kvm_main.c 106.7 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>
58
#include <asm/pgtable.h>
A
Avi Kivity 已提交
59

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

152 153
static bool largepages_enabled = true;

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

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

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

	return is_zone_device_page(pfn_to_page(pfn));
}

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

	return true;
}

194 195 196 197 198 199 200 201 202 203
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
204 205 206
/*
 * Switches to specified vcpu, until a matching vcpu_put()
 */
207
void vcpu_load(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
208
{
209
	int cpu = get_cpu();
210 211

	__this_cpu_write(kvm_running_vcpu, vcpu);
212
	preempt_notifier_register(&vcpu->preempt_notifier);
213
	kvm_arch_vcpu_load(vcpu, cpu);
214
	put_cpu();
A
Avi Kivity 已提交
215
}
216
EXPORT_SYMBOL_GPL(vcpu_load);
A
Avi Kivity 已提交
217

218
void vcpu_put(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
219
{
220
	preempt_disable();
221
	kvm_arch_vcpu_put(vcpu);
222
	preempt_notifier_unregister(&vcpu->preempt_notifier);
223
	__this_cpu_write(kvm_running_vcpu, NULL);
224
	preempt_enable();
A
Avi Kivity 已提交
225
}
226
EXPORT_SYMBOL_GPL(vcpu_put);
A
Avi Kivity 已提交
227

228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245
/* 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;
}

246 247 248 249
static void ack_flush(void *_completed)
{
}

250 251 252 253 254 255 256 257 258 259 260 261
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;
}

262 263
bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
				 unsigned long *vcpu_bitmap, cpumask_var_t tmp)
264
{
265
	int i, cpu, me;
266
	struct kvm_vcpu *vcpu;
267
	bool called;
268

269
	me = get_cpu();
270

271
	kvm_for_each_vcpu(i, vcpu, kvm) {
272
		if (vcpu_bitmap && !test_bit(i, vcpu_bitmap))
273 274
			continue;

275
		kvm_make_request(req, vcpu);
276
		cpu = vcpu->cpu;
277

278 279
		if (!(req & KVM_REQUEST_NO_WAKEUP) && kvm_vcpu_wake_up(vcpu))
			continue;
280

281
		if (tmp != NULL && cpu != -1 && cpu != me &&
282
		    kvm_request_needs_ipi(vcpu, req))
283
			__cpumask_set_cpu(cpu, tmp);
284
	}
285 286

	called = kvm_kick_many_cpus(tmp, !!(req & KVM_REQUEST_WAIT));
287
	put_cpu();
288 289 290 291 292 293 294 295 296 297 298

	return called;
}

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

	zalloc_cpumask_var(&cpus, GFP_ATOMIC);

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

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

305
#ifndef CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALL
306
void kvm_flush_remote_tlbs(struct kvm *kvm)
307
{
308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324
	/*
	 * 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.
	 */
325 326
	if (!kvm_arch_flush_remote_tlb(kvm)
	    || kvm_make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH))
327
		++kvm->stat.remote_tlb_flush;
328
	cmpxchg(&kvm->tlbs_dirty, dirty_count, 0);
329
}
330
EXPORT_SYMBOL_GPL(kvm_flush_remote_tlbs);
331
#endif
332

333 334
void kvm_reload_remote_mmus(struct kvm *kvm)
{
335
	kvm_make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD);
336
}
337

338
static void kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id)
R
Rusty Russell 已提交
339 340 341 342 343
{
	mutex_init(&vcpu->mutex);
	vcpu->cpu = -1;
	vcpu->kvm = kvm;
	vcpu->vcpu_id = id;
344
	vcpu->pid = NULL;
345
	init_swait_queue_head(&vcpu->wq);
346
	kvm_async_pf_vcpu_init(vcpu);
R
Rusty Russell 已提交
347

348 349 350
	vcpu->pre_pcpu = -1;
	INIT_LIST_HEAD(&vcpu->blocked_vcpu_list);

351 352
	kvm_vcpu_set_in_spin_loop(vcpu, false);
	kvm_vcpu_set_dy_eligible(vcpu, false);
353
	vcpu->preempted = false;
354
	vcpu->ready = false;
355
	preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);
R
Rusty Russell 已提交
356 357
}

358 359 360
void kvm_vcpu_destroy(struct kvm_vcpu *vcpu)
{
	kvm_arch_vcpu_destroy(vcpu);
361

362 363 364 365 366 367 368
	/*
	 * 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));

369
	free_page((unsigned long)vcpu->run);
370
	kmem_cache_free(kvm_vcpu_cache, vcpu);
371 372 373
}
EXPORT_SYMBOL_GPL(kvm_vcpu_destroy);

374 375 376 377 378 379
#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);
}

380 381 382 383 384 385
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);
386
	int idx;
387

388
	idx = srcu_read_lock(&kvm->srcu);
389 390
	spin_lock(&kvm->mmu_lock);
	kvm->mmu_notifier_seq++;
391 392 393 394

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

395
	spin_unlock(&kvm->mmu_lock);
396
	srcu_read_unlock(&kvm->srcu, idx);
397 398
}

399
static int kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn,
400
					const struct mmu_notifier_range *range)
401 402
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
403
	int need_tlb_flush = 0, idx;
404
	int ret;
405

406
	idx = srcu_read_lock(&kvm->srcu);
407 408 409 410 411 412 413
	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++;
414
	need_tlb_flush = kvm_unmap_hva_range(kvm, range->start, range->end);
415
	need_tlb_flush |= kvm->tlbs_dirty;
416 417 418
	/* we've to flush the tlb before the pages can be freed */
	if (need_tlb_flush)
		kvm_flush_remote_tlbs(kvm);
419 420

	spin_unlock(&kvm->mmu_lock);
421

422
	ret = kvm_arch_mmu_notifier_invalidate_range(kvm, range->start,
423 424
					range->end,
					mmu_notifier_range_blockable(range));
425

426
	srcu_read_unlock(&kvm->srcu, idx);
427 428

	return ret;
429 430 431
}

static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
432
					const struct mmu_notifier_range *range)
433 434 435 436 437 438 439 440 441 442
{
	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++;
443
	smp_wmb();
444 445
	/*
	 * The above sequence increase must be visible before the
446 447
	 * below count decrease, which is ensured by the smp_wmb above
	 * in conjunction with the smp_rmb in mmu_notifier_retry().
448 449 450 451 452 453 454 455 456
	 */
	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
457 458
					      unsigned long start,
					      unsigned long end)
459 460
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
461
	int young, idx;
462

463
	idx = srcu_read_lock(&kvm->srcu);
464 465
	spin_lock(&kvm->mmu_lock);

A
Andres Lagar-Cavilla 已提交
466
	young = kvm_age_hva(kvm, start, end);
467 468 469
	if (young)
		kvm_flush_remote_tlbs(kvm);

470 471 472
	spin_unlock(&kvm->mmu_lock);
	srcu_read_unlock(&kvm->srcu, idx);

473 474 475
	return young;
}

476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505
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
506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521
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;
}

522 523 524 525
static void kvm_mmu_notifier_release(struct mmu_notifier *mn,
				     struct mm_struct *mm)
{
	struct kvm *kvm = mmu_notifier_to_kvm(mn);
526 527 528
	int idx;

	idx = srcu_read_lock(&kvm->srcu);
529
	kvm_arch_flush_shadow_all(kvm);
530
	srcu_read_unlock(&kvm->srcu, idx);
531 532
}

533 534 535 536
static const struct mmu_notifier_ops kvm_mmu_notifier_ops = {
	.invalidate_range_start	= kvm_mmu_notifier_invalidate_range_start,
	.invalidate_range_end	= kvm_mmu_notifier_invalidate_range_end,
	.clear_flush_young	= kvm_mmu_notifier_clear_flush_young,
537
	.clear_young		= kvm_mmu_notifier_clear_young,
A
Andrea Arcangeli 已提交
538
	.test_young		= kvm_mmu_notifier_test_young,
539
	.change_pte		= kvm_mmu_notifier_change_pte,
540
	.release		= kvm_mmu_notifier_release,
541
};
542 543 544 545 546 547 548 549 550 551 552 553 554 555

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;
}

556 557
#endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */

558
static struct kvm_memslots *kvm_alloc_memslots(void)
559 560
{
	int i;
561
	struct kvm_memslots *slots;
562

563
	slots = kvzalloc(sizeof(struct kvm_memslots), GFP_KERNEL_ACCOUNT);
564 565 566
	if (!slots)
		return NULL;

567
	for (i = 0; i < KVM_MEM_SLOTS_NUM; i++)
568
		slots->id_to_index[i] = slots->memslots[i].id = i;
569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606

	return slots;
}

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

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

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

	kvm_arch_free_memslot(kvm, free, dont);

	free->npages = 0;
}

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

	if (!slots)
		return;

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

	kvfree(slots);
607 608
}

609 610 611 612 613 614 615 616 617
static void kvm_destroy_vm_debugfs(struct kvm *kvm)
{
	int i;

	if (!kvm->debugfs_dentry)
		return;

	debugfs_remove_recursive(kvm->debugfs_dentry);

618 619 620 621 622
	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);
	}
623 624 625 626 627 628 629 630 631 632 633 634
}

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);
635
	kvm->debugfs_dentry = debugfs_create_dir(dir_name, kvm_debugfs_dir);
636 637 638

	kvm->debugfs_stat_data = kcalloc(kvm_debugfs_num_entries,
					 sizeof(*kvm->debugfs_stat_data),
639
					 GFP_KERNEL_ACCOUNT);
640 641 642 643
	if (!kvm->debugfs_stat_data)
		return -ENOMEM;

	for (p = debugfs_entries; p->name; p++) {
644
		stat_data = kzalloc(sizeof(*stat_data), GFP_KERNEL_ACCOUNT);
645 646 647 648
		if (!stat_data)
			return -ENOMEM;

		stat_data->kvm = kvm;
649
		stat_data->dbgfs_item = p;
650
		kvm->debugfs_stat_data[p - debugfs_entries] = stat_data;
651 652 653
		debugfs_create_file(p->name, KVM_DBGFS_GET_MODE(p),
				    kvm->debugfs_dentry, stat_data,
				    &stat_fops_per_vm);
654 655 656 657
	}
	return 0;
}

658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674
/*
 * 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)
{
}

675
static struct kvm *kvm_create_vm(unsigned long type)
A
Avi Kivity 已提交
676
{
677
	struct kvm *kvm = kvm_arch_alloc_vm();
678 679
	int r = -ENOMEM;
	int i;
A
Avi Kivity 已提交
680

681 682 683
	if (!kvm)
		return ERR_PTR(-ENOMEM);

684
	spin_lock_init(&kvm->mmu_lock);
V
Vegard Nossum 已提交
685
	mmgrab(current->mm);
686 687 688 689 690 691 692
	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);

693 694
	BUILD_BUG_ON(KVM_MEM_SLOTS_NUM > SHRT_MAX);

695 696 697 698 699
	if (init_srcu_struct(&kvm->srcu))
		goto out_err_no_srcu;
	if (init_srcu_struct(&kvm->irq_srcu))
		goto out_err_no_irq_srcu;

700
	refcount_set(&kvm->users_count, 1);
701
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
702
		struct kvm_memslots *slots = kvm_alloc_memslots();
703

704
		if (!slots)
705
			goto out_err_no_arch_destroy_vm;
706
		/* Generations must be different for each address space. */
707
		slots->generation = i;
708
		rcu_assign_pointer(kvm->memslots[i], slots);
709
	}
710

M
Marcelo Tosatti 已提交
711
	for (i = 0; i < KVM_NR_BUSES; i++) {
712
		rcu_assign_pointer(kvm->buses[i],
713
			kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL_ACCOUNT));
714
		if (!kvm->buses[i])
715
			goto out_err_no_arch_destroy_vm;
M
Marcelo Tosatti 已提交
716
	}
717

718
	r = kvm_arch_init_vm(kvm, type);
719
	if (r)
720
		goto out_err_no_arch_destroy_vm;
721 722 723

	r = hardware_enable_all();
	if (r)
724
		goto out_err_no_disable;
725

726
#ifdef CONFIG_HAVE_KVM_IRQFD
727
	INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list);
728
#endif
A
Avi Kivity 已提交
729

730
	r = kvm_init_mmu_notifier(kvm);
731 732 733 734
	if (r)
		goto out_err_no_mmu_notifier;

	r = kvm_arch_post_init_vm(kvm);
735 736 737
	if (r)
		goto out_err;

J
Junaid Shahid 已提交
738
	mutex_lock(&kvm_lock);
739
	list_add(&kvm->vm_list, &vm_list);
J
Junaid Shahid 已提交
740
	mutex_unlock(&kvm_lock);
741

742 743
	preempt_notifier_inc();

744
	return kvm;
745 746

out_err:
747 748 749 750 751
#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:
752
	hardware_disable_all();
753
out_err_no_disable:
754 755
	kvm_arch_destroy_vm(kvm);
out_err_no_arch_destroy_vm:
756
	WARN_ON_ONCE(!refcount_dec_and_test(&kvm->users_count));
M
Marcelo Tosatti 已提交
757
	for (i = 0; i < KVM_NR_BUSES; i++)
758
		kfree(kvm_get_bus(kvm, i));
759
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
760
		kvm_free_memslots(kvm, __kvm_memslots(kvm, i));
761 762 763 764
	cleanup_srcu_struct(&kvm->irq_srcu);
out_err_no_irq_srcu:
	cleanup_srcu_struct(&kvm->srcu);
out_err_no_srcu:
765
	kvm_arch_free_vm(kvm);
766
	mmdrop(current->mm);
767
	return ERR_PTR(r);
768 769
}

770 771
static void kvm_destroy_devices(struct kvm *kvm)
{
G
Geliang Tang 已提交
772
	struct kvm_device *dev, *tmp;
773

774 775 776 777 778
	/*
	 * 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
779 780
	list_for_each_entry_safe(dev, tmp, &kvm->devices, vm_node) {
		list_del(&dev->vm_node);
781 782 783 784
		dev->ops->destroy(dev);
	}
}

785 786
static void kvm_destroy_vm(struct kvm *kvm)
{
M
Marcelo Tosatti 已提交
787
	int i;
788 789
	struct mm_struct *mm = kvm->mm;

790
	kvm_uevent_notify_change(KVM_EVENT_DESTROY_VM, kvm);
791
	kvm_destroy_vm_debugfs(kvm);
792
	kvm_arch_sync_events(kvm);
J
Junaid Shahid 已提交
793
	mutex_lock(&kvm_lock);
794
	list_del(&kvm->vm_list);
J
Junaid Shahid 已提交
795
	mutex_unlock(&kvm_lock);
796 797
	kvm_arch_pre_destroy_vm(kvm);

798
	kvm_free_irq_routing(kvm);
799
	for (i = 0; i < KVM_NR_BUSES; i++) {
800
		struct kvm_io_bus *bus = kvm_get_bus(kvm, i);
801 802 803

		if (bus)
			kvm_io_bus_destroy(bus);
804 805
		kvm->buses[i] = NULL;
	}
806
	kvm_coalesced_mmio_free(kvm);
807 808
#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
	mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
809
#else
810
	kvm_arch_flush_shadow_all(kvm);
811
#endif
812
	kvm_arch_destroy_vm(kvm);
813
	kvm_destroy_devices(kvm);
814
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
815
		kvm_free_memslots(kvm, __kvm_memslots(kvm, i));
816
	cleanup_srcu_struct(&kvm->irq_srcu);
817 818
	cleanup_srcu_struct(&kvm->srcu);
	kvm_arch_free_vm(kvm);
819
	preempt_notifier_dec();
820
	hardware_disable_all();
821
	mmdrop(mm);
822 823
}

