kvm_main.c 117.7 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>
A
Avi Kivity 已提交
58

59
#include "coalesced_mmio.h"
60
#include "async_pf.h"
P
Paolo Bonzini 已提交
61
#include "vfio.h"
62

63 64 65
#define CREATE_TRACE_POINTS
#include <trace/events/kvm.h>

66 67 68
/* Worst case buffer size needed for holding an integer. */
#define ITOA_MAX_LEN 12

A
Avi Kivity 已提交
69 70 71
MODULE_AUTHOR("Qumranet");
MODULE_LICENSE("GPL");

72
/* Architectures should define their poll value according to the halt latency */
73
unsigned int halt_poll_ns = KVM_HALT_POLL_NS_DEFAULT;
74
module_param(halt_poll_ns, uint, 0644);
75
EXPORT_SYMBOL_GPL(halt_poll_ns);
76

W
Wanpeng Li 已提交
77
/* Default doubles per-vcpu halt_poll_ns. */
78
unsigned int halt_poll_ns_grow = 2;
79
module_param(halt_poll_ns_grow, uint, 0644);
80
EXPORT_SYMBOL_GPL(halt_poll_ns_grow);
W
Wanpeng Li 已提交
81

82 83 84 85 86
/* 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
87
/* Default resets per-vcpu halt_poll_ns . */
88
unsigned int halt_poll_ns_shrink;
89
module_param(halt_poll_ns_shrink, uint, 0644);
90
EXPORT_SYMBOL_GPL(halt_poll_ns_shrink);
W
Wanpeng Li 已提交
91

92 93 94
/*
 * Ordering of locks:
 *
95
 *	kvm->lock --> kvm->slots_lock --> kvm->irq_lock
96 97
 */

J
Junaid Shahid 已提交
98
DEFINE_MUTEX(kvm_lock);
99
static DEFINE_RAW_SPINLOCK(kvm_count_lock);
100
LIST_HEAD(vm_list);
101

102
static cpumask_var_t cpus_hardware_enabled;
103
static int kvm_usage_count;
104
static atomic_t hardware_enable_failed;
105

106
static struct kmem_cache *kvm_vcpu_cache;
A
Avi Kivity 已提交
107

108
static __read_mostly struct preempt_ops kvm_preempt_ops;
109
static DEFINE_PER_CPU(struct kvm_vcpu *, kvm_running_vcpu);
110

111
struct dentry *kvm_debugfs_dir;
112
EXPORT_SYMBOL_GPL(kvm_debugfs_dir);
A
Avi Kivity 已提交
113

114
static int kvm_debugfs_num_entries;
115
static const struct file_operations stat_fops_per_vm;
116

A
Avi Kivity 已提交
117 118
static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl,
			   unsigned long arg);
119
#ifdef CONFIG_KVM_COMPAT
120 121
static long kvm_vcpu_compat_ioctl(struct file *file, unsigned int ioctl,
				  unsigned long arg);
122 123
#define KVM_COMPAT(c)	.compat_ioctl	= (c)
#else
124 125 126 127 128 129 130
/*
 * 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.
 */
131 132
static long kvm_no_compat_ioctl(struct file *file, unsigned int ioctl,
				unsigned long arg) { return -EINVAL; }
133 134 135 136 137 138 139

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
140
#endif
141 142
static int hardware_enable_all(void);
static void hardware_disable_all(void);
A
Avi Kivity 已提交
143

M
Marcelo Tosatti 已提交
144
static void kvm_io_bus_destroy(struct kvm_io_bus *bus);
145

146
static void mark_page_dirty_in_slot(struct kvm_memory_slot *memslot, gfn_t gfn);
M
Marcelo Tosatti 已提交
147

148
__visible bool kvm_rebooting;
149
EXPORT_SYMBOL_GPL(kvm_rebooting);
150

151 152 153 154 155 156
#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;

157 158
__weak void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
						   unsigned long start, unsigned long end)
159 160 161
{
}

162 163 164 165 166 167 168 169 170 171 172 173 174 175
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
176
bool kvm_is_reserved_pfn(kvm_pfn_t pfn)
B
Ben-Ami Yassour 已提交
177
{
178 179 180 181 182
	/*
	 * ZONE_DEVICE pages currently set PG_reserved, but from a refcounting
	 * perspective they are "normal" pages, albeit with slightly different
	 * usage rules.
	 */
183
	if (pfn_valid(pfn))
184
		return PageReserved(pfn_to_page(pfn)) &&
185
		       !is_zero_pfn(pfn) &&
186
		       !kvm_is_zone_device_pfn(pfn);
B
Ben-Ami Yassour 已提交
187 188 189 190

	return true;
}

191 192 193 194 195 196 197 198 199 200
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
201 202 203
/*
 * Switches to specified vcpu, until a matching vcpu_put()
 */
204
void vcpu_load(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
205
{
206
	int cpu = get_cpu();
207 208

	__this_cpu_write(kvm_running_vcpu, vcpu);
209
	preempt_notifier_register(&vcpu->preempt_notifier);
210
	kvm_arch_vcpu_load(vcpu, cpu);
211
	put_cpu();
A
Avi Kivity 已提交
212
}
213
EXPORT_SYMBOL_GPL(vcpu_load);
A
Avi Kivity 已提交
214

215
void vcpu_put(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
216
{
217
	preempt_disable();
218
	kvm_arch_vcpu_put(vcpu);
219
	preempt_notifier_unregister(&vcpu->preempt_notifier);
220
	__this_cpu_write(kvm_running_vcpu, NULL);
221
	preempt_enable();
A
Avi Kivity 已提交
222
}
223
EXPORT_SYMBOL_GPL(vcpu_put);
A
Avi Kivity 已提交
224

225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242
/* 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;
}

243 244 245 246
static void ack_flush(void *_completed)
{
}

247 248 249 250 251 252 253 254 255 256 257 258
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;
}

259
bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
260
				 struct kvm_vcpu *except,
261
				 unsigned long *vcpu_bitmap, cpumask_var_t tmp)
262
{
263
	int i, cpu, me;
264
	struct kvm_vcpu *vcpu;
265
	bool called;
266

267
	me = get_cpu();
268

269
	kvm_for_each_vcpu(i, vcpu, kvm) {
270 271
		if ((vcpu_bitmap && !test_bit(i, vcpu_bitmap)) ||
		    vcpu == except)
272 273
			continue;

274
		kvm_make_request(req, vcpu);
275
		cpu = vcpu->cpu;
276

277 278
		if (!(req & KVM_REQUEST_NO_WAKEUP) && kvm_vcpu_wake_up(vcpu))
			continue;
279

280
		if (tmp != NULL && cpu != -1 && cpu != me &&
281
		    kvm_request_needs_ipi(vcpu, req))
282
			__cpumask_set_cpu(cpu, tmp);
283
	}
284 285

	called = kvm_kick_many_cpus(tmp, !!(req & KVM_REQUEST_WAIT));
286
	put_cpu();
287 288 289 290

	return called;
}

291 292
bool kvm_make_all_cpus_request_except(struct kvm *kvm, unsigned int req,
				      struct kvm_vcpu *except)
293 294 295 296 297 298
{
	cpumask_var_t cpus;
	bool called;

	zalloc_cpumask_var(&cpus, GFP_ATOMIC);

299
	called = kvm_make_vcpus_request_mask(kvm, req, except, NULL, cpus);
300

301
	free_cpumask_var(cpus);
302
	return called;
303 304
}

305 306 307 308 309
bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req)
{
	return kvm_make_all_cpus_request_except(kvm, req, NULL);
}

310
#ifndef CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALL
311
void kvm_flush_remote_tlbs(struct kvm *kvm)
312
{
313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329
	/*
	 * 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.
	 */
330 331
	if (!kvm_arch_flush_remote_tlb(kvm)
	    || kvm_make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH))
332
		++kvm->stat.remote_tlb_flush;
333
	cmpxchg(&kvm->tlbs_dirty, dirty_count, 0);
334
}
335
EXPORT_SYMBOL_GPL(kvm_flush_remote_tlbs);
336
#endif
337

338 339
void kvm_reload_remote_mmus(struct kvm *kvm)
{
340
	kvm_make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD);
341
}
342

343
static void kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id)
R
Rusty Russell 已提交
344 345 346 347 348
{
	mutex_init(&vcpu->mutex);
	vcpu->cpu = -1;
	vcpu->kvm = kvm;
	vcpu->vcpu_id = id;
349
	vcpu->pid = NULL;
350
	rcuwait_init(&vcpu->wait);
351
	kvm_async_pf_vcpu_init(vcpu);
R
Rusty Russell 已提交
352

353 354 355
	vcpu->pre_pcpu = -1;
	INIT_LIST_HEAD(&vcpu->blocked_vcpu_list);

356 357
	kvm_vcpu_set_in_spin_loop(vcpu, false);
	kvm_vcpu_set_dy_eligible(vcpu, false);
358
	vcpu->preempted = false;
359
	vcpu->ready = false;
360
	preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);
R
Rusty Russell 已提交
361 362
}

363 364 365
void kvm_vcpu_destroy(struct kvm_vcpu *vcpu)
{
	kvm_arch_vcpu_destroy(vcpu);
366

367 368 369 370 371 372 373
	/*
	 * 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));

374
	free_page((unsigned long)vcpu->run);
375
	kmem_cache_free(kvm_vcpu_cache, vcpu);
376 377 378
}
EXPORT_SYMBOL_GPL(kvm_vcpu_destroy);

379 380 381 382 383 384
#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);
}

385 386 387 388 389 390 391 392 393 394 395 396
static void kvm_mmu_notifier_invalidate_range(struct mmu_notifier *mn,
					      struct mm_struct *mm,
					      unsigned long start, unsigned long end)
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
	int idx;

	idx = srcu_read_lock(&kvm->srcu);
	kvm_arch_mmu_notifier_invalidate_range(kvm, start, end);
	srcu_read_unlock(&kvm->srcu, idx);
}

397 398 399 400 401 402
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);
403
	int idx;
404

405
	idx = srcu_read_lock(&kvm->srcu);
406 407
	spin_lock(&kvm->mmu_lock);
	kvm->mmu_notifier_seq++;
408 409 410 411

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

412
	spin_unlock(&kvm->mmu_lock);
413
	srcu_read_unlock(&kvm->srcu, idx);
414 415
}

416
static int kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn,
417
					const struct mmu_notifier_range *range)
418 419
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
420
	int need_tlb_flush = 0, idx;
421

422
	idx = srcu_read_lock(&kvm->srcu);
423 424 425 426 427 428 429
	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++;
430
	need_tlb_flush = kvm_unmap_hva_range(kvm, range->start, range->end);
431
	need_tlb_flush |= kvm->tlbs_dirty;
432 433 434
	/* we've to flush the tlb before the pages can be freed */
	if (need_tlb_flush)
		kvm_flush_remote_tlbs(kvm);
435 436 437

	spin_unlock(&kvm->mmu_lock);
	srcu_read_unlock(&kvm->srcu, idx);
438

439
	return 0;
440 441 442
}

static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
443
					const struct mmu_notifier_range *range)
444 445 446 447 448 449 450 451 452 453
{
	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++;
454
	smp_wmb();
455 456
	/*
	 * The above sequence increase must be visible before the
457 458
	 * below count decrease, which is ensured by the smp_wmb above
	 * in conjunction with the smp_rmb in mmu_notifier_retry().
459 460 461 462 463 464 465 466 467
	 */
	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
468 469
					      unsigned long start,
					      unsigned long end)
470 471
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
472
	int young, idx;
473

474
	idx = srcu_read_lock(&kvm->srcu);
475 476
	spin_lock(&kvm->mmu_lock);

A
Andres Lagar-Cavilla 已提交
477
	young = kvm_age_hva(kvm, start, end);
478 479 480
	if (young)
		kvm_flush_remote_tlbs(kvm);

481 482 483
	spin_unlock(&kvm->mmu_lock);
	srcu_read_unlock(&kvm->srcu, idx);

484 485 486
	return young;
}

487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516
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
517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532
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;
}

533 534 535 536
static void kvm_mmu_notifier_release(struct mmu_notifier *mn,
				     struct mm_struct *mm)
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
537 538 539
	int idx;

	idx = srcu_read_lock(&kvm->srcu);
540
	kvm_arch_flush_shadow_all(kvm);
541
	srcu_read_unlock(&kvm->srcu, idx);
542 543
}

544
static const struct mmu_notifier_ops kvm_mmu_notifier_ops = {
545
	.invalidate_range	= kvm_mmu_notifier_invalidate_range,
546 547 548
	.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,
549
	.clear_young		= kvm_mmu_notifier_clear_young,
A
Andrea Arcangeli 已提交
550
	.test_young		= kvm_mmu_notifier_test_young,
551
	.change_pte		= kvm_mmu_notifier_change_pte,
552
	.release		= kvm_mmu_notifier_release,
553
};
554 555 556 557 558 559 560 561 562 563 564 565 566 567

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;
}

568 569
#endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */

570
static struct kvm_memslots *kvm_alloc_memslots(void)
571 572
{
	int i;
573
	struct kvm_memslots *slots;
574

575
	slots = kvzalloc(sizeof(struct kvm_memslots), GFP_KERNEL_ACCOUNT);
576 577 578
	if (!slots)
		return NULL;

579
	for (i = 0; i < KVM_MEM_SLOTS_NUM; i++)
580
		slots->id_to_index[i] = -1;
581 582 583 584 585 586 587 588 589 590 591 592 593

	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;
}

594
static void kvm_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot)
595
{
596
	kvm_destroy_dirty_bitmap(slot);
597

598
	kvm_arch_free_memslot(kvm, slot);
599

600 601
	slot->flags = 0;
	slot->npages = 0;
602 603 604 605 606 607 608 609 610 611
}

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)
612
		kvm_free_memslot(kvm, memslot);
613 614

	kvfree(slots);
615 616
}

617 618 619 620 621 622 623 624 625
static void kvm_destroy_vm_debugfs(struct kvm *kvm)
{
	int i;

	if (!kvm->debugfs_dentry)
		return;

	debugfs_remove_recursive(kvm->debugfs_dentry);

626 627 628 629 630
	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);
	}
631 632 633 634 635 636 637 638 639 640 641 642
}

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);
643
	kvm->debugfs_dentry = debugfs_create_dir(dir_name, kvm_debugfs_dir);
644 645 646

	kvm->debugfs_stat_data = kcalloc(kvm_debugfs_num_entries,
					 sizeof(*kvm->debugfs_stat_data),
647
					 GFP_KERNEL_ACCOUNT);
648 649 650 651
	if (!kvm->debugfs_stat_data)
		return -ENOMEM;

	for (p = debugfs_entries; p->name; p++) {
652
		stat_data = kzalloc(sizeof(*stat_data), GFP_KERNEL_ACCOUNT);
653 654 655 656
		if (!stat_data)
			return -ENOMEM;

		stat_data->kvm = kvm;
657
		stat_data->dbgfs_item = p;
658
		kvm->debugfs_stat_data[p - debugfs_entries] = stat_data;
659 660 661
		debugfs_create_file(p->name, KVM_DBGFS_GET_MODE(p),
				    kvm->debugfs_dentry, stat_data,
				    &stat_fops_per_vm);
662 663 664 665
	}
	return 0;
}

666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682
/*
 * 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)
{
}

683
static struct kvm *kvm_create_vm(unsigned long type)
A
Avi Kivity 已提交
684
{
685
	struct kvm *kvm = kvm_arch_alloc_vm();
686 687
	int r = -ENOMEM;
	int i;
A
Avi Kivity 已提交
688

689 690 691
	if (!kvm)
		return ERR_PTR(-ENOMEM);

692
	spin_lock_init(&kvm->mmu_lock);
V
Vegard Nossum 已提交
693
	mmgrab(current->mm);
694 695 696 697 698 699 700
	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);

701 702
	BUILD_BUG_ON(KVM_MEM_SLOTS_NUM > SHRT_MAX);

703 704 705 706 707
	if (init_srcu_struct(&kvm->srcu))
		goto out_err_no_srcu;
	if (init_srcu_struct(&kvm->irq_srcu))
		goto out_err_no_irq_srcu;

708
	refcount_set(&kvm->users_count, 1);
709
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
710
		struct kvm_memslots *slots = kvm_alloc_memslots();
711

712
		if (!slots)
713
			goto out_err_no_arch_destroy_vm;
714
		/* Generations must be different for each address space. */
715
		slots->generation = i;
716
		rcu_assign_pointer(kvm->memslots[i], slots);
717
	}
718

M
Marcelo Tosatti 已提交
719
	for (i = 0; i < KVM_NR_BUSES; i++) {
720
		rcu_assign_pointer(kvm->buses[i],
721
			kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL_ACCOUNT));
722
		if (!kvm->buses[i])
723
			goto out_err_no_arch_destroy_vm;
M
Marcelo Tosatti 已提交
724
	}
725

