kvm_main.c 116.5 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] = -1;
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
 * Delete a memslot by decrementing the number of used slots and shifting all
 * other entries in the array forward one spot.
875
 */
876 877
static inline void kvm_memslot_delete(struct kvm_memslots *slots,
				      struct kvm_memory_slot *memslot)
878
{
879
	struct kvm_memory_slot *mslots = slots->memslots;
880
	int i;
881

882 883
	if (WARN_ON(slots->id_to_index[memslot->id] == -1))
		return;
884

885 886 887
	slots->used_slots--;

	for (i = slots->id_to_index[memslot->id]; i < slots->used_slots; i++) {
888 889 890
		mslots[i] = mslots[i + 1];
		slots->id_to_index[mslots[i].id] = i;
	}
891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919
	mslots[i] = *memslot;
	slots->id_to_index[memslot->id] = -1;
}

/*
 * "Insert" a new memslot by incrementing the number of used slots.  Returns
 * the new slot's initial index into the memslots array.
 */
static inline int kvm_memslot_insert_back(struct kvm_memslots *slots)
{
	return slots->used_slots++;
}

/*
 * Move a changed memslot backwards in the array by shifting existing slots
 * with a higher GFN toward the front of the array.  Note, the changed memslot
 * itself is not preserved in the array, i.e. not swapped at this time, only
 * its new index into the array is tracked.  Returns the changed memslot's
 * current index into the memslots array.
 */
static inline int kvm_memslot_move_backward(struct kvm_memslots *slots,
					    struct kvm_memory_slot *memslot)
{
	struct kvm_memory_slot *mslots = slots->memslots;
	int i;

	if (WARN_ON_ONCE(slots->id_to_index[memslot->id] == -1) ||
	    WARN_ON_ONCE(!slots->used_slots))
		return -1;
920 921

	/*
922 923 924
	 * Move the target memslot backward in the array by shifting existing
	 * memslots with a higher GFN (than the target memslot) towards the
	 * front of the array.
925
	 */
926 927 928 929 930
	for (i = slots->id_to_index[memslot->id]; i < slots->used_slots - 1; i++) {
		if (memslot->base_gfn > mslots[i + 1].base_gfn)
			break;

		WARN_ON_ONCE(memslot->base_gfn == mslots[i + 1].base_gfn);
931

932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 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
		/* Shift the next memslot forward one and update its index. */
		mslots[i] = mslots[i + 1];
		slots->id_to_index[mslots[i].id] = i;
	}
	return i;
}

/*
 * Move a changed memslot forwards in the array by shifting existing slots with
 * a lower GFN toward the back of the array.  Note, the changed memslot itself
 * is not preserved in the array, i.e. not swapped at this time, only its new
 * index into the array is tracked.  Returns the changed memslot's final index
 * into the memslots array.
 */
static inline int kvm_memslot_move_forward(struct kvm_memslots *slots,
					   struct kvm_memory_slot *memslot,
					   int start)
{
	struct kvm_memory_slot *mslots = slots->memslots;
	int i;

	for (i = start; i > 0; i--) {
		if (memslot->base_gfn < mslots[i - 1].base_gfn)
			break;

		WARN_ON_ONCE(memslot->base_gfn == mslots[i - 1].base_gfn);

		/* Shift the next memslot back one and update its index. */
		mslots[i] = mslots[i - 1];
		slots->id_to_index[mslots[i].id] = i;
	}
	return i;
}

/*
 * Re-sort memslots based on their GFN to account for an added, deleted, or
 * moved memslot.  Sorting memslots by GFN allows using a binary search during
 * memslot lookup.
 *
 * IMPORTANT: Slots are sorted from highest GFN to lowest GFN!  I.e. the entry
 * at memslots[0] has the highest GFN.
 *
 * The sorting algorithm takes advantage of having initially sorted memslots
 * and knowing the position of the changed memslot.  Sorting is also optimized
 * by not swapping the updated memslot and instead only shifting other memslots
 * and tracking the new index for the update memslot.  Only once its final
 * index is known is the updated memslot copied into its position in the array.
 *
 *  - When deleting a memslot, the deleted memslot simply needs to be moved to
 *    the end of the array.
 *
 *  - When creating a memslot, the algorithm "inserts" the new memslot at the
 *    end of the array and then it forward to its correct location.
 *
 *  - When moving a memslot, the algorithm first moves the updated memslot
 *    backward to handle the scenario where the memslot's GFN was changed to a
 *    lower value.  update_memslots() then falls through and runs the same flow
 *    as creating a memslot to move the memslot forward to handle the scenario
 *    where its GFN was changed to a higher value.
 *
 * Note, slots are sorted from highest->lowest instead of lowest->highest for
 * historical reasons.  Originally, invalid memslots where denoted by having
 * GFN=0, thus sorting from highest->lowest naturally sorted invalid memslots
 * to the end of the array.  The current algorithm uses dedicated logic to
 * delete a memslot and thus does not rely on invalid memslots having GFN=0.
 *
 * The other historical motiviation for highest->lowest was to improve the
 * performance of memslot lookup.  KVM originally used a linear search starting
 * at memslots[0].  On x86, the largest memslot usually has one of the highest,
 * if not *the* highest, GFN, as the bulk of the guest's RAM is located in a
 * single memslot above the 4gb boundary.  As the largest memslot is also the
 * most likely to be referenced, sorting it to the front of the array was
 * advantageous.  The current binary search starts from the middle of the array
 * and uses an LRU pointer to improve performance for all memslots and GFNs.
 */
static void update_memslots(struct kvm_memslots *slots,
			    struct kvm_memory_slot *memslot,
			    enum kvm_mr_change change)
{
	int i;

	if (change == KVM_MR_DELETE) {
		kvm_memslot_delete(slots, memslot);
	} else {
		if (change == KVM_MR_CREATE)
			i = kvm_memslot_insert_back(slots);
		else
			i = kvm_memslot_move_backward(slots, memslot);
		i = kvm_memslot_move_forward(slots, memslot, i);

		/*
		 * Copy the memslot to its new position in memslots and update
		 * its index accordingly.
		 */
		slots->memslots[i] = *memslot;
		slots->id_to_index[memslot->id] = i;
	}
1029 1030
}

1031
static int check_memory_region_flags(const struct kvm_userspace_memory_region *mem)
1032
{
X
Xiao Guangrong 已提交
1033 1034
	u32 valid_flags = KVM_MEM_LOG_DIRTY_PAGES;

1035
#ifdef __KVM_HAVE_READONLY_MEM
X
Xiao Guangrong 已提交
1036 1037 1038 1039
	valid_flags |= KVM_MEM_READONLY;
#endif

	if (mem->flags & ~valid_flags)
1040 1041 1042 1043 1044
		return -EINVAL;

	return 0;
}

1045
static struct kvm_memslots *install_new_memslots(struct kvm *kvm,
1046
		int as_id, struct kvm_memslots *slots)
1047
{
1048
	struct kvm_memslots *old_memslots = __kvm_memslots(kvm, as_id);
1049
	u64 gen = old_memslots->generation;
1050

1051 1052
	WARN_ON(gen & KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS);
	slots->generation = gen | KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS;
1053

1054
	rcu_assign_pointer(kvm->memslots[as_id], slots);
1055
	synchronize_srcu_expedited(&kvm->srcu);
1056

1057
	/*
1058
	 * Increment the new memslot generation a second time, dropping the
M
Miaohe Lin 已提交
1059
	 * update in-progress flag and incrementing the generation based on
1060 1061 1062 1063 1064 1065
	 * 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;

	/*
1066 1067 1068
	 * 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
1069 1070
	 * space 0 will use generations 0, 2, 4, ... while address space 1 will
	 * use generations 1, 3, 5, ...
1071
	 */
1072
	gen += KVM_ADDRESS_SPACE_NUM;
1073

1074
	kvm_arch_memslots_updated(kvm, gen);
1075

1076
	slots->generation = gen;
1077 1078

	return old_memslots;
1079 1080
}

1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106
/*
 * Note, at a minimum, the current number of used slots must be allocated, even
 * when deleting a memslot, as we need a complete duplicate of the memslots for
 * use when invalidating a memslot prior to deleting/moving the memslot.
 */
static struct kvm_memslots *kvm_dup_memslots(struct kvm_memslots *old,
					     enum kvm_mr_change change)
{
	struct kvm_memslots *slots;
	size_t old_size, new_size;

	old_size = sizeof(struct kvm_memslots) +
		   (sizeof(struct kvm_memory_slot) * old->used_slots);

	if (change == KVM_MR_CREATE)
		new_size = old_size + sizeof(struct kvm_memory_slot);
	else
		new_size = old_size;

	slots = kvzalloc(new_size, GFP_KERNEL_ACCOUNT);
	if (likely(slots))
		memcpy(slots, old, old_size);

	return slots;
}

1107 1108
static int kvm_set_memslot(struct kvm *kvm,
			   const struct kvm_userspace_memory_region *mem,
1109
			   struct kvm_memory_slot *old,
1110 1111 1112 1113 1114 1115 1116
			   struct kvm_memory_slot *new, int as_id,
			   enum kvm_mr_change change)
{
	struct kvm_memory_slot *slot;
	struct kvm_memslots *slots;
	int r;

1117
	slots = kvm_dup_memslots(__kvm_memslots(kvm, as_id), change);
1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165
	if (!slots)
		return -ENOMEM;

	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;
}

1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182
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;

1183
	kvm_free_memslot(kvm, old);
1184 1185 1186
	return 0;
}

A
Avi Kivity 已提交
1187 1188 1189 1190 1191
/*
 * Allocate some memory and give it an address in the guest physical address
 * space.
 *
 * Discontiguous memory is allowed, mostly for framebuffers.
1192
 *
1193
 * Must be called holding kvm->slots_lock for write.
A
Avi Kivity 已提交
1194
 */
1195
int __kvm_set_memory_region(struct kvm *kvm,
1196
			    const struct kvm_userspace_memory_region *mem)
A
Avi Kivity 已提交
1197 1198
{
	struct kvm_memory_slot old, new;
1199
	struct kvm_memory_slot *tmp;
1200
	enum kvm_mr_change change;
1201 1202
	int as_id, id;
	int r;
A
Avi Kivity 已提交
1203

1204 1205
	r = check_memory_region_flags(mem);
	if (r)
1206
		return r;
1207

1208 1209 1210
	as_id = mem->slot >> 16;
	id = (u16)mem->slot;

A
Avi Kivity 已提交
1211 1212
	/* General sanity checks */
	if (mem->memory_size & (PAGE_SIZE - 1))
1213
		return -EINVAL;
A
Avi Kivity 已提交
1214
	if (mem->guest_phys_addr & (PAGE_SIZE - 1))
1215
		return -EINVAL;
1216
	/* We can read the guest memory with __xxx_user() later on. */
1217
	if ((id < KVM_USER_MEM_SLOTS) &&
1218
	    ((mem->userspace_addr & (PAGE_SIZE - 1)) ||
1219
	     !access_ok((void __user *)(unsigned long)mem->userspace_addr,
1220
			mem->memory_size)))
1221
		return -EINVAL;
1222
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_MEM_SLOTS_NUM)
1223
		return -EINVAL;
A
Avi Kivity 已提交
1224
	if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
1225
		return -EINVAL;
A
Avi Kivity 已提交
1226

1227 1228 1229 1230
	/*
	 * 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.
1231
	 * to free its resources and for arch specific behavior.
1232
	 */
1233 1234 1235 1236 1237 1238 1239 1240
	tmp = id_to_memslot(__kvm_memslots(kvm, as_id), id);
	if (tmp) {
		old = *tmp;
		tmp = NULL;
	} else {
		memset(&old, 0, sizeof(old));
		old.id = id;
	}
1241

1242 1243 1244
	if (!mem->memory_size)
		return kvm_delete_memslot(kvm, mem, &old, as_id);

1245
	new.id = id;
1246 1247
	new.base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
	new.npages = mem->memory_size >> PAGE_SHIFT;
A
Avi Kivity 已提交
1248
	new.flags = mem->flags;
1249
	new.userspace_addr = mem->userspace_addr;
A
Avi Kivity 已提交
1250

1251 1252 1253
	if (new.npages > KVM_MEM_MAX_NR_PAGES)
		return -EINVAL;

1254 1255
	if (!old.npages) {
		change = KVM_MR_CREATE;
1256 1257
		new.dirty_bitmap = NULL;
		memset(&new.arch, 0, sizeof(new.arch));
1258 1259
	} else { /* Modify an existing slot. */
		if ((new.userspace_addr != old.userspace_addr) ||
1260
		    (new.npages != old.npages) ||
1261
		    ((new.flags ^ old.flags) & KVM_MEM_READONLY))
1262
			return -EINVAL;
1263

1264
		if (new.base_gfn != old.base_gfn)
1265 1266 1267 1268 1269
			change = KVM_MR_MOVE;
		else if (new.flags != old.flags)
			change = KVM_MR_FLAGS_ONLY;
		else /* Nothing to change. */
			return 0;
1270 1271 1272 1273

		/* Copy dirty_bitmap and arch from the current memslot. */
		new.dirty_bitmap = old.dirty_bitmap;
		memcpy(&new.arch, &old.arch, sizeof(new.arch));
1274
	}
A
Avi Kivity 已提交
1275

1276
	if ((change == KVM_MR_CREATE) || (change == KVM_MR_MOVE)) {
1277
		/* Check for overlaps */
1278 1279
		kvm_for_each_memslot(tmp, __kvm_memslots(kvm, as_id)) {
			if (tmp->id == id)
1280
				continue;
1281 1282
			if (!((new.base_gfn + new.npages <= tmp->base_gfn) ||
			      (new.base_gfn >= tmp->base_gfn + tmp->npages)))
1283
				return -EEXIST;
1284
		}
A
Avi Kivity 已提交
1285 1286
	}