I
Izik Eidus 已提交
824 825
void kvm_get_kvm(struct kvm *kvm)
{
826
	refcount_inc(&kvm->users_count);
I
Izik Eidus 已提交
827 828 829 830 831
}
EXPORT_SYMBOL_GPL(kvm_get_kvm);

void kvm_put_kvm(struct kvm *kvm)
{
832
	if (refcount_dec_and_test(&kvm->users_count))
I
Izik Eidus 已提交
833 834 835 836
		kvm_destroy_vm(kvm);
}
EXPORT_SYMBOL_GPL(kvm_put_kvm);

837 838 839 840 841 842 843 844 845 846 847 848
/*
 * 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
849

850 851 852 853
static int kvm_vm_release(struct inode *inode, struct file *filp)
{
	struct kvm *kvm = filp->private_data;

G
Gregory Haskins 已提交
854 855
	kvm_irqfd_release(kvm);

I
Izik Eidus 已提交
856
	kvm_put_kvm(kvm);
A
Avi Kivity 已提交
857 858 859
	return 0;
}

860 861
/*
 * Allocation size is twice as large as the actual dirty bitmap size.
862
 * See x86's kvm_vm_ioctl_get_dirty_log() why this is needed.
863
 */
864 865
static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot)
{
866
	unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot);
867

868
	memslot->dirty_bitmap = kvzalloc(dirty_bytes, GFP_KERNEL_ACCOUNT);
869 870 871 872 873 874
	if (!memslot->dirty_bitmap)
		return -ENOMEM;

	return 0;
}

875
/*
876 877 878 879
 * Insert memslot and re-sort memslots based on their GFN,
 * so binary search could be used to lookup GFN.
 * Sorting algorithm takes advantage of having initially
 * sorted array and known changed memslot position.
880
 */
881
static void update_memslots(struct kvm_memslots *slots,
882 883
			    struct kvm_memory_slot *new,
			    enum kvm_mr_change change)
884
{
885 886
	int id = new->id;
	int i = slots->id_to_index[id];
887
	struct kvm_memory_slot *mslots = slots->memslots;
888

889
	WARN_ON(mslots[i].id != id);
890 891 892 893 894 895 896 897 898 899 900
	switch (change) {
	case KVM_MR_CREATE:
		slots->used_slots++;
		WARN_ON(mslots[i].npages || !new->npages);
		break;
	case KVM_MR_DELETE:
		slots->used_slots--;
		WARN_ON(new->npages || !mslots[i].npages);
		break;
	default:
		break;
901
	}
902

903
	while (i < KVM_MEM_SLOTS_NUM - 1 &&
904 905 906
	       new->base_gfn <= mslots[i + 1].base_gfn) {
		if (!mslots[i + 1].npages)
			break;
907 908 909 910
		mslots[i] = mslots[i + 1];
		slots->id_to_index[mslots[i].id] = i;
		i++;
	}
911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927

	/*
	 * The ">=" is needed when creating a slot with base_gfn == 0,
	 * so that it moves before all those with base_gfn == npages == 0.
	 *
	 * On the other hand, if new->npages is zero, the above loop has
	 * already left i pointing to the beginning of the empty part of
	 * mslots, and the ">=" would move the hole backwards in this
	 * case---which is wrong.  So skip the loop when deleting a slot.
	 */
	if (new->npages) {
		while (i > 0 &&
		       new->base_gfn >= mslots[i - 1].base_gfn) {
			mslots[i] = mslots[i - 1];
			slots->id_to_index[mslots[i].id] = i;
			i--;
		}
928 929
	} else
		WARN_ON_ONCE(i != slots->used_slots);
930

931 932
	mslots[i] = *new;
	slots->id_to_index[mslots[i].id] = i;
933 934
}

935
static int check_memory_region_flags(const struct kvm_userspace_memory_region *mem)
936
{
X
Xiao Guangrong 已提交
937 938
	u32 valid_flags = KVM_MEM_LOG_DIRTY_PAGES;

939
#ifdef __KVM_HAVE_READONLY_MEM
X
Xiao Guangrong 已提交
940 941 942 943
	valid_flags |= KVM_MEM_READONLY;
#endif

	if (mem->flags & ~valid_flags)
944 945 946 947 948
		return -EINVAL;

	return 0;
}

949
static struct kvm_memslots *install_new_memslots(struct kvm *kvm,
950
		int as_id, struct kvm_memslots *slots)
951
{
952
	struct kvm_memslots *old_memslots = __kvm_memslots(kvm, as_id);
953
	u64 gen = old_memslots->generation;
954

955 956
	WARN_ON(gen & KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS);
	slots->generation = gen | KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS;
957

958
	rcu_assign_pointer(kvm->memslots[as_id], slots);
959
	synchronize_srcu_expedited(&kvm->srcu);
960

961
	/*
962
	 * Increment the new memslot generation a second time, dropping the
M
Miaohe Lin 已提交
963
	 * update in-progress flag and incrementing the generation based on
964 965 966 967 968 969
	 * 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;

	/*
970 971 972
	 * 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
973 974
	 * space 0 will use generations 0, 2, 4, ... while address space 1 will
	 * use generations 1, 3, 5, ...
975
	 */
976
	gen += KVM_ADDRESS_SPACE_NUM;
977

978
	kvm_arch_memslots_updated(kvm, gen);
979

980
	slots->generation = gen;
981 982

	return old_memslots;
983 984
}

A
Avi Kivity 已提交
985 986 987 988 989
/*
 * Allocate some memory and give it an address in the guest physical address
 * space.
 *
 * Discontiguous memory is allowed, mostly for framebuffers.
990
 *
991
 * Must be called holding kvm->slots_lock for write.
A
Avi Kivity 已提交
992
 */
993
int __kvm_set_memory_region(struct kvm *kvm,
994
			    const struct kvm_userspace_memory_region *mem)
A
Avi Kivity 已提交
995
{
996
	int r;
A
Avi Kivity 已提交
997
	gfn_t base_gfn;
998
	unsigned long npages;
999
	struct kvm_memory_slot *slot;
A
Avi Kivity 已提交
1000
	struct kvm_memory_slot old, new;
1001
	struct kvm_memslots *slots = NULL, *old_memslots;
1002
	int as_id, id;
1003
	enum kvm_mr_change change;
A
Avi Kivity 已提交
1004

1005 1006 1007 1008
	r = check_memory_region_flags(mem);
	if (r)
		goto out;

A
Avi Kivity 已提交
1009
	r = -EINVAL;
1010 1011 1012
	as_id = mem->slot >> 16;
	id = (u16)mem->slot;

A
Avi Kivity 已提交
1013 1014 1015 1016 1017
	/* General sanity checks */
	if (mem->memory_size & (PAGE_SIZE - 1))
		goto out;
	if (mem->guest_phys_addr & (PAGE_SIZE - 1))
		goto out;
1018
	/* We can read the guest memory with __xxx_user() later on. */
1019
	if ((id < KVM_USER_MEM_SLOTS) &&
1020
	    ((mem->userspace_addr & (PAGE_SIZE - 1)) ||
1021
	     !access_ok((void __user *)(unsigned long)mem->userspace_addr,
1022
			mem->memory_size)))
1023
		goto out;
1024
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_MEM_SLOTS_NUM)
A
Avi Kivity 已提交
1025 1026 1027 1028
		goto out;
	if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
		goto out;

1029
	slot = id_to_memslot(__kvm_memslots(kvm, as_id), id);
A
Avi Kivity 已提交
1030 1031 1032
	base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
	npages = mem->memory_size >> PAGE_SHIFT;

1033 1034 1035
	if (npages > KVM_MEM_MAX_NR_PAGES)
		goto out;

1036
	new = old = *slot;
A
Avi Kivity 已提交
1037

1038
	new.id = id;
A
Avi Kivity 已提交
1039 1040 1041 1042
	new.base_gfn = base_gfn;
	new.npages = npages;
	new.flags = mem->flags;

1043 1044 1045 1046 1047
	if (npages) {
		if (!old.npages)
			change = KVM_MR_CREATE;
		else { /* Modify an existing slot. */
			if ((mem->userspace_addr != old.userspace_addr) ||
1048 1049
			    (npages != old.npages) ||
			    ((new.flags ^ old.flags) & KVM_MEM_READONLY))
1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060
				goto out;

			if (base_gfn != old.base_gfn)
				change = KVM_MR_MOVE;
			else if (new.flags != old.flags)
				change = KVM_MR_FLAGS_ONLY;
			else { /* Nothing to change. */
				r = 0;
				goto out;
			}
		}
1061 1062 1063 1064
	} else {
		if (!old.npages)
			goto out;

1065
		change = KVM_MR_DELETE;
1066 1067 1068
		new.base_gfn = 0;
		new.flags = 0;
	}
A
Avi Kivity 已提交
1069

1070
	if ((change == KVM_MR_CREATE) || (change == KVM_MR_MOVE)) {
1071 1072
		/* Check for overlaps */
		r = -EEXIST;
1073
		kvm_for_each_memslot(slot, __kvm_memslots(kvm, as_id)) {
1074
			if (slot->id == id)
1075 1076 1077 1078 1079
				continue;
			if (!((base_gfn + npages <= slot->base_gfn) ||
			      (base_gfn >= slot->base_gfn + slot->npages)))
				goto out;
		}
A
Avi Kivity 已提交
1080 1081 1082 1083
	}

	/* Free page dirty bitmap if unneeded */
	if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
A
Al Viro 已提交
1084
		new.dirty_bitmap = NULL;
A
Avi Kivity 已提交
1085 1086

	r = -ENOMEM;
1087
	if (change == KVM_MR_CREATE) {
1088
		new.userspace_addr = mem->userspace_addr;
1089

1090
		if (kvm_arch_create_memslot(kvm, &new, npages))
1091
			goto out_free;
A
Avi Kivity 已提交
1092
	}
1093

A
Avi Kivity 已提交
1094 1095
	/* Allocate page dirty bitmap if needed */
	if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
1096
		if (kvm_create_dirty_bitmap(&new) < 0)
1097
			goto out_free;
A
Avi Kivity 已提交
1098 1099
	}

1100
	slots = kvzalloc(sizeof(struct kvm_memslots), GFP_KERNEL_ACCOUNT);
1101 1102
	if (!slots)
		goto out_free;
1103
	memcpy(slots, __kvm_memslots(kvm, as_id), sizeof(struct kvm_memslots));
1104

1105
	if ((change == KVM_MR_DELETE) || (change == KVM_MR_MOVE)) {
1106
		slot = id_to_memslot(slots, id);
1107 1108
		slot->flags |= KVM_MEMSLOT_INVALID;

1109
		old_memslots = install_new_memslots(kvm, as_id, slots);
1110

1111 1112
		/* From this point no new shadow pages pointing to a deleted,
		 * or moved, memslot will be created.
1113 1114
		 *
		 * validation of sp->gfn happens in:
1115
		 *	- gfn_to_hva (kvm_read_guest, gfn_to_pfn)
1116
		 *	- kvm_is_visible_gfn (mmu_check_root)
1117
		 */
1118
		kvm_arch_flush_shadow_memslot(kvm, slot);
1119 1120 1121 1122 1123 1124

		/*
		 * We can re-use the old_memslots from above, the only difference
		 * from the currently installed memslots is the invalid flag.  This
		 * will get overwritten by update_memslots anyway.
		 */
1125
		slots = old_memslots;
1126
	}
1127

1128
	r = kvm_arch_prepare_memory_region(kvm, &new, mem, change);
1129
	if (r)
1130
		goto out_slots;
1131

1132
	/* actual memory is freed via old in kvm_free_memslot below */
1133
	if (change == KVM_MR_DELETE) {
1134
		new.dirty_bitmap = NULL;
1135
		memset(&new.arch, 0, sizeof(new.arch));
1136 1137
	}

1138
	update_memslots(slots, &new, change);
1139
	old_memslots = install_new_memslots(kvm, as_id, slots);
1140

1141
	kvm_arch_commit_memory_region(kvm, mem, &old, &new, change);
1142

1143
	kvm_free_memslot(kvm, &old, &new);
1144
	kvfree(old_memslots);
A
Avi Kivity 已提交
1145 1146
	return 0;

1147
out_slots:
1148
	kvfree(slots);
1149
out_free:
1150
	kvm_free_memslot(kvm, &new, &old);
A
Avi Kivity 已提交
1151 1152
out:
	return r;
1153
}
1154 1155 1156
EXPORT_SYMBOL_GPL(__kvm_set_memory_region);

int kvm_set_memory_region(struct kvm *kvm,
1157
			  const struct kvm_userspace_memory_region *mem)
1158 1159 1160
{
	int r;

1161
	mutex_lock(&kvm->slots_lock);
1162
	r = __kvm_set_memory_region(kvm, mem);
1163
	mutex_unlock(&kvm->slots_lock);
1164 1165
	return r;
}
1166 1167
EXPORT_SYMBOL_GPL(kvm_set_memory_region);

1168 1169
static int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
					  struct kvm_userspace_memory_region *mem)
1170
{
1171
	if ((u16)mem->slot >= KVM_USER_MEM_SLOTS)
1172
		return -EINVAL;
1173

1174
	return kvm_set_memory_region(kvm, mem);
A
Avi Kivity 已提交
1175 1176
}

1177 1178
int kvm_get_dirty_log(struct kvm *kvm,
			struct kvm_dirty_log *log, int *is_dirty)
A
Avi Kivity 已提交
1179
{
1180
	struct kvm_memslots *slots;
A
Avi Kivity 已提交
1181
	struct kvm_memory_slot *memslot;
1182
	int i, as_id, id;
1183
	unsigned long n;
A
Avi Kivity 已提交
1184 1185
	unsigned long any = 0;

1186 1187 1188
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1189
		return -EINVAL;
A
Avi Kivity 已提交
1190

1191 1192
	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
A
Avi Kivity 已提交
1193
	if (!memslot->dirty_bitmap)
1194
		return -ENOENT;
A
Avi Kivity 已提交
1195

1196
	n = kvm_dirty_bitmap_bytes(memslot);
A
Avi Kivity 已提交
1197

1198
	for (i = 0; !any && i < n/sizeof(long); ++i)
A
Avi Kivity 已提交
1199 1200 1201
		any = memslot->dirty_bitmap[i];

	if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
1202
		return -EFAULT;
A
Avi Kivity 已提交
1203

1204 1205
	if (any)
		*is_dirty = 1;
1206
	return 0;
A
Avi Kivity 已提交
1207
}
1208
EXPORT_SYMBOL_GPL(kvm_get_dirty_log);
A
Avi Kivity 已提交
1209

1210 1211
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
/**
J
Jiang Biao 已提交
1212
 * kvm_get_dirty_log_protect - get a snapshot of dirty pages
1213
 *	and reenable dirty page tracking for the corresponding pages.
1214 1215
 * @kvm:	pointer to kvm instance
 * @log:	slot id and address to which we copy the log
J
Jiang Biao 已提交
1216
 * @flush:	true if TLB flush is needed by caller
1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233
 *
 * We need to keep it in mind that VCPU threads can write to the bitmap
 * concurrently. So, to avoid losing track of dirty pages we keep the
 * following order:
 *
 *    1. Take a snapshot of the bit and clear it if needed.
 *    2. Write protect the corresponding page.
 *    3. Copy the snapshot to the userspace.
 *    4. Upon return caller flushes TLB's if needed.
 *
 * Between 2 and 4, the guest may write to the page using the remaining TLB
 * entry.  This is not a problem because the page is reported dirty using
 * the snapshot taken before and step 4 ensures that writes done after
 * exiting to userspace will be logged for the next call.
 *
 */
int kvm_get_dirty_log_protect(struct kvm *kvm,
1234
			struct kvm_dirty_log *log, bool *flush)
1235
{
1236
	struct kvm_memslots *slots;
1237
	struct kvm_memory_slot *memslot;
1238
	int i, as_id, id;
1239 1240 1241 1242
	unsigned long n;
	unsigned long *dirty_bitmap;
	unsigned long *dirty_bitmap_buffer;

