mmu.c 124.4 KB
Newer Older
A
Avi Kivity 已提交
1 2 3 4 5 6 7 8 9
/*
 * 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.
 *
 * MMU support
 *
 * Copyright (C) 2006 Qumranet, Inc.
N
Nicolas Kaiser 已提交
10
 * Copyright 2010 Red Hat, Inc. and/or its affiliates.
A
Avi Kivity 已提交
11 12 13 14 15 16 17 18 19
 *
 * Authors:
 *   Yaniv Kamay  <yaniv@qumranet.com>
 *   Avi Kivity   <avi@qumranet.com>
 *
 * This work is licensed under the terms of the GNU GPL, version 2.  See
 * the COPYING file in the top-level directory.
 *
 */
A
Avi Kivity 已提交
20

21
#include "irq.h"
22
#include "mmu.h"
23
#include "x86.h"
A
Avi Kivity 已提交
24
#include "kvm_cache_regs.h"
25
#include "cpuid.h"
A
Avi Kivity 已提交
26

27
#include <linux/kvm_host.h>
A
Avi Kivity 已提交
28 29 30 31 32
#include <linux/types.h>
#include <linux/string.h>
#include <linux/mm.h>
#include <linux/highmem.h>
#include <linux/module.h>
33
#include <linux/swap.h>
M
Marcelo Tosatti 已提交
34
#include <linux/hugetlb.h>
35
#include <linux/compiler.h>
36
#include <linux/srcu.h>
37
#include <linux/slab.h>
38
#include <linux/uaccess.h>
A
Avi Kivity 已提交
39

A
Avi Kivity 已提交
40 41
#include <asm/page.h>
#include <asm/cmpxchg.h>
42
#include <asm/io.h>
43
#include <asm/vmx.h>
A
Avi Kivity 已提交
44

45 46 47 48 49 50 51
/*
 * When setting this variable to true it enables Two-Dimensional-Paging
 * where the hardware walks 2 page tables:
 * 1. the guest-virtual to guest-physical
 * 2. while doing 1. it walks guest-physical to host-physical
 * If the hardware supports that we don't need to do shadow paging.
 */
52
bool tdp_enabled = false;
53

54 55 56 57
enum {
	AUDIT_PRE_PAGE_FAULT,
	AUDIT_POST_PAGE_FAULT,
	AUDIT_PRE_PTE_WRITE,
58 59 60
	AUDIT_POST_PTE_WRITE,
	AUDIT_PRE_SYNC,
	AUDIT_POST_SYNC
61
};
62

63
#undef MMU_DEBUG
64 65

#ifdef MMU_DEBUG
66 67
static bool dbg = 0;
module_param(dbg, bool, 0644);
68 69 70

#define pgprintk(x...) do { if (dbg) printk(x); } while (0)
#define rmap_printk(x...) do { if (dbg) printk(x); } while (0)
71
#define MMU_WARN_ON(x) WARN_ON(x)
72 73 74
#else
#define pgprintk(x...) do { } while (0)
#define rmap_printk(x...) do { } while (0)
75
#define MMU_WARN_ON(x) do { } while (0)
76
#endif
A
Avi Kivity 已提交
77

78 79
#define PTE_PREFETCH_NUM		8

80
#define PT_FIRST_AVAIL_BITS_SHIFT 10
A
Avi Kivity 已提交
81 82 83 84 85
#define PT64_SECOND_AVAIL_BITS_SHIFT 52

#define PT64_LEVEL_BITS 9

#define PT64_LEVEL_SHIFT(level) \
M
Mike Day 已提交
86
		(PAGE_SHIFT + (level - 1) * PT64_LEVEL_BITS)
A
Avi Kivity 已提交
87 88 89 90 91 92 93 94

#define PT64_INDEX(address, level)\
	(((address) >> PT64_LEVEL_SHIFT(level)) & ((1 << PT64_LEVEL_BITS) - 1))


#define PT32_LEVEL_BITS 10

#define PT32_LEVEL_SHIFT(level) \
M
Mike Day 已提交
95
		(PAGE_SHIFT + (level - 1) * PT32_LEVEL_BITS)
A
Avi Kivity 已提交
96

97 98 99
#define PT32_LVL_OFFSET_MASK(level) \
	(PT32_BASE_ADDR_MASK & ((1ULL << (PAGE_SHIFT + (((level) - 1) \
						* PT32_LEVEL_BITS))) - 1))
A
Avi Kivity 已提交
100 101 102 103 104

#define PT32_INDEX(address, level)\
	(((address) >> PT32_LEVEL_SHIFT(level)) & ((1 << PT32_LEVEL_BITS) - 1))


105
#define PT64_BASE_ADDR_MASK (((1ULL << 52) - 1) & ~(u64)(PAGE_SIZE-1))
A
Avi Kivity 已提交
106 107
#define PT64_DIR_BASE_ADDR_MASK \
	(PT64_BASE_ADDR_MASK & ~((1ULL << (PAGE_SHIFT + PT64_LEVEL_BITS)) - 1))
108 109 110 111 112 113
#define PT64_LVL_ADDR_MASK(level) \
	(PT64_BASE_ADDR_MASK & ~((1ULL << (PAGE_SHIFT + (((level) - 1) \
						* PT64_LEVEL_BITS))) - 1))
#define PT64_LVL_OFFSET_MASK(level) \
	(PT64_BASE_ADDR_MASK & ((1ULL << (PAGE_SHIFT + (((level) - 1) \
						* PT64_LEVEL_BITS))) - 1))
A
Avi Kivity 已提交
114 115 116 117

#define PT32_BASE_ADDR_MASK PAGE_MASK
#define PT32_DIR_BASE_ADDR_MASK \
	(PAGE_MASK & ~((1ULL << (PAGE_SHIFT + PT32_LEVEL_BITS)) - 1))
118 119 120
#define PT32_LVL_ADDR_MASK(level) \
	(PAGE_MASK & ~((1ULL << (PAGE_SHIFT + (((level) - 1) \
					    * PT32_LEVEL_BITS))) - 1))
A
Avi Kivity 已提交
121

122 123
#define PT64_PERM_MASK (PT_PRESENT_MASK | PT_WRITABLE_MASK | shadow_user_mask \
			| shadow_x_mask | shadow_nx_mask)
A
Avi Kivity 已提交
124

125 126 127 128 129
#define ACC_EXEC_MASK    1
#define ACC_WRITE_MASK   PT_WRITABLE_MASK
#define ACC_USER_MASK    PT_USER_MASK
#define ACC_ALL          (ACC_EXEC_MASK | ACC_WRITE_MASK | ACC_USER_MASK)

130 131
#include <trace/events/kvm.h>

132 133 134
#define CREATE_TRACE_POINTS
#include "mmutrace.h"

135 136
#define SPTE_HOST_WRITEABLE	(1ULL << PT_FIRST_AVAIL_BITS_SHIFT)
#define SPTE_MMU_WRITEABLE	(1ULL << (PT_FIRST_AVAIL_BITS_SHIFT + 1))
137

138 139
#define SHADOW_PT_INDEX(addr, level) PT64_INDEX(addr, level)

140 141 142
/* make pte_list_desc fit well in cache line */
#define PTE_LIST_EXT 3

143 144 145
struct pte_list_desc {
	u64 *sptes[PTE_LIST_EXT];
	struct pte_list_desc *more;
146 147
};

148 149 150 151
struct kvm_shadow_walk_iterator {
	u64 addr;
	hpa_t shadow_addr;
	u64 *sptep;
152
	int level;
153 154 155 156 157 158 159 160
	unsigned index;
};

#define for_each_shadow_entry(_vcpu, _addr, _walker)    \
	for (shadow_walk_init(&(_walker), _vcpu, _addr);	\
	     shadow_walk_okay(&(_walker));			\
	     shadow_walk_next(&(_walker)))

161 162 163 164 165 166
#define for_each_shadow_entry_lockless(_vcpu, _addr, _walker, spte)	\
	for (shadow_walk_init(&(_walker), _vcpu, _addr);		\
	     shadow_walk_okay(&(_walker)) &&				\
		({ spte = mmu_spte_get_lockless(_walker.sptep); 1; });	\
	     __shadow_walk_next(&(_walker), spte))

167
static struct kmem_cache *pte_list_desc_cache;
168
static struct kmem_cache *mmu_page_header_cache;
169
static struct percpu_counter kvm_total_used_mmu_pages;
170

S
Sheng Yang 已提交
171 172 173 174 175
static u64 __read_mostly shadow_nx_mask;
static u64 __read_mostly shadow_x_mask;	/* mutual exclusive with nx_mask */
static u64 __read_mostly shadow_user_mask;
static u64 __read_mostly shadow_accessed_mask;
static u64 __read_mostly shadow_dirty_mask;
176 177 178
static u64 __read_mostly shadow_mmio_mask;

static void mmu_spte_set(u64 *sptep, u64 spte);
179
static void mmu_free_roots(struct kvm_vcpu *vcpu);
180 181 182 183 184 185 186

void kvm_mmu_set_mmio_spte_mask(u64 mmio_mask)
{
	shadow_mmio_mask = mmio_mask;
}
EXPORT_SYMBOL_GPL(kvm_mmu_set_mmio_spte_mask);

187
/*
188 189 190 191 192 193 194
 * the low bit of the generation number is always presumed to be zero.
 * This disables mmio caching during memslot updates.  The concept is
 * similar to a seqcount but instead of retrying the access we just punt
 * and ignore the cache.
 *
 * spte bits 3-11 are used as bits 1-9 of the generation number,
 * the bits 52-61 are used as bits 10-19 of the generation number.
195
 */
196
#define MMIO_SPTE_GEN_LOW_SHIFT		2
197 198
#define MMIO_SPTE_GEN_HIGH_SHIFT	52

199 200 201
#define MMIO_GEN_SHIFT			20
#define MMIO_GEN_LOW_SHIFT		10
#define MMIO_GEN_LOW_MASK		((1 << MMIO_GEN_LOW_SHIFT) - 2)
202
#define MMIO_GEN_MASK			((1 << MMIO_GEN_SHIFT) - 1)
203 204 205 206 207

static u64 generation_mmio_spte_mask(unsigned int gen)
{
	u64 mask;

T
Tiejun Chen 已提交
208
	WARN_ON(gen & ~MMIO_GEN_MASK);
209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225

	mask = (gen & MMIO_GEN_LOW_MASK) << MMIO_SPTE_GEN_LOW_SHIFT;
	mask |= ((u64)gen >> MMIO_GEN_LOW_SHIFT) << MMIO_SPTE_GEN_HIGH_SHIFT;
	return mask;
}

static unsigned int get_mmio_spte_generation(u64 spte)
{
	unsigned int gen;

	spte &= ~shadow_mmio_mask;

	gen = (spte >> MMIO_SPTE_GEN_LOW_SHIFT) & MMIO_GEN_LOW_MASK;
	gen |= (spte >> MMIO_SPTE_GEN_HIGH_SHIFT) << MMIO_GEN_LOW_SHIFT;
	return gen;
}

226
static unsigned int kvm_current_mmio_generation(struct kvm_vcpu *vcpu)
227
{
228
	return kvm_vcpu_memslots(vcpu)->generation & MMIO_GEN_MASK;
229 230
}

231
static void mark_mmio_spte(struct kvm_vcpu *vcpu, u64 *sptep, u64 gfn,
232
			   unsigned access)
233
{
234
	unsigned int gen = kvm_current_mmio_generation(vcpu);
235
	u64 mask = generation_mmio_spte_mask(gen);
236

237
	access &= ACC_WRITE_MASK | ACC_USER_MASK;
238 239
	mask |= shadow_mmio_mask | access | gfn << PAGE_SHIFT;

240
	trace_mark_mmio_spte(sptep, gfn, access, gen);
241
	mmu_spte_set(sptep, mask);
242 243 244 245 246 247 248 249 250
}

static bool is_mmio_spte(u64 spte)
{
	return (spte & shadow_mmio_mask) == shadow_mmio_mask;
}

static gfn_t get_mmio_spte_gfn(u64 spte)
{
T
Tiejun Chen 已提交
251
	u64 mask = generation_mmio_spte_mask(MMIO_GEN_MASK) | shadow_mmio_mask;
252
	return (spte & ~mask) >> PAGE_SHIFT;
253 254 255 256
}

static unsigned get_mmio_spte_access(u64 spte)
{
T
Tiejun Chen 已提交
257
	u64 mask = generation_mmio_spte_mask(MMIO_GEN_MASK) | shadow_mmio_mask;
258
	return (spte & ~mask) & ~PAGE_MASK;
259 260
}

261
static bool set_mmio_spte(struct kvm_vcpu *vcpu, u64 *sptep, gfn_t gfn,
262
			  pfn_t pfn, unsigned access)
263 264
{
	if (unlikely(is_noslot_pfn(pfn))) {
265
		mark_mmio_spte(vcpu, sptep, gfn, access);
266 267 268 269 270
		return true;
	}

	return false;
}
271

272
static bool check_mmio_spte(struct kvm_vcpu *vcpu, u64 spte)
273
{
274 275
	unsigned int kvm_gen, spte_gen;

276
	kvm_gen = kvm_current_mmio_generation(vcpu);
277 278 279 280
	spte_gen = get_mmio_spte_generation(spte);

	trace_check_mmio_spte(spte, kvm_gen, spte_gen);
	return likely(kvm_gen == spte_gen);
281 282
}

S
Sheng Yang 已提交
283
void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
284
		u64 dirty_mask, u64 nx_mask, u64 x_mask)
S
Sheng Yang 已提交
285 286 287 288 289 290 291 292 293
{
	shadow_user_mask = user_mask;
	shadow_accessed_mask = accessed_mask;
	shadow_dirty_mask = dirty_mask;
	shadow_nx_mask = nx_mask;
	shadow_x_mask = x_mask;
}
EXPORT_SYMBOL_GPL(kvm_mmu_set_mask_ptes);

A
Avi Kivity 已提交
294 295 296 297 298
static int is_cpuid_PSE36(void)
{
	return 1;
}

299 300
static int is_nx(struct kvm_vcpu *vcpu)
{
301
	return vcpu->arch.efer & EFER_NX;
302 303
}

304 305
static int is_shadow_present_pte(u64 pte)
{
306
	return pte & PT_PRESENT_MASK && !is_mmio_spte(pte);
307 308
}

M
Marcelo Tosatti 已提交
309 310 311 312 313
static int is_large_pte(u64 pte)
{
	return pte & PT_PAGE_SIZE_MASK;
}

314 315 316 317
static int is_last_spte(u64 pte, int level)
{
	if (level == PT_PAGE_TABLE_LEVEL)
		return 1;
318
	if (is_large_pte(pte))
319 320 321 322
		return 1;
	return 0;
}

323
static pfn_t spte_to_pfn(u64 pte)
324
{
325
	return (pte & PT64_BASE_ADDR_MASK) >> PAGE_SHIFT;
326 327
}

328 329 330 331 332 333 334
static gfn_t pse36_gfn_delta(u32 gpte)
{
	int shift = 32 - PT32_DIR_PSE36_SHIFT - PAGE_SHIFT;

	return (gpte & PT32_DIR_PSE36_MASK) << shift;
}

335
#ifdef CONFIG_X86_64
A
Avi Kivity 已提交
336
static void __set_spte(u64 *sptep, u64 spte)
337
{
338
	*sptep = spte;
339 340
}

341
static void __update_clear_spte_fast(u64 *sptep, u64 spte)
342
{
343 344 345 346 347 348 349
	*sptep = spte;
}

static u64 __update_clear_spte_slow(u64 *sptep, u64 spte)
{
	return xchg(sptep, spte);
}
350 351 352 353 354

static u64 __get_spte_lockless(u64 *sptep)
{
	return ACCESS_ONCE(*sptep);
}
355
#else
356 357 358 359 360 361 362
union split_spte {
	struct {
		u32 spte_low;
		u32 spte_high;
	};
	u64 spte;
};
363

364 365 366 367 368 369 370 371 372 373 374 375
static void count_spte_clear(u64 *sptep, u64 spte)
{
	struct kvm_mmu_page *sp =  page_header(__pa(sptep));

	if (is_shadow_present_pte(spte))
		return;

	/* Ensure the spte is completely set before we increase the count */
	smp_wmb();
	sp->clear_spte_count++;
}

376 377 378
static void __set_spte(u64 *sptep, u64 spte)
{
	union split_spte *ssptep, sspte;
379

380 381 382 383 384 385 386 387 388 389 390 391 392
	ssptep = (union split_spte *)sptep;
	sspte = (union split_spte)spte;

	ssptep->spte_high = sspte.spte_high;

	/*
	 * If we map the spte from nonpresent to present, We should store
	 * the high bits firstly, then set present bit, so cpu can not
	 * fetch this spte while we are setting the spte.
	 */
	smp_wmb();

	ssptep->spte_low = sspte.spte_low;
393 394
}

395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410
static void __update_clear_spte_fast(u64 *sptep, u64 spte)
{
	union split_spte *ssptep, sspte;

	ssptep = (union split_spte *)sptep;
	sspte = (union split_spte)spte;

	ssptep->spte_low = sspte.spte_low;

	/*
	 * If we map the spte from present to nonpresent, we should clear
	 * present bit firstly to avoid vcpu fetch the old high bits.
	 */
	smp_wmb();

	ssptep->spte_high = sspte.spte_high;
411
	count_spte_clear(sptep, spte);
412 413 414 415 416 417 418 419 420 421 422
}

static u64 __update_clear_spte_slow(u64 *sptep, u64 spte)
{
	union split_spte *ssptep, sspte, orig;

	ssptep = (union split_spte *)sptep;
	sspte = (union split_spte)spte;

	/* xchg acts as a barrier before the setting of the high bits */
	orig.spte_low = xchg(&ssptep->spte_low, sspte.spte_low);
423 424
	orig.spte_high = ssptep->spte_high;
	ssptep->spte_high = sspte.spte_high;
425
	count_spte_clear(sptep, spte);
426 427 428

	return orig.spte;
}
429 430 431 432

/*
 * The idea using the light way get the spte on x86_32 guest is from
 * gup_get_pte(arch/x86/mm/gup.c).
433 434 435 436 437 438 439 440 441 442 443 444 445 446
 *
 * An spte tlb flush may be pending, because kvm_set_pte_rmapp
 * coalesces them and we are running out of the MMU lock.  Therefore
 * we need to protect against in-progress updates of the spte.
 *
 * Reading the spte while an update is in progress may get the old value
 * for the high part of the spte.  The race is fine for a present->non-present
 * change (because the high part of the spte is ignored for non-present spte),
 * but for a present->present change we must reread the spte.
 *
 * All such changes are done in two steps (present->non-present and
 * non-present->present), hence it is enough to count the number of
 * present->non-present updates: if it changed while reading the spte,
 * we might have hit the race.  This is done using clear_spte_count.
447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469
 */
static u64 __get_spte_lockless(u64 *sptep)
{
	struct kvm_mmu_page *sp =  page_header(__pa(sptep));
	union split_spte spte, *orig = (union split_spte *)sptep;
	int count;

retry:
	count = sp->clear_spte_count;
	smp_rmb();

	spte.spte_low = orig->spte_low;
	smp_rmb();

	spte.spte_high = orig->spte_high;
	smp_rmb();

	if (unlikely(spte.spte_low != orig->spte_low ||
	      count != sp->clear_spte_count))
		goto retry;

	return spte.spte;
}
470 471
#endif

472 473
static bool spte_is_locklessly_modifiable(u64 spte)
{
474 475
	return (spte & (SPTE_HOST_WRITEABLE | SPTE_MMU_WRITEABLE)) ==
		(SPTE_HOST_WRITEABLE | SPTE_MMU_WRITEABLE);
476 477
}

478 479
static bool spte_has_volatile_bits(u64 spte)
{
480 481 482 483 484 485 486 487 488
	/*
	 * Always atomicly update spte if it can be updated
	 * out of mmu-lock, it can ensure dirty bit is not lost,
	 * also, it can help us to get a stable is_writable_pte()
	 * to ensure tlb flush is not missed.
	 */
	if (spte_is_locklessly_modifiable(spte))
		return true;

489 490 491 492 493 494
	if (!shadow_accessed_mask)
		return false;

	if (!is_shadow_present_pte(spte))
		return false;

495 496
	if ((spte & shadow_accessed_mask) &&
	      (!is_writable_pte(spte) || (spte & shadow_dirty_mask)))
497 498 499 500 501
		return false;

	return true;
}

502 503 504 505 506
static bool spte_is_bit_cleared(u64 old_spte, u64 new_spte, u64 bit_mask)
{
	return (old_spte & bit_mask) && !(new_spte & bit_mask);
}

507 508 509 510 511
static bool spte_is_bit_changed(u64 old_spte, u64 new_spte, u64 bit_mask)
{
	return (old_spte & bit_mask) != (new_spte & bit_mask);
}

512 513 514 515 516 517 518 519 520 521 522 523 524 525
/* Rules for using mmu_spte_set:
 * Set the sptep from nonpresent to present.
 * Note: the sptep being assigned *must* be either not present
 * or in a state where the hardware will not attempt to update
 * the spte.
 */
static void mmu_spte_set(u64 *sptep, u64 new_spte)
{
	WARN_ON(is_shadow_present_pte(*sptep));
	__set_spte(sptep, new_spte);
}

/* Rules for using mmu_spte_update:
 * Update the state bits, it means the mapped pfn is not changged.
526 527 528 529 530 531
 *
 * Whenever we overwrite a writable spte with a read-only one we
 * should flush remote TLBs. Otherwise rmap_write_protect
 * will find a read-only spte, even though the writable spte
 * might be cached on a CPU's TLB, the return value indicates this
 * case.
532
 */
533
static bool mmu_spte_update(u64 *sptep, u64 new_spte)
534
{
535
	u64 old_spte = *sptep;
536
	bool ret = false;
537

538
	WARN_ON(!is_shadow_present_pte(new_spte));
539

540 541 542 543
	if (!is_shadow_present_pte(old_spte)) {
		mmu_spte_set(sptep, new_spte);
		return ret;
	}
544

545
	if (!spte_has_volatile_bits(old_spte))
546
		__update_clear_spte_fast(sptep, new_spte);
547
	else
548
		old_spte = __update_clear_spte_slow(sptep, new_spte);
549

550 551 552 553 554
	/*
	 * For the spte updated out of mmu-lock is safe, since
	 * we always atomicly update it, see the comments in
	 * spte_has_volatile_bits().
	 */
555 556
	if (spte_is_locklessly_modifiable(old_spte) &&
	      !is_writable_pte(new_spte))
557 558
		ret = true;

559
	if (!shadow_accessed_mask)
560
		return ret;
561

562 563 564 565 566 567 568 569
	/*
	 * Flush TLB when accessed/dirty bits are changed in the page tables,
	 * to guarantee consistency between TLB and page tables.
	 */
	if (spte_is_bit_changed(old_spte, new_spte,
                                shadow_accessed_mask | shadow_dirty_mask))
		ret = true;

570 571 572 573
	if (spte_is_bit_cleared(old_spte, new_spte, shadow_accessed_mask))
		kvm_set_pfn_accessed(spte_to_pfn(old_spte));
	if (spte_is_bit_cleared(old_spte, new_spte, shadow_dirty_mask))
		kvm_set_pfn_dirty(spte_to_pfn(old_spte));
574 575

	return ret;
576 577
}

578 579 580 581 582 583 584 585 586 587 588
/*
 * Rules for using mmu_spte_clear_track_bits:
 * It sets the sptep from present to nonpresent, and track the
 * state bits, it is used to clear the last level sptep.
 */
static int mmu_spte_clear_track_bits(u64 *sptep)
{
	pfn_t pfn;
	u64 old_spte = *sptep;

	if (!spte_has_volatile_bits(old_spte))
589
		__update_clear_spte_fast(sptep, 0ull);
590
	else
591
		old_spte = __update_clear_spte_slow(sptep, 0ull);
592

593
	if (!is_shadow_present_pte(old_spte))
594 595 596
		return 0;

	pfn = spte_to_pfn(old_spte);
597 598 599 600 601 602

	/*
	 * KVM does not hold the refcount of the page used by
	 * kvm mmu, before reclaiming the page, we should
	 * unmap it from mmu first.
	 */
603
	WARN_ON(!kvm_is_reserved_pfn(pfn) && !page_count(pfn_to_page(pfn)));
604

605 606 607 608 609 610 611 612 613 614 615 616 617 618
	if (!shadow_accessed_mask || old_spte & shadow_accessed_mask)
		kvm_set_pfn_accessed(pfn);
	if (!shadow_dirty_mask || (old_spte & shadow_dirty_mask))
		kvm_set_pfn_dirty(pfn);
	return 1;
}

/*
 * Rules for using mmu_spte_clear_no_track:
 * Directly clear spte without caring the state bits of sptep,
 * it is used to set the upper level spte.
 */
static void mmu_spte_clear_no_track(u64 *sptep)
{
619
	__update_clear_spte_fast(sptep, 0ull);
620 621
}