1287 1288 1289 1290
	/* Allocate/free page dirty bitmap as needed */
	if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
		new.dirty_bitmap = NULL;
	else if (!new.dirty_bitmap) {
1291 1292 1293
		r = kvm_create_dirty_bitmap(&new);
		if (r)
			return r;
A
Avi Kivity 已提交
1294 1295
	}

1296 1297 1298
	r = kvm_set_memslot(kvm, mem, &old, &new, as_id, change);
	if (r)
		goto out_bitmap;
1299

1300 1301
	if (old.dirty_bitmap && !new.dirty_bitmap)
		kvm_destroy_dirty_bitmap(&old);
A
Avi Kivity 已提交
1302 1303
	return 0;

1304 1305 1306
out_bitmap:
	if (new.dirty_bitmap && !old.dirty_bitmap)
		kvm_destroy_dirty_bitmap(&new);
A
Avi Kivity 已提交
1307
	return r;
1308
}
1309 1310 1311
EXPORT_SYMBOL_GPL(__kvm_set_memory_region);

int kvm_set_memory_region(struct kvm *kvm,
1312
			  const struct kvm_userspace_memory_region *mem)
1313 1314 1315
{
	int r;

1316
	mutex_lock(&kvm->slots_lock);
1317
	r = __kvm_set_memory_region(kvm, mem);
1318
	mutex_unlock(&kvm->slots_lock);
1319 1320
	return r;
}
1321 1322
EXPORT_SYMBOL_GPL(kvm_set_memory_region);

1323 1324
static int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
					  struct kvm_userspace_memory_region *mem)
1325
{
1326
	if ((u16)mem->slot >= KVM_USER_MEM_SLOTS)
1327
		return -EINVAL;
1328

1329
	return kvm_set_memory_region(kvm, mem);
A
Avi Kivity 已提交
1330 1331
}

1332
#ifndef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
1333 1334 1335 1336 1337 1338 1339 1340 1341
/**
 * kvm_get_dirty_log - get a snapshot of dirty pages
 * @kvm:	pointer to kvm instance
 * @log:	slot id and address to which we copy the log
 * @is_dirty:	set to '1' if any dirty pages were found
 * @memslot:	set to the associated memslot, always valid on success
 */
int kvm_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log,
		      int *is_dirty, struct kvm_memory_slot **memslot)
A
Avi Kivity 已提交
1342
{
1343
	struct kvm_memslots *slots;
1344
	int i, as_id, id;
1345
	unsigned long n;
A
Avi Kivity 已提交
1346 1347
	unsigned long any = 0;

1348 1349 1350
	*memslot = NULL;
	*is_dirty = 0;

1351 1352 1353
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1354
		return -EINVAL;
A
Avi Kivity 已提交
1355

1356
	slots = __kvm_memslots(kvm, as_id);
1357
	*memslot = id_to_memslot(slots, id);
1358
	if (!(*memslot) || !(*memslot)->dirty_bitmap)
1359
		return -ENOENT;
A
Avi Kivity 已提交
1360

1361 1362 1363
	kvm_arch_sync_dirty_log(kvm, *memslot);

	n = kvm_dirty_bitmap_bytes(*memslot);
A
Avi Kivity 已提交
1364

1365
	for (i = 0; !any && i < n/sizeof(long); ++i)
1366
		any = (*memslot)->dirty_bitmap[i];
A
Avi Kivity 已提交
1367

1368
	if (copy_to_user(log->dirty_bitmap, (*memslot)->dirty_bitmap, n))
1369
		return -EFAULT;
A
Avi Kivity 已提交
1370

1371 1372
	if (any)
		*is_dirty = 1;
1373
	return 0;
A
Avi Kivity 已提交
1374
}
1375
EXPORT_SYMBOL_GPL(kvm_get_dirty_log);
A
Avi Kivity 已提交
1376

1377
#else /* CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
1378
/**
J
Jiang Biao 已提交
1379
 * kvm_get_dirty_log_protect - get a snapshot of dirty pages
1380
 *	and reenable dirty page tracking for the corresponding pages.
1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398
 * @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.
 *
 */
1399
static int kvm_get_dirty_log_protect(struct kvm *kvm, struct kvm_dirty_log *log)
1400
{
1401
	struct kvm_memslots *slots;
1402
	struct kvm_memory_slot *memslot;
1403
	int i, as_id, id;
1404 1405 1406
	unsigned long n;
	unsigned long *dirty_bitmap;
	unsigned long *dirty_bitmap_buffer;
1407
	bool flush;
1408

1409 1410 1411
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1412
		return -EINVAL;
1413

1414 1415
	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
1416 1417
	if (!memslot || !memslot->dirty_bitmap)
		return -ENOENT;
1418 1419 1420

	dirty_bitmap = memslot->dirty_bitmap;

1421 1422
	kvm_arch_sync_dirty_log(kvm, memslot);

1423
	n = kvm_dirty_bitmap_bytes(memslot);
1424
	flush = false;
1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437
	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);
1438

1439 1440 1441 1442
		spin_lock(&kvm->mmu_lock);
		for (i = 0; i < n / sizeof(long); i++) {
			unsigned long mask;
			gfn_t offset;
1443

1444 1445 1446
			if (!dirty_bitmap[i])
				continue;

1447
			flush = true;
1448 1449 1450
			mask = xchg(&dirty_bitmap[i], 0);
			dirty_bitmap_buffer[i] = mask;

1451 1452 1453
			offset = i * BITS_PER_LONG;
			kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot,
								offset, mask);
1454 1455 1456 1457
		}
		spin_unlock(&kvm->mmu_lock);
	}

1458 1459 1460
	if (flush)
		kvm_arch_flush_remote_tlbs_memslot(kvm, memslot);

1461 1462 1463 1464
	if (copy_to_user(log->dirty_bitmap, dirty_bitmap_buffer, n))
		return -EFAULT;
	return 0;
}
1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497


/**
 * 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;
}
1498 1499 1500 1501 1502 1503 1504

/**
 * 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
 */
1505 1506
static int kvm_clear_dirty_log_protect(struct kvm *kvm,
				       struct kvm_clear_dirty_log *log)
1507 1508 1509
{
	struct kvm_memslots *slots;
	struct kvm_memory_slot *memslot;
1510
	int as_id, id;
1511
	gfn_t offset;
1512
	unsigned long i, n;
1513 1514
	unsigned long *dirty_bitmap;
	unsigned long *dirty_bitmap_buffer;
1515
	bool flush;
1516 1517 1518 1519 1520 1521

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

1522
	if (log->first_page & 63)
1523 1524 1525 1526
		return -EINVAL;

	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
1527 1528
	if (!memslot || !memslot->dirty_bitmap)
		return -ENOENT;
1529 1530 1531

	dirty_bitmap = memslot->dirty_bitmap;

1532
	n = ALIGN(log->num_pages, BITS_PER_LONG) / 8;
1533 1534

	if (log->first_page > memslot->npages ||
1535 1536 1537
	    log->num_pages > memslot->npages - log->first_page ||
	    (log->num_pages < memslot->npages - log->first_page && (log->num_pages & 63)))
	    return -EINVAL;
1538

1539 1540 1541
	kvm_arch_sync_dirty_log(kvm, memslot);

	flush = false;
1542 1543 1544
	dirty_bitmap_buffer = kvm_second_dirty_bitmap(memslot);
	if (copy_from_user(dirty_bitmap_buffer, log->dirty_bitmap, n))
		return -EFAULT;
1545

1546
	spin_lock(&kvm->mmu_lock);
1547 1548
	for (offset = log->first_page, i = offset / BITS_PER_LONG,
		 n = DIV_ROUND_UP(log->num_pages, BITS_PER_LONG); n--;
1549 1550 1551 1552
	     i++, offset += BITS_PER_LONG) {
		unsigned long mask = *dirty_bitmap_buffer++;
		atomic_long_t *p = (atomic_long_t *) &dirty_bitmap[i];
		if (!mask)
1553 1554
			continue;

1555
		mask &= atomic_long_fetch_andnot(mask, p);
1556

1557 1558 1559 1560 1561 1562
		/*
		 * 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.
		*/
1563
		if (mask) {
1564
			flush = true;
1565 1566 1567
			kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot,
								offset, mask);
		}
1568 1569
	}
	spin_unlock(&kvm->mmu_lock);
1570

1571 1572 1573
	if (flush)
		kvm_arch_flush_remote_tlbs_memslot(kvm, memslot);

1574
	return 0;
1575
}
1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589

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 */
1590

1591 1592 1593 1594 1595
bool kvm_largepages_enabled(void)
{
	return largepages_enabled;
}

1596 1597 1598 1599 1600 1601
void kvm_disable_largepages(void)
{
	largepages_enabled = false;
}
EXPORT_SYMBOL_GPL(kvm_disable_largepages);

1602 1603 1604 1605
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
1606
EXPORT_SYMBOL_GPL(gfn_to_memslot);
A
Avi Kivity 已提交
1607

1608 1609 1610 1611 1612
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);
}

1613
bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
1614
{
1615
	struct kvm_memory_slot *memslot = gfn_to_memslot(kvm, gfn);
1616

1617
	if (!memslot || memslot->id >= KVM_USER_MEM_SLOTS ||
1618
	      memslot->flags & KVM_MEMSLOT_INVALID)
1619
		return false;
1620

1621
	return true;
1622 1623 1624
}
EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);

1625
unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn)
J
Joerg Roedel 已提交
1626 1627 1628 1629 1630 1631
{
	struct vm_area_struct *vma;
	unsigned long addr, size;

	size = PAGE_SIZE;

1632
	addr = kvm_vcpu_gfn_to_hva_prot(vcpu, gfn, NULL);
J
Joerg Roedel 已提交
1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648
	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
1649 1650 1651 1652 1653 1654 1655
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
1656
{
1657
	if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
X
Xiao Guangrong 已提交
1658
		return KVM_HVA_ERR_BAD;
1659

X
Xiao Guangrong 已提交
1660 1661
	if (memslot_is_readonly(slot) && write)
		return KVM_HVA_ERR_RO_BAD;
1662 1663 1664 1665

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

X
Xiao Guangrong 已提交
1666
	return __gfn_to_hva_memslot(slot, gfn);
I
Izik Eidus 已提交
1667
}
1668

X
Xiao Guangrong 已提交
1669 1670 1671 1672
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
1673
}
1674

X
Xiao Guangrong 已提交
1675
unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot,
1676
					gfn_t gfn)
X
Xiao Guangrong 已提交
1677 1678 1679 1680 1681
{
	return gfn_to_hva_many(slot, gfn, NULL);
}
EXPORT_SYMBOL_GPL(gfn_to_hva_memslot);

1682 1683
unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
{
1684
	return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL);
1685
}
1686
EXPORT_SYMBOL_GPL(gfn_to_hva);
I
Izik Eidus 已提交
1687

1688 1689 1690 1691 1692 1693
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);

1694
/*
1695 1696 1697 1698 1699 1700
 * 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
1701
 */
1702 1703
unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot,
				      gfn_t gfn, bool *writable)
1704
{
1705 1706 1707
	unsigned long hva = __gfn_to_hva_many(slot, gfn, NULL, false);

	if (!kvm_is_error_hva(hva) && writable)
1708 1709
		*writable = !memslot_is_readonly(slot);

1710
	return hva;
1711 1712
}

1713 1714 1715 1716 1717 1718 1719
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);
}

1720 1721 1722 1723 1724 1725 1726
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);
}

1727 1728
static inline int check_user_page_hwpoison(unsigned long addr)
{
L
Lorenzo Stoakes 已提交
1729
	int rc, flags = FOLL_HWPOISON | FOLL_WRITE;
1730

L
Lorenzo Stoakes 已提交
1731
	rc = get_user_pages(addr, 1, flags, NULL, NULL);
1732 1733 1734
	return rc == -EHWPOISON;
}

X
Xiao Guangrong 已提交
1735
/*
1736 1737
 * 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
1738
 * only part that runs if we can in atomic context.
X
Xiao Guangrong 已提交
1739
 */
1740 1741
static bool hva_to_pfn_fast(unsigned long addr, bool write_fault,
			    bool *writable, kvm_pfn_t *pfn)
A
Avi Kivity 已提交
1742
{
1743
	struct page *page[1];
X
Xiao Guangrong 已提交
1744
	int npages;
A
Avi Kivity 已提交
1745

1746 1747 1748 1749 1750 1751 1752
	/*
	 * 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;
1753

X
Xiao Guangrong 已提交
1754 1755 1756
	npages = __get_user_pages_fast(addr, 1, 1, page);
	if (npages == 1) {
		*pfn = page_to_pfn(page[0]);
1757

X
Xiao Guangrong 已提交
1758 1759 1760 1761
		if (writable)
			*writable = true;
		return true;
	}
1762

X
Xiao Guangrong 已提交
1763 1764
	return false;
}
1765

X
Xiao Guangrong 已提交
1766 1767 1768 1769 1770
/*
 * 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
1771
			   bool *writable, kvm_pfn_t *pfn)
X
Xiao Guangrong 已提交
1772
{
1773 1774
	unsigned int flags = FOLL_HWPOISON;
	struct page *page;
X
Xiao Guangrong 已提交
1775
	int npages = 0;
1776

X
Xiao Guangrong 已提交
1777 1778 1779 1780 1781
	might_sleep();

	if (writable)
		*writable = write_fault;

1782 1783 1784 1785
	if (write_fault)
		flags |= FOLL_WRITE;
	if (async)
		flags |= FOLL_NOWAIT;
1786

1787
	npages = get_user_pages_unlocked(addr, 1, &page, flags);
X
Xiao Guangrong 已提交
1788 1789 1790 1791
	if (npages != 1)
		return npages;

	/* map read fault as writable if possible */
1792
	if (unlikely(!write_fault) && writable) {
1793
		struct page *wpage;
X
Xiao Guangrong 已提交
1794

