mmu.c 123.5 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,
D
Dan Williams 已提交
262
			  kvm_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;
}

D
Dan Williams 已提交
323
static kvm_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
/*
 * 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)
{
D
Dan Williams 已提交
585
	kvm_pfn_t pfn;
586 587 588
	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 1011
static struct kvm_rmap_head *__gfn_to_rmap(gfn_t gfn, int level,
					   struct kvm_memory_slot *slot)
1012
{
1013
	unsigned long idx;
1014

1015
	idx = gfn_to_index(gfn, slot->base_gfn, level);
1016
	return &slot->arch.rmap[level - PT_PAGE_TABLE_LEVEL][idx];
1017 1018
}

1019 1020
static struct kvm_rmap_head *gfn_to_rmap(struct kvm *kvm, gfn_t gfn,
					 struct kvm_mmu_page *sp)
1021
{
1022
	struct kvm_memslots *slots;
1023 1024
	struct kvm_memory_slot *slot;

1025 1026
	slots = kvm_memslots_for_spte_role(kvm, sp->role);
	slot = __gfn_to_memslot(slots, gfn);
1027
	return __gfn_to_rmap(gfn, sp->role.level, slot);
1028 1029
}

1030 1031 1032 1033 1034 1035 1036 1037
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);
}

1038 1039 1040
static int rmap_add(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn)
{
	struct kvm_mmu_page *sp;
1041
	struct kvm_rmap_head *rmap_head;
1042 1043 1044

	sp = page_header(__pa(spte));
	kvm_mmu_page_set_gfn(sp, spte - sp->spt, gfn);
1045 1046
	rmap_head = gfn_to_rmap(vcpu->kvm, gfn, sp);
	return pte_list_add(vcpu, spte, rmap_head);
1047 1048 1049 1050 1051 1052
}

static void rmap_remove(struct kvm *kvm, u64 *spte)
{
	struct kvm_mmu_page *sp;
	gfn_t gfn;
1053
	struct kvm_rmap_head *rmap_head;
1054 1055 1056

	sp = page_header(__pa(spte));
	gfn = kvm_mmu_page_get_gfn(sp, spte - sp->spt);
1057 1058
	rmap_head = gfn_to_rmap(kvm, gfn, sp);
	pte_list_remove(spte, rmap_head);
1059 1060
}

1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077
/*
 * 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.
 */
1078 1079
static u64 *rmap_get_first(struct kvm_rmap_head *rmap_head,
			   struct rmap_iterator *iter)
1080
{
1081 1082
	u64 *sptep;

1083
	if (!rmap_head->val)
1084 1085
		return NULL;

1086
	if (!(rmap_head->val & 1)) {
1087
		iter->desc = NULL;
1088 1089
		sptep = (u64 *)rmap_head->val;
		goto out;
1090 1091
	}

1092
	iter->desc = (struct pte_list_desc *)(rmap_head->val & ~1ul);
1093
	iter->pos = 0;
1094 1095 1096 1097
	sptep = iter->desc->sptes[iter->pos];
out:
	BUG_ON(!is_shadow_present_pte(*sptep));
	return sptep;
1098 1099 1100 1101 1102 1103 1104 1105 1106
}

/*
 * 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)
{
1107 1108
	u64 *sptep;

1109 1110 1111 1112 1113
	if (iter->desc) {
		if (iter->pos < PTE_LIST_EXT - 1) {
			++iter->pos;
			sptep = iter->desc->sptes[iter->pos];
			if (sptep)
1114
				goto out;
1115 1116 1117 1118 1119 1120 1121
		}

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

		if (iter->desc) {
			iter->pos = 0;
			/* desc->sptes[0] cannot be NULL */
1122 1123
			sptep = iter->desc->sptes[iter->pos];
			goto out;
1124 1125 1126 1127
		}
	}

	return NULL;
1128 1129 1130
out:
	BUG_ON(!is_shadow_present_pte(*sptep));
	return sptep;
1131 1132
}

1133 1134
#define for_each_rmap_spte(_rmap_head_, _iter_, _spte_)			\
	for (_spte_ = rmap_get_first(_rmap_head_, _iter_);		\
1135
	     _spte_; _spte_ = rmap_get_next(_iter_))
1136

1137
static void drop_spte(struct kvm *kvm, u64 *sptep)
1138
{
1139
	if (mmu_spte_clear_track_bits(sptep))
1140
		rmap_remove(kvm, sptep);
A
Avi Kivity 已提交
1141 1142
}

1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163

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);
}

/*
1164
 * Write-protect on the specified @sptep, @pt_protect indicates whether
1165
 * spte write-protection is caused by protecting shadow page table.
1166
 *
T
Tiejun Chen 已提交
1167
 * Note: write protection is difference between dirty logging and spte
1168 1169 1170 1171 1172
 * 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.
1173
 *
1174
 * Return true if tlb need be flushed.
1175
 */
1176
static bool spte_write_protect(struct kvm *kvm, u64 *sptep, bool pt_protect)
1177 1178 1179
{
	u64 spte = *sptep;

1180 1181
	if (!is_writable_pte(spte) &&
	      !(pt_protect && spte_is_locklessly_modifiable(spte)))
1182 1183 1184 1185
		return false;

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

1186 1187
	if (pt_protect)
		spte &= ~SPTE_MMU_WRITEABLE;
1188
	spte = spte & ~PT_WRITABLE_MASK;
1189

1190
	return mmu_spte_update(sptep, spte);
1191 1192
}

1193 1194
static bool __rmap_write_protect(struct kvm *kvm,
				 struct kvm_rmap_head *rmap_head,
1195
				 bool pt_protect)
1196
{
1197 1198
	u64 *sptep;
	struct rmap_iterator iter;
1199
	bool flush = false;
1200

1201
	for_each_rmap_spte(rmap_head, &iter, sptep)
1202
		flush |= spte_write_protect(kvm, sptep, pt_protect);
1203

1204
	return flush;
1205 1206
}

1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217
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);
}

1218
static bool __rmap_clear_dirty(struct kvm *kvm, struct kvm_rmap_head *rmap_head)
1219 1220 1221 1222 1223
{
	u64 *sptep;
	struct rmap_iterator iter;
	bool flush = false;

1224
	for_each_rmap_spte(rmap_head, &iter, sptep)
1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240
		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);
}

1241
static bool __rmap_set_dirty(struct kvm *kvm, struct kvm_rmap_head *rmap_head)
1242 1243 1244 1245 1246
{
	u64 *sptep;
	struct rmap_iterator iter;
	bool flush = false;

1247
	for_each_rmap_spte(rmap_head, &iter, sptep)
1248 1249 1250 1251 1252
		flush |= spte_set_dirty(kvm, sptep);

	return flush;
}

1253
/**
1254
 * kvm_mmu_write_protect_pt_masked - write protect selected PT level pages
1255 1256 1257 1258 1259 1260 1261 1262
 * @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.
 */
1263
static void kvm_mmu_write_protect_pt_masked(struct kvm *kvm,
1264 1265
				     struct kvm_memory_slot *slot,
				     gfn_t gfn_offset, unsigned long mask)
1266
{
1267
	struct kvm_rmap_head *rmap_head;
1268

1269
	while (mask) {
1270 1271 1272
		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
1273

1274 1275 1276
		/* clear the first set bit */
		mask &= mask - 1;
	}
1277 1278
}

1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291
/**
 * 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)
{
1292
	struct kvm_rmap_head *rmap_head;
1293 1294

	while (mask) {
1295 1296 1297
		rmap_head = __gfn_to_rmap(slot->base_gfn + gfn_offset + __ffs(mask),
					  PT_PAGE_TABLE_LEVEL, slot);
		__rmap_clear_dirty(kvm, rmap_head);
1298 1299 1300 1301 1302 1303 1304

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

1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318
/**
 * 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)
{
1319 1320 1321 1322 1323
	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);
1324 1325
}

1326
static bool rmap_write_protect(struct kvm_vcpu *vcpu, u64 gfn)
1327 1328
{
	struct kvm_memory_slot *slot;
1329
	struct kvm_rmap_head *rmap_head;
1330
	int i;
1331
	bool write_protected = false;
1332

1333
	slot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);
1334

1335
	for (i = PT_PAGE_TABLE_LEVEL; i <= PT_MAX_HUGEPAGE_LEVEL; ++i) {
1336 1337
		rmap_head = __gfn_to_rmap(gfn, i, slot);
		write_protected |= __rmap_write_protect(vcpu->kvm, rmap_head, true);
1338 1339 1340
	}

	return write_protected;
1341 1342
}

1343
static bool kvm_zap_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head)
1344
{
1345 1346
	u64 *sptep;
	struct rmap_iterator iter;
1347
	bool flush = false;
1348

1349
	while ((sptep = rmap_get_first(rmap_head, &iter))) {
1350
		rmap_printk("%s: spte %p %llx.\n", __func__, sptep, *sptep);
1351 1352

		drop_spte(kvm, sptep);
1353
		flush = true;
1354
	}
1355

1356 1357 1358
	return flush;
}

1359
static int kvm_unmap_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
1360 1361 1362
			   struct kvm_memory_slot *slot, gfn_t gfn, int level,
			   unsigned long data)
{
1363
	return kvm_zap_rmapp(kvm, rmap_head);
1364 1365
}

1366
static int kvm_set_pte_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
1367 1368
			     struct kvm_memory_slot *slot, gfn_t gfn, int level,
			     unsigned long data)
1369
{
1370 1371
	u64 *sptep;
	struct rmap_iterator iter;
1372
	int need_flush = 0;
1373
	u64 new_spte;
1374
	pte_t *ptep = (pte_t *)data;
D
Dan Williams 已提交
1375
	kvm_pfn_t new_pfn;
1376 1377 1378

	WARN_ON(pte_huge(*ptep));
	new_pfn = pte_pfn(*ptep);
1379

1380
restart:
1381
	for_each_rmap_spte(rmap_head, &iter, sptep) {
1382 1383
		rmap_printk("kvm_set_pte_rmapp: spte %p %llx gfn %llx (%d)\n",
			     sptep, *sptep, gfn, level);
1384

1385
		need_flush = 1;
1386

1387
		if (pte_write(*ptep)) {
1388
			drop_spte(kvm, sptep);
1389
			goto restart;
1390
		} else {
1391
			new_spte = *sptep & ~PT64_BASE_ADDR_MASK;
1392 1393 1394 1395
			new_spte |= (u64)new_pfn << PAGE_SHIFT;

			new_spte &= ~PT_WRITABLE_MASK;
			new_spte &= ~SPTE_HOST_WRITEABLE;
1396
			new_spte &= ~shadow_accessed_mask;
1397 1398 1399

			mmu_spte_clear_track_bits(sptep);
			mmu_spte_set(sptep, new_spte);
1400 1401
		}
	}
1402

1403 1404 1405 1406 1407 1408
	if (need_flush)
		kvm_flush_remote_tlbs(kvm);

	return 0;
}

1409 1410 1411 1412 1413 1414 1415 1416 1417 1418
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;
1419
	struct kvm_rmap_head *rmap;
1420 1421 1422
	int level;

	/* private field. */
1423
	struct kvm_rmap_head *end_rmap;
1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 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
};

