kvm_main.c 108.8 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-only
A
Avi Kivity 已提交
2 3 4 5 6 7 8
/*
 * Kernel-based Virtual Machine driver for Linux
 *
 * This module enables machines with Intel VT-x extensions to run virtual
 * machines without emulation or binary translation.
 *
 * Copyright (C) 2006 Qumranet, Inc.
N
Nicolas Kaiser 已提交
9
 * Copyright 2010 Red Hat, Inc. and/or its affiliates.
A
Avi Kivity 已提交
10 11 12 13 14 15
 *
 * Authors:
 *   Avi Kivity   <avi@qumranet.com>
 *   Yaniv Kamay  <yaniv@qumranet.com>
 */

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

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

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

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

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

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

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

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

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

83 84 85 86 87
/* The start value to grow halt_poll_ns from */
unsigned int halt_poll_ns_grow_start = 10000; /* 10us */
module_param(halt_poll_ns_grow_start, uint, 0644);
EXPORT_SYMBOL_GPL(halt_poll_ns_grow_start);

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

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

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

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

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

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

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

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

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

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

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

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

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

152 153
static bool largepages_enabled = true;

154 155 156 157 158 159
#define KVM_EVENT_CREATE_VM 0
#define KVM_EVENT_DESTROY_VM 1
static void kvm_uevent_notify_change(unsigned int type, struct kvm *kvm);
static unsigned long long kvm_createvm_count;
static unsigned long long kvm_active_vms;

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

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

	return is_zone_device_page(pfn_to_page(pfn));
}

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

	return true;
}

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

	if (!PageTransCompoundMap(page))
		return false;

	return is_transparent_hugepage(compound_head(page));
}

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

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

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

229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246
/* TODO: merge with kvm_arch_vcpu_should_kick */
static bool kvm_request_needs_ipi(struct kvm_vcpu *vcpu, unsigned req)
{
	int mode = kvm_vcpu_exiting_guest_mode(vcpu);

	/*
	 * We need to wait for the VCPU to reenable interrupts and get out of
	 * READING_SHADOW_PAGE_TABLES mode.
	 */
	if (req & KVM_REQUEST_WAIT)
		return mode != OUTSIDE_GUEST_MODE;

	/*
	 * Need to kick a running VCPU, but otherwise there is nothing to do.
	 */
	return mode == IN_GUEST_MODE;
}

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

251 252 253 254 255 256 257 258 259 260 261 262
static inline bool kvm_kick_many_cpus(const struct cpumask *cpus, bool wait)
{
	if (unlikely(!cpus))
		cpus = cpu_online_mask;

	if (cpumask_empty(cpus))
		return false;

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

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

270
	me = get_cpu();
271

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

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

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

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

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

	return called;
}

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

	zalloc_cpumask_var(&cpus, GFP_ATOMIC);

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

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

306
#ifndef CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALL
307
void kvm_flush_remote_tlbs(struct kvm *kvm)
308
{
309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
	/*
	 * Read tlbs_dirty before setting KVM_REQ_TLB_FLUSH in
	 * kvm_make_all_cpus_request.
	 */
	long dirty_count = smp_load_acquire(&kvm->tlbs_dirty);

	/*
	 * We want to publish modifications to the page tables before reading
	 * mode. Pairs with a memory barrier in arch-specific code.
	 * - x86: smp_mb__after_srcu_read_unlock in vcpu_enter_guest
	 * and smp_mb in walk_shadow_page_lockless_begin/end.
	 * - powerpc: smp_mb in kvmppc_prepare_to_enter.
	 *
	 * There is already an smp_mb__after_atomic() before
	 * kvm_make_all_cpus_request() reads vcpu->mode. We reuse that
	 * barrier here.
	 */
326 327
	if (!kvm_arch_flush_remote_tlb(kvm)
	    || kvm_make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH))
328
		++kvm->stat.remote_tlb_flush;
329
	cmpxchg(&kvm->tlbs_dirty, dirty_count, 0);
330
}
331
EXPORT_SYMBOL_GPL(kvm_flush_remote_tlbs);
332
#endif
333

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

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

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

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

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

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

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

375 376 377 378 379 380
#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn)
{
	return container_of(mn, struct kvm, mmu_notifier);
}

381 382 383 384 385 386
static void kvm_mmu_notifier_change_pte(struct mmu_notifier *mn,
					struct mm_struct *mm,
					unsigned long address,
					pte_t pte)
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
387
	int idx;
388

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

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

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

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

407
	idx = srcu_read_lock(&kvm->srcu);
408 409 410 411 412 413 414
	spin_lock(&kvm->mmu_lock);
	/*
	 * The count increase must become visible at unlock time as no
	 * spte can be established without taking the mmu_lock and
	 * count is also read inside the mmu_lock critical section.
	 */
	kvm->mmu_notifier_count++;
415
	need_tlb_flush = kvm_unmap_hva_range(kvm, range->start, range->end);
416
	need_tlb_flush |= kvm->tlbs_dirty;
417 418 419
	/* we've to flush the tlb before the pages can be freed */
	if (need_tlb_flush)
		kvm_flush_remote_tlbs(kvm);
420 421

	spin_unlock(&kvm->mmu_lock);
422

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

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

	return ret;
430 431 432
}

static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
433
					const struct mmu_notifier_range *range)
434 435 436 437 438 439 440 441 442 443
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);

	spin_lock(&kvm->mmu_lock);
	/*
	 * This sequence increase will notify the kvm page fault that
	 * the page that is going to be mapped in the spte could have
	 * been freed.
	 */
	kvm->mmu_notifier_seq++;
444
	smp_wmb();
445 446
	/*
	 * The above sequence increase must be visible before the
447 448
	 * below count decrease, which is ensured by the smp_wmb above
	 * in conjunction with the smp_rmb in mmu_notifier_retry().
449 450 451 452 453 454 455 456 457
	 */
	kvm->mmu_notifier_count--;
	spin_unlock(&kvm->mmu_lock);

	BUG_ON(kvm->mmu_notifier_count < 0);
}

static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
					      struct mm_struct *mm,
A
Andres Lagar-Cavilla 已提交
458 459
					      unsigned long start,
					      unsigned long end)
460 461
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
462
	int young, idx;
463

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

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

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

474 475 476
	return young;
}

477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506
static int kvm_mmu_notifier_clear_young(struct mmu_notifier *mn,
					struct mm_struct *mm,
					unsigned long start,
					unsigned long end)
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
	int young, idx;

	idx = srcu_read_lock(&kvm->srcu);
	spin_lock(&kvm->mmu_lock);
	/*
	 * Even though we do not flush TLB, this will still adversely
	 * affect performance on pre-Haswell Intel EPT, where there is
	 * no EPT Access Bit to clear so that we have to tear down EPT
	 * tables instead. If we find this unacceptable, we can always
	 * add a parameter to kvm_age_hva so that it effectively doesn't
	 * do anything on clear_young.
	 *
	 * Also note that currently we never issue secondary TLB flushes
	 * from clear_young, leaving this job up to the regular system
	 * cadence. If we find this inaccurate, we might come up with a
	 * more sophisticated heuristic later.
	 */
	young = kvm_age_hva(kvm, start, end);
	spin_unlock(&kvm->mmu_lock);
	srcu_read_unlock(&kvm->srcu, idx);

	return young;
}

A
Andrea Arcangeli 已提交
507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522
static int kvm_mmu_notifier_test_young(struct mmu_notifier *mn,
				       struct mm_struct *mm,
				       unsigned long address)
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
	int young, idx;

	idx = srcu_read_lock(&kvm->srcu);
	spin_lock(&kvm->mmu_lock);
	young = kvm_test_age_hva(kvm, address);
	spin_unlock(&kvm->mmu_lock);
	srcu_read_unlock(&kvm->srcu, idx);

	return young;
}

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

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

534 535 536 537
static const struct mmu_notifier_ops kvm_mmu_notifier_ops = {
	.invalidate_range_start	= kvm_mmu_notifier_invalidate_range_start,
	.invalidate_range_end	= kvm_mmu_notifier_invalidate_range_end,
	.clear_flush_young	= kvm_mmu_notifier_clear_flush_young,
538
	.clear_young		= kvm_mmu_notifier_clear_young,
A
Andrea Arcangeli 已提交
539
	.test_young		= kvm_mmu_notifier_test_young,
540
	.change_pte		= kvm_mmu_notifier_change_pte,
541
	.release		= kvm_mmu_notifier_release,
542
};
543 544 545 546 547 548 549 550 551 552 553 554 555 556

static int kvm_init_mmu_notifier(struct kvm *kvm)
{
	kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops;
	return mmu_notifier_register(&kvm->mmu_notifier, current->mm);
}

#else  /* !(CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER) */

static int kvm_init_mmu_notifier(struct kvm *kvm)
{
	return 0;
}

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

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

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

568
	for (i = 0; i < KVM_MEM_SLOTS_NUM; i++)
569
		slots->id_to_index[i] = slots->memslots[i].id = i;
570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607

	return slots;
}

static void kvm_destroy_dirty_bitmap(struct kvm_memory_slot *memslot)
{
	if (!memslot->dirty_bitmap)
		return;

	kvfree(memslot->dirty_bitmap);
	memslot->dirty_bitmap = NULL;
}

/*
 * Free any memory in @free but not in @dont.
 */
static void kvm_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
			      struct kvm_memory_slot *dont)
{
	if (!dont || free->dirty_bitmap != dont->dirty_bitmap)
		kvm_destroy_dirty_bitmap(free);

	kvm_arch_free_memslot(kvm, free, dont);

	free->npages = 0;
}

static void kvm_free_memslots(struct kvm *kvm, struct kvm_memslots *slots)
{
	struct kvm_memory_slot *memslot;

	if (!slots)
		return;

	kvm_for_each_memslot(memslot, slots)
		kvm_free_memslot(kvm, memslot, NULL);

	kvfree(slots);
608 609
}

610 611 612 613 614 615 616 617 618
static void kvm_destroy_vm_debugfs(struct kvm *kvm)
{
	int i;

	if (!kvm->debugfs_dentry)
		return;

	debugfs_remove_recursive(kvm->debugfs_dentry);

619 620 621 622 623
	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);
	}
624 625 626 627 628 629 630 631 632 633 634 635
}

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);
636
	kvm->debugfs_dentry = debugfs_create_dir(dir_name, kvm_debugfs_dir);
637 638 639

	kvm->debugfs_stat_data = kcalloc(kvm_debugfs_num_entries,
					 sizeof(*kvm->debugfs_stat_data),
640
					 GFP_KERNEL_ACCOUNT);
641 642 643 644
	if (!kvm->debugfs_stat_data)
		return -ENOMEM;

	for (p = debugfs_entries; p->name; p++) {
645
		stat_data = kzalloc(sizeof(*stat_data), GFP_KERNEL_ACCOUNT);
646 647 648 649
		if (!stat_data)
			return -ENOMEM;

		stat_data->kvm = kvm;
650
		stat_data->dbgfs_item = p;
651
		kvm->debugfs_stat_data[p - debugfs_entries] = stat_data;
652 653 654
		debugfs_create_file(p->name, KVM_DBGFS_GET_MODE(p),
				    kvm->debugfs_dentry, stat_data,
				    &stat_fops_per_vm);
655 656 657 658
	}
	return 0;
}

659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675
/*
 * 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)
{
}

676
static struct kvm *kvm_create_vm(unsigned long type)
A
Avi Kivity 已提交
677
{
678
	struct kvm *kvm = kvm_arch_alloc_vm();
679 680
	int r = -ENOMEM;
	int i;
A
Avi Kivity 已提交
681

682 683 684
	if (!kvm)
		return ERR_PTR(-ENOMEM);

685
	spin_lock_init(&kvm->mmu_lock);
V
Vegard Nossum 已提交
686
	mmgrab(current->mm);
687 688 689 690 691 692 693
	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);

694 695
	BUILD_BUG_ON(KVM_MEM_SLOTS_NUM > SHRT_MAX);

696 697 698 699 700
	if (init_srcu_struct(&kvm->srcu))
		goto out_err_no_srcu;
	if (init_srcu_struct(&kvm->irq_srcu))
		goto out_err_no_irq_srcu;

701
	refcount_set(&kvm->users_count, 1);
702
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
703
		struct kvm_memslots *slots = kvm_alloc_memslots();
704

705
		if (!slots)
706
			goto out_err_no_arch_destroy_vm;
707
		/* Generations must be different for each address space. */
708
		slots->generation = i;
709
		rcu_assign_pointer(kvm->memslots[i], slots);
710
	}
711

M
Marcelo Tosatti 已提交
712
	for (i = 0; i < KVM_NR_BUSES; i++) {
713
		rcu_assign_pointer(kvm->buses[i],
714
			kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL_ACCOUNT));
715
		if (!kvm->buses[i])
716
			goto out_err_no_arch_destroy_vm;
M
Marcelo Tosatti 已提交
717
	}
718

719
	r = kvm_arch_init_vm(kvm, type);
720
	if (r)
721
		goto out_err_no_arch_destroy_vm;
722 723 724

	r = hardware_enable_all();
	if (r)
725
		goto out_err_no_disable;
726

727
#ifdef CONFIG_HAVE_KVM_IRQFD
728
	INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list);
729
#endif
A
Avi Kivity 已提交
730

731
	r = kvm_init_mmu_notifier(kvm);
732 733 734 735
	if (r)
		goto out_err_no_mmu_notifier;

	r = kvm_arch_post_init_vm(kvm);
736 737 738
	if (r)
		goto out_err;

J
Junaid Shahid 已提交
739
	mutex_lock(&kvm_lock);
740
	list_add(&kvm->vm_list, &vm_list);
J
Junaid Shahid 已提交
741
	mutex_unlock(&kvm_lock);
742

743 744
	preempt_notifier_inc();

745
	return kvm;
746 747

out_err:
748 749 750 751 752
#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:
753
	hardware_disable_all();
754
out_err_no_disable:
755 756
	kvm_arch_destroy_vm(kvm);
out_err_no_arch_destroy_vm:
757
	WARN_ON_ONCE(!refcount_dec_and_test(&kvm->users_count));
M
Marcelo Tosatti 已提交
758
	for (i = 0; i < KVM_NR_BUSES; i++)
759
		kfree(kvm_get_bus(kvm, i));
760
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
761
		kvm_free_memslots(kvm, __kvm_memslots(kvm, i));
762 763 764 765
	cleanup_srcu_struct(&kvm->irq_srcu);
out_err_no_irq_srcu:
	cleanup_srcu_struct(&kvm->srcu);
out_err_no_srcu:
766
	kvm_arch_free_vm(kvm);
767
	mmdrop(current->mm);
768
	return ERR_PTR(r);
769 770
}

771 772
static void kvm_destroy_devices(struct kvm *kvm)
{
G
Geliang Tang 已提交
773
	struct kvm_device *dev, *tmp;
774

775 776 777 778 779
	/*
	 * 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
780 781
	list_for_each_entry_safe(dev, tmp, &kvm->devices, vm_node) {
		list_del(&dev->vm_node);
782 783 784 785
		dev->ops->destroy(dev);
	}
}

786 787
static void kvm_destroy_vm(struct kvm *kvm)
{
M
Marcelo Tosatti 已提交
788
	int i;
789 790
	struct mm_struct *mm = kvm->mm;

791
	kvm_uevent_notify_change(KVM_EVENT_DESTROY_VM, kvm);
792
	kvm_destroy_vm_debugfs(kvm);
793
	kvm_arch_sync_events(kvm);
J
Junaid Shahid 已提交
794
	mutex_lock(&kvm_lock);
795
	list_del(&kvm->vm_list);
J
Junaid Shahid 已提交
796
	mutex_unlock(&kvm_lock);
797 798
	kvm_arch_pre_destroy_vm(kvm);

