	return r;
}

3627
#ifdef CONFIG_KVM_COMPAT
3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651
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)))
3652
			return -EFAULT;
3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667
		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

3668
static struct file_operations kvm_vm_fops = {
3669 3670
	.release        = kvm_vm_release,
	.unlocked_ioctl = kvm_vm_ioctl,
3671
	.llseek		= noop_llseek,
3672
	KVM_COMPAT(kvm_vm_compat_ioctl),
3673 3674
};

3675
static int kvm_dev_ioctl_create_vm(unsigned long type)
3676
{
3677
	int r;
3678
	struct kvm *kvm;
3679
	struct file *file;
3680

3681
	kvm = kvm_create_vm(type);
3682 3683
	if (IS_ERR(kvm))
		return PTR_ERR(kvm);
3684
#ifdef CONFIG_KVM_MMIO
3685
	r = kvm_coalesced_mmio_init(kvm);
3686 3687
	if (r < 0)
		goto put_kvm;
3688
#endif
3689
	r = get_unused_fd_flags(O_CLOEXEC);
3690 3691 3692
	if (r < 0)
		goto put_kvm;

3693 3694 3695
	file = anon_inode_getfile("kvm-vm", &kvm_vm_fops, kvm, O_RDWR);
	if (IS_ERR(file)) {
		put_unused_fd(r);
3696 3697
		r = PTR_ERR(file);
		goto put_kvm;
3698
	}
3699

3700 3701 3702 3703 3704 3705
	/*
	 * 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).
	 */
3706
	if (kvm_create_vm_debugfs(kvm, r) < 0) {
3707 3708
		put_unused_fd(r);
		fput(file);
3709 3710
		return -ENOMEM;
	}
3711
	kvm_uevent_notify_change(KVM_EVENT_CREATE_VM, kvm);
3712

3713
	fd_install(r, file);
3714
	return r;
3715 3716 3717 3718

put_kvm:
	kvm_put_kvm(kvm);
	return r;
3719 3720 3721 3722 3723
}

static long kvm_dev_ioctl(struct file *filp,
			  unsigned int ioctl, unsigned long arg)
{
3724
	long r = -EINVAL;
3725 3726 3727

	switch (ioctl) {
	case KVM_GET_API_VERSION:
3728 3729
		if (arg)
			goto out;
3730 3731 3732
		r = KVM_API_VERSION;
		break;
	case KVM_CREATE_VM:
3733
		r = kvm_dev_ioctl_create_vm(arg);
3734
		break;
3735
	case KVM_CHECK_EXTENSION:
3736
		r = kvm_vm_ioctl_check_extension_generic(NULL, arg);
3737
		break;
3738 3739 3740
	case KVM_GET_VCPU_MMAP_SIZE:
		if (arg)
			goto out;
3741 3742 3743
		r = PAGE_SIZE;     /* struct kvm_run */
#ifdef CONFIG_X86
		r += PAGE_SIZE;    /* pio data page */
3744
#endif
3745
#ifdef CONFIG_KVM_MMIO
3746
		r += PAGE_SIZE;    /* coalesced mmio ring page */
3747
#endif
3748
		break;
3749 3750 3751
	case KVM_TRACE_ENABLE:
	case KVM_TRACE_PAUSE:
	case KVM_TRACE_DISABLE:
3752
		r = -EOPNOTSUPP;
3753
		break;
A
Avi Kivity 已提交
3754
	default:
3755
		return kvm_arch_dev_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
3756 3757 3758 3759 3760 3761 3762
	}
out:
	return r;
}

static struct file_operations kvm_chardev_ops = {
	.unlocked_ioctl = kvm_dev_ioctl,
3763
	.llseek		= noop_llseek,
3764
	KVM_COMPAT(kvm_dev_ioctl),
A
Avi Kivity 已提交
3765 3766 3767
};

static struct miscdevice kvm_dev = {
A
Avi Kivity 已提交
3768
	KVM_MINOR,
A
Avi Kivity 已提交
3769 3770 3771 3772
	"kvm",
	&kvm_chardev_ops,
};