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_))

1477 1478 1479 1480 1481
static int kvm_handle_hva_range(struct kvm *kvm,
				unsigned long start,
				unsigned long end,
				unsigned long data,
				int (*handler)(struct kvm *kvm,
1482
					       struct kvm_rmap_head *rmap_head,
1483
					       struct kvm_memory_slot *slot,
1484 1485
					       gfn_t gfn,
					       int level,
1486
					       unsigned long data))
1487
{
1488
	struct kvm_memslots *slots;
1489
	struct kvm_memory_slot *memslot;
1490 1491
	struct slot_rmap_walk_iterator iterator;
	int ret = 0;
1492
	int i;
1493

1494 1495 1496 1497 1498
	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;
1499

1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518
			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);
		}
1519 1520
	}

1521
	return ret;
1522 1523
}

1524 1525
static int kvm_handle_hva(struct kvm *kvm, unsigned long hva,
			  unsigned long data,
1526 1527
			  int (*handler)(struct kvm *kvm,
					 struct kvm_rmap_head *rmap_head,
1528
					 struct kvm_memory_slot *slot,
1529
					 gfn_t gfn, int level,
1530 1531 1532
					 unsigned long data))
{
	return kvm_handle_hva_range(kvm, hva, hva + 1, data, handler);
1533 1534 1535 1536
}

int kvm_unmap_hva(struct kvm *kvm, unsigned long hva)
{
1537 1538 1539
	return kvm_handle_hva(kvm, hva, 0, kvm_unmap_rmapp);
}

1540 1541 1542 1543 1544
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);
}

1545 1546
void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte)
{
F
Frederik Deweerdt 已提交
1547
	kvm_handle_hva(kvm, hva, (unsigned long)&pte, kvm_set_pte_rmapp);
1548 1549
}

1550
static int kvm_age_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
1551 1552
			 struct kvm_memory_slot *slot, gfn_t gfn, int level,
			 unsigned long data)
1553
{
1554
	u64 *sptep;
1555
	struct rmap_iterator uninitialized_var(iter);
1556 1557
	int young = 0;

A
Andres Lagar-Cavilla 已提交
1558
	BUG_ON(!shadow_accessed_mask);
1559

1560
	for_each_rmap_spte(rmap_head, &iter, sptep) {
1561
		if (*sptep & shadow_accessed_mask) {
1562
			young = 1;
1563 1564
			clear_bit((ffs(shadow_accessed_mask) - 1),
				 (unsigned long *)sptep);
1565
		}
1566
	}
1567

1568
	trace_kvm_age_page(gfn, level, slot, young);
1569 1570 1571
	return young;
}

1572
static int kvm_test_age_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
1573 1574
			      struct kvm_memory_slot *slot, gfn_t gfn,
			      int level, unsigned long data)
A
Andrea Arcangeli 已提交
1575
{
1576 1577
	u64 *sptep;
	struct rmap_iterator iter;
A
Andrea Arcangeli 已提交
1578 1579 1580 1581 1582 1583 1584 1585 1586 1587
	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;

1588
	for_each_rmap_spte(rmap_head, &iter, sptep) {
1589
		if (*sptep & shadow_accessed_mask) {
A
Andrea Arcangeli 已提交
1590 1591 1592
			young = 1;
			break;
		}
1593
	}
A
Andrea Arcangeli 已提交
1594 1595 1596 1597
out:
	return young;
}

1598 1599
#define RMAP_RECYCLE_THRESHOLD 1000

1600
static void rmap_recycle(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn)
1601
{
1602
	struct kvm_rmap_head *rmap_head;
1603 1604 1605
	struct kvm_mmu_page *sp;

	sp = page_header(__pa(spte));
1606

1607
	rmap_head = gfn_to_rmap(vcpu->kvm, gfn, sp);
1608

1609
	kvm_unmap_rmapp(vcpu->kvm, rmap_head, NULL, gfn, sp->role.level, 0);
1610 1611 1612
	kvm_flush_remote_tlbs(vcpu->kvm);
}

A
Andres Lagar-Cavilla 已提交
1613
int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
1614
{
A
Andres Lagar-Cavilla 已提交
1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635
	/*
	 * 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);
1636 1637
}

A
Andrea Arcangeli 已提交
1638 1639 1640 1641 1642
int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
{
	return kvm_handle_hva(kvm, hva, 0, kvm_test_age_rmapp);
}

1643
#ifdef MMU_DEBUG
1644
static int is_empty_shadow_page(u64 *spt)
A
Avi Kivity 已提交
1645
{
1646 1647 1648
	u64 *pos;
	u64 *end;

1649
	for (pos = spt, end = pos + PAGE_SIZE / sizeof(u64); pos != end; pos++)
1650
		if (is_shadow_present_pte(*pos)) {
1651
			printk(KERN_ERR "%s: %p %llx\n", __func__,
1652
			       pos, *pos);
A
Avi Kivity 已提交
1653
			return 0;
1654
		}
A
Avi Kivity 已提交
1655 1656
	return 1;
}
1657
#endif
A
Avi Kivity 已提交
1658

1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670
/*
 * 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);
}

1671
static void kvm_mmu_free_page(struct kvm_mmu_page *sp)
1672
{
1673
	MMU_WARN_ON(!is_empty_shadow_page(sp->spt));
1674
	hlist_del(&sp->hash_link);
1675 1676
	list_del(&sp->link);
	free_page((unsigned long)sp->spt);
1677 1678
	if (!sp->role.direct)
		free_page((unsigned long)sp->gfns);
1679
	kmem_cache_free(mmu_page_header_cache, sp);
1680 1681
}

1682 1683
static unsigned kvm_page_table_hashfn(gfn_t gfn)
{
1684
	return gfn & ((1 << KVM_MMU_HASH_SHIFT) - 1);
1685 1686
}

1687
static void mmu_page_add_parent_pte(struct kvm_vcpu *vcpu,
1688
				    struct kvm_mmu_page *sp, u64 *parent_pte)
1689 1690 1691 1692
{
	if (!parent_pte)
		return;

1693
	pte_list_add(vcpu, parent_pte, &sp->parent_ptes);
1694 1695
}

1696
static void mmu_page_remove_parent_pte(struct kvm_mmu_page *sp,
1697 1698
				       u64 *parent_pte)
{
1699
	pte_list_remove(parent_pte, &sp->parent_ptes);
1700 1701
}

1702 1703 1704 1705
static void drop_parent_pte(struct kvm_mmu_page *sp,
			    u64 *parent_pte)
{
	mmu_page_remove_parent_pte(sp, parent_pte);
1706
	mmu_spte_clear_no_track(parent_pte);
1707 1708
}

1709
static struct kvm_mmu_page *kvm_mmu_alloc_page(struct kvm_vcpu *vcpu, int direct)
M
Marcelo Tosatti 已提交
1710
{
1711
	struct kvm_mmu_page *sp;
1712

1713 1714
	sp = mmu_memory_cache_alloc(&vcpu->arch.mmu_page_header_cache);
	sp->spt = mmu_memory_cache_alloc(&vcpu->arch.mmu_page_cache);
1715
	if (!direct)
1716
		sp->gfns = mmu_memory_cache_alloc(&vcpu->arch.mmu_page_cache);
1717
	set_page_private(virt_to_page(sp->spt), (unsigned long)sp);
1718 1719 1720 1721 1722 1723

	/*
	 * 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().
	 */
1724 1725 1726
	list_add(&sp->link, &vcpu->kvm->arch.active_mmu_pages);
	kvm_mod_used_mmu_pages(vcpu->kvm, +1);
	return sp;
M
Marcelo Tosatti 已提交
1727 1728
}

1729
static void mark_unsync(u64 *spte);
1730
static void kvm_mmu_mark_parents_unsync(struct kvm_mmu_page *sp)
1731
{
1732 1733 1734 1735 1736 1737
	u64 *sptep;
	struct rmap_iterator iter;

	for_each_rmap_spte(&sp->parent_ptes, &iter, sptep) {
		mark_unsync(sptep);
	}
1738 1739
}

1740
static void mark_unsync(u64 *spte)
1741
{
1742
	struct kvm_mmu_page *sp;
1743
	unsigned int index;
1744

1745
	sp = page_header(__pa(spte));
1746 1747
	index = spte - sp->spt;
	if (__test_and_set_bit(index, sp->unsync_child_bitmap))
1748
		return;
1749
	if (sp->unsync_children++)
1750
		return;
1751
	kvm_mmu_mark_parents_unsync(sp);
1752 1753
}

1754
static int nonpaging_sync_page(struct kvm_vcpu *vcpu,
1755
			       struct kvm_mmu_page *sp)
1756 1757 1758 1759
{
	return 1;
}

M
Marcelo Tosatti 已提交
1760 1761 1762 1763
static void nonpaging_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
{
}

1764 1765
static void nonpaging_update_pte(struct kvm_vcpu *vcpu,
				 struct kvm_mmu_page *sp, u64 *spte,
1766
				 const void *pte)
1767 1768 1769 1770
{
	WARN_ON(1);
}

1771 1772 1773 1774 1775 1776 1777 1778 1779 1780
#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;
};

1781 1782
static int mmu_pages_add(struct kvm_mmu_pages *pvec, struct kvm_mmu_page *sp,
			 int idx)
1783
{
1784
	int i;
1785

1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796
	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);
}

1797 1798 1799 1800 1801 1802 1803
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);
}

1804 1805 1806 1807
static int __mmu_unsync_walk(struct kvm_mmu_page *sp,
			   struct kvm_mmu_pages *pvec)
{
	int i, ret, nr_unsync_leaf = 0;
1808

1809
	for_each_set_bit(i, sp->unsync_child_bitmap, 512) {
1810
		struct kvm_mmu_page *child;
1811 1812
		u64 ent = sp->spt[i];

1813 1814 1815 1816
		if (!is_shadow_present_pte(ent) || is_large_pte(ent)) {
			clear_unsync_child_bit(sp, i);
			continue;
		}
1817 1818 1819 1820 1821 1822 1823 1824

		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);
1825 1826 1827 1828
			if (!ret) {
				clear_unsync_child_bit(sp, i);
				continue;
			} else if (ret > 0) {
1829
				nr_unsync_leaf += ret;
1830
			} else
1831 1832 1833 1834 1835 1836
				return ret;
		} else if (child->unsync) {
			nr_unsync_leaf++;
			if (mmu_pages_add(pvec, child, i))
				return -ENOSPC;
		} else
1837
			clear_unsync_child_bit(sp, i);
1838 1839
	}

1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850
	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);
1851 1852 1853 1854 1855
}

static void kvm_unlink_unsync_page(struct kvm *kvm, struct kvm_mmu_page *sp)
{
	WARN_ON(!sp->unsync);
1856
	trace_kvm_mmu_sync_page(sp);
1857 1858 1859 1860
	sp->unsync = 0;
	--kvm->stat.mmu_unsync;
}

1861 1862 1863 1864
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);
1865

1866 1867 1868 1869 1870 1871 1872 1873 1874 1875
/*
 * 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.
 */
1876 1877 1878 1879 1880 1881 1882 1883
#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
1884