622 623 624 625 626 627 628
static u64 mmu_spte_get_lockless(u64 *sptep)
{
	return __get_spte_lockless(sptep);
}

static void walk_shadow_page_lockless_begin(struct kvm_vcpu *vcpu)
{
629 630 631 632 633 634 635 636 637 638 639
	/*
	 * Prevent page table teardown by making any free-er wait during
	 * kvm_flush_remote_tlbs() IPI to all active vcpus.
	 */
	local_irq_disable();
	vcpu->mode = READING_SHADOW_PAGE_TABLES;
	/*
	 * Make sure a following spte read is not reordered ahead of the write
	 * to vcpu->mode.
	 */
	smp_mb();
640 641 642 643
}

static void walk_shadow_page_lockless_end(struct kvm_vcpu *vcpu)
{
644 645 646 647 648 649 650 651
	/*
	 * Make sure the write to vcpu->mode is not reordered in front of
	 * reads to sptes.  If it does, kvm_commit_zap_page() can see us
	 * OUTSIDE_GUEST_MODE and proceed to free the shadow page table.
	 */
	smp_mb();
	vcpu->mode = OUTSIDE_GUEST_MODE;
	local_irq_enable();
652 653
}

654
static int mmu_topup_memory_cache(struct kvm_mmu_memory_cache *cache,
655
				  struct kmem_cache *base_cache, int min)
656 657 658 659
{
	void *obj;

	if (cache->nobjs >= min)
660
		return 0;
661
	while (cache->nobjs < ARRAY_SIZE(cache->objects)) {
662
		obj = kmem_cache_zalloc(base_cache, GFP_KERNEL);
663
		if (!obj)
664
			return -ENOMEM;
665 666
		cache->objects[cache->nobjs++] = obj;
	}
667
	return 0;
668 669
}

670 671 672 673 674
static int mmu_memory_cache_free_objects(struct kvm_mmu_memory_cache *cache)
{
	return cache->nobjs;
}

675 676
static void mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc,
				  struct kmem_cache *cache)
677 678
{
	while (mc->nobjs)
679
		kmem_cache_free(cache, mc->objects[--mc->nobjs]);
680 681
}

A
Avi Kivity 已提交
682
static int mmu_topup_memory_cache_page(struct kvm_mmu_memory_cache *cache,
683
				       int min)
A
Avi Kivity 已提交
684
{
685
	void *page;
A
Avi Kivity 已提交
686 687 688 689

	if (cache->nobjs >= min)
		return 0;
	while (cache->nobjs < ARRAY_SIZE(cache->objects)) {
690
		page = (void *)__get_free_page(GFP_KERNEL);
A
Avi Kivity 已提交
691 692
		if (!page)
			return -ENOMEM;
693
		cache->objects[cache->nobjs++] = page;
A
Avi Kivity 已提交
694 695 696 697 698 699 700
	}
	return 0;
}

static void mmu_free_memory_cache_page(struct kvm_mmu_memory_cache *mc)
{
	while (mc->nobjs)
701
		free_page((unsigned long)mc->objects[--mc->nobjs]);
A
Avi Kivity 已提交
702 703
}

704
static int mmu_topup_memory_caches(struct kvm_vcpu *vcpu)
705
{
706 707
	int r;

708
	r = mmu_topup_memory_cache(&vcpu->arch.mmu_pte_list_desc_cache,
709
				   pte_list_desc_cache, 8 + PTE_PREFETCH_NUM);
710 711
	if (r)
		goto out;
712
	r = mmu_topup_memory_cache_page(&vcpu->arch.mmu_page_cache, 8);
713 714
	if (r)
		goto out;
715
	r = mmu_topup_memory_cache(&vcpu->arch.mmu_page_header_cache,
716
				   mmu_page_header_cache, 4);
717 718
out:
	return r;
719 720 721 722
}

static void mmu_free_memory_caches(struct kvm_vcpu *vcpu)
{
723 724
	mmu_free_memory_cache(&vcpu->arch.mmu_pte_list_desc_cache,
				pte_list_desc_cache);
725
	mmu_free_memory_cache_page(&vcpu->arch.mmu_page_cache);
726 727
	mmu_free_memory_cache(&vcpu->arch.mmu_page_header_cache,
				mmu_page_header_cache);
728 729
}

730
static void *mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc)
731 732 733 734 735 736 737 738
{
	void *p;

	BUG_ON(!mc->nobjs);
	p = mc->objects[--mc->nobjs];
	return p;
}

739
static struct pte_list_desc *mmu_alloc_pte_list_desc(struct kvm_vcpu *vcpu)
740
{
741
	return mmu_memory_cache_alloc(&vcpu->arch.mmu_pte_list_desc_cache);
742 743
}

744
static void mmu_free_pte_list_desc(struct pte_list_desc *pte_list_desc)
745
{
746
	kmem_cache_free(pte_list_desc_cache, pte_list_desc);
747 748
}

749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764
static gfn_t kvm_mmu_page_get_gfn(struct kvm_mmu_page *sp, int index)
{
	if (!sp->role.direct)
		return sp->gfns[index];

	return sp->gfn + (index << ((sp->role.level - 1) * PT64_LEVEL_BITS));
}

static void kvm_mmu_page_set_gfn(struct kvm_mmu_page *sp, int index, gfn_t gfn)
{
	if (sp->role.direct)
		BUG_ON(gfn != kvm_mmu_page_get_gfn(sp, index));
	else
		sp->gfns[index] = gfn;
}

M
Marcelo Tosatti 已提交
765
/*
766 767
 * Return the pointer to the large page information for a given gfn,
 * handling slots that are not large page aligned.
M
Marcelo Tosatti 已提交
768
 */
769 770 771
static struct kvm_lpage_info *lpage_info_slot(gfn_t gfn,
					      struct kvm_memory_slot *slot,
					      int level)
M
Marcelo Tosatti 已提交
772 773 774
{
	unsigned long idx;

775
	idx = gfn_to_index(gfn, slot->base_gfn, level);
776
	return &slot->arch.lpage_info[level - 2][idx];
M
Marcelo Tosatti 已提交
777 778
}

779
static void account_shadowed(struct kvm *kvm, struct kvm_mmu_page *sp)
M
Marcelo Tosatti 已提交
780
{
781
	struct kvm_memslots *slots;
782
	struct kvm_memory_slot *slot;
783
	struct kvm_lpage_info *linfo;
784
	gfn_t gfn;
785
	int i;
M
Marcelo Tosatti 已提交
786

787
	gfn = sp->gfn;
788 789
	slots = kvm_memslots_for_spte_role(kvm, sp->role);
	slot = __gfn_to_memslot(slots, gfn);
790
	for (i = PT_DIRECTORY_LEVEL; i <= PT_MAX_HUGEPAGE_LEVEL; ++i) {
791 792
		linfo = lpage_info_slot(gfn, slot, i);
		linfo->write_count += 1;
793
	}
794
	kvm->arch.indirect_shadow_pages++;
M
Marcelo Tosatti 已提交
795 796
}

797
static void unaccount_shadowed(struct kvm *kvm, struct kvm_mmu_page *sp)
M
Marcelo Tosatti 已提交
798
{
799
	struct kvm_memslots *slots;
800
	struct kvm_memory_slot *slot;
801
	struct kvm_lpage_info *linfo;
802
	gfn_t gfn;
803
	int i;
M
Marcelo Tosatti 已提交
804

805
	gfn = sp->gfn;
806 807
	slots = kvm_memslots_for_spte_role(kvm, sp->role);
	slot = __gfn_to_memslot(slots, gfn);
808
	for (i = PT_DIRECTORY_LEVEL; i <= PT_MAX_HUGEPAGE_LEVEL; ++i) {
809 810 811
		linfo = lpage_info_slot(gfn, slot, i);
		linfo->write_count -= 1;
		WARN_ON(linfo->write_count < 0);
812
	}
813
	kvm->arch.indirect_shadow_pages--;
M
Marcelo Tosatti 已提交
814 815
}

816 817
static int __has_wrprotected_page(gfn_t gfn, int level,
				  struct kvm_memory_slot *slot)
M
Marcelo Tosatti 已提交
818
{
819
	struct kvm_lpage_info *linfo;
M
Marcelo Tosatti 已提交
820 821

	if (slot) {
822 823
		linfo = lpage_info_slot(gfn, slot, level);
		return linfo->write_count;
M
Marcelo Tosatti 已提交
824 825 826 827 828
	}

	return 1;
}

829 830 831 832 833 834 835 836
static int has_wrprotected_page(struct kvm_vcpu *vcpu, gfn_t gfn, int level)
{
	struct kvm_memory_slot *slot;

	slot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);
	return __has_wrprotected_page(gfn, level, slot);
}

837
static int host_mapping_level(struct kvm *kvm, gfn_t gfn)
M
Marcelo Tosatti 已提交
838
{
J
Joerg Roedel 已提交
839
	unsigned long page_size;
840
	int i, ret = 0;
M
Marcelo Tosatti 已提交
841

J
Joerg Roedel 已提交
842
	page_size = kvm_host_page_size(kvm, gfn);
M
Marcelo Tosatti 已提交
843

844
	for (i = PT_PAGE_TABLE_LEVEL; i <= PT_MAX_HUGEPAGE_LEVEL; ++i) {
845 846 847 848 849 850
		if (page_size >= KVM_HPAGE_SIZE(i))
			ret = i;
		else
			break;
	}

851
	return ret;
M
Marcelo Tosatti 已提交
852 853
}

854 855 856 857 858 859 860 861 862 863 864
static inline bool memslot_valid_for_gpte(struct kvm_memory_slot *slot,
					  bool no_dirty_log)
{
	if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
		return false;
	if (no_dirty_log && slot->dirty_bitmap)
		return false;

	return true;
}

865 866 867
static struct kvm_memory_slot *
gfn_to_memslot_dirty_bitmap(struct kvm_vcpu *vcpu, gfn_t gfn,
			    bool no_dirty_log)
M
Marcelo Tosatti 已提交
868 869
{
	struct kvm_memory_slot *slot;
870

871
	slot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);
872
	if (!memslot_valid_for_gpte(slot, no_dirty_log))
873 874 875 876 877
		slot = NULL;

	return slot;
}

878 879
static int mapping_level(struct kvm_vcpu *vcpu, gfn_t large_gfn,
			 bool *force_pt_level)
880 881
{
	int host_level, level, max_level;
882 883
	struct kvm_memory_slot *slot;

884 885
	if (unlikely(*force_pt_level))
		return PT_PAGE_TABLE_LEVEL;
M
Marcelo Tosatti 已提交
886

887 888
	slot = kvm_vcpu_gfn_to_memslot(vcpu, large_gfn);
	*force_pt_level = !memslot_valid_for_gpte(slot, true);
889 890 891
	if (unlikely(*force_pt_level))
		return PT_PAGE_TABLE_LEVEL;

892 893 894 895 896
	host_level = host_mapping_level(vcpu->kvm, large_gfn);

	if (host_level == PT_PAGE_TABLE_LEVEL)
		return host_level;

X
Xiao Guangrong 已提交
897
	max_level = min(kvm_x86_ops->get_lpage_level(), host_level);
898 899

	for (level = PT_DIRECTORY_LEVEL; level <= max_level; ++level)
900
		if (__has_wrprotected_page(large_gfn, level, slot))
901 902 903
			break;

	return level - 1;
M
Marcelo Tosatti 已提交
904 905
}

906
/*
907
 * About rmap_head encoding:
908
 *
909 910
 * If the bit zero of rmap_head->val is clear, then it points to the only spte
 * in this rmap chain. Otherwise, (rmap_head->val & ~1) points to a struct
911
 * pte_list_desc containing more mappings.
912 913 914 915
 */

/*
 * Returns the number of pointers in the rmap chain, not counting the new one.
916
 */
917
static int pte_list_add(struct kvm_vcpu *vcpu, u64 *spte,
918
			struct kvm_rmap_head *rmap_head)
919
{
920
	struct pte_list_desc *desc;
921
	int i, count = 0;
922

923
	if (!rmap_head->val) {
924
		rmap_printk("pte_list_add: %p %llx 0->1\n", spte, *spte);
925 926
		rmap_head->val = (unsigned long)spte;
	} else if (!(rmap_head->val & 1)) {
927 928
		rmap_printk("pte_list_add: %p %llx 1->many\n", spte, *spte);
		desc = mmu_alloc_pte_list_desc(vcpu);
929
		desc->sptes[0] = (u64 *)rmap_head->val;
A
Avi Kivity 已提交
930
		desc->sptes[1] = spte;
931
		rmap_head->val = (unsigned long)desc | 1;
932
		++count;
933
	} else {
934
		rmap_printk("pte_list_add: %p %llx many->many\n", spte, *spte);
935
		desc = (struct pte_list_desc *)(rmap_head->val & ~1ul);
936
		while (desc->sptes[PTE_LIST_EXT-1] && desc->more) {
937
			desc = desc->more;
938
			count += PTE_LIST_EXT;
939
		}
940 941
		if (desc->sptes[PTE_LIST_EXT-1]) {
			desc->more = mmu_alloc_pte_list_desc(vcpu);
942 943
			desc = desc->more;
		}
A
Avi Kivity 已提交
944
		for (i = 0; desc->sptes[i]; ++i)
945
			++count;
A
Avi Kivity 已提交
946
		desc->sptes[i] = spte;
947
	}
948
	return count;
949 950
}

951
static void
952 953 954
pte_list_desc_remove_entry(struct kvm_rmap_head *rmap_head,
			   struct pte_list_desc *desc, int i,
			   struct pte_list_desc *prev_desc)
955 956 957
{
	int j;

958
	for (j = PTE_LIST_EXT - 1; !desc->sptes[j] && j > i; --j)
959
		;
A
Avi Kivity 已提交
960 961
	desc->sptes[i] = desc->sptes[j];
	desc->sptes[j] = NULL;
962 963 964
	if (j != 0)
		return;
	if (!prev_desc && !desc->more)
965
		rmap_head->val = (unsigned long)desc->sptes[0];
966 967 968 969
	else
		if (prev_desc)
			prev_desc->more = desc->more;
		else
970
			rmap_head->val = (unsigned long)desc->more | 1;
971
	mmu_free_pte_list_desc(desc);
972 973
}

974
static void pte_list_remove(u64 *spte, struct kvm_rmap_head *rmap_head)
975
{
976 977
	struct pte_list_desc *desc;
	struct pte_list_desc *prev_desc;
978 979
	int i;

980
	if (!rmap_head->val) {
981
		printk(KERN_ERR "pte_list_remove: %p 0->BUG\n", spte);
982
		BUG();
983
	} else if (!(rmap_head->val & 1)) {
984
		rmap_printk("pte_list_remove:  %p 1->0\n", spte);
985
		if ((u64 *)rmap_head->val != spte) {
986
			printk(KERN_ERR "pte_list_remove:  %p 1->BUG\n", spte);
987 988
			BUG();
		}
989
		rmap_head->val = 0;
990
	} else {
991
		rmap_printk("pte_list_remove:  %p many->many\n", spte);
992
		desc = (struct pte_list_desc *)(rmap_head->val & ~1ul);
993 994
		prev_desc = NULL;
		while (desc) {
995
			for (i = 0; i < PTE_LIST_EXT && desc->sptes[i]; ++i) {
A
Avi Kivity 已提交
996
				if (desc->sptes[i] == spte) {
997 998
					pte_list_desc_remove_entry(rmap_head,
							desc, i, prev_desc);
999 1000
					return;
				}
1001
			}
1002 1003 1004
			prev_desc = desc;
			desc = desc->more;
		}
1005
		pr_err("pte_list_remove: %p many->many\n", spte);
1006 1007 1008 1009
		BUG();
	}
}

1010
typedef void (*pte_list_walk_fn) (u64 *spte);
1011
static void pte_list_walk(struct kvm_rmap_head *rmap_head, pte_list_walk_fn fn)
1012 1013 1014 1015
{
	struct pte_list_desc *desc;
	int i;

1016
	if (!rmap_head->val)
1017 1018
		return;

1019 1020
	if (!(rmap_head->val & 1))
		return fn((u64 *)rmap_head->val);
1021

1022
	desc = (struct pte_list_desc *)(rmap_head->val & ~1ul);
1023 1024 1025 1026 1027 1028 1029
	while (desc) {
		for (i = 0; i < PTE_LIST_EXT && desc->sptes[i]; ++i)
			fn(desc->sptes[i]);
		desc = desc->more;
	}
}

1030 1031
static struct kvm_rmap_head *__gfn_to_rmap(gfn_t gfn, int level,
					   struct kvm_memory_slot *slot)
1032
{
1033
	unsigned long idx;
1034

1035
	idx = gfn_to_index(gfn, slot->base_gfn, level);
1036
	return &slot->arch.rmap[level - PT_PAGE_TABLE_LEVEL][idx];
1037 1038
}

1039 1040
static struct kvm_rmap_head *gfn_to_rmap(struct kvm *kvm, gfn_t gfn,
					 struct kvm_mmu_page *sp)
1041
{
1042
	struct kvm_memslots *slots;
1043 1044
	struct kvm_memory_slot *slot;

1045 1046
	slots = kvm_memslots_for_spte_role(kvm, sp->role);
	slot = __gfn_to_memslot(slots, gfn);
1047
	return __gfn_to_rmap(gfn, sp->role.level, slot);
1048 1049
}

1050 1051 1052 1053 1054 1055 1056 1057
static bool rmap_can_add(struct kvm_vcpu *vcpu)
{
	struct kvm_mmu_memory_cache *cache;

	cache = &vcpu->arch.mmu_pte_list_desc_cache;
	return mmu_memory_cache_free_objects(cache);
}

1058 1059 1060
static int rmap_add(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn)
{
	struct kvm_mmu_page *sp;
1061
	struct kvm_rmap_head *rmap_head;
1062 1063 1064

	sp = page_header(__pa(spte));
	kvm_mmu_page_set_gfn(sp, spte - sp->spt, gfn);
1065 1066
	rmap_head = gfn_to_rmap(vcpu->kvm, gfn, sp);
	return pte_list_add(vcpu, spte, rmap_head);
1067 1068 1069 1070 1071 1072
}

static void rmap_remove(struct kvm *kvm, u64 *spte)
{
	struct kvm_mmu_page *sp;
	gfn_t gfn;
1073
	struct kvm_rmap_head *rmap_head;
1074 1075 1076

	sp = page_header(__pa(spte));
	gfn = kvm_mmu_page_get_gfn(sp, spte - sp->spt);
1077 1078
	rmap_head = gfn_to_rmap(kvm, gfn, sp);
	pte_list_remove(spte, rmap_head);
1079 1080
}

1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097
/*
 * Used by the following functions to iterate through the sptes linked by a
 * rmap.  All fields are private and not assumed to be used outside.
 */
struct rmap_iterator {
	/* private fields */
	struct pte_list_desc *desc;	/* holds the sptep if not NULL */
	int pos;			/* index of the sptep */
};

/*
 * Iteration must be started by this function.  This should also be used after
 * removing/dropping sptes from the rmap link because in such cases the
 * information in the itererator may not be valid.
 *
 * Returns sptep if found, NULL otherwise.
 */
1098 1099
static u64 *rmap_get_first(struct kvm_rmap_head *rmap_head,
			   struct rmap_iterator *iter)
1100
{
1101
	if (!rmap_head->val)
1102 1103
		return NULL;

1104
	if (!(rmap_head->val & 1)) {
1105
		iter->desc = NULL;
1106
		return (u64 *)rmap_head->val;
1107 1108
	}

1109
	iter->desc = (struct pte_list_desc *)(rmap_head->val & ~1ul);
1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142
	iter->pos = 0;
	return iter->desc->sptes[iter->pos];
}

/*
 * Must be used with a valid iterator: e.g. after rmap_get_first().
 *
 * Returns sptep if found, NULL otherwise.
 */
static u64 *rmap_get_next(struct rmap_iterator *iter)
{
	if (iter->desc) {
		if (iter->pos < PTE_LIST_EXT - 1) {
			u64 *sptep;

			++iter->pos;
			sptep = iter->desc->sptes[iter->pos];
			if (sptep)
				return sptep;
		}

		iter->desc = iter->desc->more;

		if (iter->desc) {
			iter->pos = 0;
			/* desc->sptes[0] cannot be NULL */
			return iter->desc->sptes[iter->pos];
		}
	}

	return NULL;
}

1143 1144 1145 1146
#define for_each_rmap_spte(_rmap_head_, _iter_, _spte_)			\
	for (_spte_ = rmap_get_first(_rmap_head_, _iter_);		\
	     _spte_ && ({BUG_ON(!is_shadow_present_pte(*_spte_)); 1;});	\
	     _spte_ = rmap_get_next(_iter_))
1147

1148
static void drop_spte(struct kvm *kvm, u64 *sptep)
1149
{
1150
	if (mmu_spte_clear_track_bits(sptep))
1151
		rmap_remove(kvm, sptep);
A
Avi Kivity 已提交
1152 1153
}

1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174

static bool __drop_large_spte(struct kvm *kvm, u64 *sptep)
{
	if (is_large_pte(*sptep)) {
		WARN_ON(page_header(__pa(sptep))->role.level ==
			PT_PAGE_TABLE_LEVEL);
		drop_spte(kvm, sptep);
		--kvm->stat.lpages;
		return true;
	}

	return false;
}

static void drop_large_spte(struct kvm_vcpu *vcpu, u64 *sptep)
{
	if (__drop_large_spte(vcpu->kvm, sptep))
		kvm_flush_remote_tlbs(vcpu->kvm);
}

/*
1175
 * Write-protect on the specified @sptep, @pt_protect indicates whether
1176
 * spte write-protection is caused by protecting shadow page table.
1177
 *
T
Tiejun Chen 已提交
1178
 * Note: write protection is difference between dirty logging and spte
1179 1180 1181 1182 1183
 * protection:
 * - for dirty logging, the spte can be set to writable at anytime if
 *   its dirty bitmap is properly set.
 * - for spte protection, the spte can be writable only after unsync-ing
 *   shadow page.
1184
 *
1185
 * Return true if tlb need be flushed.
1186
 */
1187
static bool spte_write_protect(struct kvm *kvm, u64 *sptep, bool pt_protect)
1188 1189 1190
{
	u64 spte = *sptep;

1191 1192
	if (!is_writable_pte(spte) &&
	      !(pt_protect && spte_is_locklessly_modifiable(spte)))
1193 1194 1195 1196
		return false;

	rmap_printk("rmap_write_protect: spte %p %llx\n", sptep, *sptep);

1197 1198
	if (pt_protect)
		spte &= ~SPTE_MMU_WRITEABLE;
1199
	spte = spte & ~PT_WRITABLE_MASK;
1200

1201
	return mmu_spte_update(sptep, spte);
1202 1203
}

1204 1205
static bool __rmap_write_protect(struct kvm *kvm,
				 struct kvm_rmap_head *rmap_head,
1206
				 bool pt_protect)
1207
{
1208 1209
	u64 *sptep;
	struct rmap_iterator iter;
1210
	bool flush = false;
1211

1212
	for_each_rmap_spte(rmap_head, &iter, sptep)
1213
		flush |= spte_write_protect(kvm, sptep, pt_protect);
1214

1215
	return flush;
1216 1217
}

1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228
static bool spte_clear_dirty(struct kvm *kvm, u64 *sptep)
{
	u64 spte = *sptep;

	rmap_printk("rmap_clear_dirty: spte %p %llx\n", sptep, *sptep);

	spte &= ~shadow_dirty_mask;

	return mmu_spte_update(sptep, spte);
}

1229
static bool __rmap_clear_dirty(struct kvm *kvm, struct kvm_rmap_head *rmap_head)
1230 1231 1232 1233 1234
{
	u64 *sptep;
	struct rmap_iterator iter;
	bool flush = false;

1235
	for_each_rmap_spte(rmap_head, &iter, sptep)
1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251
		flush |= spte_clear_dirty(kvm, sptep);

	return flush;
}

static bool spte_set_dirty(struct kvm *kvm, u64 *sptep)
{
	u64 spte = *sptep;

	rmap_printk("rmap_set_dirty: spte %p %llx\n", sptep, *sptep);

	spte |= shadow_dirty_mask;

	return mmu_spte_update(sptep, spte);
}

1252
static bool __rmap_set_dirty(struct kvm *kvm, struct kvm_rmap_head *rmap_head)
1253 1254 1255 1256 1257
{
	u64 *sptep;
	struct rmap_iterator iter;
	bool flush = false;

1258
	for_each_rmap_spte(rmap_head, &iter, sptep)
1259 1260 1261 1262 1263
		flush |= spte_set_dirty(kvm, sptep);

	return flush;
}

1264
/**
1265
 * kvm_mmu_write_protect_pt_masked - write protect selected PT level pages
1266 1267 1268 1269 1270 1271 1272 1273
 * @kvm: kvm instance
 * @slot: slot to protect
 * @gfn_offset: start of the BITS_PER_LONG pages we care about
 * @mask: indicates which pages we should protect
 *
 * Used when we do not need to care about huge page mappings: e.g. during dirty
 * logging we do not have any such mappings.
 */
1274
static void kvm_mmu_write_protect_pt_masked(struct kvm *kvm,
1275 1276
				     struct kvm_memory_slot *slot,
				     gfn_t gfn_offset, unsigned long mask)
1277
{
1278
	struct kvm_rmap_head *rmap_head;
1279

1280
	while (mask) {
1281 1282 1283
		rmap_head = __gfn_to_rmap(slot->base_gfn + gfn_offset + __ffs(mask),
					  PT_PAGE_TABLE_LEVEL, slot);
		__rmap_write_protect(kvm, rmap_head, false);
M
Marcelo Tosatti 已提交
1284

1285 1286 1287
		/* clear the first set bit */
		mask &= mask - 1;
	}
1288 1289
}