3773
static void hardware_enable_nolock(void *junk)
3774 3775
{
	int cpu = raw_smp_processor_id();
3776
	int r;
3777

3778
	if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
3779
		return;
3780

3781
	cpumask_set_cpu(cpu, cpus_hardware_enabled);
3782

3783
	r = kvm_arch_hardware_enable();
3784 3785 3786 3787

	if (r) {
		cpumask_clear_cpu(cpu, cpus_hardware_enabled);
		atomic_inc(&hardware_enable_failed);
X
Xiubo Li 已提交
3788
		pr_info("kvm: enabling virtualization on CPU%d failed\n", cpu);
3789
	}
3790 3791
}

3792
static int kvm_starting_cpu(unsigned int cpu)
3793
{
3794
	raw_spin_lock(&kvm_count_lock);
3795 3796
	if (kvm_usage_count)
		hardware_enable_nolock(NULL);
3797
	raw_spin_unlock(&kvm_count_lock);
3798
	return 0;
3799 3800 3801
}

static void hardware_disable_nolock(void *junk)
3802 3803 3804
{
	int cpu = raw_smp_processor_id();

3805
	if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
3806
		return;
3807
	cpumask_clear_cpu(cpu, cpus_hardware_enabled);
3808
	kvm_arch_hardware_disable();
3809 3810
}

3811
static int kvm_dying_cpu(unsigned int cpu)
3812
{
3813
	raw_spin_lock(&kvm_count_lock);
3814 3815
	if (kvm_usage_count)
		hardware_disable_nolock(NULL);
3816
	raw_spin_unlock(&kvm_count_lock);
3817
	return 0;
3818 3819
}

3820 3821 3822 3823 3824 3825
static void hardware_disable_all_nolock(void)
{
	BUG_ON(!kvm_usage_count);

	kvm_usage_count--;
	if (!kvm_usage_count)
3826
		on_each_cpu(hardware_disable_nolock, NULL, 1);
3827 3828 3829 3830
}

static void hardware_disable_all(void)
{
3831
	raw_spin_lock(&kvm_count_lock);
3832
	hardware_disable_all_nolock();
3833
	raw_spin_unlock(&kvm_count_lock);
3834 3835 3836 3837 3838 3839
}

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

3840
	raw_spin_lock(&kvm_count_lock);
3841 3842 3843 3844

	kvm_usage_count++;
	if (kvm_usage_count == 1) {
		atomic_set(&hardware_enable_failed, 0);
3845
		on_each_cpu(hardware_enable_nolock, NULL, 1);
3846 3847 3848 3849 3850 3851 3852

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

3853
	raw_spin_unlock(&kvm_count_lock);
3854 3855 3856 3857

	return r;
}

3858
static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
M
Mike Day 已提交
3859
		      void *v)
3860
{
3861 3862 3863 3864 3865 3866
	/*
	 * 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
3867
	pr_info("kvm: exiting hardware virtualization\n");
3868
	kvm_rebooting = true;
3869
	on_each_cpu(hardware_disable_nolock, NULL, 1);
3870 3871 3872 3873 3874 3875 3876 3877
	return NOTIFY_OK;
}

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

M
Marcelo Tosatti 已提交
3878
static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
3879 3880 3881 3882
{
	int i;

	for (i = 0; i < bus->dev_count; i++) {
3883
		struct kvm_io_device *pos = bus->range[i].dev;
3884 3885 3886

		kvm_iodevice_destructor(pos);
	}
M
Marcelo Tosatti 已提交
3887
	kfree(bus);
3888 3889
}

3890
static inline int kvm_io_bus_cmp(const struct kvm_io_range *r1,
X
Xiubo Li 已提交
3891
				 const struct kvm_io_range *r2)
3892
{
J
Jason Wang 已提交
3893 3894 3895 3896
	gpa_t addr1 = r1->addr;
	gpa_t addr2 = r2->addr;

	if (addr1 < addr2)
3897
		return -1;
J
Jason Wang 已提交
3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909

	/* 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)
3910
		return 1;
J
Jason Wang 已提交
3911