726 727
	kvm->max_halt_poll_ns = halt_poll_ns;

728
	r = kvm_arch_init_vm(kvm, type);
729
	if (r)
730
		goto out_err_no_arch_destroy_vm;
731 732 733

	r = hardware_enable_all();
	if (r)
734
		goto out_err_no_disable;
735

736
#ifdef CONFIG_HAVE_KVM_IRQFD
737
	INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list);
738
#endif
A
Avi Kivity 已提交
739

740
	r = kvm_init_mmu_notifier(kvm);
741 742 743 744
	if (r)
		goto out_err_no_mmu_notifier;

	r = kvm_arch_post_init_vm(kvm);
745 746 747
	if (r)
		goto out_err;

J
Junaid Shahid 已提交
748
	mutex_lock(&kvm_lock);
749
	list_add(&kvm->vm_list, &vm_list);
J
Junaid Shahid 已提交
750
	mutex_unlock(&kvm_lock);
751

752 753
	preempt_notifier_inc();

754
	return kvm;
755 756

out_err:
757 758 759 760 761
#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:
762
	hardware_disable_all();
763
out_err_no_disable:
764 765
	kvm_arch_destroy_vm(kvm);
out_err_no_arch_destroy_vm:
766
	WARN_ON_ONCE(!refcount_dec_and_test(&kvm->users_count));
M
Marcelo Tosatti 已提交
767
	for (i = 0; i < KVM_NR_BUSES; i++)
768
		kfree(kvm_get_bus(kvm, i));
769
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
770
		kvm_free_memslots(kvm, __kvm_memslots(kvm, i));
771 772 773 774
	cleanup_srcu_struct(&kvm->irq_srcu);
out_err_no_irq_srcu:
	cleanup_srcu_struct(&kvm->srcu);
out_err_no_srcu:
775
	kvm_arch_free_vm(kvm);
776
	mmdrop(current->mm);
777
	return ERR_PTR(r);
778 779
}

780 781
static void kvm_destroy_devices(struct kvm *kvm)
{
G
Geliang Tang 已提交
782
	struct kvm_device *dev, *tmp;
783

784 785 786 787 788
	/*
	 * 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
789 790
	list_for_each_entry_safe(dev, tmp, &kvm->devices, vm_node) {
		list_del(&dev->vm_node);
791 792 793 794
		dev->ops->destroy(dev);
	}
}

795 796
static void kvm_destroy_vm(struct kvm *kvm)
{
M
Marcelo Tosatti 已提交
797
	int i;
798 799
	struct mm_struct *mm = kvm->mm;

800
	kvm_uevent_notify_change(KVM_EVENT_DESTROY_VM, kvm);
801
	kvm_destroy_vm_debugfs(kvm);
802
	kvm_arch_sync_events(kvm);
J
Junaid Shahid 已提交
803
	mutex_lock(&kvm_lock);
804
	list_del(&kvm->vm_list);
J
Junaid Shahid 已提交
805
	mutex_unlock(&kvm_lock);
806 807
	kvm_arch_pre_destroy_vm(kvm);

808
	kvm_free_irq_routing(kvm);
809
	for (i = 0; i < KVM_NR_BUSES; i++) {
810
		struct kvm_io_bus *bus = kvm_get_bus(kvm, i);
811 812 813

		if (bus)
			kvm_io_bus_destroy(bus);
814 815
		kvm->buses[i] = NULL;
	}
816
	kvm_coalesced_mmio_free(kvm);
817 818
#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
	mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
819
#else
820
	kvm_arch_flush_shadow_all(kvm);
821
#endif
822
	kvm_arch_destroy_vm(kvm);
823
	kvm_destroy_devices(kvm);
824
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
825
		kvm_free_memslots(kvm, __kvm_memslots(kvm, i));
826
	cleanup_srcu_struct(&kvm->irq_srcu);
827 828
	cleanup_srcu_struct(&kvm->srcu);
	kvm_arch_free_vm(kvm);
829
	preempt_notifier_dec();
830
	hardware_disable_all();
831
	mmdrop(mm);
832 833
}

I
Izik Eidus 已提交
834 835
void kvm_get_kvm(struct kvm *kvm)
{
836
	refcount_inc(&kvm->users_count);
I
Izik Eidus 已提交
837 838 839 840 841
}
EXPORT_SYMBOL_GPL(kvm_get_kvm);

void kvm_put_kvm(struct kvm *kvm)
{
842
	if (refcount_dec_and_test(&kvm->users_count))
I
Izik Eidus 已提交
843 844 845 846
		kvm_destroy_vm(kvm);
}
EXPORT_SYMBOL_GPL(kvm_put_kvm);

847 848 849 850 851 852 853 854 855 856 857 858
/*
 * 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
859

860 861 862 863
static int kvm_vm_release(struct inode *inode, struct file *filp)
{
	struct kvm *kvm = filp->private_data;

G
Gregory Haskins 已提交
864 865
	kvm_irqfd_release(kvm);

I
Izik Eidus 已提交
866
	kvm_put_kvm(kvm);
A
Avi Kivity 已提交
867 868 869
	return 0;
}

870 871
/*
 * Allocation size is twice as large as the actual dirty bitmap size.
872
 * See kvm_vm_ioctl_get_dirty_log() why this is needed.
873
 */
874
static int kvm_alloc_dirty_bitmap(struct kvm_memory_slot *memslot)
875
{
876
	unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot);
877

878
	memslot->dirty_bitmap = kvzalloc(dirty_bytes, GFP_KERNEL_ACCOUNT);
879 880 881 882 883 884
	if (!memslot->dirty_bitmap)
		return -ENOMEM;

	return 0;
}

885
/*
886 887
 * Delete a memslot by decrementing the number of used slots and shifting all
 * other entries in the array forward one spot.
888
 */
889 890
static inline void kvm_memslot_delete(struct kvm_memslots *slots,
				      struct kvm_memory_slot *memslot)
891
{
892
	struct kvm_memory_slot *mslots = slots->memslots;
893
	int i;
894

895 896
	if (WARN_ON(slots->id_to_index[memslot->id] == -1))
		return;
897

898 899
	slots->used_slots--;

900 901 902
	if (atomic_read(&slots->lru_slot) >= slots->used_slots)
		atomic_set(&slots->lru_slot, 0);

903
	for (i = slots->id_to_index[memslot->id]; i < slots->used_slots; i++) {
904 905 906
		mslots[i] = mslots[i + 1];
		slots->id_to_index[mslots[i].id] = i;
	}
907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935
	mslots[i] = *memslot;
	slots->id_to_index[memslot->id] = -1;
}

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

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

	if (WARN_ON_ONCE(slots->id_to_index[memslot->id] == -1) ||
	    WARN_ON_ONCE(!slots->used_slots))
		return -1;
936 937

	/*
938 939 940
	 * Move the target memslot backward in the array by shifting existing
	 * memslots with a higher GFN (than the target memslot) towards the
	 * front of the array.
941
	 */
942 943 944 945 946
	for (i = slots->id_to_index[memslot->id]; i < slots->used_slots - 1; i++) {
		if (memslot->base_gfn > mslots[i + 1].base_gfn)
			break;

		WARN_ON_ONCE(memslot->base_gfn == mslots[i + 1].base_gfn);
947

948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044
		/* Shift the next memslot forward one and update its index. */
		mslots[i] = mslots[i + 1];
		slots->id_to_index[mslots[i].id] = i;
	}
	return i;
}

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

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

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

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

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

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

		/*
		 * Copy the memslot to its new position in memslots and update
		 * its index accordingly.
		 */
		slots->memslots[i] = *memslot;
		slots->id_to_index[memslot->id] = i;
	}
1045 1046
}

1047
static int check_memory_region_flags(const struct kvm_userspace_memory_region *mem)
1048
{
X
Xiao Guangrong 已提交
1049 1050
	u32 valid_flags = KVM_MEM_LOG_DIRTY_PAGES;

1051
#ifdef __KVM_HAVE_READONLY_MEM
X
Xiao Guangrong 已提交
1052 1053 1054 1055
	valid_flags |= KVM_MEM_READONLY;
#endif

	if (mem->flags & ~valid_flags)
1056 1057 1058 1059 1060
		return -EINVAL;

	return 0;
}

1061
static struct kvm_memslots *install_new_memslots(struct kvm *kvm,
1062
		int as_id, struct kvm_memslots *slots)
1063
{
1064
	struct kvm_memslots *old_memslots = __kvm_memslots(kvm, as_id);
1065
	u64 gen = old_memslots->generation;
1066

1067 1068
	WARN_ON(gen & KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS);
	slots->generation = gen | KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS;
1069

1070
	rcu_assign_pointer(kvm->memslots[as_id], slots);
1071
	synchronize_srcu_expedited(&kvm->srcu);
1072

1073
	/*
1074
	 * Increment the new memslot generation a second time, dropping the
M
Miaohe Lin 已提交
1075
	 * update in-progress flag and incrementing the generation based on
1076 1077 1078 1079 1080 1081
	 * 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;

	/*
1082 1083 1084
	 * 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
1085 1086
	 * space 0 will use generations 0, 2, 4, ... while address space 1 will
	 * use generations 1, 3, 5, ...
1087
	 */
1088
	gen += KVM_ADDRESS_SPACE_NUM;
1089

1090
	kvm_arch_memslots_updated(kvm, gen);
1091

1092
	slots->generation = gen;
1093 1094

	return old_memslots;
1095 1096
}

1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122
/*
 * Note, at a minimum, the current number of used slots must be allocated, even
 * when deleting a memslot, as we need a complete duplicate of the memslots for
 * use when invalidating a memslot prior to deleting/moving the memslot.
 */
static struct kvm_memslots *kvm_dup_memslots(struct kvm_memslots *old,
					     enum kvm_mr_change change)
{
	struct kvm_memslots *slots;
	size_t old_size, new_size;

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

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

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

	return slots;
}

1123 1124
static int kvm_set_memslot(struct kvm *kvm,
			   const struct kvm_userspace_memory_region *mem,
1125
			   struct kvm_memory_slot *old,
1126 1127 1128 1129 1130 1131 1132
			   struct kvm_memory_slot *new, int as_id,
			   enum kvm_mr_change change)
{
	struct kvm_memory_slot *slot;
	struct kvm_memslots *slots;
	int r;

1133
	slots = kvm_dup_memslots(__kvm_memslots(kvm, as_id), change);
1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181
	if (!slots)
		return -ENOMEM;

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

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

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

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

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

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

	kvfree(slots);
	return 0;

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

1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198
static int kvm_delete_memslot(struct kvm *kvm,
			      const struct kvm_userspace_memory_region *mem,
			      struct kvm_memory_slot *old, int as_id)
{
	struct kvm_memory_slot new;
	int r;

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

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

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

1199
	kvm_free_memslot(kvm, old);
1200 1201 1202
	return 0;
}

A
Avi Kivity 已提交
1203 1204 1205 1206 1207
/*
 * Allocate some memory and give it an address in the guest physical address
 * space.
 *
 * Discontiguous memory is allowed, mostly for framebuffers.
1208
 *
1209
 * Must be called holding kvm->slots_lock for write.
A
Avi Kivity 已提交
1210
 */
1211
int __kvm_set_memory_region(struct kvm *kvm,
1212
			    const struct kvm_userspace_memory_region *mem)
A
Avi Kivity 已提交
1213 1214
{
	struct kvm_memory_slot old, new;
1215
	struct kvm_memory_slot *tmp;
1216
	enum kvm_mr_change change;
1217 1218
	int as_id, id;
	int r;
A
Avi Kivity 已提交
1219

1220 1221
	r = check_memory_region_flags(mem);
	if (r)
1222
		return r;
1223

1224 1225 1226
	as_id = mem->slot >> 16;
	id = (u16)mem->slot;

A
Avi Kivity 已提交
1227 1228
	/* General sanity checks */
	if (mem->memory_size & (PAGE_SIZE - 1))
1229
		return -EINVAL;
A
Avi Kivity 已提交
1230
	if (mem->guest_phys_addr & (PAGE_SIZE - 1))
1231
		return -EINVAL;
1232
	/* We can read the guest memory with __xxx_user() later on. */
1233
	if ((mem->userspace_addr & (PAGE_SIZE - 1)) ||
1234
	     !access_ok((void __user *)(unsigned long)mem->userspace_addr,
1235
			mem->memory_size))
1236
		return -EINVAL;
1237
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_MEM_SLOTS_NUM)
1238
		return -EINVAL;
A
Avi Kivity 已提交
1239
	if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
1240
		return -EINVAL;
A
Avi Kivity 已提交
1241

1242 1243 1244 1245
	/*
	 * Make a full copy of the old memslot, the pointer will become stale
	 * when the memslots are re-sorted by update_memslots(), and the old
	 * memslot needs to be referenced after calling update_memslots(), e.g.
1246
	 * to free its resources and for arch specific behavior.
1247
	 */
1248 1249 1250 1251 1252 1253 1254 1255
	tmp = id_to_memslot(__kvm_memslots(kvm, as_id), id);
	if (tmp) {
		old = *tmp;
		tmp = NULL;
	} else {
		memset(&old, 0, sizeof(old));
		old.id = id;
	}
1256

1257 1258 1259
	if (!mem->memory_size)
		return kvm_delete_memslot(kvm, mem, &old, as_id);

1260
	new.id = id;
1261 1262
	new.base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
	new.npages = mem->memory_size >> PAGE_SHIFT;
A
Avi Kivity 已提交
1263
	new.flags = mem->flags;
1264
	new.userspace_addr = mem->userspace_addr;
A
Avi Kivity 已提交
1265

1266 1267 1268
	if (new.npages > KVM_MEM_MAX_NR_PAGES)
		return -EINVAL;

1269 1270
	if (!old.npages) {
		change = KVM_MR_CREATE;
1271 1272
		new.dirty_bitmap = NULL;
		memset(&new.arch, 0, sizeof(new.arch));
1273 1274
	} else { /* Modify an existing slot. */
		if ((new.userspace_addr != old.userspace_addr) ||
1275
		    (new.npages != old.npages) ||
1276
		    ((new.flags ^ old.flags) & KVM_MEM_READONLY))
1277
			return -EINVAL;
1278

1279
		if (new.base_gfn != old.base_gfn)
1280 1281 1282 1283 1284
			change = KVM_MR_MOVE;
		else if (new.flags != old.flags)
			change = KVM_MR_FLAGS_ONLY;
		else /* Nothing to change. */
			return 0;
1285 1286 1287 1288

		/* Copy dirty_bitmap and arch from the current memslot. */
		new.dirty_bitmap = old.dirty_bitmap;
		memcpy(&new.arch, &old.arch, sizeof(new.arch));
1289
	}
A
Avi Kivity 已提交
1290

1291
	if ((change == KVM_MR_CREATE) || (change == KVM_MR_MOVE)) {
1292
		/* Check for overlaps */
1293 1294
		kvm_for_each_memslot(tmp, __kvm_memslots(kvm, as_id)) {
			if (tmp->id == id)
1295
				continue;
1296 1297
			if (!((new.base_gfn + new.npages <= tmp->base_gfn) ||
			      (new.base_gfn >= tmp->base_gfn + tmp->npages)))
1298
				return -EEXIST;
1299
		}
A
Avi Kivity 已提交
1300 1301
	}

1302 1303 1304 1305
	/* Allocate/free page dirty bitmap as needed */
	if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
		new.dirty_bitmap = NULL;
	else if (!new.dirty_bitmap) {
1306
		r = kvm_alloc_dirty_bitmap(&new);
1307 1308
		if (r)
			return r;
1309 1310 1311

		if (kvm_dirty_log_manual_protect_and_init_set(kvm))
			bitmap_set(new.dirty_bitmap, 0, new.npages);
A
Avi Kivity 已提交
1312 1313
	}

1314 1315 1316
	r = kvm_set_memslot(kvm, mem, &old, &new, as_id, change);
	if (r)
		goto out_bitmap;
1317

1318 1319
	if (old.dirty_bitmap && !new.dirty_bitmap)
		kvm_destroy_dirty_bitmap(&old);
A
Avi Kivity 已提交
1320 1321
	return 0;

1322 1323 1324
out_bitmap:
	if (new.dirty_bitmap && !old.dirty_bitmap)
		kvm_destroy_dirty_bitmap(&new);
A
Avi Kivity 已提交
1325
	return r;
1326
}
1327 1328 1329
EXPORT_SYMBOL_GPL(__kvm_set_memory_region);

int kvm_set_memory_region(struct kvm *kvm,
1330
			  const struct kvm_userspace_memory_region *mem)
1331 1332 1333
{
	int r;