1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302
/**
 * kvm_mmu_clear_dirty_pt_masked - clear MMU D-bit for PT level pages
 * @kvm: kvm instance
 * @slot: slot to clear D-bit
 * @gfn_offset: start of the BITS_PER_LONG pages we care about
 * @mask: indicates which pages we should clear D-bit
 *
 * Used for PML to re-log the dirty GPAs after userspace querying dirty_bitmap.
 */
void kvm_mmu_clear_dirty_pt_masked(struct kvm *kvm,
				     struct kvm_memory_slot *slot,
				     gfn_t gfn_offset, unsigned long mask)
{
1303
	struct kvm_rmap_head *rmap_head;
1304 1305

	while (mask) {
1306 1307 1308
		rmap_head = __gfn_to_rmap(slot->base_gfn + gfn_offset + __ffs(mask),
					  PT_PAGE_TABLE_LEVEL, slot);
		__rmap_clear_dirty(kvm, rmap_head);
1309 1310 1311 1312 1313 1314 1315

		/* clear the first set bit */
		mask &= mask - 1;
	}
}
EXPORT_SYMBOL_GPL(kvm_mmu_clear_dirty_pt_masked);

1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329
/**
 * kvm_arch_mmu_enable_log_dirty_pt_masked - enable dirty logging for selected
 * PT level pages.
 *
 * It calls kvm_mmu_write_protect_pt_masked to write protect selected pages to
 * enable dirty logging for them.
 *
 * Used when we do not need to care about huge page mappings: e.g. during dirty
 * logging we do not have any such mappings.
 */
void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
				struct kvm_memory_slot *slot,
				gfn_t gfn_offset, unsigned long mask)
{
1330 1331 1332 1333 1334
	if (kvm_x86_ops->enable_log_dirty_pt_masked)
		kvm_x86_ops->enable_log_dirty_pt_masked(kvm, slot, gfn_offset,
				mask);
	else
		kvm_mmu_write_protect_pt_masked(kvm, slot, gfn_offset, mask);
1335 1336
}

1337
static bool rmap_write_protect(struct kvm_vcpu *vcpu, u64 gfn)
1338 1339
{
	struct kvm_memory_slot *slot;
1340
	struct kvm_rmap_head *rmap_head;
1341
	int i;
1342
	bool write_protected = false;
1343

1344
	slot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);
1345

1346
	for (i = PT_PAGE_TABLE_LEVEL; i <= PT_MAX_HUGEPAGE_LEVEL; ++i) {
1347 1348
		rmap_head = __gfn_to_rmap(gfn, i, slot);
		write_protected |= __rmap_write_protect(vcpu->kvm, rmap_head, true);
1349 1350 1351
	}

	return write_protected;
1352 1353
}

1354
static bool kvm_zap_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head)
1355
{
1356 1357
	u64 *sptep;
	struct rmap_iterator iter;
1358
	bool flush = false;
1359

1360
	while ((sptep = rmap_get_first(rmap_head, &iter))) {
1361
		BUG_ON(!(*sptep & PT_PRESENT_MASK));
1362
		rmap_printk("%s: spte %p %llx.\n", __func__, sptep, *sptep);
1363 1364

		drop_spte(kvm, sptep);
1365
		flush = true;
1366
	}
1367

1368 1369 1370
	return flush;
}

1371
static int kvm_unmap_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
1372 1373 1374
			   struct kvm_memory_slot *slot, gfn_t gfn, int level,
			   unsigned long data)
{
1375
	return kvm_zap_rmapp(kvm, rmap_head);
1376 1377
}

1378
static int kvm_set_pte_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
1379 1380
			     struct kvm_memory_slot *slot, gfn_t gfn, int level,
			     unsigned long data)
1381
{
1382 1383
	u64 *sptep;
	struct rmap_iterator iter;
1384
	int need_flush = 0;
1385
	u64 new_spte;
1386 1387 1388 1389 1390
	pte_t *ptep = (pte_t *)data;
	pfn_t new_pfn;

	WARN_ON(pte_huge(*ptep));
	new_pfn = pte_pfn(*ptep);
1391

1392
restart:
1393
	for_each_rmap_spte(rmap_head, &iter, sptep) {
1394 1395
		rmap_printk("kvm_set_pte_rmapp: spte %p %llx gfn %llx (%d)\n",
			     sptep, *sptep, gfn, level);
1396

1397
		need_flush = 1;
1398

1399
		if (pte_write(*ptep)) {
1400
			drop_spte(kvm, sptep);
1401
			goto restart;
1402
		} else {
1403
			new_spte = *sptep & ~PT64_BASE_ADDR_MASK;
1404 1405 1406 1407
			new_spte |= (u64)new_pfn << PAGE_SHIFT;

			new_spte &= ~PT_WRITABLE_MASK;
			new_spte &= ~SPTE_HOST_WRITEABLE;
1408
			new_spte &= ~shadow_accessed_mask;
1409 1410 1411

			mmu_spte_clear_track_bits(sptep);
			mmu_spte_set(sptep, new_spte);
1412 1413
		}
	}
1414

1415 1416 1417 1418 1419 1420
	if (need_flush)
		kvm_flush_remote_tlbs(kvm);

	return 0;
}

1421 1422 1423 1424 1425 1426 1427 1428 1429 1430
struct slot_rmap_walk_iterator {
	/* input fields. */
	struct kvm_memory_slot *slot;
	gfn_t start_gfn;
	gfn_t end_gfn;
	int start_level;
	int end_level;

	/* output fields. */
	gfn_t gfn;
1431
	struct kvm_rmap_head *rmap;
1432 1433 1434
	int level;

	/* private field. */
1435
	struct kvm_rmap_head *end_rmap;
1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488
};

static void
rmap_walk_init_level(struct slot_rmap_walk_iterator *iterator, int level)
{
	iterator->level = level;
	iterator->gfn = iterator->start_gfn;
	iterator->rmap = __gfn_to_rmap(iterator->gfn, level, iterator->slot);
	iterator->end_rmap = __gfn_to_rmap(iterator->end_gfn, level,
					   iterator->slot);
}

static void
slot_rmap_walk_init(struct slot_rmap_walk_iterator *iterator,
		    struct kvm_memory_slot *slot, int start_level,
		    int end_level, gfn_t start_gfn, gfn_t end_gfn)
{
	iterator->slot = slot;
	iterator->start_level = start_level;
	iterator->end_level = end_level;
	iterator->start_gfn = start_gfn;
	iterator->end_gfn = end_gfn;

	rmap_walk_init_level(iterator, iterator->start_level);
}

static bool slot_rmap_walk_okay(struct slot_rmap_walk_iterator *iterator)
{
	return !!iterator->rmap;
}

static void slot_rmap_walk_next(struct slot_rmap_walk_iterator *iterator)
{
	if (++iterator->rmap <= iterator->end_rmap) {
		iterator->gfn += (1UL << KVM_HPAGE_GFN_SHIFT(iterator->level));
		return;
	}

	if (++iterator->level > iterator->end_level) {
		iterator->rmap = NULL;
		return;
	}

	rmap_walk_init_level(iterator, iterator->level);
}

#define for_each_slot_rmap_range(_slot_, _start_level_, _end_level_,	\
	   _start_gfn, _end_gfn, _iter_)				\
	for (slot_rmap_walk_init(_iter_, _slot_, _start_level_,		\
				 _end_level_, _start_gfn, _end_gfn);	\
	     slot_rmap_walk_okay(_iter_);				\
	     slot_rmap_walk_next(_iter_))

1489 1490 1491 1492 1493
static int kvm_handle_hva_range(struct kvm *kvm,
				unsigned long start,
				unsigned long end,
				unsigned long data,
				int (*handler)(struct kvm *kvm,
1494
					       struct kvm_rmap_head *rmap_head,
1495
					       struct kvm_memory_slot *slot,
1496 1497
					       gfn_t gfn,
					       int level,
1498
					       unsigned long data))
1499
{
1500
	struct kvm_memslots *slots;
1501
	struct kvm_memory_slot *memslot;
1502 1503
	struct slot_rmap_walk_iterator iterator;
	int ret = 0;
1504
	int i;
1505

1506 1507 1508 1509 1510
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
		slots = __kvm_memslots(kvm, i);
		kvm_for_each_memslot(memslot, slots) {
			unsigned long hva_start, hva_end;
			gfn_t gfn_start, gfn_end;
1511

1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530
			hva_start = max(start, memslot->userspace_addr);
			hva_end = min(end, memslot->userspace_addr +
				      (memslot->npages << PAGE_SHIFT));
			if (hva_start >= hva_end)
				continue;
			/*
			 * {gfn(page) | page intersects with [hva_start, hva_end)} =
			 * {gfn_start, gfn_start+1, ..., gfn_end-1}.
			 */
			gfn_start = hva_to_gfn_memslot(hva_start, memslot);
			gfn_end = hva_to_gfn_memslot(hva_end + PAGE_SIZE - 1, memslot);

			for_each_slot_rmap_range(memslot, PT_PAGE_TABLE_LEVEL,
						 PT_MAX_HUGEPAGE_LEVEL,
						 gfn_start, gfn_end - 1,
						 &iterator)
				ret |= handler(kvm, iterator.rmap, memslot,
					       iterator.gfn, iterator.level, data);
		}
1531 1532
	}

1533
	return ret;
1534 1535
}

1536 1537
static int kvm_handle_hva(struct kvm *kvm, unsigned long hva,
			  unsigned long data,
1538 1539
			  int (*handler)(struct kvm *kvm,
					 struct kvm_rmap_head *rmap_head,
1540
					 struct kvm_memory_slot *slot,
1541
					 gfn_t gfn, int level,
1542 1543 1544
					 unsigned long data))
{
	return kvm_handle_hva_range(kvm, hva, hva + 1, data, handler);
1545 1546 1547 1548
}

int kvm_unmap_hva(struct kvm *kvm, unsigned long hva)
{
1549 1550 1551
	return kvm_handle_hva(kvm, hva, 0, kvm_unmap_rmapp);
}

1552 1553 1554 1555 1556
int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end)
{
	return kvm_handle_hva_range(kvm, start, end, 0, kvm_unmap_rmapp);
}

1557 1558
void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte)
{
F
Frederik Deweerdt 已提交
1559
	kvm_handle_hva(kvm, hva, (unsigned long)&pte, kvm_set_pte_rmapp);
1560 1561
}

1562
static int kvm_age_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
1563 1564
			 struct kvm_memory_slot *slot, gfn_t gfn, int level,
			 unsigned long data)
1565
{
1566
	u64 *sptep;
1567
	struct rmap_iterator uninitialized_var(iter);
1568 1569
	int young = 0;

A
Andres Lagar-Cavilla 已提交
1570
	BUG_ON(!shadow_accessed_mask);
1571

1572
	for_each_rmap_spte(rmap_head, &iter, sptep) {
1573
		if (*sptep & shadow_accessed_mask) {
1574
			young = 1;
1575 1576
			clear_bit((ffs(shadow_accessed_mask) - 1),
				 (unsigned long *)sptep);
1577
		}
1578
	}
1579

1580
	trace_kvm_age_page(gfn, level, slot, young);
1581 1582 1583
	return young;
}

1584
static int kvm_test_age_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
1585 1586
			      struct kvm_memory_slot *slot, gfn_t gfn,
			      int level, unsigned long data)
A
Andrea Arcangeli 已提交
1587
{
1588 1589
	u64 *sptep;
	struct rmap_iterator iter;
A
Andrea Arcangeli 已提交
1590 1591 1592 1593 1594 1595 1596 1597 1598 1599
	int young = 0;

	/*
	 * If there's no access bit in the secondary pte set by the
	 * hardware it's up to gup-fast/gup to set the access bit in
	 * the primary pte or in the page structure.
	 */
	if (!shadow_accessed_mask)
		goto out;

1600
	for_each_rmap_spte(rmap_head, &iter, sptep) {
1601
		if (*sptep & shadow_accessed_mask) {
A
Andrea Arcangeli 已提交
1602 1603 1604
			young = 1;
			break;
		}
1605
	}
A
Andrea Arcangeli 已提交
1606 1607 1608 1609
out:
	return young;
}

1610 1611
#define RMAP_RECYCLE_THRESHOLD 1000

1612
static void rmap_recycle(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn)
1613
{
1614
	struct kvm_rmap_head *rmap_head;
1615 1616 1617
	struct kvm_mmu_page *sp;

	sp = page_header(__pa(spte));
1618

1619
	rmap_head = gfn_to_rmap(vcpu->kvm, gfn, sp);
1620

1621
	kvm_unmap_rmapp(vcpu->kvm, rmap_head, NULL, gfn, sp->role.level, 0);
1622 1623 1624
	kvm_flush_remote_tlbs(vcpu->kvm);
}

A
Andres Lagar-Cavilla 已提交
1625
int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
1626
{
A
Andres Lagar-Cavilla 已提交
1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647
	/*
	 * In case of absence of EPT Access and Dirty Bits supports,
	 * emulate the accessed bit for EPT, by checking if this page has
	 * an EPT mapping, and clearing it if it does. On the next access,
	 * a new EPT mapping will be established.
	 * This has some overhead, but not as much as the cost of swapping
	 * out actively used pages or breaking up actively used hugepages.
	 */
	if (!shadow_accessed_mask) {
		/*
		 * We are holding the kvm->mmu_lock, and we are blowing up
		 * shadow PTEs. MMU notifier consumers need to be kept at bay.
		 * This is correct as long as we don't decouple the mmu_lock
		 * protected regions (like invalidate_range_start|end does).
		 */
		kvm->mmu_notifier_seq++;
		return kvm_handle_hva_range(kvm, start, end, 0,
					    kvm_unmap_rmapp);
	}

	return kvm_handle_hva_range(kvm, start, end, 0, kvm_age_rmapp);
1648 1649
}

A
Andrea Arcangeli 已提交
1650 1651 1652 1653 1654
int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
{
	return kvm_handle_hva(kvm, hva, 0, kvm_test_age_rmapp);
}

1655
#ifdef MMU_DEBUG
1656
static int is_empty_shadow_page(u64 *spt)
A
Avi Kivity 已提交
1657
{
1658 1659 1660
	u64 *pos;
	u64 *end;

1661
	for (pos = spt, end = pos + PAGE_SIZE / sizeof(u64); pos != end; pos++)
1662
		if (is_shadow_present_pte(*pos)) {
1663
			printk(KERN_ERR "%s: %p %llx\n", __func__,
1664
			       pos, *pos);
A
Avi Kivity 已提交
1665
			return 0;
1666
		}
A
Avi Kivity 已提交
1667 1668
	return 1;
}
1669
#endif
A
Avi Kivity 已提交
1670

1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682
/*
 * This value is the sum of all of the kvm instances's
 * kvm->arch.n_used_mmu_pages values.  We need a global,
 * aggregate version in order to make the slab shrinker
 * faster
 */
static inline void kvm_mod_used_mmu_pages(struct kvm *kvm, int nr)
{
	kvm->arch.n_used_mmu_pages += nr;
	percpu_counter_add(&kvm_total_used_mmu_pages, nr);
}

1683
static void kvm_mmu_free_page(struct kvm_mmu_page *sp)
1684
{
1685
	MMU_WARN_ON(!is_empty_shadow_page(sp->spt));
1686
	hlist_del(&sp->hash_link);
1687 1688
	list_del(&sp->link);
	free_page((unsigned long)sp->spt);
1689 1690
	if (!sp->role.direct)
		free_page((unsigned long)sp->gfns);
1691
	kmem_cache_free(mmu_page_header_cache, sp);
1692 1693
}

1694 1695
static unsigned kvm_page_table_hashfn(gfn_t gfn)
{
1696
	return gfn & ((1 << KVM_MMU_HASH_SHIFT) - 1);
1697 1698
}

1699
static void mmu_page_add_parent_pte(struct kvm_vcpu *vcpu,
1700
				    struct kvm_mmu_page *sp, u64 *parent_pte)
1701 1702 1703 1704
{
	if (!parent_pte)
		return;

1705
	pte_list_add(vcpu, parent_pte, &sp->parent_ptes);
1706 1707
}

1708
static void mmu_page_remove_parent_pte(struct kvm_mmu_page *sp,
1709 1710
				       u64 *parent_pte)
{
1711
	pte_list_remove(parent_pte, &sp->parent_ptes);
1712 1713
}

1714 1715 1716 1717
static void drop_parent_pte(struct kvm_mmu_page *sp,
			    u64 *parent_pte)
{
	mmu_page_remove_parent_pte(sp, parent_pte);
1718
	mmu_spte_clear_no_track(parent_pte);
1719 1720
}

1721 1722
static struct kvm_mmu_page *kvm_mmu_alloc_page(struct kvm_vcpu *vcpu,
					       u64 *parent_pte, int direct)
M
Marcelo Tosatti 已提交
1723
{
1724
	struct kvm_mmu_page *sp;
1725

1726 1727
	sp = mmu_memory_cache_alloc(&vcpu->arch.mmu_page_header_cache);
	sp->spt = mmu_memory_cache_alloc(&vcpu->arch.mmu_page_cache);
1728
	if (!direct)
1729
		sp->gfns = mmu_memory_cache_alloc(&vcpu->arch.mmu_page_cache);
1730
	set_page_private(virt_to_page(sp->spt), (unsigned long)sp);
1731 1732 1733 1734 1735 1736

	/*
	 * The active_mmu_pages list is the FIFO list, do not move the
	 * page until it is zapped. kvm_zap_obsolete_pages depends on
	 * this feature. See the comments in kvm_zap_obsolete_pages().
	 */
1737
	list_add(&sp->link, &vcpu->kvm->arch.active_mmu_pages);
1738
	sp->parent_ptes.val = 0;
1739 1740 1741
	mmu_page_add_parent_pte(vcpu, sp, parent_pte);
	kvm_mod_used_mmu_pages(vcpu->kvm, +1);
	return sp;
M
Marcelo Tosatti 已提交
1742 1743
}

1744
static void mark_unsync(u64 *spte);
1745
static void kvm_mmu_mark_parents_unsync(struct kvm_mmu_page *sp)
1746
{
1747
	pte_list_walk(&sp->parent_ptes, mark_unsync);
1748 1749
}

1750
static void mark_unsync(u64 *spte)
1751
{
1752
	struct kvm_mmu_page *sp;
1753
	unsigned int index;
1754

1755
	sp = page_header(__pa(spte));
1756 1757
	index = spte - sp->spt;
	if (__test_and_set_bit(index, sp->unsync_child_bitmap))
1758
		return;
1759
	if (sp->unsync_children++)
1760
		return;
1761
	kvm_mmu_mark_parents_unsync(sp);
1762 1763
}

1764
static int nonpaging_sync_page(struct kvm_vcpu *vcpu,
1765
			       struct kvm_mmu_page *sp)
1766 1767 1768 1769
{
	return 1;
}

M
Marcelo Tosatti 已提交
1770 1771 1772 1773
static void nonpaging_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
{
}

1774 1775
static void nonpaging_update_pte(struct kvm_vcpu *vcpu,
				 struct kvm_mmu_page *sp, u64 *spte,
1776
				 const void *pte)
1777 1778 1779 1780
{
	WARN_ON(1);
}

1781 1782 1783 1784 1785 1786 1787 1788 1789 1790
#define KVM_PAGE_ARRAY_NR 16

struct kvm_mmu_pages {
	struct mmu_page_and_offset {
		struct kvm_mmu_page *sp;
		unsigned int idx;
	} page[KVM_PAGE_ARRAY_NR];
	unsigned int nr;
};

1791 1792
static int mmu_pages_add(struct kvm_mmu_pages *pvec, struct kvm_mmu_page *sp,
			 int idx)
1793
{
1794
	int i;
1795

1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806
	if (sp->unsync)
		for (i=0; i < pvec->nr; i++)
			if (pvec->page[i].sp == sp)
				return 0;

	pvec->page[pvec->nr].sp = sp;
	pvec->page[pvec->nr].idx = idx;
	pvec->nr++;
	return (pvec->nr == KVM_PAGE_ARRAY_NR);
}

1807 1808 1809 1810 1811 1812 1813
static inline void clear_unsync_child_bit(struct kvm_mmu_page *sp, int idx)
{
	--sp->unsync_children;
	WARN_ON((int)sp->unsync_children < 0);
	__clear_bit(idx, sp->unsync_child_bitmap);
}

1814 1815 1816 1817
static int __mmu_unsync_walk(struct kvm_mmu_page *sp,
			   struct kvm_mmu_pages *pvec)
{
	int i, ret, nr_unsync_leaf = 0;
1818

1819
	for_each_set_bit(i, sp->unsync_child_bitmap, 512) {
1820
		struct kvm_mmu_page *child;
1821 1822
		u64 ent = sp->spt[i];

1823 1824 1825 1826
		if (!is_shadow_present_pte(ent) || is_large_pte(ent)) {
			clear_unsync_child_bit(sp, i);
			continue;
		}
1827 1828 1829 1830 1831 1832 1833 1834

		child = page_header(ent & PT64_BASE_ADDR_MASK);

		if (child->unsync_children) {
			if (mmu_pages_add(pvec, child, i))
				return -ENOSPC;

			ret = __mmu_unsync_walk(child, pvec);
1835 1836 1837 1838
			if (!ret) {
				clear_unsync_child_bit(sp, i);
				continue;
			} else if (ret > 0) {
1839
				nr_unsync_leaf += ret;
1840
			} else
1841 1842 1843 1844 1845 1846
				return ret;
		} else if (child->unsync) {
			nr_unsync_leaf++;
			if (mmu_pages_add(pvec, child, i))
				return -ENOSPC;
		} else
1847
			clear_unsync_child_bit(sp, i);
1848 1849
	}

1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860
	return nr_unsync_leaf;
}

static int mmu_unsync_walk(struct kvm_mmu_page *sp,
			   struct kvm_mmu_pages *pvec)
{
	if (!sp->unsync_children)
		return 0;

	mmu_pages_add(pvec, sp, 0);
	return __mmu_unsync_walk(sp, pvec);
1861 1862 1863 1864 1865
}

static void kvm_unlink_unsync_page(struct kvm *kvm, struct kvm_mmu_page *sp)
{
	WARN_ON(!sp->unsync);
1866
	trace_kvm_mmu_sync_page(sp);
1867 1868 1869 1870
	sp->unsync = 0;
	--kvm->stat.mmu_unsync;
}

1871 1872 1873 1874
static int kvm_mmu_prepare_zap_page(struct kvm *kvm, struct kvm_mmu_page *sp,
				    struct list_head *invalid_list);
static void kvm_mmu_commit_zap_page(struct kvm *kvm,
				    struct list_head *invalid_list);
1875

1876 1877 1878 1879 1880 1881 1882 1883 1884 1885
/*
 * NOTE: we should pay more attention on the zapped-obsolete page
 * (is_obsolete_sp(sp) && sp->role.invalid) when you do hash list walk
 * since it has been deleted from active_mmu_pages but still can be found
 * at hast list.
 *
 * for_each_gfn_indirect_valid_sp has skipped that kind of page and
 * kvm_mmu_get_page(), the only user of for_each_gfn_sp(), has skipped
 * all the obsolete pages.
 */
1886 1887 1888 1889 1890 1891 1892 1893
#define for_each_gfn_sp(_kvm, _sp, _gfn)				\
	hlist_for_each_entry(_sp,					\
	  &(_kvm)->arch.mmu_page_hash[kvm_page_table_hashfn(_gfn)], hash_link) \
		if ((_sp)->gfn != (_gfn)) {} else

#define for_each_gfn_indirect_valid_sp(_kvm, _sp, _gfn)			\
	for_each_gfn_sp(_kvm, _sp, _gfn)				\
		if ((_sp)->role.direct || (_sp)->role.invalid) {} else
1894

1895
/* @sp->gfn should be write-protected at the call site */
1896
static int __kvm_sync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
1897
			   struct list_head *invalid_list, bool clear_unsync)
1898
{
1899
	if (sp->role.cr4_pae != !!is_pae(vcpu)) {
1900
		kvm_mmu_prepare_zap_page(vcpu->kvm, sp, invalid_list);
1901 1902 1903
		return 1;
	}

1904
	if (clear_unsync)
1905 1906
		kvm_unlink_unsync_page(vcpu->kvm, sp);

1907
	if (vcpu->arch.mmu.sync_page(vcpu, sp)) {
1908
		kvm_mmu_prepare_zap_page(vcpu->kvm, sp, invalid_list);
1909 1910 1911
		return 1;
	}

1912
	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1913 1914 1915
	return 0;
}

1916 1917 1918
static int kvm_sync_page_transient(struct kvm_vcpu *vcpu,
				   struct kvm_mmu_page *sp)
{
1919
	LIST_HEAD(invalid_list);
1920 1921
	int ret;

1922
	ret = __kvm_sync_page(vcpu, sp, &invalid_list, false);
1923
	if (ret)
1924 1925
		kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);

1926 1927 1928
	return ret;
}

1929 1930 1931 1932 1933 1934 1935
#ifdef CONFIG_KVM_MMU_AUDIT
#include "mmu_audit.c"
#else
static void kvm_mmu_audit(struct kvm_vcpu *vcpu, int point) { }
static void mmu_audit_disable(void) { }
#endif

1936 1937
static int kvm_sync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
			 struct list_head *invalid_list)
1938
{
1939
	return __kvm_sync_page(vcpu, sp, invalid_list, true);
1940 1941
}