3912 3913 3914
	return 0;
}

3915 3916
static int kvm_io_bus_sort_cmp(const void *p1, const void *p2)
{
3917
	return kvm_io_bus_cmp(p1, p2);
3918 3919
}

G
Geoff Levand 已提交
3920
static int kvm_io_bus_get_first_dev(struct kvm_io_bus *bus,
3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937
			     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;

3938
	while (off > 0 && kvm_io_bus_cmp(&key, &bus->range[off-1]) == 0)
3939 3940 3941 3942 3943
		off--;

	return off;
}

3944
static int __kvm_io_bus_write(struct kvm_vcpu *vcpu, struct kvm_io_bus *bus,
C
Cornelia Huck 已提交
3945 3946 3947 3948 3949 3950 3951 3952 3953
			      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 &&
3954
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
3955
		if (!kvm_iodevice_write(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
3956 3957 3958 3959 3960 3961 3962 3963
					range->len, val))
			return idx;
		idx++;
	}

	return -EOPNOTSUPP;
}

3964
/* kvm_io_bus_write - called under kvm->slots_lock */
3965
int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
3966
		     int len, const void *val)
3967
{
3968
	struct kvm_io_bus *bus;
3969
	struct kvm_io_range range;
C
Cornelia Huck 已提交
3970
	int r;
3971 3972 3973 3974 3975

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
3976

3977
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3978 3979
	if (!bus)
		return -ENOMEM;
3980
	r = __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3981 3982
	return r < 0 ? r : 0;
}
L
Leo Yan 已提交
3983
EXPORT_SYMBOL_GPL(kvm_io_bus_write);
C
Cornelia Huck 已提交
3984 3985

/* kvm_io_bus_write_cookie - called under kvm->slots_lock */
3986 3987
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
3988 3989 3990 3991 3992 3993 3994 3995 3996
{
	struct kvm_io_bus *bus;
	struct kvm_io_range range;

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

3997
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3998 3999
	if (!bus)
		return -ENOMEM;
C
Cornelia Huck 已提交
4000 4001 4002

	/* First try the device referenced by cookie. */
	if ((cookie >= 0) && (cookie < bus->dev_count) &&
4003
	    (kvm_io_bus_cmp(&range, &bus->range[cookie]) == 0))
4004
		if (!kvm_iodevice_write(vcpu, bus->range[cookie].dev, addr, len,
C
Cornelia Huck 已提交
4005 4006 4007 4008 4009 4010 4011
					val))
			return cookie;

	/*
	 * cookie contained garbage; fall back to search and return the
	 * correct cookie value.
	 */
4012
	return __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
4013 4014
}

4015 4016
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
4017 4018 4019 4020
{
	int idx;

	idx = kvm_io_bus_get_first_dev(bus, range->addr, range->len);
4021 4022 4023 4024
	if (idx < 0)
		return -EOPNOTSUPP;

	while (idx < bus->dev_count &&
4025
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
4026
		if (!kvm_iodevice_read(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
4027 4028
				       range->len, val))
			return idx;
4029 4030 4031
		idx++;
	}

4032 4033
	return -EOPNOTSUPP;
}
4034

4035
/* kvm_io_bus_read - called under kvm->slots_lock */
4036
int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
M
Marcelo Tosatti 已提交
4037
		    int len, void *val)
4038
{
4039
	struct kvm_io_bus *bus;
4040
	struct kvm_io_range range;
C
Cornelia Huck 已提交
4041
	int r;
4042 4043 4044 4045 4046

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
M
Marcelo Tosatti 已提交
4047

4048
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
4049 4050
	if (!bus)
		return -ENOMEM;
4051
	r = __kvm_io_bus_read(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
4052 4053
	return r < 0 ? r : 0;
}
4054

4055
/* Caller must hold slots_lock. */
4056 4057
int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
			    int len, struct kvm_io_device *dev)
4058
{
4059
	int i;
M
Marcelo Tosatti 已提交
4060
	struct kvm_io_bus *new_bus, *bus;
4061
	struct kvm_io_range range;
4062

4063
	bus = kvm_get_bus(kvm, bus_idx);
4064 4065 4066
	if (!bus)
		return -ENOMEM;