799
	kvm_free_irq_routing(kvm);
800
	for (i = 0; i < KVM_NR_BUSES; i++) {
801
		struct kvm_io_bus *bus = kvm_get_bus(kvm, i);
802 803 804

		if (bus)
			kvm_io_bus_destroy(bus);
805 806
		kvm->buses[i] = NULL;
	}
807
	kvm_coalesced_mmio_free(kvm);
808 809
#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
	mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
810
#else
811
	kvm_arch_flush_shadow_all(kvm);
812
#endif
813
	kvm_arch_destroy_vm(kvm);
814
	kvm_destroy_devices(kvm);
815
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
816
		kvm_free_memslots(kvm, __kvm_memslots(kvm, i));
817
	cleanup_srcu_struct(&kvm->irq_srcu);
818 819
	cleanup_srcu_struct(&kvm->srcu);
	kvm_arch_free_vm(kvm);
820
	preempt_notifier_dec();
821
	hardware_disable_all();
822
	mmdrop(mm);
823 824
}

I
Izik Eidus 已提交
825 826
void kvm_get_kvm(struct kvm *kvm)
{
827
	refcount_inc(&kvm->users_count);
I
Izik Eidus 已提交
828 829 830 831 832
}
EXPORT_SYMBOL_GPL(kvm_get_kvm);

void kvm_put_kvm(struct kvm *kvm)
{
833
	if (refcount_dec_and_test(&kvm->users_count))
I
Izik Eidus 已提交
834 835 836 837
		kvm_destroy_vm(kvm);
}
EXPORT_SYMBOL_GPL(kvm_put_kvm);

838 839 840 841 842 843 844 845 846 847 848 849
/*
 * 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
850

851 852 853 854
static int kvm_vm_release(struct inode *inode, struct file *filp)
{
	struct kvm *kvm = filp->private_data;

G
Gregory Haskins 已提交
855 856
	kvm_irqfd_release(kvm);

I
Izik Eidus 已提交
857
	kvm_put_kvm(kvm);
A
Avi Kivity 已提交
858 859 860
	return 0;
}

861 862
/*
 * Allocation size is twice as large as the actual dirty bitmap size.
863
 * See x86's kvm_vm_ioctl_get_dirty_log() why this is needed.
864
 */
865 866
static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot)
{
867
	unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot);
868

869
	memslot->dirty_bitmap = kvzalloc(dirty_bytes, GFP_KERNEL_ACCOUNT);
870 871 872 873 874 875
	if (!memslot->dirty_bitmap)
		return -ENOMEM;

	return 0;
}

876
/*
877 878 879 880
 * Insert memslot and re-sort memslots based on their GFN,
 * so binary search could be used to lookup GFN.
 * Sorting algorithm takes advantage of having initially
 * sorted array and known changed memslot position.
881
 */
882
static void update_memslots(struct kvm_memslots *slots,
883 884
			    struct kvm_memory_slot *new,
			    enum kvm_mr_change change)
885
{
886 887
	int id = new->id;
	int i = slots->id_to_index[id];
888
	struct kvm_memory_slot *mslots = slots->memslots;
889

890
	WARN_ON(mslots[i].id != id);
891 892 893 894 895 896 897 898 899 900 901
	switch (change) {
	case KVM_MR_CREATE:
		slots->used_slots++;
		WARN_ON(mslots[i].npages || !new->npages);
		break;
	case KVM_MR_DELETE:
		slots->used_slots--;
		WARN_ON(new->npages || !mslots[i].npages);
		break;
	default:
		break;
902
	}
903

904
	while (i < KVM_MEM_SLOTS_NUM - 1 &&
905 906 907
	       new->base_gfn <= mslots[i + 1].base_gfn) {
		if (!mslots[i + 1].npages)
			break;
908 909 910 911
		mslots[i] = mslots[i + 1];
		slots->id_to_index[mslots[i].id] = i;
		i++;
	}
912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928

	/*
	 * The ">=" is needed when creating a slot with base_gfn == 0,
	 * so that it moves before all those with base_gfn == npages == 0.
	 *
	 * On the other hand, if new->npages is zero, the above loop has
	 * already left i pointing to the beginning of the empty part of
	 * mslots, and the ">=" would move the hole backwards in this
	 * case---which is wrong.  So skip the loop when deleting a slot.
	 */
	if (new->npages) {
		while (i > 0 &&
		       new->base_gfn >= mslots[i - 1].base_gfn) {
			mslots[i] = mslots[i - 1];
			slots->id_to_index[mslots[i].id] = i;
			i--;
		}
929 930
	} else
		WARN_ON_ONCE(i != slots->used_slots);
931

932 933
	mslots[i] = *new;
	slots->id_to_index[mslots[i].id] = i;
934 935
}

936
static int check_memory_region_flags(const struct kvm_userspace_memory_region *mem)
937
{
X
Xiao Guangrong 已提交
938 939
	u32 valid_flags = KVM_MEM_LOG_DIRTY_PAGES;

940
#ifdef __KVM_HAVE_READONLY_MEM
X
Xiao Guangrong 已提交
941 942 943 944
	valid_flags |= KVM_MEM_READONLY;
#endif

	if (mem->flags & ~valid_flags)
945 946 947 948 949
		return -EINVAL;

	return 0;
}

950
static struct kvm_memslots *install_new_memslots(struct kvm *kvm,
951
		int as_id, struct kvm_memslots *slots)
952
{
953
	struct kvm_memslots *old_memslots = __kvm_memslots(kvm, as_id);
954
	u64 gen = old_memslots->generation;
955

956 957
	WARN_ON(gen & KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS);
	slots->generation = gen | KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS;
958

959
	rcu_assign_pointer(kvm->memslots[as_id], slots);
960
	synchronize_srcu_expedited(&kvm->srcu);
961

962
	/*
963
	 * Increment the new memslot generation a second time, dropping the
M
Miaohe Lin 已提交
964
	 * update in-progress flag and incrementing the generation based on
965 966 967 968 969 970
	 * 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;

	/*
971 972 973
	 * 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
974 975
	 * space 0 will use generations 0, 2, 4, ... while address space 1 will
	 * use generations 1, 3, 5, ...
976
	 */
977
	gen += KVM_ADDRESS_SPACE_NUM;
978

979
	kvm_arch_memslots_updated(kvm, gen);
980

981
	slots->generation = gen;
982 983

	return old_memslots;
984 985
}

A
Avi Kivity 已提交
986 987 988 989 990
/*
 * Allocate some memory and give it an address in the guest physical address
 * space.
 *
 * Discontiguous memory is allowed, mostly for framebuffers.
991
 *
992
 * Must be called holding kvm->slots_lock for write.
A
Avi Kivity 已提交
993
 */
994
int __kvm_set_memory_region(struct kvm *kvm,
995
			    const struct kvm_userspace_memory_region *mem)
A
Avi Kivity 已提交
996
{
997
	int r;
A
Avi Kivity 已提交
998
	gfn_t base_gfn;
999
	unsigned long npages;
1000
	struct kvm_memory_slot *slot;
A
Avi Kivity 已提交
1001
	struct kvm_memory_slot old, new;
1002
	struct kvm_memslots *slots;
1003
	int as_id, id;
1004
	enum kvm_mr_change change;
A
Avi Kivity 已提交
1005

1006 1007 1008 1009
	r = check_memory_region_flags(mem);
	if (r)
		goto out;

A
Avi Kivity 已提交
1010
	r = -EINVAL;
1011 1012 1013
	as_id = mem->slot >> 16;
	id = (u16)mem->slot;

A
Avi Kivity 已提交
1014 1015 1016 1017 1018
	/* General sanity checks */
	if (mem->memory_size & (PAGE_SIZE - 1))
		goto out;
	if (mem->guest_phys_addr & (PAGE_SIZE - 1))
		goto out;
1019
	/* We can read the guest memory with __xxx_user() later on. */
1020
	if ((id < KVM_USER_MEM_SLOTS) &&
1021
	    ((mem->userspace_addr & (PAGE_SIZE - 1)) ||
1022
	     !access_ok((void __user *)(unsigned long)mem->userspace_addr,
1023
			mem->memory_size)))
1024
		goto out;
1025
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_MEM_SLOTS_NUM)
A
Avi Kivity 已提交
1026 1027 1028 1029
		goto out;
	if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
		goto out;

1030
	slot = id_to_memslot(__kvm_memslots(kvm, as_id), id);
A
Avi Kivity 已提交
1031 1032 1033
	base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
	npages = mem->memory_size >> PAGE_SHIFT;

1034 1035 1036
	if (npages > KVM_MEM_MAX_NR_PAGES)
		goto out;

1037
	new = old = *slot;
A
Avi Kivity 已提交
1038

1039
	new.id = id;
A
Avi Kivity 已提交
1040 1041 1042
	new.base_gfn = base_gfn;
	new.npages = npages;
	new.flags = mem->flags;
1043
	new.userspace_addr = mem->userspace_addr;
A
Avi Kivity 已提交
1044

1045 1046 1047 1048
	if (npages) {
		if (!old.npages)
			change = KVM_MR_CREATE;
		else { /* Modify an existing slot. */
1049
			if ((new.userspace_addr != old.userspace_addr) ||
1050 1051
			    (npages != old.npages) ||
			    ((new.flags ^ old.flags) & KVM_MEM_READONLY))
1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062
				goto out;

			if (base_gfn != old.base_gfn)
				change = KVM_MR_MOVE;
			else if (new.flags != old.flags)
				change = KVM_MR_FLAGS_ONLY;
			else { /* Nothing to change. */
				r = 0;
				goto out;
			}
		}
1063 1064 1065 1066
	} else {
		if (!old.npages)
			goto out;

1067
		change = KVM_MR_DELETE;
1068 1069 1070
		new.base_gfn = 0;
		new.flags = 0;
	}
A
Avi Kivity 已提交
1071

1072
	if ((change == KVM_MR_CREATE) || (change == KVM_MR_MOVE)) {
1073 1074
		/* Check for overlaps */
		r = -EEXIST;
1075
		kvm_for_each_memslot(slot, __kvm_memslots(kvm, as_id)) {
1076
			if (slot->id == id)
1077 1078 1079 1080 1081
				continue;
			if (!((base_gfn + npages <= slot->base_gfn) ||
			      (base_gfn >= slot->base_gfn + slot->npages)))
				goto out;
		}
A
Avi Kivity 已提交
1082 1083 1084
	}

	r = -ENOMEM;
1085

1086 1087 1088 1089
	/* Allocate/free page dirty bitmap as needed */
	if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
		new.dirty_bitmap = NULL;
	else if (!new.dirty_bitmap) {
1090
		if (kvm_create_dirty_bitmap(&new) < 0)
1091
			goto out;
A
Avi Kivity 已提交
1092 1093
	}

1094
	slots = kvzalloc(sizeof(struct kvm_memslots), GFP_KERNEL_ACCOUNT);
1095 1096
	if (!slots)
		goto out_free;
1097
	memcpy(slots, __kvm_memslots(kvm, as_id), sizeof(struct kvm_memslots));
1098

1099
	if ((change == KVM_MR_DELETE) || (change == KVM_MR_MOVE)) {
1100
		slot = id_to_memslot(slots, id);
1101 1102
		slot->flags |= KVM_MEMSLOT_INVALID;

1103 1104 1105 1106 1107 1108 1109
		/*
		 * 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);
1110

1111 1112
		/* From this point no new shadow pages pointing to a deleted,
		 * or moved, memslot will be created.
1113 1114
		 *
		 * validation of sp->gfn happens in:
1115
		 *	- gfn_to_hva (kvm_read_guest, gfn_to_pfn)
1116
		 *	- kvm_is_visible_gfn (mmu_check_root)
1117
		 */
1118
		kvm_arch_flush_shadow_memslot(kvm, slot);
1119
	}
1120

1121
	r = kvm_arch_prepare_memory_region(kvm, &new, mem, change);
1122
	if (r)
1123
		goto out_slots;
1124

1125
	/* actual memory is freed via old in kvm_free_memslot below */
1126
	if (change == KVM_MR_DELETE) {
1127
		new.dirty_bitmap = NULL;
1128
		memset(&new.arch, 0, sizeof(new.arch));
1129 1130
	}

1131
	update_memslots(slots, &new, change);
1132
	slots = install_new_memslots(kvm, as_id, slots);
1133

1134
	kvm_arch_commit_memory_region(kvm, mem, &old, &new, change);
1135

1136
	kvm_free_memslot(kvm, &old, &new);
1137
	kvfree(slots);
A
Avi Kivity 已提交
1138 1139
	return 0;

1140
out_slots:
1141 1142
	if (change == KVM_MR_DELETE || change == KVM_MR_MOVE)
		slots = install_new_memslots(kvm, as_id, slots);
1143
	kvfree(slots);
1144
out_free:
1145
	kvm_free_memslot(kvm, &new, &old);
A
Avi Kivity 已提交
1146 1147
out:
	return r;
1148
}
1149 1150 1151
EXPORT_SYMBOL_GPL(__kvm_set_memory_region);

int kvm_set_memory_region(struct kvm *kvm,
1152
			  const struct kvm_userspace_memory_region *mem)
1153 1154 1155
{
	int r;

1156
	mutex_lock(&kvm->slots_lock);
1157
	r = __kvm_set_memory_region(kvm, mem);
1158
	mutex_unlock(&kvm->slots_lock);
1159 1160
	return r;
}
1161 1162
EXPORT_SYMBOL_GPL(kvm_set_memory_region);

1163 1164
static int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
					  struct kvm_userspace_memory_region *mem)
1165
{
1166
	if ((u16)mem->slot >= KVM_USER_MEM_SLOTS)
1167
		return -EINVAL;
1168

1169
	return kvm_set_memory_region(kvm, mem);
A
Avi Kivity 已提交
1170 1171
}

1172 1173
int kvm_get_dirty_log(struct kvm *kvm,
			struct kvm_dirty_log *log, int *is_dirty)
A
Avi Kivity 已提交
1174
{
1175
	struct kvm_memslots *slots;
A
Avi Kivity 已提交
1176
	struct kvm_memory_slot *memslot;
1177
	int i, as_id, id;
1178
	unsigned long n;
A
Avi Kivity 已提交
1179 1180
	unsigned long any = 0;

1181 1182 1183
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1184
		return -EINVAL;
A
Avi Kivity 已提交
1185

1186 1187
	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
A
Avi Kivity 已提交
1188
	if (!memslot->dirty_bitmap)
1189
		return -ENOENT;
A
Avi Kivity 已提交
1190

1191
	n = kvm_dirty_bitmap_bytes(memslot);
A
Avi Kivity 已提交
1192

1193
	for (i = 0; !any && i < n/sizeof(long); ++i)
A
Avi Kivity 已提交
1194 1195 1196
		any = memslot->dirty_bitmap[i];

	if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
1197
		return -EFAULT;
A
Avi Kivity 已提交
1198

1199 1200
	if (any)
		*is_dirty = 1;
1201
	return 0;
A
Avi Kivity 已提交
1202
}
1203
EXPORT_SYMBOL_GPL(kvm_get_dirty_log);
A
Avi Kivity 已提交
1204