1334
	mutex_lock(&kvm->slots_lock);
1335
	r = __kvm_set_memory_region(kvm, mem);
1336
	mutex_unlock(&kvm->slots_lock);
1337 1338
	return r;
}
1339 1340
EXPORT_SYMBOL_GPL(kvm_set_memory_region);

1341 1342
static int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
					  struct kvm_userspace_memory_region *mem)
1343
{
1344
	if ((u16)mem->slot >= KVM_USER_MEM_SLOTS)
1345
		return -EINVAL;
1346

1347
	return kvm_set_memory_region(kvm, mem);
A
Avi Kivity 已提交
1348 1349
}

1350
#ifndef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
1351 1352 1353 1354 1355 1356 1357 1358 1359
/**
 * kvm_get_dirty_log - get a snapshot of dirty pages
 * @kvm:	pointer to kvm instance
 * @log:	slot id and address to which we copy the log
 * @is_dirty:	set to '1' if any dirty pages were found
 * @memslot:	set to the associated memslot, always valid on success
 */
int kvm_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log,
		      int *is_dirty, struct kvm_memory_slot **memslot)
A
Avi Kivity 已提交
1360
{
1361
	struct kvm_memslots *slots;
1362
	int i, as_id, id;
1363
	unsigned long n;
A
Avi Kivity 已提交
1364 1365
	unsigned long any = 0;

1366 1367 1368
	*memslot = NULL;
	*is_dirty = 0;

1369 1370 1371
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1372
		return -EINVAL;
A
Avi Kivity 已提交
1373

1374
	slots = __kvm_memslots(kvm, as_id);
1375
	*memslot = id_to_memslot(slots, id);
1376
	if (!(*memslot) || !(*memslot)->dirty_bitmap)
1377
		return -ENOENT;
A
Avi Kivity 已提交
1378

1379 1380 1381
	kvm_arch_sync_dirty_log(kvm, *memslot);

	n = kvm_dirty_bitmap_bytes(*memslot);
A
Avi Kivity 已提交
1382

1383
	for (i = 0; !any && i < n/sizeof(long); ++i)
1384
		any = (*memslot)->dirty_bitmap[i];
A
Avi Kivity 已提交
1385

1386
	if (copy_to_user(log->dirty_bitmap, (*memslot)->dirty_bitmap, n))
1387
		return -EFAULT;
A
Avi Kivity 已提交
1388

1389 1390
	if (any)
		*is_dirty = 1;
1391
	return 0;
A
Avi Kivity 已提交
1392
}
1393
EXPORT_SYMBOL_GPL(kvm_get_dirty_log);
A
Avi Kivity 已提交
1394

1395
#else /* CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
1396
/**
J
Jiang Biao 已提交
1397
 * kvm_get_dirty_log_protect - get a snapshot of dirty pages
1398
 *	and reenable dirty page tracking for the corresponding pages.
1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416
 * @kvm:	pointer to kvm instance
 * @log:	slot id and address to which we copy the log
 *
 * We need to keep it in mind that VCPU threads can write to the bitmap
 * concurrently. So, to avoid losing track of dirty pages we keep the
 * following order:
 *
 *    1. Take a snapshot of the bit and clear it if needed.
 *    2. Write protect the corresponding page.
 *    3. Copy the snapshot to the userspace.
 *    4. Upon return caller flushes TLB's if needed.
 *
 * Between 2 and 4, the guest may write to the page using the remaining TLB
 * entry.  This is not a problem because the page is reported dirty using
 * the snapshot taken before and step 4 ensures that writes done after
 * exiting to userspace will be logged for the next call.
 *
 */
1417
static int kvm_get_dirty_log_protect(struct kvm *kvm, struct kvm_dirty_log *log)
1418
{
1419
	struct kvm_memslots *slots;
1420
	struct kvm_memory_slot *memslot;
1421
	int i, as_id, id;
1422 1423 1424
	unsigned long n;
	unsigned long *dirty_bitmap;
	unsigned long *dirty_bitmap_buffer;
1425
	bool flush;
1426

1427 1428 1429
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1430
		return -EINVAL;
1431

1432 1433
	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
1434 1435
	if (!memslot || !memslot->dirty_bitmap)
		return -ENOENT;
1436 1437 1438

	dirty_bitmap = memslot->dirty_bitmap;

1439 1440
	kvm_arch_sync_dirty_log(kvm, memslot);

1441
	n = kvm_dirty_bitmap_bytes(memslot);
1442
	flush = false;
1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455
	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);
1456

1457 1458 1459 1460
		spin_lock(&kvm->mmu_lock);
		for (i = 0; i < n / sizeof(long); i++) {
			unsigned long mask;
			gfn_t offset;
1461

1462 1463 1464
			if (!dirty_bitmap[i])
				continue;

1465
			flush = true;
1466 1467 1468
			mask = xchg(&dirty_bitmap[i], 0);
			dirty_bitmap_buffer[i] = mask;

1469 1470 1471
			offset = i * BITS_PER_LONG;
			kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot,
								offset, mask);
1472 1473 1474 1475
		}
		spin_unlock(&kvm->mmu_lock);
	}

1476 1477 1478
	if (flush)
		kvm_arch_flush_remote_tlbs_memslot(kvm, memslot);

1479 1480 1481 1482
	if (copy_to_user(log->dirty_bitmap, dirty_bitmap_buffer, n))
		return -EFAULT;
	return 0;
}
1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515


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

	mutex_lock(&kvm->slots_lock);

	r = kvm_get_dirty_log_protect(kvm, log);

	mutex_unlock(&kvm->slots_lock);
	return r;
}
1516 1517 1518 1519 1520 1521 1522

/**
 * 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
 */
1523 1524
static int kvm_clear_dirty_log_protect(struct kvm *kvm,
				       struct kvm_clear_dirty_log *log)
1525 1526 1527
{
	struct kvm_memslots *slots;
	struct kvm_memory_slot *memslot;
1528
	int as_id, id;
1529
	gfn_t offset;
1530
	unsigned long i, n;
1531 1532
	unsigned long *dirty_bitmap;
	unsigned long *dirty_bitmap_buffer;
1533
	bool flush;
1534 1535 1536 1537 1538 1539

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

1540
	if (log->first_page & 63)
1541 1542 1543 1544
		return -EINVAL;

	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
1545 1546
	if (!memslot || !memslot->dirty_bitmap)
		return -ENOENT;
1547 1548 1549

	dirty_bitmap = memslot->dirty_bitmap;

1550
	n = ALIGN(log->num_pages, BITS_PER_LONG) / 8;
1551 1552

	if (log->first_page > memslot->npages ||
1553 1554 1555
	    log->num_pages > memslot->npages - log->first_page ||
	    (log->num_pages < memslot->npages - log->first_page && (log->num_pages & 63)))
	    return -EINVAL;
1556

1557 1558 1559
	kvm_arch_sync_dirty_log(kvm, memslot);

	flush = false;
1560 1561 1562
	dirty_bitmap_buffer = kvm_second_dirty_bitmap(memslot);
	if (copy_from_user(dirty_bitmap_buffer, log->dirty_bitmap, n))
		return -EFAULT;
1563

1564
	spin_lock(&kvm->mmu_lock);
1565 1566
	for (offset = log->first_page, i = offset / BITS_PER_LONG,
		 n = DIV_ROUND_UP(log->num_pages, BITS_PER_LONG); n--;
1567 1568 1569 1570
	     i++, offset += BITS_PER_LONG) {
		unsigned long mask = *dirty_bitmap_buffer++;
		atomic_long_t *p = (atomic_long_t *) &dirty_bitmap[i];
		if (!mask)
1571 1572
			continue;

1573
		mask &= atomic_long_fetch_andnot(mask, p);
1574

1575 1576 1577 1578 1579 1580
		/*
		 * 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.
		*/
1581
		if (mask) {
1582
			flush = true;
1583 1584 1585
			kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot,
								offset, mask);
		}
1586 1587
	}
	spin_unlock(&kvm->mmu_lock);
1588

1589 1590 1591
	if (flush)
		kvm_arch_flush_remote_tlbs_memslot(kvm, memslot);

1592
	return 0;
1593
}
1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607

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

	mutex_lock(&kvm->slots_lock);

	r = kvm_clear_dirty_log_protect(kvm, log);

	mutex_unlock(&kvm->slots_lock);
	return r;
}
#endif /* CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
1608

1609 1610 1611 1612
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
1613
EXPORT_SYMBOL_GPL(gfn_to_memslot);
A
Avi Kivity 已提交
1614

1615 1616 1617 1618
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);
}
1619
EXPORT_SYMBOL_GPL(kvm_vcpu_gfn_to_memslot);
1620

1621
bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
1622
{
1623
	struct kvm_memory_slot *memslot = gfn_to_memslot(kvm, gfn);
1624

1625
	return kvm_is_visible_memslot(memslot);
1626 1627 1628
}
EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);

1629
unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn)
J
Joerg Roedel 已提交
1630 1631 1632 1633 1634 1635
{
	struct vm_area_struct *vma;
	unsigned long addr, size;

	size = PAGE_SIZE;

1636
	addr = kvm_vcpu_gfn_to_hva_prot(vcpu, gfn, NULL);
J
Joerg Roedel 已提交
1637 1638 1639
	if (kvm_is_error_hva(addr))
		return PAGE_SIZE;

1640
	mmap_read_lock(current->mm);
J
Joerg Roedel 已提交
1641 1642 1643 1644 1645 1646 1647
	vma = find_vma(current->mm, addr);
	if (!vma)
		goto out;

	size = vma_kernel_pagesize(vma);

out:
1648
	mmap_read_unlock(current->mm);
J
Joerg Roedel 已提交
1649 1650 1651 1652

	return size;
}

X
Xiao Guangrong 已提交
1653 1654 1655 1656 1657 1658 1659
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
1660
{
1661
	if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
X
Xiao Guangrong 已提交
1662
		return KVM_HVA_ERR_BAD;
1663

X
Xiao Guangrong 已提交
1664 1665
	if (memslot_is_readonly(slot) && write)
		return KVM_HVA_ERR_RO_BAD;
1666 1667 1668 1669

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

X
Xiao Guangrong 已提交
1670
	return __gfn_to_hva_memslot(slot, gfn);
I
Izik Eidus 已提交
1671
}
1672

X
Xiao Guangrong 已提交
1673 1674 1675 1676
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
1677
}
1678

X
Xiao Guangrong 已提交
1679
unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot,
1680
					gfn_t gfn)
X
Xiao Guangrong 已提交
1681 1682 1683 1684 1685
{
	return gfn_to_hva_many(slot, gfn, NULL);
}
EXPORT_SYMBOL_GPL(gfn_to_hva_memslot);

1686 1687
unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
{
1688
	return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL);
1689
}
1690
EXPORT_SYMBOL_GPL(gfn_to_hva);
I
Izik Eidus 已提交
1691

1692 1693 1694 1695 1696 1697
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);

1698
/*
1699 1700 1701 1702 1703 1704
 * 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
1705
 */
1706 1707
unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot,
				      gfn_t gfn, bool *writable)
1708
{
1709 1710 1711
	unsigned long hva = __gfn_to_hva_many(slot, gfn, NULL, false);

	if (!kvm_is_error_hva(hva) && writable)
1712 1713
		*writable = !memslot_is_readonly(slot);

1714
	return hva;
1715 1716
}

1717 1718 1719 1720 1721 1722 1723
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);
}

1724 1725 1726 1727 1728 1729 1730
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);
}

1731 1732
static inline int check_user_page_hwpoison(unsigned long addr)
{
L
Lorenzo Stoakes 已提交
1733
	int rc, flags = FOLL_HWPOISON | FOLL_WRITE;
1734

L
Lorenzo Stoakes 已提交
1735
	rc = get_user_pages(addr, 1, flags, NULL, NULL);
1736 1737 1738
	return rc == -EHWPOISON;
}

X
Xiao Guangrong 已提交
1739
/*
1740 1741
 * 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
1742
 * only part that runs if we can in atomic context.
X
Xiao Guangrong 已提交
1743
 */
1744 1745
static bool hva_to_pfn_fast(unsigned long addr, bool write_fault,
			    bool *writable, kvm_pfn_t *pfn)
A
Avi Kivity 已提交
1746
{
1747
	struct page *page[1];
A
Avi Kivity 已提交
1748

1749 1750 1751 1752 1753 1754 1755
	/*
	 * 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;
1756

1757
	if (get_user_page_fast_only(addr, FOLL_WRITE, page)) {
X
Xiao Guangrong 已提交
1758
		*pfn = page_to_pfn(page[0]);
1759

X
Xiao Guangrong 已提交
1760 1761 1762 1763
		if (writable)
			*writable = true;
		return true;
	}
1764

X
Xiao Guangrong 已提交
1765 1766
	return false;
}
1767

X
Xiao Guangrong 已提交
1768 1769 1770 1771 1772
/*
 * 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
1773
			   bool *writable, kvm_pfn_t *pfn)
X
Xiao Guangrong 已提交
1774
{
1775 1776
	unsigned int flags = FOLL_HWPOISON;
	struct page *page;
X
Xiao Guangrong 已提交
1777
	int npages = 0;
1778

X
Xiao Guangrong 已提交
1779 1780 1781 1782 1783
	might_sleep();

	if (writable)
		*writable = write_fault;

1784 1785 1786 1787
	if (write_fault)
		flags |= FOLL_WRITE;
	if (async)
		flags |= FOLL_NOWAIT;
1788

1789
	npages = get_user_pages_unlocked(addr, 1, &page, flags);
X
Xiao Guangrong 已提交
1790 1791 1792 1793
	if (npages != 1)
		return npages;

	/* map read fault as writable if possible */
1794
	if (unlikely(!write_fault) && writable) {
1795
		struct page *wpage;
X
Xiao Guangrong 已提交
1796

1797
		if (get_user_page_fast_only(addr, FOLL_WRITE, &wpage)) {
X
Xiao Guangrong 已提交
1798
			*writable = true;
1799 1800
			put_page(page);
			page = wpage;
1801
		}
1802
	}
1803
	*pfn = page_to_pfn(page);
X
Xiao Guangrong 已提交
1804 1805
	return npages;
}
I
Izik Eidus 已提交
1806

X
Xiao Guangrong 已提交
1807 1808 1809 1810
static bool vma_is_valid(struct vm_area_struct *vma, bool write_fault)
{
	if (unlikely(!(vma->vm_flags & VM_READ)))
		return false;
1811

X
Xiao Guangrong 已提交
1812 1813
	if (write_fault && (unlikely(!(vma->vm_flags & VM_WRITE))))
		return false;
1814

X
Xiao Guangrong 已提交
1815 1816
	return true;
}
1817

1818 1819
static int hva_to_pfn_remapped(struct vm_area_struct *vma,
			       unsigned long addr, bool *async,
1820 1821
			       bool write_fault, bool *writable,
			       kvm_pfn_t *p_pfn)
1822
{
1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835
	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);
1836 1837
		if (unlocked)
			return -EAGAIN;
1838 1839 1840 1841 1842 1843 1844 1845 1846
		if (r)
			return r;

		r = follow_pfn(vma, addr, &pfn);
		if (r)
			return r;

	}

1847 1848
	if (writable)
		*writable = true;
1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863

	/*
	 * 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;
1864 1865 1866
	return 0;
}

1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880
/*
 * 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
1881
static kvm_pfn_t hva_to_pfn(unsigned long addr, bool atomic, bool *async,
X
Xiao Guangrong 已提交
1882 1883 1884
			bool write_fault, bool *writable)
{
	struct vm_area_struct *vma;
D
Dan Williams 已提交
1885
	kvm_pfn_t pfn = 0;
1886
	int npages, r;
1887

X
Xiao Guangrong 已提交
1888 1889
	/* we can do it either atomically or asynchronously, not both */
	BUG_ON(atomic && async);
1890

1891
	if (hva_to_pfn_fast(addr, write_fault, writable, &pfn))
X
Xiao Guangrong 已提交
1892 1893 1894 1895 1896 1897 1898 1899
		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;
1900

1901
	mmap_read_lock(current->mm);
X
Xiao Guangrong 已提交
1902 1903 1904 1905 1906 1907
	if (npages == -EHWPOISON ||
	      (!async && check_user_page_hwpoison(addr))) {
		pfn = KVM_PFN_ERR_HWPOISON;
		goto exit;
	}