4067 4068
	/* exclude ioeventfd which is limited by maximum fd */
	if (bus->dev_count - bus->ioeventfd_count > NR_IOBUS_DEVS - 1)
4069
		return -ENOSPC;
4070

4071
	new_bus = kmalloc(struct_size(bus, range, bus->dev_count + 1),
4072
			  GFP_KERNEL_ACCOUNT);
M
Marcelo Tosatti 已提交
4073 4074
	if (!new_bus)
		return -ENOMEM;
4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090

	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
4091 4092 4093
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
4094 4095 4096 4097

	return 0;
}

4098
/* Caller must hold slots_lock. */
4099 4100
void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
			       struct kvm_io_device *dev)
4101
{
4102
	int i;
M
Marcelo Tosatti 已提交
4103
	struct kvm_io_bus *new_bus, *bus;
4104

4105
	bus = kvm_get_bus(kvm, bus_idx);
4106
	if (!bus)
4107
		return;
4108

4109 4110
	for (i = 0; i < bus->dev_count; i++)
		if (bus->range[i].dev == dev) {
4111 4112
			break;
		}
M
Marcelo Tosatti 已提交
4113

4114 4115
	if (i == bus->dev_count)
		return;
4116

4117
	new_bus = kmalloc(struct_size(bus, range, bus->dev_count - 1),
4118
			  GFP_KERNEL_ACCOUNT);
4119 4120 4121 4122
	if (!new_bus)  {
		pr_err("kvm: failed to shrink bus, removing it completely\n");
		goto broken;
	}
4123 4124 4125 4126 4127

	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
4128

4129
broken:
M
Marcelo Tosatti 已提交
4130 4131 4132
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
4133
	return;
4134 4135
}

4136 4137 4138 4139 4140 4141 4142 4143 4144 4145
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);
4146 4147
	if (!bus)
		goto out_unlock;
4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161

	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);

4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173
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.
	 */
4174
	if (!refcount_inc_not_zero(&stat_data->kvm->users_count))
4175 4176
		return -ENOENT;

4177
	if (simple_attr_open(inode, file, get,
4178 4179 4180
		    KVM_DBGFS_GET_MODE(stat_data->dbgfs_item) & 0222
		    ? set : NULL,
		    fmt)) {
4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198
		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;
}

4199
static int kvm_get_stat_per_vm(struct kvm *kvm, size_t offset, u64 *val)
4200
{
4201
	*val = *(ulong *)((void *)kvm + offset);
4202

4203 4204 4205 4206 4207 4208
	return 0;
}

static int kvm_clear_stat_per_vm(struct kvm *kvm, size_t offset)
{
	*(ulong *)((void *)kvm + offset) = 0;
4209 4210 4211 4212

	return 0;
}

4213
static int kvm_get_stat_per_vcpu(struct kvm *kvm, size_t offset, u64 *val)
4214
{
4215 4216
	int i;
	struct kvm_vcpu *vcpu;
4217

4218
	*val = 0;
4219

4220 4221
	kvm_for_each_vcpu(i, vcpu, kvm)
		*val += *(u64 *)((void *)vcpu + offset);
4222 4223 4224 4225

	return 0;
}

4226
static int kvm_clear_stat_per_vcpu(struct kvm *kvm, size_t offset)
4227
{
4228 4229
	int i;
	struct kvm_vcpu *vcpu;
4230

4231 4232 4233 4234 4235
	kvm_for_each_vcpu(i, vcpu, kvm)
		*(u64 *)((void *)vcpu + offset) = 0;

	return 0;
}
4236

4237
static int kvm_stat_data_get(void *data, u64 *val)
4238
{
4239
	int r = -EFAULT;
4240 4241
	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)data;

4242 4243 4244 4245 4246 4247 4248 4249 4250 4251
	switch (stat_data->dbgfs_item->kind) {
	case KVM_STAT_VM:
		r = kvm_get_stat_per_vm(stat_data->kvm,
					stat_data->dbgfs_item->offset, val);
		break;
	case KVM_STAT_VCPU:
		r = kvm_get_stat_per_vcpu(stat_data->kvm,
					  stat_data->dbgfs_item->offset, val);
		break;
	}
4252