1885
/* @sp->gfn should be write-protected at the call site */
1886
static int __kvm_sync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
1887
			   struct list_head *invalid_list, bool clear_unsync)
1888
{
1889
	if (sp->role.cr4_pae != !!is_pae(vcpu)) {
1890
		kvm_mmu_prepare_zap_page(vcpu->kvm, sp, invalid_list);
1891 1892 1893
		return 1;
	}

1894
	if (clear_unsync)
1895 1896
		kvm_unlink_unsync_page(vcpu->kvm, sp);

1897
	if (vcpu->arch.mmu.sync_page(vcpu, sp)) {
1898
		kvm_mmu_prepare_zap_page(vcpu->kvm, sp, invalid_list);
1899 1900 1901
		return 1;
	}

1902
	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1903 1904 1905
	return 0;
}

1906 1907 1908
static int kvm_sync_page_transient(struct kvm_vcpu *vcpu,
				   struct kvm_mmu_page *sp)
{
1909
	LIST_HEAD(invalid_list);
1910 1911
	int ret;

1912
	ret = __kvm_sync_page(vcpu, sp, &invalid_list, false);
1913
	if (ret)
1914 1915
		kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);

1916 1917 1918
	return ret;
}

1919 1920 1921 1922 1923 1924 1925
#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

1926 1927
static int kvm_sync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
			 struct list_head *invalid_list)
1928
{
1929
	return __kvm_sync_page(vcpu, sp, invalid_list, true);
1930 1931
}

1932 1933 1934 1935
/* @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;
1936
	LIST_HEAD(invalid_list);
1937 1938
	bool flush = false;

1939
	for_each_gfn_indirect_valid_sp(vcpu->kvm, s, gfn) {
1940
		if (!s->unsync)
1941 1942 1943
			continue;

		WARN_ON(s->role.level != PT_PAGE_TABLE_LEVEL);
1944
		kvm_unlink_unsync_page(vcpu->kvm, s);
1945
		if ((s->role.cr4_pae != !!is_pae(vcpu)) ||
1946
			(vcpu->arch.mmu.sync_page(vcpu, s))) {
1947
			kvm_mmu_prepare_zap_page(vcpu->kvm, s, &invalid_list);
1948 1949 1950 1951 1952
			continue;
		}
		flush = true;
	}

1953
	kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
1954
	if (flush)
1955
		kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1956 1957
}

1958 1959 1960
struct mmu_page_path {
	struct kvm_mmu_page *parent[PT64_ROOT_LEVEL-1];
	unsigned int idx[PT64_ROOT_LEVEL-1];
1961 1962
};

1963 1964 1965 1966 1967 1968
#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))

1969 1970 1971
static int mmu_pages_next(struct kvm_mmu_pages *pvec,
			  struct mmu_page_path *parents,
			  int i)
1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989
{
	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;
}

1990
static void mmu_pages_clear_parents(struct mmu_page_path *parents)
1991
{
1992 1993 1994 1995 1996
	struct kvm_mmu_page *sp;
	unsigned int level = 0;

	do {
		unsigned int idx = parents->idx[level];
1997

1998 1999 2000 2001
		sp = parents->parent[level];
		if (!sp)
			return;

2002
		clear_unsync_child_bit(sp, idx);
2003 2004
		level++;
	} while (level < PT64_ROOT_LEVEL-1 && !sp->unsync_children);
2005 2006
}

2007 2008 2009
static void kvm_mmu_pages_init(struct kvm_mmu_page *parent,
			       struct mmu_page_path *parents,
			       struct kvm_mmu_pages *pvec)
2010
{
2011 2012 2013
	parents->parent[parent->role.level-1] = NULL;
	pvec->nr = 0;
}
2014

2015 2016 2017 2018 2019 2020 2021
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;
2022
	LIST_HEAD(invalid_list);
2023 2024 2025

	kvm_mmu_pages_init(parent, &parents, &pages);
	while (mmu_unsync_walk(parent, &pages)) {
2026
		bool protected = false;
2027 2028

		for_each_sp(pages, sp, parents, i)
2029
			protected |= rmap_write_protect(vcpu, sp->gfn);
2030 2031 2032 2033

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

2034
		for_each_sp(pages, sp, parents, i) {
2035
			kvm_sync_page(vcpu, sp, &invalid_list);
2036 2037
			mmu_pages_clear_parents(&parents);
		}
2038
		kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
2039
		cond_resched_lock(&vcpu->kvm->mmu_lock);
2040 2041
		kvm_mmu_pages_init(parent, &parents, &pages);
	}
2042 2043
}

2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055
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);
}

2056 2057 2058 2059 2060
static bool is_obsolete_sp(struct kvm *kvm, struct kvm_mmu_page *sp)
{
	return unlikely(sp->mmu_valid_gen != kvm->arch.mmu_valid_gen);
}

2061 2062 2063 2064
static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
					     gfn_t gfn,
					     gva_t gaddr,
					     unsigned level,
2065
					     int direct,
2066
					     unsigned access)
2067 2068 2069
{
	union kvm_mmu_page_role role;
	unsigned quadrant;
2070 2071
	struct kvm_mmu_page *sp;
	bool need_sync = false;
2072

2073
	role = vcpu->arch.mmu.base_role;
2074
	role.level = level;
2075
	role.direct = direct;
2076
	if (role.direct)
2077
		role.cr4_pae = 0;
2078
	role.access = access;
2079 2080
	if (!vcpu->arch.mmu.direct_map
	    && vcpu->arch.mmu.root_level <= PT32_ROOT_LEVEL) {
2081 2082 2083 2084
		quadrant = gaddr >> (PAGE_SHIFT + (PT64_PT_BITS * level));
		quadrant &= (1 << ((PT32_PT_BITS - PT64_PT_BITS) * level)) - 1;
		role.quadrant = quadrant;
	}
2085
	for_each_gfn_sp(vcpu->kvm, sp, gfn) {
2086 2087 2088
		if (is_obsolete_sp(vcpu->kvm, sp))
			continue;

2089 2090
		if (!need_sync && sp->unsync)
			need_sync = true;
2091

2092 2093
		if (sp->role.word != role.word)
			continue;
2094

2095 2096
		if (sp->unsync && kvm_sync_page_transient(vcpu, sp))
			break;
2097

2098
		if (sp->unsync_children)
2099
			kvm_make_request(KVM_REQ_MMU_SYNC, vcpu);
2100

2101
		__clear_sp_write_flooding_count(sp);
2102 2103 2104
		trace_kvm_mmu_get_page(sp, false);
		return sp;
	}
2105

A
Avi Kivity 已提交
2106
	++vcpu->kvm->stat.mmu_cache_miss;
2107 2108 2109

	sp = kvm_mmu_alloc_page(vcpu, direct);

2110 2111
	sp->gfn = gfn;
	sp->role = role;
2112 2113
	hlist_add_head(&sp->hash_link,
		&vcpu->kvm->arch.mmu_page_hash[kvm_page_table_hashfn(gfn)]);
2114
	if (!direct) {
2115
		if (rmap_write_protect(vcpu, gfn))
2116
			kvm_flush_remote_tlbs(vcpu->kvm);
2117 2118 2119
		if (level > PT_PAGE_TABLE_LEVEL && need_sync)
			kvm_sync_pages(vcpu, gfn);

2120
		account_shadowed(vcpu->kvm, sp);
2121
	}
2122
	sp->mmu_valid_gen = vcpu->kvm->arch.mmu_valid_gen;
2123
	clear_page(sp->spt);
A
Avi Kivity 已提交
2124
	trace_kvm_mmu_get_page(sp, true);
2125
	return sp;
2126 2127
}

2128 2129 2130 2131 2132 2133
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;
2134 2135 2136 2137 2138 2139

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

2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153
	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;
2154

2155 2156 2157 2158 2159
	iterator->index = SHADOW_PT_INDEX(iterator->addr, iterator->level);
	iterator->sptep	= ((u64 *)__va(iterator->shadow_addr)) + iterator->index;
	return true;
}

2160 2161
static void __shadow_walk_next(struct kvm_shadow_walk_iterator *iterator,
			       u64 spte)
2162
{
2163
	if (is_last_spte(spte, iterator->level)) {
2164 2165 2166 2167
		iterator->level = 0;
		return;
	}

2168
	iterator->shadow_addr = spte & PT64_BASE_ADDR_MASK;
2169 2170 2171
	--iterator->level;
}

2172 2173 2174 2175 2176
static void shadow_walk_next(struct kvm_shadow_walk_iterator *iterator)
{
	return __shadow_walk_next(iterator, *iterator->sptep);
}

2177 2178
static void link_shadow_page(struct kvm_vcpu *vcpu, u64 *sptep,
			     struct kvm_mmu_page *sp)
2179 2180 2181
{
	u64 spte;

2182 2183 2184
	BUILD_BUG_ON(VMX_EPT_READABLE_MASK != PT_PRESENT_MASK ||
			VMX_EPT_WRITABLE_MASK != PT_WRITABLE_MASK);

X
Xiao Guangrong 已提交
2185
	spte = __pa(sp->spt) | PT_PRESENT_MASK | PT_WRITABLE_MASK |
2186
	       shadow_user_mask | shadow_x_mask | shadow_accessed_mask;
X
Xiao Guangrong 已提交
2187

2188
	mmu_spte_set(sptep, spte);
2189 2190 2191 2192 2193

	mmu_page_add_parent_pte(vcpu, sp, sptep);

	if (sp->unsync_children || sp->unsync)
		mark_unsync(sptep);
2194 2195
}

2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212
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;

2213
		drop_parent_pte(child, sptep);
2214 2215 2216 2217
		kvm_flush_remote_tlbs(vcpu->kvm);
	}
}

X
Xiao Guangrong 已提交
2218
static bool mmu_page_zap_pte(struct kvm *kvm, struct kvm_mmu_page *sp,
2219 2220 2221 2222 2223 2224 2225
			     u64 *spte)
{
	u64 pte;
	struct kvm_mmu_page *child;

	pte = *spte;
	if (is_shadow_present_pte(pte)) {
X
Xiao Guangrong 已提交
2226
		if (is_last_spte(pte, sp->role.level)) {
2227
			drop_spte(kvm, spte);
X
Xiao Guangrong 已提交
2228 2229 2230
			if (is_large_pte(pte))
				--kvm->stat.lpages;
		} else {
2231
			child = page_header(pte & PT64_BASE_ADDR_MASK);
2232
			drop_parent_pte(child, spte);
2233
		}
X
Xiao Guangrong 已提交
2234 2235 2236 2237
		return true;
	}

	if (is_mmio_spte(pte))
2238
		mmu_spte_clear_no_track(spte);
2239

X
Xiao Guangrong 已提交
2240
	return false;
2241 2242
}

2243
static void kvm_mmu_page_unlink_children(struct kvm *kvm,
2244
					 struct kvm_mmu_page *sp)
2245
{
2246 2247
	unsigned i;

2248 2249
	for (i = 0; i < PT64_ENT_PER_PAGE; ++i)
		mmu_page_zap_pte(kvm, sp, sp->spt + i);
2250 2251
}