1908
retry:
X
Xiao Guangrong 已提交
1909 1910 1911 1912
	vma = find_vma_intersection(current->mm, addr, addr + 1);

	if (vma == NULL)
		pfn = KVM_PFN_ERR_FAULT;
1913
	else if (vma->vm_flags & (VM_IO | VM_PFNMAP)) {
1914
		r = hva_to_pfn_remapped(vma, addr, async, write_fault, writable, &pfn);
1915 1916
		if (r == -EAGAIN)
			goto retry;
1917 1918
		if (r < 0)
			pfn = KVM_PFN_ERR_FAULT;
X
Xiao Guangrong 已提交
1919
	} else {
X
Xiao Guangrong 已提交
1920
		if (async && vma_is_valid(vma, write_fault))
X
Xiao Guangrong 已提交
1921 1922 1923 1924
			*async = true;
		pfn = KVM_PFN_ERR_FAULT;
	}
exit:
1925
	mmap_read_unlock(current->mm);
1926
	return pfn;
1927 1928
}

D
Dan Williams 已提交
1929 1930 1931
kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
			       bool atomic, bool *async, bool write_fault,
			       bool *writable)
1932
{
X
Xiao Guangrong 已提交
1933 1934
	unsigned long addr = __gfn_to_hva_many(slot, gfn, NULL, write_fault);

1935 1936 1937
	if (addr == KVM_HVA_ERR_RO_BAD) {
		if (writable)
			*writable = false;
X
Xiao Guangrong 已提交
1938
		return KVM_PFN_ERR_RO_FAULT;
1939
	}
X
Xiao Guangrong 已提交
1940

1941 1942 1943
	if (kvm_is_error_hva(addr)) {
		if (writable)
			*writable = false;
1944
		return KVM_PFN_NOSLOT;
1945
	}
X
Xiao Guangrong 已提交
1946 1947 1948 1949 1950 1951 1952 1953 1954

	/* 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);
1955
}
1956
EXPORT_SYMBOL_GPL(__gfn_to_pfn_memslot);
1957

D
Dan Williams 已提交
1958
kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1959 1960
		      bool *writable)
{
P
Paolo Bonzini 已提交
1961 1962
	return __gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn, false, NULL,
				    write_fault, writable);
1963 1964 1965
}
EXPORT_SYMBOL_GPL(gfn_to_pfn_prot);

D
Dan Williams 已提交
1966
kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1967
{
X
Xiao Guangrong 已提交
1968
	return __gfn_to_pfn_memslot(slot, gfn, false, NULL, true, NULL);
1969
}
P
Paolo Bonzini 已提交
1970
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot);
1971

D
Dan Williams 已提交
1972
kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn)
1973
{
X
Xiao Guangrong 已提交
1974
	return __gfn_to_pfn_memslot(slot, gfn, true, NULL, true, NULL);
1975
}
1976
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot_atomic);
1977

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

D
Dan Williams 已提交
1990
kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn)
1991 1992 1993 1994 1995
{
	return gfn_to_pfn_memslot(kvm_vcpu_gfn_to_memslot(vcpu, gfn), gfn);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_gfn_to_pfn);

1996 1997
int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
			    struct page **pages, int nr_pages)
1998 1999
{
	unsigned long addr;
2000
	gfn_t entry = 0;
2001

2002
	addr = gfn_to_hva_many(slot, gfn, &entry);
2003 2004 2005 2006 2007 2008
	if (kvm_is_error_hva(addr))
		return -1;

	if (entry < nr_pages)
		return 0;

2009
	return get_user_pages_fast_only(addr, nr_pages, FOLL_WRITE, pages);
2010 2011 2012
}
EXPORT_SYMBOL_GPL(gfn_to_page_many_atomic);

D
Dan Williams 已提交
2013
static struct page *kvm_pfn_to_page(kvm_pfn_t pfn)
X
Xiao Guangrong 已提交
2014
{
2015
	if (is_error_noslot_pfn(pfn))
2016
		return KVM_ERR_PTR_BAD_PAGE;
X
Xiao Guangrong 已提交
2017

2018
	if (kvm_is_reserved_pfn(pfn)) {
2019
		WARN_ON(1);
2020
		return KVM_ERR_PTR_BAD_PAGE;
2021
	}
X
Xiao Guangrong 已提交
2022 2023 2024 2025

	return pfn_to_page(pfn);
}

2026 2027
struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
{
D
Dan Williams 已提交
2028
	kvm_pfn_t pfn;
2029 2030 2031

	pfn = gfn_to_pfn(kvm, gfn);

X
Xiao Guangrong 已提交
2032
	return kvm_pfn_to_page(pfn);
A
Avi Kivity 已提交
2033 2034 2035
}
EXPORT_SYMBOL_GPL(gfn_to_page);

2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060
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;
}

2061
static int __kvm_map_gfn(struct kvm_memslots *slots, gfn_t gfn,
2062 2063 2064
			 struct kvm_host_map *map,
			 struct gfn_to_pfn_cache *cache,
			 bool atomic)
2065 2066 2067 2068
{
	kvm_pfn_t pfn;
	void *hva = NULL;
	struct page *page = KVM_UNMAPPED_PAGE;
2069
	struct kvm_memory_slot *slot = __gfn_to_memslot(slots, gfn);
2070
	u64 gen = slots->generation;
2071 2072 2073 2074

	if (!map)
		return -EINVAL;

2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087
	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);
	}
2088 2089 2090 2091 2092
	if (is_error_noslot_pfn(pfn))
		return -EINVAL;

	if (pfn_valid(pfn)) {
		page = pfn_to_page(pfn);
2093 2094 2095 2096
		if (atomic)
			hva = kmap_atomic(page);
		else
			hva = kmap(page);
P
Paolo Bonzini 已提交
2097
#ifdef CONFIG_HAS_IOMEM
2098
	} else if (!atomic) {
2099
		hva = memremap(pfn_to_hpa(pfn), PAGE_SIZE, MEMREMAP_WB);
2100 2101
	} else {
		return -EINVAL;
P
Paolo Bonzini 已提交
2102
#endif
2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115
	}

	if (!hva)
		return -EFAULT;

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

	return 0;
}

2116 2117
int kvm_map_gfn(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map,
		struct gfn_to_pfn_cache *cache, bool atomic)
2118
{
2119 2120
	return __kvm_map_gfn(kvm_memslots(vcpu->kvm), gfn, map,
			cache, atomic);
2121 2122 2123
}
EXPORT_SYMBOL_GPL(kvm_map_gfn);

2124 2125
int kvm_vcpu_map(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map)
{
2126 2127
	return __kvm_map_gfn(kvm_vcpu_memslots(vcpu), gfn, map,
		NULL, false);
2128 2129 2130
}
EXPORT_SYMBOL_GPL(kvm_vcpu_map);

2131
static void __kvm_unmap_gfn(struct kvm_memory_slot *memslot,
2132 2133 2134
			struct kvm_host_map *map,
			struct gfn_to_pfn_cache *cache,
			bool dirty, bool atomic)
2135 2136 2137 2138 2139 2140 2141
{
	if (!map)
		return;

	if (!map->hva)
		return;

2142 2143 2144 2145 2146 2147
	if (map->page != KVM_UNMAPPED_PAGE) {
		if (atomic)
			kunmap_atomic(map->hva);
		else
			kunmap(map->page);
	}
2148
#ifdef CONFIG_HAS_IOMEM
2149
	else if (!atomic)
2150
		memunmap(map->hva);
2151 2152
	else
		WARN_ONCE(1, "Unexpected unmapping in atomic context");
2153
#endif
2154

2155
	if (dirty)
2156
		mark_page_dirty_in_slot(memslot, map->gfn);
2157 2158 2159 2160 2161

	if (cache)
		cache->dirty |= dirty;
	else
		kvm_release_pfn(map->pfn, dirty, NULL);
2162 2163 2164 2165

	map->hva = NULL;
	map->page = NULL;
}
2166

2167 2168
int kvm_unmap_gfn(struct kvm_vcpu *vcpu, struct kvm_host_map *map, 
		  struct gfn_to_pfn_cache *cache, bool dirty, bool atomic)
2169
{
2170 2171
	__kvm_unmap_gfn(gfn_to_memslot(vcpu->kvm, map->gfn), map,
			cache, dirty, atomic);
2172 2173 2174 2175 2176 2177
	return 0;
}
EXPORT_SYMBOL_GPL(kvm_unmap_gfn);

void kvm_vcpu_unmap(struct kvm_vcpu *vcpu, struct kvm_host_map *map, bool dirty)
{
2178 2179
	__kvm_unmap_gfn(kvm_vcpu_gfn_to_memslot(vcpu, map->gfn), map, NULL,
			dirty, false);
2180
}
2181 2182
EXPORT_SYMBOL_GPL(kvm_vcpu_unmap);

2183 2184
struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn)
{
D
Dan Williams 已提交
2185
	kvm_pfn_t pfn;
2186 2187 2188 2189 2190 2191 2192

	pfn = kvm_vcpu_gfn_to_pfn(vcpu, gfn);

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

2193 2194
void kvm_release_page_clean(struct page *page)
{
2195 2196
	WARN_ON(is_error_page(page));

2197
	kvm_release_pfn_clean(page_to_pfn(page));
2198 2199 2200
}
EXPORT_SYMBOL_GPL(kvm_release_page_clean);

D
Dan Williams 已提交
2201
void kvm_release_pfn_clean(kvm_pfn_t pfn)
2202
{
2203
	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn))
2204
		put_page(pfn_to_page(pfn));
2205 2206 2207
}
EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);

2208
void kvm_release_page_dirty(struct page *page)
2209
{
X
Xiao Guangrong 已提交
2210 2211
	WARN_ON(is_error_page(page));

2212 2213 2214 2215
	kvm_release_pfn_dirty(page_to_pfn(page));
}
EXPORT_SYMBOL_GPL(kvm_release_page_dirty);

2216
void kvm_release_pfn_dirty(kvm_pfn_t pfn)
2217 2218 2219 2220
{
	kvm_set_pfn_dirty(pfn);
	kvm_release_pfn_clean(pfn);
}
2221
EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
2222

D
Dan Williams 已提交
2223
void kvm_set_pfn_dirty(kvm_pfn_t pfn)
2224
{
2225 2226
	if (!kvm_is_reserved_pfn(pfn) && !kvm_is_zone_device_pfn(pfn))
		SetPageDirty(pfn_to_page(pfn));
2227
}
2228 2229
EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);

D
Dan Williams 已提交
2230
void kvm_set_pfn_accessed(kvm_pfn_t pfn)
2231
{
2232
	if (!kvm_is_reserved_pfn(pfn) && !kvm_is_zone_device_pfn(pfn))
2233
		mark_page_accessed(pfn_to_page(pfn));
2234 2235 2236
}
EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);

D
Dan Williams 已提交
2237
void kvm_get_pfn(kvm_pfn_t pfn)
2238
{
2239
	if (!kvm_is_reserved_pfn(pfn))
2240
		get_page(pfn_to_page(pfn));
2241 2242
}
EXPORT_SYMBOL_GPL(kvm_get_pfn);
2243

2244 2245 2246 2247 2248 2249 2250 2251
static int next_segment(unsigned long len, int offset)
{
	if (len > PAGE_SIZE - offset)
		return PAGE_SIZE - offset;
	else
		return len;
}

2252 2253
static int __kvm_read_guest_page(struct kvm_memory_slot *slot, gfn_t gfn,
				 void *data, int offset, int len)
2254
{
2255 2256
	int r;
	unsigned long addr;
2257

2258
	addr = gfn_to_hva_memslot_prot(slot, gfn, NULL);
2259 2260
	if (kvm_is_error_hva(addr))
		return -EFAULT;
2261
	r = __copy_from_user(data, (void __user *)addr + offset, len);
2262
	if (r)
2263 2264 2265
		return -EFAULT;
	return 0;
}
2266 2267 2268 2269 2270 2271 2272 2273

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);
}
2274 2275
EXPORT_SYMBOL_GPL(kvm_read_guest_page);

2276 2277 2278 2279 2280 2281 2282 2283 2284
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);

2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304
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);

2305
int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data, unsigned long len)
2306 2307
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
2308
	int seg;
2309
	int offset = offset_in_page(gpa);
2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323
	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);
2324

2325 2326 2327 2328 2329 2330 2331
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);
2332 2333
	if (kvm_is_error_hva(addr))
		return -EFAULT;
2334
	pagefault_disable();
2335
	r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
2336
	pagefault_enable();
2337 2338 2339 2340 2341
	if (r)
		return -EFAULT;
	return 0;
}

2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354
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)
2355
{
2356 2357
	int r;
	unsigned long addr;
2358

2359
	addr = gfn_to_hva_memslot(memslot, gfn);
2360 2361
	if (kvm_is_error_hva(addr))
		return -EFAULT;
2362
	r = __copy_to_user((void __user *)addr + offset, data, len);
2363
	if (r)
2364
		return -EFAULT;
2365
	mark_page_dirty_in_slot(memslot, gfn);
2366 2367
	return 0;
}
2368 2369 2370 2371 2372 2373 2374 2375

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);
}
2376 2377
EXPORT_SYMBOL_GPL(kvm_write_guest_page);

2378 2379 2380 2381 2382 2383 2384 2385 2386
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);

2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405
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;
}
2406
EXPORT_SYMBOL_GPL(kvm_write_guest);
2407

2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428
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);

2429 2430 2431
static int __kvm_gfn_to_hva_cache_init(struct kvm_memslots *slots,
				       struct gfn_to_hva_cache *ghc,
				       gpa_t gpa, unsigned long len)
2432 2433
{
	int offset = offset_in_page(gpa);
2434 2435 2436 2437
	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;
2438

2439
	/* Update ghc->generation before performing any error checks. */
2440
	ghc->generation = slots->generation;
2441 2442 2443 2444 2445

	if (start_gfn > end_gfn) {
		ghc->hva = KVM_HVA_ERR_BAD;
		return -EINVAL;
	}
2446 2447 2448 2449 2450

	/*
	 * If the requested region crosses two memslots, we still
	 * verify that the entire region is valid here.
	 */
2451
	for ( ; start_gfn <= end_gfn; start_gfn += nr_pages_avail) {
2452 2453 2454 2455
		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))
2456
			return -EFAULT;
2457 2458 2459
	}

	/* Use the slow path for cross page reads and writes. */
2460
	if (nr_pages_needed == 1)
2461
		ghc->hva += offset;
2462
	else
2463
		ghc->memslot = NULL;
2464

2465 2466 2467
	ghc->gpa = gpa;
	ghc->len = len;
	return 0;
2468
}
2469

2470
int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
2471 2472
			      gpa_t gpa, unsigned long len)
{
2473
	struct kvm_memslots *slots = kvm_memslots(kvm);
2474 2475
	return __kvm_gfn_to_hva_cache_init(slots, ghc, gpa, len);
}
2476
EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init);
2477

2478
int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
2479 2480
				  void *data, unsigned int offset,
				  unsigned long len)
2481
{
2482
	struct kvm_memslots *slots = kvm_memslots(kvm);
2483
	int r;
2484
	gpa_t gpa = ghc->gpa + offset;
2485

2486
	BUG_ON(len + offset > ghc->len);
2487

2488 2489 2490 2491
	if (slots->generation != ghc->generation) {
		if (__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len))
			return -EFAULT;
	}
2492

2493 2494 2495
	if (kvm_is_error_hva(ghc->hva))
		return -EFAULT;

2496 2497 2498
	if (unlikely(!ghc->memslot))
		return kvm_write_guest(kvm, gpa, data, len);

2499
	r = __copy_to_user((void __user *)ghc->hva + offset, data, len);
2500 2501
	if (r)
		return -EFAULT;
2502
	mark_page_dirty_in_slot(ghc->memslot, gpa >> PAGE_SHIFT);
2503 2504 2505

	return 0;
}
2506
EXPORT_SYMBOL_GPL(kvm_write_guest_offset_cached);
2507

2508 2509
int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
2510
{
2511
	return kvm_write_guest_offset_cached(kvm, ghc, data, 0, len);
2512
}
2513
EXPORT_SYMBOL_GPL(kvm_write_guest_cached);
2514

2515 2516 2517
int kvm_read_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
				 void *data, unsigned int offset,
				 unsigned long len)
2518
{
2519
	struct kvm_memslots *slots = kvm_memslots(kvm);
2520
	int r;
2521
	gpa_t gpa = ghc->gpa + offset;
2522

2523
	BUG_ON(len + offset > ghc->len);
2524

2525 2526 2527 2528
	if (slots->generation != ghc->generation) {
		if (__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len))
			return -EFAULT;
	}
2529

2530 2531 2532
	if (kvm_is_error_hva(ghc->hva))
		return -EFAULT;

2533
	if (unlikely(!ghc->memslot))
2534
		return kvm_read_guest(kvm, gpa, data, len);
2535