1205 1206
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
/**
J
Jiang Biao 已提交
1207
 * kvm_get_dirty_log_protect - get a snapshot of dirty pages
1208
 *	and reenable dirty page tracking for the corresponding pages.
1209 1210
 * @kvm:	pointer to kvm instance
 * @log:	slot id and address to which we copy the log
J
Jiang Biao 已提交
1211
 * @flush:	true if TLB flush is needed by caller
1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228
 *
 * We need to keep it in mind that VCPU threads can write to the bitmap
 * concurrently. So, to avoid losing track of dirty pages we keep the
 * following order:
 *
 *    1. Take a snapshot of the bit and clear it if needed.
 *    2. Write protect the corresponding page.
 *    3. Copy the snapshot to the userspace.
 *    4. Upon return caller flushes TLB's if needed.
 *
 * Between 2 and 4, the guest may write to the page using the remaining TLB
 * entry.  This is not a problem because the page is reported dirty using
 * the snapshot taken before and step 4 ensures that writes done after
 * exiting to userspace will be logged for the next call.
 *
 */
int kvm_get_dirty_log_protect(struct kvm *kvm,
1229
			struct kvm_dirty_log *log, bool *flush)
1230
{
1231
	struct kvm_memslots *slots;
1232
	struct kvm_memory_slot *memslot;
1233
	int i, as_id, id;
1234 1235 1236 1237
	unsigned long n;
	unsigned long *dirty_bitmap;
	unsigned long *dirty_bitmap_buffer;

1238 1239 1240
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1241
		return -EINVAL;
1242

1243 1244
	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
1245 1246 1247

	dirty_bitmap = memslot->dirty_bitmap;
	if (!dirty_bitmap)
1248
		return -ENOENT;
1249 1250

	n = kvm_dirty_bitmap_bytes(memslot);
1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264
	*flush = false;
	if (kvm->manual_dirty_log_protect) {
		/*
		 * Unlike kvm_get_dirty_log, we always return false in *flush,
		 * because no flush is needed until KVM_CLEAR_DIRTY_LOG.  There
		 * is some code duplication between this function and
		 * kvm_get_dirty_log, but hopefully all architecture
		 * transition to kvm_get_dirty_log_protect and kvm_get_dirty_log
		 * can be eliminated.
		 */
		dirty_bitmap_buffer = dirty_bitmap;
	} else {
		dirty_bitmap_buffer = kvm_second_dirty_bitmap(memslot);
		memset(dirty_bitmap_buffer, 0, n);
1265

1266 1267 1268 1269
		spin_lock(&kvm->mmu_lock);
		for (i = 0; i < n / sizeof(long); i++) {
			unsigned long mask;
			gfn_t offset;
1270

1271 1272 1273 1274 1275 1276 1277
			if (!dirty_bitmap[i])
				continue;

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

1278 1279 1280
			offset = i * BITS_PER_LONG;
			kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot,
								offset, mask);
1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295
		}
		spin_unlock(&kvm->mmu_lock);
	}

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

/**
 * kvm_clear_dirty_log_protect - clear dirty bits in the bitmap
 *	and reenable dirty page tracking for the corresponding pages.
 * @kvm:	pointer to kvm instance
 * @log:	slot id and address from which to fetch the bitmap of dirty pages
J
Jiang Biao 已提交
1296
 * @flush:	true if TLB flush is needed by caller
1297 1298 1299 1300 1301 1302
 */
int kvm_clear_dirty_log_protect(struct kvm *kvm,
				struct kvm_clear_dirty_log *log, bool *flush)
{
	struct kvm_memslots *slots;
	struct kvm_memory_slot *memslot;
1303
	int as_id, id;
1304
	gfn_t offset;
1305
	unsigned long i, n;
1306 1307 1308 1309 1310 1311 1312 1313
	unsigned long *dirty_bitmap;
	unsigned long *dirty_bitmap_buffer;

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

1314
	if (log->first_page & 63)
1315 1316 1317 1318 1319 1320 1321 1322 1323
		return -EINVAL;

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

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

1324
	n = ALIGN(log->num_pages, BITS_PER_LONG) / 8;
1325 1326

	if (log->first_page > memslot->npages ||
1327 1328 1329
	    log->num_pages > memslot->npages - log->first_page ||
	    (log->num_pages < memslot->npages - log->first_page && (log->num_pages & 63)))
	    return -EINVAL;
1330

1331
	*flush = false;
1332 1333 1334
	dirty_bitmap_buffer = kvm_second_dirty_bitmap(memslot);
	if (copy_from_user(dirty_bitmap_buffer, log->dirty_bitmap, n))
		return -EFAULT;
1335

1336
	spin_lock(&kvm->mmu_lock);
1337 1338
	for (offset = log->first_page, i = offset / BITS_PER_LONG,
		 n = DIV_ROUND_UP(log->num_pages, BITS_PER_LONG); n--;
1339 1340 1341 1342
	     i++, offset += BITS_PER_LONG) {
		unsigned long mask = *dirty_bitmap_buffer++;
		atomic_long_t *p = (atomic_long_t *) &dirty_bitmap[i];
		if (!mask)
1343 1344
			continue;

1345
		mask &= atomic_long_fetch_andnot(mask, p);
1346

1347 1348 1349 1350 1351 1352
		/*
		 * 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.
		*/
1353
		if (mask) {
1354
			*flush = true;
1355 1356 1357
			kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot,
								offset, mask);
		}
1358 1359
	}
	spin_unlock(&kvm->mmu_lock);
1360

1361
	return 0;
1362
}
1363
EXPORT_SYMBOL_GPL(kvm_clear_dirty_log_protect);
1364 1365
#endif

1366 1367 1368 1369 1370
bool kvm_largepages_enabled(void)
{
	return largepages_enabled;
}

1371 1372 1373 1374 1375 1376
void kvm_disable_largepages(void)
{
	largepages_enabled = false;
}
EXPORT_SYMBOL_GPL(kvm_disable_largepages);

1377 1378 1379 1380
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
1381
EXPORT_SYMBOL_GPL(gfn_to_memslot);
A
Avi Kivity 已提交
1382

1383 1384 1385 1386 1387
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);
}

1388
bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
1389
{
1390
	struct kvm_memory_slot *memslot = gfn_to_memslot(kvm, gfn);
1391

1392
	if (!memslot || memslot->id >= KVM_USER_MEM_SLOTS ||
1393
	      memslot->flags & KVM_MEMSLOT_INVALID)
1394
		return false;
1395

1396
	return true;
1397 1398 1399
}
EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);

1400
unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn)
J
Joerg Roedel 已提交
1401 1402 1403 1404 1405 1406
{
	struct vm_area_struct *vma;
	unsigned long addr, size;

	size = PAGE_SIZE;

1407
	addr = kvm_vcpu_gfn_to_hva_prot(vcpu, gfn, NULL);
J
Joerg Roedel 已提交
1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423
	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
1424 1425 1426 1427 1428 1429 1430
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
1431
{
1432
	if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
X
Xiao Guangrong 已提交
1433
		return KVM_HVA_ERR_BAD;
1434

X
Xiao Guangrong 已提交
1435 1436
	if (memslot_is_readonly(slot) && write)
		return KVM_HVA_ERR_RO_BAD;
1437 1438 1439 1440

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

X
Xiao Guangrong 已提交
1441
	return __gfn_to_hva_memslot(slot, gfn);
I
Izik Eidus 已提交
1442
}
1443

X
Xiao Guangrong 已提交
1444 1445 1446 1447
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
1448
}
1449

X
Xiao Guangrong 已提交
1450
unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot,
1451
					gfn_t gfn)
X
Xiao Guangrong 已提交
1452 1453 1454 1455 1456
{
	return gfn_to_hva_many(slot, gfn, NULL);
}
EXPORT_SYMBOL_GPL(gfn_to_hva_memslot);

1457 1458
unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
{
1459
	return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL);
1460
}
1461
EXPORT_SYMBOL_GPL(gfn_to_hva);
I
Izik Eidus 已提交
1462

1463 1464 1465 1466 1467 1468
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);

1469
/*
1470 1471 1472 1473 1474 1475
 * 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
1476
 */
1477 1478
unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot,
				      gfn_t gfn, bool *writable)
1479
{
1480 1481 1482
	unsigned long hva = __gfn_to_hva_many(slot, gfn, NULL, false);

	if (!kvm_is_error_hva(hva) && writable)
1483 1484
		*writable = !memslot_is_readonly(slot);

1485
	return hva;
1486 1487
}

1488 1489 1490 1491 1492 1493 1494
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);
}

1495 1496 1497 1498 1499 1500 1501
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);
}

1502 1503
static inline int check_user_page_hwpoison(unsigned long addr)
{
L
Lorenzo Stoakes 已提交
1504
	int rc, flags = FOLL_HWPOISON | FOLL_WRITE;
1505

L
Lorenzo Stoakes 已提交
1506
	rc = get_user_pages(addr, 1, flags, NULL, NULL);
1507 1508 1509
	return rc == -EHWPOISON;
}

X
Xiao Guangrong 已提交
1510
/*
1511 1512
 * 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
1513
 * only part that runs if we can in atomic context.
X
Xiao Guangrong 已提交
1514
 */
1515 1516
static bool hva_to_pfn_fast(unsigned long addr, bool write_fault,
			    bool *writable, kvm_pfn_t *pfn)
A
Avi Kivity 已提交
1517
{
1518
	struct page *page[1];
X
Xiao Guangrong 已提交
1519
	int npages;
A
Avi Kivity 已提交
1520

1521 1522 1523 1524 1525 1526 1527
	/*
	 * 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;
1528

X
Xiao Guangrong 已提交
1529 1530 1531
	npages = __get_user_pages_fast(addr, 1, 1, page);
	if (npages == 1) {
		*pfn = page_to_pfn(page[0]);
1532

X
Xiao Guangrong 已提交
1533 1534 1535 1536
		if (writable)
			*writable = true;
		return true;
	}
1537

X
Xiao Guangrong 已提交
1538 1539
	return false;
}
1540

X
Xiao Guangrong 已提交
1541 1542 1543 1544 1545
/*
 * 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
1546
			   bool *writable, kvm_pfn_t *pfn)
X
Xiao Guangrong 已提交
1547
{
1548 1549
	unsigned int flags = FOLL_HWPOISON;
	struct page *page;
X
Xiao Guangrong 已提交
1550
	int npages = 0;
1551

X
Xiao Guangrong 已提交
1552 1553 1554 1555 1556
	might_sleep();

	if (writable)
		*writable = write_fault;

1557 1558 1559 1560
	if (write_fault)
		flags |= FOLL_WRITE;
	if (async)
		flags |= FOLL_NOWAIT;
1561

1562
	npages = get_user_pages_unlocked(addr, 1, &page, flags);
X
Xiao Guangrong 已提交
1563 1564 1565 1566
	if (npages != 1)
		return npages;

	/* map read fault as writable if possible */
1567
	if (unlikely(!write_fault) && writable) {
1568
		struct page *wpage;
X
Xiao Guangrong 已提交
1569

1570
		if (__get_user_pages_fast(addr, 1, 1, &wpage) == 1) {
X
Xiao Guangrong 已提交
1571
			*writable = true;
1572 1573
			put_page(page);
			page = wpage;
1574
		}
1575
	}
1576
	*pfn = page_to_pfn(page);
X
Xiao Guangrong 已提交
1577 1578
	return npages;
}
I
Izik Eidus 已提交
1579

X
Xiao Guangrong 已提交
1580 1581 1582 1583
static bool vma_is_valid(struct vm_area_struct *vma, bool write_fault)
{
	if (unlikely(!(vma->vm_flags & VM_READ)))
		return false;
1584

X
Xiao Guangrong 已提交
1585 1586
	if (write_fault && (unlikely(!(vma->vm_flags & VM_WRITE))))
		return false;
1587

X
Xiao Guangrong 已提交
1588 1589
	return true;
}
1590

1591 1592
static int hva_to_pfn_remapped(struct vm_area_struct *vma,
			       unsigned long addr, bool *async,
1593 1594
			       bool write_fault, bool *writable,
			       kvm_pfn_t *p_pfn)
1595
{
1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619
	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;

	}

1620 1621
	if (writable)
		*writable = true;
1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636

	/*
	 * 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;
1637 1638 1639
	return 0;
}

1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653
/*
 * 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
1654
static kvm_pfn_t hva_to_pfn(unsigned long addr, bool atomic, bool *async,
X
Xiao Guangrong 已提交
1655 1656 1657
			bool write_fault, bool *writable)
{
	struct vm_area_struct *vma;
D
Dan Williams 已提交
1658
	kvm_pfn_t pfn = 0;
1659
	int npages, r;
1660

X
Xiao Guangrong 已提交
1661 1662
	/* we can do it either atomically or asynchronously, not both */
	BUG_ON(atomic && async);
1663

1664
	if (hva_to_pfn_fast(addr, write_fault, writable, &pfn))
X
Xiao Guangrong 已提交
1665 1666 1667 1668 1669 1670 1671 1672
		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;
1673

X
Xiao Guangrong 已提交
1674 1675 1676 1677 1678 1679 1680
	down_read(&current->mm->mmap_sem);
	if (npages == -EHWPOISON ||
	      (!async && check_user_page_hwpoison(addr))) {
		pfn = KVM_PFN_ERR_HWPOISON;
		goto exit;
	}

1681
retry:
X
Xiao Guangrong 已提交
1682 1683 1684 1685
	vma = find_vma_intersection(current->mm, addr, addr + 1);

	if (vma == NULL)
		pfn = KVM_PFN_ERR_FAULT;
1686
	else if (vma->vm_flags & (VM_IO | VM_PFNMAP)) {
1687
		r = hva_to_pfn_remapped(vma, addr, async, write_fault, writable, &pfn);
1688 1689
		if (r == -EAGAIN)
			goto retry;
1690 1691
		if (r < 0)
			pfn = KVM_PFN_ERR_FAULT;
X
Xiao Guangrong 已提交
1692
	} else {
X
Xiao Guangrong 已提交
1693
		if (async && vma_is_valid(vma, write_fault))
X
Xiao Guangrong 已提交
1694 1695 1696 1697 1698
			*async = true;
		pfn = KVM_PFN_ERR_FAULT;
	}
exit:
	up_read(&current->mm->mmap_sem);
1699
	return pfn;
1700 1701
}

D
Dan Williams 已提交
1702 1703 1704
kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
			       bool atomic, bool *async, bool write_fault,
			       bool *writable)
1705
{
X
Xiao Guangrong 已提交
1706 1707
	unsigned long addr = __gfn_to_hva_many(slot, gfn, NULL, write_fault);

1708 1709 1710
	if (addr == KVM_HVA_ERR_RO_BAD) {
		if (writable)
			*writable = false;
X
Xiao Guangrong 已提交
1711
		return KVM_PFN_ERR_RO_FAULT;
1712
	}
X
Xiao Guangrong 已提交
1713

1714 1715 1716
	if (kvm_is_error_hva(addr)) {
		if (writable)
			*writable = false;
1717
		return KVM_PFN_NOSLOT;
1718
	}
X
Xiao Guangrong 已提交
1719 1720 1721 1722 1723 1724 1725 1726 1727

	/* 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);
1728
}
1729
EXPORT_SYMBOL_GPL(__gfn_to_pfn_memslot);
1730

D
Dan Williams 已提交
1731
kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1732 1733
		      bool *writable)
{
P
Paolo Bonzini 已提交
1734 1735
	return __gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn, false, NULL,
				    write_fault, writable);
1736 1737 1738
}
EXPORT_SYMBOL_GPL(gfn_to_pfn_prot);

D
Dan Williams 已提交
1739
kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1740
{
X
Xiao Guangrong 已提交
1741
	return __gfn_to_pfn_memslot(slot, gfn, false, NULL, true, NULL);
1742
}
P
Paolo Bonzini 已提交
1743
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot);
1744

D
Dan Williams 已提交
1745
kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn)
1746
{
X
Xiao Guangrong 已提交
1747
	return __gfn_to_pfn_memslot(slot, gfn, true, NULL, true, NULL);
1748
}
1749
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot_atomic);
1750