2252
static void kvm_mmu_unlink_parents(struct kvm *kvm, struct kvm_mmu_page *sp)
2253
{
2254 2255
	u64 *sptep;
	struct rmap_iterator iter;
2256

2257
	while ((sptep = rmap_get_first(&sp->parent_ptes, &iter)))
2258
		drop_parent_pte(sp, sptep);
2259 2260
}

2261
static int mmu_zap_unsync_children(struct kvm *kvm,
2262 2263
				   struct kvm_mmu_page *parent,
				   struct list_head *invalid_list)
2264
{
2265 2266 2267
	int i, zapped = 0;
	struct mmu_page_path parents;
	struct kvm_mmu_pages pages;
2268

2269
	if (parent->role.level == PT_PAGE_TABLE_LEVEL)
2270
		return 0;
2271 2272 2273 2274 2275 2276

	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) {
2277
			kvm_mmu_prepare_zap_page(kvm, sp, invalid_list);
2278
			mmu_pages_clear_parents(&parents);
2279
			zapped++;
2280 2281 2282 2283 2284
		}
		kvm_mmu_pages_init(parent, &parents, &pages);
	}

	return zapped;
2285 2286
}

2287 2288
static int kvm_mmu_prepare_zap_page(struct kvm *kvm, struct kvm_mmu_page *sp,
				    struct list_head *invalid_list)
2289
{
2290
	int ret;
A
Avi Kivity 已提交
2291

2292
	trace_kvm_mmu_prepare_zap_page(sp);
2293
	++kvm->stat.mmu_shadow_zapped;
2294
	ret = mmu_zap_unsync_children(kvm, sp, invalid_list);
2295
	kvm_mmu_page_unlink_children(kvm, sp);
2296
	kvm_mmu_unlink_parents(kvm, sp);
2297

2298
	if (!sp->role.invalid && !sp->role.direct)
2299
		unaccount_shadowed(kvm, sp);
2300

2301 2302
	if (sp->unsync)
		kvm_unlink_unsync_page(kvm, sp);
2303
	if (!sp->root_count) {
2304 2305
		/* Count self */
		ret++;
2306
		list_move(&sp->link, invalid_list);
2307
		kvm_mod_used_mmu_pages(kvm, -1);
2308
	} else {
A
Avi Kivity 已提交
2309
		list_move(&sp->link, &kvm->arch.active_mmu_pages);
2310 2311 2312 2313 2314 2315 2316

		/*
		 * 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);
2317
	}
2318 2319

	sp->role.invalid = 1;
2320
	return ret;
2321 2322
}

2323 2324 2325
static void kvm_mmu_commit_zap_page(struct kvm *kvm,
				    struct list_head *invalid_list)
{
2326
	struct kvm_mmu_page *sp, *nsp;
2327 2328 2329 2330

	if (list_empty(invalid_list))
		return;

2331 2332 2333 2334 2335
	/*
	 * 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
2336

2337 2338 2339 2340 2341
	/*
	 * Wait for all vcpus to exit guest mode and/or lockless shadow
	 * page table walks.
	 */
	kvm_flush_remote_tlbs(kvm);
2342

2343
	list_for_each_entry_safe(sp, nsp, invalid_list, link) {
2344
		WARN_ON(!sp->role.invalid || sp->root_count);
2345
		kvm_mmu_free_page(sp);
2346
	}
2347 2348
}

2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363
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;
}

2364 2365
/*
 * Changing the number of mmu pages allocated to the vm
2366
 * Note: if goal_nr_mmu_pages is too small, you will get dead lock
2367
 */
2368
void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int goal_nr_mmu_pages)
2369
{
2370
	LIST_HEAD(invalid_list);
2371

2372 2373
	spin_lock(&kvm->mmu_lock);

2374
	if (kvm->arch.n_used_mmu_pages > goal_nr_mmu_pages) {
2375 2376 2377 2378
		/* 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;
2379

2380
		kvm_mmu_commit_zap_page(kvm, &invalid_list);
2381
		goal_nr_mmu_pages = kvm->arch.n_used_mmu_pages;
2382 2383
	}

2384
	kvm->arch.n_max_mmu_pages = goal_nr_mmu_pages;
2385 2386

	spin_unlock(&kvm->mmu_lock);
2387 2388
}

2389
int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn)
2390
{
2391
	struct kvm_mmu_page *sp;
2392
	LIST_HEAD(invalid_list);
2393 2394
	int r;

2395
	pgprintk("%s: looking for gfn %llx\n", __func__, gfn);
2396
	r = 0;
2397
	spin_lock(&kvm->mmu_lock);
2398
	for_each_gfn_indirect_valid_sp(kvm, sp, gfn) {
2399
		pgprintk("%s: gfn %llx role %x\n", __func__, gfn,
2400 2401
			 sp->role.word);
		r = 1;
2402
		kvm_mmu_prepare_zap_page(kvm, sp, &invalid_list);
2403
	}
2404
	kvm_mmu_commit_zap_page(kvm, &invalid_list);
2405 2406
	spin_unlock(&kvm->mmu_lock);

2407
	return r;
2408
}
2409
EXPORT_SYMBOL_GPL(kvm_mmu_unprotect_page);
2410

2411 2412 2413 2414 2415 2416 2417 2418 2419 2420
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)
2421 2422
{
	struct kvm_mmu_page *s;
2423

2424
	for_each_gfn_indirect_valid_sp(vcpu->kvm, s, gfn) {
2425
		if (s->unsync)
2426
			continue;
2427 2428
		WARN_ON(s->role.level != PT_PAGE_TABLE_LEVEL);
		__kvm_unsync_page(vcpu, s);
2429 2430 2431 2432 2433 2434
	}
}

static int mmu_need_write_protect(struct kvm_vcpu *vcpu, gfn_t gfn,
				  bool can_unsync)
{
2435 2436 2437
	struct kvm_mmu_page *s;
	bool need_unsync = false;

2438
	for_each_gfn_indirect_valid_sp(vcpu->kvm, s, gfn) {
2439 2440 2441
		if (!can_unsync)
			return 1;

2442
		if (s->role.level != PT_PAGE_TABLE_LEVEL)
2443
			return 1;
2444

G
Gleb Natapov 已提交
2445
		if (!s->unsync)
2446
			need_unsync = true;
2447
	}
2448 2449
	if (need_unsync)
		kvm_unsync_pages(vcpu, gfn);
2450 2451 2452
	return 0;
}

D
Dan Williams 已提交
2453
static bool kvm_is_mmio_pfn(kvm_pfn_t pfn)
2454 2455 2456 2457 2458 2459 2460
{
	if (pfn_valid(pfn))
		return !is_zero_pfn(pfn) && PageReserved(pfn_to_page(pfn));

	return true;
}

A
Avi Kivity 已提交
2461
static int set_spte(struct kvm_vcpu *vcpu, u64 *sptep,
2462
		    unsigned pte_access, int level,
D
Dan Williams 已提交
2463
		    gfn_t gfn, kvm_pfn_t pfn, bool speculative,
2464
		    bool can_unsync, bool host_writable)
2465
{
2466
	u64 spte;
M
Marcelo Tosatti 已提交
2467
	int ret = 0;
S
Sheng Yang 已提交
2468

2469
	if (set_mmio_spte(vcpu, sptep, gfn, pfn, pte_access))
2470 2471
		return 0;

2472
	spte = PT_PRESENT_MASK;
2473
	if (!speculative)
2474
		spte |= shadow_accessed_mask;
2475

S
Sheng Yang 已提交
2476 2477 2478 2479
	if (pte_access & ACC_EXEC_MASK)
		spte |= shadow_x_mask;
	else
		spte |= shadow_nx_mask;
2480

2481
	if (pte_access & ACC_USER_MASK)
S
Sheng Yang 已提交
2482
		spte |= shadow_user_mask;
2483

2484
	if (level > PT_PAGE_TABLE_LEVEL)
M
Marcelo Tosatti 已提交
2485
		spte |= PT_PAGE_SIZE_MASK;
2486
	if (tdp_enabled)
2487
		spte |= kvm_x86_ops->get_mt_mask(vcpu, gfn,
2488
			kvm_is_mmio_pfn(pfn));
2489

2490
	if (host_writable)
2491
		spte |= SPTE_HOST_WRITEABLE;
2492 2493
	else
		pte_access &= ~ACC_WRITE_MASK;
2494

2495
	spte |= (u64)pfn << PAGE_SHIFT;
2496

2497
	if (pte_access & ACC_WRITE_MASK) {
2498

X
Xiao Guangrong 已提交
2499
		/*
2500 2501 2502 2503
		 * 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
2504
		 */
2505
		if (level > PT_PAGE_TABLE_LEVEL &&
2506
		    has_wrprotected_page(vcpu, gfn, level))
A
Avi Kivity 已提交
2507
			goto done;
2508

2509
		spte |= PT_WRITABLE_MASK | SPTE_MMU_WRITEABLE;
2510

2511 2512 2513 2514 2515 2516
		/*
		 * 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.
		 */
2517
		if (!can_unsync && is_writable_pte(*sptep))
2518 2519
			goto set_pte;

2520
		if (mmu_need_write_protect(vcpu, gfn, can_unsync)) {
2521
			pgprintk("%s: found shadow page for %llx, marking ro\n",
2522
				 __func__, gfn);
M
Marcelo Tosatti 已提交
2523
			ret = 1;
2524
			pte_access &= ~ACC_WRITE_MASK;
2525
			spte &= ~(PT_WRITABLE_MASK | SPTE_MMU_WRITEABLE);
2526 2527 2528
		}
	}

2529
	if (pte_access & ACC_WRITE_MASK) {
2530
		kvm_vcpu_mark_page_dirty(vcpu, gfn);
2531 2532
		spte |= shadow_dirty_mask;
	}
2533

2534
set_pte:
2535
	if (mmu_spte_update(sptep, spte))
2536
		kvm_flush_remote_tlbs(vcpu->kvm);
A
Avi Kivity 已提交
2537
done:
M
Marcelo Tosatti 已提交
2538 2539 2540
	return ret;
}

2541
static bool mmu_set_spte(struct kvm_vcpu *vcpu, u64 *sptep, unsigned pte_access,
D
Dan Williams 已提交
2542
			 int write_fault, int level, gfn_t gfn, kvm_pfn_t pfn,
2543
			 bool speculative, bool host_writable)
M
Marcelo Tosatti 已提交
2544 2545
{
	int was_rmapped = 0;
2546
	int rmap_count;
2547
	bool emulate = false;
M
Marcelo Tosatti 已提交
2548

2549 2550
	pgprintk("%s: spte %llx write_fault %d gfn %llx\n", __func__,
		 *sptep, write_fault, gfn);
M
Marcelo Tosatti 已提交
2551