1942 1943 1944 1945
/* @gfn should be write-protected at the call site */
static void kvm_sync_pages(struct kvm_vcpu *vcpu,  gfn_t gfn)
{
	struct kvm_mmu_page *s;
1946
	LIST_HEAD(invalid_list);
1947 1948
	bool flush = false;

1949
	for_each_gfn_indirect_valid_sp(vcpu->kvm, s, gfn) {
1950
		if (!s->unsync)
1951 1952 1953
			continue;

		WARN_ON(s->role.level != PT_PAGE_TABLE_LEVEL);
1954
		kvm_unlink_unsync_page(vcpu->kvm, s);
1955
		if ((s->role.cr4_pae != !!is_pae(vcpu)) ||
1956
			(vcpu->arch.mmu.sync_page(vcpu, s))) {
1957
			kvm_mmu_prepare_zap_page(vcpu->kvm, s, &invalid_list);
1958 1959 1960 1961 1962
			continue;
		}
		flush = true;
	}

1963
	kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
1964
	if (flush)
1965
		kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1966 1967
}

1968 1969 1970
struct mmu_page_path {
	struct kvm_mmu_page *parent[PT64_ROOT_LEVEL-1];
	unsigned int idx[PT64_ROOT_LEVEL-1];
1971 1972
};

1973 1974 1975 1976 1977 1978
#define for_each_sp(pvec, sp, parents, i)			\
		for (i = mmu_pages_next(&pvec, &parents, -1),	\
			sp = pvec.page[i].sp;			\
			i < pvec.nr && ({ sp = pvec.page[i].sp; 1;});	\
			i = mmu_pages_next(&pvec, &parents, i))

1979 1980 1981
static int mmu_pages_next(struct kvm_mmu_pages *pvec,
			  struct mmu_page_path *parents,
			  int i)
1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
{
	int n;

	for (n = i+1; n < pvec->nr; n++) {
		struct kvm_mmu_page *sp = pvec->page[n].sp;

		if (sp->role.level == PT_PAGE_TABLE_LEVEL) {
			parents->idx[0] = pvec->page[n].idx;
			return n;
		}

		parents->parent[sp->role.level-2] = sp;
		parents->idx[sp->role.level-1] = pvec->page[n].idx;
	}

	return n;
}

2000
static void mmu_pages_clear_parents(struct mmu_page_path *parents)
2001
{
2002 2003 2004 2005 2006
	struct kvm_mmu_page *sp;
	unsigned int level = 0;

	do {
		unsigned int idx = parents->idx[level];
2007

2008 2009 2010 2011
		sp = parents->parent[level];
		if (!sp)
			return;

2012
		clear_unsync_child_bit(sp, idx);
2013 2014
		level++;
	} while (level < PT64_ROOT_LEVEL-1 && !sp->unsync_children);
2015 2016
}

2017 2018 2019
static void kvm_mmu_pages_init(struct kvm_mmu_page *parent,
			       struct mmu_page_path *parents,
			       struct kvm_mmu_pages *pvec)
2020
{
2021 2022 2023
	parents->parent[parent->role.level-1] = NULL;
	pvec->nr = 0;
}
2024

2025 2026 2027 2028 2029 2030 2031
static void mmu_sync_children(struct kvm_vcpu *vcpu,
			      struct kvm_mmu_page *parent)
{
	int i;
	struct kvm_mmu_page *sp;
	struct mmu_page_path parents;
	struct kvm_mmu_pages pages;
2032
	LIST_HEAD(invalid_list);
2033 2034 2035

	kvm_mmu_pages_init(parent, &parents, &pages);
	while (mmu_unsync_walk(parent, &pages)) {
2036
		bool protected = false;
2037 2038

		for_each_sp(pages, sp, parents, i)
2039
			protected |= rmap_write_protect(vcpu, sp->gfn);
2040 2041 2042 2043

		if (protected)
			kvm_flush_remote_tlbs(vcpu->kvm);

2044
		for_each_sp(pages, sp, parents, i) {
2045
			kvm_sync_page(vcpu, sp, &invalid_list);
2046 2047
			mmu_pages_clear_parents(&parents);
		}
2048
		kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
2049
		cond_resched_lock(&vcpu->kvm->mmu_lock);
2050 2051
		kvm_mmu_pages_init(parent, &parents, &pages);
	}
2052 2053
}

2054 2055 2056 2057 2058 2059 2060 2061
static void init_shadow_page_table(struct kvm_mmu_page *sp)
{
	int i;

	for (i = 0; i < PT64_ENT_PER_PAGE; ++i)
		sp->spt[i] = 0ull;
}

2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073
static void __clear_sp_write_flooding_count(struct kvm_mmu_page *sp)
{
	sp->write_flooding_count = 0;
}

static void clear_sp_write_flooding_count(u64 *spte)
{
	struct kvm_mmu_page *sp =  page_header(__pa(spte));

	__clear_sp_write_flooding_count(sp);
}

2074 2075 2076 2077 2078
static bool is_obsolete_sp(struct kvm *kvm, struct kvm_mmu_page *sp)
{
	return unlikely(sp->mmu_valid_gen != kvm->arch.mmu_valid_gen);
}

2079 2080 2081 2082
static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
					     gfn_t gfn,
					     gva_t gaddr,
					     unsigned level,
2083
					     int direct,
2084
					     unsigned access,
2085
					     u64 *parent_pte)
2086 2087 2088
{
	union kvm_mmu_page_role role;
	unsigned quadrant;
2089 2090
	struct kvm_mmu_page *sp;
	bool need_sync = false;
2091

2092
	role = vcpu->arch.mmu.base_role;
2093
	role.level = level;
2094
	role.direct = direct;
2095
	if (role.direct)
2096
		role.cr4_pae = 0;
2097
	role.access = access;
2098 2099
	if (!vcpu->arch.mmu.direct_map
	    && vcpu->arch.mmu.root_level <= PT32_ROOT_LEVEL) {
2100 2101 2102 2103
		quadrant = gaddr >> (PAGE_SHIFT + (PT64_PT_BITS * level));
		quadrant &= (1 << ((PT32_PT_BITS - PT64_PT_BITS) * level)) - 1;
		role.quadrant = quadrant;
	}
2104
	for_each_gfn_sp(vcpu->kvm, sp, gfn) {
2105 2106 2107
		if (is_obsolete_sp(vcpu->kvm, sp))
			continue;

2108 2109
		if (!need_sync && sp->unsync)
			need_sync = true;
2110

2111 2112
		if (sp->role.word != role.word)
			continue;
2113

2114 2115
		if (sp->unsync && kvm_sync_page_transient(vcpu, sp))
			break;
2116

2117 2118
		mmu_page_add_parent_pte(vcpu, sp, parent_pte);
		if (sp->unsync_children) {
2119
			kvm_make_request(KVM_REQ_MMU_SYNC, vcpu);
2120 2121 2122
			kvm_mmu_mark_parents_unsync(sp);
		} else if (sp->unsync)
			kvm_mmu_mark_parents_unsync(sp);
2123

2124
		__clear_sp_write_flooding_count(sp);
2125 2126 2127
		trace_kvm_mmu_get_page(sp, false);
		return sp;
	}
A
Avi Kivity 已提交
2128
	++vcpu->kvm->stat.mmu_cache_miss;
2129
	sp = kvm_mmu_alloc_page(vcpu, parent_pte, direct);
2130 2131 2132 2133
	if (!sp)
		return sp;
	sp->gfn = gfn;
	sp->role = role;
2134 2135
	hlist_add_head(&sp->hash_link,
		&vcpu->kvm->arch.mmu_page_hash[kvm_page_table_hashfn(gfn)]);
2136
	if (!direct) {
2137
		if (rmap_write_protect(vcpu, gfn))
2138
			kvm_flush_remote_tlbs(vcpu->kvm);
2139 2140 2141
		if (level > PT_PAGE_TABLE_LEVEL && need_sync)
			kvm_sync_pages(vcpu, gfn);

2142
		account_shadowed(vcpu->kvm, sp);
2143
	}
2144
	sp->mmu_valid_gen = vcpu->kvm->arch.mmu_valid_gen;
2145
	init_shadow_page_table(sp);
A
Avi Kivity 已提交
2146
	trace_kvm_mmu_get_page(sp, true);
2147
	return sp;
2148 2149
}

2150 2151 2152 2153 2154 2155
static void shadow_walk_init(struct kvm_shadow_walk_iterator *iterator,
			     struct kvm_vcpu *vcpu, u64 addr)
{
	iterator->addr = addr;
	iterator->shadow_addr = vcpu->arch.mmu.root_hpa;
	iterator->level = vcpu->arch.mmu.shadow_root_level;
2156 2157 2158 2159 2160 2161

	if (iterator->level == PT64_ROOT_LEVEL &&
	    vcpu->arch.mmu.root_level < PT64_ROOT_LEVEL &&
	    !vcpu->arch.mmu.direct_map)
		--iterator->level;

2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175
	if (iterator->level == PT32E_ROOT_LEVEL) {
		iterator->shadow_addr
			= vcpu->arch.mmu.pae_root[(addr >> 30) & 3];
		iterator->shadow_addr &= PT64_BASE_ADDR_MASK;
		--iterator->level;
		if (!iterator->shadow_addr)
			iterator->level = 0;
	}
}

static bool shadow_walk_okay(struct kvm_shadow_walk_iterator *iterator)
{
	if (iterator->level < PT_PAGE_TABLE_LEVEL)
		return false;
2176

2177 2178 2179 2180 2181
	iterator->index = SHADOW_PT_INDEX(iterator->addr, iterator->level);
	iterator->sptep	= ((u64 *)__va(iterator->shadow_addr)) + iterator->index;
	return true;
}

2182 2183
static void __shadow_walk_next(struct kvm_shadow_walk_iterator *iterator,
			       u64 spte)
2184
{
2185
	if (is_last_spte(spte, iterator->level)) {
2186 2187 2188 2189
		iterator->level = 0;
		return;
	}

2190
	iterator->shadow_addr = spte & PT64_BASE_ADDR_MASK;
2191 2192 2193
	--iterator->level;
}

2194 2195 2196 2197 2198
static void shadow_walk_next(struct kvm_shadow_walk_iterator *iterator)
{
	return __shadow_walk_next(iterator, *iterator->sptep);
}

2199
static void link_shadow_page(u64 *sptep, struct kvm_mmu_page *sp)
2200 2201 2202
{
	u64 spte;

2203 2204 2205
	BUILD_BUG_ON(VMX_EPT_READABLE_MASK != PT_PRESENT_MASK ||
			VMX_EPT_WRITABLE_MASK != PT_WRITABLE_MASK);

X
Xiao Guangrong 已提交
2206
	spte = __pa(sp->spt) | PT_PRESENT_MASK | PT_WRITABLE_MASK |
2207
	       shadow_user_mask | shadow_x_mask | shadow_accessed_mask;
X
Xiao Guangrong 已提交
2208

2209
	mmu_spte_set(sptep, spte);
2210 2211
}

2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228
static void validate_direct_spte(struct kvm_vcpu *vcpu, u64 *sptep,
				   unsigned direct_access)
{
	if (is_shadow_present_pte(*sptep) && !is_large_pte(*sptep)) {
		struct kvm_mmu_page *child;

		/*
		 * For the direct sp, if the guest pte's dirty bit
		 * changed form clean to dirty, it will corrupt the
		 * sp's access: allow writable in the read-only sp,
		 * so we should update the spte at this point to get
		 * a new sp with the correct access.
		 */
		child = page_header(*sptep & PT64_BASE_ADDR_MASK);
		if (child->role.access == direct_access)
			return;

2229
		drop_parent_pte(child, sptep);
2230 2231 2232 2233
		kvm_flush_remote_tlbs(vcpu->kvm);
	}
}

X
Xiao Guangrong 已提交
2234
static bool mmu_page_zap_pte(struct kvm *kvm, struct kvm_mmu_page *sp,
2235 2236 2237 2238 2239 2240 2241
			     u64 *spte)
{
	u64 pte;
	struct kvm_mmu_page *child;

	pte = *spte;
	if (is_shadow_present_pte(pte)) {
X
Xiao Guangrong 已提交
2242
		if (is_last_spte(pte, sp->role.level)) {
2243
			drop_spte(kvm, spte);
X
Xiao Guangrong 已提交
2244 2245 2246
			if (is_large_pte(pte))
				--kvm->stat.lpages;
		} else {
2247
			child = page_header(pte & PT64_BASE_ADDR_MASK);
2248
			drop_parent_pte(child, spte);
2249
		}
X
Xiao Guangrong 已提交
2250 2251 2252 2253
		return true;
	}

	if (is_mmio_spte(pte))
2254
		mmu_spte_clear_no_track(spte);
2255

X
Xiao Guangrong 已提交
2256
	return false;
2257 2258
}

2259
static void kvm_mmu_page_unlink_children(struct kvm *kvm,
2260
					 struct kvm_mmu_page *sp)
2261
{
2262 2263
	unsigned i;

2264 2265
	for (i = 0; i < PT64_ENT_PER_PAGE; ++i)
		mmu_page_zap_pte(kvm, sp, sp->spt + i);
2266 2267
}

2268
static void kvm_mmu_put_page(struct kvm_mmu_page *sp, u64 *parent_pte)
2269
{
2270
	mmu_page_remove_parent_pte(sp, parent_pte);
2271 2272
}

2273
static void kvm_mmu_unlink_parents(struct kvm *kvm, struct kvm_mmu_page *sp)
2274
{
2275 2276
	u64 *sptep;
	struct rmap_iterator iter;
2277

2278
	while ((sptep = rmap_get_first(&sp->parent_ptes, &iter)))
2279
		drop_parent_pte(sp, sptep);
2280 2281
}

2282
static int mmu_zap_unsync_children(struct kvm *kvm,
2283 2284
				   struct kvm_mmu_page *parent,
				   struct list_head *invalid_list)
2285
{
2286 2287 2288
	int i, zapped = 0;
	struct mmu_page_path parents;
	struct kvm_mmu_pages pages;
2289

2290
	if (parent->role.level == PT_PAGE_TABLE_LEVEL)
2291
		return 0;
2292 2293 2294 2295 2296 2297

	kvm_mmu_pages_init(parent, &parents, &pages);
	while (mmu_unsync_walk(parent, &pages)) {
		struct kvm_mmu_page *sp;

		for_each_sp(pages, sp, parents, i) {
2298
			kvm_mmu_prepare_zap_page(kvm, sp, invalid_list);
2299
			mmu_pages_clear_parents(&parents);
2300
			zapped++;
2301 2302 2303 2304 2305
		}
		kvm_mmu_pages_init(parent, &parents, &pages);
	}

	return zapped;
2306 2307
}

2308 2309
static int kvm_mmu_prepare_zap_page(struct kvm *kvm, struct kvm_mmu_page *sp,
				    struct list_head *invalid_list)
2310
{
2311
	int ret;
A
Avi Kivity 已提交
2312

2313
	trace_kvm_mmu_prepare_zap_page(sp);
2314
	++kvm->stat.mmu_shadow_zapped;
2315
	ret = mmu_zap_unsync_children(kvm, sp, invalid_list);
2316
	kvm_mmu_page_unlink_children(kvm, sp);
2317
	kvm_mmu_unlink_parents(kvm, sp);
2318

2319
	if (!sp->role.invalid && !sp->role.direct)
2320
		unaccount_shadowed(kvm, sp);
2321

2322 2323
	if (sp->unsync)
		kvm_unlink_unsync_page(kvm, sp);
2324
	if (!sp->root_count) {
2325 2326
		/* Count self */
		ret++;
2327
		list_move(&sp->link, invalid_list);
2328
		kvm_mod_used_mmu_pages(kvm, -1);
2329
	} else {
A
Avi Kivity 已提交
2330
		list_move(&sp->link, &kvm->arch.active_mmu_pages);
2331 2332 2333 2334 2335 2336 2337

		/*
		 * The obsolete pages can not be used on any vcpus.
		 * See the comments in kvm_mmu_invalidate_zap_all_pages().
		 */
		if (!sp->role.invalid && !is_obsolete_sp(kvm, sp))
			kvm_reload_remote_mmus(kvm);
2338
	}
2339 2340

	sp->role.invalid = 1;
2341
	return ret;
2342 2343
}

2344 2345 2346
static void kvm_mmu_commit_zap_page(struct kvm *kvm,
				    struct list_head *invalid_list)
{
2347
	struct kvm_mmu_page *sp, *nsp;
2348 2349 2350 2351

	if (list_empty(invalid_list))
		return;

2352 2353 2354 2355 2356
	/*
	 * wmb: make sure everyone sees our modifications to the page tables
	 * rmb: make sure we see changes to vcpu->mode
	 */
	smp_mb();
X
Xiao Guangrong 已提交
2357

2358 2359 2360 2361 2362
	/*
	 * Wait for all vcpus to exit guest mode and/or lockless shadow
	 * page table walks.
	 */
	kvm_flush_remote_tlbs(kvm);
2363

2364
	list_for_each_entry_safe(sp, nsp, invalid_list, link) {
2365
		WARN_ON(!sp->role.invalid || sp->root_count);
2366
		kvm_mmu_free_page(sp);
2367
	}
2368 2369
}

2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384
static bool prepare_zap_oldest_mmu_page(struct kvm *kvm,
					struct list_head *invalid_list)
{
	struct kvm_mmu_page *sp;

	if (list_empty(&kvm->arch.active_mmu_pages))
		return false;

	sp = list_entry(kvm->arch.active_mmu_pages.prev,
			struct kvm_mmu_page, link);
	kvm_mmu_prepare_zap_page(kvm, sp, invalid_list);

	return true;
}

2385 2386
/*
 * Changing the number of mmu pages allocated to the vm
2387
 * Note: if goal_nr_mmu_pages is too small, you will get dead lock
2388
 */
2389
void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int goal_nr_mmu_pages)
2390
{
2391
	LIST_HEAD(invalid_list);
2392

2393 2394
	spin_lock(&kvm->mmu_lock);

2395
	if (kvm->arch.n_used_mmu_pages > goal_nr_mmu_pages) {
2396 2397 2398 2399
		/* Need to free some mmu pages to achieve the goal. */
		while (kvm->arch.n_used_mmu_pages > goal_nr_mmu_pages)
			if (!prepare_zap_oldest_mmu_page(kvm, &invalid_list))
				break;
2400

2401
		kvm_mmu_commit_zap_page(kvm, &invalid_list);
2402
		goal_nr_mmu_pages = kvm->arch.n_used_mmu_pages;
2403 2404
	}

2405
	kvm->arch.n_max_mmu_pages = goal_nr_mmu_pages;
2406 2407

	spin_unlock(&kvm->mmu_lock);
2408 2409
}

2410
int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn)
2411
{
2412
	struct kvm_mmu_page *sp;
2413
	LIST_HEAD(invalid_list);
2414 2415
	int r;

2416
	pgprintk("%s: looking for gfn %llx\n", __func__, gfn);
2417
	r = 0;
2418
	spin_lock(&kvm->mmu_lock);
2419
	for_each_gfn_indirect_valid_sp(kvm, sp, gfn) {
2420
		pgprintk("%s: gfn %llx role %x\n", __func__, gfn,
2421 2422
			 sp->role.word);
		r = 1;
2423
		kvm_mmu_prepare_zap_page(kvm, sp, &invalid_list);
2424
	}
2425
	kvm_mmu_commit_zap_page(kvm, &invalid_list);
2426 2427
	spin_unlock(&kvm->mmu_lock);

2428
	return r;
2429
}
2430
EXPORT_SYMBOL_GPL(kvm_mmu_unprotect_page);
2431

2432 2433 2434 2435 2436 2437 2438 2439 2440 2441
static void __kvm_unsync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp)
{
	trace_kvm_mmu_unsync_page(sp);
	++vcpu->kvm->stat.mmu_unsync;
	sp->unsync = 1;

	kvm_mmu_mark_parents_unsync(sp);
}

static void kvm_unsync_pages(struct kvm_vcpu *vcpu,  gfn_t gfn)
2442 2443
{
	struct kvm_mmu_page *s;
2444

2445
	for_each_gfn_indirect_valid_sp(vcpu->kvm, s, gfn) {
2446
		if (s->unsync)
2447
			continue;
2448 2449
		WARN_ON(s->role.level != PT_PAGE_TABLE_LEVEL);
		__kvm_unsync_page(vcpu, s);
2450 2451 2452 2453 2454 2455
	}
}

static int mmu_need_write_protect(struct kvm_vcpu *vcpu, gfn_t gfn,
				  bool can_unsync)
{
2456 2457 2458
	struct kvm_mmu_page *s;
	bool need_unsync = false;

2459
	for_each_gfn_indirect_valid_sp(vcpu->kvm, s, gfn) {
2460 2461 2462
		if (!can_unsync)
			return 1;

2463
		if (s->role.level != PT_PAGE_TABLE_LEVEL)
2464
			return 1;
2465

G
Gleb Natapov 已提交
2466
		if (!s->unsync)
2467
			need_unsync = true;
2468
	}
2469 2470
	if (need_unsync)
		kvm_unsync_pages(vcpu, gfn);
2471 2472 2473
	return 0;
}

2474 2475 2476 2477 2478 2479 2480 2481
static bool kvm_is_mmio_pfn(pfn_t pfn)
{
	if (pfn_valid(pfn))
		return !is_zero_pfn(pfn) && PageReserved(pfn_to_page(pfn));

	return true;
}

A
Avi Kivity 已提交
2482
static int set_spte(struct kvm_vcpu *vcpu, u64 *sptep,
2483
		    unsigned pte_access, int level,
2484
		    gfn_t gfn, pfn_t pfn, bool speculative,
2485
		    bool can_unsync, bool host_writable)
2486
{
2487
	u64 spte;
M
Marcelo Tosatti 已提交
2488
	int ret = 0;
S
Sheng Yang 已提交
2489

2490
	if (set_mmio_spte(vcpu, sptep, gfn, pfn, pte_access))
2491 2492
		return 0;

2493
	spte = PT_PRESENT_MASK;
2494
	if (!speculative)
2495
		spte |= shadow_accessed_mask;
2496

S
Sheng Yang 已提交
2497 2498 2499 2500
	if (pte_access & ACC_EXEC_MASK)
		spte |= shadow_x_mask;
	else
		spte |= shadow_nx_mask;
2501

2502
	if (pte_access & ACC_USER_MASK)
S
Sheng Yang 已提交
2503
		spte |= shadow_user_mask;
2504

2505
	if (level > PT_PAGE_TABLE_LEVEL)
M
Marcelo Tosatti 已提交
2506
		spte |= PT_PAGE_SIZE_MASK;
2507
	if (tdp_enabled)
2508
		spte |= kvm_x86_ops->get_mt_mask(vcpu, gfn,
2509
			kvm_is_mmio_pfn(pfn));
2510

2511
	if (host_writable)
2512
		spte |= SPTE_HOST_WRITEABLE;
2513 2514
	else
		pte_access &= ~ACC_WRITE_MASK;
2515

2516
	spte |= (u64)pfn << PAGE_SHIFT;
2517

2518
	if (pte_access & ACC_WRITE_MASK) {
2519

X
Xiao Guangrong 已提交
2520
		/*
2521 2522 2523 2524
		 * Other vcpu creates new sp in the window between
		 * mapping_level() and acquiring mmu-lock. We can
		 * allow guest to retry the access, the mapping can
		 * be fixed if guest refault.
X
Xiao Guangrong 已提交
2525
		 */
2526
		if (level > PT_PAGE_TABLE_LEVEL &&
2527
		    has_wrprotected_page(vcpu, gfn, level))
A
Avi Kivity 已提交
2528
			goto done;
2529

2530
		spte |= PT_WRITABLE_MASK | SPTE_MMU_WRITEABLE;
2531

2532 2533 2534 2535 2536 2537
		/*
		 * Optimization: for pte sync, if spte was writable the hash
		 * lookup is unnecessary (and expensive). Write protection
		 * is responsibility of mmu_get_page / kvm_sync_page.
		 * Same reasoning can be applied to dirty page accounting.
		 */
2538
		if (!can_unsync && is_writable_pte(*sptep))
2539 2540
			goto set_pte;

2541
		if (mmu_need_write_protect(vcpu, gfn, can_unsync)) {
2542
			pgprintk("%s: found shadow page for %llx, marking ro\n",
2543
				 __func__, gfn);
M
Marcelo Tosatti 已提交
2544
			ret = 1;
2545
			pte_access &= ~ACC_WRITE_MASK;
2546
			spte &= ~(PT_WRITABLE_MASK | SPTE_MMU_WRITEABLE);
2547 2548 2549
		}
	}

2550
	if (pte_access & ACC_WRITE_MASK) {
2551
		kvm_vcpu_mark_page_dirty(vcpu, gfn);
2552 2553
		spte |= shadow_dirty_mask;
	}
2554

2555
set_pte:
2556
	if (mmu_spte_update(sptep, spte))
2557
		kvm_flush_remote_tlbs(vcpu->kvm);
A
Avi Kivity 已提交
2558
done:
M
Marcelo Tosatti 已提交
2559 2560 2561
	return ret;
}

2562 2563 2564
static bool mmu_set_spte(struct kvm_vcpu *vcpu, u64 *sptep, unsigned pte_access,
			 int write_fault, int level, gfn_t gfn, pfn_t pfn,
			 bool speculative, bool host_writable)
M
Marcelo Tosatti 已提交
2565 2566
{
	int was_rmapped = 0;
2567
	int rmap_count;
2568
	bool emulate = false;
M
Marcelo Tosatti 已提交
2569