4253
	return r;
4254 4255
}

4256
static int kvm_stat_data_clear(void *data, u64 val)
4257
{
4258
	int r = -EFAULT;
4259 4260 4261 4262 4263
	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)data;

	if (val)
		return -EINVAL;

4264 4265 4266 4267 4268 4269 4270 4271 4272 4273
	switch (stat_data->dbgfs_item->kind) {
	case KVM_STAT_VM:
		r = kvm_clear_stat_per_vm(stat_data->kvm,
					  stat_data->dbgfs_item->offset);
		break;
	case KVM_STAT_VCPU:
		r = kvm_clear_stat_per_vcpu(stat_data->kvm,
					    stat_data->dbgfs_item->offset);
		break;
	}
4274

4275
	return r;
4276 4277
}

4278
static int kvm_stat_data_open(struct inode *inode, struct file *file)
4279 4280
{
	__simple_attr_check_format("%llu\n", 0ull);
4281 4282
	return kvm_debugfs_open(inode, file, kvm_stat_data_get,
				kvm_stat_data_clear, "%llu\n");
4283 4284
}

4285 4286 4287
static const struct file_operations stat_fops_per_vm = {
	.owner = THIS_MODULE,
	.open = kvm_stat_data_open,
4288
	.release = kvm_debugfs_release,
4289 4290 4291
	.read = simple_attr_read,
	.write = simple_attr_write,
	.llseek = no_llseek,
4292 4293
};

4294
static int vm_stat_get(void *_offset, u64 *val)
4295 4296 4297
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
4298
	u64 tmp_val;
4299

4300
	*val = 0;
J
Junaid Shahid 已提交
4301
	mutex_lock(&kvm_lock);
4302
	list_for_each_entry(kvm, &vm_list, vm_list) {
4303
		kvm_get_stat_per_vm(kvm, offset, &tmp_val);
4304 4305
		*val += tmp_val;
	}
J
Junaid Shahid 已提交
4306
	mutex_unlock(&kvm_lock);
4307
	return 0;
4308 4309
}

4310 4311 4312 4313 4314 4315 4316 4317
static int vm_stat_clear(void *_offset, u64 val)
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;

	if (val)
		return -EINVAL;

J
Junaid Shahid 已提交
4318
	mutex_lock(&kvm_lock);
4319
	list_for_each_entry(kvm, &vm_list, vm_list) {
4320
		kvm_clear_stat_per_vm(kvm, offset);
4321
	}
J
Junaid Shahid 已提交
4322
	mutex_unlock(&kvm_lock);
4323 4324 4325 4326 4327

	return 0;
}

DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, vm_stat_clear, "%llu\n");
4328

4329
static int vcpu_stat_get(void *_offset, u64 *val)
A
Avi Kivity 已提交
4330 4331 4332
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
4333
	u64 tmp_val;
A
Avi Kivity 已提交
4334

4335
	*val = 0;
J
Junaid Shahid 已提交
4336
	mutex_lock(&kvm_lock);
4337
	list_for_each_entry(kvm, &vm_list, vm_list) {
4338
		kvm_get_stat_per_vcpu(kvm, offset, &tmp_val);
4339 4340
		*val += tmp_val;
	}
J
Junaid Shahid 已提交
4341
	mutex_unlock(&kvm_lock);
4342
	return 0;
A
Avi Kivity 已提交
4343 4344
}

4345 4346 4347 4348 4349 4350 4351 4352
static int vcpu_stat_clear(void *_offset, u64 val)
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;

	if (val)
		return -EINVAL;

J
Junaid Shahid 已提交
4353
	mutex_lock(&kvm_lock);
4354
	list_for_each_entry(kvm, &vm_list, vm_list) {
4355
		kvm_clear_stat_per_vcpu(kvm, offset);
4356
	}
J
Junaid Shahid 已提交
4357
	mutex_unlock(&kvm_lock);
4358 4359 4360 4361 4362 4363

	return 0;
}

DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, vcpu_stat_clear,
			"%llu\n");
4364