1795
		if (__get_user_pages_fast(addr, 1, 1, &wpage) == 1) {
X
Xiao Guangrong 已提交
1796
			*writable = true;
1797 1798
			put_page(page);
			page = wpage;
1799
		}
1800
	}
1801
	*pfn = page_to_pfn(page);
X
Xiao Guangrong 已提交
1802 1803
	return npages;
}
I
Izik Eidus 已提交
1804

X
Xiao Guangrong 已提交
1805 1806 1807 1808
static bool vma_is_valid(struct vm_area_struct *vma, bool write_fault)
{
	if (unlikely(!(vma->vm_flags & VM_READ)))
		return false;
1809

X
Xiao Guangrong 已提交
1810 1811
	if (write_fault && (unlikely(!(vma->vm_flags & VM_WRITE))))
		return false;
1812

X
Xiao Guangrong 已提交
1813 1814
	return true;
}
1815

1816 1817
static int hva_to_pfn_remapped(struct vm_area_struct *vma,
			       unsigned long addr, bool *async,
1818 1819
			       bool write_fault, bool *writable,
			       kvm_pfn_t *p_pfn)
1820
{
1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844
	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;

	}

1845 1846
	if (writable)
		*writable = true;
1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861

	/*
	 * 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;
1862 1863 1864
	return 0;
}

1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878
/*
 * 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
1879
static kvm_pfn_t hva_to_pfn(unsigned long addr, bool atomic, bool *async,
X
Xiao Guangrong 已提交
1880 1881 1882
			bool write_fault, bool *writable)
{
	struct vm_area_struct *vma;
D
Dan Williams 已提交
1883
	kvm_pfn_t pfn = 0;
1884
	int npages, r;
1885

X
Xiao Guangrong 已提交
1886 1887
	/* we can do it either atomically or asynchronously, not both */
	BUG_ON(atomic && async);
1888

1889
	if (hva_to_pfn_fast(addr, write_fault, writable, &pfn))
X
Xiao Guangrong 已提交
1890 1891 1892 1893 1894 1895 1896 1897
		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;
1898

X
Xiao Guangrong 已提交
1899 1900 1901 1902 1903 1904 1905
	down_read(&current->mm->mmap_sem);
	if (npages == -EHWPOISON ||
	      (!async && check_user_page_hwpoison(addr))) {
		pfn = KVM_PFN_ERR_HWPOISON;
		goto exit;
	}

1906
retry:
X
Xiao Guangrong 已提交
1907 1908 1909 1910
	vma = find_vma_intersection(current->mm, addr, addr + 1);

	if (vma == NULL)
		pfn = KVM_PFN_ERR_FAULT;
1911
	else if (vma->vm_flags & (VM_IO | VM_PFNMAP)) {
1912
		r = hva_to_pfn_remapped(vma, addr, async, write_fault, writable, &pfn);
1913 1914
		if (r == -EAGAIN)
			goto retry;
1915 1916
		if (r < 0)
			pfn = KVM_PFN_ERR_FAULT;
X
Xiao Guangrong 已提交
1917
	} else {
X
Xiao Guangrong 已提交
1918
		if (async && vma_is_valid(vma, write_fault))
X
Xiao Guangrong 已提交
1919 1920 1921 1922 1923
			*async = true;
		pfn = KVM_PFN_ERR_FAULT;
	}
exit:
	up_read(&current->mm->mmap_sem);
1924
	return pfn;
1925 1926
}

D
Dan Williams 已提交
1927 1928 1929
kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
			       bool atomic, bool *async, bool write_fault,
			       bool *writable)
1930
{
X
Xiao Guangrong 已提交
1931 1932
	unsigned long addr = __gfn_to_hva_many(slot, gfn, NULL, write_fault);

1933 1934 1935
	if (addr == KVM_HVA_ERR_RO_BAD) {
		if (writable)
			*writable = false;
X
Xiao Guangrong 已提交
1936
		return KVM_PFN_ERR_RO_FAULT;
1937
	}
X
Xiao Guangrong 已提交
1938

1939 1940 1941
	if (kvm_is_error_hva(addr)) {
		if (writable)
			*writable = false;
1942
		return KVM_PFN_NOSLOT;
1943
	}
X
Xiao Guangrong 已提交
1944 1945 1946 1947 1948 1949 1950 1951 1952

	/* 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);
1953
}
1954
EXPORT_SYMBOL_GPL(__gfn_to_pfn_memslot);
1955

D
Dan Williams 已提交
1956
kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1957 1958
		      bool *writable)
{
P
Paolo Bonzini 已提交
1959 1960
	return __gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn, false, NULL,
				    write_fault, writable);
1961 1962 1963
}
EXPORT_SYMBOL_GPL(gfn_to_pfn_prot);

D
Dan Williams 已提交
1964
kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1965
{
X
Xiao Guangrong 已提交
1966
	return __gfn_to_pfn_memslot(slot, gfn, false, NULL, true, NULL);
1967
}
P
Paolo Bonzini 已提交
1968
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot);
1969

D
Dan Williams 已提交
1970
kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn)
1971
{
X
Xiao Guangrong 已提交
1972
	return __gfn_to_pfn_memslot(slot, gfn, true, NULL, true, NULL);
1973
}
1974
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot_atomic);
1975

D
Dan Williams 已提交
1976
kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1977 1978 1979 1980 1981
{
	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
1982
kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn)
1983 1984 1985 1986 1987
{
	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
1988
kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1989 1990 1991 1992 1993
{
	return gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn);
}
EXPORT_SYMBOL_GPL(gfn_to_pfn);

D
Dan Williams 已提交
1994
kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn)
1995 1996 1997 1998 1999
{
	return gfn_to_pfn_memslot(kvm_vcpu_gfn_to_memslot(vcpu, gfn), gfn);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_gfn_to_pfn);

2000 2001
int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
			    struct page **pages, int nr_pages)
2002 2003
{
	unsigned long addr;
2004
	gfn_t entry = 0;
2005

2006
	addr = gfn_to_hva_many(slot, gfn, &entry);
2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
	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
2017
static struct page *kvm_pfn_to_page(kvm_pfn_t pfn)
X
Xiao Guangrong 已提交
2018
{
2019
	if (is_error_noslot_pfn(pfn))
2020
		return KVM_ERR_PTR_BAD_PAGE;
X
Xiao Guangrong 已提交
2021

2022
	if (kvm_is_reserved_pfn(pfn)) {
2023
		WARN_ON(1);
2024
		return KVM_ERR_PTR_BAD_PAGE;
2025
	}
X
Xiao Guangrong 已提交
2026 2027 2028 2029

	return pfn_to_page(pfn);
}

2030 2031
struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
{
D
Dan Williams 已提交
2032
	kvm_pfn_t pfn;
2033 2034 2035

	pfn = gfn_to_pfn(kvm, gfn);

X
Xiao Guangrong 已提交
2036
	return kvm_pfn_to_page(pfn);
A
Avi Kivity 已提交
2037 2038 2039
}
EXPORT_SYMBOL_GPL(gfn_to_page);

2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064
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;
}

2065
static int __kvm_map_gfn(struct kvm_memslots *slots, gfn_t gfn,
2066 2067 2068
			 struct kvm_host_map *map,
			 struct gfn_to_pfn_cache *cache,
			 bool atomic)
2069 2070 2071 2072
{
	kvm_pfn_t pfn;
	void *hva = NULL;
	struct page *page = KVM_UNMAPPED_PAGE;
2073
	struct kvm_memory_slot *slot = __gfn_to_memslot(slots, gfn);
2074
	u64 gen = slots->generation;
2075 2076 2077 2078

	if (!map)
		return -EINVAL;

2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091
	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);
	}
2092 2093 2094 2095 2096
	if (is_error_noslot_pfn(pfn))
		return -EINVAL;

	if (pfn_valid(pfn)) {
		page = pfn_to_page(pfn);
2097 2098 2099 2100
		if (atomic)
			hva = kmap_atomic(page);
		else
			hva = kmap(page);
P
Paolo Bonzini 已提交
2101
#ifdef CONFIG_HAS_IOMEM
2102
	} else if (!atomic) {
2103
		hva = memremap(pfn_to_hpa(pfn), PAGE_SIZE, MEMREMAP_WB);
2104 2105
	} else {
		return -EINVAL;
P
Paolo Bonzini 已提交
2106
#endif
2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119
	}

	if (!hva)
		return -EFAULT;

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

	return 0;
}

2120 2121
int kvm_map_gfn(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map,
		struct gfn_to_pfn_cache *cache, bool atomic)
2122
{
2123 2124
	return __kvm_map_gfn(kvm_memslots(vcpu->kvm), gfn, map,
			cache, atomic);
2125 2126 2127
}
EXPORT_SYMBOL_GPL(kvm_map_gfn);

2128 2129
int kvm_vcpu_map(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map)
{
2130 2131
	return __kvm_map_gfn(kvm_vcpu_memslots(vcpu), gfn, map,
		NULL, false);
2132 2133 2134
}
EXPORT_SYMBOL_GPL(kvm_vcpu_map);

2135
static void __kvm_unmap_gfn(struct kvm_memory_slot *memslot,
2136 2137 2138
			struct kvm_host_map *map,
			struct gfn_to_pfn_cache *cache,
			bool dirty, bool atomic)
2139 2140 2141 2142 2143 2144 2145
{
	if (!map)
		return;

	if (!map->hva)
		return;

2146 2147 2148 2149 2150 2151
	if (map->page != KVM_UNMAPPED_PAGE) {
		if (atomic)
			kunmap_atomic(map->hva);
		else
			kunmap(map->page);
	}
2152
#ifdef CONFIG_HAS_IOMEM
2153
	else if (!atomic)
2154
		memunmap(map->hva);
2155 2156
	else
		WARN_ONCE(1, "Unexpected unmapping in atomic context");
2157
#endif
2158

2159
	if (dirty)
2160
		mark_page_dirty_in_slot(memslot, map->gfn);
2161 2162 2163 2164 2165

	if (cache)
		cache->dirty |= dirty;
	else
		kvm_release_pfn(map->pfn, dirty, NULL);
2166 2167 2168 2169

	map->hva = NULL;
	map->page = NULL;
}
2170

2171 2172
int kvm_unmap_gfn(struct kvm_vcpu *vcpu, struct kvm_host_map *map, 
		  struct gfn_to_pfn_cache *cache, bool dirty, bool atomic)
2173
{
2174 2175
	__kvm_unmap_gfn(gfn_to_memslot(vcpu->kvm, map->gfn), map,
			cache, dirty, atomic);
2176 2177 2178 2179 2180 2181
	return 0;
}
EXPORT_SYMBOL_GPL(kvm_unmap_gfn);

void kvm_vcpu_unmap(struct kvm_vcpu *vcpu, struct kvm_host_map *map, bool dirty)
{
2182 2183
	__kvm_unmap_gfn(kvm_vcpu_gfn_to_memslot(vcpu, map->gfn), map, NULL,
			dirty, false);
2184
}
2185 2186
EXPORT_SYMBOL_GPL(kvm_vcpu_unmap);

2187 2188
struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn)
{
D
Dan Williams 已提交
2189
	kvm_pfn_t pfn;
2190 2191 2192 2193 2194 2195 2196

	pfn = kvm_vcpu_gfn_to_pfn(vcpu, gfn);

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

2197 2198
void kvm_release_page_clean(struct page *page)
{
2199 2200
	WARN_ON(is_error_page(page));

2201
	kvm_release_pfn_clean(page_to_pfn(page));
2202 2203 2204
}
EXPORT_SYMBOL_GPL(kvm_release_page_clean);

D
Dan Williams 已提交
2205
void kvm_release_pfn_clean(kvm_pfn_t pfn)
2206
{
2207
	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn))
2208
		put_page(pfn_to_page(pfn));
2209 2210 2211
}
EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);

2212
void kvm_release_page_dirty(struct page *page)
2213
{
X
Xiao Guangrong 已提交
2214 2215
	WARN_ON(is_error_page(page));

2216 2217 2218 2219
	kvm_release_pfn_dirty(page_to_pfn(page));
}
EXPORT_SYMBOL_GPL(kvm_release_page_dirty);

2220
void kvm_release_pfn_dirty(kvm_pfn_t pfn)
2221 2222 2223 2224
{
	kvm_set_pfn_dirty(pfn);
	kvm_release_pfn_clean(pfn);
}
2225
EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
2226

D
Dan Williams 已提交
2227
void kvm_set_pfn_dirty(kvm_pfn_t pfn)
2228
{
2229 2230
	if (!kvm_is_reserved_pfn(pfn) && !kvm_is_zone_device_pfn(pfn))
		SetPageDirty(pfn_to_page(pfn));
2231
}
2232 2233
EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);

D
Dan Williams 已提交
2234
void kvm_set_pfn_accessed(kvm_pfn_t pfn)
2235
{
2236
	if (!kvm_is_reserved_pfn(pfn) && !kvm_is_zone_device_pfn(pfn))
2237
		mark_page_accessed(pfn_to_page(pfn));
2238 2239 2240
}
EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);

D
Dan Williams 已提交
2241
void kvm_get_pfn(kvm_pfn_t pfn)
2242
{
2243
	if (!kvm_is_reserved_pfn(pfn))
2244
		get_page(pfn_to_page(pfn));
2245 2246
}
EXPORT_SYMBOL_GPL(kvm_get_pfn);
2247

2248 2249 2250 2251 2252 2253 2254 2255
static int next_segment(unsigned long len, int offset)
{
	if (len > PAGE_SIZE - offset)
		return PAGE_SIZE - offset;
	else
		return len;
}

2256 2257
static int __kvm_read_guest_page(struct kvm_memory_slot *slot, gfn_t gfn,
				 void *data, int offset, int len)
2258
{
2259 2260
	int r;
	unsigned long addr;
2261

2262
	addr = gfn_to_hva_memslot_prot(slot, gfn, NULL);
2263 2264
	if (kvm_is_error_hva(addr))
		return -EFAULT;
2265
	r = __copy_from_user(data, (void __user *)addr + offset, len);
2266
	if (r)
2267 2268 2269
		return -EFAULT;
	return 0;
}
2270 2271 2272 2273 2274 2275 2276 2277