2570 2571
	pgprintk("%s: spte %llx write_fault %d gfn %llx\n", __func__,
		 *sptep, write_fault, gfn);
M
Marcelo Tosatti 已提交
2572

2573
	if (is_shadow_present_pte(*sptep)) {
M
Marcelo Tosatti 已提交
2574 2575 2576 2577
		/*
		 * If we overwrite a PTE page pointer with a 2MB PMD, unlink
		 * the parent of the now unreachable PTE.
		 */
2578 2579
		if (level > PT_PAGE_TABLE_LEVEL &&
		    !is_large_pte(*sptep)) {
M
Marcelo Tosatti 已提交
2580
			struct kvm_mmu_page *child;
A
Avi Kivity 已提交
2581
			u64 pte = *sptep;
M
Marcelo Tosatti 已提交
2582 2583

			child = page_header(pte & PT64_BASE_ADDR_MASK);
2584
			drop_parent_pte(child, sptep);
2585
			kvm_flush_remote_tlbs(vcpu->kvm);
A
Avi Kivity 已提交
2586
		} else if (pfn != spte_to_pfn(*sptep)) {
2587
			pgprintk("hfn old %llx new %llx\n",
A
Avi Kivity 已提交
2588
				 spte_to_pfn(*sptep), pfn);
2589
			drop_spte(vcpu->kvm, sptep);
2590
			kvm_flush_remote_tlbs(vcpu->kvm);
2591 2592
		} else
			was_rmapped = 1;
M
Marcelo Tosatti 已提交
2593
	}
2594

2595 2596
	if (set_spte(vcpu, sptep, pte_access, level, gfn, pfn, speculative,
	      true, host_writable)) {
M
Marcelo Tosatti 已提交
2597
		if (write_fault)
2598
			emulate = true;
2599
		kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
2600
	}
M
Marcelo Tosatti 已提交
2601

2602 2603
	if (unlikely(is_mmio_spte(*sptep)))
		emulate = true;
2604

A
Avi Kivity 已提交
2605
	pgprintk("%s: setting spte %llx\n", __func__, *sptep);
2606
	pgprintk("instantiating %s PTE (%s) at %llx (%llx) addr %p\n",
A
Avi Kivity 已提交
2607
		 is_large_pte(*sptep)? "2MB" : "4kB",
2608 2609
		 *sptep & PT_PRESENT_MASK ?"RW":"R", gfn,
		 *sptep, sptep);
A
Avi Kivity 已提交
2610
	if (!was_rmapped && is_large_pte(*sptep))
M
Marcelo Tosatti 已提交
2611 2612
		++vcpu->kvm->stat.lpages;

2613 2614 2615 2616 2617 2618
	if (is_shadow_present_pte(*sptep)) {
		if (!was_rmapped) {
			rmap_count = rmap_add(vcpu, sptep, gfn);
			if (rmap_count > RMAP_RECYCLE_THRESHOLD)
				rmap_recycle(vcpu, sptep, gfn);
		}
2619
	}
2620

X
Xiao Guangrong 已提交
2621
	kvm_release_pfn_clean(pfn);
2622 2623

	return emulate;
2624 2625
}

2626 2627 2628 2629 2630
static pfn_t pte_prefetch_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn,
				     bool no_dirty_log)
{
	struct kvm_memory_slot *slot;

2631
	slot = gfn_to_memslot_dirty_bitmap(vcpu, gfn, no_dirty_log);
2632
	if (!slot)
2633
		return KVM_PFN_ERR_FAULT;
2634

2635
	return gfn_to_pfn_memslot_atomic(slot, gfn);
2636 2637 2638 2639 2640 2641 2642
}

static int direct_pte_prefetch_many(struct kvm_vcpu *vcpu,
				    struct kvm_mmu_page *sp,
				    u64 *start, u64 *end)
{
	struct page *pages[PTE_PREFETCH_NUM];
2643
	struct kvm_memory_slot *slot;
2644 2645 2646 2647 2648
	unsigned access = sp->role.access;
	int i, ret;
	gfn_t gfn;

	gfn = kvm_mmu_page_get_gfn(sp, start - sp->spt);
2649 2650
	slot = gfn_to_memslot_dirty_bitmap(vcpu, gfn, access & ACC_WRITE_MASK);
	if (!slot)
2651 2652
		return -1;

2653
	ret = gfn_to_page_many_atomic(slot, gfn, pages, end - start);
2654 2655 2656 2657
	if (ret <= 0)
		return -1;

	for (i = 0; i < ret; i++, gfn++, start++)
2658 2659
		mmu_set_spte(vcpu, start, access, 0, sp->role.level, gfn,
			     page_to_pfn(pages[i]), true, true);
2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675

	return 0;
}

static void __direct_pte_prefetch(struct kvm_vcpu *vcpu,
				  struct kvm_mmu_page *sp, u64 *sptep)
{
	u64 *spte, *start = NULL;
	int i;

	WARN_ON(!sp->role.direct);

	i = (sptep - sp->spt) & ~(PTE_PREFETCH_NUM - 1);
	spte = sp->spt + i;

	for (i = 0; i < PTE_PREFETCH_NUM; i++, spte++) {
2676
		if (is_shadow_present_pte(*spte) || spte == sptep) {
2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706
			if (!start)
				continue;
			if (direct_pte_prefetch_many(vcpu, sp, start, spte) < 0)
				break;
			start = NULL;
		} else if (!start)
			start = spte;
	}
}

static void direct_pte_prefetch(struct kvm_vcpu *vcpu, u64 *sptep)
{
	struct kvm_mmu_page *sp;

	/*
	 * Since it's no accessed bit on EPT, it's no way to
	 * distinguish between actually accessed translations
	 * and prefetched, so disable pte prefetch if EPT is
	 * enabled.
	 */
	if (!shadow_accessed_mask)
		return;

	sp = page_header(__pa(sptep));
	if (sp->role.level > PT_PAGE_TABLE_LEVEL)
		return;

	__direct_pte_prefetch(vcpu, sp, sptep);
}

2707 2708
static int __direct_map(struct kvm_vcpu *vcpu, int write, int map_writable,
			int level, gfn_t gfn, pfn_t pfn, bool prefault)
2709
{
2710
	struct kvm_shadow_walk_iterator iterator;
2711
	struct kvm_mmu_page *sp;
2712
	int emulate = 0;
2713
	gfn_t pseudo_gfn;
A
Avi Kivity 已提交
2714

2715 2716 2717
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
		return 0;

2718
	for_each_shadow_entry(vcpu, (u64)gfn << PAGE_SHIFT, iterator) {
2719
		if (iterator.level == level) {
2720 2721 2722
			emulate = mmu_set_spte(vcpu, iterator.sptep, ACC_ALL,
					       write, level, gfn, pfn, prefault,
					       map_writable);
2723
			direct_pte_prefetch(vcpu, iterator.sptep);
2724 2725
			++vcpu->stat.pf_fixed;
			break;
A
Avi Kivity 已提交
2726 2727
		}

2728
		drop_large_spte(vcpu, iterator.sptep);
2729
		if (!is_shadow_present_pte(*iterator.sptep)) {
2730 2731 2732 2733
			u64 base_addr = iterator.addr;

			base_addr &= PT64_LVL_ADDR_MASK(iterator.level);
			pseudo_gfn = base_addr >> PAGE_SHIFT;
2734 2735 2736
			sp = kvm_mmu_get_page(vcpu, pseudo_gfn, iterator.addr,
					      iterator.level - 1,
					      1, ACC_ALL, iterator.sptep);
2737

2738
			link_shadow_page(iterator.sptep, sp);
2739 2740
		}
	}
2741
	return emulate;
A
Avi Kivity 已提交
2742 2743
}

H
Huang Ying 已提交
2744
static void kvm_send_hwpoison_signal(unsigned long address, struct task_struct *tsk)
2745
{
H
Huang Ying 已提交
2746 2747 2748 2749 2750 2751 2752
	siginfo_t info;

	info.si_signo	= SIGBUS;
	info.si_errno	= 0;
	info.si_code	= BUS_MCEERR_AR;
	info.si_addr	= (void __user *)address;
	info.si_addr_lsb = PAGE_SHIFT;
2753

H
Huang Ying 已提交
2754
	send_sig_info(SIGBUS, &info, tsk);
2755 2756
}

2757
static int kvm_handle_bad_page(struct kvm_vcpu *vcpu, gfn_t gfn, pfn_t pfn)
2758
{
X
Xiao Guangrong 已提交
2759 2760 2761 2762 2763 2764 2765 2766 2767
	/*
	 * Do not cache the mmio info caused by writing the readonly gfn
	 * into the spte otherwise read access on readonly gfn also can
	 * caused mmio page fault and treat it as mmio access.
	 * Return 1 to tell kvm to emulate it.
	 */
	if (pfn == KVM_PFN_ERR_RO_FAULT)
		return 1;

2768
	if (pfn == KVM_PFN_ERR_HWPOISON) {
2769
		kvm_send_hwpoison_signal(kvm_vcpu_gfn_to_hva(vcpu, gfn), current);
2770
		return 0;
2771
	}
2772

2773
	return -EFAULT;
2774 2775
}

2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788
static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
					gfn_t *gfnp, pfn_t *pfnp, int *levelp)
{
	pfn_t pfn = *pfnp;
	gfn_t gfn = *gfnp;
	int level = *levelp;

	/*
	 * Check if it's a transparent hugepage. If this would be an
	 * hugetlbfs page, level wouldn't be set to
	 * PT_PAGE_TABLE_LEVEL and there would be no adjustment done
	 * here.
	 */
2789
	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn) &&
2790 2791
	    level == PT_PAGE_TABLE_LEVEL &&
	    PageTransCompound(pfn_to_page(pfn)) &&
2792
	    !has_wrprotected_page(vcpu, gfn, PT_DIRECTORY_LEVEL)) {
2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810
		unsigned long mask;
		/*
		 * mmu_notifier_retry was successful and we hold the
		 * mmu_lock here, so the pmd can't become splitting
		 * from under us, and in turn
		 * __split_huge_page_refcount() can't run from under
		 * us and we can safely transfer the refcount from
		 * PG_tail to PG_head as we switch the pfn to tail to
		 * head.
		 */
		*levelp = level = PT_DIRECTORY_LEVEL;
		mask = KVM_PAGES_PER_HPAGE(level) - 1;
		VM_BUG_ON((gfn & mask) != (pfn & mask));
		if (pfn & mask) {
			gfn &= ~mask;
			*gfnp = gfn;
			kvm_release_pfn_clean(pfn);
			pfn &= ~mask;
2811
			kvm_get_pfn(pfn);
2812 2813 2814 2815 2816
			*pfnp = pfn;
		}
	}
}

2817 2818 2819 2820 2821 2822
static bool handle_abnormal_pfn(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
				pfn_t pfn, unsigned access, int *ret_val)
{
	bool ret = true;

	/* The pfn is invalid, report the error! */
2823
	if (unlikely(is_error_pfn(pfn))) {
2824 2825 2826 2827
		*ret_val = kvm_handle_bad_page(vcpu, gfn, pfn);
		goto exit;
	}

2828
	if (unlikely(is_noslot_pfn(pfn)))
2829 2830 2831 2832 2833 2834 2835
		vcpu_cache_mmio_info(vcpu, gva, gfn, access);

	ret = false;
exit:
	return ret;
}

2836
static bool page_fault_can_be_fast(u32 error_code)
2837
{
2838 2839 2840 2841 2842 2843 2844
	/*
	 * Do not fix the mmio spte with invalid generation number which
	 * need to be updated by slow page fault path.
	 */
	if (unlikely(error_code & PFERR_RSVD_MASK))
		return false;

2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857
	/*
	 * #PF can be fast only if the shadow page table is present and it
	 * is caused by write-protect, that means we just need change the
	 * W bit of the spte which can be done out of mmu-lock.
	 */
	if (!(error_code & PFERR_PRESENT_MASK) ||
	      !(error_code & PFERR_WRITE_MASK))
		return false;

	return true;
}

static bool
2858 2859
fast_pf_fix_direct_spte(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
			u64 *sptep, u64 spte)
2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870
{
	gfn_t gfn;

	WARN_ON(!sp->role.direct);

	/*
	 * The gfn of direct spte is stable since it is calculated
	 * by sp->gfn.
	 */
	gfn = kvm_mmu_page_get_gfn(sp, sptep - sp->spt);

2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882
	/*
	 * Theoretically we could also set dirty bit (and flush TLB) here in
	 * order to eliminate unnecessary PML logging. See comments in
	 * set_spte. But fast_page_fault is very unlikely to happen with PML
	 * enabled, so we do not do this. This might result in the same GPA
	 * to be logged in PML buffer again when the write really happens, and
	 * eventually to be called by mark_page_dirty twice. But it's also no
	 * harm. This also avoids the TLB flush needed after setting dirty bit
	 * so non-PML cases won't be impacted.
	 *
	 * Compare with set_spte where instead shadow_dirty_mask is set.
	 */
2883
	if (cmpxchg64(sptep, spte, spte | PT_WRITABLE_MASK) == spte)
2884
		kvm_vcpu_mark_page_dirty(vcpu, gfn);
2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897

	return true;
}

/*
 * Return value:
 * - true: let the vcpu to access on the same address again.
 * - false: let the real page fault path to fix it.
 */
static bool fast_page_fault(struct kvm_vcpu *vcpu, gva_t gva, int level,
			    u32 error_code)
{
	struct kvm_shadow_walk_iterator iterator;
2898
	struct kvm_mmu_page *sp;
2899 2900 2901
	bool ret = false;
	u64 spte = 0ull;

2902 2903 2904
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
		return false;

2905
	if (!page_fault_can_be_fast(error_code))
2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916
		return false;

	walk_shadow_page_lockless_begin(vcpu);
	for_each_shadow_entry_lockless(vcpu, gva, iterator, spte)
		if (!is_shadow_present_pte(spte) || iterator.level < level)
			break;

	/*
	 * If the mapping has been changed, let the vcpu fault on the
	 * same address again.
	 */
2917
	if (!is_shadow_present_pte(spte)) {
2918 2919 2920 2921
		ret = true;
		goto exit;
	}

2922 2923
	sp = page_header(__pa(iterator.sptep));
	if (!is_last_spte(spte, sp->role.level))
2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943
		goto exit;

	/*
	 * Check if it is a spurious fault caused by TLB lazily flushed.
	 *
	 * Need not check the access of upper level table entries since
	 * they are always ACC_ALL.
	 */
	 if (is_writable_pte(spte)) {
		ret = true;
		goto exit;
	}

	/*
	 * Currently, to simplify the code, only the spte write-protected
	 * by dirty-log can be fast fixed.
	 */
	if (!spte_is_locklessly_modifiable(spte))
		goto exit;

2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956
	/*
	 * Do not fix write-permission on the large spte since we only dirty
	 * the first page into the dirty-bitmap in fast_pf_fix_direct_spte()
	 * that means other pages are missed if its slot is dirty-logged.
	 *
	 * Instead, we let the slow page fault path create a normal spte to
	 * fix the access.
	 *
	 * See the comments in kvm_arch_commit_memory_region().
	 */
	if (sp->role.level > PT_PAGE_TABLE_LEVEL)
		goto exit;

2957 2958 2959 2960 2961
	/*
	 * Currently, fast page fault only works for direct mapping since
	 * the gfn is not stable for indirect shadow page.
	 * See Documentation/virtual/kvm/locking.txt to get more detail.
	 */
2962
	ret = fast_pf_fix_direct_spte(vcpu, sp, iterator.sptep, spte);
2963
exit:
X
Xiao Guangrong 已提交
2964 2965
	trace_fast_page_fault(vcpu, gva, error_code, iterator.sptep,
			      spte, ret);
2966 2967 2968 2969 2970
	walk_shadow_page_lockless_end(vcpu);

	return ret;
}

2971
static bool try_async_pf(struct kvm_vcpu *vcpu, bool prefault, gfn_t gfn,
2972
			 gva_t gva, pfn_t *pfn, bool write, bool *writable);
2973
static void make_mmu_pages_available(struct kvm_vcpu *vcpu);
2974

2975 2976
static int nonpaging_map(struct kvm_vcpu *vcpu, gva_t v, u32 error_code,
			 gfn_t gfn, bool prefault)
2977 2978
{
	int r;
2979
	int level;
2980
	bool force_pt_level = false;
2981
	pfn_t pfn;
2982
	unsigned long mmu_seq;
2983
	bool map_writable, write = error_code & PFERR_WRITE_MASK;
2984

2985
	level = mapping_level(vcpu, gfn, &force_pt_level);
2986 2987 2988 2989 2990 2991 2992 2993
	if (likely(!force_pt_level)) {
		/*
		 * This path builds a PAE pagetable - so we can map
		 * 2mb pages at maximum. Therefore check if the level
		 * is larger than that.
		 */
		if (level > PT_DIRECTORY_LEVEL)
			level = PT_DIRECTORY_LEVEL;
2994

2995
		gfn &= ~(KVM_PAGES_PER_HPAGE(level) - 1);
2996
	}
M
Marcelo Tosatti 已提交
2997

2998 2999 3000
	if (fast_page_fault(vcpu, v, level, error_code))
		return 0;

3001
	mmu_seq = vcpu->kvm->mmu_notifier_seq;
3002
	smp_rmb();
3003

3004
	if (try_async_pf(vcpu, prefault, gfn, v, &pfn, write, &map_writable))
3005
		return 0;
3006

3007 3008
	if (handle_abnormal_pfn(vcpu, v, gfn, pfn, ACC_ALL, &r))
		return r;
3009

3010
	spin_lock(&vcpu->kvm->mmu_lock);
3011
	if (mmu_notifier_retry(vcpu->kvm, mmu_seq))
3012
		goto out_unlock;
3013
	make_mmu_pages_available(vcpu);
3014 3015
	if (likely(!force_pt_level))
		transparent_hugepage_adjust(vcpu, &gfn, &pfn, &level);
3016
	r = __direct_map(vcpu, write, map_writable, level, gfn, pfn, prefault);
3017 3018
	spin_unlock(&vcpu->kvm->mmu_lock);

3019
	return r;
3020 3021 3022 3023 3024

out_unlock:
	spin_unlock(&vcpu->kvm->mmu_lock);
	kvm_release_pfn_clean(pfn);
	return 0;
3025 3026 3027
}


3028 3029 3030
static void mmu_free_roots(struct kvm_vcpu *vcpu)
{
	int i;
3031
	struct kvm_mmu_page *sp;
3032
	LIST_HEAD(invalid_list);
3033

3034
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
A
Avi Kivity 已提交
3035
		return;
3036

3037 3038 3039
	if (vcpu->arch.mmu.shadow_root_level == PT64_ROOT_LEVEL &&
	    (vcpu->arch.mmu.root_level == PT64_ROOT_LEVEL ||
	     vcpu->arch.mmu.direct_map)) {
3040
		hpa_t root = vcpu->arch.mmu.root_hpa;
3041

3042
		spin_lock(&vcpu->kvm->mmu_lock);
3043 3044
		sp = page_header(root);
		--sp->root_count;
3045 3046 3047 3048
		if (!sp->root_count && sp->role.invalid) {
			kvm_mmu_prepare_zap_page(vcpu->kvm, sp, &invalid_list);
			kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
		}
3049
		spin_unlock(&vcpu->kvm->mmu_lock);
3050
		vcpu->arch.mmu.root_hpa = INVALID_PAGE;
3051 3052
		return;
	}
3053 3054

	spin_lock(&vcpu->kvm->mmu_lock);
3055
	for (i = 0; i < 4; ++i) {
3056
		hpa_t root = vcpu->arch.mmu.pae_root[i];
3057

A
Avi Kivity 已提交
3058 3059
		if (root) {
			root &= PT64_BASE_ADDR_MASK;
3060 3061
			sp = page_header(root);
			--sp->root_count;
3062
			if (!sp->root_count && sp->role.invalid)
3063 3064
				kvm_mmu_prepare_zap_page(vcpu->kvm, sp,
							 &invalid_list);
A
Avi Kivity 已提交
3065
		}
3066
		vcpu->arch.mmu.pae_root[i] = INVALID_PAGE;
3067
	}
3068
	kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
3069
	spin_unlock(&vcpu->kvm->mmu_lock);
3070
	vcpu->arch.mmu.root_hpa = INVALID_PAGE;
3071 3072
}

3073 3074 3075 3076 3077
static int mmu_check_root(struct kvm_vcpu *vcpu, gfn_t root_gfn)
{
	int ret = 0;

	if (!kvm_is_visible_gfn(vcpu->kvm, root_gfn)) {
3078
		kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu);
3079 3080 3081 3082 3083 3084
		ret = 1;
	}

	return ret;
}

3085 3086 3087
static int mmu_alloc_direct_roots(struct kvm_vcpu *vcpu)
{
	struct kvm_mmu_page *sp;
3088
	unsigned i;
3089 3090 3091

	if (vcpu->arch.mmu.shadow_root_level == PT64_ROOT_LEVEL) {
		spin_lock(&vcpu->kvm->mmu_lock);
3092
		make_mmu_pages_available(vcpu);
3093 3094 3095 3096 3097 3098 3099 3100 3101
		sp = kvm_mmu_get_page(vcpu, 0, 0, PT64_ROOT_LEVEL,
				      1, ACC_ALL, NULL);
		++sp->root_count;
		spin_unlock(&vcpu->kvm->mmu_lock);
		vcpu->arch.mmu.root_hpa = __pa(sp->spt);
	} else if (vcpu->arch.mmu.shadow_root_level == PT32E_ROOT_LEVEL) {
		for (i = 0; i < 4; ++i) {
			hpa_t root = vcpu->arch.mmu.pae_root[i];

3102
			MMU_WARN_ON(VALID_PAGE(root));
3103
			spin_lock(&vcpu->kvm->mmu_lock);
3104
			make_mmu_pages_available(vcpu);
3105 3106
			sp = kvm_mmu_get_page(vcpu, i << (30 - PAGE_SHIFT),
					      i << 30,
3107 3108 3109 3110 3111 3112 3113
					      PT32_ROOT_LEVEL, 1, ACC_ALL,
					      NULL);
			root = __pa(sp->spt);
			++sp->root_count;
			spin_unlock(&vcpu->kvm->mmu_lock);
			vcpu->arch.mmu.pae_root[i] = root | PT_PRESENT_MASK;
		}
3114
		vcpu->arch.mmu.root_hpa = __pa(vcpu->arch.mmu.pae_root);
3115 3116 3117 3118 3119 3120 3121
	} else
		BUG();

	return 0;
}

static int mmu_alloc_shadow_roots(struct kvm_vcpu *vcpu)
3122
{
3123
	struct kvm_mmu_page *sp;
3124 3125 3126
	u64 pdptr, pm_mask;
	gfn_t root_gfn;
	int i;
3127

3128
	root_gfn = vcpu->arch.mmu.get_cr3(vcpu) >> PAGE_SHIFT;
3129

3130 3131 3132 3133 3134 3135 3136 3137
	if (mmu_check_root(vcpu, root_gfn))
		return 1;

	/*
	 * Do we shadow a long mode page table? If so we need to
	 * write-protect the guests page table root.
	 */
	if (vcpu->arch.mmu.root_level == PT64_ROOT_LEVEL) {
3138
		hpa_t root = vcpu->arch.mmu.root_hpa;
3139

3140
		MMU_WARN_ON(VALID_PAGE(root));
3141

3142
		spin_lock(&vcpu->kvm->mmu_lock);
3143
		make_mmu_pages_available(vcpu);
3144 3145
		sp = kvm_mmu_get_page(vcpu, root_gfn, 0, PT64_ROOT_LEVEL,
				      0, ACC_ALL, NULL);
3146 3147
		root = __pa(sp->spt);
		++sp->root_count;
3148
		spin_unlock(&vcpu->kvm->mmu_lock);
3149
		vcpu->arch.mmu.root_hpa = root;
3150
		return 0;
3151
	}
3152

3153 3154
	/*
	 * We shadow a 32 bit page table. This may be a legacy 2-level
3155 3156
	 * or a PAE 3-level page table. In either case we need to be aware that
	 * the shadow page table may be a PAE or a long mode page table.
3157
	 */
3158 3159 3160 3161
	pm_mask = PT_PRESENT_MASK;
	if (vcpu->arch.mmu.shadow_root_level == PT64_ROOT_LEVEL)
		pm_mask |= PT_ACCESSED_MASK | PT_WRITABLE_MASK | PT_USER_MASK;

3162
	for (i = 0; i < 4; ++i) {
3163
		hpa_t root = vcpu->arch.mmu.pae_root[i];
3164

3165
		MMU_WARN_ON(VALID_PAGE(root));
3166
		if (vcpu->arch.mmu.root_level == PT32E_ROOT_LEVEL) {
3167
			pdptr = vcpu->arch.mmu.get_pdptr(vcpu, i);
3168
			if (!is_present_gpte(pdptr)) {
3169
				vcpu->arch.mmu.pae_root[i] = 0;
A
Avi Kivity 已提交
3170 3171
				continue;
			}
A
Avi Kivity 已提交
3172
			root_gfn = pdptr >> PAGE_SHIFT;
3173 3174
			if (mmu_check_root(vcpu, root_gfn))
				return 1;
3175
		}
3176
		spin_lock(&vcpu->kvm->mmu_lock);
3177
		make_mmu_pages_available(vcpu);
3178
		sp = kvm_mmu_get_page(vcpu, root_gfn, i << 30,
3179
				      PT32_ROOT_LEVEL, 0,
3180
				      ACC_ALL, NULL);
3181 3182
		root = __pa(sp->spt);
		++sp->root_count;
3183 3184
		spin_unlock(&vcpu->kvm->mmu_lock);