4365
static const struct file_operations *stat_fops[] = {
4366 4367 4368
	[KVM_STAT_VCPU] = &vcpu_stat_fops,
	[KVM_STAT_VM]   = &vm_stat_fops,
};
A
Avi Kivity 已提交
4369

4370 4371 4372 4373 4374 4375 4376 4377
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;

J
Junaid Shahid 已提交
4378
	mutex_lock(&kvm_lock);
4379 4380 4381 4382 4383 4384 4385 4386
	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;
J
Junaid Shahid 已提交
4387
	mutex_unlock(&kvm_lock);
4388

4389
	env = kzalloc(sizeof(*env), GFP_KERNEL_ACCOUNT);
4390 4391 4392 4393 4394 4395
	if (!env)
		return;

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

4396
	if (type == KVM_EVENT_CREATE_VM) {
4397
		add_uevent_var(env, "EVENT=create");
4398 4399
		kvm->userspace_pid = task_pid_nr(current);
	} else if (type == KVM_EVENT_DESTROY_VM) {
4400
		add_uevent_var(env, "EVENT=destroy");
4401 4402
	}
	add_uevent_var(env, "PID=%d", kvm->userspace_pid);
4403

4404
	if (!IS_ERR_OR_NULL(kvm->debugfs_dentry)) {
4405
		char *tmp, *p = kmalloc(PATH_MAX, GFP_KERNEL_ACCOUNT);
4406 4407 4408 4409 4410 4411

		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);
4412 4413 4414 4415 4416 4417 4418 4419
		}
	}
	/* 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);
}

4420
static void kvm_init_debug(void)
A
Avi Kivity 已提交
4421 4422 4423
{
	struct kvm_stats_debugfs_item *p;

4424
	kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
4425

4426 4427
	kvm_debugfs_num_entries = 0;
	for (p = debugfs_entries; p->name; ++p, kvm_debugfs_num_entries++) {
4428 4429
		debugfs_create_file(p->name, KVM_DBGFS_GET_MODE(p),
				    kvm_debugfs_dir, (void *)(long)p->offset,
4430
				    stat_fops[p->kind]);
4431
	}
A
Avi Kivity 已提交
4432 4433
}

4434
static int kvm_suspend(void)
4435
{
4436
	if (kvm_usage_count)
4437
		hardware_disable_nolock(NULL);
4438 4439 4440
	return 0;
}

4441
static void kvm_resume(void)
4442
{
4443
	if (kvm_usage_count) {
4444 4445 4446
#ifdef CONFIG_LOCKDEP
		WARN_ON(lockdep_is_held(&kvm_count_lock));
#endif
4447
		hardware_enable_nolock(NULL);
4448
	}
4449 4450
}

4451
static struct syscore_ops kvm_syscore_ops = {
4452 4453 4454 4455
	.suspend = kvm_suspend,
	.resume = kvm_resume,
};

4456 4457 4458 4459 4460 4461 4462 4463 4464
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);
4465

4466
	WRITE_ONCE(vcpu->preempted, false);
4467
	WRITE_ONCE(vcpu->ready, false);
4468

4469
	__this_cpu_write(kvm_running_vcpu, vcpu);
R
Radim Krčmář 已提交
4470
	kvm_arch_sched_in(vcpu, cpu);
4471
	kvm_arch_vcpu_load(vcpu, cpu);
4472 4473 4474 4475 4476 4477 4478
}

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

4479
	if (current->state == TASK_RUNNING) {
4480
		WRITE_ONCE(vcpu->preempted, true);
4481 4482
		WRITE_ONCE(vcpu->ready, true);
	}
4483
	kvm_arch_vcpu_put(vcpu);
4484 4485 4486 4487 4488
	__this_cpu_write(kvm_running_vcpu, NULL);
}