D
Dan Williams 已提交
1751
kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1752 1753 1754 1755 1756
{
	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
1757
kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn)
1758 1759 1760 1761 1762
{
	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
1763
kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1764 1765 1766 1767 1768
{
	return gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn);
}
EXPORT_SYMBOL_GPL(gfn_to_pfn);

D
Dan Williams 已提交
1769
kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn)
1770 1771 1772 1773 1774
{
	return gfn_to_pfn_memslot(kvm_vcpu_gfn_to_memslot(vcpu, gfn), gfn);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_gfn_to_pfn);

1775 1776
int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
			    struct page **pages, int nr_pages)
1777 1778
{
	unsigned long addr;
1779
	gfn_t entry = 0;
1780

1781
	addr = gfn_to_hva_many(slot, gfn, &entry);
1782 1783 1784 1785 1786 1787 1788 1789 1790 1791
	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
1792
static struct page *kvm_pfn_to_page(kvm_pfn_t pfn)
X
Xiao Guangrong 已提交
1793
{
1794
	if (is_error_noslot_pfn(pfn))
1795
		return KVM_ERR_PTR_BAD_PAGE;
X
Xiao Guangrong 已提交
1796

1797
	if (kvm_is_reserved_pfn(pfn)) {
1798
		WARN_ON(1);
1799
		return KVM_ERR_PTR_BAD_PAGE;
1800
	}
X
Xiao Guangrong 已提交
1801 1802 1803 1804

	return pfn_to_page(pfn);
}

1805 1806
struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
{
D
Dan Williams 已提交
1807
	kvm_pfn_t pfn;
1808 1809 1810

	pfn = gfn_to_pfn(kvm, gfn);

X
Xiao Guangrong 已提交
1811
	return kvm_pfn_to_page(pfn);
A
Avi Kivity 已提交
1812 1813 1814
}
EXPORT_SYMBOL_GPL(gfn_to_page);

1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839
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;
}

1840
static int __kvm_map_gfn(struct kvm_memslots *slots, gfn_t gfn,
1841 1842 1843
			 struct kvm_host_map *map,
			 struct gfn_to_pfn_cache *cache,
			 bool atomic)
1844 1845 1846 1847
{
	kvm_pfn_t pfn;
	void *hva = NULL;
	struct page *page = KVM_UNMAPPED_PAGE;
1848
	struct kvm_memory_slot *slot = __gfn_to_memslot(slots, gfn);
1849
	u64 gen = slots->generation;
1850 1851 1852 1853

	if (!map)
		return -EINVAL;

1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866
	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);
	}
1867 1868 1869 1870 1871
	if (is_error_noslot_pfn(pfn))
		return -EINVAL;

	if (pfn_valid(pfn)) {
		page = pfn_to_page(pfn);
1872 1873 1874 1875
		if (atomic)
			hva = kmap_atomic(page);
		else
			hva = kmap(page);
P
Paolo Bonzini 已提交
1876
#ifdef CONFIG_HAS_IOMEM
1877
	} else if (!atomic) {
1878
		hva = memremap(pfn_to_hpa(pfn), PAGE_SIZE, MEMREMAP_WB);
1879 1880
	} else {
		return -EINVAL;
P
Paolo Bonzini 已提交
1881
#endif
1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894
	}

	if (!hva)
		return -EFAULT;

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

	return 0;
}

1895 1896
int kvm_map_gfn(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map,
		struct gfn_to_pfn_cache *cache, bool atomic)
1897
{
1898 1899
	return __kvm_map_gfn(kvm_memslots(vcpu->kvm), gfn, map,
			cache, atomic);
1900 1901 1902
}
EXPORT_SYMBOL_GPL(kvm_map_gfn);

1903 1904
int kvm_vcpu_map(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map)
{
1905 1906
	return __kvm_map_gfn(kvm_vcpu_memslots(vcpu), gfn, map,
		NULL, false);
1907 1908 1909
}
EXPORT_SYMBOL_GPL(kvm_vcpu_map);

1910
static void __kvm_unmap_gfn(struct kvm_memory_slot *memslot,
1911 1912 1913
			struct kvm_host_map *map,
			struct gfn_to_pfn_cache *cache,
			bool dirty, bool atomic)
1914 1915 1916 1917 1918 1919 1920
{
	if (!map)
		return;

	if (!map->hva)
		return;

1921 1922 1923 1924 1925 1926
	if (map->page != KVM_UNMAPPED_PAGE) {
		if (atomic)
			kunmap_atomic(map->hva);
		else
			kunmap(map->page);
	}
1927
#ifdef CONFIG_HAS_IOMEM
1928
	else if (!atomic)
1929
		memunmap(map->hva);
1930 1931
	else
		WARN_ONCE(1, "Unexpected unmapping in atomic context");
1932
#endif
1933

1934
	if (dirty)
1935
		mark_page_dirty_in_slot(memslot, map->gfn);
1936 1937 1938 1939 1940

	if (cache)
		cache->dirty |= dirty;
	else
		kvm_release_pfn(map->pfn, dirty, NULL);
1941 1942 1943 1944

	map->hva = NULL;
	map->page = NULL;
}
1945

1946 1947
int kvm_unmap_gfn(struct kvm_vcpu *vcpu, struct kvm_host_map *map, 
		  struct gfn_to_pfn_cache *cache, bool dirty, bool atomic)
1948
{
1949 1950
	__kvm_unmap_gfn(gfn_to_memslot(vcpu->kvm, map->gfn), map,
			cache, dirty, atomic);
1951 1952 1953 1954 1955 1956
	return 0;
}
EXPORT_SYMBOL_GPL(kvm_unmap_gfn);

void kvm_vcpu_unmap(struct kvm_vcpu *vcpu, struct kvm_host_map *map, bool dirty)
{
1957 1958
	__kvm_unmap_gfn(kvm_vcpu_gfn_to_memslot(vcpu, map->gfn), map, NULL,
			dirty, false);
1959
}
1960 1961
EXPORT_SYMBOL_GPL(kvm_vcpu_unmap);

1962 1963
struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn)
{
D
Dan Williams 已提交
1964
	kvm_pfn_t pfn;
1965 1966 1967 1968 1969 1970 1971

	pfn = kvm_vcpu_gfn_to_pfn(vcpu, gfn);

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

1972 1973
void kvm_release_page_clean(struct page *page)
{
1974 1975
	WARN_ON(is_error_page(page));

1976
	kvm_release_pfn_clean(page_to_pfn(page));
1977 1978 1979
}
EXPORT_SYMBOL_GPL(kvm_release_page_clean);

D
Dan Williams 已提交
1980
void kvm_release_pfn_clean(kvm_pfn_t pfn)
1981
{
1982
	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn))
1983
		put_page(pfn_to_page(pfn));
1984 1985 1986
}
EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);

1987
void kvm_release_page_dirty(struct page *page)
1988
{
X
Xiao Guangrong 已提交
1989 1990
	WARN_ON(is_error_page(page));

1991 1992 1993 1994
	kvm_release_pfn_dirty(page_to_pfn(page));
}
EXPORT_SYMBOL_GPL(kvm_release_page_dirty);

1995
void kvm_release_pfn_dirty(kvm_pfn_t pfn)
1996 1997 1998 1999
{
	kvm_set_pfn_dirty(pfn);
	kvm_release_pfn_clean(pfn);
}
2000
EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
2001

D
Dan Williams 已提交
2002
void kvm_set_pfn_dirty(kvm_pfn_t pfn)
2003
{
2004 2005
	if (!kvm_is_reserved_pfn(pfn) && !kvm_is_zone_device_pfn(pfn))
		SetPageDirty(pfn_to_page(pfn));
2006
}
2007 2008
EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);

D
Dan Williams 已提交
2009
void kvm_set_pfn_accessed(kvm_pfn_t pfn)
2010
{
2011
	if (!kvm_is_reserved_pfn(pfn) && !kvm_is_zone_device_pfn(pfn))
2012
		mark_page_accessed(pfn_to_page(pfn));
2013 2014 2015
}
EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);

D
Dan Williams 已提交
2016
void kvm_get_pfn(kvm_pfn_t pfn)
2017
{
2018
	if (!kvm_is_reserved_pfn(pfn))
2019
		get_page(pfn_to_page(pfn));
2020 2021
}
EXPORT_SYMBOL_GPL(kvm_get_pfn);
2022

2023 2024 2025 2026 2027 2028 2029 2030
static int next_segment(unsigned long len, int offset)
{
	if (len > PAGE_SIZE - offset)
		return PAGE_SIZE - offset;
	else
		return len;
}

2031 2032
static int __kvm_read_guest_page(struct kvm_memory_slot *slot, gfn_t gfn,
				 void *data, int offset, int len)
2033
{
2034 2035
	int r;
	unsigned long addr;
2036

2037
	addr = gfn_to_hva_memslot_prot(slot, gfn, NULL);
2038 2039
	if (kvm_is_error_hva(addr))
		return -EFAULT;
2040
	r = __copy_from_user(data, (void __user *)addr + offset, len);
2041
	if (r)
2042 2043 2044
		return -EFAULT;
	return 0;
}
2045 2046 2047 2048 2049 2050 2051 2052

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);
}
2053 2054
EXPORT_SYMBOL_GPL(kvm_read_guest_page);

2055 2056 2057 2058 2059 2060 2061 2062 2063
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);

2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083
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);

2084
int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data, unsigned long len)
2085 2086
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
2087
	int seg;
2088
	int offset = offset_in_page(gpa);
2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102
	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);
2103

2104 2105 2106 2107 2108 2109 2110
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);
2111 2112
	if (kvm_is_error_hva(addr))
		return -EFAULT;
2113
	pagefault_disable();
2114
	r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
2115
	pagefault_enable();
2116 2117 2118 2119 2120
	if (r)
		return -EFAULT;
	return 0;
}

2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133
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)
2134
{
2135 2136
	int r;
	unsigned long addr;
2137

2138
	addr = gfn_to_hva_memslot(memslot, gfn);
2139 2140
	if (kvm_is_error_hva(addr))
		return -EFAULT;
2141
	r = __copy_to_user((void __user *)addr + offset, data, len);
2142
	if (r)
2143
		return -EFAULT;
2144
	mark_page_dirty_in_slot(memslot, gfn);
2145 2146
	return 0;
}
2147 2148 2149 2150 2151 2152 2153 2154

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);
}
2155 2156
EXPORT_SYMBOL_GPL(kvm_write_guest_page);

2157 2158 2159 2160 2161 2162 2163 2164 2165
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);

2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184
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;
}
2185
EXPORT_SYMBOL_GPL(kvm_write_guest);
2186

2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207
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);

2208 2209 2210
static int __kvm_gfn_to_hva_cache_init(struct kvm_memslots *slots,
				       struct gfn_to_hva_cache *ghc,
				       gpa_t gpa, unsigned long len)
2211 2212
{
	int offset = offset_in_page(gpa);
2213 2214 2215 2216
	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;
2217

2218
	/* Update ghc->generation before performing any error checks. */
2219
	ghc->generation = slots->generation;
2220 2221 2222 2223 2224

	if (start_gfn > end_gfn) {
		ghc->hva = KVM_HVA_ERR_BAD;
		return -EINVAL;
	}
2225 2226 2227 2228 2229

	/*
	 * If the requested region crosses two memslots, we still
	 * verify that the entire region is valid here.
	 */
2230
	for ( ; start_gfn <= end_gfn; start_gfn += nr_pages_avail) {
2231 2232 2233 2234
		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))
2235
			return -EFAULT;
2236 2237 2238
	}

	/* Use the slow path for cross page reads and writes. */
2239
	if (nr_pages_needed == 1)
2240
		ghc->hva += offset;
2241
	else
2242
		ghc->memslot = NULL;
2243

2244 2245 2246
	ghc->gpa = gpa;
	ghc->len = len;
	return 0;
2247
}
2248

2249
int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
2250 2251
			      gpa_t gpa, unsigned long len)
{
2252
	struct kvm_memslots *slots = kvm_memslots(kvm);
2253 2254
	return __kvm_gfn_to_hva_cache_init(slots, ghc, gpa, len);
}
2255
EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init);
2256

2257
int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
2258 2259
				  void *data, unsigned int offset,
				  unsigned long len)
2260
{
2261
	struct kvm_memslots *slots = kvm_memslots(kvm);
2262
	int r;
2263
	gpa_t gpa = ghc->gpa + offset;
2264

2265
	BUG_ON(len + offset > ghc->len);
2266

2267 2268 2269 2270
	if (slots->generation != ghc->generation) {
		if (__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len))
			return -EFAULT;
	}
2271

2272 2273 2274
	if (kvm_is_error_hva(ghc->hva))
		return -EFAULT;

2275 2276 2277
	if (unlikely(!ghc->memslot))
		return kvm_write_guest(kvm, gpa, data, len);

2278
	r = __copy_to_user((void __user *)ghc->hva + offset, data, len);
2279 2280
	if (r)
		return -EFAULT;
2281
	mark_page_dirty_in_slot(ghc->memslot, gpa >> PAGE_SHIFT);
2282 2283 2284

	return 0;
}
2285
EXPORT_SYMBOL_GPL(kvm_write_guest_offset_cached);
2286

2287 2288
int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
2289
{
2290
	return kvm_write_guest_offset_cached(kvm, ghc, data, 0, len);
2291
}
2292
EXPORT_SYMBOL_GPL(kvm_write_guest_cached);
2293

2294 2295
int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
2296
{
2297
	struct kvm_memslots *slots = kvm_memslots(kvm);
2298 2299
	int r;

2300 2301
	BUG_ON(len > ghc->len);

2302 2303 2304 2305
	if (slots->generation != ghc->generation) {
		if (__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len))
			return -EFAULT;
	}
2306

2307 2308 2309
	if (kvm_is_error_hva(ghc->hva))
		return -EFAULT;

2310 2311 2312
	if (unlikely(!ghc->memslot))
		return kvm_read_guest(kvm, ghc->gpa, data, len);

2313 2314 2315 2316 2317 2318
	r = __copy_from_user(data, (void __user *)ghc->hva, len);
	if (r)
		return -EFAULT;

	return 0;
}
2319
EXPORT_SYMBOL_GPL(kvm_read_guest_cached);
2320

2321 2322
int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
{
2323 2324 2325
	const void *zero_page = (const void *) __va(page_to_phys(ZERO_PAGE(0)));

	return kvm_write_guest_page(kvm, gfn, zero_page, offset, len);
2326 2327 2328 2329 2330 2331 2332 2333 2334 2335
}
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;

2336
	while ((seg = next_segment(len, offset)) != 0) {
2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347
		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);

2348
static void mark_page_dirty_in_slot(struct kvm_memory_slot *memslot,
2349
				    gfn_t gfn)
A
Avi Kivity 已提交
2350
{
R
Rusty Russell 已提交
2351 2352
	if (memslot && memslot->dirty_bitmap) {
		unsigned long rel_gfn = gfn - memslot->base_gfn;
A
Avi Kivity 已提交
2353

2354
		set_bit_le(rel_gfn, memslot->dirty_bitmap);
A
Avi Kivity 已提交
2355 2356 2357
	}
}

2358 2359 2360 2361 2362
void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
{
	struct kvm_memory_slot *memslot;

	memslot = gfn_to_memslot(kvm, gfn);
2363
	mark_page_dirty_in_slot(memslot, gfn);
2364
}
2365
EXPORT_SYMBOL_GPL(mark_page_dirty);
2366

2367 2368 2369 2370 2371 2372 2373 2374 2375
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);

2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398
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
2399 2400
static void grow_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2401
	unsigned int old, val, grow, grow_start;
W
Wanpeng Li 已提交
2402