1243 1244 1245
	as_id = log->slot >> 16;
	id = (u16)log->slot;
	if (as_id >= KVM_ADDRESS_SPACE_NUM || id >= KVM_USER_MEM_SLOTS)
1246
		return -EINVAL;
1247

1248 1249
	slots = __kvm_memslots(kvm, as_id);
	memslot = id_to_memslot(slots, id);
1250 1251 1252

	dirty_bitmap = memslot->dirty_bitmap;
	if (!dirty_bitmap)
1253
		return -ENOENT;
1254 1255

	n = kvm_dirty_bitmap_bytes(memslot);
1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269
	*flush = false;
	if (kvm->manual_dirty_log_protect) {
		/*
		 * Unlike kvm_get_dirty_log, we always return false in *flush,
		 * because no flush is needed until KVM_CLEAR_DIRTY_LOG.  There
		 * is some code duplication between this function and
		 * kvm_get_dirty_log, but hopefully all architecture
		 * transition to kvm_get_dirty_log_protect and kvm_get_dirty_log
		 * can be eliminated.
		 */
		dirty_bitmap_buffer = dirty_bitmap;
	} else {
		dirty_bitmap_buffer = kvm_second_dirty_bitmap(memslot);
		memset(dirty_bitmap_buffer, 0, n);
1270

1271 1272 1273 1274
		spin_lock(&kvm->mmu_lock);
		for (i = 0; i < n / sizeof(long); i++) {
			unsigned long mask;
			gfn_t offset;
1275

1276 1277 1278 1279 1280 1281 1282
			if (!dirty_bitmap[i])
				continue;

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

1283 1284 1285
			offset = i * BITS_PER_LONG;
			kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot,
								offset, mask);
1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300
		}
		spin_unlock(&kvm->mmu_lock);
	}

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

/**
 * kvm_clear_dirty_log_protect - clear dirty bits in the bitmap
 *	and reenable dirty page tracking for the corresponding pages.
 * @kvm:	pointer to kvm instance
 * @log:	slot id and address from which to fetch the bitmap of dirty pages
J
Jiang Biao 已提交
1301
 * @flush:	true if TLB flush is needed by caller
1302 1303 1304 1305 1306 1307
 */
int kvm_clear_dirty_log_protect(struct kvm *kvm,
				struct kvm_clear_dirty_log *log, bool *flush)
{
	struct kvm_memslots *slots;
	struct kvm_memory_slot *memslot;
1308
	int as_id, id;
1309
	gfn_t offset;
1310
	unsigned long i, n;
1311 1312 1313 1314 1315 1316 1317 1318
	unsigned long *dirty_bitmap;
	unsigned long *dirty_bitmap_buffer;

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

1319
	if (log->first_page & 63)
1320 1321 1322 1323 1324 1325 1326 1327 1328
		return -EINVAL;

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

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

1329
	n = ALIGN(log->num_pages, BITS_PER_LONG) / 8;
1330 1331

	if (log->first_page > memslot->npages ||
1332 1333 1334
	    log->num_pages > memslot->npages - log->first_page ||
	    (log->num_pages < memslot->npages - log->first_page && (log->num_pages & 63)))
	    return -EINVAL;
1335

1336
	*flush = false;
1337 1338 1339
	dirty_bitmap_buffer = kvm_second_dirty_bitmap(memslot);
	if (copy_from_user(dirty_bitmap_buffer, log->dirty_bitmap, n))
		return -EFAULT;
1340

1341
	spin_lock(&kvm->mmu_lock);
1342 1343
	for (offset = log->first_page, i = offset / BITS_PER_LONG,
		 n = DIV_ROUND_UP(log->num_pages, BITS_PER_LONG); n--;
1344 1345 1346 1347
	     i++, offset += BITS_PER_LONG) {
		unsigned long mask = *dirty_bitmap_buffer++;
		atomic_long_t *p = (atomic_long_t *) &dirty_bitmap[i];
		if (!mask)
1348 1349
			continue;

1350
		mask &= atomic_long_fetch_andnot(mask, p);
1351

1352 1353 1354 1355 1356 1357
		/*
		 * 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.
		*/
1358
		if (mask) {
1359
			*flush = true;
1360 1361 1362
			kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot,
								offset, mask);
		}
1363 1364
	}
	spin_unlock(&kvm->mmu_lock);
1365

1366
	return 0;
1367
}
1368
EXPORT_SYMBOL_GPL(kvm_clear_dirty_log_protect);
1369 1370
#endif

1371 1372 1373 1374 1375
bool kvm_largepages_enabled(void)
{
	return largepages_enabled;
}

1376 1377 1378 1379 1380 1381
void kvm_disable_largepages(void)
{
	largepages_enabled = false;
}
EXPORT_SYMBOL_GPL(kvm_disable_largepages);

1382 1383 1384 1385
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
1386
EXPORT_SYMBOL_GPL(gfn_to_memslot);
A
Avi Kivity 已提交
1387

1388 1389 1390 1391 1392
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);
}

1393
bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
1394
{
1395
	struct kvm_memory_slot *memslot = gfn_to_memslot(kvm, gfn);
1396

1397
	if (!memslot || memslot->id >= KVM_USER_MEM_SLOTS ||
1398
	      memslot->flags & KVM_MEMSLOT_INVALID)
1399
		return false;
1400

1401
	return true;
1402 1403 1404
}
EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);

J
Joerg Roedel 已提交
1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428
unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn)
{
	struct vm_area_struct *vma;
	unsigned long addr, size;

	size = PAGE_SIZE;

	addr = gfn_to_hva(kvm, gfn);
	if (kvm_is_error_hva(addr))
		return PAGE_SIZE;

	down_read(&current->mm->mmap_sem);
	vma = find_vma(current->mm, addr);
	if (!vma)
		goto out;

	size = vma_kernel_pagesize(vma);

out:
	up_read(&current->mm->mmap_sem);

	return size;
}

X
Xiao Guangrong 已提交
1429 1430 1431 1432 1433 1434 1435
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
1436
{
1437
	if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
X
Xiao Guangrong 已提交
1438
		return KVM_HVA_ERR_BAD;
1439

X
Xiao Guangrong 已提交
1440 1441
	if (memslot_is_readonly(slot) && write)
		return KVM_HVA_ERR_RO_BAD;
1442 1443 1444 1445

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

X
Xiao Guangrong 已提交
1446
	return __gfn_to_hva_memslot(slot, gfn);
I
Izik Eidus 已提交
1447
}
1448

X
Xiao Guangrong 已提交
1449 1450 1451 1452
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
1453
}
1454

X
Xiao Guangrong 已提交
1455
unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot,
1456
					gfn_t gfn)
X
Xiao Guangrong 已提交
1457 1458 1459 1460 1461
{
	return gfn_to_hva_many(slot, gfn, NULL);
}
EXPORT_SYMBOL_GPL(gfn_to_hva_memslot);

1462 1463
unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
{
1464
	return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL);
1465
}
1466
EXPORT_SYMBOL_GPL(gfn_to_hva);
I
Izik Eidus 已提交
1467

1468 1469 1470 1471 1472 1473
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);

1474
/*
1475 1476 1477 1478 1479 1480
 * 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
1481
 */
1482 1483
unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot,
				      gfn_t gfn, bool *writable)
1484
{
1485 1486 1487
	unsigned long hva = __gfn_to_hva_many(slot, gfn, NULL, false);

	if (!kvm_is_error_hva(hva) && writable)
1488 1489
		*writable = !memslot_is_readonly(slot);

1490
	return hva;
1491 1492
}

1493 1494 1495 1496 1497 1498 1499
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);
}

1500 1501 1502 1503 1504 1505 1506
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);
}

1507 1508
static inline int check_user_page_hwpoison(unsigned long addr)
{
L
Lorenzo Stoakes 已提交
1509
	int rc, flags = FOLL_HWPOISON | FOLL_WRITE;
1510

L
Lorenzo Stoakes 已提交
1511
	rc = get_user_pages(addr, 1, flags, NULL, NULL);
1512 1513 1514
	return rc == -EHWPOISON;
}

X
Xiao Guangrong 已提交
1515
/*
1516 1517
 * 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
1518
 * only part that runs if we can in atomic context.
X
Xiao Guangrong 已提交
1519
 */
1520 1521
static bool hva_to_pfn_fast(unsigned long addr, bool write_fault,
			    bool *writable, kvm_pfn_t *pfn)
A
Avi Kivity 已提交
1522
{
1523
	struct page *page[1];
X
Xiao Guangrong 已提交
1524
	int npages;
A
Avi Kivity 已提交
1525

1526 1527 1528 1529 1530 1531 1532
	/*
	 * 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;
1533

X
Xiao Guangrong 已提交
1534 1535 1536
	npages = __get_user_pages_fast(addr, 1, 1, page);
	if (npages == 1) {
		*pfn = page_to_pfn(page[0]);
1537

X
Xiao Guangrong 已提交
1538 1539 1540 1541
		if (writable)
			*writable = true;
		return true;
	}
1542

X
Xiao Guangrong 已提交
1543 1544
	return false;
}
1545

X
Xiao Guangrong 已提交
1546 1547 1548 1549 1550
/*
 * 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
1551
			   bool *writable, kvm_pfn_t *pfn)
X
Xiao Guangrong 已提交
1552
{
1553 1554
	unsigned int flags = FOLL_HWPOISON;
	struct page *page;
X
Xiao Guangrong 已提交
1555
	int npages = 0;
1556

X
Xiao Guangrong 已提交
1557 1558 1559 1560 1561
	might_sleep();

	if (writable)
		*writable = write_fault;

1562 1563 1564 1565
	if (write_fault)
		flags |= FOLL_WRITE;
	if (async)
		flags |= FOLL_NOWAIT;
1566

1567
	npages = get_user_pages_unlocked(addr, 1, &page, flags);
X
Xiao Guangrong 已提交
1568 1569 1570 1571
	if (npages != 1)
		return npages;

	/* map read fault as writable if possible */
1572
	if (unlikely(!write_fault) && writable) {
1573
		struct page *wpage;
X
Xiao Guangrong 已提交
1574

1575
		if (__get_user_pages_fast(addr, 1, 1, &wpage) == 1) {
X
Xiao Guangrong 已提交
1576
			*writable = true;
1577 1578
			put_page(page);
			page = wpage;
1579
		}
1580
	}
1581
	*pfn = page_to_pfn(page);
X
Xiao Guangrong 已提交
1582 1583
	return npages;
}
I
Izik Eidus 已提交
1584

X
Xiao Guangrong 已提交
1585 1586 1587 1588
static bool vma_is_valid(struct vm_area_struct *vma, bool write_fault)
{
	if (unlikely(!(vma->vm_flags & VM_READ)))
		return false;
1589

X
Xiao Guangrong 已提交
1590 1591
	if (write_fault && (unlikely(!(vma->vm_flags & VM_WRITE))))
		return false;
1592

X
Xiao Guangrong 已提交
1593 1594
	return true;
}
1595

1596 1597
static int hva_to_pfn_remapped(struct vm_area_struct *vma,
			       unsigned long addr, bool *async,
1598 1599
			       bool write_fault, bool *writable,
			       kvm_pfn_t *p_pfn)
1600
{
1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624
	unsigned long pfn;
	int r;

	r = follow_pfn(vma, addr, &pfn);
	if (r) {
		/*
		 * get_user_pages fails for VM_IO and VM_PFNMAP vmas and does
		 * not call the fault handler, so do it here.
		 */
		bool unlocked = false;
		r = fixup_user_fault(current, current->mm, addr,
				     (write_fault ? FAULT_FLAG_WRITE : 0),
				     &unlocked);
		if (unlocked)
			return -EAGAIN;
		if (r)
			return r;

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

	}

1625 1626
	if (writable)
		*writable = true;
1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641

	/*
	 * 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;
1642 1643 1644
	return 0;
}

1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658
/*
 * 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
1659
static kvm_pfn_t hva_to_pfn(unsigned long addr, bool atomic, bool *async,
X
Xiao Guangrong 已提交
1660 1661 1662
			bool write_fault, bool *writable)
{
	struct vm_area_struct *vma;
D
Dan Williams 已提交
1663
	kvm_pfn_t pfn = 0;
1664
	int npages, r;
1665

X
Xiao Guangrong 已提交
1666 1667
	/* we can do it either atomically or asynchronously, not both */
	BUG_ON(atomic && async);
1668

1669
	if (hva_to_pfn_fast(addr, write_fault, writable, &pfn))
X
Xiao Guangrong 已提交
1670 1671 1672 1673 1674 1675 1676 1677
		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;
1678

X
Xiao Guangrong 已提交
1679 1680 1681 1682 1683 1684 1685
	down_read(&current->mm->mmap_sem);
	if (npages == -EHWPOISON ||
	      (!async && check_user_page_hwpoison(addr))) {
		pfn = KVM_PFN_ERR_HWPOISON;
		goto exit;
	}

1686
retry:
X
Xiao Guangrong 已提交
1687 1688 1689 1690
	vma = find_vma_intersection(current->mm, addr, addr + 1);

	if (vma == NULL)
		pfn = KVM_PFN_ERR_FAULT;
1691
	else if (vma->vm_flags & (VM_IO | VM_PFNMAP)) {
1692
		r = hva_to_pfn_remapped(vma, addr, async, write_fault, writable, &pfn);
1693 1694
		if (r == -EAGAIN)
			goto retry;
1695 1696
		if (r < 0)
			pfn = KVM_PFN_ERR_FAULT;
X
Xiao Guangrong 已提交
1697
	} else {
X
Xiao Guangrong 已提交
1698
		if (async && vma_is_valid(vma, write_fault))
X
Xiao Guangrong 已提交
1699 1700 1701 1702 1703
			*async = true;
		pfn = KVM_PFN_ERR_FAULT;
	}
exit:
	up_read(&current->mm->mmap_sem);
1704
	return pfn;
1705 1706
}

D
Dan Williams 已提交
1707 1708 1709
kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
			       bool atomic, bool *async, bool write_fault,
			       bool *writable)
1710
{
X
Xiao Guangrong 已提交
1711 1712
	unsigned long addr = __gfn_to_hva_many(slot, gfn, NULL, write_fault);

1713 1714 1715
	if (addr == KVM_HVA_ERR_RO_BAD) {
		if (writable)
			*writable = false;
X
Xiao Guangrong 已提交
1716
		return KVM_PFN_ERR_RO_FAULT;
1717
	}
X
Xiao Guangrong 已提交
1718

1719 1720 1721
	if (kvm_is_error_hva(addr)) {
		if (writable)
			*writable = false;
1722
		return KVM_PFN_NOSLOT;
1723
	}
X
Xiao Guangrong 已提交
1724 1725 1726 1727 1728 1729 1730 1731 1732

	/* 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);
1733
}
1734
EXPORT_SYMBOL_GPL(__gfn_to_pfn_memslot);
1735

D
Dan Williams 已提交
1736
kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1737 1738
		      bool *writable)
{
P
Paolo Bonzini 已提交
1739 1740
	return __gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn, false, NULL,
				    write_fault, writable);
1741 1742 1743
}
EXPORT_SYMBOL_GPL(gfn_to_pfn_prot);

D
Dan Williams 已提交
1744
kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1745
{
X
Xiao Guangrong 已提交
1746
	return __gfn_to_pfn_memslot(slot, gfn, false, NULL, true, NULL);
1747
}
P
Paolo Bonzini 已提交
1748
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot);
1749

D
Dan Williams 已提交
1750
kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn)
1751
{
X
Xiao Guangrong 已提交
1752
	return __gfn_to_pfn_memslot(slot, gfn, true, NULL, true, NULL);
1753
}
1754
EXPORT_SYMBOL_GPL(gfn_to_pfn_memslot_atomic);
1755

D
Dan Williams 已提交
1756
kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1757 1758 1759 1760 1761
{
	return gfn_to_pfn_memslot_atomic(gfn_to_memslot(kvm, gfn), gfn);
}
EXPORT_SYMBOL_GPL(gfn_to_pfn_atomic);