/**
 * kvm_get_running_vcpu - get the vcpu running on the current CPU.
4489 4490 4491 4492 4493 4494
 *
 * We can disable preemption locally around accessing the per-CPU variable,
 * and use the resolved vcpu pointer after enabling preemption again,
 * because even if the current thread is migrated to another CPU, reading
 * the per-CPU value later will give us the same value as we update the
 * per-CPU variable in the preempt notifier handlers.
4495 4496 4497
 */
struct kvm_vcpu *kvm_get_running_vcpu(void)
{
4498 4499 4500 4501 4502 4503 4504
	struct kvm_vcpu *vcpu;

	preempt_disable();
	vcpu = __this_cpu_read(kvm_running_vcpu);
	preempt_enable();

	return vcpu;
4505 4506 4507 4508 4509 4510 4511 4512
}

/**
 * kvm_get_running_vcpus - get the per-CPU array of currently running vcpus.
 */
struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void)
{
        return &kvm_running_vcpu;
4513 4514
}

4515 4516 4517 4518 4519
static void check_processor_compat(void *rtn)
{
	*(int *)rtn = kvm_arch_check_processor_compat();
}

4520
int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
4521
		  struct module *module)
A
Avi Kivity 已提交
4522 4523
{
	int r;
Y
Yang, Sheng 已提交
4524
	int cpu;
A
Avi Kivity 已提交
4525

4526 4527
	r = kvm_arch_init(opaque);
	if (r)
4528
		goto out_fail;
4529

4530 4531 4532 4533
	/*
	 * 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
4534 4535
	 * kvm_arch_init must be called before kvm_irqfd_init to avoid creating
	 * conflicts in case kvm is already setup for another implementation.
4536
	 */
P
Paolo Bonzini 已提交
4537 4538 4539
	r = kvm_irqfd_init();
	if (r)
		goto out_irqfd;
4540

4541
	if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
4542 4543 4544 4545
		r = -ENOMEM;
		goto out_free_0;
	}

4546
	r = kvm_arch_hardware_setup();
A
Avi Kivity 已提交
4547
	if (r < 0)
4548
		goto out_free_1;
A
Avi Kivity 已提交
4549

Y
Yang, Sheng 已提交
4550
	for_each_online_cpu(cpu) {
4551
		smp_call_function_single(cpu, check_processor_compat, &r, 1);
Y
Yang, Sheng 已提交
4552
		if (r < 0)
4553
			goto out_free_2;
Y
Yang, Sheng 已提交
4554 4555
	}

T
Thomas Gleixner 已提交
4556
	r = cpuhp_setup_state_nocalls(CPUHP_AP_KVM_STARTING, "kvm/cpu:starting",
4557
				      kvm_starting_cpu, kvm_dying_cpu);
A
Avi Kivity 已提交
4558
	if (r)
4559
		goto out_free_2;
A
Avi Kivity 已提交
4560 4561
	register_reboot_notifier(&kvm_reboot_notifier);

4562
	/* A kmem cache lets us meet the alignment requirements of fx_save. */
4563 4564
	if (!vcpu_align)
		vcpu_align = __alignof__(struct kvm_vcpu);
4565 4566 4567 4568 4569 4570
	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);
4571 4572
	if (!kvm_vcpu_cache) {
		r = -ENOMEM;
4573
		goto out_free_3;
4574 4575
	}

4576 4577 4578 4579
	r = kvm_async_pf_init();
	if (r)
		goto out_free;

A
Avi Kivity 已提交
4580
	kvm_chardev_ops.owner = module;
4581 4582
	kvm_vm_fops.owner = module;
	kvm_vcpu_fops.owner = module;
A
Avi Kivity 已提交
4583 4584 4585

	r = misc_register(&kvm_dev);
	if (r) {
X
Xiubo Li 已提交
4586
		pr_err("kvm: misc device register failed\n");
4587
		goto out_unreg;
A
Avi Kivity 已提交
4588 4589
	}

4590 4591
	register_syscore_ops(&kvm_syscore_ops);

4592 4593 4594
	kvm_preempt_ops.sched_in = kvm_sched_in;
	kvm_preempt_ops.sched_out = kvm_sched_out;

4595
	kvm_init_debug();
4596

P
Paolo Bonzini 已提交
4597 4598 4599
	r = kvm_vfio_ops_init();
	WARN_ON(r);