2403
	old = val = vcpu->halt_poll_ns;
2404
	grow_start = READ_ONCE(halt_poll_ns_grow_start);
2405
	grow = READ_ONCE(halt_poll_ns_grow);
2406 2407 2408
	if (!grow)
		goto out;

2409 2410 2411
	val *= grow;
	if (val < grow_start)
		val = grow_start;
W
Wanpeng Li 已提交
2412

2413 2414 2415
	if (val > halt_poll_ns)
		val = halt_poll_ns;

W
Wanpeng Li 已提交
2416
	vcpu->halt_poll_ns = val;
2417
out:
2418
	trace_kvm_halt_poll_ns_grow(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2419 2420 2421 2422
}

static void shrink_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2423
	unsigned int old, val, shrink;
W
Wanpeng Li 已提交
2424

2425
	old = val = vcpu->halt_poll_ns;
2426 2427
	shrink = READ_ONCE(halt_poll_ns_shrink);
	if (shrink == 0)
W
Wanpeng Li 已提交
2428 2429
		val = 0;
	else
2430
		val /= shrink;
W
Wanpeng Li 已提交
2431 2432

	vcpu->halt_poll_ns = val;
2433
	trace_kvm_halt_poll_ns_shrink(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2434 2435
}

2436 2437
static int kvm_vcpu_check_block(struct kvm_vcpu *vcpu)
{
2438 2439 2440
	int ret = -EINTR;
	int idx = srcu_read_lock(&vcpu->kvm->srcu);

2441 2442
	if (kvm_arch_vcpu_runnable(vcpu)) {
		kvm_make_request(KVM_REQ_UNHALT, vcpu);
2443
		goto out;
2444 2445
	}
	if (kvm_cpu_has_pending_timer(vcpu))
2446
		goto out;
2447
	if (signal_pending(current))
2448
		goto out;
2449

2450 2451 2452 2453
	ret = 0;
out:
	srcu_read_unlock(&vcpu->kvm->srcu, idx);
	return ret;
2454 2455
}

E
Eddie Dong 已提交
2456 2457 2458
/*
 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
 */
2459
void kvm_vcpu_block(struct kvm_vcpu *vcpu)
2460
{
2461
	ktime_t start, cur;
2462
	DECLARE_SWAITQUEUE(wait);
2463
	bool waited = false;
W
Wanpeng Li 已提交
2464
	u64 block_ns;
2465

2466 2467
	kvm_arch_vcpu_blocking(vcpu);

2468
	start = cur = ktime_get();
2469
	if (vcpu->halt_poll_ns && !kvm_arch_no_poll(vcpu)) {
W
Wanpeng Li 已提交
2470
		ktime_t stop = ktime_add_ns(ktime_get(), vcpu->halt_poll_ns);
2471

2472
		++vcpu->stat.halt_attempted_poll;
2473 2474 2475 2476 2477 2478 2479
		do {
			/*
			 * This sets KVM_REQ_UNHALT if an interrupt
			 * arrives.
			 */
			if (kvm_vcpu_check_block(vcpu) < 0) {
				++vcpu->stat.halt_successful_poll;
2480 2481
				if (!vcpu_valid_wakeup(vcpu))
					++vcpu->stat.halt_poll_invalid;
2482 2483 2484 2485 2486
				goto out;
			}
			cur = ktime_get();
		} while (single_task_running() && ktime_before(cur, stop));
	}
2487 2488

	for (;;) {
2489
		prepare_to_swait_exclusive(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
2490

2491
		if (kvm_vcpu_check_block(vcpu) < 0)
2492 2493
			break;

2494
		waited = true;
E
Eddie Dong 已提交
2495 2496
		schedule();
	}
2497

2498
	finish_swait(&vcpu->wq, &wait);
2499 2500
	cur = ktime_get();
out:
2501
	kvm_arch_vcpu_unblocking(vcpu);
W
Wanpeng Li 已提交
2502 2503
	block_ns = ktime_to_ns(cur) - ktime_to_ns(start);

2504 2505
	if (!kvm_arch_no_poll(vcpu)) {
		if (!vcpu_valid_wakeup(vcpu)) {
W
Wanpeng Li 已提交
2506
			shrink_halt_poll_ns(vcpu);
2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520
		} 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
2521

2522 2523
	trace_kvm_vcpu_wakeup(block_ns, waited, vcpu_valid_wakeup(vcpu));
	kvm_arch_vcpu_block_finish(vcpu);
E
Eddie Dong 已提交
2524
}
2525
EXPORT_SYMBOL_GPL(kvm_vcpu_block);
E
Eddie Dong 已提交
2526

2527
bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu)
2528
{
2529
	struct swait_queue_head *wqp;
2530 2531

	wqp = kvm_arch_vcpu_wq(vcpu);
2532
	if (swq_has_sleeper(wqp)) {
2533
		swake_up_one(wqp);
2534
		WRITE_ONCE(vcpu->ready, true);
2535
		++vcpu->stat.halt_wakeup;
2536
		return true;
2537 2538
	}

2539
	return false;
2540 2541 2542
}
EXPORT_SYMBOL_GPL(kvm_vcpu_wake_up);

2543
#ifndef CONFIG_S390
2544 2545 2546 2547 2548 2549 2550 2551
/*
 * 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;

2552 2553 2554
	if (kvm_vcpu_wake_up(vcpu))
		return;

2555 2556 2557 2558 2559 2560
	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();
}
2561
EXPORT_SYMBOL_GPL(kvm_vcpu_kick);
2562
#endif /* !CONFIG_S390 */
2563

2564
int kvm_vcpu_yield_to(struct kvm_vcpu *target)
2565 2566 2567
{
	struct pid *pid;
	struct task_struct *task = NULL;
2568
	int ret = 0;
2569 2570 2571 2572

	rcu_read_lock();
	pid = rcu_dereference(target->pid);
	if (pid)
2573
		task = get_pid_task(pid, PIDTYPE_PID);
2574 2575
	rcu_read_unlock();
	if (!task)
2576 2577
		return ret;
	ret = yield_to(task, 1);
2578
	put_task_struct(task);
2579 2580

	return ret;
2581 2582 2583
}
EXPORT_SYMBOL_GPL(kvm_vcpu_yield_to);

2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605
/*
 * 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.
 */
2606
static bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
2607
{
2608
#ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
2609 2610 2611
	bool eligible;

	eligible = !vcpu->spin_loop.in_spin_loop ||
2612
		    vcpu->spin_loop.dy_eligible;
2613 2614 2615 2616 2617

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

	return eligible;
2618 2619
#else
	return true;
2620
#endif
2621
}
2622

2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645
/*
 * 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;
}

2646
void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
Z
Zhai, Edwin 已提交
2647
{
2648 2649 2650 2651
	struct kvm *kvm = me->kvm;
	struct kvm_vcpu *vcpu;
	int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
	int yielded = 0;
2652
	int try = 3;
2653 2654
	int pass;
	int i;
Z
Zhai, Edwin 已提交
2655

2656
	kvm_vcpu_set_in_spin_loop(me, true);
2657 2658 2659 2660 2661 2662 2663
	/*
	 * 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.
	 */
2664
	for (pass = 0; pass < 2 && !yielded && try; pass++) {
2665
		kvm_for_each_vcpu(i, vcpu, kvm) {
2666
			if (!pass && i <= last_boosted_vcpu) {
2667 2668 2669 2670
				i = last_boosted_vcpu;
				continue;
			} else if (pass && i > last_boosted_vcpu)
				break;
2671
			if (!READ_ONCE(vcpu->ready))
2672
				continue;
2673 2674
			if (vcpu == me)
				continue;
2675
			if (swait_active(&vcpu->wq) && !vcpu_dy_runnable(vcpu))
2676
				continue;
2677 2678
			if (READ_ONCE(vcpu->preempted) && yield_to_kernel_mode &&
				!kvm_arch_vcpu_in_kernel(vcpu))
2679
				continue;
2680 2681
			if (!kvm_vcpu_eligible_for_directed_yield(vcpu))
				continue;
2682 2683 2684

			yielded = kvm_vcpu_yield_to(vcpu);
			if (yielded > 0) {
2685 2686
				kvm->last_boosted_vcpu = i;
				break;
2687 2688 2689 2690
			} else if (yielded < 0) {
				try--;
				if (!try)
					break;
2691 2692 2693
			}
		}
	}
2694
	kvm_vcpu_set_in_spin_loop(me, false);
2695 2696 2697

	/* Ensure vcpu is not eligible during next spinloop */
	kvm_vcpu_set_dy_eligible(me, false);
Z
Zhai, Edwin 已提交
2698 2699 2700
}
EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin);

2701
static vm_fault_t kvm_vcpu_fault(struct vm_fault *vmf)
2702
{
2703
	struct kvm_vcpu *vcpu = vmf->vma->vm_file->private_data;
2704 2705
	struct page *page;

2706
	if (vmf->pgoff == 0)
2707
		page = virt_to_page(vcpu->run);
A
Avi Kivity 已提交
2708
#ifdef CONFIG_X86
2709
	else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
2710
		page = virt_to_page(vcpu->arch.pio_data);
2711
#endif
2712
#ifdef CONFIG_KVM_MMIO
2713 2714
	else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
		page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
A
Avi Kivity 已提交
2715
#endif
2716
	else
2717
		return kvm_arch_vcpu_fault(vcpu, vmf);
2718
	get_page(page);
2719 2720
	vmf->page = page;
	return 0;
2721 2722
}

2723
static const struct vm_operations_struct kvm_vcpu_vm_ops = {
2724
	.fault = kvm_vcpu_fault,
2725 2726 2727 2728 2729 2730 2731 2732
};

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
2733 2734 2735 2736
static int kvm_vcpu_release(struct inode *inode, struct file *filp)
{
	struct kvm_vcpu *vcpu = filp->private_data;

2737
	debugfs_remove_recursive(vcpu->debugfs_dentry);
A
Al Viro 已提交
2738
	kvm_put_kvm(vcpu->kvm);
A
Avi Kivity 已提交
2739 2740 2741
	return 0;
}

2742
static struct file_operations kvm_vcpu_fops = {
A
Avi Kivity 已提交
2743 2744
	.release        = kvm_vcpu_release,
	.unlocked_ioctl = kvm_vcpu_ioctl,
2745
	.mmap           = kvm_vcpu_mmap,
2746
	.llseek		= noop_llseek,
2747
	KVM_COMPAT(kvm_vcpu_compat_ioctl),
A
Avi Kivity 已提交
2748 2749 2750 2751 2752 2753 2754
};

/*
 * Allocates an inode for the vcpu.
 */
static int create_vcpu_fd(struct kvm_vcpu *vcpu)
{
2755 2756 2757 2758
	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
2759 2760
}

2761
static void kvm_create_vcpu_debugfs(struct kvm_vcpu *vcpu)
2762
{
2763
#ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
2764 2765 2766
	char dir_name[ITOA_MAX_LEN * 2];

	if (!debugfs_initialized())
2767
		return;
2768 2769 2770

	snprintf(dir_name, sizeof(dir_name), "vcpu%d", vcpu->vcpu_id);
	vcpu->debugfs_dentry = debugfs_create_dir(dir_name,
2771
						  vcpu->kvm->debugfs_dentry);
2772

2773
	kvm_arch_create_vcpu_debugfs(vcpu);
2774
#endif
2775 2776
}

2777 2778 2779
/*
 * Creates some virtual cpus.  Good luck creating more than one.
 */
2780
static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
2781 2782
{
	int r;
2783
	struct kvm_vcpu *vcpu;
2784
	struct page *page;
2785

G
Greg Kurz 已提交
2786
	if (id >= KVM_MAX_VCPU_ID)
2787 2788
		return -EINVAL;

2789 2790 2791 2792 2793 2794 2795 2796 2797
	mutex_lock(&kvm->lock);
	if (kvm->created_vcpus == KVM_MAX_VCPUS) {
		mutex_unlock(&kvm->lock);
		return -EINVAL;
	}

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

2798 2799 2800 2801
	r = kvm_arch_vcpu_precreate(kvm, id);
	if (r)
		goto vcpu_decrement;

2802 2803 2804
	vcpu = kmem_cache_zalloc(kvm_vcpu_cache, GFP_KERNEL);
	if (!vcpu) {
		r = -ENOMEM;
2805 2806
		goto vcpu_decrement;
	}
2807

2808
	BUILD_BUG_ON(sizeof(struct kvm_run) > PAGE_SIZE);
2809 2810 2811
	page = alloc_page(GFP_KERNEL | __GFP_ZERO);
	if (!page) {
		r = -ENOMEM;
2812
		goto vcpu_free;
2813 2814 2815 2816
	}
	vcpu->run = page_address(page);

	kvm_vcpu_init(vcpu, kvm, id);
2817 2818 2819

	r = kvm_arch_vcpu_create(vcpu);
	if (r)
2820
		goto vcpu_free_run_page;
2821

2822
	kvm_create_vcpu_debugfs(vcpu);
2823

S
Shaohua Li 已提交
2824
	mutex_lock(&kvm->lock);
2825 2826 2827 2828
	if (kvm_get_vcpu_by_id(kvm, id)) {
		r = -EEXIST;
		goto unlock_vcpu_destroy;
	}
2829

2830 2831
	vcpu->vcpu_idx = atomic_read(&kvm->online_vcpus);
	BUG_ON(kvm->vcpus[vcpu->vcpu_idx]);
2832

R
Rusty Russell 已提交
2833
	/* Now it's all set up, let userspace reach it */
A
Al Viro 已提交
2834
	kvm_get_kvm(kvm);
A
Avi Kivity 已提交
2835
	r = create_vcpu_fd(vcpu);
2836
	if (r < 0) {
2837
		kvm_put_kvm_no_destroy(kvm);
2838
		goto unlock_vcpu_destroy;
2839 2840
	}

2841
	kvm->vcpus[vcpu->vcpu_idx] = vcpu;
2842 2843 2844 2845 2846

	/*
	 * Pairs with smp_rmb() in kvm_get_vcpu.  Write kvm->vcpus
	 * before kvm->online_vcpu's incremented value.
	 */
2847 2848 2849 2850
	smp_wmb();
	atomic_inc(&kvm->online_vcpus);

	mutex_unlock(&kvm->lock);
2851
	kvm_arch_vcpu_postcreate(vcpu);
R
Rusty Russell 已提交
2852
	return r;
2853

2854
unlock_vcpu_destroy:
2855
	mutex_unlock(&kvm->lock);
2856
	debugfs_remove_recursive(vcpu->debugfs_dentry);
2857
	kvm_arch_vcpu_destroy(vcpu);
2858 2859
vcpu_free_run_page:
	free_page((unsigned long)vcpu->run);
2860 2861
vcpu_free:
	kmem_cache_free(kvm_vcpu_cache, vcpu);
2862 2863 2864 2865
vcpu_decrement:
	mutex_lock(&kvm->lock);
	kvm->created_vcpus--;
	mutex_unlock(&kvm->lock);
2866 2867 2868
	return r;
}

A
Avi Kivity 已提交
2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879
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
2880 2881
static long kvm_vcpu_ioctl(struct file *filp,
			   unsigned int ioctl, unsigned long arg)
A
Avi Kivity 已提交
2882
{
A
Avi Kivity 已提交
2883
	struct kvm_vcpu *vcpu = filp->private_data;
A
Al Viro 已提交
2884
	void __user *argp = (void __user *)arg;
2885
	int r;
2886 2887
	struct kvm_fpu *fpu = NULL;
	struct kvm_sregs *kvm_sregs = NULL;
A
Avi Kivity 已提交
2888