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);
}
2278 2279
EXPORT_SYMBOL_GPL(kvm_read_guest_page);

2280 2281 2282 2283 2284 2285 2286 2287 2288
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);

2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308
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);

2309
int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data, unsigned long len)
2310 2311
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
2312
	int seg;
2313
	int offset = offset_in_page(gpa);
2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327
	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);
2328

2329 2330 2331 2332 2333 2334 2335
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);
2336 2337
	if (kvm_is_error_hva(addr))
		return -EFAULT;
2338
	pagefault_disable();
2339
	r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
2340
	pagefault_enable();
2341 2342 2343 2344 2345
	if (r)
		return -EFAULT;
	return 0;
}

2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358
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)
2359
{
2360 2361
	int r;
	unsigned long addr;
2362

2363
	addr = gfn_to_hva_memslot(memslot, gfn);
2364 2365
	if (kvm_is_error_hva(addr))
		return -EFAULT;
2366
	r = __copy_to_user((void __user *)addr + offset, data, len);
2367
	if (r)
2368
		return -EFAULT;
2369
	mark_page_dirty_in_slot(memslot, gfn);
2370 2371
	return 0;
}
2372 2373 2374 2375 2376 2377 2378 2379

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);
}
2380 2381
EXPORT_SYMBOL_GPL(kvm_write_guest_page);

2382 2383 2384 2385 2386 2387 2388 2389 2390
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);

2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409
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;
}
2410
EXPORT_SYMBOL_GPL(kvm_write_guest);
2411

2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432
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);

2433 2434 2435
static int __kvm_gfn_to_hva_cache_init(struct kvm_memslots *slots,
				       struct gfn_to_hva_cache *ghc,
				       gpa_t gpa, unsigned long len)
2436 2437
{
	int offset = offset_in_page(gpa);
2438 2439 2440 2441
	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;
2442

2443
	/* Update ghc->generation before performing any error checks. */
2444
	ghc->generation = slots->generation;
2445 2446 2447 2448 2449

	if (start_gfn > end_gfn) {
		ghc->hva = KVM_HVA_ERR_BAD;
		return -EINVAL;
	}
2450 2451 2452 2453 2454

	/*
	 * If the requested region crosses two memslots, we still
	 * verify that the entire region is valid here.
	 */
2455
	for ( ; start_gfn <= end_gfn; start_gfn += nr_pages_avail) {
2456 2457 2458 2459
		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))
2460
			return -EFAULT;
2461 2462 2463
	}

	/* Use the slow path for cross page reads and writes. */
2464
	if (nr_pages_needed == 1)
2465
		ghc->hva += offset;
2466
	else
2467
		ghc->memslot = NULL;
2468

2469 2470 2471
	ghc->gpa = gpa;
	ghc->len = len;
	return 0;
2472
}
2473

2474
int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
2475 2476
			      gpa_t gpa, unsigned long len)
{
2477
	struct kvm_memslots *slots = kvm_memslots(kvm);
2478 2479
	return __kvm_gfn_to_hva_cache_init(slots, ghc, gpa, len);
}
2480
EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init);
2481

2482
int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
2483 2484
				  void *data, unsigned int offset,
				  unsigned long len)
2485
{
2486
	struct kvm_memslots *slots = kvm_memslots(kvm);
2487
	int r;
2488
	gpa_t gpa = ghc->gpa + offset;
2489

2490
	BUG_ON(len + offset > ghc->len);
2491

2492 2493 2494 2495
	if (slots->generation != ghc->generation) {
		if (__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len))
			return -EFAULT;
	}
2496

2497 2498 2499
	if (kvm_is_error_hva(ghc->hva))
		return -EFAULT;

2500 2501 2502
	if (unlikely(!ghc->memslot))
		return kvm_write_guest(kvm, gpa, data, len);

2503
	r = __copy_to_user((void __user *)ghc->hva + offset, data, len);
2504 2505
	if (r)
		return -EFAULT;
2506
	mark_page_dirty_in_slot(ghc->memslot, gpa >> PAGE_SHIFT);
2507 2508 2509

	return 0;
}
2510
EXPORT_SYMBOL_GPL(kvm_write_guest_offset_cached);
2511

2512 2513
int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
2514
{
2515
	return kvm_write_guest_offset_cached(kvm, ghc, data, 0, len);
2516
}
2517
EXPORT_SYMBOL_GPL(kvm_write_guest_cached);
2518

2519 2520
int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
2521
{
2522
	struct kvm_memslots *slots = kvm_memslots(kvm);
2523 2524
	int r;

2525 2526
	BUG_ON(len > ghc->len);

2527 2528 2529 2530
	if (slots->generation != ghc->generation) {
		if (__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len))
			return -EFAULT;
	}
2531

2532 2533 2534
	if (kvm_is_error_hva(ghc->hva))
		return -EFAULT;

2535 2536 2537
	if (unlikely(!ghc->memslot))
		return kvm_read_guest(kvm, ghc->gpa, data, len);

2538 2539 2540 2541 2542 2543
	r = __copy_from_user(data, (void __user *)ghc->hva, len);
	if (r)
		return -EFAULT;

	return 0;
}
2544
EXPORT_SYMBOL_GPL(kvm_read_guest_cached);
2545

2546 2547
int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
{
2548 2549 2550
	const void *zero_page = (const void *) __va(page_to_phys(ZERO_PAGE(0)));

	return kvm_write_guest_page(kvm, gfn, zero_page, offset, len);
2551 2552 2553 2554 2555 2556 2557 2558 2559 2560
}
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;

2561
	while ((seg = next_segment(len, offset)) != 0) {
2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572
		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);

2573
static void mark_page_dirty_in_slot(struct kvm_memory_slot *memslot,
2574
				    gfn_t gfn)
A
Avi Kivity 已提交
2575
{
R
Rusty Russell 已提交
2576 2577
	if (memslot && memslot->dirty_bitmap) {
		unsigned long rel_gfn = gfn - memslot->base_gfn;
A
Avi Kivity 已提交
2578

2579
		set_bit_le(rel_gfn, memslot->dirty_bitmap);
A
Avi Kivity 已提交
2580 2581 2582
	}
}

2583 2584 2585 2586 2587
void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
{
	struct kvm_memory_slot *memslot;

	memslot = gfn_to_memslot(kvm, gfn);
2588
	mark_page_dirty_in_slot(memslot, gfn);
2589
}
2590
EXPORT_SYMBOL_GPL(mark_page_dirty);
2591

2592 2593 2594 2595 2596 2597 2598 2599 2600
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);

2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623
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
2624 2625
static void grow_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2626
	unsigned int old, val, grow, grow_start;
W
Wanpeng Li 已提交
2627

2628
	old = val = vcpu->halt_poll_ns;
2629
	grow_start = READ_ONCE(halt_poll_ns_grow_start);
2630
	grow = READ_ONCE(halt_poll_ns_grow);
2631 2632 2633
	if (!grow)
		goto out;

2634 2635 2636
	val *= grow;
	if (val < grow_start)
		val = grow_start;
W
Wanpeng Li 已提交
2637

2638 2639 2640
	if (val > halt_poll_ns)
		val = halt_poll_ns;

W
Wanpeng Li 已提交
2641
	vcpu->halt_poll_ns = val;
2642
out:
2643
	trace_kvm_halt_poll_ns_grow(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2644 2645 2646 2647
}

static void shrink_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2648
	unsigned int old, val, shrink;
W
Wanpeng Li 已提交
2649

2650
	old = val = vcpu->halt_poll_ns;
2651 2652
	shrink = READ_ONCE(halt_poll_ns_shrink);
	if (shrink == 0)
W
Wanpeng Li 已提交
2653 2654
		val = 0;
	else
2655
		val /= shrink;
W
Wanpeng Li 已提交
2656 2657

	vcpu->halt_poll_ns = val;
2658
	trace_kvm_halt_poll_ns_shrink(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2659 2660
}

2661 2662
static int kvm_vcpu_check_block(struct kvm_vcpu *vcpu)
{
2663 2664 2665
	int ret = -EINTR;
	int idx = srcu_read_lock(&vcpu->kvm->srcu);

2666 2667
	if (kvm_arch_vcpu_runnable(vcpu)) {
		kvm_make_request(KVM_REQ_UNHALT, vcpu);
2668
		goto out;
2669 2670
	}
	if (kvm_cpu_has_pending_timer(vcpu))
2671
		goto out;
2672
	if (signal_pending(current))
2673
		goto out;
2674

2675 2676 2677 2678
	ret = 0;
out:
	srcu_read_unlock(&vcpu->kvm->srcu, idx);
	return ret;
2679 2680
}

E
Eddie Dong 已提交
2681 2682 2683
/*
 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
 */
2684
void kvm_vcpu_block(struct kvm_vcpu *vcpu)
2685
{
2686
	ktime_t start, cur;
2687
	DECLARE_SWAITQUEUE(wait);
2688
	bool waited = false;
W
Wanpeng Li 已提交
2689
	u64 block_ns;
2690

2691 2692
	kvm_arch_vcpu_blocking(vcpu);

2693
	start = cur = ktime_get();
2694
	if (vcpu->halt_poll_ns && !kvm_arch_no_poll(vcpu)) {
W
Wanpeng Li 已提交
2695
		ktime_t stop = ktime_add_ns(ktime_get(), vcpu->halt_poll_ns);
2696

2697
		++vcpu->stat.halt_attempted_poll;
2698 2699 2700 2701 2702 2703 2704
		do {
			/*
			 * This sets KVM_REQ_UNHALT if an interrupt
			 * arrives.
			 */
			if (kvm_vcpu_check_block(vcpu) < 0) {
				++vcpu->stat.halt_successful_poll;
2705 2706
				if (!vcpu_valid_wakeup(vcpu))
					++vcpu->stat.halt_poll_invalid;
2707 2708 2709 2710 2711
				goto out;
			}
			cur = ktime_get();
		} while (single_task_running() && ktime_before(cur, stop));
	}
2712 2713

	for (;;) {
2714
		prepare_to_swait_exclusive(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
2715

2716
		if (kvm_vcpu_check_block(vcpu) < 0)
2717 2718
			break;

2719
		waited = true;
E
Eddie Dong 已提交
2720 2721
		schedule();
	}
2722

2723
	finish_swait(&vcpu->wq, &wait);
2724 2725
	cur = ktime_get();
out:
2726
	kvm_arch_vcpu_unblocking(vcpu);
W
Wanpeng Li 已提交
2727 2728
	block_ns = ktime_to_ns(cur) - ktime_to_ns(start);

2729 2730
	if (!kvm_arch_no_poll(vcpu)) {
		if (!vcpu_valid_wakeup(vcpu)) {
W
Wanpeng Li 已提交
2731
			shrink_halt_poll_ns(vcpu);
2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745
		} 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
2746

2747 2748
	trace_kvm_vcpu_wakeup(block_ns, waited, vcpu_valid_wakeup(vcpu));
	kvm_arch_vcpu_block_finish(vcpu);
E
Eddie Dong 已提交
2749
}
2750
EXPORT_SYMBOL_GPL(kvm_vcpu_block);
E
Eddie Dong 已提交
2751

2752
bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu)
2753
{
2754
	struct swait_queue_head *wqp;
2755 2756

	wqp = kvm_arch_vcpu_wq(vcpu);
2757
	if (swq_has_sleeper(wqp)) {
2758
		swake_up_one(wqp);
2759
		WRITE_ONCE(vcpu->ready, true);
2760
		++vcpu->stat.halt_wakeup;
2761
		return true;
2762 2763
	}

2764
	return false;
2765 2766 2767
}
EXPORT_SYMBOL_GPL(kvm_vcpu_wake_up);

2768
#ifndef CONFIG_S390
2769 2770 2771 2772 2773 2774 2775 2776
/*
 * 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;

2777 2778 2779
	if (kvm_vcpu_wake_up(vcpu))
		return;

2780 2781 2782 2783 2784 2785
	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();
}
2786
EXPORT_SYMBOL_GPL(kvm_vcpu_kick);
2787
#endif /* !CONFIG_S390 */
2788

2789
int kvm_vcpu_yield_to(struct kvm_vcpu *target)
2790 2791 2792
{
	struct pid *pid;
	struct task_struct *task = NULL;
2793
	int ret = 0;
2794 2795 2796 2797

	rcu_read_lock();
	pid = rcu_dereference(target->pid);
	if (pid)
2798
		task = get_pid_task(pid, PIDTYPE_PID);
2799 2800
	rcu_read_unlock();
	if (!task)
2801 2802
		return ret;
	ret = yield_to(task, 1);
2803
	put_task_struct(task);
2804 2805

	return ret;
2806 2807 2808
}
EXPORT_SYMBOL_GPL(kvm_vcpu_yield_to);

2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830
/*
 * 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.
 */
2831
static bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
2832
{
2833
#ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
2834 2835 2836
	bool eligible;

	eligible = !vcpu->spin_loop.in_spin_loop ||
2837
		    vcpu->spin_loop.dy_eligible;
2838 2839 2840 2841 2842

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

	return eligible;
2843 2844
#else
	return true;
2845
#endif
2846
}
2847

2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870
/*
 * 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;
}

2871
void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
Z
Zhai, Edwin 已提交
2872
{
2873 2874 2875 2876
	struct kvm *kvm = me->kvm;
	struct kvm_vcpu *vcpu;
	int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
	int yielded = 0;
2877
	int try = 3;
2878 2879
	int pass;
	int i;
Z
Zhai, Edwin 已提交
2880