2552
	if (is_shadow_present_pte(*sptep)) {
M
Marcelo Tosatti 已提交
2553 2554 2555 2556
		/*
		 * If we overwrite a PTE page pointer with a 2MB PMD, unlink
		 * the parent of the now unreachable PTE.
		 */
2557 2558
		if (level > PT_PAGE_TABLE_LEVEL &&
		    !is_large_pte(*sptep)) {
M
Marcelo Tosatti 已提交
2559
			struct kvm_mmu_page *child;
A
Avi Kivity 已提交
2560
			u64 pte = *sptep;
M
Marcelo Tosatti 已提交
2561 2562

			child = page_header(pte & PT64_BASE_ADDR_MASK);
2563
			drop_parent_pte(child, sptep);
2564
			kvm_flush_remote_tlbs(vcpu->kvm);
A
Avi Kivity 已提交
2565
		} else if (pfn != spte_to_pfn(*sptep)) {
2566
			pgprintk("hfn old %llx new %llx\n",
A
Avi Kivity 已提交
2567
				 spte_to_pfn(*sptep), pfn);
2568
			drop_spte(vcpu->kvm, sptep);
2569
			kvm_flush_remote_tlbs(vcpu->kvm);
2570 2571
		} else
			was_rmapped = 1;
M
Marcelo Tosatti 已提交
2572
	}
2573

2574 2575
	if (set_spte(vcpu, sptep, pte_access, level, gfn, pfn, speculative,
	      true, host_writable)) {
M
Marcelo Tosatti 已提交
2576
		if (write_fault)
2577
			emulate = true;
2578
		kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
2579
	}
M
Marcelo Tosatti 已提交
2580

2581 2582
	if (unlikely(is_mmio_spte(*sptep)))
		emulate = true;
2583

A
Avi Kivity 已提交
2584
	pgprintk("%s: setting spte %llx\n", __func__, *sptep);
2585
	pgprintk("instantiating %s PTE (%s) at %llx (%llx) addr %p\n",
A
Avi Kivity 已提交
2586
		 is_large_pte(*sptep)? "2MB" : "4kB",
2587 2588
		 *sptep & PT_PRESENT_MASK ?"RW":"R", gfn,
		 *sptep, sptep);
A
Avi Kivity 已提交
2589
	if (!was_rmapped && is_large_pte(*sptep))
M
Marcelo Tosatti 已提交
2590 2591
		++vcpu->kvm->stat.lpages;

2592 2593 2594 2595 2596 2597
	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);
		}
2598
	}
2599

X
Xiao Guangrong 已提交
2600
	kvm_release_pfn_clean(pfn);
2601 2602

	return emulate;
2603 2604
}

D
Dan Williams 已提交
2605
static kvm_pfn_t pte_prefetch_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn,
2606 2607 2608 2609
				     bool no_dirty_log)
{
	struct kvm_memory_slot *slot;

2610
	slot = gfn_to_memslot_dirty_bitmap(vcpu, gfn, no_dirty_log);
2611
	if (!slot)
2612
		return KVM_PFN_ERR_FAULT;
2613

2614
	return gfn_to_pfn_memslot_atomic(slot, gfn);
2615 2616 2617 2618 2619 2620 2621
}

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];
2622
	struct kvm_memory_slot *slot;
2623 2624 2625 2626 2627
	unsigned access = sp->role.access;
	int i, ret;
	gfn_t gfn;

	gfn = kvm_mmu_page_get_gfn(sp, start - sp->spt);
2628 2629
	slot = gfn_to_memslot_dirty_bitmap(vcpu, gfn, access & ACC_WRITE_MASK);
	if (!slot)
2630 2631
		return -1;

2632
	ret = gfn_to_page_many_atomic(slot, gfn, pages, end - start);
2633 2634 2635 2636
	if (ret <= 0)
		return -1;

	for (i = 0; i < ret; i++, gfn++, start++)
2637 2638
		mmu_set_spte(vcpu, start, access, 0, sp->role.level, gfn,
			     page_to_pfn(pages[i]), true, true);
2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654

	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++) {
2655
		if (is_shadow_present_pte(*spte) || spte == sptep) {
2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685
			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);
}

2686
static int __direct_map(struct kvm_vcpu *vcpu, int write, int map_writable,
D
Dan Williams 已提交
2687
			int level, gfn_t gfn, kvm_pfn_t pfn, bool prefault)
2688
{
2689
	struct kvm_shadow_walk_iterator iterator;
2690
	struct kvm_mmu_page *sp;
2691
	int emulate = 0;
2692
	gfn_t pseudo_gfn;
A
Avi Kivity 已提交
2693

2694 2695 2696
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
		return 0;

2697
	for_each_shadow_entry(vcpu, (u64)gfn << PAGE_SHIFT, iterator) {
2698
		if (iterator.level == level) {
2699 2700 2701
			emulate = mmu_set_spte(vcpu, iterator.sptep, ACC_ALL,
					       write, level, gfn, pfn, prefault,
					       map_writable);
2702
			direct_pte_prefetch(vcpu, iterator.sptep);
2703 2704
			++vcpu->stat.pf_fixed;
			break;
A
Avi Kivity 已提交
2705 2706
		}

2707
		drop_large_spte(vcpu, iterator.sptep);
2708
		if (!is_shadow_present_pte(*iterator.sptep)) {
2709 2710 2711 2712
			u64 base_addr = iterator.addr;

			base_addr &= PT64_LVL_ADDR_MASK(iterator.level);
			pseudo_gfn = base_addr >> PAGE_SHIFT;
2713
			sp = kvm_mmu_get_page(vcpu, pseudo_gfn, iterator.addr,
2714
					      iterator.level - 1, 1, ACC_ALL);
2715

2716
			link_shadow_page(vcpu, iterator.sptep, sp);
2717 2718
		}
	}
2719
	return emulate;
A
Avi Kivity 已提交
2720 2721
}

H
Huang Ying 已提交
2722
static void kvm_send_hwpoison_signal(unsigned long address, struct task_struct *tsk)
2723
{
H
Huang Ying 已提交
2724 2725 2726 2727 2728 2729 2730
	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;
2731

H
Huang Ying 已提交
2732
	send_sig_info(SIGBUS, &info, tsk);
2733 2734
}

D
Dan Williams 已提交
2735
static int kvm_handle_bad_page(struct kvm_vcpu *vcpu, gfn_t gfn, kvm_pfn_t pfn)
2736
{
X
Xiao Guangrong 已提交
2737 2738 2739 2740 2741 2742 2743 2744 2745
	/*
	 * 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;

2746
	if (pfn == KVM_PFN_ERR_HWPOISON) {
2747
		kvm_send_hwpoison_signal(kvm_vcpu_gfn_to_hva(vcpu, gfn), current);
2748
		return 0;
2749
	}
2750

2751
	return -EFAULT;
2752 2753
}

2754
static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
D
Dan Williams 已提交
2755 2756
					gfn_t *gfnp, kvm_pfn_t *pfnp,
					int *levelp)
2757
{
D
Dan Williams 已提交
2758
	kvm_pfn_t pfn = *pfnp;
2759 2760 2761 2762 2763 2764 2765 2766 2767
	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.
	 */
2768
	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn) &&
2769 2770
	    level == PT_PAGE_TABLE_LEVEL &&
	    PageTransCompound(pfn_to_page(pfn)) &&
2771
	    !has_wrprotected_page(vcpu, gfn, PT_DIRECTORY_LEVEL)) {
2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789
		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;
2790
			kvm_get_pfn(pfn);
2791 2792 2793 2794 2795
			*pfnp = pfn;
		}
	}
}

2796
static bool handle_abnormal_pfn(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
D
Dan Williams 已提交
2797
				kvm_pfn_t pfn, unsigned access, int *ret_val)
2798 2799 2800 2801
{
	bool ret = true;

	/* The pfn is invalid, report the error! */
2802
	if (unlikely(is_error_pfn(pfn))) {
2803 2804 2805 2806
		*ret_val = kvm_handle_bad_page(vcpu, gfn, pfn);
		goto exit;
	}

2807
	if (unlikely(is_noslot_pfn(pfn)))
2808 2809 2810 2811 2812 2813 2814
		vcpu_cache_mmio_info(vcpu, gva, gfn, access);

	ret = false;
exit:
	return ret;
}

2815
static bool page_fault_can_be_fast(u32 error_code)
2816
{
2817 2818 2819 2820 2821 2822 2823
	/*
	 * 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;

2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836
	/*
	 * #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
2837 2838
fast_pf_fix_direct_spte(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
			u64 *sptep, u64 spte)
2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849
{
	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);

2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861
	/*
	 * 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.
	 */
2862
	if (cmpxchg64(sptep, spte, spte | PT_WRITABLE_MASK) == spte)
2863
		kvm_vcpu_mark_page_dirty(vcpu, gfn);
2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876

	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;
2877
	struct kvm_mmu_page *sp;
2878 2879 2880
	bool ret = false;
	u64 spte = 0ull;

2881 2882 2883
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
		return false;

2884
	if (!page_fault_can_be_fast(error_code))
2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895
		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.
	 */
2896
	if (!is_shadow_present_pte(spte)) {
2897 2898 2899 2900
		ret = true;
		goto exit;
	}

2901 2902
	sp = page_header(__pa(iterator.sptep));
	if (!is_last_spte(spte, sp->role.level))
2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922
		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;

2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935
	/*
	 * 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;

2936 2937 2938 2939 2940
	/*
	 * 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.
	 */
2941
	ret = fast_pf_fix_direct_spte(vcpu, sp, iterator.sptep, spte);
2942
exit:
X
Xiao Guangrong 已提交
2943 2944
	trace_fast_page_fault(vcpu, gva, error_code, iterator.sptep,
			      spte, ret);
2945 2946 2947 2948 2949
	walk_shadow_page_lockless_end(vcpu);

	return ret;
}

2950
static bool try_async_pf(struct kvm_vcpu *vcpu, bool prefault, gfn_t gfn,
D
Dan Williams 已提交
2951
			 gva_t gva, kvm_pfn_t *pfn, bool write, bool *writable);
2952
static void make_mmu_pages_available(struct kvm_vcpu *vcpu);
2953

2954 2955
static int nonpaging_map(struct kvm_vcpu *vcpu, gva_t v, u32 error_code,
			 gfn_t gfn, bool prefault)
2956 2957
{
	int r;
2958
	int level;
2959
	bool force_pt_level = false;
D
Dan Williams 已提交
2960
	kvm_pfn_t pfn;
2961
	unsigned long mmu_seq;
2962
	bool map_writable, write = error_code & PFERR_WRITE_MASK;
2963

2964
	level = mapping_level(vcpu, gfn, &force_pt_level);
2965 2966 2967 2968 2969 2970 2971 2972
	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;
2973

2974
		gfn &= ~(KVM_PAGES_PER_HPAGE(level) - 1);
2975
	}
M
Marcelo Tosatti 已提交
2976

2977 2978 2979
	if (fast_page_fault(vcpu, v, level, error_code))
		return 0;

2980
	mmu_seq = vcpu->kvm->mmu_notifier_seq;
2981
	smp_rmb();
2982

2983
	if (try_async_pf(vcpu, prefault, gfn, v, &pfn, write, &map_writable))
2984
		return 0;
2985

2986 2987
	if (handle_abnormal_pfn(vcpu, v, gfn, pfn, ACC_ALL, &r))
		return r;
2988