2536
	r = __copy_from_user(data, (void __user *)ghc->hva + offset, len);
2537 2538 2539 2540 2541
	if (r)
		return -EFAULT;

	return 0;
}
2542 2543 2544 2545 2546 2547 2548
EXPORT_SYMBOL_GPL(kvm_read_guest_offset_cached);

int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			  void *data, unsigned long len)
{
	return kvm_read_guest_offset_cached(kvm, ghc, data, 0, len);
}
2549
EXPORT_SYMBOL_GPL(kvm_read_guest_cached);
2550

2551 2552
int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
{
2553 2554 2555
	const void *zero_page = (const void *) __va(page_to_phys(ZERO_PAGE(0)));

	return kvm_write_guest_page(kvm, gfn, zero_page, offset, len);
2556 2557 2558 2559 2560 2561 2562 2563 2564 2565
}
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;

2566
	while ((seg = next_segment(len, offset)) != 0) {
2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577
		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);

2578
static void mark_page_dirty_in_slot(struct kvm_memory_slot *memslot,
2579
				    gfn_t gfn)
A
Avi Kivity 已提交
2580
{
R
Rusty Russell 已提交
2581 2582
	if (memslot && memslot->dirty_bitmap) {
		unsigned long rel_gfn = gfn - memslot->base_gfn;
A
Avi Kivity 已提交
2583

2584
		set_bit_le(rel_gfn, memslot->dirty_bitmap);
A
Avi Kivity 已提交
2585 2586 2587
	}
}

2588 2589 2590 2591 2592
void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
{
	struct kvm_memory_slot *memslot;

	memslot = gfn_to_memslot(kvm, gfn);
2593
	mark_page_dirty_in_slot(memslot, gfn);
2594
}
2595
EXPORT_SYMBOL_GPL(mark_page_dirty);
2596

2597 2598 2599 2600 2601 2602 2603 2604 2605
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);

2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628
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
2629 2630
static void grow_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2631
	unsigned int old, val, grow, grow_start;
W
Wanpeng Li 已提交
2632

2633
	old = val = vcpu->halt_poll_ns;
2634
	grow_start = READ_ONCE(halt_poll_ns_grow_start);
2635
	grow = READ_ONCE(halt_poll_ns_grow);
2636 2637 2638
	if (!grow)
		goto out;

2639 2640 2641
	val *= grow;
	if (val < grow_start)
		val = grow_start;
W
Wanpeng Li 已提交
2642

2643 2644 2645
	if (val > halt_poll_ns)
		val = halt_poll_ns;

W
Wanpeng Li 已提交
2646
	vcpu->halt_poll_ns = val;
2647
out:
2648
	trace_kvm_halt_poll_ns_grow(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2649 2650 2651 2652
}

static void shrink_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2653
	unsigned int old, val, shrink;
W
Wanpeng Li 已提交
2654

2655
	old = val = vcpu->halt_poll_ns;
2656 2657
	shrink = READ_ONCE(halt_poll_ns_shrink);
	if (shrink == 0)
W
Wanpeng Li 已提交
2658 2659
		val = 0;
	else
2660
		val /= shrink;
W
Wanpeng Li 已提交
2661 2662

	vcpu->halt_poll_ns = val;
2663
	trace_kvm_halt_poll_ns_shrink(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2664 2665
}

2666 2667
static int kvm_vcpu_check_block(struct kvm_vcpu *vcpu)
{
2668 2669 2670
	int ret = -EINTR;
	int idx = srcu_read_lock(&vcpu->kvm->srcu);

2671 2672
	if (kvm_arch_vcpu_runnable(vcpu)) {
		kvm_make_request(KVM_REQ_UNHALT, vcpu);
2673
		goto out;
2674 2675
	}
	if (kvm_cpu_has_pending_timer(vcpu))
2676
		goto out;
2677
	if (signal_pending(current))
2678
		goto out;
2679

2680 2681 2682 2683
	ret = 0;
out:
	srcu_read_unlock(&vcpu->kvm->srcu, idx);
	return ret;
2684 2685
}

2686 2687 2688 2689 2690 2691 2692 2693 2694
static inline void
update_halt_poll_stats(struct kvm_vcpu *vcpu, u64 poll_ns, bool waited)
{
	if (waited)
		vcpu->stat.halt_poll_fail_ns += poll_ns;
	else
		vcpu->stat.halt_poll_success_ns += poll_ns;
}

E
Eddie Dong 已提交
2695 2696 2697
/*
 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
 */
2698
void kvm_vcpu_block(struct kvm_vcpu *vcpu)
2699
{
2700
	ktime_t start, cur, poll_end;
2701
	bool waited = false;
W
Wanpeng Li 已提交
2702
	u64 block_ns;
2703

2704 2705
	kvm_arch_vcpu_blocking(vcpu);

2706
	start = cur = poll_end = ktime_get();
2707
	if (vcpu->halt_poll_ns && !kvm_arch_no_poll(vcpu)) {
W
Wanpeng Li 已提交
2708
		ktime_t stop = ktime_add_ns(ktime_get(), vcpu->halt_poll_ns);
2709

2710
		++vcpu->stat.halt_attempted_poll;
2711 2712 2713 2714 2715 2716 2717
		do {
			/*
			 * This sets KVM_REQ_UNHALT if an interrupt
			 * arrives.
			 */
			if (kvm_vcpu_check_block(vcpu) < 0) {
				++vcpu->stat.halt_successful_poll;
2718 2719
				if (!vcpu_valid_wakeup(vcpu))
					++vcpu->stat.halt_poll_invalid;
2720 2721
				goto out;
			}
2722
			poll_end = cur = ktime_get();
2723 2724
		} while (single_task_running() && ktime_before(cur, stop));
	}
2725

2726
	prepare_to_rcuwait(&vcpu->wait);
2727
	for (;;) {
2728
		set_current_state(TASK_INTERRUPTIBLE);
2729

2730
		if (kvm_vcpu_check_block(vcpu) < 0)
2731 2732
			break;

2733
		waited = true;
E
Eddie Dong 已提交
2734 2735
		schedule();
	}
2736
	finish_rcuwait(&vcpu->wait);
2737 2738
	cur = ktime_get();
out:
2739
	kvm_arch_vcpu_unblocking(vcpu);
W
Wanpeng Li 已提交
2740 2741
	block_ns = ktime_to_ns(cur) - ktime_to_ns(start);

2742 2743 2744
	update_halt_poll_stats(
		vcpu, ktime_to_ns(ktime_sub(poll_end, start)), waited);

2745 2746
	if (!kvm_arch_no_poll(vcpu)) {
		if (!vcpu_valid_wakeup(vcpu)) {
W
Wanpeng Li 已提交
2747
			shrink_halt_poll_ns(vcpu);
2748
		} else if (vcpu->kvm->max_halt_poll_ns) {
2749 2750 2751
			if (block_ns <= vcpu->halt_poll_ns)
				;
			/* we had a long block, shrink polling */
2752 2753
			else if (vcpu->halt_poll_ns &&
					block_ns > vcpu->kvm->max_halt_poll_ns)
2754 2755
				shrink_halt_poll_ns(vcpu);
			/* we had a short halt and our poll time is too small */
2756 2757
			else if (vcpu->halt_poll_ns < vcpu->kvm->max_halt_poll_ns &&
					block_ns < vcpu->kvm->max_halt_poll_ns)
2758 2759 2760 2761 2762
				grow_halt_poll_ns(vcpu);
		} else {
			vcpu->halt_poll_ns = 0;
		}
	}
W
Wanpeng Li 已提交
2763

2764 2765
	trace_kvm_vcpu_wakeup(block_ns, waited, vcpu_valid_wakeup(vcpu));
	kvm_arch_vcpu_block_finish(vcpu);
E
Eddie Dong 已提交
2766
}
2767
EXPORT_SYMBOL_GPL(kvm_vcpu_block);
E
Eddie Dong 已提交
2768

2769
bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu)
2770
{
2771
	struct rcuwait *waitp;
2772

2773 2774
	waitp = kvm_arch_vcpu_get_wait(vcpu);
	if (rcuwait_wake_up(waitp)) {
2775
		WRITE_ONCE(vcpu->ready, true);
2776
		++vcpu->stat.halt_wakeup;
2777
		return true;
2778 2779
	}

2780
	return false;
2781 2782 2783
}
EXPORT_SYMBOL_GPL(kvm_vcpu_wake_up);

2784
#ifndef CONFIG_S390
2785 2786 2787 2788 2789 2790 2791 2792
/*
 * 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;

2793 2794 2795
	if (kvm_vcpu_wake_up(vcpu))
		return;

2796 2797 2798 2799 2800 2801
	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();
}
2802
EXPORT_SYMBOL_GPL(kvm_vcpu_kick);
2803
#endif /* !CONFIG_S390 */
2804

2805
int kvm_vcpu_yield_to(struct kvm_vcpu *target)
2806 2807 2808
{
	struct pid *pid;
	struct task_struct *task = NULL;
2809
	int ret = 0;
2810 2811 2812 2813

	rcu_read_lock();
	pid = rcu_dereference(target->pid);
	if (pid)
2814
		task = get_pid_task(pid, PIDTYPE_PID);
2815 2816
	rcu_read_unlock();
	if (!task)
2817 2818
		return ret;
	ret = yield_to(task, 1);
2819
	put_task_struct(task);
2820 2821

	return ret;
2822 2823 2824
}
EXPORT_SYMBOL_GPL(kvm_vcpu_yield_to);

2825 2826 2827 2828 2829 2830
/*
 * Helper that checks whether a VCPU is eligible for directed yield.
 * Most eligible candidate to yield is decided by following heuristics:
 *
 *  (a) VCPU which has not done pl-exit or cpu relax intercepted recently
 *  (preempted lock holder), indicated by @in_spin_loop.
F
Fuad Tabba 已提交
2831
 *  Set at the beginning and cleared at the end of interception/PLE handler.
2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846
 *
 *  (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.
 */
2847
static bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
2848
{
2849
#ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
2850 2851 2852
	bool eligible;

	eligible = !vcpu->spin_loop.in_spin_loop ||
2853
		    vcpu->spin_loop.dy_eligible;
2854 2855 2856 2857 2858

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

	return eligible;
2859 2860
#else
	return true;
2861
#endif
2862
}
2863

2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886
/*
 * 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;
}

2887
void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
Z
Zhai, Edwin 已提交
2888
{
2889 2890 2891 2892
	struct kvm *kvm = me->kvm;
	struct kvm_vcpu *vcpu;
	int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
	int yielded = 0;
2893
	int try = 3;
2894 2895
	int pass;
	int i;
Z
Zhai, Edwin 已提交
2896

2897
	kvm_vcpu_set_in_spin_loop(me, true);
2898 2899 2900 2901 2902 2903 2904
	/*
	 * 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.
	 */
2905
	for (pass = 0; pass < 2 && !yielded && try; pass++) {
2906
		kvm_for_each_vcpu(i, vcpu, kvm) {
2907
			if (!pass && i <= last_boosted_vcpu) {
2908 2909 2910 2911
				i = last_boosted_vcpu;
				continue;
			} else if (pass && i > last_boosted_vcpu)
				break;
2912
			if (!READ_ONCE(vcpu->ready))
2913
				continue;
2914 2915
			if (vcpu == me)
				continue;
2916 2917
			if (rcuwait_active(&vcpu->wait) &&
			    !vcpu_dy_runnable(vcpu))
2918
				continue;
2919 2920
			if (READ_ONCE(vcpu->preempted) && yield_to_kernel_mode &&
				!kvm_arch_vcpu_in_kernel(vcpu))
2921
				continue;
2922 2923
			if (!kvm_vcpu_eligible_for_directed_yield(vcpu))
				continue;
2924 2925 2926

			yielded = kvm_vcpu_yield_to(vcpu);
			if (yielded > 0) {
2927 2928
				kvm->last_boosted_vcpu = i;
				break;
2929 2930 2931 2932
			} else if (yielded < 0) {
				try--;
				if (!try)
					break;
2933 2934 2935
			}
		}
	}
2936
	kvm_vcpu_set_in_spin_loop(me, false);
2937 2938 2939

	/* Ensure vcpu is not eligible during next spinloop */
	kvm_vcpu_set_dy_eligible(me, false);
Z
Zhai, Edwin 已提交
2940 2941 2942
}
EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin);

2943
static vm_fault_t kvm_vcpu_fault(struct vm_fault *vmf)
2944
{
2945
	struct kvm_vcpu *vcpu = vmf->vma->vm_file->private_data;
2946 2947
	struct page *page;

2948
	if (vmf->pgoff == 0)
2949
		page = virt_to_page(vcpu->run);
A
Avi Kivity 已提交
2950
#ifdef CONFIG_X86
2951
	else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
2952
		page = virt_to_page(vcpu->arch.pio_data);
2953
#endif
2954
#ifdef CONFIG_KVM_MMIO
2955 2956
	else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
		page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
A
Avi Kivity 已提交
2957
#endif
2958
	else
2959
		return kvm_arch_vcpu_fault(vcpu, vmf);
2960
	get_page(page);
2961 2962
	vmf->page = page;
	return 0;
2963 2964
}

2965
static const struct vm_operations_struct kvm_vcpu_vm_ops = {
2966
	.fault = kvm_vcpu_fault,
2967 2968 2969 2970 2971 2972 2973 2974
};

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
2975 2976 2977 2978
static int kvm_vcpu_release(struct inode *inode, struct file *filp)
{
	struct kvm_vcpu *vcpu = filp->private_data;

A
Al Viro 已提交
2979
	kvm_put_kvm(vcpu->kvm);
A
Avi Kivity 已提交
2980 2981 2982
	return 0;
}

2983
static struct file_operations kvm_vcpu_fops = {
A
Avi Kivity 已提交
2984 2985
	.release        = kvm_vcpu_release,
	.unlocked_ioctl = kvm_vcpu_ioctl,
2986
	.mmap           = kvm_vcpu_mmap,
2987
	.llseek		= noop_llseek,
2988
	KVM_COMPAT(kvm_vcpu_compat_ioctl),
A
Avi Kivity 已提交
2989 2990 2991 2992 2993 2994 2995
};

/*
 * Allocates an inode for the vcpu.
 */
static int create_vcpu_fd(struct kvm_vcpu *vcpu)
{
2996 2997 2998 2999
	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
3000 3001
}

3002
static void kvm_create_vcpu_debugfs(struct kvm_vcpu *vcpu)
3003
{
3004
#ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
3005
	struct dentry *debugfs_dentry;
3006 3007 3008
	char dir_name[ITOA_MAX_LEN * 2];

	if (!debugfs_initialized())
3009
		return;
3010 3011

	snprintf(dir_name, sizeof(dir_name), "vcpu%d", vcpu->vcpu_id);
3012 3013
	debugfs_dentry = debugfs_create_dir(dir_name,
					    vcpu->kvm->debugfs_dentry);
3014

3015
	kvm_arch_create_vcpu_debugfs(vcpu, debugfs_dentry);
3016
#endif
3017 3018
}

3019 3020 3021
/*
 * Creates some virtual cpus.  Good luck creating more than one.
 */
3022
static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
3023 3024
{
	int r;
3025
	struct kvm_vcpu *vcpu;
3026
	struct page *page;
3027

G
Greg Kurz 已提交
3028
	if (id >= KVM_MAX_VCPU_ID)
3029 3030
		return -EINVAL;

3031 3032 3033 3034 3035 3036 3037 3038 3039
	mutex_lock(&kvm->lock);
	if (kvm->created_vcpus == KVM_MAX_VCPUS) {
		mutex_unlock(&kvm->lock);
		return -EINVAL;
	}

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

3040 3041 3042 3043
	r = kvm_arch_vcpu_precreate(kvm, id);
	if (r)
		goto vcpu_decrement;

3044 3045 3046
	vcpu = kmem_cache_zalloc(kvm_vcpu_cache, GFP_KERNEL);
	if (!vcpu) {
		r = -ENOMEM;
3047 3048
		goto vcpu_decrement;
	}
3049

3050
	BUILD_BUG_ON(sizeof(struct kvm_run) > PAGE_SIZE);
3051 3052 3053
	page = alloc_page(GFP_KERNEL | __GFP_ZERO);
	if (!page) {
		r = -ENOMEM;
3054
		goto vcpu_free;
3055 3056 3057 3058
	}
	vcpu->run = page_address(page);

	kvm_vcpu_init(vcpu, kvm, id);
3059 3060 3061

	r = kvm_arch_vcpu_create(vcpu);
	if (r)
3062
		goto vcpu_free_run_page;
3063

S
Shaohua Li 已提交
3064
	mutex_lock(&kvm->lock);
3065 3066 3067 3068
	if (kvm_get_vcpu_by_id(kvm, id)) {
		r = -EEXIST;
		goto unlock_vcpu_destroy;
	}
3069

3070 3071
	vcpu->vcpu_idx = atomic_read(&kvm->online_vcpus);
	BUG_ON(kvm->vcpus[vcpu->vcpu_idx]);
3072