D
Dan Williams 已提交
1762
kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn)
1763 1764 1765 1766 1767
{
	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
1768
kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
P
Paolo Bonzini 已提交
1769 1770 1771 1772 1773
{
	return gfn_to_pfn_memslot(gfn_to_memslot(kvm, gfn), gfn);
}
EXPORT_SYMBOL_GPL(gfn_to_pfn);

D
Dan Williams 已提交
1774
kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn)
1775 1776 1777 1778 1779
{
	return gfn_to_pfn_memslot(kvm_vcpu_gfn_to_memslot(vcpu, gfn), gfn);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_gfn_to_pfn);

1780 1781
int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
			    struct page **pages, int nr_pages)
1782 1783
{
	unsigned long addr;
1784
	gfn_t entry = 0;
1785

1786
	addr = gfn_to_hva_many(slot, gfn, &entry);
1787 1788 1789 1790 1791 1792 1793 1794 1795 1796
	if (kvm_is_error_hva(addr))
		return -1;

	if (entry < nr_pages)
		return 0;

	return __get_user_pages_fast(addr, nr_pages, 1, pages);
}
EXPORT_SYMBOL_GPL(gfn_to_page_many_atomic);

D
Dan Williams 已提交
1797
static struct page *kvm_pfn_to_page(kvm_pfn_t pfn)
X
Xiao Guangrong 已提交
1798
{
1799
	if (is_error_noslot_pfn(pfn))
1800
		return KVM_ERR_PTR_BAD_PAGE;
X
Xiao Guangrong 已提交
1801

1802
	if (kvm_is_reserved_pfn(pfn)) {
1803
		WARN_ON(1);
1804
		return KVM_ERR_PTR_BAD_PAGE;
1805
	}
X
Xiao Guangrong 已提交
1806 1807 1808 1809

	return pfn_to_page(pfn);
}

1810 1811
struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
{
D
Dan Williams 已提交
1812
	kvm_pfn_t pfn;
1813 1814 1815

	pfn = gfn_to_pfn(kvm, gfn);

X
Xiao Guangrong 已提交
1816
	return kvm_pfn_to_page(pfn);
A
Avi Kivity 已提交
1817 1818 1819
}
EXPORT_SYMBOL_GPL(gfn_to_page);

1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836
static int __kvm_map_gfn(struct kvm_memory_slot *slot, gfn_t gfn,
			 struct kvm_host_map *map)
{
	kvm_pfn_t pfn;
	void *hva = NULL;
	struct page *page = KVM_UNMAPPED_PAGE;

	if (!map)
		return -EINVAL;

	pfn = gfn_to_pfn_memslot(slot, gfn);
	if (is_error_noslot_pfn(pfn))
		return -EINVAL;

	if (pfn_valid(pfn)) {
		page = pfn_to_page(pfn);
		hva = kmap(page);
P
Paolo Bonzini 已提交
1837
#ifdef CONFIG_HAS_IOMEM
1838 1839
	} else {
		hva = memremap(pfn_to_hpa(pfn), PAGE_SIZE, MEMREMAP_WB);
P
Paolo Bonzini 已提交
1840
#endif
1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868
	}

	if (!hva)
		return -EFAULT;

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

	return 0;
}

int kvm_vcpu_map(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map)
{
	return __kvm_map_gfn(kvm_vcpu_gfn_to_memslot(vcpu, gfn), gfn, map);
}
EXPORT_SYMBOL_GPL(kvm_vcpu_map);

void kvm_vcpu_unmap(struct kvm_vcpu *vcpu, struct kvm_host_map *map,
		    bool dirty)
{
	if (!map)
		return;

	if (!map->hva)
		return;

1869
	if (map->page != KVM_UNMAPPED_PAGE)
1870
		kunmap(map->page);
1871
#ifdef CONFIG_HAS_IOMEM
1872 1873
	else
		memunmap(map->hva);
1874
#endif
1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887

	if (dirty) {
		kvm_vcpu_mark_page_dirty(vcpu, map->gfn);
		kvm_release_pfn_dirty(map->pfn);
	} else {
		kvm_release_pfn_clean(map->pfn);
	}

	map->hva = NULL;
	map->page = NULL;
}
EXPORT_SYMBOL_GPL(kvm_vcpu_unmap);

1888 1889
struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn)
{
D
Dan Williams 已提交
1890
	kvm_pfn_t pfn;
1891 1892 1893 1894 1895 1896 1897

	pfn = kvm_vcpu_gfn_to_pfn(vcpu, gfn);

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

1898 1899
void kvm_release_page_clean(struct page *page)
{
1900 1901
	WARN_ON(is_error_page(page));

1902
	kvm_release_pfn_clean(page_to_pfn(page));
1903 1904 1905
}
EXPORT_SYMBOL_GPL(kvm_release_page_clean);

D
Dan Williams 已提交
1906
void kvm_release_pfn_clean(kvm_pfn_t pfn)
1907
{
1908
	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn))
1909
		put_page(pfn_to_page(pfn));
1910 1911 1912
}
EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);

1913
void kvm_release_page_dirty(struct page *page)
1914
{
X
Xiao Guangrong 已提交
1915 1916
	WARN_ON(is_error_page(page));

1917 1918 1919 1920
	kvm_release_pfn_dirty(page_to_pfn(page));
}
EXPORT_SYMBOL_GPL(kvm_release_page_dirty);

1921
void kvm_release_pfn_dirty(kvm_pfn_t pfn)
1922 1923 1924 1925
{
	kvm_set_pfn_dirty(pfn);
	kvm_release_pfn_clean(pfn);
}
1926
EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
1927

D
Dan Williams 已提交
1928
void kvm_set_pfn_dirty(kvm_pfn_t pfn)
1929
{
1930 1931
	if (!kvm_is_reserved_pfn(pfn) && !kvm_is_zone_device_pfn(pfn))
		SetPageDirty(pfn_to_page(pfn));
1932
}
1933 1934
EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);

D
Dan Williams 已提交
1935
void kvm_set_pfn_accessed(kvm_pfn_t pfn)
1936
{
1937
	if (!kvm_is_reserved_pfn(pfn) && !kvm_is_zone_device_pfn(pfn))
1938
		mark_page_accessed(pfn_to_page(pfn));
1939 1940 1941
}
EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);

D
Dan Williams 已提交
1942
void kvm_get_pfn(kvm_pfn_t pfn)
1943
{
1944
	if (!kvm_is_reserved_pfn(pfn))
1945
		get_page(pfn_to_page(pfn));
1946 1947
}
EXPORT_SYMBOL_GPL(kvm_get_pfn);
1948

1949 1950 1951 1952 1953 1954 1955 1956
static int next_segment(unsigned long len, int offset)
{
	if (len > PAGE_SIZE - offset)
		return PAGE_SIZE - offset;
	else
		return len;
}

1957 1958
static int __kvm_read_guest_page(struct kvm_memory_slot *slot, gfn_t gfn,
				 void *data, int offset, int len)
1959
{
1960 1961
	int r;
	unsigned long addr;
1962

1963
	addr = gfn_to_hva_memslot_prot(slot, gfn, NULL);
1964 1965
	if (kvm_is_error_hva(addr))
		return -EFAULT;
1966
	r = __copy_from_user(data, (void __user *)addr + offset, len);
1967
	if (r)
1968 1969 1970
		return -EFAULT;
	return 0;
}
1971 1972 1973 1974 1975 1976 1977 1978

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);
}
1979 1980
EXPORT_SYMBOL_GPL(kvm_read_guest_page);

1981 1982 1983 1984 1985 1986 1987 1988 1989
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);

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
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);

2010
int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data, unsigned long len)
2011 2012
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
2013
	int seg;
2014
	int offset = offset_in_page(gpa);
2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028
	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);
2029

2030 2031 2032 2033 2034 2035 2036
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);
2037 2038
	if (kvm_is_error_hva(addr))
		return -EFAULT;
2039
	pagefault_disable();
2040
	r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
2041
	pagefault_enable();
2042 2043 2044 2045 2046
	if (r)
		return -EFAULT;
	return 0;
}

2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059
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)
2060
{
2061 2062
	int r;
	unsigned long addr;
2063

2064
	addr = gfn_to_hva_memslot(memslot, gfn);
2065 2066
	if (kvm_is_error_hva(addr))
		return -EFAULT;
2067
	r = __copy_to_user((void __user *)addr + offset, data, len);
2068
	if (r)
2069
		return -EFAULT;
2070
	mark_page_dirty_in_slot(memslot, gfn);
2071 2072
	return 0;
}
2073 2074 2075 2076 2077 2078 2079 2080

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);
}
2081 2082
EXPORT_SYMBOL_GPL(kvm_write_guest_page);

2083 2084 2085 2086 2087 2088 2089 2090 2091
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);

2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110
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;
}
2111
EXPORT_SYMBOL_GPL(kvm_write_guest);
2112

2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133
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);

2134 2135 2136
static int __kvm_gfn_to_hva_cache_init(struct kvm_memslots *slots,
				       struct gfn_to_hva_cache *ghc,
				       gpa_t gpa, unsigned long len)
2137 2138
{
	int offset = offset_in_page(gpa);
2139 2140 2141 2142
	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;
2143

2144
	/* Update ghc->generation before performing any error checks. */
2145
	ghc->generation = slots->generation;
2146 2147 2148 2149 2150

	if (start_gfn > end_gfn) {
		ghc->hva = KVM_HVA_ERR_BAD;
		return -EINVAL;
	}
2151 2152 2153 2154 2155

	/*
	 * If the requested region crosses two memslots, we still
	 * verify that the entire region is valid here.
	 */
2156
	for ( ; start_gfn <= end_gfn; start_gfn += nr_pages_avail) {
2157 2158 2159 2160
		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))
2161
			return -EFAULT;
2162 2163 2164
	}

	/* Use the slow path for cross page reads and writes. */
2165
	if (nr_pages_needed == 1)
2166
		ghc->hva += offset;
2167
	else
2168
		ghc->memslot = NULL;
2169

2170 2171 2172
	ghc->gpa = gpa;
	ghc->len = len;
	return 0;
2173
}
2174

2175
int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
2176 2177
			      gpa_t gpa, unsigned long len)
{
2178
	struct kvm_memslots *slots = kvm_memslots(kvm);
2179 2180
	return __kvm_gfn_to_hva_cache_init(slots, ghc, gpa, len);
}
2181
EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init);
2182

2183
int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
2184 2185
				  void *data, unsigned int offset,
				  unsigned long len)
2186
{
2187
	struct kvm_memslots *slots = kvm_memslots(kvm);
2188
	int r;
2189
	gpa_t gpa = ghc->gpa + offset;
2190

2191
	BUG_ON(len + offset > ghc->len);
2192

2193 2194 2195 2196
	if (slots->generation != ghc->generation) {
		if (__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len))
			return -EFAULT;
	}
2197

2198 2199 2200
	if (kvm_is_error_hva(ghc->hva))
		return -EFAULT;

2201 2202 2203
	if (unlikely(!ghc->memslot))
		return kvm_write_guest(kvm, gpa, data, len);

2204
	r = __copy_to_user((void __user *)ghc->hva + offset, data, len);
2205 2206
	if (r)
		return -EFAULT;
2207
	mark_page_dirty_in_slot(ghc->memslot, gpa >> PAGE_SHIFT);
2208 2209 2210

	return 0;
}
2211
EXPORT_SYMBOL_GPL(kvm_write_guest_offset_cached);
2212

2213 2214
int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
2215
{
2216
	return kvm_write_guest_offset_cached(kvm, ghc, data, 0, len);
2217
}
2218
EXPORT_SYMBOL_GPL(kvm_write_guest_cached);
2219

2220 2221
int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
			   void *data, unsigned long len)
2222
{
2223
	struct kvm_memslots *slots = kvm_memslots(kvm);
2224 2225
	int r;

2226 2227
	BUG_ON(len > ghc->len);

2228 2229 2230 2231
	if (slots->generation != ghc->generation) {
		if (__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len))
			return -EFAULT;
	}
2232

2233 2234 2235
	if (kvm_is_error_hva(ghc->hva))
		return -EFAULT;

2236 2237 2238
	if (unlikely(!ghc->memslot))
		return kvm_read_guest(kvm, ghc->gpa, data, len);

2239 2240 2241 2242 2243 2244
	r = __copy_from_user(data, (void __user *)ghc->hva, len);
	if (r)
		return -EFAULT;

	return 0;
}
2245
EXPORT_SYMBOL_GPL(kvm_read_guest_cached);
2246

2247 2248
int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
{
2249 2250 2251
	const void *zero_page = (const void *) __va(page_to_phys(ZERO_PAGE(0)));

	return kvm_write_guest_page(kvm, gfn, zero_page, offset, len);
2252 2253 2254 2255 2256 2257 2258 2259 2260 2261
}
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;

2262
	while ((seg = next_segment(len, offset)) != 0) {
2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273
		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);

2274
static void mark_page_dirty_in_slot(struct kvm_memory_slot *memslot,
2275
				    gfn_t gfn)
A
Avi Kivity 已提交
2276
{
R
Rusty Russell 已提交
2277 2278
	if (memslot && memslot->dirty_bitmap) {
		unsigned long rel_gfn = gfn - memslot->base_gfn;
A
Avi Kivity 已提交
2279

2280
		set_bit_le(rel_gfn, memslot->dirty_bitmap);
A
Avi Kivity 已提交
2281 2282 2283
	}
}

2284 2285 2286 2287 2288
void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
{
	struct kvm_memory_slot *memslot;

	memslot = gfn_to_memslot(kvm, gfn);
2289
	mark_page_dirty_in_slot(memslot, gfn);
2290
}
2291
EXPORT_SYMBOL_GPL(mark_page_dirty);
2292

2293 2294 2295 2296 2297 2298 2299 2300 2301
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);

2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324
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
2325 2326
static void grow_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2327
	unsigned int old, val, grow, grow_start;
W
Wanpeng Li 已提交
2328

2329
	old = val = vcpu->halt_poll_ns;
2330
	grow_start = READ_ONCE(halt_poll_ns_grow_start);
2331
	grow = READ_ONCE(halt_poll_ns_grow);
2332 2333 2334
	if (!grow)
		goto out;

2335 2336 2337
	val *= grow;
	if (val < grow_start)
		val = grow_start;
W
Wanpeng Li 已提交
2338

2339 2340 2341
	if (val > halt_poll_ns)
		val = halt_poll_ns;

W
Wanpeng Li 已提交
2342
	vcpu->halt_poll_ns = val;
2343
out:
2344
	trace_kvm_halt_poll_ns_grow(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2345 2346 2347 2348
}

static void shrink_halt_poll_ns(struct kvm_vcpu *vcpu)
{
2349
	unsigned int old, val, shrink;
W
Wanpeng Li 已提交
2350

2351
	old = val = vcpu->halt_poll_ns;
2352 2353
	shrink = READ_ONCE(halt_poll_ns_shrink);
	if (shrink == 0)
W
Wanpeng Li 已提交
2354 2355
		val = 0;
	else
2356
		val /= shrink;
W
Wanpeng Li 已提交
2357 2358

	vcpu->halt_poll_ns = val;
2359
	trace_kvm_halt_poll_ns_shrink(vcpu->vcpu_id, val, old);
W
Wanpeng Li 已提交
2360 2361
}

2362 2363
static int kvm_vcpu_check_block(struct kvm_vcpu *vcpu)
{
2364 2365 2366
	int ret = -EINTR;
	int idx = srcu_read_lock(&vcpu->kvm->srcu);

2367 2368
	if (kvm_arch_vcpu_runnable(vcpu)) {
		kvm_make_request(KVM_REQ_UNHALT, vcpu);
2369
		goto out;
2370 2371
	}
	if (kvm_cpu_has_pending_timer(vcpu))
2372
		goto out;
2373
	if (signal_pending(current))
2374
		goto out;
2375