2889 2890
	if (vcpu->kvm->mm != current->mm)
		return -EIO;
2891

2892 2893 2894
	if (unlikely(_IOC_TYPE(ioctl) != KVMIO))
		return -EINVAL;

2895
	/*
2896 2897
	 * Some architectures have vcpu ioctls that are asynchronous to vcpu
	 * execution; mutex_lock() would break them.
2898
	 */
2899 2900
	r = kvm_arch_vcpu_async_ioctl(filp, ioctl, arg);
	if (r != -ENOIOCTLCMD)
2901
		return r;
2902

2903 2904
	if (mutex_lock_killable(&vcpu->mutex))
		return -EINTR;
A
Avi Kivity 已提交
2905
	switch (ioctl) {
2906 2907
	case KVM_RUN: {
		struct pid *oldpid;
2908 2909 2910
		r = -EINVAL;
		if (arg)
			goto out;
2911
		oldpid = rcu_access_pointer(vcpu->pid);
2912
		if (unlikely(oldpid != task_pid(current))) {
2913
			/* The thread running this VCPU changed. */
2914
			struct pid *newpid;
2915

2916 2917 2918 2919 2920
			r = kvm_arch_vcpu_run_pid_change(vcpu);
			if (r)
				break;

			newpid = get_task_pid(current, PIDTYPE_PID);
2921 2922 2923 2924 2925
			rcu_assign_pointer(vcpu->pid, newpid);
			if (oldpid)
				synchronize_rcu();
			put_pid(oldpid);
		}
2926
		r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
2927
		trace_kvm_userspace_exit(vcpu->run->exit_reason, r);
A
Avi Kivity 已提交
2928
		break;
2929
	}
A
Avi Kivity 已提交
2930
	case KVM_GET_REGS: {
2931
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
2932

2933
		r = -ENOMEM;
2934
		kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL_ACCOUNT);
2935
		if (!kvm_regs)
A
Avi Kivity 已提交
2936
			goto out;
2937 2938 2939
		r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
		if (r)
			goto out_free1;
A
Avi Kivity 已提交
2940
		r = -EFAULT;
2941 2942
		if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
			goto out_free1;
A
Avi Kivity 已提交
2943
		r = 0;
2944 2945
out_free1:
		kfree(kvm_regs);
A
Avi Kivity 已提交
2946 2947 2948
		break;
	}
	case KVM_SET_REGS: {
2949
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
2950

2951
		r = -ENOMEM;
2952 2953 2954
		kvm_regs = memdup_user(argp, sizeof(*kvm_regs));
		if (IS_ERR(kvm_regs)) {
			r = PTR_ERR(kvm_regs);
A
Avi Kivity 已提交
2955
			goto out;
2956
		}
2957 2958
		r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
		kfree(kvm_regs);
A
Avi Kivity 已提交
2959 2960 2961
		break;
	}
	case KVM_GET_SREGS: {
2962 2963
		kvm_sregs = kzalloc(sizeof(struct kvm_sregs),
				    GFP_KERNEL_ACCOUNT);
2964 2965 2966 2967
		r = -ENOMEM;
		if (!kvm_sregs)
			goto out;
		r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
2968 2969 2970
		if (r)
			goto out;
		r = -EFAULT;
2971
		if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
A
Avi Kivity 已提交
2972 2973 2974 2975 2976
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_SREGS: {
2977 2978 2979
		kvm_sregs = memdup_user(argp, sizeof(*kvm_sregs));
		if (IS_ERR(kvm_sregs)) {
			r = PTR_ERR(kvm_sregs);
G
Guo Chao 已提交
2980
			kvm_sregs = NULL;
A
Avi Kivity 已提交
2981
			goto out;
2982
		}
2983
		r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
2984 2985
		break;
	}
2986 2987 2988 2989 2990 2991 2992
	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;
2993
		if (copy_to_user(argp, &mp_state, sizeof(mp_state)))
2994 2995 2996 2997 2998 2999 3000 3001
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_MP_STATE: {
		struct kvm_mp_state mp_state;

		r = -EFAULT;
3002
		if (copy_from_user(&mp_state, argp, sizeof(mp_state)))
3003 3004 3005 3006
			goto out;
		r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
		break;
	}
A
Avi Kivity 已提交
3007 3008 3009 3010
	case KVM_TRANSLATE: {
		struct kvm_translation tr;

		r = -EFAULT;
3011
		if (copy_from_user(&tr, argp, sizeof(tr)))
A
Avi Kivity 已提交
3012
			goto out;
3013
		r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
A
Avi Kivity 已提交
3014 3015 3016
		if (r)
			goto out;
		r = -EFAULT;
3017
		if (copy_to_user(argp, &tr, sizeof(tr)))
A
Avi Kivity 已提交
3018 3019 3020 3021
			goto out;
		r = 0;
		break;
	}
J
Jan Kiszka 已提交
3022 3023
	case KVM_SET_GUEST_DEBUG: {
		struct kvm_guest_debug dbg;
A
Avi Kivity 已提交
3024 3025

		r = -EFAULT;
3026
		if (copy_from_user(&dbg, argp, sizeof(dbg)))
A
Avi Kivity 已提交
3027
			goto out;
J
Jan Kiszka 已提交
3028
		r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
A
Avi Kivity 已提交
3029 3030
		break;
	}
A
Avi Kivity 已提交
3031 3032 3033 3034 3035 3036 3037 3038 3039
	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,
3040
					   sizeof(kvm_sigmask)))
A
Avi Kivity 已提交
3041 3042
				goto out;
			r = -EINVAL;
3043
			if (kvm_sigmask.len != sizeof(sigset))
A
Avi Kivity 已提交
3044 3045 3046
				goto out;
			r = -EFAULT;
			if (copy_from_user(&sigset, sigmask_arg->sigset,
3047
					   sizeof(sigset)))
A
Avi Kivity 已提交
3048 3049 3050
				goto out;
			p = &sigset;
		}
3051
		r = kvm_vcpu_ioctl_set_sigmask(vcpu, p);
A
Avi Kivity 已提交
3052 3053
		break;
	}
A
Avi Kivity 已提交
3054
	case KVM_GET_FPU: {
3055
		fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL_ACCOUNT);
3056 3057 3058 3059
		r = -ENOMEM;
		if (!fpu)
			goto out;
		r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu);
A
Avi Kivity 已提交
3060 3061 3062
		if (r)
			goto out;
		r = -EFAULT;
3063
		if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
A
Avi Kivity 已提交
3064 3065 3066 3067 3068
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_FPU: {
3069 3070 3071
		fpu = memdup_user(argp, sizeof(*fpu));
		if (IS_ERR(fpu)) {
			r = PTR_ERR(fpu);
G
Guo Chao 已提交
3072
			fpu = NULL;
A
Avi Kivity 已提交
3073
			goto out;
3074
		}
3075
		r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
A
Avi Kivity 已提交
3076 3077
		break;
	}
A
Avi Kivity 已提交
3078
	default:
3079
		r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
3080 3081
	}
out:
3082
	mutex_unlock(&vcpu->mutex);
3083 3084
	kfree(fpu);
	kfree(kvm_sregs);
A
Avi Kivity 已提交
3085 3086 3087
	return r;
}

3088
#ifdef CONFIG_KVM_COMPAT
3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107
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,
3108
					   sizeof(kvm_sigmask)))
3109 3110
				goto out;
			r = -EINVAL;
A
Al Viro 已提交
3111
			if (kvm_sigmask.len != sizeof(compat_sigset_t))
3112 3113
				goto out;
			r = -EFAULT;
A
Al Viro 已提交
3114
			if (get_compat_sigset(&sigset, (void *)sigmask_arg->sigset))
3115
				goto out;
3116 3117 3118
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
		} else
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, NULL);
3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129
		break;
	}
	default:
		r = kvm_vcpu_ioctl(filp, ioctl, arg);
	}

out:
	return r;
}
#endif

3130 3131 3132 3133 3134 3135 3136 3137 3138 3139
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
3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160
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;

3161 3162 3163
	if (dev->kvm->mm != current->mm)
		return -EIO;

S
Scott Wood 已提交
3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183
	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;

3184 3185 3186 3187 3188 3189 3190
	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
3191 3192 3193 3194 3195 3196 3197
	kvm_put_kvm(kvm);
	return 0;
}

static const struct file_operations kvm_device_fops = {
	.unlocked_ioctl = kvm_device_ioctl,
	.release = kvm_device_release,
3198
	KVM_COMPAT(kvm_device_ioctl),
3199
	.mmap = kvm_device_mmap,
S
Scott Wood 已提交
3200 3201 3202 3203 3204 3205 3206 3207 3208 3209
};

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

	return filp->private_data;
}

3210
static const struct kvm_device_ops *kvm_device_ops_table[KVM_DEV_TYPE_MAX] = {
3211
#ifdef CONFIG_KVM_MPIC
3212 3213
	[KVM_DEV_TYPE_FSL_MPIC_20]	= &kvm_mpic_ops,
	[KVM_DEV_TYPE_FSL_MPIC_42]	= &kvm_mpic_ops,
3214
#endif
3215 3216
};

3217
int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type)
3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228
{
	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;
}

3229 3230 3231 3232 3233 3234
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
3235 3236 3237
static int kvm_ioctl_create_device(struct kvm *kvm,
				   struct kvm_create_device *cd)
{
3238
	const struct kvm_device_ops *ops = NULL;
S
Scott Wood 已提交
3239 3240
	struct kvm_device *dev;
	bool test = cd->flags & KVM_CREATE_DEVICE_TEST;
P
Paolo Bonzini 已提交
3241
	int type;
S
Scott Wood 已提交
3242 3243
	int ret;

3244 3245 3246
	if (cd->type >= ARRAY_SIZE(kvm_device_ops_table))
		return -ENODEV;

P
Paolo Bonzini 已提交
3247 3248
	type = array_index_nospec(cd->type, ARRAY_SIZE(kvm_device_ops_table));
	ops = kvm_device_ops_table[type];
3249
	if (ops == NULL)
S
Scott Wood 已提交
3250 3251 3252 3253 3254
		return -ENODEV;

	if (test)
		return 0;

3255
	dev = kzalloc(sizeof(*dev), GFP_KERNEL_ACCOUNT);
S
Scott Wood 已提交
3256 3257 3258 3259 3260 3261
	if (!dev)
		return -ENOMEM;

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

3262
	mutex_lock(&kvm->lock);
P
Paolo Bonzini 已提交
3263
	ret = ops->create(dev, type);
S
Scott Wood 已提交
3264
	if (ret < 0) {
3265
		mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
3266 3267 3268
		kfree(dev);
		return ret;
	}
3269 3270
	list_add(&dev->vm_node, &kvm->devices);
	mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
3271

3272 3273 3274
	if (ops->init)
		ops->init(dev);

3275
	kvm_get_kvm(kvm);
3276
	ret = anon_inode_getfd(ops->name, &kvm_device_fops, dev, O_RDWR | O_CLOEXEC);
S
Scott Wood 已提交
3277
	if (ret < 0) {
3278
		kvm_put_kvm_no_destroy(kvm);
3279 3280 3281
		mutex_lock(&kvm->lock);
		list_del(&dev->vm_node);
		mutex_unlock(&kvm->lock);
3282
		ops->destroy(dev);
S
Scott Wood 已提交
3283 3284 3285 3286 3287 3288 3289
		return ret;
	}

	cd->fd = ret;
	return 0;
}

3290 3291 3292 3293 3294 3295 3296 3297 3298 3299
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
3300
#ifdef CONFIG_HAVE_KVM_IRQFD
3301
	case KVM_CAP_IRQFD:
3302 3303
	case KVM_CAP_IRQFD_RESAMPLE:
#endif
3304
	case KVM_CAP_IOEVENTFD_ANY_LENGTH:
3305
	case KVM_CAP_CHECK_EXTENSION_VM:
3306
	case KVM_CAP_ENABLE_CAP_VM:
3307
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
3308
	case KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2:
3309
#endif
3310
		return 1;
3311
#ifdef CONFIG_KVM_MMIO
3312 3313
	case KVM_CAP_COALESCED_MMIO:
		return KVM_COALESCED_MMIO_PAGE_OFFSET;
P
Peng Hao 已提交
3314 3315
	case KVM_CAP_COALESCED_PIO:
		return 1;
3316
#endif
3317 3318 3319
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_CAP_IRQ_ROUTING:
		return KVM_MAX_IRQ_ROUTES;
3320 3321 3322 3323
#endif
#if KVM_ADDRESS_SPACE_NUM > 1
	case KVM_CAP_MULTI_ADDRESS_SPACE:
		return KVM_ADDRESS_SPACE_NUM;
3324
#endif
3325 3326
	case KVM_CAP_NR_MEMSLOTS:
		return KVM_USER_MEM_SLOTS;
3327 3328 3329 3330 3331 3332
	default:
		break;
	}
	return kvm_vm_ioctl_check_extension(kvm, arg);
}

3333 3334 3335 3336 3337 3338 3339 3340 3341 3342
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) {
3343
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
3344
	case KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2:
3345 3346 3347 3348 3349
		if (cap->flags || (cap->args[0] & ~1))
			return -EINVAL;
		kvm->manual_dirty_log_protect = cap->args[0];
		return 0;
#endif
3350 3351 3352 3353 3354
	default:
		return kvm_vm_ioctl_enable_cap(kvm, cap);
	}
}

A
Avi Kivity 已提交
3355 3356 3357 3358 3359
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;
3360
	int r;
A
Avi Kivity 已提交
3361

3362 3363
	if (kvm->mm != current->mm)
		return -EIO;
A
Avi Kivity 已提交
3364 3365 3366 3367
	switch (ioctl) {
	case KVM_CREATE_VCPU:
		r = kvm_vm_ioctl_create_vcpu(kvm, arg);
		break;
3368 3369 3370 3371 3372 3373 3374 3375 3376
	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;
	}
3377 3378 3379 3380 3381
	case KVM_SET_USER_MEMORY_REGION: {
		struct kvm_userspace_memory_region kvm_userspace_mem;

		r = -EFAULT;
		if (copy_from_user(&kvm_userspace_mem, argp,
3382
						sizeof(kvm_userspace_mem)))
3383 3384
			goto out;

3385
		r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem);
A
Avi Kivity 已提交
3386 3387 3388 3389 3390 3391
		break;
	}
	case KVM_GET_DIRTY_LOG: {
		struct kvm_dirty_log log;

		r = -EFAULT;
3392
		if (copy_from_user(&log, argp, sizeof(log)))
A
Avi Kivity 已提交
3393
			goto out;
3394
		r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
A
Avi Kivity 已提交
3395 3396
		break;
	}
3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407
#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
3408
#ifdef CONFIG_KVM_MMIO
3409 3410
	case KVM_REGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3411

3412
		r = -EFAULT;
3413
		if (copy_from_user(&zone, argp, sizeof(zone)))
3414 3415 3416 3417 3418 3419
			goto out;
		r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
		break;
	}
	case KVM_UNREGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3420