2989
	spin_lock(&vcpu->kvm->mmu_lock);
2990
	if (mmu_notifier_retry(vcpu->kvm, mmu_seq))
2991
		goto out_unlock;
2992
	make_mmu_pages_available(vcpu);
2993 2994
	if (likely(!force_pt_level))
		transparent_hugepage_adjust(vcpu, &gfn, &pfn, &level);
2995
	r = __direct_map(vcpu, write, map_writable, level, gfn, pfn, prefault);
2996 2997
	spin_unlock(&vcpu->kvm->mmu_lock);

2998
	return r;
2999 3000 3001 3002 3003

out_unlock:
	spin_unlock(&vcpu->kvm->mmu_lock);
	kvm_release_pfn_clean(pfn);
	return 0;
3004 3005 3006
}


3007 3008 3009
static void mmu_free_roots(struct kvm_vcpu *vcpu)
{
	int i;
3010
	struct kvm_mmu_page *sp;
3011
	LIST_HEAD(invalid_list);
3012

3013
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
A
Avi Kivity 已提交
3014
		return;
3015

3016 3017 3018
	if (vcpu->arch.mmu.shadow_root_level == PT64_ROOT_LEVEL &&
	    (vcpu->arch.mmu.root_level == PT64_ROOT_LEVEL ||
	     vcpu->arch.mmu.direct_map)) {
3019
		hpa_t root = vcpu->arch.mmu.root_hpa;
3020

3021
		spin_lock(&vcpu->kvm->mmu_lock);
3022 3023
		sp = page_header(root);
		--sp->root_count;
3024 3025 3026 3027
		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);
		}
3028
		spin_unlock(&vcpu->kvm->mmu_lock);
3029
		vcpu->arch.mmu.root_hpa = INVALID_PAGE;
3030 3031
		return;
	}
3032 3033

	spin_lock(&vcpu->kvm->mmu_lock);
3034
	for (i = 0; i < 4; ++i) {
3035
		hpa_t root = vcpu->arch.mmu.pae_root[i];
3036

A
Avi Kivity 已提交
3037 3038
		if (root) {
			root &= PT64_BASE_ADDR_MASK;
3039 3040
			sp = page_header(root);
			--sp->root_count;
3041
			if (!sp->root_count && sp->role.invalid)
3042 3043
				kvm_mmu_prepare_zap_page(vcpu->kvm, sp,
							 &invalid_list);
A
Avi Kivity 已提交
3044
		}
3045
		vcpu->arch.mmu.pae_root[i] = INVALID_PAGE;
3046
	}
3047
	kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
3048
	spin_unlock(&vcpu->kvm->mmu_lock);
3049
	vcpu->arch.mmu.root_hpa = INVALID_PAGE;
3050 3051
}

3052 3053 3054 3055 3056
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)) {
3057
		kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu);
3058 3059 3060 3061 3062 3063
		ret = 1;
	}

	return ret;
}

3064 3065 3066
static int mmu_alloc_direct_roots(struct kvm_vcpu *vcpu)
{
	struct kvm_mmu_page *sp;
3067
	unsigned i;
3068 3069 3070

	if (vcpu->arch.mmu.shadow_root_level == PT64_ROOT_LEVEL) {
		spin_lock(&vcpu->kvm->mmu_lock);
3071
		make_mmu_pages_available(vcpu);
3072
		sp = kvm_mmu_get_page(vcpu, 0, 0, PT64_ROOT_LEVEL, 1, ACC_ALL);
3073 3074 3075 3076 3077 3078 3079
		++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];

3080
			MMU_WARN_ON(VALID_PAGE(root));
3081
			spin_lock(&vcpu->kvm->mmu_lock);
3082
			make_mmu_pages_available(vcpu);
3083
			sp = kvm_mmu_get_page(vcpu, i << (30 - PAGE_SHIFT),
3084
					i << 30, PT32_ROOT_LEVEL, 1, ACC_ALL);
3085 3086 3087 3088 3089
			root = __pa(sp->spt);
			++sp->root_count;
			spin_unlock(&vcpu->kvm->mmu_lock);
			vcpu->arch.mmu.pae_root[i] = root | PT_PRESENT_MASK;
		}
3090
		vcpu->arch.mmu.root_hpa = __pa(vcpu->arch.mmu.pae_root);
3091 3092 3093 3094 3095 3096 3097
	} else
		BUG();

	return 0;
}

static int mmu_alloc_shadow_roots(struct kvm_vcpu *vcpu)
3098
{
3099
	struct kvm_mmu_page *sp;
3100 3101 3102
	u64 pdptr, pm_mask;
	gfn_t root_gfn;
	int i;
3103

3104
	root_gfn = vcpu->arch.mmu.get_cr3(vcpu) >> PAGE_SHIFT;
3105

3106 3107 3108 3109 3110 3111 3112 3113
	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) {
3114
		hpa_t root = vcpu->arch.mmu.root_hpa;
3115

3116
		MMU_WARN_ON(VALID_PAGE(root));
3117

3118
		spin_lock(&vcpu->kvm->mmu_lock);
3119
		make_mmu_pages_available(vcpu);
3120
		sp = kvm_mmu_get_page(vcpu, root_gfn, 0, PT64_ROOT_LEVEL,
3121
				      0, ACC_ALL);
3122 3123
		root = __pa(sp->spt);
		++sp->root_count;
3124
		spin_unlock(&vcpu->kvm->mmu_lock);
3125
		vcpu->arch.mmu.root_hpa = root;
3126
		return 0;
3127
	}
3128

3129 3130
	/*
	 * We shadow a 32 bit page table. This may be a legacy 2-level
3131 3132
	 * 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.
3133
	 */
3134 3135 3136 3137
	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;

3138
	for (i = 0; i < 4; ++i) {
3139
		hpa_t root = vcpu->arch.mmu.pae_root[i];
3140

3141
		MMU_WARN_ON(VALID_PAGE(root));
3142
		if (vcpu->arch.mmu.root_level == PT32E_ROOT_LEVEL) {
3143
			pdptr = vcpu->arch.mmu.get_pdptr(vcpu, i);
3144
			if (!is_present_gpte(pdptr)) {
3145
				vcpu->arch.mmu.pae_root[i] = 0;
A
Avi Kivity 已提交
3146 3147
				continue;
			}
A
Avi Kivity 已提交
3148
			root_gfn = pdptr >> PAGE_SHIFT;
3149 3150
			if (mmu_check_root(vcpu, root_gfn))
				return 1;
3151
		}
3152
		spin_lock(&vcpu->kvm->mmu_lock);
3153
		make_mmu_pages_available(vcpu);
3154 3155
		sp = kvm_mmu_get_page(vcpu, root_gfn, i << 30, PT32_ROOT_LEVEL,
				      0, ACC_ALL);
3156 3157
		root = __pa(sp->spt);
		++sp->root_count;
3158 3159
		spin_unlock(&vcpu->kvm->mmu_lock);

3160
		vcpu->arch.mmu.pae_root[i] = root | pm_mask;
3161
	}
3162
	vcpu->arch.mmu.root_hpa = __pa(vcpu->arch.mmu.pae_root);
3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188

	/*
	 * 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);
	}

3189
	return 0;
3190 3191
}

3192 3193 3194 3195 3196 3197 3198 3199
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);
}

3200 3201 3202 3203 3204
static void mmu_sync_roots(struct kvm_vcpu *vcpu)
{
	int i;
	struct kvm_mmu_page *sp;

3205 3206 3207
	if (vcpu->arch.mmu.direct_map)
		return;

3208 3209
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
		return;
3210

3211
	vcpu_clear_mmio_info(vcpu, MMIO_GVA_ANY);
3212
	kvm_mmu_audit(vcpu, AUDIT_PRE_SYNC);
3213
	if (vcpu->arch.mmu.root_level == PT64_ROOT_LEVEL) {
3214 3215 3216
		hpa_t root = vcpu->arch.mmu.root_hpa;
		sp = page_header(root);
		mmu_sync_children(vcpu, sp);
3217
		kvm_mmu_audit(vcpu, AUDIT_POST_SYNC);
3218 3219 3220 3221 3222
		return;
	}
	for (i = 0; i < 4; ++i) {
		hpa_t root = vcpu->arch.mmu.pae_root[i];

3223
		if (root && VALID_PAGE(root)) {
3224 3225 3226 3227 3228
			root &= PT64_BASE_ADDR_MASK;
			sp = page_header(root);
			mmu_sync_children(vcpu, sp);
		}
	}
3229
	kvm_mmu_audit(vcpu, AUDIT_POST_SYNC);
3230 3231 3232 3233 3234 3235
}

void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu)
{
	spin_lock(&vcpu->kvm->mmu_lock);
	mmu_sync_roots(vcpu);
3236
	spin_unlock(&vcpu->kvm->mmu_lock);
3237
}
N
Nadav Har'El 已提交
3238
EXPORT_SYMBOL_GPL(kvm_mmu_sync_roots);
3239

3240
static gpa_t nonpaging_gva_to_gpa(struct kvm_vcpu *vcpu, gva_t vaddr,
3241
				  u32 access, struct x86_exception *exception)
A
Avi Kivity 已提交
3242
{
3243 3244
	if (exception)
		exception->error_code = 0;
A
Avi Kivity 已提交
3245 3246 3247
	return vaddr;
}

3248
static gpa_t nonpaging_gva_to_gpa_nested(struct kvm_vcpu *vcpu, gva_t vaddr,
3249 3250
					 u32 access,
					 struct x86_exception *exception)
3251
{
3252 3253
	if (exception)
		exception->error_code = 0;
3254
	return vcpu->arch.nested_mmu.translate_gpa(vcpu, vaddr, access, exception);
3255 3256
}

3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275
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);
}

3276
static bool mmio_info_in_cache(struct kvm_vcpu *vcpu, u64 addr, bool direct)
3277 3278 3279 3280 3281 3282 3283
{
	if (direct)
		return vcpu_match_mmio_gpa(vcpu, addr);

	return vcpu_match_mmio_gva(vcpu, addr);
}

3284 3285 3286
/* 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)
3287 3288
{
	struct kvm_shadow_walk_iterator iterator;
3289 3290 3291
	u64 sptes[PT64_ROOT_LEVEL], spte = 0ull;
	int root, leaf;
	bool reserved = false;
3292

3293
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
3294
		goto exit;
3295

3296
	walk_shadow_page_lockless_begin(vcpu);
3297

3298 3299
	for (shadow_walk_init(&iterator, vcpu, addr),
		 leaf = root = iterator.level;
3300 3301 3302 3303 3304
	     shadow_walk_okay(&iterator);
	     __shadow_walk_next(&iterator, spte)) {
		spte = mmu_spte_get_lockless(iterator.sptep);

		sptes[leaf - 1] = spte;
3305
		leaf--;
3306

3307 3308
		if (!is_shadow_present_pte(spte))
			break;
3309 3310

		reserved |= is_shadow_zero_bits_set(&vcpu->arch.mmu, spte,
3311
						    iterator.level);
3312 3313
	}

3314 3315
	walk_shadow_page_lockless_end(vcpu);

3316 3317 3318
	if (reserved) {
		pr_err("%s: detect reserved bits on spte, addr 0x%llx, dump hierarchy:\n",
		       __func__, addr);
3319
		while (root > leaf) {
3320 3321 3322 3323 3324 3325 3326 3327
			pr_err("------ spte 0x%llx level %d.\n",
			       sptes[root - 1], root);
			root--;
		}
	}
exit:
	*sptep = spte;
	return reserved;
3328 3329
}