2376 2377 2378 2379
	ret = 0;
out:
	srcu_read_unlock(&vcpu->kvm->srcu, idx);
	return ret;
2380 2381
}

E
Eddie Dong 已提交
2382 2383 2384
/*
 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
 */
2385
void kvm_vcpu_block(struct kvm_vcpu *vcpu)
2386
{
2387
	ktime_t start, cur;
2388
	DECLARE_SWAITQUEUE(wait);
2389
	bool waited = false;
W
Wanpeng Li 已提交
2390
	u64 block_ns;
2391

2392 2393
	kvm_arch_vcpu_blocking(vcpu);

2394
	start = cur = ktime_get();
2395
	if (vcpu->halt_poll_ns && !kvm_arch_no_poll(vcpu)) {
W
Wanpeng Li 已提交
2396
		ktime_t stop = ktime_add_ns(ktime_get(), vcpu->halt_poll_ns);
2397

2398
		++vcpu->stat.halt_attempted_poll;
2399 2400 2401 2402 2403 2404 2405
		do {
			/*
			 * This sets KVM_REQ_UNHALT if an interrupt
			 * arrives.
			 */
			if (kvm_vcpu_check_block(vcpu) < 0) {
				++vcpu->stat.halt_successful_poll;
2406 2407
				if (!vcpu_valid_wakeup(vcpu))
					++vcpu->stat.halt_poll_invalid;
2408 2409 2410 2411 2412
				goto out;
			}
			cur = ktime_get();
		} while (single_task_running() && ktime_before(cur, stop));
	}
2413 2414

	for (;;) {
2415
		prepare_to_swait_exclusive(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
2416

2417
		if (kvm_vcpu_check_block(vcpu) < 0)
2418 2419
			break;

2420
		waited = true;
E
Eddie Dong 已提交
2421 2422
		schedule();
	}
2423

2424
	finish_swait(&vcpu->wq, &wait);
2425 2426
	cur = ktime_get();
out:
2427
	kvm_arch_vcpu_unblocking(vcpu);
W
Wanpeng Li 已提交
2428 2429
	block_ns = ktime_to_ns(cur) - ktime_to_ns(start);

2430 2431
	if (!kvm_arch_no_poll(vcpu)) {
		if (!vcpu_valid_wakeup(vcpu)) {
W
Wanpeng Li 已提交
2432
			shrink_halt_poll_ns(vcpu);
2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446
		} else if (halt_poll_ns) {
			if (block_ns <= vcpu->halt_poll_ns)
				;
			/* we had a long block, shrink polling */
			else if (vcpu->halt_poll_ns && block_ns > halt_poll_ns)
				shrink_halt_poll_ns(vcpu);
			/* we had a short halt and our poll time is too small */
			else if (vcpu->halt_poll_ns < halt_poll_ns &&
				block_ns < halt_poll_ns)
				grow_halt_poll_ns(vcpu);
		} else {
			vcpu->halt_poll_ns = 0;
		}
	}
W
Wanpeng Li 已提交
2447

2448 2449
	trace_kvm_vcpu_wakeup(block_ns, waited, vcpu_valid_wakeup(vcpu));
	kvm_arch_vcpu_block_finish(vcpu);
E
Eddie Dong 已提交
2450
}
2451
EXPORT_SYMBOL_GPL(kvm_vcpu_block);
E
Eddie Dong 已提交
2452

2453
bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu)
2454
{
2455
	struct swait_queue_head *wqp;
2456 2457

	wqp = kvm_arch_vcpu_wq(vcpu);
2458
	if (swq_has_sleeper(wqp)) {
2459
		swake_up_one(wqp);
2460
		WRITE_ONCE(vcpu->ready, true);
2461
		++vcpu->stat.halt_wakeup;
2462
		return true;
2463 2464
	}

2465
	return false;
2466 2467 2468
}
EXPORT_SYMBOL_GPL(kvm_vcpu_wake_up);

2469
#ifndef CONFIG_S390
2470 2471 2472 2473 2474 2475 2476 2477
/*
 * 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;

2478 2479 2480
	if (kvm_vcpu_wake_up(vcpu))
		return;

2481 2482 2483 2484 2485 2486
	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();
}
2487
EXPORT_SYMBOL_GPL(kvm_vcpu_kick);
2488
#endif /* !CONFIG_S390 */
2489

2490
int kvm_vcpu_yield_to(struct kvm_vcpu *target)
2491 2492 2493
{
	struct pid *pid;
	struct task_struct *task = NULL;
2494
	int ret = 0;
2495 2496 2497 2498

	rcu_read_lock();
	pid = rcu_dereference(target->pid);
	if (pid)
2499
		task = get_pid_task(pid, PIDTYPE_PID);
2500 2501
	rcu_read_unlock();
	if (!task)
2502 2503
		return ret;
	ret = yield_to(task, 1);
2504
	put_task_struct(task);
2505 2506

	return ret;
2507 2508 2509
}
EXPORT_SYMBOL_GPL(kvm_vcpu_yield_to);

2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531
/*
 * Helper that checks whether a VCPU is eligible for directed yield.
 * Most eligible candidate to yield is decided by following heuristics:
 *
 *  (a) VCPU which has not done pl-exit or cpu relax intercepted recently
 *  (preempted lock holder), indicated by @in_spin_loop.
 *  Set at the beiginning and cleared at the end of interception/PLE handler.
 *
 *  (b) VCPU which has done pl-exit/ cpu relax intercepted but did not get
 *  chance last time (mostly it has become eligible now since we have probably
 *  yielded to lockholder in last iteration. This is done by toggling
 *  @dy_eligible each time a VCPU checked for eligibility.)
 *
 *  Yielding to a recently pl-exited/cpu relax intercepted VCPU before yielding
 *  to preempted lock-holder could result in wrong VCPU selection and CPU
 *  burning. Giving priority for a potential lock-holder increases lock
 *  progress.
 *
 *  Since algorithm is based on heuristics, accessing another VCPU data without
 *  locking does not harm. It may result in trying to yield to  same VCPU, fail
 *  and continue with next VCPU and so on.
 */
2532
static bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
2533
{
2534
#ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
2535 2536 2537
	bool eligible;

	eligible = !vcpu->spin_loop.in_spin_loop ||
2538
		    vcpu->spin_loop.dy_eligible;
2539 2540 2541 2542 2543

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

	return eligible;
2544 2545
#else
	return true;
2546
#endif
2547
}
2548

2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571
/*
 * 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;
}

2572
void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
Z
Zhai, Edwin 已提交
2573
{
2574 2575 2576 2577
	struct kvm *kvm = me->kvm;
	struct kvm_vcpu *vcpu;
	int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
	int yielded = 0;
2578
	int try = 3;
2579 2580
	int pass;
	int i;
Z
Zhai, Edwin 已提交
2581

2582
	kvm_vcpu_set_in_spin_loop(me, true);
2583 2584 2585 2586 2587 2588 2589
	/*
	 * 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.
	 */
2590
	for (pass = 0; pass < 2 && !yielded && try; pass++) {
2591
		kvm_for_each_vcpu(i, vcpu, kvm) {
2592
			if (!pass && i <= last_boosted_vcpu) {
2593 2594 2595 2596
				i = last_boosted_vcpu;
				continue;
			} else if (pass && i > last_boosted_vcpu)
				break;
2597
			if (!READ_ONCE(vcpu->ready))
2598
				continue;
2599 2600
			if (vcpu == me)
				continue;
2601
			if (swait_active(&vcpu->wq) && !vcpu_dy_runnable(vcpu))
2602
				continue;
2603 2604
			if (READ_ONCE(vcpu->preempted) && yield_to_kernel_mode &&
				!kvm_arch_vcpu_in_kernel(vcpu))
2605
				continue;
2606 2607
			if (!kvm_vcpu_eligible_for_directed_yield(vcpu))
				continue;
2608 2609 2610

			yielded = kvm_vcpu_yield_to(vcpu);
			if (yielded > 0) {
2611 2612
				kvm->last_boosted_vcpu = i;
				break;
2613 2614 2615 2616
			} else if (yielded < 0) {
				try--;
				if (!try)
					break;
2617 2618 2619
			}
		}
	}
2620
	kvm_vcpu_set_in_spin_loop(me, false);
2621 2622 2623

	/* Ensure vcpu is not eligible during next spinloop */
	kvm_vcpu_set_dy_eligible(me, false);
Z
Zhai, Edwin 已提交
2624 2625 2626
}
EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin);

2627
static vm_fault_t kvm_vcpu_fault(struct vm_fault *vmf)
2628
{
2629
	struct kvm_vcpu *vcpu = vmf->vma->vm_file->private_data;
2630 2631
	struct page *page;

2632
	if (vmf->pgoff == 0)
2633
		page = virt_to_page(vcpu->run);
A
Avi Kivity 已提交
2634
#ifdef CONFIG_X86
2635
	else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
2636
		page = virt_to_page(vcpu->arch.pio_data);
2637
#endif
2638
#ifdef CONFIG_KVM_MMIO
2639 2640
	else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
		page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
A
Avi Kivity 已提交
2641
#endif
2642
	else
2643
		return kvm_arch_vcpu_fault(vcpu, vmf);
2644
	get_page(page);
2645 2646
	vmf->page = page;
	return 0;
2647 2648
}

2649
static const struct vm_operations_struct kvm_vcpu_vm_ops = {
2650
	.fault = kvm_vcpu_fault,
2651 2652 2653 2654 2655 2656 2657 2658
};

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
2659 2660 2661 2662
static int kvm_vcpu_release(struct inode *inode, struct file *filp)
{
	struct kvm_vcpu *vcpu = filp->private_data;

2663
	debugfs_remove_recursive(vcpu->debugfs_dentry);
A
Al Viro 已提交
2664
	kvm_put_kvm(vcpu->kvm);
A
Avi Kivity 已提交
2665 2666 2667
	return 0;
}

2668
static struct file_operations kvm_vcpu_fops = {
A
Avi Kivity 已提交
2669 2670
	.release        = kvm_vcpu_release,
	.unlocked_ioctl = kvm_vcpu_ioctl,
2671
	.mmap           = kvm_vcpu_mmap,
2672
	.llseek		= noop_llseek,
2673
	KVM_COMPAT(kvm_vcpu_compat_ioctl),
A
Avi Kivity 已提交
2674 2675 2676 2677 2678 2679 2680
};

/*
 * Allocates an inode for the vcpu.
 */
static int create_vcpu_fd(struct kvm_vcpu *vcpu)
{
2681 2682 2683 2684
	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
2685 2686
}

2687
static void kvm_create_vcpu_debugfs(struct kvm_vcpu *vcpu)
2688
{
2689
#ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
2690 2691 2692
	char dir_name[ITOA_MAX_LEN * 2];

	if (!debugfs_initialized())
2693
		return;
2694 2695 2696

	snprintf(dir_name, sizeof(dir_name), "vcpu%d", vcpu->vcpu_id);
	vcpu->debugfs_dentry = debugfs_create_dir(dir_name,
2697
						  vcpu->kvm->debugfs_dentry);
2698

2699
	kvm_arch_create_vcpu_debugfs(vcpu);
2700
#endif
2701 2702
}

2703 2704 2705
/*
 * Creates some virtual cpus.  Good luck creating more than one.
 */
2706
static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
2707 2708
{
	int r;
2709
	struct kvm_vcpu *vcpu;
2710
	struct page *page;
2711

G
Greg Kurz 已提交
2712
	if (id >= KVM_MAX_VCPU_ID)
2713 2714
		return -EINVAL;

2715 2716 2717 2718 2719 2720 2721 2722 2723
	mutex_lock(&kvm->lock);
	if (kvm->created_vcpus == KVM_MAX_VCPUS) {
		mutex_unlock(&kvm->lock);
		return -EINVAL;
	}

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

2724 2725 2726 2727
	r = kvm_arch_vcpu_precreate(kvm, id);
	if (r)
		goto vcpu_decrement;

2728 2729 2730
	vcpu = kmem_cache_zalloc(kvm_vcpu_cache, GFP_KERNEL);
	if (!vcpu) {
		r = -ENOMEM;
2731 2732
		goto vcpu_decrement;
	}
2733

2734
	BUILD_BUG_ON(sizeof(struct kvm_run) > PAGE_SIZE);
2735 2736 2737
	page = alloc_page(GFP_KERNEL | __GFP_ZERO);
	if (!page) {
		r = -ENOMEM;
2738
		goto vcpu_free;
2739 2740 2741 2742
	}
	vcpu->run = page_address(page);

	kvm_vcpu_init(vcpu, kvm, id);
2743 2744 2745

	r = kvm_arch_vcpu_create(vcpu);
	if (r)
2746
		goto vcpu_free_run_page;
2747

2748
	kvm_create_vcpu_debugfs(vcpu);
2749

S
Shaohua Li 已提交
2750
	mutex_lock(&kvm->lock);
2751 2752 2753 2754
	if (kvm_get_vcpu_by_id(kvm, id)) {
		r = -EEXIST;
		goto unlock_vcpu_destroy;
	}
2755

2756 2757
	vcpu->vcpu_idx = atomic_read(&kvm->online_vcpus);
	BUG_ON(kvm->vcpus[vcpu->vcpu_idx]);
2758

R
Rusty Russell 已提交
2759
	/* Now it's all set up, let userspace reach it */
A
Al Viro 已提交
2760
	kvm_get_kvm(kvm);
A
Avi Kivity 已提交
2761
	r = create_vcpu_fd(vcpu);
2762
	if (r < 0) {
2763
		kvm_put_kvm_no_destroy(kvm);
2764
		goto unlock_vcpu_destroy;
2765 2766
	}

2767
	kvm->vcpus[vcpu->vcpu_idx] = vcpu;
2768 2769 2770 2771 2772

	/*
	 * Pairs with smp_rmb() in kvm_get_vcpu.  Write kvm->vcpus
	 * before kvm->online_vcpu's incremented value.
	 */
2773 2774 2775 2776
	smp_wmb();
	atomic_inc(&kvm->online_vcpus);

	mutex_unlock(&kvm->lock);
2777
	kvm_arch_vcpu_postcreate(vcpu);
R
Rusty Russell 已提交
2778
	return r;
2779

2780
unlock_vcpu_destroy:
2781
	mutex_unlock(&kvm->lock);
2782
	debugfs_remove_recursive(vcpu->debugfs_dentry);
2783
	kvm_arch_vcpu_destroy(vcpu);
2784 2785
vcpu_free_run_page:
	free_page((unsigned long)vcpu->run);
2786 2787
vcpu_free:
	kmem_cache_free(kvm_vcpu_cache, vcpu);
2788 2789 2790 2791
vcpu_decrement:
	mutex_lock(&kvm->lock);
	kvm->created_vcpus--;
	mutex_unlock(&kvm->lock);
2792 2793 2794
	return r;
}

A
Avi Kivity 已提交
2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805
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
2806 2807
static long kvm_vcpu_ioctl(struct file *filp,
			   unsigned int ioctl, unsigned long arg)
A
Avi Kivity 已提交
2808
{
A
Avi Kivity 已提交
2809
	struct kvm_vcpu *vcpu = filp->private_data;
A
Al Viro 已提交
2810
	void __user *argp = (void __user *)arg;
2811
	int r;
2812 2813
	struct kvm_fpu *fpu = NULL;
	struct kvm_sregs *kvm_sregs = NULL;
A
Avi Kivity 已提交
2814

2815 2816
	if (vcpu->kvm->mm != current->mm)
		return -EIO;
2817

2818 2819 2820
	if (unlikely(_IOC_TYPE(ioctl) != KVMIO))
		return -EINVAL;

2821
	/*
2822 2823
	 * Some architectures have vcpu ioctls that are asynchronous to vcpu
	 * execution; mutex_lock() would break them.
2824
	 */
2825 2826
	r = kvm_arch_vcpu_async_ioctl(filp, ioctl, arg);
	if (r != -ENOIOCTLCMD)
2827
		return r;
2828