R
Rusty Russell 已提交
3073
	/* Now it's all set up, let userspace reach it */
A
Al Viro 已提交
3074
	kvm_get_kvm(kvm);
A
Avi Kivity 已提交
3075
	r = create_vcpu_fd(vcpu);
3076
	if (r < 0) {
3077
		kvm_put_kvm_no_destroy(kvm);
3078
		goto unlock_vcpu_destroy;
3079 3080
	}

3081
	kvm->vcpus[vcpu->vcpu_idx] = vcpu;
3082 3083 3084 3085 3086

	/*
	 * Pairs with smp_rmb() in kvm_get_vcpu.  Write kvm->vcpus
	 * before kvm->online_vcpu's incremented value.
	 */
3087 3088 3089 3090
	smp_wmb();
	atomic_inc(&kvm->online_vcpus);

	mutex_unlock(&kvm->lock);
3091
	kvm_arch_vcpu_postcreate(vcpu);
3092
	kvm_create_vcpu_debugfs(vcpu);
R
Rusty Russell 已提交
3093
	return r;
3094

3095
unlock_vcpu_destroy:
3096
	mutex_unlock(&kvm->lock);
3097
	kvm_arch_vcpu_destroy(vcpu);
3098 3099
vcpu_free_run_page:
	free_page((unsigned long)vcpu->run);
3100 3101
vcpu_free:
	kmem_cache_free(kvm_vcpu_cache, vcpu);
3102 3103 3104 3105
vcpu_decrement:
	mutex_lock(&kvm->lock);
	kvm->created_vcpus--;
	mutex_unlock(&kvm->lock);
3106 3107 3108
	return r;
}

A
Avi Kivity 已提交
3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119
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
3120 3121
static long kvm_vcpu_ioctl(struct file *filp,
			   unsigned int ioctl, unsigned long arg)
A
Avi Kivity 已提交
3122
{
A
Avi Kivity 已提交
3123
	struct kvm_vcpu *vcpu = filp->private_data;
A
Al Viro 已提交
3124
	void __user *argp = (void __user *)arg;
3125
	int r;
3126 3127
	struct kvm_fpu *fpu = NULL;
	struct kvm_sregs *kvm_sregs = NULL;
A
Avi Kivity 已提交
3128

3129 3130
	if (vcpu->kvm->mm != current->mm)
		return -EIO;
3131

3132 3133 3134
	if (unlikely(_IOC_TYPE(ioctl) != KVMIO))
		return -EINVAL;

3135
	/*
3136 3137
	 * Some architectures have vcpu ioctls that are asynchronous to vcpu
	 * execution; mutex_lock() would break them.
3138
	 */
3139 3140
	r = kvm_arch_vcpu_async_ioctl(filp, ioctl, arg);
	if (r != -ENOIOCTLCMD)
3141
		return r;
3142

3143 3144
	if (mutex_lock_killable(&vcpu->mutex))
		return -EINTR;
A
Avi Kivity 已提交
3145
	switch (ioctl) {
3146 3147
	case KVM_RUN: {
		struct pid *oldpid;
3148 3149 3150
		r = -EINVAL;
		if (arg)
			goto out;
3151
		oldpid = rcu_access_pointer(vcpu->pid);
3152
		if (unlikely(oldpid != task_pid(current))) {
3153
			/* The thread running this VCPU changed. */
3154
			struct pid *newpid;
3155

3156 3157 3158 3159 3160
			r = kvm_arch_vcpu_run_pid_change(vcpu);
			if (r)
				break;

			newpid = get_task_pid(current, PIDTYPE_PID);
3161 3162 3163 3164 3165
			rcu_assign_pointer(vcpu->pid, newpid);
			if (oldpid)
				synchronize_rcu();
			put_pid(oldpid);
		}
3166
		r = kvm_arch_vcpu_ioctl_run(vcpu);
3167
		trace_kvm_userspace_exit(vcpu->run->exit_reason, r);
A
Avi Kivity 已提交
3168
		break;
3169
	}
A
Avi Kivity 已提交
3170
	case KVM_GET_REGS: {
3171
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
3172

3173
		r = -ENOMEM;
3174
		kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL_ACCOUNT);
3175
		if (!kvm_regs)
A
Avi Kivity 已提交
3176
			goto out;
3177 3178 3179
		r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
		if (r)
			goto out_free1;
A
Avi Kivity 已提交
3180
		r = -EFAULT;
3181 3182
		if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
			goto out_free1;
A
Avi Kivity 已提交
3183
		r = 0;
3184 3185
out_free1:
		kfree(kvm_regs);
A
Avi Kivity 已提交
3186 3187 3188
		break;
	}
	case KVM_SET_REGS: {
3189
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
3190

3191 3192 3193
		kvm_regs = memdup_user(argp, sizeof(*kvm_regs));
		if (IS_ERR(kvm_regs)) {
			r = PTR_ERR(kvm_regs);
A
Avi Kivity 已提交
3194
			goto out;
3195
		}
3196 3197
		r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
		kfree(kvm_regs);
A
Avi Kivity 已提交
3198 3199 3200
		break;
	}
	case KVM_GET_SREGS: {
3201 3202
		kvm_sregs = kzalloc(sizeof(struct kvm_sregs),
				    GFP_KERNEL_ACCOUNT);
3203 3204 3205 3206
		r = -ENOMEM;
		if (!kvm_sregs)
			goto out;
		r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
3207 3208 3209
		if (r)
			goto out;
		r = -EFAULT;
3210
		if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
A
Avi Kivity 已提交
3211 3212 3213 3214 3215
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_SREGS: {
3216 3217 3218
		kvm_sregs = memdup_user(argp, sizeof(*kvm_sregs));
		if (IS_ERR(kvm_sregs)) {
			r = PTR_ERR(kvm_sregs);
G
Guo Chao 已提交
3219
			kvm_sregs = NULL;
A
Avi Kivity 已提交
3220
			goto out;
3221
		}
3222
		r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
3223 3224
		break;
	}
3225 3226 3227 3228 3229 3230 3231
	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;
3232
		if (copy_to_user(argp, &mp_state, sizeof(mp_state)))
3233 3234 3235 3236 3237 3238 3239 3240
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_MP_STATE: {
		struct kvm_mp_state mp_state;

		r = -EFAULT;
3241
		if (copy_from_user(&mp_state, argp, sizeof(mp_state)))
3242 3243 3244 3245
			goto out;
		r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
		break;
	}
A
Avi Kivity 已提交
3246 3247 3248 3249
	case KVM_TRANSLATE: {
		struct kvm_translation tr;

		r = -EFAULT;
3250
		if (copy_from_user(&tr, argp, sizeof(tr)))
A
Avi Kivity 已提交
3251
			goto out;
3252
		r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
A
Avi Kivity 已提交
3253 3254 3255
		if (r)
			goto out;
		r = -EFAULT;
3256
		if (copy_to_user(argp, &tr, sizeof(tr)))
A
Avi Kivity 已提交
3257 3258 3259 3260
			goto out;
		r = 0;
		break;
	}
J
Jan Kiszka 已提交
3261 3262
	case KVM_SET_GUEST_DEBUG: {
		struct kvm_guest_debug dbg;
A
Avi Kivity 已提交
3263 3264

		r = -EFAULT;
3265
		if (copy_from_user(&dbg, argp, sizeof(dbg)))
A
Avi Kivity 已提交
3266
			goto out;
J
Jan Kiszka 已提交
3267
		r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
A
Avi Kivity 已提交
3268 3269
		break;
	}
A
Avi Kivity 已提交
3270 3271 3272 3273 3274 3275 3276 3277 3278
	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,
3279
					   sizeof(kvm_sigmask)))
A
Avi Kivity 已提交
3280 3281
				goto out;
			r = -EINVAL;
3282
			if (kvm_sigmask.len != sizeof(sigset))
A
Avi Kivity 已提交
3283 3284 3285
				goto out;
			r = -EFAULT;
			if (copy_from_user(&sigset, sigmask_arg->sigset,
3286
					   sizeof(sigset)))
A
Avi Kivity 已提交
3287 3288 3289
				goto out;
			p = &sigset;
		}
3290
		r = kvm_vcpu_ioctl_set_sigmask(vcpu, p);
A
Avi Kivity 已提交
3291 3292
		break;
	}
A
Avi Kivity 已提交
3293
	case KVM_GET_FPU: {
3294
		fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL_ACCOUNT);
3295 3296 3297 3298
		r = -ENOMEM;
		if (!fpu)
			goto out;
		r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu);
A
Avi Kivity 已提交
3299 3300 3301
		if (r)
			goto out;
		r = -EFAULT;
3302
		if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
A
Avi Kivity 已提交
3303 3304 3305 3306 3307
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_FPU: {
3308 3309 3310
		fpu = memdup_user(argp, sizeof(*fpu));
		if (IS_ERR(fpu)) {
			r = PTR_ERR(fpu);
G
Guo Chao 已提交
3311
			fpu = NULL;
A
Avi Kivity 已提交
3312
			goto out;
3313
		}
3314
		r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
A
Avi Kivity 已提交
3315 3316
		break;
	}
A
Avi Kivity 已提交
3317
	default:
3318
		r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
3319 3320
	}
out:
3321
	mutex_unlock(&vcpu->mutex);
3322 3323
	kfree(fpu);
	kfree(kvm_sregs);
A
Avi Kivity 已提交
3324 3325 3326
	return r;
}

3327
#ifdef CONFIG_KVM_COMPAT
3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346
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,
3347
					   sizeof(kvm_sigmask)))
3348 3349
				goto out;
			r = -EINVAL;
A
Al Viro 已提交
3350
			if (kvm_sigmask.len != sizeof(compat_sigset_t))
3351 3352
				goto out;
			r = -EFAULT;
3353 3354
			if (get_compat_sigset(&sigset,
					      (compat_sigset_t __user *)sigmask_arg->sigset))
3355
				goto out;
3356 3357 3358
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
		} else
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, NULL);
3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369
		break;
	}
	default:
		r = kvm_vcpu_ioctl(filp, ioctl, arg);
	}

out:
	return r;
}
#endif

3370 3371 3372 3373 3374 3375 3376 3377 3378 3379
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
3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400
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;

3401 3402 3403
	if (dev->kvm->mm != current->mm)
		return -EIO;

S
Scott Wood 已提交
3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423
	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;

3424 3425 3426 3427 3428 3429 3430
	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
3431 3432 3433 3434 3435 3436 3437
	kvm_put_kvm(kvm);
	return 0;
}

static const struct file_operations kvm_device_fops = {
	.unlocked_ioctl = kvm_device_ioctl,
	.release = kvm_device_release,
3438
	KVM_COMPAT(kvm_device_ioctl),
3439
	.mmap = kvm_device_mmap,
S
Scott Wood 已提交
3440 3441 3442 3443 3444 3445 3446 3447 3448 3449
};

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

	return filp->private_data;
}

3450
static const struct kvm_device_ops *kvm_device_ops_table[KVM_DEV_TYPE_MAX] = {
3451
#ifdef CONFIG_KVM_MPIC
3452 3453
	[KVM_DEV_TYPE_FSL_MPIC_20]	= &kvm_mpic_ops,
	[KVM_DEV_TYPE_FSL_MPIC_42]	= &kvm_mpic_ops,
3454
#endif
3455 3456
};

3457
int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type)
3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468
{
	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;
}

3469 3470 3471 3472 3473 3474
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
3475 3476 3477
static int kvm_ioctl_create_device(struct kvm *kvm,
				   struct kvm_create_device *cd)
{
3478
	const struct kvm_device_ops *ops = NULL;
S
Scott Wood 已提交
3479 3480
	struct kvm_device *dev;
	bool test = cd->flags & KVM_CREATE_DEVICE_TEST;
P
Paolo Bonzini 已提交
3481
	int type;
S
Scott Wood 已提交
3482 3483
	int ret;

3484 3485 3486
	if (cd->type >= ARRAY_SIZE(kvm_device_ops_table))
		return -ENODEV;

P
Paolo Bonzini 已提交
3487 3488
	type = array_index_nospec(cd->type, ARRAY_SIZE(kvm_device_ops_table));
	ops = kvm_device_ops_table[type];
3489
	if (ops == NULL)
S
Scott Wood 已提交
3490 3491 3492 3493 3494
		return -ENODEV;

	if (test)
		return 0;

3495
	dev = kzalloc(sizeof(*dev), GFP_KERNEL_ACCOUNT);
S
Scott Wood 已提交
3496 3497 3498 3499 3500 3501
	if (!dev)
		return -ENOMEM;

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

3502
	mutex_lock(&kvm->lock);
P
Paolo Bonzini 已提交
3503
	ret = ops->create(dev, type);
S
Scott Wood 已提交
3504
	if (ret < 0) {
3505
		mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
3506 3507 3508
		kfree(dev);
		return ret;
	}
3509 3510
	list_add(&dev->vm_node, &kvm->devices);
	mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
3511

3512 3513 3514
	if (ops->init)
		ops->init(dev);

3515
	kvm_get_kvm(kvm);
3516
	ret = anon_inode_getfd(ops->name, &kvm_device_fops, dev, O_RDWR | O_CLOEXEC);
S
Scott Wood 已提交
3517
	if (ret < 0) {
3518
		kvm_put_kvm_no_destroy(kvm);
3519 3520 3521
		mutex_lock(&kvm->lock);
		list_del(&dev->vm_node);
		mutex_unlock(&kvm->lock);
3522
		ops->destroy(dev);
S
Scott Wood 已提交
3523 3524 3525 3526 3527 3528 3529
		return ret;
	}

	cd->fd = ret;
	return 0;
}

3530 3531 3532 3533 3534 3535 3536 3537 3538 3539
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
3540
#ifdef CONFIG_HAVE_KVM_IRQFD
3541
	case KVM_CAP_IRQFD:
3542 3543
	case KVM_CAP_IRQFD_RESAMPLE:
#endif
3544
	case KVM_CAP_IOEVENTFD_ANY_LENGTH:
3545
	case KVM_CAP_CHECK_EXTENSION_VM:
3546
	case KVM_CAP_ENABLE_CAP_VM:
3547
	case KVM_CAP_HALT_POLL:
3548
		return 1;
3549
#ifdef CONFIG_KVM_MMIO
3550 3551
	case KVM_CAP_COALESCED_MMIO:
		return KVM_COALESCED_MMIO_PAGE_OFFSET;
P
Peng Hao 已提交
3552 3553
	case KVM_CAP_COALESCED_PIO:
		return 1;
3554
#endif
3555 3556 3557 3558
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
	case KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2:
		return KVM_DIRTY_LOG_MANUAL_CAPS;
#endif
3559 3560 3561
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_CAP_IRQ_ROUTING:
		return KVM_MAX_IRQ_ROUTES;
3562 3563 3564 3565
#endif
#if KVM_ADDRESS_SPACE_NUM > 1
	case KVM_CAP_MULTI_ADDRESS_SPACE:
		return KVM_ADDRESS_SPACE_NUM;
3566
#endif
3567 3568
	case KVM_CAP_NR_MEMSLOTS:
		return KVM_USER_MEM_SLOTS;
3569 3570 3571 3572 3573 3574
	default:
		break;
	}
	return kvm_vm_ioctl_check_extension(kvm, arg);
}

3575 3576 3577 3578 3579 3580 3581 3582 3583 3584
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) {
3585
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
3586 3587 3588 3589 3590 3591 3592
	case KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2: {
		u64 allowed_options = KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE;

		if (cap->args[0] & KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE)
			allowed_options = KVM_DIRTY_LOG_MANUAL_CAPS;

		if (cap->flags || (cap->args[0] & ~allowed_options))
3593 3594 3595
			return -EINVAL;
		kvm->manual_dirty_log_protect = cap->args[0];
		return 0;
3596
	}
3597
#endif
3598 3599 3600 3601 3602 3603 3604
	case KVM_CAP_HALT_POLL: {
		if (cap->flags || cap->args[0] != (unsigned int)cap->args[0])
			return -EINVAL;

		kvm->max_halt_poll_ns = cap->args[0];
		return 0;
	}
3605 3606 3607 3608 3609
	default:
		return kvm_vm_ioctl_enable_cap(kvm, cap);
	}
}

A
Avi Kivity 已提交
3610 3611 3612 3613 3614
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;
3615
	int r;
A
Avi Kivity 已提交
3616

3617 3618
	if (kvm->mm != current->mm)
		return -EIO;
A
Avi Kivity 已提交
3619 3620 3621 3622
	switch (ioctl) {
	case KVM_CREATE_VCPU:
		r = kvm_vm_ioctl_create_vcpu(kvm, arg);
		break;
3623 3624 3625 3626 3627 3628 3629 3630 3631
	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;
	}
3632 3633 3634 3635 3636
	case KVM_SET_USER_MEMORY_REGION: {
		struct kvm_userspace_memory_region kvm_userspace_mem;

		r = -EFAULT;
		if (copy_from_user(&kvm_userspace_mem, argp,
3637
						sizeof(kvm_userspace_mem)))
3638 3639
			goto out;