4600
	return 0;
A
Avi Kivity 已提交
4601

4602 4603
out_unreg:
	kvm_async_pf_deinit();
A
Avi Kivity 已提交
4604
out_free:
4605
	kmem_cache_destroy(kvm_vcpu_cache);
4606
out_free_3:
A
Avi Kivity 已提交
4607
	unregister_reboot_notifier(&kvm_reboot_notifier);
4608
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4609
out_free_2:
4610
	kvm_arch_hardware_unsetup();
4611
out_free_1:
4612
	free_cpumask_var(cpus_hardware_enabled);
4613
out_free_0:
4614
	kvm_irqfd_exit();
P
Paolo Bonzini 已提交
4615
out_irqfd:
4616 4617
	kvm_arch_exit();
out_fail:
A
Avi Kivity 已提交
4618 4619
	return r;
}
4620
EXPORT_SYMBOL_GPL(kvm_init);
A
Avi Kivity 已提交
4621

4622
void kvm_exit(void)
A
Avi Kivity 已提交
4623
{
4624
	debugfs_remove_recursive(kvm_debugfs_dir);
A
Avi Kivity 已提交
4625
	misc_deregister(&kvm_dev);
4626
	kmem_cache_destroy(kvm_vcpu_cache);
4627
	kvm_async_pf_deinit();
4628
	unregister_syscore_ops(&kvm_syscore_ops);
A
Avi Kivity 已提交
4629
	unregister_reboot_notifier(&kvm_reboot_notifier);
4630
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4631
	on_each_cpu(hardware_disable_nolock, NULL, 1);
4632
	kvm_arch_hardware_unsetup();
4633
	kvm_arch_exit();
4634
	kvm_irqfd_exit();
4635
	free_cpumask_var(cpus_hardware_enabled);
4636
	kvm_vfio_ops_exit();
A
Avi Kivity 已提交
4637
}
4638
EXPORT_SYMBOL_GPL(kvm_exit);
4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716 4717 4718 4719 4720 4721

struct kvm_vm_worker_thread_context {
	struct kvm *kvm;
	struct task_struct *parent;
	struct completion init_done;
	kvm_vm_thread_fn_t thread_fn;
	uintptr_t data;
	int err;
};

static int kvm_vm_worker_thread(void *context)
{
	/*
	 * The init_context is allocated on the stack of the parent thread, so
	 * we have to locally copy anything that is needed beyond initialization
	 */
	struct kvm_vm_worker_thread_context *init_context = context;
	struct kvm *kvm = init_context->kvm;
	kvm_vm_thread_fn_t thread_fn = init_context->thread_fn;
	uintptr_t data = init_context->data;
	int err;

	err = kthread_park(current);
	/* kthread_park(current) is never supposed to return an error */
	WARN_ON(err != 0);
	if (err)
		goto init_complete;

	err = cgroup_attach_task_all(init_context->parent, current);
	if (err) {
		kvm_err("%s: cgroup_attach_task_all failed with err %d\n",
			__func__, err);
		goto init_complete;
	}

	set_user_nice(current, task_nice(init_context->parent));

init_complete:
	init_context->err = err;
	complete(&init_context->init_done);
	init_context = NULL;

	if (err)
		return err;

	/* Wait to be woken up by the spawner before proceeding. */
	kthread_parkme();

	if (!kthread_should_stop())
		err = thread_fn(kvm, data);

	return err;
}

int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn,
				uintptr_t data, const char *name,
				struct task_struct **thread_ptr)
{
	struct kvm_vm_worker_thread_context init_context = {};
	struct task_struct *thread;

	*thread_ptr = NULL;
	init_context.kvm = kvm;
	init_context.parent = current;
	init_context.thread_fn = thread_fn;
	init_context.data = data;
	init_completion(&init_context.init_done);

	thread = kthread_run(kvm_vm_worker_thread, &init_context,
			     "%s-%d", name, task_pid_nr(current));
	if (IS_ERR(thread))
		return PTR_ERR(thread);

	/* kthread_run is never supposed to return NULL */
	WARN_ON(thread == NULL);

	wait_for_completion(&init_context.init_done);

	if (!init_context.err)
		*thread_ptr = thread;

	return init_context.err;
}