3330
int handle_mmio_page_fault(struct kvm_vcpu *vcpu, u64 addr, bool direct)
3331 3332
{
	u64 spte;
3333
	bool reserved;
3334

3335
	if (mmio_info_in_cache(vcpu, addr, direct))
3336
		return RET_MMIO_PF_EMULATE;
3337

3338
	reserved = walk_shadow_page_get_mmio_spte(vcpu, addr, &spte);
3339
	if (WARN_ON(reserved))
3340
		return RET_MMIO_PF_BUG;
3341 3342 3343 3344 3345

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

3346
		if (!check_mmio_spte(vcpu, spte))
3347 3348
			return RET_MMIO_PF_INVALID;

3349 3350
		if (direct)
			addr = 0;
X
Xiao Guangrong 已提交
3351 3352

		trace_handle_mmio_page_fault(addr, gfn, access);
3353
		vcpu_cache_mmio_info(vcpu, addr, gfn, access);
3354
		return RET_MMIO_PF_EMULATE;
3355 3356 3357 3358 3359 3360
	}

	/*
	 * If the page table is zapped by other cpus, let CPU fault again on
	 * the address.
	 */
3361
	return RET_MMIO_PF_RETRY;
3362
}
3363
EXPORT_SYMBOL_GPL(handle_mmio_page_fault);
3364

A
Avi Kivity 已提交
3365
static int nonpaging_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
3366
				u32 error_code, bool prefault)
A
Avi Kivity 已提交
3367
{
3368
	gfn_t gfn;
3369
	int r;
A
Avi Kivity 已提交
3370

3371
	pgprintk("%s: gva %lx error %x\n", __func__, gva, error_code);
3372

3373 3374 3375
	r = mmu_topup_memory_caches(vcpu);
	if (r)
		return r;
3376

3377
	MMU_WARN_ON(!VALID_PAGE(vcpu->arch.mmu.root_hpa));
A
Avi Kivity 已提交
3378

3379
	gfn = gva >> PAGE_SHIFT;
A
Avi Kivity 已提交
3380

3381
	return nonpaging_map(vcpu, gva & PAGE_MASK,
3382
			     error_code, gfn, prefault);
A
Avi Kivity 已提交
3383 3384
}

3385
static int kvm_arch_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn)
3386 3387
{
	struct kvm_arch_async_pf arch;
X
Xiao Guangrong 已提交
3388

3389
	arch.token = (vcpu->arch.apf.id++ << 12) | vcpu->vcpu_id;
3390
	arch.gfn = gfn;
3391
	arch.direct_map = vcpu->arch.mmu.direct_map;
X
Xiao Guangrong 已提交
3392
	arch.cr3 = vcpu->arch.mmu.get_cr3(vcpu);
3393

3394
	return kvm_setup_async_pf(vcpu, gva, kvm_vcpu_gfn_to_hva(vcpu, gfn), &arch);
3395 3396 3397 3398
}

static bool can_do_async_pf(struct kvm_vcpu *vcpu)
{
3399
	if (unlikely(!lapic_in_kernel(vcpu) ||
3400 3401 3402 3403 3404 3405
		     kvm_event_needs_reinjection(vcpu)))
		return false;

	return kvm_x86_ops->interrupt_allowed(vcpu);
}

3406
static bool try_async_pf(struct kvm_vcpu *vcpu, bool prefault, gfn_t gfn,
D
Dan Williams 已提交
3407
			 gva_t gva, kvm_pfn_t *pfn, bool write, bool *writable)
3408
{
3409
	struct kvm_memory_slot *slot;
3410 3411
	bool async;

3412
	slot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);
3413 3414
	async = false;
	*pfn = __gfn_to_pfn_memslot(slot, gfn, false, &async, write, writable);
3415 3416 3417
	if (!async)
		return false; /* *pfn has correct page already */

3418
	if (!prefault && can_do_async_pf(vcpu)) {
3419
		trace_kvm_try_async_get_page(gva, gfn);
3420 3421 3422 3423 3424 3425 3426 3427
		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;
	}

3428
	*pfn = __gfn_to_pfn_memslot(slot, gfn, false, NULL, write, writable);
3429 3430 3431
	return false;
}

3432 3433 3434 3435 3436 3437 3438 3439 3440 3441
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
3442
static int tdp_page_fault(struct kvm_vcpu *vcpu, gva_t gpa, u32 error_code,
3443
			  bool prefault)
3444
{
D
Dan Williams 已提交
3445
	kvm_pfn_t pfn;
3446
	int r;
3447
	int level;
3448
	bool force_pt_level;
M
Marcelo Tosatti 已提交
3449
	gfn_t gfn = gpa >> PAGE_SHIFT;
3450
	unsigned long mmu_seq;
3451 3452
	int write = error_code & PFERR_WRITE_MASK;
	bool map_writable;
3453

3454
	MMU_WARN_ON(!VALID_PAGE(vcpu->arch.mmu.root_hpa));
3455 3456 3457 3458 3459

	r = mmu_topup_memory_caches(vcpu);
	if (r)
		return r;

3460 3461 3462
	force_pt_level = !check_hugepage_cache_consistency(vcpu, gfn,
							   PT_DIRECTORY_LEVEL);
	level = mapping_level(vcpu, gfn, &force_pt_level);
3463
	if (likely(!force_pt_level)) {
3464 3465 3466
		if (level > PT_DIRECTORY_LEVEL &&
		    !check_hugepage_cache_consistency(vcpu, gfn, level))
			level = PT_DIRECTORY_LEVEL;
3467
		gfn &= ~(KVM_PAGES_PER_HPAGE(level) - 1);
3468
	}
3469

3470 3471 3472
	if (fast_page_fault(vcpu, gpa, level, error_code))
		return 0;

3473
	mmu_seq = vcpu->kvm->mmu_notifier_seq;
3474
	smp_rmb();
3475

3476
	if (try_async_pf(vcpu, prefault, gfn, gpa, &pfn, write, &map_writable))
3477 3478
		return 0;

3479 3480 3481
	if (handle_abnormal_pfn(vcpu, 0, gfn, pfn, ACC_ALL, &r))
		return r;

3482
	spin_lock(&vcpu->kvm->mmu_lock);
3483
	if (mmu_notifier_retry(vcpu->kvm, mmu_seq))
3484
		goto out_unlock;
3485
	make_mmu_pages_available(vcpu);
3486 3487
	if (likely(!force_pt_level))
		transparent_hugepage_adjust(vcpu, &gfn, &pfn, &level);
3488
	r = __direct_map(vcpu, write, map_writable, level, gfn, pfn, prefault);
3489 3490 3491
	spin_unlock(&vcpu->kvm->mmu_lock);

	return r;
3492 3493 3494 3495 3496

out_unlock:
	spin_unlock(&vcpu->kvm->mmu_lock);
	kvm_release_pfn_clean(pfn);
	return 0;
3497 3498
}

3499 3500
static void nonpaging_init_context(struct kvm_vcpu *vcpu,
				   struct kvm_mmu *context)
A
Avi Kivity 已提交
3501 3502 3503
{
	context->page_fault = nonpaging_page_fault;
	context->gva_to_gpa = nonpaging_gva_to_gpa;
3504
	context->sync_page = nonpaging_sync_page;
M
Marcelo Tosatti 已提交
3505
	context->invlpg = nonpaging_invlpg;
3506
	context->update_pte = nonpaging_update_pte;
3507
	context->root_level = 0;
A
Avi Kivity 已提交
3508
	context->shadow_root_level = PT32E_ROOT_LEVEL;
A
Avi Kivity 已提交
3509
	context->root_hpa = INVALID_PAGE;
3510
	context->direct_map = true;
3511
	context->nx = false;
A
Avi Kivity 已提交
3512 3513
}

3514
void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
3515
{
3516
	mmu_free_roots(vcpu);
A
Avi Kivity 已提交
3517 3518
}

3519 3520
static unsigned long get_cr3(struct kvm_vcpu *vcpu)
{
3521
	return kvm_read_cr3(vcpu);
3522 3523
}

3524 3525
static void inject_page_fault(struct kvm_vcpu *vcpu,
			      struct x86_exception *fault)
A
Avi Kivity 已提交
3526
{
3527
	vcpu->arch.mmu.inject_page_fault(vcpu, fault);
A
Avi Kivity 已提交
3528 3529
}

3530
static bool sync_mmio_spte(struct kvm_vcpu *vcpu, u64 *sptep, gfn_t gfn,
3531
			   unsigned access, int *nr_present)
3532 3533 3534 3535 3536 3537 3538 3539
{
	if (unlikely(is_mmio_spte(*sptep))) {
		if (gfn != get_mmio_spte_gfn(*sptep)) {
			mmu_spte_clear_no_track(sptep);
			return true;
		}

		(*nr_present)++;
3540
		mark_mmio_spte(vcpu, sptep, gfn, access);
3541 3542 3543 3544 3545 3546
		return true;
	}

	return false;
}

A
Avi Kivity 已提交
3547 3548 3549 3550 3551 3552 3553 3554 3555
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);
}

3556 3557 3558 3559 3560
#define PTTYPE_EPT 18 /* arbitrary */
#define PTTYPE PTTYPE_EPT
#include "paging_tmpl.h"
#undef PTTYPE

A
Avi Kivity 已提交
3561 3562 3563 3564 3565 3566 3567 3568
#define PTTYPE 64
#include "paging_tmpl.h"
#undef PTTYPE

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

3569 3570 3571 3572
static void
__reset_rsvds_bits_mask(struct kvm_vcpu *vcpu,
			struct rsvd_bits_validate *rsvd_check,
			int maxphyaddr, int level, bool nx, bool gbpages,
3573
			bool pse, bool amd)
3574 3575
{
	u64 exb_bit_rsvd = 0;
3576
	u64 gbpages_bit_rsvd = 0;
3577
	u64 nonleaf_bit8_rsvd = 0;
3578

3579
	rsvd_check->bad_mt_xwr = 0;
3580

3581
	if (!nx)
3582
		exb_bit_rsvd = rsvd_bits(63, 63);
3583
	if (!gbpages)
3584
		gbpages_bit_rsvd = rsvd_bits(7, 7);
3585 3586 3587 3588 3589

	/*
	 * Non-leaf PML4Es and PDPEs reserve bit 8 (which would be the G bit for
	 * leaf entries) on AMD CPUs only.
	 */
3590
	if (amd)
3591 3592
		nonleaf_bit8_rsvd = rsvd_bits(8, 8);

3593
	switch (level) {
3594 3595
	case PT32_ROOT_LEVEL:
		/* no rsvd bits for 2 level 4K page table entries */
3596 3597 3598 3599
		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];
3600

3601
		if (!pse) {
3602
			rsvd_check->rsvd_bits_mask[1][1] = 0;
3603 3604 3605
			break;
		}