2829 2830
	if (mutex_lock_killable(&vcpu->mutex))
		return -EINTR;
A
Avi Kivity 已提交
2831
	switch (ioctl) {
2832 2833
	case KVM_RUN: {
		struct pid *oldpid;
2834 2835 2836
		r = -EINVAL;
		if (arg)
			goto out;
2837
		oldpid = rcu_access_pointer(vcpu->pid);
2838
		if (unlikely(oldpid != task_pid(current))) {
2839
			/* The thread running this VCPU changed. */
2840
			struct pid *newpid;
2841

2842 2843 2844 2845 2846
			r = kvm_arch_vcpu_run_pid_change(vcpu);
			if (r)
				break;

			newpid = get_task_pid(current, PIDTYPE_PID);
2847 2848 2849 2850 2851
			rcu_assign_pointer(vcpu->pid, newpid);
			if (oldpid)
				synchronize_rcu();
			put_pid(oldpid);
		}
2852
		r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
2853
		trace_kvm_userspace_exit(vcpu->run->exit_reason, r);
A
Avi Kivity 已提交
2854
		break;
2855
	}
A
Avi Kivity 已提交
2856
	case KVM_GET_REGS: {
2857
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
2858

2859
		r = -ENOMEM;
2860
		kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL_ACCOUNT);
2861
		if (!kvm_regs)
A
Avi Kivity 已提交
2862
			goto out;
2863 2864 2865
		r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
		if (r)
			goto out_free1;
A
Avi Kivity 已提交
2866
		r = -EFAULT;
2867 2868
		if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
			goto out_free1;
A
Avi Kivity 已提交
2869
		r = 0;
2870 2871
out_free1:
		kfree(kvm_regs);
A
Avi Kivity 已提交
2872 2873 2874
		break;
	}
	case KVM_SET_REGS: {
2875
		struct kvm_regs *kvm_regs;
A
Avi Kivity 已提交
2876

2877
		r = -ENOMEM;
2878 2879 2880
		kvm_regs = memdup_user(argp, sizeof(*kvm_regs));
		if (IS_ERR(kvm_regs)) {
			r = PTR_ERR(kvm_regs);
A
Avi Kivity 已提交
2881
			goto out;
2882
		}
2883 2884
		r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
		kfree(kvm_regs);
A
Avi Kivity 已提交
2885 2886 2887
		break;
	}
	case KVM_GET_SREGS: {
2888 2889
		kvm_sregs = kzalloc(sizeof(struct kvm_sregs),
				    GFP_KERNEL_ACCOUNT);
2890 2891 2892 2893
		r = -ENOMEM;
		if (!kvm_sregs)
			goto out;
		r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
2894 2895 2896
		if (r)
			goto out;
		r = -EFAULT;
2897
		if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
A
Avi Kivity 已提交
2898 2899 2900 2901 2902
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_SREGS: {
2903 2904 2905
		kvm_sregs = memdup_user(argp, sizeof(*kvm_sregs));
		if (IS_ERR(kvm_sregs)) {
			r = PTR_ERR(kvm_sregs);
G
Guo Chao 已提交
2906
			kvm_sregs = NULL;
A
Avi Kivity 已提交
2907
			goto out;
2908
		}
2909
		r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
A
Avi Kivity 已提交
2910 2911
		break;
	}
2912 2913 2914 2915 2916 2917 2918
	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;
2919
		if (copy_to_user(argp, &mp_state, sizeof(mp_state)))
2920 2921 2922 2923 2924 2925 2926 2927
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_MP_STATE: {
		struct kvm_mp_state mp_state;

		r = -EFAULT;
2928
		if (copy_from_user(&mp_state, argp, sizeof(mp_state)))
2929 2930 2931 2932
			goto out;
		r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
		break;
	}
A
Avi Kivity 已提交
2933 2934 2935 2936
	case KVM_TRANSLATE: {
		struct kvm_translation tr;

		r = -EFAULT;
2937
		if (copy_from_user(&tr, argp, sizeof(tr)))
A
Avi Kivity 已提交
2938
			goto out;
2939
		r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
A
Avi Kivity 已提交
2940 2941 2942
		if (r)
			goto out;
		r = -EFAULT;
2943
		if (copy_to_user(argp, &tr, sizeof(tr)))
A
Avi Kivity 已提交
2944 2945 2946 2947
			goto out;
		r = 0;
		break;
	}
J
Jan Kiszka 已提交
2948 2949
	case KVM_SET_GUEST_DEBUG: {
		struct kvm_guest_debug dbg;
A
Avi Kivity 已提交
2950 2951

		r = -EFAULT;
2952
		if (copy_from_user(&dbg, argp, sizeof(dbg)))
A
Avi Kivity 已提交
2953
			goto out;
J
Jan Kiszka 已提交
2954
		r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
A
Avi Kivity 已提交
2955 2956
		break;
	}
A
Avi Kivity 已提交
2957 2958 2959 2960 2961 2962 2963 2964 2965
	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,
2966
					   sizeof(kvm_sigmask)))
A
Avi Kivity 已提交
2967 2968
				goto out;
			r = -EINVAL;
2969
			if (kvm_sigmask.len != sizeof(sigset))
A
Avi Kivity 已提交
2970 2971 2972
				goto out;
			r = -EFAULT;
			if (copy_from_user(&sigset, sigmask_arg->sigset,
2973
					   sizeof(sigset)))
A
Avi Kivity 已提交
2974 2975 2976
				goto out;
			p = &sigset;
		}
2977
		r = kvm_vcpu_ioctl_set_sigmask(vcpu, p);
A
Avi Kivity 已提交
2978 2979
		break;
	}
A
Avi Kivity 已提交
2980
	case KVM_GET_FPU: {
2981
		fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL_ACCOUNT);
2982 2983 2984 2985
		r = -ENOMEM;
		if (!fpu)
			goto out;
		r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu);
A
Avi Kivity 已提交
2986 2987 2988
		if (r)
			goto out;
		r = -EFAULT;
2989
		if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
A
Avi Kivity 已提交
2990 2991 2992 2993 2994
			goto out;
		r = 0;
		break;
	}
	case KVM_SET_FPU: {
2995 2996 2997
		fpu = memdup_user(argp, sizeof(*fpu));
		if (IS_ERR(fpu)) {
			r = PTR_ERR(fpu);
G
Guo Chao 已提交
2998
			fpu = NULL;
A
Avi Kivity 已提交
2999
			goto out;
3000
		}
3001
		r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
A
Avi Kivity 已提交
3002 3003
		break;
	}
A
Avi Kivity 已提交
3004
	default:
3005
		r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
3006 3007
	}
out:
3008
	mutex_unlock(&vcpu->mutex);
3009 3010
	kfree(fpu);
	kfree(kvm_sregs);
A
Avi Kivity 已提交
3011 3012 3013
	return r;
}

3014
#ifdef CONFIG_KVM_COMPAT
3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033
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,
3034
					   sizeof(kvm_sigmask)))
3035 3036
				goto out;
			r = -EINVAL;
A
Al Viro 已提交
3037
			if (kvm_sigmask.len != sizeof(compat_sigset_t))
3038 3039
				goto out;
			r = -EFAULT;
A
Al Viro 已提交
3040
			if (get_compat_sigset(&sigset, (void *)sigmask_arg->sigset))
3041
				goto out;
3042 3043 3044
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
		} else
			r = kvm_vcpu_ioctl_set_sigmask(vcpu, NULL);
3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055
		break;
	}
	default:
		r = kvm_vcpu_ioctl(filp, ioctl, arg);
	}

out:
	return r;
}
#endif

3056 3057 3058 3059 3060 3061 3062 3063 3064 3065
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
3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086
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;

3087 3088 3089
	if (dev->kvm->mm != current->mm)
		return -EIO;

S
Scott Wood 已提交
3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109
	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;

3110 3111 3112 3113 3114 3115 3116
	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
3117 3118 3119 3120 3121 3122 3123
	kvm_put_kvm(kvm);
	return 0;
}

static const struct file_operations kvm_device_fops = {
	.unlocked_ioctl = kvm_device_ioctl,
	.release = kvm_device_release,
3124
	KVM_COMPAT(kvm_device_ioctl),
3125
	.mmap = kvm_device_mmap,
S
Scott Wood 已提交
3126 3127 3128 3129 3130 3131 3132 3133 3134 3135
};

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

	return filp->private_data;
}

3136
static const struct kvm_device_ops *kvm_device_ops_table[KVM_DEV_TYPE_MAX] = {
3137
#ifdef CONFIG_KVM_MPIC
3138 3139
	[KVM_DEV_TYPE_FSL_MPIC_20]	= &kvm_mpic_ops,
	[KVM_DEV_TYPE_FSL_MPIC_42]	= &kvm_mpic_ops,
3140
#endif
3141 3142
};

3143
int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type)
3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154
{
	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;
}

3155 3156 3157 3158 3159 3160
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
3161 3162 3163
static int kvm_ioctl_create_device(struct kvm *kvm,
				   struct kvm_create_device *cd)
{
3164
	const struct kvm_device_ops *ops = NULL;
S
Scott Wood 已提交
3165 3166
	struct kvm_device *dev;
	bool test = cd->flags & KVM_CREATE_DEVICE_TEST;
P
Paolo Bonzini 已提交
3167
	int type;
S
Scott Wood 已提交
3168 3169
	int ret;

3170 3171 3172
	if (cd->type >= ARRAY_SIZE(kvm_device_ops_table))
		return -ENODEV;

P
Paolo Bonzini 已提交
3173 3174
	type = array_index_nospec(cd->type, ARRAY_SIZE(kvm_device_ops_table));
	ops = kvm_device_ops_table[type];
3175
	if (ops == NULL)
S
Scott Wood 已提交
3176 3177 3178 3179 3180
		return -ENODEV;

	if (test)
		return 0;

3181
	dev = kzalloc(sizeof(*dev), GFP_KERNEL_ACCOUNT);
S
Scott Wood 已提交
3182 3183 3184 3185 3186 3187
	if (!dev)
		return -ENOMEM;

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

3188
	mutex_lock(&kvm->lock);
P
Paolo Bonzini 已提交
3189
	ret = ops->create(dev, type);
S
Scott Wood 已提交
3190
	if (ret < 0) {
3191
		mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
3192 3193 3194
		kfree(dev);
		return ret;
	}
3195 3196
	list_add(&dev->vm_node, &kvm->devices);
	mutex_unlock(&kvm->lock);
S
Scott Wood 已提交
3197

3198 3199 3200
	if (ops->init)
		ops->init(dev);

3201
	kvm_get_kvm(kvm);
3202
	ret = anon_inode_getfd(ops->name, &kvm_device_fops, dev, O_RDWR | O_CLOEXEC);
S
Scott Wood 已提交
3203
	if (ret < 0) {
3204
		kvm_put_kvm_no_destroy(kvm);
3205 3206 3207
		mutex_lock(&kvm->lock);
		list_del(&dev->vm_node);
		mutex_unlock(&kvm->lock);
3208
		ops->destroy(dev);
S
Scott Wood 已提交
3209 3210 3211 3212 3213 3214 3215
		return ret;
	}

	cd->fd = ret;
	return 0;
}

3216 3217 3218 3219 3220 3221 3222 3223 3224 3225
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
3226
#ifdef CONFIG_HAVE_KVM_IRQFD
3227
	case KVM_CAP_IRQFD:
3228 3229
	case KVM_CAP_IRQFD_RESAMPLE:
#endif
3230
	case KVM_CAP_IOEVENTFD_ANY_LENGTH:
3231
	case KVM_CAP_CHECK_EXTENSION_VM:
3232
	case KVM_CAP_ENABLE_CAP_VM:
3233
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
3234
	case KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2:
3235
#endif
3236
		return 1;
3237
#ifdef CONFIG_KVM_MMIO
3238 3239
	case KVM_CAP_COALESCED_MMIO:
		return KVM_COALESCED_MMIO_PAGE_OFFSET;
P
Peng Hao 已提交
3240 3241
	case KVM_CAP_COALESCED_PIO:
		return 1;
3242
#endif
3243 3244 3245
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_CAP_IRQ_ROUTING:
		return KVM_MAX_IRQ_ROUTES;
3246 3247 3248 3249
#endif
#if KVM_ADDRESS_SPACE_NUM > 1
	case KVM_CAP_MULTI_ADDRESS_SPACE:
		return KVM_ADDRESS_SPACE_NUM;
3250
#endif
3251 3252
	case KVM_CAP_NR_MEMSLOTS:
		return KVM_USER_MEM_SLOTS;
3253 3254 3255 3256 3257 3258
	default:
		break;
	}
	return kvm_vm_ioctl_check_extension(kvm, arg);
}

3259 3260 3261 3262 3263 3264 3265 3266 3267 3268
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) {
3269
#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
3270
	case KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2:
3271 3272 3273 3274 3275
		if (cap->flags || (cap->args[0] & ~1))
			return -EINVAL;
		kvm->manual_dirty_log_protect = cap->args[0];
		return 0;
#endif
3276 3277 3278 3279 3280
	default:
		return kvm_vm_ioctl_enable_cap(kvm, cap);
	}
}

A
Avi Kivity 已提交
3281 3282 3283 3284 3285
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;
3286
	int r;
A
Avi Kivity 已提交
3287

3288 3289
	if (kvm->mm != current->mm)
		return -EIO;
A
Avi Kivity 已提交
3290 3291 3292 3293
	switch (ioctl) {
	case KVM_CREATE_VCPU:
		r = kvm_vm_ioctl_create_vcpu(kvm, arg);
		break;
3294 3295 3296 3297 3298 3299 3300 3301 3302
	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;
	}
3303 3304 3305 3306 3307
	case KVM_SET_USER_MEMORY_REGION: {
		struct kvm_userspace_memory_region kvm_userspace_mem;

		r = -EFAULT;
		if (copy_from_user(&kvm_userspace_mem, argp,
3308
						sizeof(kvm_userspace_mem)))
3309 3310
			goto out;

3311
		r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem);
A
Avi Kivity 已提交
3312 3313 3314 3315 3316 3317
		break;
	}
	case KVM_GET_DIRTY_LOG: {
		struct kvm_dirty_log log;

		r = -EFAULT;
3318
		if (copy_from_user(&log, argp, sizeof(log)))
A
Avi Kivity 已提交
3319
			goto out;
3320
		r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
A
Avi Kivity 已提交
3321 3322
		break;
	}
3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333
#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
3334
#ifdef CONFIG_KVM_MMIO
3335 3336
	case KVM_REGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3337

3338
		r = -EFAULT;
3339
		if (copy_from_user(&zone, argp, sizeof(zone)))
3340 3341 3342 3343 3344 3345
			goto out;
		r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
		break;
	}
	case KVM_UNREGISTER_COALESCED_MMIO: {
		struct kvm_coalesced_mmio_zone zone;
3346

3347
		r = -EFAULT;
3348
		if (copy_from_user(&zone, argp, sizeof(zone)))
3349 3350 3351 3352 3353
			goto out;
		r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
		break;
	}
#endif
G
Gregory Haskins 已提交
3354 3355 3356 3357
	case KVM_IRQFD: {
		struct kvm_irqfd data;

		r = -EFAULT;
3358
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3359
			goto out;
3360
		r = kvm_irqfd(kvm, &data);
G
Gregory Haskins 已提交
3361 3362
		break;
	}
G
Gregory Haskins 已提交
3363 3364 3365 3366
	case KVM_IOEVENTFD: {
		struct kvm_ioeventfd data;

		r = -EFAULT;
3367
		if (copy_from_user(&data, argp, sizeof(data)))
G
Gregory Haskins 已提交
3368 3369 3370 3371
			goto out;
		r = kvm_ioeventfd(kvm, &data);
		break;
	}
3372 3373 3374 3375 3376
#ifdef CONFIG_HAVE_KVM_MSI
	case KVM_SIGNAL_MSI: {
		struct kvm_msi msi;

		r = -EFAULT;
3377
		if (copy_from_user(&msi, argp, sizeof(msi)))
3378 3379 3380 3381
			goto out;
		r = kvm_send_userspace_msi(kvm, &msi);
		break;
	}
3382 3383 3384 3385 3386 3387 3388
#endif
#ifdef __KVM_HAVE_IRQ_LINE
	case KVM_IRQ_LINE_STATUS:
	case KVM_IRQ_LINE: {
		struct kvm_irq_level irq_event;

		r = -EFAULT;
3389
		if (copy_from_user(&irq_event, argp, sizeof(irq_event)))
3390 3391
			goto out;