3640
		r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem);
A
Avi Kivity 已提交
3641 3642 3643 3644 3645 3646
		break;
	}
	case KVM_GET_DIRTY_LOG: {
		struct kvm_dirty_log log;

		r = -EFAULT;
3647
		if (copy_from_user(&log, argp, sizeof(log)))
A
Avi Kivity 已提交
3648
			goto out;
3649
		r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
A
Avi Kivity 已提交
3650 3651
		break;
	}
3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662
#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
3663
#ifdef CONFIG_KVM_MMIO
3664 3665
	case KVM_REGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3666

3667
		r = -EFAULT;
3668
		if (copy_from_user(&zone, argp, sizeof(zone)))
3669 3670 3671 3672 3673 3674
			goto out;
		r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
		break;
	}
	case KVM_UNREGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3675

3676
		r = -EFAULT;
3677
		if (copy_from_user(&zone, argp, sizeof(zone)))
3678 3679 3680 3681 3682
			goto out;
		r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
		break;
	}
#endif
G
Gregory Haskins 已提交
3683 3684 3685 3686
	case KVM_IRQFD: {
		struct kvm_irqfd data;

		r = -EFAULT;
3687
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3688
			goto out;
3689
		r = kvm_irqfd(kvm, &data);
G
Gregory Haskins 已提交
3690 3691
		break;
	}
G
Gregory Haskins 已提交
3692 3693 3694 3695
	case KVM_IOEVENTFD: {
		struct kvm_ioeventfd data;

		r = -EFAULT;
3696
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3697 3698 3699 3700
			goto out;
		r = kvm_ioeventfd(kvm, &data);
		break;
	}
3701 3702 3703 3704 3705
#ifdef CONFIG_HAVE_KVM_MSI
	case KVM_SIGNAL_MSI: {
		struct kvm_msi msi;

		r = -EFAULT;
3706
		if (copy_from_user(&msi, argp, sizeof(msi)))
3707 3708 3709 3710
			goto out;
		r = kvm_send_userspace_msi(kvm, &msi);
		break;
	}
3711 3712 3713 3714 3715 3716 3717
#endif
#ifdef __KVM_HAVE_IRQ_LINE
	case KVM_IRQ_LINE_STATUS:
	case KVM_IRQ_LINE: {
		struct kvm_irq_level irq_event;

		r = -EFAULT;
3718
		if (copy_from_user(&irq_event, argp, sizeof(irq_event)))
3719 3720
			goto out;

3721 3722
		r = kvm_vm_ioctl_irq_line(kvm, &irq_event,
					ioctl == KVM_IRQ_LINE_STATUS);
3723 3724 3725 3726 3727
		if (r)
			goto out;

		r = -EFAULT;
		if (ioctl == KVM_IRQ_LINE_STATUS) {
3728
			if (copy_to_user(argp, &irq_event, sizeof(irq_event)))
3729 3730 3731 3732 3733 3734
				goto out;
		}

		r = 0;
		break;
	}
3735
#endif
3736 3737 3738 3739
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_SET_GSI_ROUTING: {
		struct kvm_irq_routing routing;
		struct kvm_irq_routing __user *urouting;
3740
		struct kvm_irq_routing_entry *entries = NULL;
3741 3742 3743 3744 3745

		r = -EFAULT;
		if (copy_from_user(&routing, argp, sizeof(routing)))
			goto out;
		r = -EINVAL;
3746 3747
		if (!kvm_arch_can_set_irq_routing(kvm))
			goto out;
3748
		if (routing.nr > KVM_MAX_IRQ_ROUTES)
3749 3750 3751
			goto out;
		if (routing.flags)
			goto out;
3752 3753
		if (routing.nr) {
			urouting = argp;
D
Denis Efremov 已提交
3754 3755 3756 3757 3758 3759 3760
			entries = vmemdup_user(urouting->entries,
					       array_size(sizeof(*entries),
							  routing.nr));
			if (IS_ERR(entries)) {
				r = PTR_ERR(entries);
				goto out;
			}
3761
		}
3762 3763
		r = kvm_set_irq_routing(kvm, entries, routing.nr,
					routing.flags);
D
Denis Efremov 已提交
3764
		kvfree(entries);
3765 3766 3767
		break;
	}
#endif /* CONFIG_HAVE_KVM_IRQ_ROUTING */
S
Scott Wood 已提交
3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785
	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;
	}
3786 3787 3788
	case KVM_CHECK_EXTENSION:
		r = kvm_vm_ioctl_check_extension_generic(kvm, arg);
		break;
3789
	default:
3790
		r = kvm_arch_vm_ioctl(filp, ioctl, arg);
3791 3792 3793 3794 3795
	}
out:
	return r;
}

3796
#ifdef CONFIG_KVM_COMPAT
3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820
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)))
3821
			return -EFAULT;
3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836
		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

3837
static struct file_operations kvm_vm_fops = {
3838 3839
	.release        = kvm_vm_release,
	.unlocked_ioctl = kvm_vm_ioctl,
3840
	.llseek		= noop_llseek,
3841
	KVM_COMPAT(kvm_vm_compat_ioctl),
3842 3843
};

3844
static int kvm_dev_ioctl_create_vm(unsigned long type)
3845
{
3846
	int r;
3847
	struct kvm *kvm;
3848
	struct file *file;
3849

3850
	kvm = kvm_create_vm(type);
3851 3852
	if (IS_ERR(kvm))
		return PTR_ERR(kvm);
3853
#ifdef CONFIG_KVM_MMIO
3854
	r = kvm_coalesced_mmio_init(kvm);
3855 3856
	if (r < 0)
		goto put_kvm;
3857
#endif
3858
	r = get_unused_fd_flags(O_CLOEXEC);
3859 3860 3861
	if (r < 0)
		goto put_kvm;

3862 3863 3864
	file = anon_inode_getfile("kvm-vm", &kvm_vm_fops, kvm, O_RDWR);
	if (IS_ERR(file)) {
		put_unused_fd(r);
3865 3866
		r = PTR_ERR(file);
		goto put_kvm;
3867
	}
3868

3869 3870 3871 3872 3873 3874
	/*
	 * 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).
	 */
3875
	if (kvm_create_vm_debugfs(kvm, r) < 0) {
3876 3877
		put_unused_fd(r);
		fput(file);
3878 3879
		return -ENOMEM;
	}
3880
	kvm_uevent_notify_change(KVM_EVENT_CREATE_VM, kvm);
3881

3882
	fd_install(r, file);
3883
	return r;
3884 3885 3886 3887

put_kvm:
	kvm_put_kvm(kvm);
	return r;
3888 3889 3890 3891 3892
}

static long kvm_dev_ioctl(struct file *filp,
			  unsigned int ioctl, unsigned long arg)
{
3893
	long r = -EINVAL;
3894 3895 3896

	switch (ioctl) {
	case KVM_GET_API_VERSION:
3897 3898
		if (arg)
			goto out;
3899 3900 3901
		r = KVM_API_VERSION;
		break;
	case KVM_CREATE_VM:
3902
		r = kvm_dev_ioctl_create_vm(arg);
3903
		break;
3904
	case KVM_CHECK_EXTENSION:
3905
		r = kvm_vm_ioctl_check_extension_generic(NULL, arg);
3906
		break;
3907 3908 3909
	case KVM_GET_VCPU_MMAP_SIZE:
		if (arg)
			goto out;
3910 3911 3912
		r = PAGE_SIZE;     /* struct kvm_run */
#ifdef CONFIG_X86
		r += PAGE_SIZE;    /* pio data page */
3913
#endif
3914
#ifdef CONFIG_KVM_MMIO
3915
		r += PAGE_SIZE;    /* coalesced mmio ring page */
3916
#endif
3917
		break;
3918 3919 3920
	case KVM_TRACE_ENABLE:
	case KVM_TRACE_PAUSE:
	case KVM_TRACE_DISABLE:
3921
		r = -EOPNOTSUPP;
3922
		break;
A
Avi Kivity 已提交
3923
	default:
3924
		return kvm_arch_dev_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
3925 3926 3927 3928 3929 3930 3931
	}
out:
	return r;
}

static struct file_operations kvm_chardev_ops = {
	.unlocked_ioctl = kvm_dev_ioctl,
3932
	.llseek		= noop_llseek,
3933
	KVM_COMPAT(kvm_dev_ioctl),
A
Avi Kivity 已提交
3934 3935 3936
};

static struct miscdevice kvm_dev = {
A
Avi Kivity 已提交
3937
	KVM_MINOR,
A
Avi Kivity 已提交
3938 3939 3940 3941
	"kvm",
	&kvm_chardev_ops,
};

3942
static void hardware_enable_nolock(void *junk)
3943 3944
{
	int cpu = raw_smp_processor_id();
3945
	int r;
3946

3947
	if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
3948
		return;
3949

3950
	cpumask_set_cpu(cpu, cpus_hardware_enabled);
3951

3952
	r = kvm_arch_hardware_enable();
3953 3954 3955 3956

	if (r) {
		cpumask_clear_cpu(cpu, cpus_hardware_enabled);
		atomic_inc(&hardware_enable_failed);
X
Xiubo Li 已提交
3957
		pr_info("kvm: enabling virtualization on CPU%d failed\n", cpu);
3958
	}
3959 3960
}

3961
static int kvm_starting_cpu(unsigned int cpu)
3962
{
3963
	raw_spin_lock(&kvm_count_lock);
3964 3965
	if (kvm_usage_count)
		hardware_enable_nolock(NULL);
3966
	raw_spin_unlock(&kvm_count_lock);
3967
	return 0;
3968 3969 3970
}

static void hardware_disable_nolock(void *junk)
3971 3972 3973
{
	int cpu = raw_smp_processor_id();

3974
	if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
3975
		return;
3976
	cpumask_clear_cpu(cpu, cpus_hardware_enabled);
3977
	kvm_arch_hardware_disable();
3978 3979
}

3980
static int kvm_dying_cpu(unsigned int cpu)
3981
{
3982
	raw_spin_lock(&kvm_count_lock);
3983 3984
	if (kvm_usage_count)
		hardware_disable_nolock(NULL);
3985
	raw_spin_unlock(&kvm_count_lock);
3986
	return 0;
3987 3988
}

3989 3990 3991 3992 3993 3994
static void hardware_disable_all_nolock(void)
{
	BUG_ON(!kvm_usage_count);

	kvm_usage_count--;
	if (!kvm_usage_count)
3995
		on_each_cpu(hardware_disable_nolock, NULL, 1);
3996 3997 3998 3999
}

static void hardware_disable_all(void)
{
4000
	raw_spin_lock(&kvm_count_lock);
4001
	hardware_disable_all_nolock();
4002
	raw_spin_unlock(&kvm_count_lock);
4003 4004 4005 4006 4007 4008
}

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

4009
	raw_spin_lock(&kvm_count_lock);
4010 4011 4012 4013

	kvm_usage_count++;
	if (kvm_usage_count == 1) {
		atomic_set(&hardware_enable_failed, 0);
4014
		on_each_cpu(hardware_enable_nolock, NULL, 1);
4015 4016 4017 4018 4019 4020 4021

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

4022
	raw_spin_unlock(&kvm_count_lock);
4023 4024 4025 4026

	return r;
}

4027
static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
M
Mike Day 已提交
4028
		      void *v)
4029
{
4030 4031 4032 4033 4034 4035
	/*
	 * 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
4036
	pr_info("kvm: exiting hardware virtualization\n");
4037
	kvm_rebooting = true;
4038
	on_each_cpu(hardware_disable_nolock, NULL, 1);
4039 4040 4041 4042 4043 4044 4045 4046
	return NOTIFY_OK;
}

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

M
Marcelo Tosatti 已提交
4047
static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
4048 4049 4050 4051
{
	int i;

	for (i = 0; i < bus->dev_count; i++) {
4052
		struct kvm_io_device *pos = bus->range[i].dev;
4053 4054 4055

		kvm_iodevice_destructor(pos);
	}
M
Marcelo Tosatti 已提交
4056
	kfree(bus);
4057 4058
}

4059
static inline int kvm_io_bus_cmp(const struct kvm_io_range *r1,
X
Xiubo Li 已提交
4060
				 const struct kvm_io_range *r2)
4061
{
J
Jason Wang 已提交
4062 4063 4064 4065
	gpa_t addr1 = r1->addr;
	gpa_t addr2 = r2->addr;

	if (addr1 < addr2)
4066
		return -1;
J
Jason Wang 已提交
4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078

	/* 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)
4079
		return 1;
J
Jason Wang 已提交
4080

4081 4082 4083
	return 0;
}

4084 4085
static int kvm_io_bus_sort_cmp(const void *p1, const void *p2)
{
4086
	return kvm_io_bus_cmp(p1, p2);
4087 4088
}

G
Geoff Levand 已提交
4089
static int kvm_io_bus_get_first_dev(struct kvm_io_bus *bus,
4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106
			     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;

4107
	while (off > 0 && kvm_io_bus_cmp(&key, &bus->range[off-1]) == 0)
4108 4109 4110 4111 4112
		off--;

	return off;
}

4113
static int __kvm_io_bus_write(struct kvm_vcpu *vcpu, struct kvm_io_bus *bus,
C
Cornelia Huck 已提交
4114 4115 4116 4117 4118 4119 4120 4121 4122
			      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 &&
4123
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
4124
		if (!kvm_iodevice_write(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
4125 4126 4127 4128 4129 4130 4131 4132
					range->len, val))
			return idx;
		idx++;
	}

	return -EOPNOTSUPP;
}

4133
/* kvm_io_bus_write - called under kvm->slots_lock */
4134
int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
4135
		     int len, const void *val)
4136
{
4137
	struct kvm_io_bus *bus;
4138
	struct kvm_io_range range;
C
Cornelia Huck 已提交
4139
	int r;
4140 4141 4142 4143 4144

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
4145

4146
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
4147 4148
	if (!bus)
		return -ENOMEM;
4149
	r = __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
4150 4151
	return r < 0 ? r : 0;
}
L
Leo Yan 已提交
4152
EXPORT_SYMBOL_GPL(kvm_io_bus_write);
C
Cornelia Huck 已提交
4153 4154

/* kvm_io_bus_write_cookie - called under kvm->slots_lock */
4155 4156
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
4157 4158 4159 4160 4161 4162 4163 4164 4165
{
	struct kvm_io_bus *bus;
	struct kvm_io_range range;

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

4166
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
4167 4168
	if (!bus)
		return -ENOMEM;
C
Cornelia Huck 已提交
4169 4170 4171

	/* First try the device referenced by cookie. */
	if ((cookie >= 0) && (cookie < bus->dev_count) &&
4172
	    (kvm_io_bus_cmp(&range, &bus->range[cookie]) == 0))
4173
		if (!kvm_iodevice_write(vcpu, bus->range[cookie].dev, addr, len,
C
Cornelia Huck 已提交
4174 4175 4176 4177 4178 4179 4180
					val))
			return cookie;

	/*
	 * cookie contained garbage; fall back to search and return the
	 * correct cookie value.
	 */
4181
	return __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
4182 4183
}

4184 4185
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
4186 4187 4188 4189
{
	int idx;

	idx = kvm_io_bus_get_first_dev(bus, range->addr, range->len);
4190 4191 4192 4193
	if (idx < 0)
		return -EOPNOTSUPP;

	while (idx < bus->dev_count &&
4194
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
4195
		if (!kvm_iodevice_read(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
4196 4197
				       range->len, val))
			return idx;
4198 4199 4200
		idx++;
	}

4201 4202
	return -EOPNOTSUPP;
}
4203

4204
/* kvm_io_bus_read - called under kvm->slots_lock */
4205
int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
M
Marcelo Tosatti 已提交
4206
		    int len, void *val)
4207
{
4208
	struct kvm_io_bus *bus;
4209
	struct kvm_io_range range;
C
Cornelia Huck 已提交
4210
	int r;
4211 4212 4213 4214 4215

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
M
Marcelo Tosatti 已提交
4216

4217
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
4218 4219
	if (!bus)
		return -ENOMEM;
4220
	r = __kvm_io_bus_read(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
4221 4222
	return r < 0 ? r : 0;
}
4223

4224
/* Caller must hold slots_lock. */
4225 4226
int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
			    int len, struct kvm_io_device *dev)
4227
{
4228
	int i;
M
Marcelo Tosatti 已提交
4229
	struct kvm_io_bus *new_bus, *bus;
4230
	struct kvm_io_range range;
4231