3185
		vcpu->arch.mmu.pae_root[i] = root | pm_mask;
3186
	}
3187
	vcpu->arch.mmu.root_hpa = __pa(vcpu->arch.mmu.pae_root);
3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213

	/*
	 * If we shadow a 32 bit page table with a long mode page
	 * table we enter this path.
	 */
	if (vcpu->arch.mmu.shadow_root_level == PT64_ROOT_LEVEL) {
		if (vcpu->arch.mmu.lm_root == NULL) {
			/*
			 * The additional page necessary for this is only
			 * allocated on demand.
			 */

			u64 *lm_root;

			lm_root = (void*)get_zeroed_page(GFP_KERNEL);
			if (lm_root == NULL)
				return 1;

			lm_root[0] = __pa(vcpu->arch.mmu.pae_root) | pm_mask;

			vcpu->arch.mmu.lm_root = lm_root;
		}

		vcpu->arch.mmu.root_hpa = __pa(vcpu->arch.mmu.lm_root);
	}

3214
	return 0;
3215 3216
}

3217 3218 3219 3220 3221 3222 3223 3224
static int mmu_alloc_roots(struct kvm_vcpu *vcpu)
{
	if (vcpu->arch.mmu.direct_map)
		return mmu_alloc_direct_roots(vcpu);
	else
		return mmu_alloc_shadow_roots(vcpu);
}

3225 3226 3227 3228 3229
static void mmu_sync_roots(struct kvm_vcpu *vcpu)
{
	int i;
	struct kvm_mmu_page *sp;

3230 3231 3232
	if (vcpu->arch.mmu.direct_map)
		return;

3233 3234
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
		return;
3235

3236
	vcpu_clear_mmio_info(vcpu, MMIO_GVA_ANY);
3237
	kvm_mmu_audit(vcpu, AUDIT_PRE_SYNC);
3238
	if (vcpu->arch.mmu.root_level == PT64_ROOT_LEVEL) {
3239 3240 3241
		hpa_t root = vcpu->arch.mmu.root_hpa;
		sp = page_header(root);
		mmu_sync_children(vcpu, sp);
3242
		kvm_mmu_audit(vcpu, AUDIT_POST_SYNC);
3243 3244 3245 3246 3247
		return;
	}
	for (i = 0; i < 4; ++i) {
		hpa_t root = vcpu->arch.mmu.pae_root[i];

3248
		if (root && VALID_PAGE(root)) {
3249 3250 3251 3252 3253
			root &= PT64_BASE_ADDR_MASK;
			sp = page_header(root);
			mmu_sync_children(vcpu, sp);
		}
	}
3254
	kvm_mmu_audit(vcpu, AUDIT_POST_SYNC);
3255 3256 3257 3258 3259 3260
}

void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu)
{
	spin_lock(&vcpu->kvm->mmu_lock);
	mmu_sync_roots(vcpu);
3261
	spin_unlock(&vcpu->kvm->mmu_lock);
3262
}
N
Nadav Har'El 已提交
3263
EXPORT_SYMBOL_GPL(kvm_mmu_sync_roots);
3264

3265
static gpa_t nonpaging_gva_to_gpa(struct kvm_vcpu *vcpu, gva_t vaddr,
3266
				  u32 access, struct x86_exception *exception)
A
Avi Kivity 已提交
3267
{
3268 3269
	if (exception)
		exception->error_code = 0;
A
Avi Kivity 已提交
3270 3271 3272
	return vaddr;
}

3273
static gpa_t nonpaging_gva_to_gpa_nested(struct kvm_vcpu *vcpu, gva_t vaddr,
3274 3275
					 u32 access,
					 struct x86_exception *exception)
3276
{
3277 3278
	if (exception)
		exception->error_code = 0;
3279
	return vcpu->arch.nested_mmu.translate_gpa(vcpu, vaddr, access, exception);
3280 3281
}

3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300
static bool
__is_rsvd_bits_set(struct rsvd_bits_validate *rsvd_check, u64 pte, int level)
{
	int bit7 = (pte >> 7) & 1, low6 = pte & 0x3f;

	return (pte & rsvd_check->rsvd_bits_mask[bit7][level-1]) |
		((rsvd_check->bad_mt_xwr & (1ull << low6)) != 0);
}

static bool is_rsvd_bits_set(struct kvm_mmu *mmu, u64 gpte, int level)
{
	return __is_rsvd_bits_set(&mmu->guest_rsvd_check, gpte, level);
}

static bool is_shadow_zero_bits_set(struct kvm_mmu *mmu, u64 spte, int level)
{
	return __is_rsvd_bits_set(&mmu->shadow_zero_check, spte, level);
}

3301 3302 3303 3304 3305 3306 3307 3308
static bool quickly_check_mmio_pf(struct kvm_vcpu *vcpu, u64 addr, bool direct)
{
	if (direct)
		return vcpu_match_mmio_gpa(vcpu, addr);

	return vcpu_match_mmio_gva(vcpu, addr);
}

3309 3310 3311
/* return true if reserved bit is detected on spte. */
static bool
walk_shadow_page_get_mmio_spte(struct kvm_vcpu *vcpu, u64 addr, u64 *sptep)
3312 3313
{
	struct kvm_shadow_walk_iterator iterator;
3314 3315 3316
	u64 sptes[PT64_ROOT_LEVEL], spte = 0ull;
	int root, leaf;
	bool reserved = false;
3317

3318
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
3319
		goto exit;
3320

3321
	walk_shadow_page_lockless_begin(vcpu);
3322

3323 3324
	for (shadow_walk_init(&iterator, vcpu, addr),
		 leaf = root = iterator.level;
3325 3326 3327 3328 3329
	     shadow_walk_okay(&iterator);
	     __shadow_walk_next(&iterator, spte)) {
		spte = mmu_spte_get_lockless(iterator.sptep);

		sptes[leaf - 1] = spte;
3330
		leaf--;
3331

3332 3333
		if (!is_shadow_present_pte(spte))
			break;
3334 3335

		reserved |= is_shadow_zero_bits_set(&vcpu->arch.mmu, spte,
3336
						    iterator.level);
3337 3338
	}

3339 3340
	walk_shadow_page_lockless_end(vcpu);

3341 3342 3343
	if (reserved) {
		pr_err("%s: detect reserved bits on spte, addr 0x%llx, dump hierarchy:\n",
		       __func__, addr);
3344
		while (root > leaf) {
3345 3346 3347 3348 3349 3350 3351 3352
			pr_err("------ spte 0x%llx level %d.\n",
			       sptes[root - 1], root);
			root--;
		}
	}
exit:
	*sptep = spte;
	return reserved;
3353 3354
}

3355
int handle_mmio_page_fault(struct kvm_vcpu *vcpu, u64 addr, bool direct)
3356 3357
{
	u64 spte;
3358
	bool reserved;
3359 3360

	if (quickly_check_mmio_pf(vcpu, addr, direct))
3361
		return RET_MMIO_PF_EMULATE;
3362

3363
	reserved = walk_shadow_page_get_mmio_spte(vcpu, addr, &spte);
3364
	if (WARN_ON(reserved))
3365
		return RET_MMIO_PF_BUG;
3366 3367 3368 3369 3370

	if (is_mmio_spte(spte)) {
		gfn_t gfn = get_mmio_spte_gfn(spte);
		unsigned access = get_mmio_spte_access(spte);

3371
		if (!check_mmio_spte(vcpu, spte))
3372 3373
			return RET_MMIO_PF_INVALID;

3374 3375
		if (direct)
			addr = 0;
X
Xiao Guangrong 已提交
3376 3377

		trace_handle_mmio_page_fault(addr, gfn, access);
3378
		vcpu_cache_mmio_info(vcpu, addr, gfn, access);
3379
		return RET_MMIO_PF_EMULATE;
3380 3381 3382 3383 3384 3385
	}

	/*
	 * If the page table is zapped by other cpus, let CPU fault again on
	 * the address.
	 */
3386
	return RET_MMIO_PF_RETRY;
3387
}
3388
EXPORT_SYMBOL_GPL(handle_mmio_page_fault);
3389

A
Avi Kivity 已提交
3390
static int nonpaging_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
3391
				u32 error_code, bool prefault)
A
Avi Kivity 已提交
3392
{
3393
	gfn_t gfn;
3394
	int r;
A
Avi Kivity 已提交
3395

3396
	pgprintk("%s: gva %lx error %x\n", __func__, gva, error_code);
3397

3398
	if (unlikely(error_code & PFERR_RSVD_MASK)) {
3399
		r = handle_mmio_page_fault(vcpu, gva, true);
3400 3401 3402 3403

		if (likely(r != RET_MMIO_PF_INVALID))
			return r;
	}
3404

3405 3406 3407
	r = mmu_topup_memory_caches(vcpu);
	if (r)
		return r;
3408

3409
	MMU_WARN_ON(!VALID_PAGE(vcpu->arch.mmu.root_hpa));
A
Avi Kivity 已提交
3410

3411
	gfn = gva >> PAGE_SHIFT;
A
Avi Kivity 已提交
3412

3413
	return nonpaging_map(vcpu, gva & PAGE_MASK,
3414
			     error_code, gfn, prefault);
A
Avi Kivity 已提交
3415 3416
}

3417
static int kvm_arch_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn)
3418 3419
{
	struct kvm_arch_async_pf arch;
X
Xiao Guangrong 已提交
3420

3421
	arch.token = (vcpu->arch.apf.id++ << 12) | vcpu->vcpu_id;
3422
	arch.gfn = gfn;
3423
	arch.direct_map = vcpu->arch.mmu.direct_map;
X
Xiao Guangrong 已提交
3424
	arch.cr3 = vcpu->arch.mmu.get_cr3(vcpu);
3425

3426
	return kvm_setup_async_pf(vcpu, gva, kvm_vcpu_gfn_to_hva(vcpu, gfn), &arch);
3427 3428 3429 3430
}

static bool can_do_async_pf(struct kvm_vcpu *vcpu)
{
3431
	if (unlikely(!lapic_in_kernel(vcpu) ||
3432 3433 3434 3435 3436 3437
		     kvm_event_needs_reinjection(vcpu)))
		return false;

	return kvm_x86_ops->interrupt_allowed(vcpu);
}

3438
static bool try_async_pf(struct kvm_vcpu *vcpu, bool prefault, gfn_t gfn,
3439
			 gva_t gva, pfn_t *pfn, bool write, bool *writable)
3440
{
3441
	struct kvm_memory_slot *slot;
3442 3443
	bool async;

3444
	slot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);
3445 3446
	async = false;
	*pfn = __gfn_to_pfn_memslot(slot, gfn, false, &async, write, writable);
3447 3448 3449
	if (!async)
		return false; /* *pfn has correct page already */

3450
	if (!prefault && can_do_async_pf(vcpu)) {
3451
		trace_kvm_try_async_get_page(gva, gfn);
3452 3453 3454 3455 3456 3457 3458 3459
		if (kvm_find_async_pf_gfn(vcpu, gfn)) {
			trace_kvm_async_pf_doublefault(gva, gfn);
			kvm_make_request(KVM_REQ_APF_HALT, vcpu);
			return true;
		} else if (kvm_arch_setup_async_pf(vcpu, gva, gfn))
			return true;
	}

3460
	*pfn = __gfn_to_pfn_memslot(slot, gfn, false, NULL, write, writable);
3461 3462 3463
	return false;
}

3464 3465 3466 3467 3468 3469 3470 3471 3472 3473
static bool
check_hugepage_cache_consistency(struct kvm_vcpu *vcpu, gfn_t gfn, int level)
{
	int page_num = KVM_PAGES_PER_HPAGE(level);

	gfn &= ~(page_num - 1);

	return kvm_mtrr_check_gfn_range_consistency(vcpu, gfn, page_num);
}

G
Gleb Natapov 已提交
3474
static int tdp_page_fault(struct kvm_vcpu *vcpu, gva_t gpa, u32 error_code,
3475
			  bool prefault)
3476
{
3477
	pfn_t pfn;
3478
	int r;
3479
	int level;
3480
	bool force_pt_level;
M
Marcelo Tosatti 已提交
3481
	gfn_t gfn = gpa >> PAGE_SHIFT;
3482
	unsigned long mmu_seq;
3483 3484
	int write = error_code & PFERR_WRITE_MASK;
	bool map_writable;
3485

3486
	MMU_WARN_ON(!VALID_PAGE(vcpu->arch.mmu.root_hpa));
3487

3488
	if (unlikely(error_code & PFERR_RSVD_MASK)) {
3489
		r = handle_mmio_page_fault(vcpu, gpa, true);
3490 3491 3492 3493

		if (likely(r != RET_MMIO_PF_INVALID))
			return r;
	}
3494

3495 3496 3497 3498
	r = mmu_topup_memory_caches(vcpu);
	if (r)
		return r;

3499 3500 3501
	force_pt_level = !check_hugepage_cache_consistency(vcpu, gfn,
							   PT_DIRECTORY_LEVEL);
	level = mapping_level(vcpu, gfn, &force_pt_level);
3502
	if (likely(!force_pt_level)) {
3503 3504 3505
		if (level > PT_DIRECTORY_LEVEL &&
		    !check_hugepage_cache_consistency(vcpu, gfn, level))
			level = PT_DIRECTORY_LEVEL;
3506
		gfn &= ~(KVM_PAGES_PER_HPAGE(level) - 1);
3507
	}
3508

3509 3510 3511
	if (fast_page_fault(vcpu, gpa, level, error_code))
		return 0;

3512
	mmu_seq = vcpu->kvm->mmu_notifier_seq;
3513
	smp_rmb();
3514

3515
	if (try_async_pf(vcpu, prefault, gfn, gpa, &pfn, write, &map_writable))
3516 3517
		return 0;

3518 3519 3520
	if (handle_abnormal_pfn(vcpu, 0, gfn, pfn, ACC_ALL, &r))
		return r;

3521
	spin_lock(&vcpu->kvm->mmu_lock);
3522
	if (mmu_notifier_retry(vcpu->kvm, mmu_seq))
3523
		goto out_unlock;
3524
	make_mmu_pages_available(vcpu);
3525 3526
	if (likely(!force_pt_level))
		transparent_hugepage_adjust(vcpu, &gfn, &pfn, &level);
3527
	r = __direct_map(vcpu, write, map_writable, level, gfn, pfn, prefault);
3528 3529 3530
	spin_unlock(&vcpu->kvm->mmu_lock);

	return r;
3531 3532 3533 3534 3535

out_unlock:
	spin_unlock(&vcpu->kvm->mmu_lock);
	kvm_release_pfn_clean(pfn);
	return 0;
3536 3537
}

3538 3539
static void nonpaging_init_context(struct kvm_vcpu *vcpu,
				   struct kvm_mmu *context)
A
Avi Kivity 已提交
3540 3541 3542
{
	context->page_fault = nonpaging_page_fault;
	context->gva_to_gpa = nonpaging_gva_to_gpa;
3543
	context->sync_page = nonpaging_sync_page;
M
Marcelo Tosatti 已提交
3544
	context->invlpg = nonpaging_invlpg;
3545
	context->update_pte = nonpaging_update_pte;
3546
	context->root_level = 0;
A
Avi Kivity 已提交
3547
	context->shadow_root_level = PT32E_ROOT_LEVEL;
A
Avi Kivity 已提交
3548
	context->root_hpa = INVALID_PAGE;
3549
	context->direct_map = true;
3550
	context->nx = false;
A
Avi Kivity 已提交
3551 3552
}

3553
void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
3554
{
3555
	mmu_free_roots(vcpu);
A
Avi Kivity 已提交
3556 3557
}

3558 3559
static unsigned long get_cr3(struct kvm_vcpu *vcpu)
{
3560
	return kvm_read_cr3(vcpu);
3561 3562
}

3563 3564
static void inject_page_fault(struct kvm_vcpu *vcpu,
			      struct x86_exception *fault)
A
Avi Kivity 已提交
3565
{
3566
	vcpu->arch.mmu.inject_page_fault(vcpu, fault);
A
Avi Kivity 已提交
3567 3568
}

3569
static bool sync_mmio_spte(struct kvm_vcpu *vcpu, u64 *sptep, gfn_t gfn,
3570
			   unsigned access, int *nr_present)
3571 3572 3573 3574 3575 3576 3577 3578
{
	if (unlikely(is_mmio_spte(*sptep))) {
		if (gfn != get_mmio_spte_gfn(*sptep)) {
			mmu_spte_clear_no_track(sptep);
			return true;
		}

		(*nr_present)++;
3579
		mark_mmio_spte(vcpu, sptep, gfn, access);
3580 3581 3582 3583 3584 3585
		return true;
	}

	return false;
}

A
Avi Kivity 已提交
3586 3587 3588 3589 3590 3591 3592 3593 3594
static inline bool is_last_gpte(struct kvm_mmu *mmu, unsigned level, unsigned gpte)
{
	unsigned index;

	index = level - 1;
	index |= (gpte & PT_PAGE_SIZE_MASK) >> (PT_PAGE_SIZE_SHIFT - 2);
	return mmu->last_pte_bitmap & (1 << index);
}

3595 3596 3597 3598 3599
#define PTTYPE_EPT 18 /* arbitrary */
#define PTTYPE PTTYPE_EPT
#include "paging_tmpl.h"
#undef PTTYPE

A
Avi Kivity 已提交
3600 3601 3602 3603 3604 3605 3606 3607
#define PTTYPE 64
#include "paging_tmpl.h"
#undef PTTYPE

#define PTTYPE 32
#include "paging_tmpl.h"
#undef PTTYPE

3608 3609 3610 3611
static void
__reset_rsvds_bits_mask(struct kvm_vcpu *vcpu,
			struct rsvd_bits_validate *rsvd_check,
			int maxphyaddr, int level, bool nx, bool gbpages,
3612
			bool pse, bool amd)
3613 3614
{
	u64 exb_bit_rsvd = 0;
3615
	u64 gbpages_bit_rsvd = 0;
3616
	u64 nonleaf_bit8_rsvd = 0;
3617

3618
	rsvd_check->bad_mt_xwr = 0;
3619

3620
	if (!nx)
3621
		exb_bit_rsvd = rsvd_bits(63, 63);
3622
	if (!gbpages)
3623
		gbpages_bit_rsvd = rsvd_bits(7, 7);
3624 3625 3626 3627 3628

	/*
	 * Non-leaf PML4Es and PDPEs reserve bit 8 (which would be the G bit for
	 * leaf entries) on AMD CPUs only.
	 */
3629
	if (amd)
3630 3631
		nonleaf_bit8_rsvd = rsvd_bits(8, 8);

3632
	switch (level) {
3633 3634
	case PT32_ROOT_LEVEL:
		/* no rsvd bits for 2 level 4K page table entries */
3635 3636 3637 3638
		rsvd_check->rsvd_bits_mask[0][1] = 0;
		rsvd_check->rsvd_bits_mask[0][0] = 0;
		rsvd_check->rsvd_bits_mask[1][0] =
			rsvd_check->rsvd_bits_mask[0][0];
3639

3640
		if (!pse) {
3641
			rsvd_check->rsvd_bits_mask[1][1] = 0;
3642 3643 3644
			break;
		}

3645 3646
		if (is_cpuid_PSE36())
			/* 36bits PSE 4MB page */
3647
			rsvd_check->rsvd_bits_mask[1][1] = rsvd_bits(17, 21);
3648 3649
		else
			/* 32 bits PSE 4MB page */
3650
			rsvd_check->rsvd_bits_mask[1][1] = rsvd_bits(13, 21);
3651 3652
		break;
	case PT32E_ROOT_LEVEL:
3653
		rsvd_check->rsvd_bits_mask[0][2] =
3654
			rsvd_bits(maxphyaddr, 63) |
3655
			rsvd_bits(5, 8) | rsvd_bits(1, 2);	/* PDPTE */
3656
		rsvd_check->rsvd_bits_mask[0][1] = exb_bit_rsvd |
3657
			rsvd_bits(maxphyaddr, 62);	/* PDE */
3658
		rsvd_check->rsvd_bits_mask[0][0] = exb_bit_rsvd |
3659
			rsvd_bits(maxphyaddr, 62); 	/* PTE */
3660
		rsvd_check->rsvd_bits_mask[1][1] = exb_bit_rsvd |
3661 3662
			rsvd_bits(maxphyaddr, 62) |
			rsvd_bits(13, 20);		/* large page */
3663 3664
		rsvd_check->rsvd_bits_mask[1][0] =
			rsvd_check->rsvd_bits_mask[0][0];
3665 3666
		break;
	case PT64_ROOT_LEVEL:
3667 3668
		rsvd_check->rsvd_bits_mask[0][3] = exb_bit_rsvd |
			nonleaf_bit8_rsvd | rsvd_bits(7, 7) |
3669
			rsvd_bits(maxphyaddr, 51);
3670 3671
		rsvd_check->rsvd_bits_mask[0][2] = exb_bit_rsvd |
			nonleaf_bit8_rsvd | gbpages_bit_rsvd |
3672
			rsvd_bits(maxphyaddr, 51);
3673 3674 3675 3676 3677 3678 3679
		rsvd_check->rsvd_bits_mask[0][1] = exb_bit_rsvd |
			rsvd_bits(maxphyaddr, 51);
		rsvd_check->rsvd_bits_mask[0][0] = exb_bit_rsvd |
			rsvd_bits(maxphyaddr, 51);
		rsvd_check->rsvd_bits_mask[1][3] =
			rsvd_check->rsvd_bits_mask[0][3];
		rsvd_check->rsvd_bits_mask[1][2] = exb_bit_rsvd |
3680
			gbpages_bit_rsvd | rsvd_bits(maxphyaddr, 51) |
3681
			rsvd_bits(13, 29);
3682
		rsvd_check->rsvd_bits_mask[1][1] = exb_bit_rsvd |
3683 3684
			rsvd_bits(maxphyaddr, 51) |
			rsvd_bits(13, 20);		/* large page */
3685 3686
		rsvd_check->rsvd_bits_mask[1][0] =
			rsvd_check->rsvd_bits_mask[0][0];
3687 3688 3689 3690
		break;
	}
}

3691 3692 3693 3694 3695 3696
static void reset_rsvds_bits_mask(struct kvm_vcpu *vcpu,
				  struct kvm_mmu *context)
{
	__reset_rsvds_bits_mask(vcpu, &context->guest_rsvd_check,
				cpuid_maxphyaddr(vcpu), context->root_level,
				context->nx, guest_cpuid_has_gbpages(vcpu),
3697
				is_pse(vcpu), guest_cpuid_is_amd(vcpu));
3698 3699
}

3700 3701 3702
static void
__reset_rsvds_bits_mask_ept(struct rsvd_bits_validate *rsvd_check,
			    int maxphyaddr, bool execonly)
3703
{
3704
	u64 bad_mt_xwr;
3705

3706
	rsvd_check->rsvd_bits_mask[0][3] =
3707
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(3, 7);
3708
	rsvd_check->rsvd_bits_mask[0][2] =
3709
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(3, 6);
3710
	rsvd_check->rsvd_bits_mask[0][1] =
3711
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(3, 6);
3712
	rsvd_check->rsvd_bits_mask[0][0] = rsvd_bits(maxphyaddr, 51);
3713 3714

	/* large page */
3715 3716
	rsvd_check->rsvd_bits_mask[1][3] = rsvd_check->rsvd_bits_mask[0][3];
	rsvd_check->rsvd_bits_mask[1][2] =
3717
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(12, 29);
3718
	rsvd_check->rsvd_bits_mask[1][1] =
3719
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(12, 20);
3720
	rsvd_check->rsvd_bits_mask[1][0] = rsvd_check->rsvd_bits_mask[0][0];
3721

3722 3723 3724 3725 3726 3727 3728 3729
	bad_mt_xwr = 0xFFull << (2 * 8);	/* bits 3..5 must not be 2 */
	bad_mt_xwr |= 0xFFull << (3 * 8);	/* bits 3..5 must not be 3 */
	bad_mt_xwr |= 0xFFull << (7 * 8);	/* bits 3..5 must not be 7 */
	bad_mt_xwr |= REPEAT_BYTE(1ull << 2);	/* bits 0..2 must not be 010 */
	bad_mt_xwr |= REPEAT_BYTE(1ull << 6);	/* bits 0..2 must not be 110 */
	if (!execonly) {
		/* bits 0..2 must not be 100 unless VMX capabilities allow it */
		bad_mt_xwr |= REPEAT_BYTE(1ull << 4);
3730
	}
3731
	rsvd_check->bad_mt_xwr = bad_mt_xwr;
3732 3733
}

3734 3735 3736 3737 3738 3739 3740
static void reset_rsvds_bits_mask_ept(struct kvm_vcpu *vcpu,
		struct kvm_mmu *context, bool execonly)
{
	__reset_rsvds_bits_mask_ept(&context->guest_rsvd_check,
				    cpuid_maxphyaddr(vcpu), execonly);
}