3392 3393
		r = kvm_vm_ioctl_irq_line(kvm, &irq_event,
					ioctl == KVM_IRQ_LINE_STATUS);
3394 3395 3396 3397 3398
		if (r)
			goto out;

		r = -EFAULT;
		if (ioctl == KVM_IRQ_LINE_STATUS) {
3399
			if (copy_to_user(argp, &irq_event, sizeof(irq_event)))
3400 3401 3402 3403 3404 3405
				goto out;
		}

		r = 0;
		break;
	}
3406
#endif
3407 3408 3409 3410
#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
	case KVM_SET_GSI_ROUTING: {
		struct kvm_irq_routing routing;
		struct kvm_irq_routing __user *urouting;
3411
		struct kvm_irq_routing_entry *entries = NULL;
3412 3413 3414 3415 3416

		r = -EFAULT;
		if (copy_from_user(&routing, argp, sizeof(routing)))
			goto out;
		r = -EINVAL;
3417 3418
		if (!kvm_arch_can_set_irq_routing(kvm))
			goto out;
3419
		if (routing.nr > KVM_MAX_IRQ_ROUTES)
3420 3421 3422
			goto out;
		if (routing.flags)
			goto out;
3423 3424
		if (routing.nr) {
			r = -ENOMEM;
3425 3426
			entries = vmalloc(array_size(sizeof(*entries),
						     routing.nr));
3427 3428 3429 3430 3431 3432 3433 3434
			if (!entries)
				goto out;
			r = -EFAULT;
			urouting = argp;
			if (copy_from_user(entries, urouting->entries,
					   routing.nr * sizeof(*entries)))
				goto out_free_irq_routing;
		}
3435 3436
		r = kvm_set_irq_routing(kvm, entries, routing.nr,
					routing.flags);
3437
out_free_irq_routing:
3438 3439 3440 3441
		vfree(entries);
		break;
	}
#endif /* CONFIG_HAVE_KVM_IRQ_ROUTING */
S
Scott Wood 已提交
3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459
	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;
	}
3460 3461 3462
	case KVM_CHECK_EXTENSION:
		r = kvm_vm_ioctl_check_extension_generic(kvm, arg);
		break;
3463
	default:
3464
		r = kvm_arch_vm_ioctl(filp, ioctl, arg);
3465 3466 3467 3468 3469
	}
out:
	return r;
}

3470
#ifdef CONFIG_KVM_COMPAT
3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494
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)))
3495
			return -EFAULT;
3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510
		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

3511
static struct file_operations kvm_vm_fops = {
3512 3513
	.release        = kvm_vm_release,
	.unlocked_ioctl = kvm_vm_ioctl,
3514
	.llseek		= noop_llseek,
3515
	KVM_COMPAT(kvm_vm_compat_ioctl),
3516 3517
};

3518
static int kvm_dev_ioctl_create_vm(unsigned long type)
3519
{
3520
	int r;
3521
	struct kvm *kvm;
3522
	struct file *file;
3523

3524
	kvm = kvm_create_vm(type);
3525 3526
	if (IS_ERR(kvm))
		return PTR_ERR(kvm);
3527
#ifdef CONFIG_KVM_MMIO
3528
	r = kvm_coalesced_mmio_init(kvm);
3529 3530
	if (r < 0)
		goto put_kvm;
3531
#endif
3532
	r = get_unused_fd_flags(O_CLOEXEC);
3533 3534 3535
	if (r < 0)
		goto put_kvm;

3536 3537 3538
	file = anon_inode_getfile("kvm-vm", &kvm_vm_fops, kvm, O_RDWR);
	if (IS_ERR(file)) {
		put_unused_fd(r);
3539 3540
		r = PTR_ERR(file);
		goto put_kvm;
3541
	}
3542

3543 3544 3545 3546 3547 3548
	/*
	 * 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).
	 */
3549
	if (kvm_create_vm_debugfs(kvm, r) < 0) {
3550 3551
		put_unused_fd(r);
		fput(file);
3552 3553
		return -ENOMEM;
	}
3554
	kvm_uevent_notify_change(KVM_EVENT_CREATE_VM, kvm);
3555

3556
	fd_install(r, file);
3557
	return r;
3558 3559 3560 3561

put_kvm:
	kvm_put_kvm(kvm);
	return r;
3562 3563 3564 3565 3566
}

static long kvm_dev_ioctl(struct file *filp,
			  unsigned int ioctl, unsigned long arg)
{
3567
	long r = -EINVAL;
3568 3569 3570

	switch (ioctl) {
	case KVM_GET_API_VERSION:
3571 3572
		if (arg)
			goto out;
3573 3574 3575
		r = KVM_API_VERSION;
		break;
	case KVM_CREATE_VM:
3576
		r = kvm_dev_ioctl_create_vm(arg);
3577
		break;
3578
	case KVM_CHECK_EXTENSION:
3579
		r = kvm_vm_ioctl_check_extension_generic(NULL, arg);
3580
		break;
3581 3582 3583
	case KVM_GET_VCPU_MMAP_SIZE:
		if (arg)
			goto out;
3584 3585 3586
		r = PAGE_SIZE;     /* struct kvm_run */
#ifdef CONFIG_X86
		r += PAGE_SIZE;    /* pio data page */
3587
#endif
3588
#ifdef CONFIG_KVM_MMIO
3589
		r += PAGE_SIZE;    /* coalesced mmio ring page */
3590
#endif
3591
		break;
3592 3593 3594
	case KVM_TRACE_ENABLE:
	case KVM_TRACE_PAUSE:
	case KVM_TRACE_DISABLE:
3595
		r = -EOPNOTSUPP;
3596
		break;
A
Avi Kivity 已提交
3597
	default:
3598
		return kvm_arch_dev_ioctl(filp, ioctl, arg);
A
Avi Kivity 已提交
3599 3600 3601 3602 3603 3604 3605
	}
out:
	return r;
}

static struct file_operations kvm_chardev_ops = {
	.unlocked_ioctl = kvm_dev_ioctl,
3606
	.llseek		= noop_llseek,
3607
	KVM_COMPAT(kvm_dev_ioctl),
A
Avi Kivity 已提交
3608 3609 3610
};

static struct miscdevice kvm_dev = {
A
Avi Kivity 已提交
3611
	KVM_MINOR,
A
Avi Kivity 已提交
3612 3613 3614 3615
	"kvm",
	&kvm_chardev_ops,
};

3616
static void hardware_enable_nolock(void *junk)
3617 3618
{
	int cpu = raw_smp_processor_id();
3619
	int r;
3620

3621
	if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
3622
		return;
3623

3624
	cpumask_set_cpu(cpu, cpus_hardware_enabled);
3625

3626
	r = kvm_arch_hardware_enable();
3627 3628 3629 3630

	if (r) {
		cpumask_clear_cpu(cpu, cpus_hardware_enabled);
		atomic_inc(&hardware_enable_failed);
X
Xiubo Li 已提交
3631
		pr_info("kvm: enabling virtualization on CPU%d failed\n", cpu);
3632
	}
3633 3634
}

3635
static int kvm_starting_cpu(unsigned int cpu)
3636
{
3637
	raw_spin_lock(&kvm_count_lock);
3638 3639
	if (kvm_usage_count)
		hardware_enable_nolock(NULL);
3640
	raw_spin_unlock(&kvm_count_lock);
3641
	return 0;
3642 3643 3644
}

static void hardware_disable_nolock(void *junk)
3645 3646 3647
{
	int cpu = raw_smp_processor_id();

3648
	if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
3649
		return;
3650
	cpumask_clear_cpu(cpu, cpus_hardware_enabled);
3651
	kvm_arch_hardware_disable();
3652 3653
}

3654
static int kvm_dying_cpu(unsigned int cpu)
3655
{
3656
	raw_spin_lock(&kvm_count_lock);
3657 3658
	if (kvm_usage_count)
		hardware_disable_nolock(NULL);
3659
	raw_spin_unlock(&kvm_count_lock);
3660
	return 0;
3661 3662
}

3663 3664 3665 3666 3667 3668
static void hardware_disable_all_nolock(void)
{
	BUG_ON(!kvm_usage_count);

	kvm_usage_count--;
	if (!kvm_usage_count)
3669
		on_each_cpu(hardware_disable_nolock, NULL, 1);
3670 3671 3672 3673
}

static void hardware_disable_all(void)
{
3674
	raw_spin_lock(&kvm_count_lock);
3675
	hardware_disable_all_nolock();
3676
	raw_spin_unlock(&kvm_count_lock);
3677 3678 3679 3680 3681 3682
}

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

3683
	raw_spin_lock(&kvm_count_lock);
3684 3685 3686 3687

	kvm_usage_count++;
	if (kvm_usage_count == 1) {
		atomic_set(&hardware_enable_failed, 0);
3688
		on_each_cpu(hardware_enable_nolock, NULL, 1);
3689 3690 3691 3692 3693 3694 3695

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

3696
	raw_spin_unlock(&kvm_count_lock);
3697 3698 3699 3700

	return r;
}

3701
static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
M
Mike Day 已提交
3702
		      void *v)
3703
{
3704 3705 3706 3707 3708 3709
	/*
	 * 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
3710
	pr_info("kvm: exiting hardware virtualization\n");
3711
	kvm_rebooting = true;
3712
	on_each_cpu(hardware_disable_nolock, NULL, 1);
3713 3714 3715 3716 3717 3718 3719 3720
	return NOTIFY_OK;
}

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

M
Marcelo Tosatti 已提交
3721
static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
3722 3723 3724 3725
{
	int i;

	for (i = 0; i < bus->dev_count; i++) {
3726
		struct kvm_io_device *pos = bus->range[i].dev;
3727 3728 3729

		kvm_iodevice_destructor(pos);
	}
M
Marcelo Tosatti 已提交
3730
	kfree(bus);
3731 3732
}

3733
static inline int kvm_io_bus_cmp(const struct kvm_io_range *r1,
X
Xiubo Li 已提交
3734
				 const struct kvm_io_range *r2)
3735
{
J
Jason Wang 已提交
3736 3737 3738 3739
	gpa_t addr1 = r1->addr;
	gpa_t addr2 = r2->addr;

	if (addr1 < addr2)
3740
		return -1;
J
Jason Wang 已提交
3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752

	/* 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)
3753
		return 1;
J
Jason Wang 已提交
3754

3755 3756 3757
	return 0;
}

3758 3759
static int kvm_io_bus_sort_cmp(const void *p1, const void *p2)
{
3760
	return kvm_io_bus_cmp(p1, p2);
3761 3762
}

G
Geoff Levand 已提交
3763
static int kvm_io_bus_get_first_dev(struct kvm_io_bus *bus,
3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780
			     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;

3781
	while (off > 0 && kvm_io_bus_cmp(&key, &bus->range[off-1]) == 0)
3782 3783 3784 3785 3786
		off--;

	return off;
}

3787
static int __kvm_io_bus_write(struct kvm_vcpu *vcpu, struct kvm_io_bus *bus,
C
Cornelia Huck 已提交
3788 3789 3790 3791 3792 3793 3794 3795 3796
			      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 &&
3797
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
3798
		if (!kvm_iodevice_write(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
3799 3800 3801 3802 3803 3804 3805 3806
					range->len, val))
			return idx;
		idx++;
	}

	return -EOPNOTSUPP;
}

3807
/* kvm_io_bus_write - called under kvm->slots_lock */
3808
int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
3809
		     int len, const void *val)
3810
{
3811
	struct kvm_io_bus *bus;
3812
	struct kvm_io_range range;
C
Cornelia Huck 已提交
3813
	int r;
3814 3815 3816 3817 3818

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
3819

3820
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3821 3822
	if (!bus)
		return -ENOMEM;
3823
	r = __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3824 3825
	return r < 0 ? r : 0;
}
L
Leo Yan 已提交
3826
EXPORT_SYMBOL_GPL(kvm_io_bus_write);
C
Cornelia Huck 已提交
3827 3828

/* kvm_io_bus_write_cookie - called under kvm->slots_lock */
3829 3830
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
3831 3832 3833 3834 3835 3836 3837 3838 3839
{
	struct kvm_io_bus *bus;
	struct kvm_io_range range;

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

3840
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3841 3842
	if (!bus)
		return -ENOMEM;
C
Cornelia Huck 已提交
3843 3844 3845

	/* First try the device referenced by cookie. */
	if ((cookie >= 0) && (cookie < bus->dev_count) &&
3846
	    (kvm_io_bus_cmp(&range, &bus->range[cookie]) == 0))
3847
		if (!kvm_iodevice_write(vcpu, bus->range[cookie].dev, addr, len,
C
Cornelia Huck 已提交
3848 3849 3850 3851 3852 3853 3854
					val))
			return cookie;

	/*
	 * cookie contained garbage; fall back to search and return the
	 * correct cookie value.
	 */
3855
	return __kvm_io_bus_write(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3856 3857
}

3858 3859
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
3860 3861 3862 3863
{
	int idx;

	idx = kvm_io_bus_get_first_dev(bus, range->addr, range->len);
3864 3865 3866 3867
	if (idx < 0)
		return -EOPNOTSUPP;

	while (idx < bus->dev_count &&
3868
		kvm_io_bus_cmp(range, &bus->range[idx]) == 0) {
3869
		if (!kvm_iodevice_read(vcpu, bus->range[idx].dev, range->addr,
C
Cornelia Huck 已提交
3870 3871
				       range->len, val))
			return idx;
3872 3873 3874
		idx++;
	}

3875 3876
	return -EOPNOTSUPP;
}
3877

3878
/* kvm_io_bus_read - called under kvm->slots_lock */
3879
int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
M
Marcelo Tosatti 已提交
3880
		    int len, void *val)
3881
{
3882
	struct kvm_io_bus *bus;
3883
	struct kvm_io_range range;
C
Cornelia Huck 已提交
3884
	int r;
3885 3886 3887 3888 3889

	range = (struct kvm_io_range) {
		.addr = addr,
		.len = len,
	};
M
Marcelo Tosatti 已提交
3890

3891
	bus = srcu_dereference(vcpu->kvm->buses[bus_idx], &vcpu->kvm->srcu);
3892 3893
	if (!bus)
		return -ENOMEM;
3894
	r = __kvm_io_bus_read(vcpu, bus, &range, val);
C
Cornelia Huck 已提交
3895 3896
	return r < 0 ? r : 0;
}
3897

3898
/* Caller must hold slots_lock. */
3899 3900
int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
			    int len, struct kvm_io_device *dev)
3901
{
3902
	int i;
M
Marcelo Tosatti 已提交
3903
	struct kvm_io_bus *new_bus, *bus;
3904
	struct kvm_io_range range;
3905

3906
	bus = kvm_get_bus(kvm, bus_idx);
3907 3908 3909
	if (!bus)
		return -ENOMEM;

3910 3911
	/* exclude ioeventfd which is limited by maximum fd */
	if (bus->dev_count - bus->ioeventfd_count > NR_IOBUS_DEVS - 1)
3912
		return -ENOSPC;
3913