4232
	bus = kvm_get_bus(kvm, bus_idx);
4233 4234 4235
	if (!bus)
		return -ENOMEM;

4236 4237
	/* exclude ioeventfd which is limited by maximum fd */
	if (bus->dev_count - bus->ioeventfd_count > NR_IOBUS_DEVS - 1)
4238
		return -ENOSPC;
4239

4240
	new_bus = kmalloc(struct_size(bus, range, bus->dev_count + 1),
4241
			  GFP_KERNEL_ACCOUNT);
M
Marcelo Tosatti 已提交
4242 4243
	if (!new_bus)
		return -ENOMEM;
4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259

	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
4260 4261 4262
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
4263 4264 4265 4266

	return 0;
}

4267
/* Caller must hold slots_lock. */
4268 4269
void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
			       struct kvm_io_device *dev)
4270
{
4271
	int i;
M
Marcelo Tosatti 已提交
4272
	struct kvm_io_bus *new_bus, *bus;
4273

4274
	bus = kvm_get_bus(kvm, bus_idx);
4275
	if (!bus)
4276
		return;
4277

4278 4279
	for (i = 0; i < bus->dev_count; i++)
		if (bus->range[i].dev == dev) {
4280 4281
			break;
		}
M
Marcelo Tosatti 已提交
4282

4283 4284
	if (i == bus->dev_count)
		return;
4285

4286
	new_bus = kmalloc(struct_size(bus, range, bus->dev_count - 1),
4287
			  GFP_KERNEL_ACCOUNT);
4288 4289 4290 4291
	if (!new_bus)  {
		pr_err("kvm: failed to shrink bus, removing it completely\n");
		goto broken;
	}
4292 4293 4294 4295 4296

	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
4297

4298
broken:
M
Marcelo Tosatti 已提交
4299 4300 4301
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
4302
	return;
4303 4304
}

4305 4306 4307 4308 4309 4310 4311 4312 4313 4314
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);
4315 4316
	if (!bus)
		goto out_unlock;
4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330

	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);

4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342
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.
	 */
4343
	if (!refcount_inc_not_zero(&stat_data->kvm->users_count))
4344 4345
		return -ENOENT;

4346
	if (simple_attr_open(inode, file, get,
4347 4348 4349
		    KVM_DBGFS_GET_MODE(stat_data->dbgfs_item) & 0222
		    ? set : NULL,
		    fmt)) {
4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367
		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;
}

4368
static int kvm_get_stat_per_vm(struct kvm *kvm, size_t offset, u64 *val)
4369
{
4370
	*val = *(ulong *)((void *)kvm + offset);
4371

4372 4373 4374 4375 4376 4377
	return 0;
}

static int kvm_clear_stat_per_vm(struct kvm *kvm, size_t offset)
{
	*(ulong *)((void *)kvm + offset) = 0;
4378 4379 4380 4381

	return 0;
}

4382
static int kvm_get_stat_per_vcpu(struct kvm *kvm, size_t offset, u64 *val)
4383
{
4384 4385
	int i;
	struct kvm_vcpu *vcpu;
4386

4387
	*val = 0;
4388

4389 4390
	kvm_for_each_vcpu(i, vcpu, kvm)
		*val += *(u64 *)((void *)vcpu + offset);
4391 4392 4393 4394

	return 0;
}

4395
static int kvm_clear_stat_per_vcpu(struct kvm *kvm, size_t offset)
4396
{
4397 4398
	int i;
	struct kvm_vcpu *vcpu;
4399

4400 4401 4402 4403 4404
	kvm_for_each_vcpu(i, vcpu, kvm)
		*(u64 *)((void *)vcpu + offset) = 0;

	return 0;
}
4405

4406
static int kvm_stat_data_get(void *data, u64 *val)
4407
{
4408
	int r = -EFAULT;
4409 4410
	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)data;

4411 4412 4413 4414 4415 4416 4417 4418 4419 4420
	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;
	}
4421

4422
	return r;
4423 4424
}

4425
static int kvm_stat_data_clear(void *data, u64 val)
4426
{
4427
	int r = -EFAULT;
4428 4429 4430 4431 4432
	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)data;

	if (val)
		return -EINVAL;

4433 4434 4435 4436 4437 4438 4439 4440 4441 4442
	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;
	}
4443

4444
	return r;
4445 4446
}

4447
static int kvm_stat_data_open(struct inode *inode, struct file *file)
4448 4449
{
	__simple_attr_check_format("%llu\n", 0ull);
4450 4451
	return kvm_debugfs_open(inode, file, kvm_stat_data_get,
				kvm_stat_data_clear, "%llu\n");
4452 4453
}

4454 4455 4456
static const struct file_operations stat_fops_per_vm = {
	.owner = THIS_MODULE,
	.open = kvm_stat_data_open,
4457
	.release = kvm_debugfs_release,
4458 4459 4460
	.read = simple_attr_read,
	.write = simple_attr_write,
	.llseek = no_llseek,
4461 4462
};

4463
static int vm_stat_get(void *_offset, u64 *val)
4464 4465 4466
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
4467
	u64 tmp_val;
4468

4469
	*val = 0;
J
Junaid Shahid 已提交
4470
	mutex_lock(&kvm_lock);
4471
	list_for_each_entry(kvm, &vm_list, vm_list) {
4472
		kvm_get_stat_per_vm(kvm, offset, &tmp_val);
4473 4474
		*val += tmp_val;
	}
J
Junaid Shahid 已提交
4475
	mutex_unlock(&kvm_lock);
4476
	return 0;
4477 4478
}

4479 4480 4481 4482 4483 4484 4485 4486
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
4487
	mutex_lock(&kvm_lock);
4488
	list_for_each_entry(kvm, &vm_list, vm_list) {
4489
		kvm_clear_stat_per_vm(kvm, offset);
4490
	}
J
Junaid Shahid 已提交
4491
	mutex_unlock(&kvm_lock);
4492 4493 4494 4495 4496

	return 0;
}

DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, vm_stat_clear, "%llu\n");
4497

4498
static int vcpu_stat_get(void *_offset, u64 *val)
A
Avi Kivity 已提交
4499 4500 4501
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
4502
	u64 tmp_val;
A
Avi Kivity 已提交
4503

4504
	*val = 0;
J
Junaid Shahid 已提交
4505
	mutex_lock(&kvm_lock);
4506
	list_for_each_entry(kvm, &vm_list, vm_list) {
4507
		kvm_get_stat_per_vcpu(kvm, offset, &tmp_val);
4508 4509
		*val += tmp_val;
	}
J
Junaid Shahid 已提交
4510
	mutex_unlock(&kvm_lock);
4511
	return 0;
A
Avi Kivity 已提交
4512 4513
}

4514 4515 4516 4517 4518 4519 4520 4521
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
4522
	mutex_lock(&kvm_lock);
4523
	list_for_each_entry(kvm, &vm_list, vm_list) {
4524
		kvm_clear_stat_per_vcpu(kvm, offset);
4525
	}
J
Junaid Shahid 已提交
4526
	mutex_unlock(&kvm_lock);
4527 4528 4529 4530 4531 4532

	return 0;
}

DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, vcpu_stat_clear,
			"%llu\n");
4533

4534
static const struct file_operations *stat_fops[] = {
4535 4536 4537
	[KVM_STAT_VCPU] = &vcpu_stat_fops,
	[KVM_STAT_VM]   = &vm_stat_fops,
};
A
Avi Kivity 已提交
4538

4539 4540 4541 4542 4543 4544 4545 4546
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
4547
	mutex_lock(&kvm_lock);
4548 4549 4550 4551 4552 4553 4554 4555
	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
4556
	mutex_unlock(&kvm_lock);
4557

4558
	env = kzalloc(sizeof(*env), GFP_KERNEL_ACCOUNT);
4559 4560 4561 4562 4563 4564
	if (!env)
		return;

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

4565
	if (type == KVM_EVENT_CREATE_VM) {
4566
		add_uevent_var(env, "EVENT=create");
4567 4568
		kvm->userspace_pid = task_pid_nr(current);
	} else if (type == KVM_EVENT_DESTROY_VM) {
4569
		add_uevent_var(env, "EVENT=destroy");
4570 4571
	}
	add_uevent_var(env, "PID=%d", kvm->userspace_pid);
4572

4573
	if (!IS_ERR_OR_NULL(kvm->debugfs_dentry)) {
4574
		char *tmp, *p = kmalloc(PATH_MAX, GFP_KERNEL_ACCOUNT);
4575 4576 4577 4578 4579 4580

		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);
4581 4582 4583 4584 4585 4586 4587 4588
		}
	}
	/* 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);
}

4589
static void kvm_init_debug(void)
A
Avi Kivity 已提交
4590 4591 4592
{
	struct kvm_stats_debugfs_item *p;

4593
	kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
4594

4595 4596
	kvm_debugfs_num_entries = 0;
	for (p = debugfs_entries; p->name; ++p, kvm_debugfs_num_entries++) {
4597 4598
		debugfs_create_file(p->name, KVM_DBGFS_GET_MODE(p),
				    kvm_debugfs_dir, (void *)(long)p->offset,
4599
				    stat_fops[p->kind]);
4600
	}
A
Avi Kivity 已提交
4601 4602
}

4603
static int kvm_suspend(void)
4604
{
4605
	if (kvm_usage_count)
4606
		hardware_disable_nolock(NULL);
4607 4608 4609
	return 0;
}

4610
static void kvm_resume(void)
4611
{
4612
	if (kvm_usage_count) {
4613 4614 4615
#ifdef CONFIG_LOCKDEP
		WARN_ON(lockdep_is_held(&kvm_count_lock));
#endif
4616
		hardware_enable_nolock(NULL);
4617
	}
4618 4619
}

4620
static struct syscore_ops kvm_syscore_ops = {
4621 4622 4623 4624
	.suspend = kvm_suspend,
	.resume = kvm_resume,
};

4625 4626 4627 4628 4629 4630 4631 4632 4633
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);
4634

4635
	WRITE_ONCE(vcpu->preempted, false);
4636
	WRITE_ONCE(vcpu->ready, false);
4637

4638
	__this_cpu_write(kvm_running_vcpu, vcpu);
R
Radim Krčmář 已提交
4639
	kvm_arch_sched_in(vcpu, cpu);
4640
	kvm_arch_vcpu_load(vcpu, cpu);
4641 4642 4643 4644 4645 4646 4647
}

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

4648
	if (current->state == TASK_RUNNING) {
4649
		WRITE_ONCE(vcpu->preempted, true);
4650 4651
		WRITE_ONCE(vcpu->ready, true);
	}
4652
	kvm_arch_vcpu_put(vcpu);
4653 4654 4655 4656 4657
	__this_cpu_write(kvm_running_vcpu, NULL);
}

/**
 * kvm_get_running_vcpu - get the vcpu running on the current CPU.
4658 4659 4660 4661 4662 4663
 *
 * 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.
4664 4665 4666
 */
struct kvm_vcpu *kvm_get_running_vcpu(void)
{
4667 4668 4669 4670 4671 4672 4673
	struct kvm_vcpu *vcpu;

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

	return vcpu;
4674
}
4675
EXPORT_SYMBOL_GPL(kvm_get_running_vcpu);
4676 4677 4678 4679 4680 4681 4682

/**
 * 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;
4683 4684
}

4685 4686 4687 4688 4689 4690
struct kvm_cpu_compat_check {
	void *opaque;
	int *ret;
};

static void check_processor_compat(void *data)
4691
{
4692 4693 4694
	struct kvm_cpu_compat_check *c = data;

	*c->ret = kvm_arch_check_processor_compat(c->opaque);
4695 4696
}

4697
int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
4698
		  struct module *module)
A
Avi Kivity 已提交
4699
{
4700
	struct kvm_cpu_compat_check c;
A
Avi Kivity 已提交
4701
	int r;
Y
Yang, Sheng 已提交
4702
	int cpu;
A
Avi Kivity 已提交
4703

4704 4705
	r = kvm_arch_init(opaque);
	if (r)
4706
		goto out_fail;
4707

4708 4709 4710 4711
	/*
	 * 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
4712 4713
	 * kvm_arch_init must be called before kvm_irqfd_init to avoid creating
	 * conflicts in case kvm is already setup for another implementation.
4714
	 */
P
Paolo Bonzini 已提交
4715 4716 4717
	r = kvm_irqfd_init();
	if (r)
		goto out_irqfd;
4718

4719
	if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
4720 4721 4722 4723
		r = -ENOMEM;
		goto out_free_0;
	}

4724
	r = kvm_arch_hardware_setup(opaque);
A
Avi Kivity 已提交
4725
	if (r < 0)
4726
		goto out_free_1;
A
Avi Kivity 已提交
4727

4728 4729
	c.ret = &r;
	c.opaque = opaque;
Y
Yang, Sheng 已提交
4730
	for_each_online_cpu(cpu) {
4731
		smp_call_function_single(cpu, check_processor_compat, &c, 1);
Y
Yang, Sheng 已提交
4732
		if (r < 0)
4733
			goto out_free_2;
Y
Yang, Sheng 已提交
4734 4735
	}

T
Thomas Gleixner 已提交
4736
	r = cpuhp_setup_state_nocalls(CPUHP_AP_KVM_STARTING, "kvm/cpu:starting",
4737
				      kvm_starting_cpu, kvm_dying_cpu);
A
Avi Kivity 已提交
4738
	if (r)
4739
		goto out_free_2;
A
Avi Kivity 已提交
4740 4741
	register_reboot_notifier(&kvm_reboot_notifier);

4742
	/* A kmem cache lets us meet the alignment requirements of fx_save. */
4743 4744
	if (!vcpu_align)
		vcpu_align = __alignof__(struct kvm_vcpu);
4745 4746 4747 4748 4749 4750
	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);
4751 4752
	if (!kvm_vcpu_cache) {
		r = -ENOMEM;
4753
		goto out_free_3;
4754 4755
	}

4756 4757 4758 4759
	r = kvm_async_pf_init();
	if (r)
		goto out_free;

A
Avi Kivity 已提交
4760
	kvm_chardev_ops.owner = module;
4761 4762
	kvm_vm_fops.owner = module;
	kvm_vcpu_fops.owner = module;
A
Avi Kivity 已提交
4763 4764 4765

	r = misc_register(&kvm_dev);
	if (r) {
X
Xiubo Li 已提交
4766
		pr_err("kvm: misc device register failed\n");
4767
		goto out_unreg;
A
Avi Kivity 已提交
4768 4769
	}

4770 4771
	register_syscore_ops(&kvm_syscore_ops);

4772 4773 4774
	kvm_preempt_ops.sched_in = kvm_sched_in;
	kvm_preempt_ops.sched_out = kvm_sched_out;

4775
	kvm_init_debug();
4776

P
Paolo Bonzini 已提交
4777 4778 4779
	r = kvm_vfio_ops_init();
	WARN_ON(r);

4780
	return 0;
A
Avi Kivity 已提交
4781

4782 4783
out_unreg:
	kvm_async_pf_deinit();
A
Avi Kivity 已提交
4784
out_free:
4785
	kmem_cache_destroy(kvm_vcpu_cache);
4786
out_free_3:
A
Avi Kivity 已提交
4787
	unregister_reboot_notifier(&kvm_reboot_notifier);
4788
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4789
out_free_2:
4790
	kvm_arch_hardware_unsetup();
4791
out_free_1:
4792
	free_cpumask_var(cpus_hardware_enabled);
4793
out_free_0:
4794
	kvm_irqfd_exit();
P
Paolo Bonzini 已提交
4795
out_irqfd:
4796 4797
	kvm_arch_exit();
out_fail:
A
Avi Kivity 已提交
4798 4799
	return r;
}
4800
EXPORT_SYMBOL_GPL(kvm_init);
A
Avi Kivity 已提交
4801

4802
void kvm_exit(void)
A
Avi Kivity 已提交
4803
{
4804
	debugfs_remove_recursive(kvm_debugfs_dir);
A
Avi Kivity 已提交
4805
	misc_deregister(&kvm_dev);
4806
	kmem_cache_destroy(kvm_vcpu_cache);
4807
	kvm_async_pf_deinit();
4808
	unregister_syscore_ops(&kvm_syscore_ops);
A
Avi Kivity 已提交
4809
	unregister_reboot_notifier(&kvm_reboot_notifier);
4810
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4811
	on_each_cpu(hardware_disable_nolock, NULL, 1);
4812
	kvm_arch_hardware_unsetup();
4813
	kvm_arch_exit();
4814
	kvm_irqfd_exit();
4815
	free_cpumask_var(cpus_hardware_enabled);
4816
	kvm_vfio_ops_exit();
A
Avi Kivity 已提交
4817
}
4818
EXPORT_SYMBOL_GPL(kvm_exit);
4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867 4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901

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;
}