3741 3742 3743 3744 3745 3746 3747 3748
/*
 * the page table on host is the shadow page table for the page
 * table in guest or amd nested guest, its mmu features completely
 * follow the features in guest.
 */
void
reset_shadow_zero_bits_mask(struct kvm_vcpu *vcpu, struct kvm_mmu *context)
{
3749 3750 3751 3752
	/*
	 * Passing "true" to the last argument is okay; it adds a check
	 * on bit 8 of the SPTEs which KVM doesn't use anyway.
	 */
3753 3754 3755
	__reset_rsvds_bits_mask(vcpu, &context->shadow_zero_check,
				boot_cpu_data.x86_phys_bits,
				context->shadow_root_level, context->nx,
3756 3757
				guest_cpuid_has_gbpages(vcpu), is_pse(vcpu),
				true);
3758 3759 3760
}
EXPORT_SYMBOL_GPL(reset_shadow_zero_bits_mask);

3761 3762 3763 3764 3765 3766
static inline bool boot_cpu_is_amd(void)
{
	WARN_ON_ONCE(!tdp_enabled);
	return shadow_x_mask == 0;
}

3767 3768 3769 3770 3771 3772 3773 3774
/*
 * the direct page table on host, use as much mmu features as
 * possible, however, kvm currently does not do execution-protection.
 */
static void
reset_tdp_shadow_zero_bits_mask(struct kvm_vcpu *vcpu,
				struct kvm_mmu *context)
{
3775
	if (boot_cpu_is_amd())
3776 3777 3778
		__reset_rsvds_bits_mask(vcpu, &context->shadow_zero_check,
					boot_cpu_data.x86_phys_bits,
					context->shadow_root_level, false,
3779
					cpu_has_gbpages, true, true);
3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798
	else
		__reset_rsvds_bits_mask_ept(&context->shadow_zero_check,
					    boot_cpu_data.x86_phys_bits,
					    false);

}

/*
 * as the comments in reset_shadow_zero_bits_mask() except it
 * is the shadow page table for intel nested guest.
 */
static void
reset_ept_shadow_zero_bits_mask(struct kvm_vcpu *vcpu,
				struct kvm_mmu *context, bool execonly)
{
	__reset_rsvds_bits_mask_ept(&context->shadow_zero_check,
				    boot_cpu_data.x86_phys_bits, execonly);
}

3799 3800
static void update_permission_bitmask(struct kvm_vcpu *vcpu,
				      struct kvm_mmu *mmu, bool ept)
3801 3802 3803
{
	unsigned bit, byte, pfec;
	u8 map;
F
Feng Wu 已提交
3804
	bool fault, x, w, u, wf, uf, ff, smapf, cr4_smap, cr4_smep, smap = 0;
3805

F
Feng Wu 已提交
3806
	cr4_smep = kvm_read_cr4_bits(vcpu, X86_CR4_SMEP);
F
Feng Wu 已提交
3807
	cr4_smap = kvm_read_cr4_bits(vcpu, X86_CR4_SMAP);
3808 3809 3810 3811 3812 3813
	for (byte = 0; byte < ARRAY_SIZE(mmu->permissions); ++byte) {
		pfec = byte << 1;
		map = 0;
		wf = pfec & PFERR_WRITE_MASK;
		uf = pfec & PFERR_USER_MASK;
		ff = pfec & PFERR_FETCH_MASK;
F
Feng Wu 已提交
3814 3815 3816 3817 3818 3819
		/*
		 * PFERR_RSVD_MASK bit is set in PFEC if the access is not
		 * subject to SMAP restrictions, and cleared otherwise. The
		 * bit is only meaningful if the SMAP bit is set in CR4.
		 */
		smapf = !(pfec & PFERR_RSVD_MASK);
3820 3821 3822 3823 3824
		for (bit = 0; bit < 8; ++bit) {
			x = bit & ACC_EXEC_MASK;
			w = bit & ACC_WRITE_MASK;
			u = bit & ACC_USER_MASK;

3825 3826 3827 3828 3829 3830
			if (!ept) {
				/* Not really needed: !nx will cause pte.nx to fault */
				x |= !mmu->nx;
				/* Allow supervisor writes if !cr0.wp */
				w |= !is_write_protection(vcpu) && !uf;
				/* Disallow supervisor fetches of user code if cr4.smep */
F
Feng Wu 已提交
3831
				x &= !(cr4_smep && u && !uf);
F
Feng Wu 已提交
3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851

				/*
				 * SMAP:kernel-mode data accesses from user-mode
				 * mappings should fault. A fault is considered
				 * as a SMAP violation if all of the following
				 * conditions are ture:
				 *   - X86_CR4_SMAP is set in CR4
				 *   - An user page is accessed
				 *   - Page fault in kernel mode
				 *   - if CPL = 3 or X86_EFLAGS_AC is clear
				 *
				 *   Here, we cover the first three conditions.
				 *   The fourth is computed dynamically in
				 *   permission_fault() and is in smapf.
				 *
				 *   Also, SMAP does not affect instruction
				 *   fetches, add the !ff check here to make it
				 *   clearer.
				 */
				smap = cr4_smap && u && !uf && !ff;
3852 3853 3854
			} else
				/* Not really needed: no U/S accesses on ept  */
				u = 1;
3855

F
Feng Wu 已提交
3856 3857
			fault = (ff && !x) || (uf && !u) || (wf && !w) ||
				(smapf && smap);
3858 3859 3860 3861 3862 3863
			map |= fault << bit;
		}
		mmu->permissions[byte] = map;
	}
}

A
Avi Kivity 已提交
3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881
static void update_last_pte_bitmap(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu)
{
	u8 map;
	unsigned level, root_level = mmu->root_level;
	const unsigned ps_set_index = 1 << 2;  /* bit 2 of index: ps */

	if (root_level == PT32E_ROOT_LEVEL)
		--root_level;
	/* PT_PAGE_TABLE_LEVEL always terminates */
	map = 1 | (1 << ps_set_index);
	for (level = PT_DIRECTORY_LEVEL; level <= root_level; ++level) {
		if (level <= PT_PDPE_LEVEL
		    && (mmu->root_level >= PT32E_ROOT_LEVEL || is_pse(vcpu)))
			map |= 1 << (ps_set_index | (level - 1));
	}
	mmu->last_pte_bitmap = map;
}

3882 3883 3884
static void paging64_init_context_common(struct kvm_vcpu *vcpu,
					 struct kvm_mmu *context,
					 int level)
A
Avi Kivity 已提交
3885
{
3886
	context->nx = is_nx(vcpu);
3887
	context->root_level = level;
3888

3889
	reset_rsvds_bits_mask(vcpu, context);
3890
	update_permission_bitmask(vcpu, context, false);
A
Avi Kivity 已提交
3891
	update_last_pte_bitmap(vcpu, context);
A
Avi Kivity 已提交
3892

3893
	MMU_WARN_ON(!is_pae(vcpu));
A
Avi Kivity 已提交
3894 3895
	context->page_fault = paging64_page_fault;
	context->gva_to_gpa = paging64_gva_to_gpa;
3896
	context->sync_page = paging64_sync_page;
M
Marcelo Tosatti 已提交
3897
	context->invlpg = paging64_invlpg;
3898
	context->update_pte = paging64_update_pte;
3899
	context->shadow_root_level = level;
A
Avi Kivity 已提交
3900
	context->root_hpa = INVALID_PAGE;
3901
	context->direct_map = false;
A
Avi Kivity 已提交
3902 3903
}

3904 3905
static void paging64_init_context(struct kvm_vcpu *vcpu,
				  struct kvm_mmu *context)
3906
{
3907
	paging64_init_context_common(vcpu, context, PT64_ROOT_LEVEL);
3908 3909
}

3910 3911
static void paging32_init_context(struct kvm_vcpu *vcpu,
				  struct kvm_mmu *context)
A
Avi Kivity 已提交
3912
{
3913
	context->nx = false;
3914
	context->root_level = PT32_ROOT_LEVEL;
3915

3916
	reset_rsvds_bits_mask(vcpu, context);
3917
	update_permission_bitmask(vcpu, context, false);
A
Avi Kivity 已提交
3918
	update_last_pte_bitmap(vcpu, context);
A
Avi Kivity 已提交
3919 3920 3921

	context->page_fault = paging32_page_fault;
	context->gva_to_gpa = paging32_gva_to_gpa;
3922
	context->sync_page = paging32_sync_page;
M
Marcelo Tosatti 已提交
3923
	context->invlpg = paging32_invlpg;
3924
	context->update_pte = paging32_update_pte;
A
Avi Kivity 已提交
3925
	context->shadow_root_level = PT32E_ROOT_LEVEL;
A
Avi Kivity 已提交
3926
	context->root_hpa = INVALID_PAGE;
3927
	context->direct_map = false;
A
Avi Kivity 已提交
3928 3929
}

3930 3931
static void paging32E_init_context(struct kvm_vcpu *vcpu,
				   struct kvm_mmu *context)
A
Avi Kivity 已提交
3932
{
3933
	paging64_init_context_common(vcpu, context, PT32E_ROOT_LEVEL);
A
Avi Kivity 已提交
3934 3935
}

3936
static void init_kvm_tdp_mmu(struct kvm_vcpu *vcpu)
3937
{
3938
	struct kvm_mmu *context = &vcpu->arch.mmu;
3939

3940
	context->base_role.word = 0;
3941
	context->base_role.smm = is_smm(vcpu);
3942
	context->page_fault = tdp_page_fault;
3943
	context->sync_page = nonpaging_sync_page;
M
Marcelo Tosatti 已提交
3944
	context->invlpg = nonpaging_invlpg;
3945
	context->update_pte = nonpaging_update_pte;
3946
	context->shadow_root_level = kvm_x86_ops->get_tdp_level();
3947
	context->root_hpa = INVALID_PAGE;
3948
	context->direct_map = true;
3949
	context->set_cr3 = kvm_x86_ops->set_tdp_cr3;
3950
	context->get_cr3 = get_cr3;
3951
	context->get_pdptr = kvm_pdptr_read;
3952
	context->inject_page_fault = kvm_inject_page_fault;
3953 3954

	if (!is_paging(vcpu)) {
3955
		context->nx = false;
3956 3957 3958
		context->gva_to_gpa = nonpaging_gva_to_gpa;
		context->root_level = 0;
	} else if (is_long_mode(vcpu)) {
3959
		context->nx = is_nx(vcpu);
3960
		context->root_level = PT64_ROOT_LEVEL;
3961 3962
		reset_rsvds_bits_mask(vcpu, context);
		context->gva_to_gpa = paging64_gva_to_gpa;
3963
	} else if (is_pae(vcpu)) {
3964
		context->nx = is_nx(vcpu);
3965
		context->root_level = PT32E_ROOT_LEVEL;
3966 3967
		reset_rsvds_bits_mask(vcpu, context);
		context->gva_to_gpa = paging64_gva_to_gpa;
3968
	} else {
3969
		context->nx = false;
3970
		context->root_level = PT32_ROOT_LEVEL;
3971 3972
		reset_rsvds_bits_mask(vcpu, context);
		context->gva_to_gpa = paging32_gva_to_gpa;
3973 3974
	}

3975
	update_permission_bitmask(vcpu, context, false);
A
Avi Kivity 已提交
3976
	update_last_pte_bitmap(vcpu, context);
3977
	reset_tdp_shadow_zero_bits_mask(vcpu, context);
3978 3979
}

3980
void kvm_init_shadow_mmu(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
3981
{
3982
	bool smep = kvm_read_cr4_bits(vcpu, X86_CR4_SMEP);
3983
	bool smap = kvm_read_cr4_bits(vcpu, X86_CR4_SMAP);
3984 3985
	struct kvm_mmu *context = &vcpu->arch.mmu;

3986
	MMU_WARN_ON(VALID_PAGE(context->root_hpa));
A
Avi Kivity 已提交
3987 3988

	if (!is_paging(vcpu))
3989
		nonpaging_init_context(vcpu, context);
A
Avi Kivity 已提交
3990
	else if (is_long_mode(vcpu))
3991
		paging64_init_context(vcpu, context);
A
Avi Kivity 已提交
3992
	else if (is_pae(vcpu))
3993
		paging32E_init_context(vcpu, context);
A
Avi Kivity 已提交
3994
	else
3995
		paging32_init_context(vcpu, context);
3996

3997 3998 3999 4000
	context->base_role.nxe = is_nx(vcpu);
	context->base_role.cr4_pae = !!is_pae(vcpu);
	context->base_role.cr0_wp  = is_write_protection(vcpu);
	context->base_role.smep_andnot_wp
4001
		= smep && !is_write_protection(vcpu);
4002 4003
	context->base_role.smap_andnot_wp
		= smap && !is_write_protection(vcpu);
4004
	context->base_role.smm = is_smm(vcpu);
4005
	reset_shadow_zero_bits_mask(vcpu, context);
4006 4007 4008
}
EXPORT_SYMBOL_GPL(kvm_init_shadow_mmu);

4009
void kvm_init_shadow_ept_mmu(struct kvm_vcpu *vcpu, bool execonly)
N
Nadav Har'El 已提交
4010
{
4011 4012
	struct kvm_mmu *context = &vcpu->arch.mmu;

4013
	MMU_WARN_ON(VALID_PAGE(context->root_hpa));
N
Nadav Har'El 已提交
4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028

	context->shadow_root_level = kvm_x86_ops->get_tdp_level();

	context->nx = true;
	context->page_fault = ept_page_fault;
	context->gva_to_gpa = ept_gva_to_gpa;
	context->sync_page = ept_sync_page;
	context->invlpg = ept_invlpg;
	context->update_pte = ept_update_pte;
	context->root_level = context->shadow_root_level;
	context->root_hpa = INVALID_PAGE;
	context->direct_map = false;

	update_permission_bitmask(vcpu, context, true);
	reset_rsvds_bits_mask_ept(vcpu, context, execonly);
4029
	reset_ept_shadow_zero_bits_mask(vcpu, context, execonly);
N
Nadav Har'El 已提交
4030 4031 4032
}
EXPORT_SYMBOL_GPL(kvm_init_shadow_ept_mmu);

4033
static void init_kvm_softmmu(struct kvm_vcpu *vcpu)
4034
{
4035 4036 4037 4038 4039 4040 4041
	struct kvm_mmu *context = &vcpu->arch.mmu;

	kvm_init_shadow_mmu(vcpu);
	context->set_cr3           = kvm_x86_ops->set_cr3;
	context->get_cr3           = get_cr3;
	context->get_pdptr         = kvm_pdptr_read;
	context->inject_page_fault = kvm_inject_page_fault;
A
Avi Kivity 已提交
4042 4043
}

4044
static void init_kvm_nested_mmu(struct kvm_vcpu *vcpu)
4045 4046 4047 4048
{
	struct kvm_mmu *g_context = &vcpu->arch.nested_mmu;

	g_context->get_cr3           = get_cr3;
4049
	g_context->get_pdptr         = kvm_pdptr_read;
4050 4051 4052 4053 4054 4055 4056 4057 4058
	g_context->inject_page_fault = kvm_inject_page_fault;

	/*
	 * Note that arch.mmu.gva_to_gpa translates l2_gva to l1_gpa. The
	 * translation of l2_gpa to l1_gpa addresses is done using the
	 * arch.nested_mmu.gva_to_gpa function. Basically the gva_to_gpa
	 * functions between mmu and nested_mmu are swapped.
	 */
	if (!is_paging(vcpu)) {
4059
		g_context->nx = false;
4060 4061 4062
		g_context->root_level = 0;
		g_context->gva_to_gpa = nonpaging_gva_to_gpa_nested;
	} else if (is_long_mode(vcpu)) {
4063
		g_context->nx = is_nx(vcpu);
4064
		g_context->root_level = PT64_ROOT_LEVEL;
4065
		reset_rsvds_bits_mask(vcpu, g_context);
4066 4067
		g_context->gva_to_gpa = paging64_gva_to_gpa_nested;
	} else if (is_pae(vcpu)) {
4068
		g_context->nx = is_nx(vcpu);
4069
		g_context->root_level = PT32E_ROOT_LEVEL;
4070
		reset_rsvds_bits_mask(vcpu, g_context);
4071 4072
		g_context->gva_to_gpa = paging64_gva_to_gpa_nested;
	} else {
4073
		g_context->nx = false;
4074
		g_context->root_level = PT32_ROOT_LEVEL;
4075
		reset_rsvds_bits_mask(vcpu, g_context);
4076 4077 4078
		g_context->gva_to_gpa = paging32_gva_to_gpa_nested;
	}

4079
	update_permission_bitmask(vcpu, g_context, false);
A
Avi Kivity 已提交
4080
	update_last_pte_bitmap(vcpu, g_context);
4081 4082
}

4083
static void init_kvm_mmu(struct kvm_vcpu *vcpu)
4084
{
4085
	if (mmu_is_nested(vcpu))
4086
		init_kvm_nested_mmu(vcpu);
4087
	else if (tdp_enabled)
4088
		init_kvm_tdp_mmu(vcpu);
4089
	else
4090
		init_kvm_softmmu(vcpu);
4091 4092
}

4093
void kvm_mmu_reset_context(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4094
{
4095
	kvm_mmu_unload(vcpu);
4096
	init_kvm_mmu(vcpu);
A
Avi Kivity 已提交
4097
}
4098
EXPORT_SYMBOL_GPL(kvm_mmu_reset_context);
A
Avi Kivity 已提交
4099 4100

int kvm_mmu_load(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4101
{
4102 4103
	int r;

4104
	r = mmu_topup_memory_caches(vcpu);
A
Avi Kivity 已提交
4105 4106
	if (r)
		goto out;
4107
	r = mmu_alloc_roots(vcpu);
4108
	kvm_mmu_sync_roots(vcpu);
4109 4110
	if (r)
		goto out;
4111
	/* set_cr3() should ensure TLB has been flushed */
4112
	vcpu->arch.mmu.set_cr3(vcpu, vcpu->arch.mmu.root_hpa);
4113 4114
out:
	return r;
A
Avi Kivity 已提交
4115
}
A
Avi Kivity 已提交
4116 4117 4118 4119 4120
EXPORT_SYMBOL_GPL(kvm_mmu_load);

void kvm_mmu_unload(struct kvm_vcpu *vcpu)
{
	mmu_free_roots(vcpu);
4121
	WARN_ON(VALID_PAGE(vcpu->arch.mmu.root_hpa));
A
Avi Kivity 已提交
4122
}
4123
EXPORT_SYMBOL_GPL(kvm_mmu_unload);
A
Avi Kivity 已提交
4124

4125
static void mmu_pte_write_new_pte(struct kvm_vcpu *vcpu,
4126 4127
				  struct kvm_mmu_page *sp, u64 *spte,
				  const void *new)
4128
{
4129
	if (sp->role.level != PT_PAGE_TABLE_LEVEL) {
4130 4131
		++vcpu->kvm->stat.mmu_pde_zapped;
		return;
4132
        }
4133

A
Avi Kivity 已提交
4134
	++vcpu->kvm->stat.mmu_pte_updated;
4135
	vcpu->arch.mmu.update_pte(vcpu, sp, spte, new);
4136 4137
}

4138 4139 4140 4141 4142 4143 4144 4145
static bool need_remote_flush(u64 old, u64 new)
{
	if (!is_shadow_present_pte(old))
		return false;
	if (!is_shadow_present_pte(new))
		return true;
	if ((old ^ new) & PT64_BASE_ADDR_MASK)
		return true;
4146 4147
	old ^= shadow_nx_mask;
	new ^= shadow_nx_mask;
4148 4149 4150
	return (old & ~new & PT64_PERM_MASK) != 0;
}

4151 4152
static void mmu_pte_write_flush_tlb(struct kvm_vcpu *vcpu, bool zap_page,
				    bool remote_flush, bool local_flush)
4153
{
4154 4155 4156 4157
	if (zap_page)
		return;

	if (remote_flush)
4158
		kvm_flush_remote_tlbs(vcpu->kvm);
4159
	else if (local_flush)
4160
		kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
4161 4162
}

4163 4164
static u64 mmu_pte_write_fetch_gpte(struct kvm_vcpu *vcpu, gpa_t *gpa,
				    const u8 *new, int *bytes)
4165
{
4166 4167
	u64 gentry;
	int r;
4168 4169 4170

	/*
	 * Assume that the pte write on a page table of the same type
4171 4172
	 * as the current vcpu paging mode since we update the sptes only
	 * when they have the same mode.
4173
	 */
4174
	if (is_pae(vcpu) && *bytes == 4) {
4175
		/* Handle a 32-bit guest writing two halves of a 64-bit gpte */
4176 4177
		*gpa &= ~(gpa_t)7;
		*bytes = 8;
4178
		r = kvm_vcpu_read_guest(vcpu, *gpa, &gentry, 8);
4179 4180
		if (r)
			gentry = 0;
4181 4182 4183
		new = (const u8 *)&gentry;
	}

4184
	switch (*bytes) {
4185 4186 4187 4188 4189 4190 4191 4192 4193
	case 4:
		gentry = *(const u32 *)new;
		break;
	case 8:
		gentry = *(const u64 *)new;
		break;
	default:
		gentry = 0;
		break;
4194 4195
	}

4196 4197 4198 4199 4200 4201 4202
	return gentry;
}

/*
 * If we're seeing too many writes to a page, it may no longer be a page table,
 * or we may be forking, in which case it is better to unmap the page.
 */
4203
static bool detect_write_flooding(struct kvm_mmu_page *sp)
4204
{
4205 4206 4207 4208
	/*
	 * Skip write-flooding detected for the sp whose level is 1, because
	 * it can become unsync, then the guest page is not write-protected.
	 */
4209
	if (sp->role.level == PT_PAGE_TABLE_LEVEL)
4210
		return false;
4211

4212
	return ++sp->write_flooding_count >= 3;
4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228
}

/*
 * Misaligned accesses are too much trouble to fix up; also, they usually
 * indicate a page is not used as a page table.
 */
static bool detect_write_misaligned(struct kvm_mmu_page *sp, gpa_t gpa,
				    int bytes)
{
	unsigned offset, pte_size, misaligned;

	pgprintk("misaligned: gpa %llx bytes %d role %x\n",
		 gpa, bytes, sp->role.word);

	offset = offset_in_page(gpa);
	pte_size = sp->role.cr4_pae ? 8 : 4;
4229 4230 4231 4232 4233 4234 4235 4236

	/*
	 * Sometimes, the OS only writes the last one bytes to update status
	 * bits, for example, in linux, andb instruction is used in clear_bit().
	 */
	if (!(offset & (pte_size - 1)) && bytes == 1)
		return false;

4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281
	misaligned = (offset ^ (offset + bytes - 1)) & ~(pte_size - 1);
	misaligned |= bytes < 4;

	return misaligned;
}

static u64 *get_written_sptes(struct kvm_mmu_page *sp, gpa_t gpa, int *nspte)
{
	unsigned page_offset, quadrant;
	u64 *spte;
	int level;

	page_offset = offset_in_page(gpa);
	level = sp->role.level;
	*nspte = 1;
	if (!sp->role.cr4_pae) {
		page_offset <<= 1;	/* 32->64 */
		/*
		 * A 32-bit pde maps 4MB while the shadow pdes map
		 * only 2MB.  So we need to double the offset again
		 * and zap two pdes instead of one.
		 */
		if (level == PT32_ROOT_LEVEL) {
			page_offset &= ~7; /* kill rounding error */
			page_offset <<= 1;
			*nspte = 2;
		}
		quadrant = page_offset >> PAGE_SHIFT;
		page_offset &= ~PAGE_MASK;
		if (quadrant != sp->role.quadrant)
			return NULL;
	}

	spte = &sp->spt[page_offset / sizeof(*spte)];
	return spte;
}

void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
		       const u8 *new, int bytes)
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
	struct kvm_mmu_page *sp;
	LIST_HEAD(invalid_list);
	u64 entry, gentry, *spte;
	int npte;
4282
	bool remote_flush, local_flush, zap_page;
4283 4284 4285 4286 4287 4288 4289
	union kvm_mmu_page_role mask = { };

	mask.cr0_wp = 1;
	mask.cr4_pae = 1;
	mask.nxe = 1;
	mask.smep_andnot_wp = 1;
	mask.smap_andnot_wp = 1;
4290
	mask.smm = 1;
4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313

	/*
	 * If we don't have indirect shadow pages, it means no page is
	 * write-protected, so we can exit simply.
	 */
	if (!ACCESS_ONCE(vcpu->kvm->arch.indirect_shadow_pages))
		return;

	zap_page = remote_flush = local_flush = false;

	pgprintk("%s: gpa %llx bytes %d\n", __func__, gpa, bytes);

	gentry = mmu_pte_write_fetch_gpte(vcpu, &gpa, new, &bytes);

	/*
	 * No need to care whether allocation memory is successful
	 * or not since pte prefetch is skiped if it does not have
	 * enough objects in the cache.
	 */
	mmu_topup_memory_caches(vcpu);

	spin_lock(&vcpu->kvm->mmu_lock);
	++vcpu->kvm->stat.mmu_pte_write;
4314
	kvm_mmu_audit(vcpu, AUDIT_PRE_PTE_WRITE);
4315

4316
	for_each_gfn_indirect_valid_sp(vcpu->kvm, sp, gfn) {
4317
		if (detect_write_misaligned(sp, gpa, bytes) ||
4318
		      detect_write_flooding(sp)) {
4319
			zap_page |= !!kvm_mmu_prepare_zap_page(vcpu->kvm, sp,
4320
						     &invalid_list);
A
Avi Kivity 已提交
4321
			++vcpu->kvm->stat.mmu_flooded;
4322 4323
			continue;
		}
4324 4325 4326 4327 4328

		spte = get_written_sptes(sp, gpa, &npte);
		if (!spte)
			continue;