2881
	kvm_vcpu_set_in_spin_loop(me, true);
2882 2883 2884 2885 2886 2887 2888
	/*
	 * 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.
	 */
2889
	for (pass = 0; pass < 2 && !yielded && try; pass++) {
2890
		kvm_for_each_vcpu(i, vcpu, kvm) {
2891
			if (!pass && i <= last_boosted_vcpu) {
2892 2893 2894 2895
				i = last_boosted_vcpu;
				continue;
			} else if (pass && i > last_boosted_vcpu)
				break;
2896
			if (!READ_ONCE(vcpu->ready))
2897
				continue;
2898 2899
			if (vcpu == me)
				continue;
2900
			if (swait_active(&vcpu->wq) && !vcpu_dy_runnable(vcpu))
2901
				continue;
2902 2903
			if (READ_ONCE(vcpu->preempted) && yield_to_kernel_mode &&
				!kvm_arch_vcpu_in_kernel(vcpu))
2904
				continue;
2905 2906
			if (!kvm_vcpu_eligible_for_directed_yield(vcpu))
				continue;
2907 2908 2909

			yielded = kvm_vcpu_yield_to(vcpu);
			if (yielded > 0) {
2910 2911
				kvm->last_boosted_vcpu = i;
				break;
2912 2913 2914 2915
			} else if (yielded < 0) {
				try--;
				if (!try)
					break;
2916 2917 2918
			}
		}
	}
2919
	kvm_vcpu_set_in_spin_loop(me, false);
2920 2921 2922

	/* Ensure vcpu is not eligible during next spinloop */
	kvm_vcpu_set_dy_eligible(me, false);
Z
Zhai, Edwin 已提交
2923 2924 2925
}
EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin);

2926
static vm_fault_t kvm_vcpu_fault(struct vm_fault *vmf)
2927
{
2928
	struct kvm_vcpu *vcpu = vmf->vma->vm_file->private_data;
2929 2930
	struct page *page;

2931
	if (vmf->pgoff == 0)
2932
		page = virt_to_page(vcpu->run);
A
Avi Kivity 已提交
2933
#ifdef CONFIG_X86
2934
	else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
2935
		page = virt_to_page(vcpu->arch.pio_data);
2936
#endif
2937
#ifdef CONFIG_KVM_MMIO
2938 2939
	else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
		page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
A
Avi Kivity 已提交
2940
#endif
2941
	else
2942
		return kvm_arch_vcpu_fault(vcpu, vmf);
2943
	get_page(page);
2944 2945
	vmf->page = page;
	return 0;
2946 2947
}

2948
static const struct vm_operations_struct kvm_vcpu_vm_ops = {
2949
	.fault = kvm_vcpu_fault,
2950 2951 2952 2953 2954 2955 2956 2957
};

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
2958 2959 2960 2961
static int kvm_vcpu_release(struct inode *inode, struct file *filp)
{
	struct kvm_vcpu *vcpu = filp->private_data;

2962
	debugfs_remove_recursive(vcpu->debugfs_dentry);
A
Al Viro 已提交
2963
	kvm_put_kvm(vcpu->kvm);
A
Avi Kivity 已提交
2964 2965 2966
	return 0;
}

2967
static struct file_operations kvm_vcpu_fops = {
A
Avi Kivity 已提交
2968 2969
	.release        = kvm_vcpu_release,
	.unlocked_ioctl = kvm_vcpu_ioctl,
2970
	.mmap           = kvm_vcpu_mmap,
2971
	.llseek		= noop_llseek,
2972
	KVM_COMPAT(kvm_vcpu_compat_ioctl),
A
Avi Kivity 已提交
2973 2974 2975 2976 2977 2978 2979
};

/*
 * Allocates an inode for the vcpu.
 */
static int create_vcpu_fd(struct kvm_vcpu *vcpu)
{
2980 2981 2982 2983
	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
2984 2985
}

2986
static void kvm_create_vcpu_debugfs(struct kvm_vcpu *vcpu)
2987
{
2988
#ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
2989 2990 2991
	char dir_name[ITOA_MAX_LEN * 2];

	if (!debugfs_initialized())
2992
		return;
2993 2994 2995

	snprintf(dir_name, sizeof(dir_name), "vcpu%d", vcpu->vcpu_id);
	vcpu->debugfs_dentry = debugfs_create_dir(dir_name,
2996
						  vcpu->kvm->debugfs_dentry);
2997

2998
	kvm_arch_create_vcpu_debugfs(vcpu);
2999
#endif
3000 3001
}

3002 3003 3004
/*
 * Creates some virtual cpus.  Good luck creating more than one.
 */
3005
static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
3006 3007
{
	int r;
3008
	struct kvm_vcpu *vcpu;
3009
	struct page *page;
3010

G
Greg Kurz 已提交
3011
	if (id >= KVM_MAX_VCPU_ID)
3012 3013
		return -EINVAL;

3014 3015 3016 3017 3018 3019 3020 3021 3022
	mutex_lock(&kvm->lock);
	if (kvm->created_vcpus == KVM_MAX_VCPUS) {
		mutex_unlock(&kvm->lock);
		return -EINVAL;
	}

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

3023 3024 3025 3026
	r = kvm_arch_vcpu_precreate(kvm, id);
	if (r)
		goto vcpu_decrement;

3027 3028 3029
	vcpu = kmem_cache_zalloc(kvm_vcpu_cache, GFP_KERNEL);
	if (!vcpu) {
		r = -ENOMEM;
3030 3031
		goto vcpu_decrement;
	}
3032

3033
	BUILD_BUG_ON(sizeof(struct kvm_run) > PAGE_SIZE);
3034 3035 3036
	page = alloc_page(GFP_KERNEL | __GFP_ZERO);
	if (!page) {
		r = -ENOMEM;
3037
		goto vcpu_free;
3038 3039 3040 3041
	}
	vcpu->run = page_address(page);

	kvm_vcpu_init(vcpu, kvm, id);
3042 3043 3044

	r = kvm_arch_vcpu_create(vcpu);
	if (r)
3045
		goto vcpu_free_run_page;
3046

3047
	kvm_create_vcpu_debugfs(vcpu);
3048

S
Shaohua Li 已提交
3049
	mutex_lock(&kvm->lock);
3050 3051 3052 3053
	if (kvm_get_vcpu_by_id(kvm, id)) {
		r = -EEXIST;
		goto unlock_vcpu_destroy;
	}
3054

3055 3056
	vcpu->vcpu_idx = atomic_read(&kvm->online_vcpus);
	BUG_ON(kvm->vcpus[vcpu->vcpu_idx]);
3057

R
Rusty Russell 已提交
3058
	/* Now it's all set up, let userspace reach it */
A
Al Viro 已提交
3059
	kvm_get_kvm(kvm);
A
Avi Kivity 已提交
3060
	r = create_vcpu_fd(vcpu);
3061
	if (r < 0) {
3062
		kvm_put_kvm_no_destroy(kvm);
3063
		goto unlock_vcpu_destroy;
3064 3065
	}

3066
	kvm->vcpus[vcpu->vcpu_idx] = vcpu;
3067 3068 3069 3070 3071

	/*
	 * Pairs with smp_rmb() in kvm_get_vcpu.  Write kvm->vcpus
	 * before kvm->online_vcpu's incremented value.
	 */
3072 3073 3074 3075
	smp_wmb();
	atomic_inc(&kvm->online_vcpus);

	mutex_unlock(&kvm->lock);
3076
	kvm_arch_vcpu_postcreate(vcpu);
R
Rusty Russell 已提交
3077
	return r;
3078

3079
unlock_vcpu_destroy:
3080
	mutex_unlock(&kvm->lock);
3081
	debugfs_remove_recursive(vcpu->debugfs_dentry);
3082
	kvm_arch_vcpu_destroy(vcpu);
3083 3084
vcpu_free_run_page:
	free_page((unsigned long)vcpu->run);
3085 3086
vcpu_free:
	kmem_cache_free(kvm_vcpu_cache, vcpu);
3087 3088 3089 3090
vcpu_decrement:
	mutex_lock(&kvm->lock);
	kvm->created_vcpus--;
	mutex_unlock(&kvm->lock);
3091 3092 3093
	return r;
}

A
Avi Kivity 已提交
3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104
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
3105 3106
static long kvm_vcpu_ioctl(struct file *filp,
			   unsigned int ioctl, unsigned long arg)
A
Avi Kivity 已提交
3107
{
A
Avi Kivity 已提交
3108
	struct kvm_vcpu *vcpu = filp->private_data;
A
Al Viro 已提交
3109
	void __user *argp = (void __user *)arg;
3110
	int r;
3111 3112
	struct kvm_fpu *fpu = NULL;
	struct kvm_sregs *kvm_sregs = NULL;
A
Avi Kivity 已提交
3113

3114 3115
	if (vcpu->kvm->mm != current->mm)
		return -EIO;
3116

3117 3118 3119
	if (unlikely(_IOC_TYPE(ioctl) != KVMIO))
		return -EINVAL;

3120
	/*
3121 3122
	 * Some architectures have vcpu ioctls that are asynchronous to vcpu
	 * execution; mutex_lock() would break them.
3123
	 */
3124 3125
	r = kvm_arch_vcpu_async_ioctl(filp, ioctl, arg);
	if (r != -ENOIOCTLCMD)
3126
		return r;
3127

3128 3129
	if (mutex_lock_killable(&vcpu->mutex))
		return -EINTR;
A
Avi Kivity 已提交
3130
	switch (ioctl) {
3131 3132
	case KVM_RUN: {
		struct pid *oldpid;
3133 3134 3135
		r = -EINVAL;
		if (arg)
			goto out;
3136
		oldpid = rcu_access_pointer(vcpu->pid);
3137
		if (unlikely(oldpid != task_pid(current))) {
3138
			/* The thread running this VCPU changed. */
3139
			struct pid *newpid;
3140

3141 3142 3143 3144 3145
			r = kvm_arch_vcpu_run_pid_change(vcpu);
			if (r)
				break;

			newpid = get_task_pid(current, PIDTYPE_PID);
3146 3147 3148 3149 3150
			rcu_assign_pointer(vcpu->pid, newpid);
			if (oldpid)
				synchronize_rcu();
			put_pid(oldpid);
		}
3151
		r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
3152
		trace_kvm_userspace_exit(vcpu->run->exit_reason, r);
A
Avi Kivity 已提交
3153
		break;
3154
	}
A
Avi Kivity 已提交
3155
	case KVM_GET_REGS: {
3156
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
3157

3158
		r = -ENOMEM;
3159
		kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL_ACCOUNT);
3160
		if (!kvm_regs)
A
Avi Kivity 已提交
3161
			goto out;
3162 3163 3164
		r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
		if (r)
			goto out_free1;
A
Avi Kivity 已提交
3165
		r = -EFAULT;
3166 3167
		if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
			goto out_free1;
A
Avi Kivity 已提交
3168
		r = 0;
3169 3170
out_free1:
		kfree(kvm_regs);
A
Avi Kivity 已提交
3171 3172 3173
		break;
	}
	case KVM_SET_REGS: {
3174
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
3175

3176
		r = -ENOMEM;
3177 3178 3179
		kvm_regs = memdup_user(argp, sizeof(*kvm_regs));
		if (IS_ERR(kvm_regs)) {
			r = PTR_ERR(kvm_regs);
A
Avi Kivity 已提交
3180
			goto out;
3181
		}
3182 3183
		r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
		kfree(kvm_regs);
A
Avi Kivity 已提交
3184 3185 3186
		break;
	}
	case KVM_GET_SREGS: {
3187 3188
		kvm_sregs = kzalloc(sizeof(struct kvm_sregs),
				    GFP_KERNEL_ACCOUNT);
3189 3190 3191 3192
		r = -ENOMEM;
		if (!kvm_sregs)
			goto out;
		r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
3193 3194 3195
		if (r)
			goto out;
		r = -EFAULT;
3196
		if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
A
Avi Kivity 已提交
3197 3198 3199 3200 3201
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_SREGS: {
3202 3203 3204
		kvm_sregs = memdup_user(argp, sizeof(*kvm_sregs));
		if (IS_ERR(kvm_sregs)) {
			r = PTR_ERR(kvm_sregs);
G
Guo Chao 已提交
3205
			kvm_sregs = NULL;
A
Avi Kivity 已提交
3206
			goto out;
3207
		}
3208
		r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
3209 3210
		break;
	}
3211 3212 3213 3214 3215 3216 3217
	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;
3218
		if (copy_to_user(argp, &mp_state, sizeof(mp_state)))
3219 3220 3221 3222 3223 3224 3225 3226
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_MP_STATE: {
		struct kvm_mp_state mp_state;

		r = -EFAULT;
3227
		if (copy_from_user(&mp_state, argp, sizeof(mp_state)))
3228 3229 3230 3231
			goto out;
		r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
		break;
	}
A
Avi Kivity 已提交
3232 3233 3234 3235
	case KVM_TRANSLATE: {
		struct kvm_translation tr;

		r = -EFAULT;
3236
		if (copy_from_user(&tr, argp, sizeof(tr)))
A
Avi Kivity 已提交
3237
			goto out;
3238
		r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
A
Avi Kivity 已提交
3239 3240 3241
		if (r)
			goto out;
		r = -EFAULT;
3242
		if (copy_to_user(argp, &tr, sizeof(tr)))
A
Avi Kivity 已提交
3243 3244 3245 3246
			goto out;
		r = 0;
		break;
	}
J
Jan Kiszka 已提交
3247 3248
	case KVM_SET_GUEST_DEBUG: {
		struct kvm_guest_debug dbg;
A
Avi Kivity 已提交
3249 3250

		r = -EFAULT;
3251
		if (copy_from_user(&dbg, argp, sizeof(dbg)))
A
Avi Kivity 已提交
3252
			goto out;
J
Jan Kiszka 已提交
3253
		r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
A
Avi Kivity 已提交
3254 3255
		break;
	}
A
Avi Kivity 已提交
3256 3257 3258 3259 3260 3261 3262 3263 3264
	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,
3265
					   sizeof(kvm_sigmask)))
A
Avi Kivity 已提交
3266 3267
				goto out;
			r = -EINVAL;
3268
			if (kvm_sigmask.len != sizeof(sigset))
A
Avi Kivity 已提交
3269 3270 3271
				goto out;
			r = -EFAULT;
			if (copy_from_user(&sigset, sigmask_arg->sigset,
3272
					   sizeof(sigset)))
A
Avi Kivity 已提交
3273 3274 3275
				goto out;
			p = &sigset;
		}
3276
		r = kvm_vcpu_ioctl_set_sigmask(vcpu, p);
A
Avi Kivity 已提交
3277 3278
		break;
	}
A
Avi Kivity 已提交
3279
	case KVM_GET_FPU: {
3280
		fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL_ACCOUNT);
3281 3282 3283 3284
		r = -ENOMEM;
		if (!fpu)
			goto out;
		r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu);
A
Avi Kivity 已提交
3285 3286 3287
		if (r)
			goto out;
		r = -EFAULT;
3288
		if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
A
Avi Kivity 已提交
3289 3290 3291 3292 3293
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_FPU: {
3294 3295 3296
		fpu = memdup_user(argp, sizeof(*fpu));
		if (IS_ERR(fpu)) {
			r = PTR_ERR(fpu);
G
Guo Chao 已提交
3297
			fpu = NULL;
A
Avi Kivity 已提交
3298
			goto out;
3299
		}
3300
		r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
A
Avi Kivity 已提交
3301 3302
		break;
	}
A
Avi Kivity 已提交
3303
	default:
3304
		r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
3305 3306
	}
out:
3307
	mutex_unlock(&vcpu->mutex);
3308 3309
	kfree(fpu);
	kfree(kvm_sregs);
A
Avi Kivity 已提交
3310 3311 3312
	return r;
}