3421
		r = -EFAULT;
3422
		if (copy_from_user(&zone, argp, sizeof(zone)))
3423 3424 3425 3426 3427
			goto out;
		r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
		break;
	}
#endif
G
Gregory Haskins 已提交
3428 3429 3430 3431
	case KVM_IRQFD: {
		struct kvm_irqfd data;

		r = -EFAULT;
3432
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3433
			goto out;
3434
		r = kvm_irqfd(kvm, &data);
G
Gregory Haskins 已提交
3435 3436
		break;
	}
G
Gregory Haskins 已提交
3437 3438 3439 3440
	case KVM_IOEVENTFD: {
		struct kvm_ioeventfd data;

		r = -EFAULT;
3441
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3442 3443 3444 3445
			goto out;
		r = kvm_ioeventfd(kvm, &data);
		break;
	}
3446 3447 3448 3449 3450
#ifdef CONFIG_HAVE_KVM_MSI
	case KVM_SIGNAL_MSI: {
		struct kvm_msi msi;

		r = -EFAULT;
3451
		if (copy_from_user(&msi, argp, sizeof(msi)))
3452 3453 3454 3455
			goto out;
		r = kvm_send_userspace_msi(kvm, &msi);
		break;
	}
3456 3457 3458 3459 3460 3461 3462
#endif
#ifdef __KVM_HAVE_IRQ_LINE
	case KVM_IRQ_LINE_STATUS:
	case KVM_IRQ_LINE: {
		struct kvm_irq_level irq_event;

		r = -EFAULT;
3463
		if (copy_from_user(&irq_event, argp, sizeof(irq_event)))
3464 3465
			goto out;

3466 3467
		r = kvm_vm_ioctl_irq_line(kvm, &irq_event,
					ioctl == KVM_IRQ_LINE_STATUS);
3468 3469 3470 3471 3472
		if (r)
			goto out;

		r = -EFAULT;
		if (ioctl == KVM_IRQ_LINE_STATUS) {
3473
			if (copy_to_user(argp, &irq_event, sizeof(irq_event)))
3474 3475 3476 3477 3478 3479
				goto out;
		}

		r = 0;
		break;
	}
3480
#endif
3481 3482 3483 3484
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_SET_GSI_ROUTING: {
		struct kvm_irq_routing routing;
		struct kvm_irq_routing __user *urouting;
3485
		struct kvm_irq_routing_entry *entries = NULL;
3486 3487 3488 3489 3490

		r = -EFAULT;
		if (copy_from_user(&routing, argp, sizeof(routing)))
			goto out;
		r = -EINVAL;
3491 3492
		if (!kvm_arch_can_set_irq_routing(kvm))
			goto out;
3493
		if (routing.nr > KVM_MAX_IRQ_ROUTES)
3494 3495 3496
			goto out;
		if (routing.flags)
			goto out;
3497 3498
		if (routing.nr) {
			r = -ENOMEM;
3499 3500
			entries = vmalloc(array_size(sizeof(*entries),
						     routing.nr));
3501 3502 3503 3504 3505 3506 3507 3508
			if (!entries)
				goto out;
			r = -EFAULT;
			urouting = argp;
			if (copy_from_user(entries, urouting->entries,
					   routing.nr * sizeof(*entries)))
				goto out_free_irq_routing;
		}
3509 3510
		r = kvm_set_irq_routing(kvm, entries, routing.nr,
					routing.flags);
3511
out_free_irq_routing:
3512 3513 3514 3515
		vfree(entries);
		break;
	}
#endif /* CONFIG_HAVE_KVM_IRQ_ROUTING */
S
Scott Wood 已提交
3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533
	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;
	}
3534 3535 3536
	case KVM_CHECK_EXTENSION:
		r = kvm_vm_ioctl_check_extension_generic(kvm, arg);
		break;
3537
	default:
3538
		r = kvm_arch_vm_ioctl(filp, ioctl, arg);
3539 3540 3541 3542 3543
	}
out:
	return r;
}

3544
#ifdef CONFIG_KVM_COMPAT
3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568
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)))
3569
			return -EFAULT;
3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584
		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

3585
static struct file_operations kvm_vm_fops = {
3586 3587
	.release        = kvm_vm_release,
	.unlocked_ioctl = kvm_vm_ioctl,
3588
	.llseek		= noop_llseek,
3589
	KVM_COMPAT(kvm_vm_compat_ioctl),
3590 3591
};

3592
static int kvm_dev_ioctl_create_vm(unsigned long type)
3593
{
3594
	int r;
3595
	struct kvm *kvm;
3596
	struct file *file;
3597

3598
	kvm = kvm_create_vm(type);
3599 3600
	if (IS_ERR(kvm))
		return PTR_ERR(kvm);
3601
#ifdef CONFIG_KVM_MMIO
3602
	r = kvm_coalesced_mmio_init(kvm);
3603 3604
	if (r < 0)
		goto put_kvm;
3605
#endif
3606
	r = get_unused_fd_flags(O_CLOEXEC);
3607 3608 3609
	if (r < 0)
		goto put_kvm;

3610 3611 3612
	file = anon_inode_getfile("kvm-vm", &kvm_vm_fops, kvm, O_RDWR);
	if (IS_ERR(file)) {
		put_unused_fd(r);
3613 3614
		r = PTR_ERR(file);
		goto put_kvm;
3615
	}
3616

3617 3618 3619 3620 3621 3622
	/*
	 * 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).
	 */
3623
	if (kvm_create_vm_debugfs(kvm, r) < 0) {
3624 3625
		put_unused_fd(r);
		fput(file);
3626 3627
		return -ENOMEM;
	}
3628
	kvm_uevent_notify_change(KVM_EVENT_CREATE_VM, kvm);
3629

3630
	fd_install(r, file);
3631
	return r;
3632 3633 3634 3635

put_kvm:
	kvm_put_kvm(kvm);
	return r;
3636 3637 3638 3639 3640
}

static long kvm_dev_ioctl(struct file *filp,
			  unsigned int ioctl, unsigned long arg)
{
3641
	long r = -EINVAL;
3642 3643 3644

	switch (ioctl) {
	case KVM_GET_API_VERSION:
3645 3646
		if (arg)
			goto out;
3647 3648 3649
		r = KVM_API_VERSION;
		break;
	case KVM_CREATE_VM:
3650
		r = kvm_dev_ioctl_create_vm(arg);
3651
		break;
3652
	case KVM_CHECK_EXTENSION:
3653
		r = kvm_vm_ioctl_check_extension_generic(NULL, arg);
3654
		break;
3655 3656 3657
	case KVM_GET_VCPU_MMAP_SIZE:
		if (arg)
			goto out;
3658 3659 3660
		r = PAGE_SIZE;     /* struct kvm_run */
#ifdef CONFIG_X86
		r += PAGE_SIZE;    /* pio data page */
3661
#endif
3662
#ifdef CONFIG_KVM_MMIO
3663
		r += PAGE_SIZE;    /* coalesced mmio ring page */
3664
#endif
3665
		break;
3666 3667 3668
	case KVM_TRACE_ENABLE:
	case KVM_TRACE_PAUSE:
	case KVM_TRACE_DISABLE:
3669
		r = -EOPNOTSUPP;
3670
		break;
A
Avi Kivity 已提交
3671
	default:
3672
		return kvm_arch_dev_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
3673 3674 3675 3676 3677 3678 3679
	}
out:
	return r;
}

static struct file_operations kvm_chardev_ops = {
	.unlocked_ioctl = kvm_dev_ioctl,
3680
	.llseek		= noop_llseek,
3681
	KVM_COMPAT(kvm_dev_ioctl),
A
Avi Kivity 已提交
3682 3683 3684
};

static struct miscdevice kvm_dev = {
A
Avi Kivity 已提交
3685
	KVM_MINOR,
A
Avi Kivity 已提交
3686 3687 3688 3689
	"kvm",
	&kvm_chardev_ops,
};

3690
static void hardware_enable_nolock(void *junk)
3691 3692
{
	int cpu = raw_smp_processor_id();
3693
	int r;
3694

3695
	if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
3696
		return;
3697

3698
	cpumask_set_cpu(cpu, cpus_hardware_enabled);
3699

3700
	r = kvm_arch_hardware_enable();
3701 3702 3703 3704

	if (r) {
		cpumask_clear_cpu(cpu, cpus_hardware_enabled);
		atomic_inc(&hardware_enable_failed);
X
Xiubo Li 已提交
3705
		pr_info("kvm: enabling virtualization on CPU%d failed\n", cpu);
3706
	}
3707 3708
}

3709
static int kvm_starting_cpu(unsigned int cpu)
3710
{
3711
	raw_spin_lock(&kvm_count_lock);
3712 3713
	if (kvm_usage_count)
		hardware_enable_nolock(NULL);
3714
	raw_spin_unlock(&kvm_count_lock);
3715
	return 0;
3716 3717 3718
}

static void hardware_disable_nolock(void *junk)
3719 3720 3721
{
	int cpu = raw_smp_processor_id();

3722
	if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
3723
		return;
3724
	cpumask_clear_cpu(cpu, cpus_hardware_enabled);
3725
	kvm_arch_hardware_disable();
3726 3727
}

3728
static int kvm_dying_cpu(unsigned int cpu)
3729
{
3730
	raw_spin_lock(&kvm_count_lock);
3731 3732
	if (kvm_usage_count)
		hardware_disable_nolock(NULL);
3733
	raw_spin_unlock(&kvm_count_lock);
3734
	return 0;
3735 3736
}

3737 3738 3739 3740 3741 3742
static void hardware_disable_all_nolock(void)
{
	BUG_ON(!kvm_usage_count);

	kvm_usage_count--;
	if (!kvm_usage_count)
3743
		on_each_cpu(hardware_disable_nolock, NULL, 1);
3744 3745 3746 3747
}

static void hardware_disable_all(void)
{
3748
	raw_spin_lock(&kvm_count_lock);
3749
	hardware_disable_all_nolock();
3750
	raw_spin_unlock(&kvm_count_lock);
3751 3752 3753 3754 3755 3756
}

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

3757
	raw_spin_lock(&kvm_count_lock);
3758 3759 3760 3761

	kvm_usage_count++;
	if (kvm_usage_count == 1) {
		atomic_set(&hardware_enable_failed, 0);
3762
		on_each_cpu(hardware_enable_nolock, NULL, 1);
3763 3764 3765 3766 3767 3768 3769

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

3770
	raw_spin_unlock(&kvm_count_lock);
3771 3772 3773 3774

	return r;
}

3775
static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
M
Mike Day 已提交
3776
		      void *v)
3777
{
3778 3779 3780 3781 3782 3783
	/*
	 * 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
3784
	pr_info("kvm: exiting hardware virtualization\n");
3785
	kvm_rebooting = true;
3786
	on_each_cpu(hardware_disable_nolock, NULL, 1);
3787 3788 3789 3790 3791 3792 3793 3794
	return NOTIFY_OK;
}

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

M
Marcelo Tosatti 已提交
3795
static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
3796 3797 3798 3799
{
	int i;

	for (i = 0; i < bus->dev_count; i++) {
3800
		struct kvm_io_device *pos = bus->range[i].dev;
3801 3802 3803

		kvm_iodevice_destructor(pos);
	}
M
Marcelo Tosatti 已提交
3804
	kfree(bus);
3805 3806
}

3807
static inline int kvm_io_bus_cmp(const struct kvm_io_range *r1,
X
Xiubo Li 已提交
3808
				 const struct kvm_io_range *r2)
3809
{
J
Jason Wang 已提交
3810 3811 3812 3813
	gpa_t addr1 = r1->addr;
	gpa_t addr2 = r2->addr;

	if (addr1 < addr2)
3814
		return -1;
J
Jason Wang 已提交
3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826

	/* 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)
3827
		return 1;
J
Jason Wang 已提交
3828

3829 3830 3831
	return 0;
}

3832 3833
static int kvm_io_bus_sort_cmp(const void *p1, const void *p2)
{
3834
	return kvm_io_bus_cmp(p1, p2);
3835 3836
}

G
Geoff Levand 已提交
3837
static int kvm_io_bus_get_first_dev(struct kvm_io_bus *bus,
3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854
			     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;

3855
	while (off > 0 && kvm_io_bus_cmp(&key, &bus->range[off-1]) == 0)
3856 3857 3858 3859 3860
		off--;

	return off;
}

3861
static int __kvm_io_bus_write(struct kvm_vcpu *vcpu, struct kvm_io_bus *bus,
C
Cornelia Huck 已提交
3862 3863 3864 3865 3866 3867 3868 3869 3870
			      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 &&
3871
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
3872
		if (!kvm_iodevice_write(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
3873 3874 3875 3876 3877 3878 3879 3880
					range->len, val))
			return idx;
		idx++;
	}

	return -EOPNOTSUPP;
}

3881
/* kvm_io_bus_write - called under kvm->slots_lock */
3882
int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
3883
		     int len, const void *val)
3884
{
3885
	struct kvm_io_bus *bus;
3886
	struct kvm_io_range range;
C
Cornelia Huck 已提交
3887
	int r;
3888 3889 3890 3891 3892

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
3893

3894
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3895 3896
	if (!bus)
		return -ENOMEM;
3897
	r = __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3898 3899
	return r < 0 ? r : 0;
}
L
Leo Yan 已提交
3900
EXPORT_SYMBOL_GPL(kvm_io_bus_write);
C
Cornelia Huck 已提交
3901 3902

/* kvm_io_bus_write_cookie - called under kvm->slots_lock */
3903 3904
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
3905 3906 3907 3908 3909 3910 3911 3912 3913
{
	struct kvm_io_bus *bus;
	struct kvm_io_range range;

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

3914
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3915 3916
	if (!bus)
		return -ENOMEM;
C
Cornelia Huck 已提交
3917 3918 3919

	/* First try the device referenced by cookie. */
	if ((cookie >= 0) && (cookie < bus->dev_count) &&
3920
	    (kvm_io_bus_cmp(&range, &bus->range[cookie]) == 0))
3921
		if (!kvm_iodevice_write(vcpu, bus->range[cookie].dev, addr, len,
C
Cornelia Huck 已提交
3922 3923 3924 3925 3926 3927 3928
					val))
			return cookie;

	/*
	 * cookie contained garbage; fall back to search and return the
	 * correct cookie value.
	 */
3929
	return __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3930 3931
}

3932 3933
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
3934 3935 3936 3937
{
	int idx;

	idx = kvm_io_bus_get_first_dev(bus, range->addr, range->len);
3938 3939 3940 3941
	if (idx < 0)
		return -EOPNOTSUPP;

	while (idx < bus->dev_count &&
3942
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
3943
		if (!kvm_iodevice_read(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
3944 3945
				       range->len, val))
			return idx;
3946 3947 3948
		idx++;
	}

3949 3950
	return -EOPNOTSUPP;
}
3951

3952
/* kvm_io_bus_read - called under kvm->slots_lock */
3953
int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
M
Marcelo Tosatti 已提交
3954
		    int len, void *val)
3955
{
3956
	struct kvm_io_bus *bus;
3957
	struct kvm_io_range range;
C
Cornelia Huck 已提交
3958
	int r;
3959 3960 3961 3962 3963

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
M
Marcelo Tosatti 已提交
3964

3965
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3966 3967
	if (!bus)
		return -ENOMEM;
3968
	r = __kvm_io_bus_read(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3969 3970
	return r < 0 ? r : 0;
}
3971

3972
/* Caller must hold slots_lock. */
3973 3974
int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
			    int len, struct kvm_io_device *dev)
3975
{
3976
	int i;
M
Marcelo Tosatti 已提交
3977
	struct kvm_io_bus *new_bus, *bus;
3978
	struct kvm_io_range range;
3979

3980
	bus = kvm_get_bus(kvm, bus_idx);
3981 3982 3983
	if (!bus)
		return -ENOMEM;

3984 3985
	/* exclude ioeventfd which is limited by maximum fd */
	if (bus->dev_count - bus->ioeventfd_count > NR_IOBUS_DEVS - 1)
3986
		return -ENOSPC;
3987

3988
	new_bus = kmalloc(struct_size(bus, range, bus->dev_count + 1),
3989
			  GFP_KERNEL_ACCOUNT);
M
Marcelo Tosatti 已提交
3990 3991
	if (!new_bus)
		return -ENOMEM;
3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007