3914
	new_bus = kmalloc(struct_size(bus, range, bus->dev_count + 1),
3915
			  GFP_KERNEL_ACCOUNT);
M
Marcelo Tosatti 已提交
3916 3917
	if (!new_bus)
		return -ENOMEM;
3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933

	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
3934 3935 3936
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
3937 3938 3939 3940

	return 0;
}

3941
/* Caller must hold slots_lock. */
3942 3943
void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
			       struct kvm_io_device *dev)
3944
{
3945
	int i;
M
Marcelo Tosatti 已提交
3946
	struct kvm_io_bus *new_bus, *bus;
3947

3948
	bus = kvm_get_bus(kvm, bus_idx);
3949
	if (!bus)
3950
		return;
3951

3952 3953
	for (i = 0; i < bus->dev_count; i++)
		if (bus->range[i].dev == dev) {
3954 3955
			break;
		}
M
Marcelo Tosatti 已提交
3956

3957 3958
	if (i == bus->dev_count)
		return;
3959

3960
	new_bus = kmalloc(struct_size(bus, range, bus->dev_count - 1),
3961
			  GFP_KERNEL_ACCOUNT);
3962 3963 3964 3965
	if (!new_bus)  {
		pr_err("kvm: failed to shrink bus, removing it completely\n");
		goto broken;
	}
3966 3967 3968 3969 3970

	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
3971

3972
broken:
M
Marcelo Tosatti 已提交
3973 3974 3975
	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
	synchronize_srcu_expedited(&kvm->srcu);
	kfree(bus);
3976
	return;
3977 3978
}

3979 3980 3981 3982 3983 3984 3985 3986 3987 3988
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);
3989 3990
	if (!bus)
		goto out_unlock;
3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004

	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);

4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016
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.
	 */
4017
	if (!refcount_inc_not_zero(&stat_data->kvm->users_count))
4018 4019
		return -ENOENT;

4020
	if (simple_attr_open(inode, file, get,
4021 4022 4023
		    KVM_DBGFS_GET_MODE(stat_data->dbgfs_item) & 0222
		    ? set : NULL,
		    fmt)) {
4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041
		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;
}

4042
static int kvm_get_stat_per_vm(struct kvm *kvm, size_t offset, u64 *val)
4043
{
4044
	*val = *(ulong *)((void *)kvm + offset);
4045

4046 4047 4048 4049 4050 4051
	return 0;
}

static int kvm_clear_stat_per_vm(struct kvm *kvm, size_t offset)
{
	*(ulong *)((void *)kvm + offset) = 0;
4052 4053 4054 4055

	return 0;
}

4056
static int kvm_get_stat_per_vcpu(struct kvm *kvm, size_t offset, u64 *val)
4057
{
4058 4059
	int i;
	struct kvm_vcpu *vcpu;
4060

4061
	*val = 0;
4062

4063 4064
	kvm_for_each_vcpu(i, vcpu, kvm)
		*val += *(u64 *)((void *)vcpu + offset);
4065 4066 4067 4068

	return 0;
}

4069
static int kvm_clear_stat_per_vcpu(struct kvm *kvm, size_t offset)
4070
{
4071 4072
	int i;
	struct kvm_vcpu *vcpu;
4073

4074 4075 4076 4077 4078
	kvm_for_each_vcpu(i, vcpu, kvm)
		*(u64 *)((void *)vcpu + offset) = 0;

	return 0;
}
4079

4080
static int kvm_stat_data_get(void *data, u64 *val)
4081
{
4082
	int r = -EFAULT;
4083 4084
	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)data;

4085 4086 4087 4088 4089 4090 4091 4092 4093 4094
	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;
	}
4095

4096
	return r;
4097 4098
}

4099
static int kvm_stat_data_clear(void *data, u64 val)
4100
{
4101
	int r = -EFAULT;
4102 4103 4104 4105 4106
	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)data;

	if (val)
		return -EINVAL;

4107 4108 4109 4110 4111 4112 4113 4114 4115 4116
	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;
	}
4117

4118
	return r;
4119 4120
}

4121
static int kvm_stat_data_open(struct inode *inode, struct file *file)
4122 4123
{
	__simple_attr_check_format("%llu\n", 0ull);
4124 4125
	return kvm_debugfs_open(inode, file, kvm_stat_data_get,
				kvm_stat_data_clear, "%llu\n");
4126 4127
}

4128 4129 4130
static const struct file_operations stat_fops_per_vm = {
	.owner = THIS_MODULE,
	.open = kvm_stat_data_open,
4131
	.release = kvm_debugfs_release,
4132 4133 4134
	.read = simple_attr_read,
	.write = simple_attr_write,
	.llseek = no_llseek,
4135 4136
};

4137
static int vm_stat_get(void *_offset, u64 *val)
4138 4139 4140
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
4141
	u64 tmp_val;
4142

4143
	*val = 0;
J
Junaid Shahid 已提交
4144
	mutex_lock(&kvm_lock);
4145
	list_for_each_entry(kvm, &vm_list, vm_list) {
4146
		kvm_get_stat_per_vm(kvm, offset, &tmp_val);
4147 4148
		*val += tmp_val;
	}
J
Junaid Shahid 已提交
4149
	mutex_unlock(&kvm_lock);
4150
	return 0;
4151 4152
}

4153 4154 4155 4156 4157 4158 4159 4160
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
4161
	mutex_lock(&kvm_lock);
4162
	list_for_each_entry(kvm, &vm_list, vm_list) {
4163
		kvm_clear_stat_per_vm(kvm, offset);
4164
	}
J
Junaid Shahid 已提交
4165
	mutex_unlock(&kvm_lock);
4166 4167 4168 4169 4170

	return 0;
}

DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, vm_stat_clear, "%llu\n");
4171

4172
static int vcpu_stat_get(void *_offset, u64 *val)
A
Avi Kivity 已提交
4173 4174 4175
{
	unsigned offset = (long)_offset;
	struct kvm *kvm;
4176
	u64 tmp_val;
A
Avi Kivity 已提交
4177

4178
	*val = 0;
J
Junaid Shahid 已提交
4179
	mutex_lock(&kvm_lock);
4180
	list_for_each_entry(kvm, &vm_list, vm_list) {
4181
		kvm_get_stat_per_vcpu(kvm, offset, &tmp_val);
4182 4183
		*val += tmp_val;
	}
J
Junaid Shahid 已提交
4184
	mutex_unlock(&kvm_lock);
4185
	return 0;
A
Avi Kivity 已提交
4186 4187
}

4188 4189 4190 4191 4192 4193 4194 4195
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
4196
	mutex_lock(&kvm_lock);
4197
	list_for_each_entry(kvm, &vm_list, vm_list) {
4198
		kvm_clear_stat_per_vcpu(kvm, offset);
4199
	}
J
Junaid Shahid 已提交
4200
	mutex_unlock(&kvm_lock);
4201 4202 4203 4204 4205 4206

	return 0;
}

DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, vcpu_stat_clear,
			"%llu\n");
4207

4208
static const struct file_operations *stat_fops[] = {
4209 4210 4211
	[KVM_STAT_VCPU] = &vcpu_stat_fops,
	[KVM_STAT_VM]   = &vm_stat_fops,
};
A
Avi Kivity 已提交
4212

4213 4214 4215 4216 4217 4218 4219 4220
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
4221
	mutex_lock(&kvm_lock);
4222 4223 4224 4225 4226 4227 4228 4229
	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
4230
	mutex_unlock(&kvm_lock);
4231

4232
	env = kzalloc(sizeof(*env), GFP_KERNEL_ACCOUNT);
4233 4234 4235 4236 4237 4238
	if (!env)
		return;

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

4239
	if (type == KVM_EVENT_CREATE_VM) {
4240
		add_uevent_var(env, "EVENT=create");
4241 4242
		kvm->userspace_pid = task_pid_nr(current);
	} else if (type == KVM_EVENT_DESTROY_VM) {
4243
		add_uevent_var(env, "EVENT=destroy");
4244 4245
	}
	add_uevent_var(env, "PID=%d", kvm->userspace_pid);
4246

4247
	if (!IS_ERR_OR_NULL(kvm->debugfs_dentry)) {
4248
		char *tmp, *p = kmalloc(PATH_MAX, GFP_KERNEL_ACCOUNT);
4249 4250 4251 4252 4253 4254

		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);
4255 4256 4257 4258 4259 4260 4261 4262
		}
	}
	/* 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);
}

4263
static void kvm_init_debug(void)
A
Avi Kivity 已提交
4264 4265 4266
{
	struct kvm_stats_debugfs_item *p;

4267
	kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
4268

4269 4270
	kvm_debugfs_num_entries = 0;
	for (p = debugfs_entries; p->name; ++p, kvm_debugfs_num_entries++) {
4271 4272
		debugfs_create_file(p->name, KVM_DBGFS_GET_MODE(p),
				    kvm_debugfs_dir, (void *)(long)p->offset,
4273
				    stat_fops[p->kind]);
4274
	}
A
Avi Kivity 已提交
4275 4276
}

4277
static int kvm_suspend(void)
4278
{
4279
	if (kvm_usage_count)
4280
		hardware_disable_nolock(NULL);
4281 4282 4283
	return 0;
}

4284
static void kvm_resume(void)
4285
{
4286
	if (kvm_usage_count) {
4287 4288 4289
#ifdef CONFIG_LOCKDEP
		WARN_ON(lockdep_is_held(&kvm_count_lock));
#endif
4290
		hardware_enable_nolock(NULL);
4291
	}
4292 4293
}

4294
static struct syscore_ops kvm_syscore_ops = {
4295 4296 4297 4298
	.suspend = kvm_suspend,
	.resume = kvm_resume,
};

4299 4300 4301 4302 4303 4304 4305 4306 4307
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);
4308

4309
	WRITE_ONCE(vcpu->preempted, false);
4310
	WRITE_ONCE(vcpu->ready, false);
4311

4312
	__this_cpu_write(kvm_running_vcpu, vcpu);
R
Radim Krčmář 已提交
4313
	kvm_arch_sched_in(vcpu, cpu);
4314
	kvm_arch_vcpu_load(vcpu, cpu);
4315 4316 4317 4318 4319 4320 4321
}

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

4322
	if (current->state == TASK_RUNNING) {
4323
		WRITE_ONCE(vcpu->preempted, true);
4324 4325
		WRITE_ONCE(vcpu->ready, true);
	}
4326
	kvm_arch_vcpu_put(vcpu);
4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345
	__this_cpu_write(kvm_running_vcpu, NULL);
}

/**
 * kvm_get_running_vcpu - get the vcpu running on the current CPU.
 * Thanks to preempt notifiers, this can also be called from
 * preemptible context.
 */
struct kvm_vcpu *kvm_get_running_vcpu(void)
{
        return __this_cpu_read(kvm_running_vcpu);
}

/**
 * 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;
4346 4347
}

4348 4349 4350 4351 4352
static void check_processor_compat(void *rtn)
{
	*(int *)rtn = kvm_arch_check_processor_compat();
}

4353
int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
4354
		  struct module *module)
A
Avi Kivity 已提交
4355 4356
{
	int r;
Y
Yang, Sheng 已提交
4357
	int cpu;
A
Avi Kivity 已提交
4358

4359 4360
	r = kvm_arch_init(opaque);
	if (r)
4361
		goto out_fail;
4362

4363 4364 4365 4366
	/*
	 * 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
4367 4368
	 * kvm_arch_init must be called before kvm_irqfd_init to avoid creating
	 * conflicts in case kvm is already setup for another implementation.
4369
	 */
P
Paolo Bonzini 已提交
4370 4371 4372
	r = kvm_irqfd_init();
	if (r)
		goto out_irqfd;
4373

4374
	if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
4375 4376 4377 4378
		r = -ENOMEM;
		goto out_free_0;
	}

4379
	r = kvm_arch_hardware_setup();
A
Avi Kivity 已提交
4380
	if (r < 0)
4381
		goto out_free_1;
A
Avi Kivity 已提交
4382

Y
Yang, Sheng 已提交
4383
	for_each_online_cpu(cpu) {
4384
		smp_call_function_single(cpu, check_processor_compat, &r, 1);
Y
Yang, Sheng 已提交
4385
		if (r < 0)
4386
			goto out_free_2;
Y
Yang, Sheng 已提交
4387 4388
	}

T
Thomas Gleixner 已提交
4389
	r = cpuhp_setup_state_nocalls(CPUHP_AP_KVM_STARTING, "kvm/cpu:starting",
4390
				      kvm_starting_cpu, kvm_dying_cpu);
A
Avi Kivity 已提交
4391
	if (r)
4392
		goto out_free_2;
A
Avi Kivity 已提交
4393 4394
	register_reboot_notifier(&kvm_reboot_notifier);

4395
	/* A kmem cache lets us meet the alignment requirements of fx_save. */
4396 4397
	if (!vcpu_align)
		vcpu_align = __alignof__(struct kvm_vcpu);
4398 4399 4400 4401 4402 4403
	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);
4404 4405
	if (!kvm_vcpu_cache) {
		r = -ENOMEM;
4406
		goto out_free_3;
4407 4408
	}

4409 4410 4411 4412
	r = kvm_async_pf_init();
	if (r)
		goto out_free;

A
Avi Kivity 已提交
4413
	kvm_chardev_ops.owner = module;
4414 4415
	kvm_vm_fops.owner = module;
	kvm_vcpu_fops.owner = module;
A
Avi Kivity 已提交
4416 4417 4418

	r = misc_register(&kvm_dev);
	if (r) {
X
Xiubo Li 已提交
4419
		pr_err("kvm: misc device register failed\n");
4420
		goto out_unreg;
A
Avi Kivity 已提交
4421 4422
	}

4423 4424
	register_syscore_ops(&kvm_syscore_ops);

4425 4426 4427
	kvm_preempt_ops.sched_in = kvm_sched_in;
	kvm_preempt_ops.sched_out = kvm_sched_out;

4428
	kvm_init_debug();
4429

P
Paolo Bonzini 已提交
4430 4431 4432
	r = kvm_vfio_ops_init();
	WARN_ON(r);

4433
	return 0;
A
Avi Kivity 已提交
4434

4435 4436
out_unreg:
	kvm_async_pf_deinit();
A
Avi Kivity 已提交
4437
out_free:
4438
	kmem_cache_destroy(kvm_vcpu_cache);
4439
out_free_3:
A
Avi Kivity 已提交
4440
	unregister_reboot_notifier(&kvm_reboot_notifier);
4441
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4442
out_free_2:
4443
	kvm_arch_hardware_unsetup();
4444
out_free_1:
4445
	free_cpumask_var(cpus_hardware_enabled);
4446
out_free_0:
4447
	kvm_irqfd_exit();
P
Paolo Bonzini 已提交
4448
out_irqfd:
4449 4450
	kvm_arch_exit();
out_fail:
A
Avi Kivity 已提交
4451 4452
	return r;
}
4453
EXPORT_SYMBOL_GPL(kvm_init);
A
Avi Kivity 已提交
4454

4455
void kvm_exit(void)
A
Avi Kivity 已提交
4456
{
4457
	debugfs_remove_recursive(kvm_debugfs_dir);
A
Avi Kivity 已提交
4458
	misc_deregister(&kvm_dev);
4459
	kmem_cache_destroy(kvm_vcpu_cache);
4460
	kvm_async_pf_deinit();
4461
	unregister_syscore_ops(&kvm_syscore_ops);
A
Avi Kivity 已提交
4462
	unregister_reboot_notifier(&kvm_reboot_notifier);
4463
	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_STARTING);
4464
	on_each_cpu(hardware_disable_nolock, NULL, 1);
4465
	kvm_arch_hardware_unsetup();
4466
	kvm_arch_exit();
4467
	kvm_irqfd_exit();
4468
	free_cpumask_var(cpus_hardware_enabled);
4469
	kvm_vfio_ops_exit();
A
Avi Kivity 已提交
4470
}
4471
EXPORT_SYMBOL_GPL(kvm_exit);
4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554

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;
}