3313
#ifdef CONFIG_KVM_COMPAT
3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332
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,
3333
					   sizeof(kvm_sigmask)))
3334 3335
				goto out;
			r = -EINVAL;
A
Al Viro 已提交
3336
			if (kvm_sigmask.len != sizeof(compat_sigset_t))
3337 3338
				goto out;
			r = -EFAULT;
A
Al Viro 已提交
3339
			if (get_compat_sigset(&sigset, (void *)sigmask_arg->sigset))
3340
				goto out;
3341 3342 3343
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
		} else
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, NULL);
3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354
		break;
	}
	default:
		r = kvm_vcpu_ioctl(filp, ioctl, arg);
	}

out:
	return r;
}
#endif

3355 3356 3357 3358 3359 3360 3361 3362 3363 3364
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
3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385
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;

3386 3387 3388
	if (dev->kvm->mm != current->mm)
		return -EIO;

S
Scott Wood 已提交
3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408
	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;

3409 3410 3411 3412 3413 3414 3415
	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
3416 3417 3418 3419 3420 3421 3422
	kvm_put_kvm(kvm);
	return 0;
}

static const struct file_operations kvm_device_fops = {
	.unlocked_ioctl = kvm_device_ioctl,
	.release = kvm_device_release,
3423
	KVM_COMPAT(kvm_device_ioctl),
3424
	.mmap = kvm_device_mmap,
S
Scott Wood 已提交
3425 3426 3427 3428 3429 3430 3431 3432 3433 3434
};

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

	return filp->private_data;
}

3435
static const struct kvm_device_ops *kvm_device_ops_table[KVM_DEV_TYPE_MAX] = {
3436
#ifdef CONFIG_KVM_MPIC
3437 3438
	[KVM_DEV_TYPE_FSL_MPIC_20]	= &kvm_mpic_ops,
	[KVM_DEV_TYPE_FSL_MPIC_42]	= &kvm_mpic_ops,
3439
#endif
3440 3441
};

3442
int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type)
3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453
{
	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;
}

3454 3455 3456 3457 3458 3459
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
3460 3461 3462
static int kvm_ioctl_create_device(struct kvm *kvm,
				   struct kvm_create_device *cd)
{
3463
	const struct kvm_device_ops *ops = NULL;
S
Scott Wood 已提交
3464 3465
	struct kvm_device *dev;
	bool test = cd->flags & KVM_CREATE_DEVICE_TEST;
P
Paolo Bonzini 已提交
3466
	int type;
S
Scott Wood 已提交
3467 3468
	int ret;

3469 3470 3471
	if (cd->type >= ARRAY_SIZE(kvm_device_ops_table))
		return -ENODEV;

P
Paolo Bonzini 已提交
3472 3473
	type = array_index_nospec(cd->type, ARRAY_SIZE(kvm_device_ops_table));
	ops = kvm_device_ops_table[type];
3474
	if (ops == NULL)
S
Scott Wood 已提交
3475 3476 3477 3478 3479
		return -ENODEV;

	if (test)
		return 0;

3480
	dev = kzalloc(sizeof(*dev), GFP_KERNEL_ACCOUNT);
S
Scott Wood 已提交
3481 3482 3483 3484 3485 3486
	if (!dev)
		return -ENOMEM;

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

3487
	mutex_lock(&kvm->lock);
P
Paolo Bonzini 已提交
3488
	ret = ops->create(dev, type);
S
Scott Wood 已提交
3489
	if (ret < 0) {
3490
		mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
3491 3492 3493
		kfree(dev);
		return ret;
	}
3494 3495
	list_add(&dev->vm_node, &kvm->devices);
	mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
3496

3497 3498 3499
	if (ops->init)
		ops->init(dev);

3500
	kvm_get_kvm(kvm);
3501
	ret = anon_inode_getfd(ops->name, &kvm_device_fops, dev, O_RDWR | O_CLOEXEC);
S
Scott Wood 已提交
3502
	if (ret < 0) {
3503
		kvm_put_kvm_no_destroy(kvm);
3504 3505 3506
		mutex_lock(&kvm->lock);
		list_del(&dev->vm_node);
		mutex_unlock(&kvm->lock);
3507
		ops->destroy(dev);
S
Scott Wood 已提交
3508 3509 3510 3511 3512 3513 3514
		return ret;
	}

	cd->fd = ret;
	return 0;
}

3515 3516 3517 3518 3519 3520 3521 3522 3523 3524
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
3525
#ifdef CONFIG_HAVE_KVM_IRQFD
3526
	case KVM_CAP_IRQFD:
3527 3528
	case KVM_CAP_IRQFD_RESAMPLE:
#endif
3529
	case KVM_CAP_IOEVENTFD_ANY_LENGTH:
3530
	case KVM_CAP_CHECK_EXTENSION_VM:
3531
	case KVM_CAP_ENABLE_CAP_VM:
3532
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
3533
	case KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2:
3534
#endif
3535
		return 1;
3536
#ifdef CONFIG_KVM_MMIO
3537 3538
	case KVM_CAP_COALESCED_MMIO:
		return KVM_COALESCED_MMIO_PAGE_OFFSET;
P
Peng Hao 已提交
3539 3540
	case KVM_CAP_COALESCED_PIO:
		return 1;
3541
#endif
3542 3543 3544
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_CAP_IRQ_ROUTING:
		return KVM_MAX_IRQ_ROUTES;
3545 3546 3547 3548
#endif
#if KVM_ADDRESS_SPACE_NUM > 1
	case KVM_CAP_MULTI_ADDRESS_SPACE:
		return KVM_ADDRESS_SPACE_NUM;
3549
#endif
3550 3551
	case KVM_CAP_NR_MEMSLOTS:
		return KVM_USER_MEM_SLOTS;
3552 3553 3554 3555 3556 3557
	default:
		break;
	}
	return kvm_vm_ioctl_check_extension(kvm, arg);
}

3558 3559 3560 3561 3562 3563 3564 3565 3566 3567
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) {
3568
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
3569
	case KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2:
3570 3571 3572 3573 3574
		if (cap->flags || (cap->args[0] & ~1))
			return -EINVAL;
		kvm->manual_dirty_log_protect = cap->args[0];
		return 0;
#endif
3575 3576 3577 3578 3579
	default:
		return kvm_vm_ioctl_enable_cap(kvm, cap);
	}
}

A
Avi Kivity 已提交
3580 3581 3582 3583 3584
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;
3585
	int r;
A
Avi Kivity 已提交
3586

3587 3588
	if (kvm->mm != current->mm)
		return -EIO;
A
Avi Kivity 已提交
3589 3590 3591 3592
	switch (ioctl) {
	case KVM_CREATE_VCPU:
		r = kvm_vm_ioctl_create_vcpu(kvm, arg);
		break;
3593 3594 3595 3596 3597 3598 3599 3600 3601
	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;
	}
3602 3603 3604 3605 3606
	case KVM_SET_USER_MEMORY_REGION: {
		struct kvm_userspace_memory_region kvm_userspace_mem;

		r = -EFAULT;
		if (copy_from_user(&kvm_userspace_mem, argp,
3607
						sizeof(kvm_userspace_mem)))
3608 3609
			goto out;

3610
		r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem);
A
Avi Kivity 已提交
3611 3612 3613 3614 3615 3616
		break;
	}
	case KVM_GET_DIRTY_LOG: {
		struct kvm_dirty_log log;

		r = -EFAULT;
3617
		if (copy_from_user(&log, argp, sizeof(log)))
A
Avi Kivity 已提交
3618
			goto out;
3619
		r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
A
Avi Kivity 已提交
3620 3621
		break;
	}
3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632
#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
3633
#ifdef CONFIG_KVM_MMIO
3634 3635
	case KVM_REGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3636

3637
		r = -EFAULT;
3638
		if (copy_from_user(&zone, argp, sizeof(zone)))
3639 3640 3641 3642 3643 3644
			goto out;
		r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
		break;
	}
	case KVM_UNREGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3645

3646
		r = -EFAULT;
3647
		if (copy_from_user(&zone, argp, sizeof(zone)))
3648 3649 3650 3651 3652
			goto out;
		r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
		break;
	}
#endif
G
Gregory Haskins 已提交
3653 3654 3655 3656
	case KVM_IRQFD: {
		struct kvm_irqfd data;

		r = -EFAULT;
3657
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3658
			goto out;
3659
		r = kvm_irqfd(kvm, &data);
G
Gregory Haskins 已提交
3660 3661
		break;
	}
G
Gregory Haskins 已提交
3662 3663 3664 3665
	case KVM_IOEVENTFD: {
		struct kvm_ioeventfd data;

		r = -EFAULT;
3666
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3667 3668 3669 3670
			goto out;
		r = kvm_ioeventfd(kvm, &data);
		break;
	}
3671 3672 3673 3674 3675
#ifdef CONFIG_HAVE_KVM_MSI
	case KVM_SIGNAL_MSI: {
		struct kvm_msi msi;

		r = -EFAULT;
3676
		if (copy_from_user(&msi, argp, sizeof(msi)))
3677 3678 3679 3680
			goto out;
		r = kvm_send_userspace_msi(kvm, &msi);
		break;
	}
3681 3682 3683 3684 3685 3686 3687
#endif
#ifdef __KVM_HAVE_IRQ_LINE
	case KVM_IRQ_LINE_STATUS:
	case KVM_IRQ_LINE: {
		struct kvm_irq_level irq_event;

		r = -EFAULT;
3688
		if (copy_from_user(&irq_event, argp, sizeof(irq_event)))
3689 3690
			goto out;

3691 3692
		r = kvm_vm_ioctl_irq_line(kvm, &irq_event,
					ioctl == KVM_IRQ_LINE_STATUS);
3693 3694 3695 3696 3697
		if (r)
			goto out;

		r = -EFAULT;
		if (ioctl == KVM_IRQ_LINE_STATUS) {
3698
			if (copy_to_user(argp, &irq_event, sizeof(irq_event)))
3699 3700 3701 3702 3703 3704
				goto out;
		}

		r = 0;
		break;
	}
3705
#endif
3706 3707 3708 3709
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_SET_GSI_ROUTING: {
		struct kvm_irq_routing routing;
		struct kvm_irq_routing __user *urouting;
3710
		struct kvm_irq_routing_entry *entries = NULL;
3711 3712 3713 3714 3715

		r = -EFAULT;
		if (copy_from_user(&routing, argp, sizeof(routing)))
			goto out;
		r = -EINVAL;
3716 3717
		if (!kvm_arch_can_set_irq_routing(kvm))
			goto out;
3718
		if (routing.nr > KVM_MAX_IRQ_ROUTES)
3719 3720 3721
			goto out;
		if (routing.flags)
			goto out;
3722 3723
		if (routing.nr) {
			r = -ENOMEM;
3724 3725
			entries = vmalloc(array_size(sizeof(*entries),
						     routing.nr));
3726 3727 3728 3729 3730 3731 3732 3733
			if (!entries)
				goto out;
			r = -EFAULT;
			urouting = argp;
			if (copy_from_user(entries, urouting->entries,
					   routing.nr * sizeof(*entries)))
				goto out_free_irq_routing;
		}
3734 3735
		r = kvm_set_irq_routing(kvm, entries, routing.nr,
					routing.flags);
3736
out_free_irq_routing:
3737 3738 3739 3740
		vfree(entries);
		break;
	}
#endif /* CONFIG_HAVE_KVM_IRQ_ROUTING */
S
Scott Wood 已提交
3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758
	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;
	}
3759 3760 3761
	case KVM_CHECK_EXTENSION:
		r = kvm_vm_ioctl_check_extension_generic(kvm, arg);
		break;
3762
	default:
3763
		r = kvm_arch_vm_ioctl(filp, ioctl, arg);
3764 3765 3766 3767 3768
	}
out:
	return r;
}

3769
#ifdef CONFIG_KVM_COMPAT
3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793
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)))
3794
			return -EFAULT;
3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809
		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

3810
static struct file_operations kvm_vm_fops = {
3811 3812
	.release        = kvm_vm_release,
	.unlocked_ioctl = kvm_vm_ioctl,
3813
	.llseek		= noop_llseek,
3814
	KVM_COMPAT(kvm_vm_compat_ioctl),
3815 3816
};