	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
4008 4009 4010
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
4011 4012 4013 4014

	return 0;
}

4015
/* Caller must hold slots_lock. */
4016 4017
void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
			       struct kvm_io_device *dev)
4018
{
4019
	int i;
M
Marcelo Tosatti 已提交
4020
	struct kvm_io_bus *new_bus, *bus;
4021

4022
	bus = kvm_get_bus(kvm, bus_idx);
4023
	if (!bus)
4024
		return;
4025

4026 4027
	for (i = 0; i < bus->dev_count; i++)
		if (bus->range[i].dev == dev) {
4028 4029
			break;
		}
M
Marcelo Tosatti 已提交
4030

4031 4032
	if (i == bus->dev_count)
		return;
4033

4034
	new_bus = kmalloc(struct_size(bus, range, bus->dev_count - 1),
4035
			  GFP_KERNEL_ACCOUNT);
4036 4037 4038 4039
	if (!new_bus)  {
		pr_err("kvm: failed to shrink bus, removing it completely\n");
		goto broken;
	}
4040 4041 4042 4043 4044

	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
4045

4046
broken:
M
Marcelo Tosatti 已提交
4047 4048 4049
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
4050
	return;
4051 4052
}

4053 4054 4055 4056 4057 4058 4059 4060 4061 4062
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);
4063 4064
	if (!bus)
		goto out_unlock;
4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078

	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);

4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090
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.
	 */
4091
	if (!refcount_inc_not_zero(&stat_data->kvm->users_count))
4092 4093
		return -ENOENT;

4094
	if (simple_attr_open(inode, file, get,
4095 4096 4097
		    KVM_DBGFS_GET_MODE(stat_data->dbgfs_item) & 0222
		    ? set : NULL,
		    fmt)) {
4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115
		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;
}

4116
static int kvm_get_stat_per_vm(struct kvm *kvm, size_t offset, u64 *val)
4117
{
4118
	*val = *(ulong *)((void *)kvm + offset);
4119

4120 4121 4122 4123 4124 4125
	return 0;
}

static int kvm_clear_stat_per_vm(struct kvm *kvm, size_t offset)
{
	*(ulong *)((void *)kvm + offset) = 0;
4126 4127 4128 4129

	return 0;
}

4130
static int kvm_get_stat_per_vcpu(struct kvm *kvm, size_t offset, u64 *val)
4131
{
4132 4133
	int i;
	struct kvm_vcpu *vcpu;
4134

4135
	*val = 0;
4136

4137 4138
	kvm_for_each_vcpu(i, vcpu, kvm)
		*val += *(u64 *)((void *)vcpu + offset);
4139 4140 4141 4142

	return 0;
}

4143
static int kvm_clear_stat_per_vcpu(struct kvm *kvm, size_t offset)
4144
{
4145 4146
	int i;
	struct kvm_vcpu *vcpu;
4147

4148 4149 4150 4151 4152
	kvm_for_each_vcpu(i, vcpu, kvm)
		*(u64 *)((void *)vcpu + offset) = 0;

	return 0;
}
4153

4154
static int kvm_stat_data_get(void *data, u64 *val)
4155
{
4156
	int r = -EFAULT;
4157 4158
	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)data;

4159 4160 4161 4162 4163 4164 4165 4166 4167 4168
	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;
	}
4169

4170
	return r;
4171 4172
}

4173
static int kvm_stat_data_clear(void *data, u64 val)
4174
{
4175
	int r = -EFAULT;
4176 4177 4178 4179 4180
	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)data;

	if (val)
		return -EINVAL;

4181 4182 4183 4184 4185 4186 4187 4188 4189 4190
	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;
	}
4191

4192
	return r;
4193 4194
}

4195
static int kvm_stat_data_open(struct inode *inode, struct file *file)
4196 4197
{
	__simple_attr_check_format("%llu\n", 0ull);
4198 4199
	return kvm_debugfs_open(inode, file, kvm_stat_data_get,
				kvm_stat_data_clear, "%llu\n");
4200 4201
}

4202 4203 4204
static const struct file_operations stat_fops_per_vm = {
	.owner = THIS_MODULE,
	.open = kvm_stat_data_open,
4205
	.release = kvm_debugfs_release,
4206 4207 4208
	.read = simple_attr_read,
	.write = simple_attr_write,
	.llseek = no_llseek,
4209 4210
};

4211
static int vm_stat_get(void *_offset, u64 *val)
4212 4213 4214
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
4215
	u64 tmp_val;
4216

4217
	*val = 0;
J
Junaid Shahid 已提交
4218
	mutex_lock(&kvm_lock);
4219
	list_for_each_entry(kvm, &vm_list, vm_list) {
4220
		kvm_get_stat_per_vm(kvm, offset, &tmp_val);
4221 4222
		*val += tmp_val;
	}
J
Junaid Shahid 已提交
4223
	mutex_unlock(&kvm_lock);
4224
	return 0;
4225 4226
}

4227 4228 4229 4230 4231 4232 4233 4234
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
4235
	mutex_lock(&kvm_lock);
4236
	list_for_each_entry(kvm, &vm_list, vm_list) {
4237
		kvm_clear_stat_per_vm(kvm, offset);
4238
	}
J
Junaid Shahid 已提交
4239
	mutex_unlock(&kvm_lock);
4240 4241 4242 4243 4244

	return 0;
}

DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, vm_stat_clear, "%llu\n");
4245

4246
static int vcpu_stat_get(void *_offset, u64 *val)
A
Avi Kivity 已提交
4247 4248 4249
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
4250
	u64 tmp_val;
A
Avi Kivity 已提交
4251

4252
	*val = 0;
J
Junaid Shahid 已提交
4253
	mutex_lock(&kvm_lock);
4254
	list_for_each_entry(kvm, &vm_list, vm_list) {
4255
		kvm_get_stat_per_vcpu(kvm, offset, &tmp_val);
4256 4257
		*val += tmp_val;
	}
J
Junaid Shahid 已提交
4258
	mutex_unlock(&kvm_lock);
4259
	return 0;
A
Avi Kivity 已提交
4260 4261
}

4262 4263 4264 4265 4266 4267 4268 4269
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
4270
	mutex_lock(&kvm_lock);
4271
	list_for_each_entry(kvm, &vm_list, vm_list) {
4272
		kvm_clear_stat_per_vcpu(kvm, offset);
4273
	}
J
Junaid Shahid 已提交
4274
	mutex_unlock(&kvm_lock);
4275 4276 4277 4278 4279 4280

	return 0;
}

DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, vcpu_stat_clear,
			"%llu\n");
4281

4282
static const struct file_operations *stat_fops[] = {
4283 4284 4285
	[KVM_STAT_VCPU] = &vcpu_stat_fops,
	[KVM_STAT_VM]   = &vm_stat_fops,
};
A
Avi Kivity 已提交
4286

4287 4288 4289 4290 4291 4292 4293 4294
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
4295
	mutex_lock(&kvm_lock);
4296 4297 4298 4299 4300 4301 4302 4303
	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
4304
	mutex_unlock(&kvm_lock);
4305

4306
	env = kzalloc(sizeof(*env), GFP_KERNEL_ACCOUNT);
4307 4308 4309 4310 4311 4312
	if (!env)
		return;

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

4313
	if (type == KVM_EVENT_CREATE_VM) {
4314
		add_uevent_var(env, "EVENT=create");
4315 4316
		kvm->userspace_pid = task_pid_nr(current);
	} else if (type == KVM_EVENT_DESTROY_VM) {
4317
		add_uevent_var(env, "EVENT=destroy");
4318 4319
	}
	add_uevent_var(env, "PID=%d", kvm->userspace_pid);
4320

4321
	if (!IS_ERR_OR_NULL(kvm->debugfs_dentry)) {
4322
		char *tmp, *p = kmalloc(PATH_MAX, GFP_KERNEL_ACCOUNT);
4323 4324 4325 4326 4327 4328

		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);
4329 4330 4331 4332 4333 4334 4335 4336
		}
	}
	/* 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);
}

4337
static void kvm_init_debug(void)
A
Avi Kivity 已提交
4338 4339 4340
{
	struct kvm_stats_debugfs_item *p;

4341
	kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
4342

4343 4344
	kvm_debugfs_num_entries = 0;
	for (p = debugfs_entries; p->name; ++p, kvm_debugfs_num_entries++) {
4345 4346
		debugfs_create_file(p->name, KVM_DBGFS_GET_MODE(p),
				    kvm_debugfs_dir, (void *)(long)p->offset,
4347
				    stat_fops[p->kind]);
4348
	}
A
Avi Kivity 已提交
4349 4350
}

4351
static int kvm_suspend(void)
4352
{
4353
	if (kvm_usage_count)
4354
		hardware_disable_nolock(NULL);
4355 4356 4357
	return 0;
}

4358
static void kvm_resume(void)
4359
{
4360
	if (kvm_usage_count) {
4361 4362 4363
#ifdef CONFIG_LOCKDEP
		WARN_ON(lockdep_is_held(&kvm_count_lock));
#endif
4364
		hardware_enable_nolock(NULL);
4365
	}
4366 4367
}

4368
static struct syscore_ops kvm_syscore_ops = {
4369 4370 4371 4372
	.suspend = kvm_suspend,
	.resume = kvm_resume,
};

4373 4374 4375 4376 4377 4378 4379 4380 4381
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);
4382

4383
	WRITE_ONCE(vcpu->preempted, false);
4384
	WRITE_ONCE(vcpu->ready, false);
4385

4386
	__this_cpu_write(kvm_running_vcpu, vcpu);
R
Radim Krčmář 已提交
4387
	kvm_arch_sched_in(vcpu, cpu);
4388
	kvm_arch_vcpu_load(vcpu, cpu);
4389 4390 4391 4392 4393 4394 4395
}

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

4396
	if (current->state == TASK_RUNNING) {
4397
		WRITE_ONCE(vcpu->preempted, true);
4398 4399
		WRITE_ONCE(vcpu->ready, true);
	}
4400
	kvm_arch_vcpu_put(vcpu);
4401 4402 4403 4404 4405
	__this_cpu_write(kvm_running_vcpu, NULL);
}

/**
 * kvm_get_running_vcpu - get the vcpu running on the current CPU.
4406 4407 4408 4409 4410 4411
 *
 * 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.
4412 4413 4414
 */
struct kvm_vcpu *kvm_get_running_vcpu(void)
{
4415 4416 4417 4418 4419 4420 4421
	struct kvm_vcpu *vcpu;

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

	return vcpu;
4422 4423 4424 4425 4426 4427 4428 4429
}

/**
 * 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;
4430 4431
}

4432 4433 4434 4435 4436
static void check_processor_compat(void *rtn)
{
	*(int *)rtn = kvm_arch_check_processor_compat();
}

4437
int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
4438
		  struct module *module)
A
Avi Kivity 已提交
4439 4440
{
	int r;
Y
Yang, Sheng 已提交
4441
	int cpu;
A
Avi Kivity 已提交
4442

4443 4444
	r = kvm_arch_init(opaque);
	if (r)
4445
		goto out_fail;
4446

4447 4448 4449 4450
	/*
	 * 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
4451 4452
	 * kvm_arch_init must be called before kvm_irqfd_init to avoid creating
	 * conflicts in case kvm is already setup for another implementation.
4453
	 */
P
Paolo Bonzini 已提交
4454 4455 4456
	r = kvm_irqfd_init();
	if (r)
		goto out_irqfd;
4457

4458
	if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
4459 4460 4461 4462
		r = -ENOMEM;
		goto out_free_0;
	}

4463
	r = kvm_arch_hardware_setup();
A
Avi Kivity 已提交
4464
	if (r < 0)
4465
		goto out_free_1;
A
Avi Kivity 已提交
4466

Y
Yang, Sheng 已提交
4467
	for_each_online_cpu(cpu) {
4468
		smp_call_function_single(cpu, check_processor_compat, &r, 1);
Y
Yang, Sheng 已提交
4469
		if (r < 0)
4470
			goto out_free_2;
Y
Yang, Sheng 已提交
4471 4472
	}

T
Thomas Gleixner 已提交
4473
	r = cpuhp_setup_state_nocalls(CPUHP_AP_KVM_STARTING, "kvm/cpu:starting",
4474
				      kvm_starting_cpu, kvm_dying_cpu);
A
Avi Kivity 已提交
4475
	if (r)
4476
		goto out_free_2;
A
Avi Kivity 已提交
4477 4478
	register_reboot_notifier(&kvm_reboot_notifier);

4479
	/* A kmem cache lets us meet the alignment requirements of fx_save. */
4480 4481
	if (!vcpu_align)
		vcpu_align = __alignof__(struct kvm_vcpu);
4482 4483 4484 4485 4486 4487
	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);
4488 4489
	if (!kvm_vcpu_cache) {
		r = -ENOMEM;
4490
		goto out_free_3;
4491 4492
	}

4493 4494 4495 4496
	r = kvm_async_pf_init();
	if (r)
		goto out_free;

A
Avi Kivity 已提交
4497
	kvm_chardev_ops.owner = module;
4498 4499
	kvm_vm_fops.owner = module;
	kvm_vcpu_fops.owner = module;
A
Avi Kivity 已提交
4500 4501 4502

	r = misc_register(&kvm_dev);
	if (r) {
X
Xiubo Li 已提交
4503
		pr_err("kvm: misc device register failed\n");
4504
		goto out_unreg;
A
Avi Kivity 已提交
4505 4506
	}

4507 4508
	register_syscore_ops(&kvm_syscore_ops);

4509 4510 4511
	kvm_preempt_ops.sched_in = kvm_sched_in;
	kvm_preempt_ops.sched_out = kvm_sched_out;

4512
	kvm_init_debug();
4513

P
Paolo Bonzini 已提交
4514 4515 4516
	r = kvm_vfio_ops_init();
	WARN_ON(r);

4517
	return 0;
A
Avi Kivity 已提交
4518

4519 4520
out_unreg:
	kvm_async_pf_deinit();
A
Avi Kivity 已提交
4521
out_free:
4522
	kmem_cache_destroy(kvm_vcpu_cache);
4523
out_free_3:
A
Avi Kivity 已提交
4524
	unregister_reboot_notifier(&kvm_reboot_notifier);
4525
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4526
out_free_2:
4527
	kvm_arch_hardware_unsetup();
4528
out_free_1:
4529
	free_cpumask_var(cpus_hardware_enabled);
4530
out_free_0:
4531
	kvm_irqfd_exit();
P
Paolo Bonzini 已提交
4532
out_irqfd:
4533 4534
	kvm_arch_exit();
out_fail:
A
Avi Kivity 已提交
4535 4536
	return r;
}
4537
EXPORT_SYMBOL_GPL(kvm_init);
A
Avi Kivity 已提交
4538

4539
void kvm_exit(void)
A
Avi Kivity 已提交
4540
{
4541
	debugfs_remove_recursive(kvm_debugfs_dir);
A
Avi Kivity 已提交
4542
	misc_deregister(&kvm_dev);
4543
	kmem_cache_destroy(kvm_vcpu_cache);
4544
	kvm_async_pf_deinit();
4545
	unregister_syscore_ops(&kvm_syscore_ops);
A
Avi Kivity 已提交
4546
	unregister_reboot_notifier(&kvm_reboot_notifier);
4547
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4548
	on_each_cpu(hardware_disable_nolock, NULL, 1);
4549
	kvm_arch_hardware_unsetup();
4550
	kvm_arch_exit();
4551
	kvm_irqfd_exit();
4552
	free_cpumask_var(cpus_hardware_enabled);
4553
	kvm_vfio_ops_exit();
A
Avi Kivity 已提交
4554
}
4555
EXPORT_SYMBOL_GPL(kvm_exit);
4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638

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;
}