4329
		local_flush = true;
4330
		while (npte--) {
4331
			entry = *spte;
4332
			mmu_page_zap_pte(vcpu->kvm, sp, spte);
4333 4334
			if (gentry &&
			      !((sp->role.word ^ vcpu->arch.mmu.base_role.word)
4335
			      & mask.word) && rmap_can_add(vcpu))
4336
				mmu_pte_write_new_pte(vcpu, sp, spte, &gentry);
G
Gleb Natapov 已提交
4337
			if (need_remote_flush(entry, *spte))
4338
				remote_flush = true;
4339
			++spte;
4340 4341
		}
	}
4342
	mmu_pte_write_flush_tlb(vcpu, zap_page, remote_flush, local_flush);
4343
	kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
4344
	kvm_mmu_audit(vcpu, AUDIT_POST_PTE_WRITE);
4345
	spin_unlock(&vcpu->kvm->mmu_lock);
4346 4347
}

4348 4349
int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva)
{
4350 4351
	gpa_t gpa;
	int r;
4352

4353
	if (vcpu->arch.mmu.direct_map)
4354 4355
		return 0;

4356
	gpa = kvm_mmu_gva_to_gpa_read(vcpu, gva, NULL);
4357 4358

	r = kvm_mmu_unprotect_page(vcpu->kvm, gpa >> PAGE_SHIFT);
4359

4360
	return r;
4361
}
4362
EXPORT_SYMBOL_GPL(kvm_mmu_unprotect_page_virt);
4363

4364
static void make_mmu_pages_available(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4365
{
4366
	LIST_HEAD(invalid_list);
4367

4368 4369 4370
	if (likely(kvm_mmu_available_pages(vcpu->kvm) >= KVM_MIN_FREE_MMU_PAGES))
		return;

4371 4372 4373
	while (kvm_mmu_available_pages(vcpu->kvm) < KVM_REFILL_PAGES) {
		if (!prepare_zap_oldest_mmu_page(vcpu->kvm, &invalid_list))
			break;
A
Avi Kivity 已提交
4374

A
Avi Kivity 已提交
4375
		++vcpu->kvm->stat.mmu_recycled;
A
Avi Kivity 已提交
4376
	}
4377
	kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
A
Avi Kivity 已提交
4378 4379
}

4380 4381 4382 4383 4384 4385 4386 4387
static bool is_mmio_page_fault(struct kvm_vcpu *vcpu, gva_t addr)
{
	if (vcpu->arch.mmu.direct_map || mmu_is_nested(vcpu))
		return vcpu_match_mmio_gpa(vcpu, addr);

	return vcpu_match_mmio_gva(vcpu, addr);
}

4388 4389
int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t cr2, u32 error_code,
		       void *insn, int insn_len)
4390
{
4391
	int r, emulation_type = EMULTYPE_RETRY;
4392 4393
	enum emulation_result er;

G
Gleb Natapov 已提交
4394
	r = vcpu->arch.mmu.page_fault(vcpu, cr2, error_code, false);
4395 4396 4397 4398 4399 4400 4401 4402
	if (r < 0)
		goto out;

	if (!r) {
		r = 1;
		goto out;
	}

4403 4404 4405 4406
	if (is_mmio_page_fault(vcpu, cr2))
		emulation_type = 0;

	er = x86_emulate_instruction(vcpu, cr2, emulation_type, insn, insn_len);
4407 4408 4409 4410

	switch (er) {
	case EMULATE_DONE:
		return 1;
P
Paolo Bonzini 已提交
4411
	case EMULATE_USER_EXIT:
4412
		++vcpu->stat.mmio_exits;
4413
		/* fall through */
4414
	case EMULATE_FAIL:
4415
		return 0;
4416 4417 4418 4419 4420 4421 4422 4423
	default:
		BUG();
	}
out:
	return r;
}
EXPORT_SYMBOL_GPL(kvm_mmu_page_fault);

M
Marcelo Tosatti 已提交
4424 4425 4426
void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
{
	vcpu->arch.mmu.invlpg(vcpu, gva);
4427
	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
M
Marcelo Tosatti 已提交
4428 4429 4430 4431
	++vcpu->stat.invlpg;
}
EXPORT_SYMBOL_GPL(kvm_mmu_invlpg);

4432 4433 4434 4435 4436 4437
void kvm_enable_tdp(void)
{
	tdp_enabled = true;
}
EXPORT_SYMBOL_GPL(kvm_enable_tdp);

4438 4439 4440 4441 4442 4443
void kvm_disable_tdp(void)
{
	tdp_enabled = false;
}
EXPORT_SYMBOL_GPL(kvm_disable_tdp);

A
Avi Kivity 已提交
4444 4445
static void free_mmu_pages(struct kvm_vcpu *vcpu)
{
4446
	free_page((unsigned long)vcpu->arch.mmu.pae_root);
4447 4448
	if (vcpu->arch.mmu.lm_root != NULL)
		free_page((unsigned long)vcpu->arch.mmu.lm_root);
A
Avi Kivity 已提交
4449 4450 4451 4452
}

static int alloc_mmu_pages(struct kvm_vcpu *vcpu)
{
4453
	struct page *page;
A
Avi Kivity 已提交
4454 4455
	int i;

4456 4457 4458 4459 4460 4461 4462
	/*
	 * When emulating 32-bit mode, cr3 is only 32 bits even on x86_64.
	 * Therefore we need to allocate shadow page tables in the first
	 * 4GB of memory, which happens to fit the DMA32 zone.
	 */
	page = alloc_page(GFP_KERNEL | __GFP_DMA32);
	if (!page)
4463 4464
		return -ENOMEM;

4465
	vcpu->arch.mmu.pae_root = page_address(page);
4466
	for (i = 0; i < 4; ++i)
4467
		vcpu->arch.mmu.pae_root[i] = INVALID_PAGE;
4468

A
Avi Kivity 已提交
4469 4470 4471
	return 0;
}

4472
int kvm_mmu_create(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4473
{
4474 4475 4476 4477
	vcpu->arch.walk_mmu = &vcpu->arch.mmu;
	vcpu->arch.mmu.root_hpa = INVALID_PAGE;
	vcpu->arch.mmu.translate_gpa = translate_gpa;
	vcpu->arch.nested_mmu.translate_gpa = translate_nested_gpa;
A
Avi Kivity 已提交
4478

4479 4480
	return alloc_mmu_pages(vcpu);
}
A
Avi Kivity 已提交
4481

4482
void kvm_mmu_setup(struct kvm_vcpu *vcpu)
4483
{
4484
	MMU_WARN_ON(VALID_PAGE(vcpu->arch.mmu.root_hpa));
4485

4486
	init_kvm_mmu(vcpu);
A
Avi Kivity 已提交
4487 4488
}

4489
/* The return value indicates if tlb flush on all vcpus is needed. */
4490
typedef bool (*slot_level_handler) (struct kvm *kvm, struct kvm_rmap_head *rmap_head);
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 4555 4556 4557

/* The caller should hold mmu-lock before calling this function. */
static bool
slot_handle_level_range(struct kvm *kvm, struct kvm_memory_slot *memslot,
			slot_level_handler fn, int start_level, int end_level,
			gfn_t start_gfn, gfn_t end_gfn, bool lock_flush_tlb)
{
	struct slot_rmap_walk_iterator iterator;
	bool flush = false;

	for_each_slot_rmap_range(memslot, start_level, end_level, start_gfn,
			end_gfn, &iterator) {
		if (iterator.rmap)
			flush |= fn(kvm, iterator.rmap);

		if (need_resched() || spin_needbreak(&kvm->mmu_lock)) {
			if (flush && lock_flush_tlb) {
				kvm_flush_remote_tlbs(kvm);
				flush = false;
			}
			cond_resched_lock(&kvm->mmu_lock);
		}
	}

	if (flush && lock_flush_tlb) {
		kvm_flush_remote_tlbs(kvm);
		flush = false;
	}

	return flush;
}

static bool
slot_handle_level(struct kvm *kvm, struct kvm_memory_slot *memslot,
		  slot_level_handler fn, int start_level, int end_level,
		  bool lock_flush_tlb)
{
	return slot_handle_level_range(kvm, memslot, fn, start_level,
			end_level, memslot->base_gfn,
			memslot->base_gfn + memslot->npages - 1,
			lock_flush_tlb);
}

static bool
slot_handle_all_level(struct kvm *kvm, struct kvm_memory_slot *memslot,
		      slot_level_handler fn, bool lock_flush_tlb)
{
	return slot_handle_level(kvm, memslot, fn, PT_PAGE_TABLE_LEVEL,
				 PT_MAX_HUGEPAGE_LEVEL, lock_flush_tlb);
}

static bool
slot_handle_large_level(struct kvm *kvm, struct kvm_memory_slot *memslot,
			slot_level_handler fn, bool lock_flush_tlb)
{
	return slot_handle_level(kvm, memslot, fn, PT_PAGE_TABLE_LEVEL + 1,
				 PT_MAX_HUGEPAGE_LEVEL, lock_flush_tlb);
}

static bool
slot_handle_leaf(struct kvm *kvm, struct kvm_memory_slot *memslot,
		 slot_level_handler fn, bool lock_flush_tlb)
{
	return slot_handle_level(kvm, memslot, fn, PT_PAGE_TABLE_LEVEL,
				 PT_PAGE_TABLE_LEVEL, lock_flush_tlb);
}

X
Xiao Guangrong 已提交
4558 4559 4560 4561
void kvm_zap_gfn_range(struct kvm *kvm, gfn_t gfn_start, gfn_t gfn_end)
{
	struct kvm_memslots *slots;
	struct kvm_memory_slot *memslot;
4562
	int i;
X
Xiao Guangrong 已提交
4563 4564

	spin_lock(&kvm->mmu_lock);
4565 4566 4567 4568 4569 4570 4571 4572 4573
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
		slots = __kvm_memslots(kvm, i);
		kvm_for_each_memslot(memslot, slots) {
			gfn_t start, end;

			start = max(gfn_start, memslot->base_gfn);
			end = min(gfn_end, memslot->base_gfn + memslot->npages);
			if (start >= end)
				continue;
X
Xiao Guangrong 已提交
4574

4575 4576 4577 4578
			slot_handle_level_range(kvm, memslot, kvm_zap_rmapp,
						PT_PAGE_TABLE_LEVEL, PT_MAX_HUGEPAGE_LEVEL,
						start, end - 1, true);
		}
X
Xiao Guangrong 已提交
4579 4580 4581 4582 4583
	}

	spin_unlock(&kvm->mmu_lock);
}

4584 4585
static bool slot_rmap_write_protect(struct kvm *kvm,
				    struct kvm_rmap_head *rmap_head)
4586
{
4587
	return __rmap_write_protect(kvm, rmap_head, false);
4588 4589
}

4590 4591
void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
				      struct kvm_memory_slot *memslot)
A
Avi Kivity 已提交
4592
{
4593
	bool flush;
A
Avi Kivity 已提交
4594

4595
	spin_lock(&kvm->mmu_lock);
4596 4597
	flush = slot_handle_all_level(kvm, memslot, slot_rmap_write_protect,
				      false);
4598
	spin_unlock(&kvm->mmu_lock);
4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617

	/*
	 * kvm_mmu_slot_remove_write_access() and kvm_vm_ioctl_get_dirty_log()
	 * which do tlb flush out of mmu-lock should be serialized by
	 * kvm->slots_lock otherwise tlb flush would be missed.
	 */
	lockdep_assert_held(&kvm->slots_lock);

	/*
	 * We can flush all the TLBs out of the mmu lock without TLB
	 * corruption since we just change the spte from writable to
	 * readonly so that we only need to care the case of changing
	 * spte from present to present (changing the spte from present
	 * to nonpresent will flush all the TLBs immediately), in other
	 * words, the only case we care is mmu_spte_update() where we
	 * haved checked SPTE_HOST_WRITEABLE | SPTE_MMU_WRITEABLE
	 * instead of PT_WRITABLE_MASK, that means it does not depend
	 * on PT_WRITABLE_MASK anymore.
	 */
4618 4619
	if (flush)
		kvm_flush_remote_tlbs(kvm);
A
Avi Kivity 已提交
4620
}
4621

4622
static bool kvm_mmu_zap_collapsible_spte(struct kvm *kvm,
4623
					 struct kvm_rmap_head *rmap_head)
4624 4625 4626 4627 4628 4629 4630
{
	u64 *sptep;
	struct rmap_iterator iter;
	int need_tlb_flush = 0;
	pfn_t pfn;
	struct kvm_mmu_page *sp;

4631
restart:
4632
	for_each_rmap_spte(rmap_head, &iter, sptep) {
4633 4634 4635 4636
		sp = page_header(__pa(sptep));
		pfn = spte_to_pfn(*sptep);

		/*
4637 4638 4639 4640 4641
		 * We cannot do huge page mapping for indirect shadow pages,
		 * which are found on the last rmap (level = 1) when not using
		 * tdp; such shadow pages are synced with the page table in
		 * the guest, and the guest page table is using 4K page size
		 * mapping if the indirect sp has level = 1.
4642 4643 4644 4645 4646 4647
		 */
		if (sp->role.direct &&
			!kvm_is_reserved_pfn(pfn) &&
			PageTransCompound(pfn_to_page(pfn))) {
			drop_spte(kvm, sptep);
			need_tlb_flush = 1;
4648 4649
			goto restart;
		}
4650 4651 4652 4653 4654 4655
	}

	return need_tlb_flush;
}

void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
4656
				   const struct kvm_memory_slot *memslot)
4657
{
4658
	/* FIXME: const-ify all uses of struct kvm_memory_slot.  */
4659
	spin_lock(&kvm->mmu_lock);
4660 4661
	slot_handle_leaf(kvm, (struct kvm_memory_slot *)memslot,
			 kvm_mmu_zap_collapsible_spte, true);
4662 4663 4664
	spin_unlock(&kvm->mmu_lock);
}

4665 4666 4667
void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
				   struct kvm_memory_slot *memslot)
{
4668
	bool flush;
4669 4670

	spin_lock(&kvm->mmu_lock);
4671
	flush = slot_handle_leaf(kvm, memslot, __rmap_clear_dirty, false);
4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689
	spin_unlock(&kvm->mmu_lock);

	lockdep_assert_held(&kvm->slots_lock);

	/*
	 * It's also safe to flush TLBs out of mmu lock here as currently this
	 * function is only used for dirty logging, in which case flushing TLB
	 * out of mmu lock also guarantees no dirty pages will be lost in
	 * dirty_bitmap.
	 */
	if (flush)
		kvm_flush_remote_tlbs(kvm);
}
EXPORT_SYMBOL_GPL(kvm_mmu_slot_leaf_clear_dirty);

void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
					struct kvm_memory_slot *memslot)
{
4690
	bool flush;
4691 4692

	spin_lock(&kvm->mmu_lock);
4693 4694
	flush = slot_handle_large_level(kvm, memslot, slot_rmap_write_protect,
					false);
4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707
	spin_unlock(&kvm->mmu_lock);

	/* see kvm_mmu_slot_remove_write_access */
	lockdep_assert_held(&kvm->slots_lock);

	if (flush)
		kvm_flush_remote_tlbs(kvm);
}
EXPORT_SYMBOL_GPL(kvm_mmu_slot_largepage_remove_write_access);

void kvm_mmu_slot_set_dirty(struct kvm *kvm,
			    struct kvm_memory_slot *memslot)
{
4708
	bool flush;
4709 4710

	spin_lock(&kvm->mmu_lock);
4711
	flush = slot_handle_all_level(kvm, memslot, __rmap_set_dirty, false);
4712 4713 4714 4715 4716 4717 4718 4719 4720 4721
	spin_unlock(&kvm->mmu_lock);

	lockdep_assert_held(&kvm->slots_lock);

	/* see kvm_mmu_slot_leaf_clear_dirty */
	if (flush)
		kvm_flush_remote_tlbs(kvm);
}
EXPORT_SYMBOL_GPL(kvm_mmu_slot_set_dirty);

X
Xiao Guangrong 已提交
4722
#define BATCH_ZAP_PAGES	10
4723 4724 4725
static void kvm_zap_obsolete_pages(struct kvm *kvm)
{
	struct kvm_mmu_page *sp, *node;
X
Xiao Guangrong 已提交
4726
	int batch = 0;
4727 4728 4729 4730

restart:
	list_for_each_entry_safe_reverse(sp, node,
	      &kvm->arch.active_mmu_pages, link) {
X
Xiao Guangrong 已提交
4731 4732
		int ret;

4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747
		/*
		 * No obsolete page exists before new created page since
		 * active_mmu_pages is the FIFO list.
		 */
		if (!is_obsolete_sp(kvm, sp))
			break;

		/*
		 * Since we are reversely walking the list and the invalid
		 * list will be moved to the head, skip the invalid page
		 * can help us to avoid the infinity list walking.
		 */
		if (sp->role.invalid)
			continue;

4748 4749 4750 4751
		/*
		 * Need not flush tlb since we only zap the sp with invalid
		 * generation number.
		 */
X
Xiao Guangrong 已提交
4752
		if (batch >= BATCH_ZAP_PAGES &&
4753
		      cond_resched_lock(&kvm->mmu_lock)) {
X
Xiao Guangrong 已提交
4754
			batch = 0;
4755 4756 4757
			goto restart;
		}

4758 4759
		ret = kvm_mmu_prepare_zap_page(kvm, sp,
				&kvm->arch.zapped_obsolete_pages);
X
Xiao Guangrong 已提交
4760 4761 4762
		batch += ret;

		if (ret)
4763 4764 4765
			goto restart;
	}

4766 4767 4768 4769
	/*
	 * Should flush tlb before free page tables since lockless-walking
	 * may use the pages.
	 */
4770
	kvm_mmu_commit_zap_page(kvm, &kvm->arch.zapped_obsolete_pages);
4771 4772 4773 4774 4775 4776 4777 4778 4779 4780 4781 4782 4783 4784
}

/*
 * Fast invalidate all shadow pages and use lock-break technique
 * to zap obsolete pages.
 *
 * It's required when memslot is being deleted or VM is being
 * destroyed, in these cases, we should ensure that KVM MMU does
 * not use any resource of the being-deleted slot or all slots
 * after calling the function.
 */
void kvm_mmu_invalidate_zap_all_pages(struct kvm *kvm)
{
	spin_lock(&kvm->mmu_lock);
4785
	trace_kvm_mmu_invalidate_zap_all_pages(kvm);
4786 4787
	kvm->arch.mmu_valid_gen++;

4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798
	/*
	 * Notify all vcpus to reload its shadow page table
	 * and flush TLB. Then all vcpus will switch to new
	 * shadow page table with the new mmu_valid_gen.
	 *
	 * Note: we should do this under the protection of
	 * mmu-lock, otherwise, vcpu would purge shadow page
	 * but miss tlb flush.
	 */
	kvm_reload_remote_mmus(kvm);

4799 4800 4801 4802
	kvm_zap_obsolete_pages(kvm);
	spin_unlock(&kvm->mmu_lock);
}

4803 4804 4805 4806 4807
static bool kvm_has_zapped_obsolete_pages(struct kvm *kvm)
{
	return unlikely(!list_empty_careful(&kvm->arch.zapped_obsolete_pages));
}

4808
void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, struct kvm_memslots *slots)
4809 4810 4811 4812 4813
{
	/*
	 * The very rare case: if the generation-number is round,
	 * zap all shadow pages.
	 */
4814
	if (unlikely((slots->generation & MMIO_GEN_MASK) == 0)) {
4815
		printk_ratelimited(KERN_DEBUG "kvm: zapping shadow pages for mmio generation wraparound\n");
4816
		kvm_mmu_invalidate_zap_all_pages(kvm);
4817
	}
4818 4819
}

4820 4821
static unsigned long
mmu_shrink_scan(struct shrinker *shrink, struct shrink_control *sc)
4822 4823
{
	struct kvm *kvm;
4824
	int nr_to_scan = sc->nr_to_scan;
4825
	unsigned long freed = 0;
4826

4827
	spin_lock(&kvm_lock);
4828 4829

	list_for_each_entry(kvm, &vm_list, vm_list) {
4830
		int idx;
4831
		LIST_HEAD(invalid_list);
4832

4833 4834 4835 4836 4837 4838 4839 4840
		/*
		 * Never scan more than sc->nr_to_scan VM instances.
		 * Will not hit this condition practically since we do not try
		 * to shrink more than one VM and it is very unlikely to see
		 * !n_used_mmu_pages so many times.
		 */
		if (!nr_to_scan--)
			break;
4841 4842 4843 4844 4845 4846
		/*
		 * n_used_mmu_pages is accessed without holding kvm->mmu_lock
		 * here. We may skip a VM instance errorneosly, but we do not
		 * want to shrink a VM that only started to populate its MMU
		 * anyway.
		 */
4847 4848
		if (!kvm->arch.n_used_mmu_pages &&
		      !kvm_has_zapped_obsolete_pages(kvm))
4849 4850
			continue;

4851
		idx = srcu_read_lock(&kvm->srcu);
4852 4853
		spin_lock(&kvm->mmu_lock);

4854 4855 4856 4857 4858 4859
		if (kvm_has_zapped_obsolete_pages(kvm)) {
			kvm_mmu_commit_zap_page(kvm,
			      &kvm->arch.zapped_obsolete_pages);
			goto unlock;
		}

4860 4861
		if (prepare_zap_oldest_mmu_page(kvm, &invalid_list))
			freed++;
4862
		kvm_mmu_commit_zap_page(kvm, &invalid_list);
4863

4864
unlock:
4865
		spin_unlock(&kvm->mmu_lock);
4866
		srcu_read_unlock(&kvm->srcu, idx);
4867

4868 4869 4870 4871 4872
		/*
		 * unfair on small ones
		 * per-vm shrinkers cry out
		 * sadness comes quickly
		 */
4873 4874
		list_move_tail(&kvm->vm_list, &vm_list);
		break;
4875 4876
	}

4877
	spin_unlock(&kvm_lock);
4878 4879 4880 4881 4882 4883
	return freed;
}

static unsigned long
mmu_shrink_count(struct shrinker *shrink, struct shrink_control *sc)
{
4884
	return percpu_counter_read_positive(&kvm_total_used_mmu_pages);
4885 4886 4887
}

static struct shrinker mmu_shrinker = {
4888 4889
	.count_objects = mmu_shrink_count,
	.scan_objects = mmu_shrink_scan,
4890 4891 4892
	.seeks = DEFAULT_SEEKS * 10,
};

I
Ingo Molnar 已提交
4893
static void mmu_destroy_caches(void)
4894
{
4895 4896
	if (pte_list_desc_cache)
		kmem_cache_destroy(pte_list_desc_cache);
4897 4898
	if (mmu_page_header_cache)
		kmem_cache_destroy(mmu_page_header_cache);
4899 4900 4901 4902
}

int kvm_mmu_module_init(void)
{
4903 4904
	pte_list_desc_cache = kmem_cache_create("pte_list_desc",
					    sizeof(struct pte_list_desc),
4905
					    0, 0, NULL);
4906
	if (!pte_list_desc_cache)
4907 4908
		goto nomem;

4909 4910
	mmu_page_header_cache = kmem_cache_create("kvm_mmu_page_header",
						  sizeof(struct kvm_mmu_page),
4911
						  0, 0, NULL);
4912 4913 4914
	if (!mmu_page_header_cache)
		goto nomem;

4915
	if (percpu_counter_init(&kvm_total_used_mmu_pages, 0, GFP_KERNEL))
4916 4917
		goto nomem;

4918 4919
	register_shrinker(&mmu_shrinker);

4920 4921 4922
	return 0;

nomem:
4923
	mmu_destroy_caches();
4924 4925 4926
	return -ENOMEM;
}

4927 4928 4929 4930 4931 4932 4933
/*
 * Caculate mmu pages needed for kvm.
 */
unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm)
{
	unsigned int nr_mmu_pages;
	unsigned int  nr_pages = 0;
4934
	struct kvm_memslots *slots;
4935
	struct kvm_memory_slot *memslot;
4936
	int i;
4937

4938 4939
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
		slots = __kvm_memslots(kvm, i);
4940

4941 4942 4943
		kvm_for_each_memslot(memslot, slots)
			nr_pages += memslot->npages;
	}
4944 4945 4946

	nr_mmu_pages = nr_pages * KVM_PERMILLE_MMU_PAGES / 1000;
	nr_mmu_pages = max(nr_mmu_pages,
4947
			   (unsigned int) KVM_MIN_ALLOC_MMU_PAGES);
4948 4949 4950 4951

	return nr_mmu_pages;
}

4952 4953
void kvm_mmu_destroy(struct kvm_vcpu *vcpu)
{
4954
	kvm_mmu_unload(vcpu);
4955 4956
	free_mmu_pages(vcpu);
	mmu_free_memory_caches(vcpu);
4957 4958 4959 4960 4961 4962 4963
}

void kvm_mmu_module_exit(void)
{
	mmu_destroy_caches();
	percpu_counter_destroy(&kvm_total_used_mmu_pages);
	unregister_shrinker(&mmu_shrinker);
4964 4965
	mmu_audit_disable();
}