3817
static int kvm_dev_ioctl_create_vm(unsigned long type)
3818
{
3819
	int r;
3820
	struct kvm *kvm;
3821
	struct file *file;
3822

3823
	kvm = kvm_create_vm(type);
3824 3825
	if (IS_ERR(kvm))
		return PTR_ERR(kvm);
3826
#ifdef CONFIG_KVM_MMIO
3827
	r = kvm_coalesced_mmio_init(kvm);
3828 3829
	if (r < 0)
		goto put_kvm;
3830
#endif
3831
	r = get_unused_fd_flags(O_CLOEXEC);
3832 3833 3834
	if (r < 0)
		goto put_kvm;

3835 3836 3837
	file = anon_inode_getfile("kvm-vm", &kvm_vm_fops, kvm, O_RDWR);
	if (IS_ERR(file)) {
		put_unused_fd(r);
3838 3839
		r = PTR_ERR(file);
		goto put_kvm;
3840
	}
3841

3842 3843 3844 3845 3846 3847
	/*
	 * 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).
	 */
3848
	if (kvm_create_vm_debugfs(kvm, r) < 0) {
3849 3850
		put_unused_fd(r);
		fput(file);
3851 3852
		return -ENOMEM;
	}
3853
	kvm_uevent_notify_change(KVM_EVENT_CREATE_VM, kvm);
3854

3855
	fd_install(r, file);
3856
	return r;
3857 3858 3859 3860

put_kvm:
	kvm_put_kvm(kvm);
	return r;
3861 3862 3863 3864 3865
}

static long kvm_dev_ioctl(struct file *filp,
			  unsigned int ioctl, unsigned long arg)
{
3866
	long r = -EINVAL;
3867 3868 3869

	switch (ioctl) {
	case KVM_GET_API_VERSION:
3870 3871
		if (arg)
			goto out;
3872 3873 3874
		r = KVM_API_VERSION;
		break;
	case KVM_CREATE_VM:
3875
		r = kvm_dev_ioctl_create_vm(arg);
3876
		break;
3877
	case KVM_CHECK_EXTENSION:
3878
		r = kvm_vm_ioctl_check_extension_generic(NULL, arg);
3879
		break;
3880 3881 3882
	case KVM_GET_VCPU_MMAP_SIZE:
		if (arg)
			goto out;
3883 3884 3885
		r = PAGE_SIZE;     /* struct kvm_run */
#ifdef CONFIG_X86
		r += PAGE_SIZE;    /* pio data page */
3886
#endif
3887
#ifdef CONFIG_KVM_MMIO
3888
		r += PAGE_SIZE;    /* coalesced mmio ring page */
3889
#endif
3890
		break;
3891 3892 3893
	case KVM_TRACE_ENABLE:
	case KVM_TRACE_PAUSE:
	case KVM_TRACE_DISABLE:
3894
		r = -EOPNOTSUPP;
3895
		break;
A
Avi Kivity 已提交
3896
	default:
3897
		return kvm_arch_dev_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
3898 3899 3900 3901 3902 3903 3904
	}
out:
	return r;
}

static struct file_operations kvm_chardev_ops = {
	.unlocked_ioctl = kvm_dev_ioctl,
3905
	.llseek		= noop_llseek,
3906
	KVM_COMPAT(kvm_dev_ioctl),
A
Avi Kivity 已提交
3907 3908 3909
};

static struct miscdevice kvm_dev = {
A
Avi Kivity 已提交
3910
	KVM_MINOR,
A
Avi Kivity 已提交
3911 3912 3913 3914
	"kvm",
	&kvm_chardev_ops,
};

3915
static void hardware_enable_nolock(void *junk)
3916 3917
{
	int cpu = raw_smp_processor_id();
3918
	int r;
3919

3920
	if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
3921
		return;
3922

3923
	cpumask_set_cpu(cpu, cpus_hardware_enabled);
3924

3925
	r = kvm_arch_hardware_enable();
3926 3927 3928 3929

	if (r) {
		cpumask_clear_cpu(cpu, cpus_hardware_enabled);
		atomic_inc(&hardware_enable_failed);
X
Xiubo Li 已提交
3930
		pr_info("kvm: enabling virtualization on CPU%d failed\n", cpu);
3931
	}
3932 3933
}

3934
static int kvm_starting_cpu(unsigned int cpu)
3935
{
3936
	raw_spin_lock(&kvm_count_lock);
3937 3938
	if (kvm_usage_count)
		hardware_enable_nolock(NULL);
3939
	raw_spin_unlock(&kvm_count_lock);
3940
	return 0;
3941 3942 3943
}

static void hardware_disable_nolock(void *junk)
3944 3945 3946
{
	int cpu = raw_smp_processor_id();

3947
	if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
3948
		return;
3949
	cpumask_clear_cpu(cpu, cpus_hardware_enabled);
3950
	kvm_arch_hardware_disable();
3951 3952
}

3953
static int kvm_dying_cpu(unsigned int cpu)
3954
{
3955
	raw_spin_lock(&kvm_count_lock);
3956 3957
	if (kvm_usage_count)
		hardware_disable_nolock(NULL);
3958
	raw_spin_unlock(&kvm_count_lock);
3959
	return 0;
3960 3961
}

3962 3963 3964 3965 3966 3967
static void hardware_disable_all_nolock(void)
{
	BUG_ON(!kvm_usage_count);

	kvm_usage_count--;
	if (!kvm_usage_count)
3968
		on_each_cpu(hardware_disable_nolock, NULL, 1);
3969 3970 3971 3972
}

static void hardware_disable_all(void)
{
3973
	raw_spin_lock(&kvm_count_lock);
3974
	hardware_disable_all_nolock();
3975
	raw_spin_unlock(&kvm_count_lock);
3976 3977 3978 3979 3980 3981
}

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

3982
	raw_spin_lock(&kvm_count_lock);
3983 3984 3985 3986

	kvm_usage_count++;
	if (kvm_usage_count == 1) {
		atomic_set(&hardware_enable_failed, 0);
3987
		on_each_cpu(hardware_enable_nolock, NULL, 1);
3988 3989 3990 3991 3992 3993 3994

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

3995
	raw_spin_unlock(&kvm_count_lock);
3996 3997 3998 3999

	return r;
}

4000
static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
M
Mike Day 已提交
4001
		      void *v)
4002
{
4003 4004 4005 4006 4007 4008
	/*
	 * 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
4009
	pr_info("kvm: exiting hardware virtualization\n");
4010
	kvm_rebooting = true;
4011
	on_each_cpu(hardware_disable_nolock, NULL, 1);
4012 4013 4014 4015 4016 4017 4018 4019
	return NOTIFY_OK;
}

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

M
Marcelo Tosatti 已提交
4020
static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
4021 4022 4023 4024
{
	int i;

	for (i = 0; i < bus->dev_count; i++) {
4025
		struct kvm_io_device *pos = bus->range[i].dev;
4026 4027 4028

		kvm_iodevice_destructor(pos);
	}
M
Marcelo Tosatti 已提交
4029
	kfree(bus);
4030 4031
}

4032
static inline int kvm_io_bus_cmp(const struct kvm_io_range *r1,
X
Xiubo Li 已提交
4033
				 const struct kvm_io_range *r2)
4034
{
J
Jason Wang 已提交
4035 4036 4037 4038
	gpa_t addr1 = r1->addr;
	gpa_t addr2 = r2->addr;

	if (addr1 < addr2)
4039
		return -1;
J
Jason Wang 已提交
4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051

	/* 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)
4052
		return 1;
J
Jason Wang 已提交
4053

4054 4055 4056
	return 0;
}

4057 4058
static int kvm_io_bus_sort_cmp(const void *p1, const void *p2)
{
4059
	return kvm_io_bus_cmp(p1, p2);
4060 4061
}

G
Geoff Levand 已提交
4062
static int kvm_io_bus_get_first_dev(struct kvm_io_bus *bus,
4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079
			     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;

4080
	while (off > 0 && kvm_io_bus_cmp(&key, &bus->range[off-1]) == 0)
4081 4082 4083 4084 4085
		off--;

	return off;
}

4086
static int __kvm_io_bus_write(struct kvm_vcpu *vcpu, struct kvm_io_bus *bus,
C
Cornelia Huck 已提交
4087 4088 4089 4090 4091 4092 4093 4094 4095
			      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 &&
4096
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
4097
		if (!kvm_iodevice_write(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
4098 4099 4100 4101 4102 4103 4104 4105
					range->len, val))
			return idx;
		idx++;
	}

	return -EOPNOTSUPP;
}

4106
/* kvm_io_bus_write - called under kvm->slots_lock */
4107
int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
4108
		     int len, const void *val)
4109
{
4110
	struct kvm_io_bus *bus;
4111
	struct kvm_io_range range;
C
Cornelia Huck 已提交
4112
	int r;
4113 4114 4115 4116 4117

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
4118

4119
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
4120 4121
	if (!bus)
		return -ENOMEM;
4122
	r = __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
4123 4124
	return r < 0 ? r : 0;
}
L
Leo Yan 已提交
4125
EXPORT_SYMBOL_GPL(kvm_io_bus_write);
C
Cornelia Huck 已提交
4126 4127

/* kvm_io_bus_write_cookie - called under kvm->slots_lock */
4128 4129
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
4130 4131 4132 4133 4134 4135 4136 4137 4138
{
	struct kvm_io_bus *bus;
	struct kvm_io_range range;

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

4139
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
4140 4141
	if (!bus)
		return -ENOMEM;
C
Cornelia Huck 已提交
4142 4143 4144

	/* First try the device referenced by cookie. */
	if ((cookie >= 0) && (cookie < bus->dev_count) &&
4145
	    (kvm_io_bus_cmp(&range, &bus->range[cookie]) == 0))
4146
		if (!kvm_iodevice_write(vcpu, bus->range[cookie].dev, addr, len,
C
Cornelia Huck 已提交
4147 4148 4149 4150 4151 4152 4153
					val))
			return cookie;

	/*
	 * cookie contained garbage; fall back to search and return the
	 * correct cookie value.
	 */
4154
	return __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
4155 4156
}

4157 4158
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
4159 4160 4161 4162
{
	int idx;

	idx = kvm_io_bus_get_first_dev(bus, range->addr, range->len);
4163 4164 4165 4166
	if (idx < 0)
		return -EOPNOTSUPP;

	while (idx < bus->dev_count &&
4167
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
4168
		if (!kvm_iodevice_read(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
4169 4170
				       range->len, val))
			return idx;
4171 4172 4173
		idx++;
	}

4174 4175
	return -EOPNOTSUPP;
}
4176

4177
/* kvm_io_bus_read - called under kvm->slots_lock */
4178
int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
M
Marcelo Tosatti 已提交
4179
		    int len, void *val)
4180
{
4181
	struct kvm_io_bus *bus;
4182
	struct kvm_io_range range;
C
Cornelia Huck 已提交
4183
	int r;
4184 4185 4186 4187 4188

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
M
Marcelo Tosatti 已提交
4189

4190
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
4191 4192
	if (!bus)
		return -ENOMEM;
4193
	r = __kvm_io_bus_read(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
4194 4195
	return r < 0 ? r : 0;
}
4196

4197
/* Caller must hold slots_lock. */
4198 4199
int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
			    int len, struct kvm_io_device *dev)
4200
{
4201
	int i;
M
Marcelo Tosatti 已提交
4202
	struct kvm_io_bus *new_bus, *bus;
4203
	struct kvm_io_range range;
4204

4205
	bus = kvm_get_bus(kvm, bus_idx);
4206 4207 4208
	if (!bus)
		return -ENOMEM;

4209 4210
	/* exclude ioeventfd which is limited by maximum fd */
	if (bus->dev_count - bus->ioeventfd_count > NR_IOBUS_DEVS - 1)
4211
		return -ENOSPC;
4212

4213
	new_bus = kmalloc(struct_size(bus, range, bus->dev_count + 1),
4214
			  GFP_KERNEL_ACCOUNT);
M
Marcelo Tosatti 已提交
4215 4216
	if (!new_bus)
		return -ENOMEM;
4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232

	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
4233 4234 4235
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
4236 4237 4238 4239

	return 0;
}

4240
/* Caller must hold slots_lock. */
4241 4242
void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
			       struct kvm_io_device *dev)
4243
{
4244
	int i;
M
Marcelo Tosatti 已提交
4245
	struct kvm_io_bus *new_bus, *bus;
4246

4247
	bus = kvm_get_bus(kvm, bus_idx);
4248
	if (!bus)
4249
		return;
4250

4251 4252
	for (i = 0; i < bus->dev_count; i++)
		if (bus->range[i].dev == dev) {
4253 4254
			break;
		}
M
Marcelo Tosatti 已提交
4255

4256 4257
	if (i == bus->dev_count)
		return;
4258

4259
	new_bus = kmalloc(struct_size(bus, range, bus->dev_count - 1),
4260
			  GFP_KERNEL_ACCOUNT);
4261 4262 4263 4264
	if (!new_bus)  {
		pr_err("kvm: failed to shrink bus, removing it completely\n");
		goto broken;
	}
4265 4266 4267 4268 4269

	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
4270

4271
broken:
M
Marcelo Tosatti 已提交
4272 4273 4274
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
4275
	return;
4276 4277
}

4278 4279 4280 4281 4282 4283 4284 4285 4286 4287
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);
4288 4289
	if (!bus)
		goto out_unlock;
4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303

	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);

4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315
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.
	 */
4316
	if (!refcount_inc_not_zero(&stat_data->kvm->users_count))
4317 4318
		return -ENOENT;

4319
	if (simple_attr_open(inode, file, get,
4320 4321 4322
		    KVM_DBGFS_GET_MODE(stat_data->dbgfs_item) & 0222
		    ? set : NULL,
		    fmt)) {
4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340
		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;
}

4341
static int kvm_get_stat_per_vm(struct kvm *kvm, size_t offset, u64 *val)
4342
{
4343
	*val = *(ulong *)((void *)kvm + offset);
4344

4345 4346 4347 4348 4349 4350
	return 0;
}

static int kvm_clear_stat_per_vm(struct kvm *kvm, size_t offset)
{
	*(ulong *)((void *)kvm + offset) = 0;
4351 4352 4353 4354

	return 0;
}

4355
static int kvm_get_stat_per_vcpu(struct kvm *kvm, size_t offset, u64 *val)
4356
{
4357 4358
	int i;
	struct kvm_vcpu *vcpu;
4359

4360
	*val = 0;
4361

4362 4363
	kvm_for_each_vcpu(i, vcpu, kvm)
		*val += *(u64 *)((void *)vcpu + offset);
4364 4365 4366 4367

	return 0;
}

4368
static int kvm_clear_stat_per_vcpu(struct kvm *kvm, size_t offset)
4369
{
4370 4371
	int i;
	struct kvm_vcpu *vcpu;
4372

4373 4374 4375 4376 4377
	kvm_for_each_vcpu(i, vcpu, kvm)
		*(u64 *)((void *)vcpu + offset) = 0;

	return 0;
}
4378

4379
static int kvm_stat_data_get(void *data, u64 *val)
4380
{
4381
	int r = -EFAULT;
4382 4383
	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)data;

4384 4385 4386 4387 4388 4389 4390 4391 4392 4393
	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;
	}
4394

4395
	return r;
4396 4397
}

4398
static int kvm_stat_data_clear(void *data, u64 val)
4399
{
4400
	int r = -EFAULT;
4401 4402 4403 4404 4405
	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)data;

	if (val)
		return -EINVAL;

4406 4407 4408 4409 4410 4411 4412 4413 4414 4415
	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;
	}
4416

4417
	return r;
4418 4419
}

4420
static int kvm_stat_data_open(struct inode *inode, struct file *file)
4421 4422
{
	__simple_attr_check_format("%llu\n", 0ull);
4423 4424
	return kvm_debugfs_open(inode, file, kvm_stat_data_get,
				kvm_stat_data_clear, "%llu\n");
4425 4426
}

4427 4428 4429
static const struct file_operations stat_fops_per_vm = {
	.owner = THIS_MODULE,
	.open = kvm_stat_data_open,
4430
	.release = kvm_debugfs_release,
4431 4432 4433
	.read = simple_attr_read,
	.write = simple_attr_write,
	.llseek = no_llseek,
4434 4435
};

4436
static int vm_stat_get(void *_offset, u64 *val)
4437 4438 4439
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
4440
	u64 tmp_val;
4441

4442
	*val = 0;
J
Junaid Shahid 已提交
4443
	mutex_lock(&kvm_lock);
4444
	list_for_each_entry(kvm, &vm_list, vm_list) {
4445
		kvm_get_stat_per_vm(kvm, offset, &tmp_val);
4446 4447
		*val += tmp_val;
	}
J
Junaid Shahid 已提交
4448
	mutex_unlock(&kvm_lock);
4449
	return 0;
4450 4451
}

4452 4453 4454 4455 4456 4457 4458 4459
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
4460
	mutex_lock(&kvm_lock);
4461
	list_for_each_entry(kvm, &vm_list, vm_list) {
4462
		kvm_clear_stat_per_vm(kvm, offset);
4463
	}
J
Junaid Shahid 已提交
4464
	mutex_unlock(&kvm_lock);
4465 4466 4467 4468 4469

	return 0;
}

DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, vm_stat_clear, "%llu\n");
4470

4471
static int vcpu_stat_get(void *_offset, u64 *val)
A
Avi Kivity 已提交
4472 4473 4474
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
4475
	u64 tmp_val;
A
Avi Kivity 已提交
4476

4477
	*val = 0;
J
Junaid Shahid 已提交
4478
	mutex_lock(&kvm_lock);
4479
	list_for_each_entry(kvm, &vm_list, vm_list) {
4480
		kvm_get_stat_per_vcpu(kvm, offset, &tmp_val);
4481 4482
		*val += tmp_val;
	}
J
Junaid Shahid 已提交
4483
	mutex_unlock(&kvm_lock);
4484
	return 0;
A
Avi Kivity 已提交
4485 4486
}

4487 4488 4489 4490 4491 4492 4493 4494
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
4495
	mutex_lock(&kvm_lock);
4496
	list_for_each_entry(kvm, &vm_list, vm_list) {
4497
		kvm_clear_stat_per_vcpu(kvm, offset);
4498
	}
J
Junaid Shahid 已提交
4499
	mutex_unlock(&kvm_lock);
4500 4501 4502 4503 4504 4505

	return 0;
}

DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, vcpu_stat_clear,
			"%llu\n");
4506

4507
static const struct file_operations *stat_fops[] = {
4508 4509 4510
	[KVM_STAT_VCPU] = &vcpu_stat_fops,
	[KVM_STAT_VM]   = &vm_stat_fops,
};
A
Avi Kivity 已提交
4511

4512 4513 4514 4515 4516 4517 4518 4519
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
4520
	mutex_lock(&kvm_lock);
4521 4522 4523 4524 4525 4526 4527 4528
	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
4529
	mutex_unlock(&kvm_lock);
4530

4531
	env = kzalloc(sizeof(*env), GFP_KERNEL_ACCOUNT);
4532 4533 4534 4535 4536 4537
	if (!env)
		return;

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

4538
	if (type == KVM_EVENT_CREATE_VM) {
4539
		add_uevent_var(env, "EVENT=create");
4540 4541
		kvm->userspace_pid = task_pid_nr(current);
	} else if (type == KVM_EVENT_DESTROY_VM) {
4542
		add_uevent_var(env, "EVENT=destroy");
4543 4544
	}
	add_uevent_var(env, "PID=%d", kvm->userspace_pid);
4545

4546
	if (!IS_ERR_OR_NULL(kvm->debugfs_dentry)) {
4547
		char *tmp, *p = kmalloc(PATH_MAX, GFP_KERNEL_ACCOUNT);
4548 4549 4550 4551 4552 4553

		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);
4554 4555 4556 4557 4558 4559 4560 4561
		}
	}
	/* 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);
}

4562
static void kvm_init_debug(void)
A
Avi Kivity 已提交
4563 4564 4565
{
	struct kvm_stats_debugfs_item *p;

4566
	kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
4567

4568 4569
	kvm_debugfs_num_entries = 0;
	for (p = debugfs_entries; p->name; ++p, kvm_debugfs_num_entries++) {
4570 4571
		debugfs_create_file(p->name, KVM_DBGFS_GET_MODE(p),
				    kvm_debugfs_dir, (void *)(long)p->offset,
4572
				    stat_fops[p->kind]);
4573
	}
A
Avi Kivity 已提交
4574 4575
}

4576
static int kvm_suspend(void)
4577
{
4578
	if (kvm_usage_count)
4579
		hardware_disable_nolock(NULL);
4580 4581 4582
	return 0;
}

4583
static void kvm_resume(void)
4584
{
4585
	if (kvm_usage_count) {
4586 4587 4588
#ifdef CONFIG_LOCKDEP
		WARN_ON(lockdep_is_held(&kvm_count_lock));
#endif
4589
		hardware_enable_nolock(NULL);
4590
	}
4591 4592
}

4593
static struct syscore_ops kvm_syscore_ops = {
4594 4595 4596 4597
	.suspend = kvm_suspend,
	.resume = kvm_resume,
};

4598 4599 4600 4601 4602 4603 4604 4605 4606
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);
4607

4608
	WRITE_ONCE(vcpu->preempted, false);
4609
	WRITE_ONCE(vcpu->ready, false);
4610

4611
	__this_cpu_write(kvm_running_vcpu, vcpu);
R
Radim Krčmář 已提交
4612
	kvm_arch_sched_in(vcpu, cpu);
4613
	kvm_arch_vcpu_load(vcpu, cpu);
4614 4615 4616 4617 4618 4619 4620
}

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

4621
	if (current->state == TASK_RUNNING) {
4622
		WRITE_ONCE(vcpu->preempted, true);
4623 4624
		WRITE_ONCE(vcpu->ready, true);
	}
4625
	kvm_arch_vcpu_put(vcpu);
4626 4627 4628 4629 4630
	__this_cpu_write(kvm_running_vcpu, NULL);
}

/**
 * kvm_get_running_vcpu - get the vcpu running on the current CPU.
4631 4632 4633 4634 4635 4636
 *
 * 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.
4637 4638 4639
 */
struct kvm_vcpu *kvm_get_running_vcpu(void)
{
4640 4641 4642 4643 4644 4645 4646
	struct kvm_vcpu *vcpu;

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

	return vcpu;
4647 4648 4649 4650 4651 4652 4653 4654
}

/**
 * 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;
4655 4656
}

4657 4658 4659 4660 4661
static void check_processor_compat(void *rtn)
{
	*(int *)rtn = kvm_arch_check_processor_compat();
}

4662
int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
4663
		  struct module *module)
A
Avi Kivity 已提交
4664 4665
{
	int r;
Y
Yang, Sheng 已提交
4666
	int cpu;
A
Avi Kivity 已提交
4667

4668 4669
	r = kvm_arch_init(opaque);
	if (r)
4670
		goto out_fail;
4671

4672 4673 4674 4675
	/*
	 * 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
4676 4677
	 * kvm_arch_init must be called before kvm_irqfd_init to avoid creating
	 * conflicts in case kvm is already setup for another implementation.
4678
	 */
P
Paolo Bonzini 已提交
4679 4680 4681
	r = kvm_irqfd_init();
	if (r)
		goto out_irqfd;
4682

4683
	if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
4684 4685 4686 4687
		r = -ENOMEM;
		goto out_free_0;
	}

4688
	r = kvm_arch_hardware_setup();
A
Avi Kivity 已提交
4689
	if (r < 0)
4690
		goto out_free_1;
A
Avi Kivity 已提交
4691

Y
Yang, Sheng 已提交
4692
	for_each_online_cpu(cpu) {
4693
		smp_call_function_single(cpu, check_processor_compat, &r, 1);
Y
Yang, Sheng 已提交
4694
		if (r < 0)
4695
			goto out_free_2;
Y
Yang, Sheng 已提交
4696 4697
	}

T
Thomas Gleixner 已提交
4698
	r = cpuhp_setup_state_nocalls(CPUHP_AP_KVM_STARTING, "kvm/cpu:starting",
4699
				      kvm_starting_cpu, kvm_dying_cpu);
A
Avi Kivity 已提交
4700
	if (r)
4701
		goto out_free_2;
A
Avi Kivity 已提交
4702 4703
	register_reboot_notifier(&kvm_reboot_notifier);

4704
	/* A kmem cache lets us meet the alignment requirements of fx_save. */
4705 4706
	if (!vcpu_align)
		vcpu_align = __alignof__(struct kvm_vcpu);
4707 4708 4709 4710 4711 4712
	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);
4713 4714
	if (!kvm_vcpu_cache) {
		r = -ENOMEM;
4715
		goto out_free_3;
4716 4717
	}

4718 4719 4720 4721
	r = kvm_async_pf_init();
	if (r)
		goto out_free;

A
Avi Kivity 已提交
4722
	kvm_chardev_ops.owner = module;
4723 4724
	kvm_vm_fops.owner = module;
	kvm_vcpu_fops.owner = module;
A
Avi Kivity 已提交
4725 4726 4727

	r = misc_register(&kvm_dev);
	if (r) {
X
Xiubo Li 已提交
4728
		pr_err("kvm: misc device register failed\n");
4729
		goto out_unreg;
A
Avi Kivity 已提交
4730 4731
	}

4732 4733
	register_syscore_ops(&kvm_syscore_ops);

4734 4735 4736
	kvm_preempt_ops.sched_in = kvm_sched_in;
	kvm_preempt_ops.sched_out = kvm_sched_out;

4737
	kvm_init_debug();
4738

P
Paolo Bonzini 已提交
4739 4740 4741
	r = kvm_vfio_ops_init();
	WARN_ON(r);

4742
	return 0;
A
Avi Kivity 已提交
4743

4744 4745
out_unreg:
	kvm_async_pf_deinit();
A
Avi Kivity 已提交
4746
out_free:
4747
	kmem_cache_destroy(kvm_vcpu_cache);
4748
out_free_3:
A
Avi Kivity 已提交
4749
	unregister_reboot_notifier(&kvm_reboot_notifier);
4750
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4751
out_free_2:
4752
	kvm_arch_hardware_unsetup();
4753
out_free_1:
4754
	free_cpumask_var(cpus_hardware_enabled);
4755
out_free_0:
4756
	kvm_irqfd_exit();
P
Paolo Bonzini 已提交
4757
out_irqfd:
4758 4759
	kvm_arch_exit();
out_fail:
A
Avi Kivity 已提交
4760 4761
	return r;
}
4762
EXPORT_SYMBOL_GPL(kvm_init);
A
Avi Kivity 已提交
4763

4764
void kvm_exit(void)
A
Avi Kivity 已提交
4765
{
4766
	debugfs_remove_recursive(kvm_debugfs_dir);
A
Avi Kivity 已提交
4767
	misc_deregister(&kvm_dev);
4768
	kmem_cache_destroy(kvm_vcpu_cache);
4769
	kvm_async_pf_deinit();
4770
	unregister_syscore_ops(&kvm_syscore_ops);
A
Avi Kivity 已提交
4771
	unregister_reboot_notifier(&kvm_reboot_notifier);
4772
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4773
	on_each_cpu(hardware_disable_nolock, NULL, 1);
4774
	kvm_arch_hardware_unsetup();
4775
	kvm_arch_exit();
4776
	kvm_irqfd_exit();
4777
	free_cpumask_var(cpus_hardware_enabled);
4778
	kvm_vfio_ops_exit();
A
Avi Kivity 已提交
4779
}
4780
EXPORT_SYMBOL_GPL(kvm_exit);
4781 4782 4783 4784 4785 4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863

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;
}