3606 3607
		if (is_cpuid_PSE36())
			/* 36bits PSE 4MB page */
3608
			rsvd_check->rsvd_bits_mask[1][1] = rsvd_bits(17, 21);
3609 3610
		else
			/* 32 bits PSE 4MB page */
3611
			rsvd_check->rsvd_bits_mask[1][1] = rsvd_bits(13, 21);
3612 3613
		break;
	case PT32E_ROOT_LEVEL:
3614
		rsvd_check->rsvd_bits_mask[0][2] =
3615
			rsvd_bits(maxphyaddr, 63) |
3616
			rsvd_bits(5, 8) | rsvd_bits(1, 2);	/* PDPTE */
3617
		rsvd_check->rsvd_bits_mask[0][1] = exb_bit_rsvd |
3618
			rsvd_bits(maxphyaddr, 62);	/* PDE */
3619
		rsvd_check->rsvd_bits_mask[0][0] = exb_bit_rsvd |
3620
			rsvd_bits(maxphyaddr, 62); 	/* PTE */
3621
		rsvd_check->rsvd_bits_mask[1][1] = exb_bit_rsvd |
3622 3623
			rsvd_bits(maxphyaddr, 62) |
			rsvd_bits(13, 20);		/* large page */
3624 3625
		rsvd_check->rsvd_bits_mask[1][0] =
			rsvd_check->rsvd_bits_mask[0][0];
3626 3627
		break;
	case PT64_ROOT_LEVEL:
3628 3629
		rsvd_check->rsvd_bits_mask[0][3] = exb_bit_rsvd |
			nonleaf_bit8_rsvd | rsvd_bits(7, 7) |
3630
			rsvd_bits(maxphyaddr, 51);
3631 3632
		rsvd_check->rsvd_bits_mask[0][2] = exb_bit_rsvd |
			nonleaf_bit8_rsvd | gbpages_bit_rsvd |
3633
			rsvd_bits(maxphyaddr, 51);
3634 3635 3636 3637 3638 3639 3640
		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 |
3641
			gbpages_bit_rsvd | rsvd_bits(maxphyaddr, 51) |
3642
			rsvd_bits(13, 29);
3643
		rsvd_check->rsvd_bits_mask[1][1] = exb_bit_rsvd |
3644 3645
			rsvd_bits(maxphyaddr, 51) |
			rsvd_bits(13, 20);		/* large page */
3646 3647
		rsvd_check->rsvd_bits_mask[1][0] =
			rsvd_check->rsvd_bits_mask[0][0];
3648 3649 3650 3651
		break;
	}
}

3652 3653 3654 3655 3656 3657
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),
3658
				is_pse(vcpu), guest_cpuid_is_amd(vcpu));
3659 3660
}

3661 3662 3663
static void
__reset_rsvds_bits_mask_ept(struct rsvd_bits_validate *rsvd_check,
			    int maxphyaddr, bool execonly)
3664
{
3665
	u64 bad_mt_xwr;
3666

3667
	rsvd_check->rsvd_bits_mask[0][3] =
3668
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(3, 7);
3669
	rsvd_check->rsvd_bits_mask[0][2] =
3670
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(3, 6);
3671
	rsvd_check->rsvd_bits_mask[0][1] =
3672
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(3, 6);
3673
	rsvd_check->rsvd_bits_mask[0][0] = rsvd_bits(maxphyaddr, 51);
3674 3675

	/* large page */
3676 3677
	rsvd_check->rsvd_bits_mask[1][3] = rsvd_check->rsvd_bits_mask[0][3];
	rsvd_check->rsvd_bits_mask[1][2] =
3678
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(12, 29);
3679
	rsvd_check->rsvd_bits_mask[1][1] =
3680
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(12, 20);
3681
	rsvd_check->rsvd_bits_mask[1][0] = rsvd_check->rsvd_bits_mask[0][0];
3682

3683 3684 3685 3686 3687 3688 3689 3690
	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);
3691
	}
3692
	rsvd_check->bad_mt_xwr = bad_mt_xwr;
3693 3694
}

3695 3696 3697 3698 3699 3700 3701
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);
}

3702 3703 3704 3705 3706 3707 3708 3709
/*
 * 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)
{
3710 3711 3712 3713
	/*
	 * Passing "true" to the last argument is okay; it adds a check
	 * on bit 8 of the SPTEs which KVM doesn't use anyway.
	 */
3714 3715 3716
	__reset_rsvds_bits_mask(vcpu, &context->shadow_zero_check,
				boot_cpu_data.x86_phys_bits,
				context->shadow_root_level, context->nx,
3717 3718
				guest_cpuid_has_gbpages(vcpu), is_pse(vcpu),
				true);
3719 3720 3721
}
EXPORT_SYMBOL_GPL(reset_shadow_zero_bits_mask);

3722 3723 3724 3725 3726 3727
static inline bool boot_cpu_is_amd(void)
{
	WARN_ON_ONCE(!tdp_enabled);
	return shadow_x_mask == 0;
}

3728 3729 3730 3731 3732 3733 3734 3735
/*
 * 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)
{
3736
	if (boot_cpu_is_amd())
3737 3738 3739
		__reset_rsvds_bits_mask(vcpu, &context->shadow_zero_check,
					boot_cpu_data.x86_phys_bits,
					context->shadow_root_level, false,
3740
					cpu_has_gbpages, true, true);
3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759
	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);
}

3760 3761
static void update_permission_bitmask(struct kvm_vcpu *vcpu,
				      struct kvm_mmu *mmu, bool ept)
3762 3763 3764
{
	unsigned bit, byte, pfec;
	u8 map;
F
Feng Wu 已提交
3765
	bool fault, x, w, u, wf, uf, ff, smapf, cr4_smap, cr4_smep, smap = 0;
3766

F
Feng Wu 已提交
3767
	cr4_smep = kvm_read_cr4_bits(vcpu, X86_CR4_SMEP);
F
Feng Wu 已提交
3768
	cr4_smap = kvm_read_cr4_bits(vcpu, X86_CR4_SMAP);
3769 3770 3771 3772 3773 3774
	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
3775 3776 3777 3778 3779 3780
		/*
		 * 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);
3781 3782 3783 3784 3785
		for (bit = 0; bit < 8; ++bit) {
			x = bit & ACC_EXEC_MASK;
			w = bit & ACC_WRITE_MASK;
			u = bit & ACC_USER_MASK;

3786 3787 3788 3789 3790 3791
			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
3792
				x &= !(cr4_smep && u && !uf);
F
Feng Wu 已提交
3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812

				/*
				 * 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;
3813 3814 3815
			} else
				/* Not really needed: no U/S accesses on ept  */
				u = 1;
3816

F
Feng Wu 已提交
3817 3818
			fault = (ff && !x) || (uf && !u) || (wf && !w) ||
				(smapf && smap);
3819 3820 3821 3822 3823 3824
			map |= fault << bit;
		}
		mmu->permissions[byte] = map;
	}
}

A
Avi Kivity 已提交
3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842
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;
}

3843 3844 3845
static void paging64_init_context_common(struct kvm_vcpu *vcpu,
					 struct kvm_mmu *context,
					 int level)
A
Avi Kivity 已提交
3846
{
3847
	context->nx = is_nx(vcpu);
3848
	context->root_level = level;
3849

3850
	reset_rsvds_bits_mask(vcpu, context);
3851
	update_permission_bitmask(vcpu, context, false);
A
Avi Kivity 已提交
3852
	update_last_pte_bitmap(vcpu, context);
A
Avi Kivity 已提交
3853

3854
	MMU_WARN_ON(!is_pae(vcpu));
A
Avi Kivity 已提交
3855 3856
	context->page_fault = paging64_page_fault;
	context->gva_to_gpa = paging64_gva_to_gpa;
3857
	context->sync_page = paging64_sync_page;
M
Marcelo Tosatti 已提交
3858
	context->invlpg = paging64_invlpg;
3859
	context->update_pte = paging64_update_pte;
3860
	context->shadow_root_level = level;
A
Avi Kivity 已提交
3861
	context->root_hpa = INVALID_PAGE;
3862
	context->direct_map = false;
A
Avi Kivity 已提交
3863 3864
}

3865 3866
static void paging64_init_context(struct kvm_vcpu *vcpu,
				  struct kvm_mmu *context)
3867
{
3868
	paging64_init_context_common(vcpu, context, PT64_ROOT_LEVEL);
3869 3870
}

3871 3872
static void paging32_init_context(struct kvm_vcpu *vcpu,
				  struct kvm_mmu *context)
A
Avi Kivity 已提交
3873
{
3874
	context->nx = false;
3875
	context->root_level = PT32_ROOT_LEVEL;
3876

3877
	reset_rsvds_bits_mask(vcpu, context);
3878
	update_permission_bitmask(vcpu, context, false);
A
Avi Kivity 已提交
3879
	update_last_pte_bitmap(vcpu, context);
A
Avi Kivity 已提交
3880 3881 3882

	context->page_fault = paging32_page_fault;
	context->gva_to_gpa = paging32_gva_to_gpa;
3883
	context->sync_page = paging32_sync_page;
M
Marcelo Tosatti 已提交
3884
	context->invlpg = paging32_invlpg;
3885
	context->update_pte = paging32_update_pte;
A
Avi Kivity 已提交
3886
	context->shadow_root_level = PT32E_ROOT_LEVEL;
A
Avi Kivity 已提交
3887
	context->root_hpa = INVALID_PAGE;
3888
	context->direct_map = false;
A
Avi Kivity 已提交
3889 3890
}

3891 3892
static void paging32E_init_context(struct kvm_vcpu *vcpu,
				   struct kvm_mmu *context)
A
Avi Kivity 已提交
3893
{
3894
	paging64_init_context_common(vcpu, context, PT32E_ROOT_LEVEL);
A
Avi Kivity 已提交
3895 3896
}

3897
static void init_kvm_tdp_mmu(struct kvm_vcpu *vcpu)
3898
{
3899
	struct kvm_mmu *context = &vcpu->arch.mmu;
3900

3901
	context->base_role.word = 0;
3902
	context->base_role.smm = is_smm(vcpu);
3903
	context->page_fault = tdp_page_fault;
3904
	context->sync_page = nonpaging_sync_page;
M
Marcelo Tosatti 已提交
3905
	context->invlpg = nonpaging_invlpg;
3906
	context->update_pte = nonpaging_update_pte;
3907
	context->shadow_root_level = kvm_x86_ops->get_tdp_level();
3908
	context->root_hpa = INVALID_PAGE;
3909
	context->direct_map = true;
3910
	context->set_cr3 = kvm_x86_ops->set_tdp_cr3;
3911
	context->get_cr3 = get_cr3;
3912
	context->get_pdptr = kvm_pdptr_read;
3913
	context->inject_page_fault = kvm_inject_page_fault;
3914 3915

	if (!is_paging(vcpu)) {
3916
		context->nx = false;
3917 3918 3919
		context->gva_to_gpa = nonpaging_gva_to_gpa;
		context->root_level = 0;
	} else if (is_long_mode(vcpu)) {
3920
		context->nx = is_nx(vcpu);
3921
		context->root_level = PT64_ROOT_LEVEL;
3922 3923
		reset_rsvds_bits_mask(vcpu, context);
		context->gva_to_gpa = paging64_gva_to_gpa;
3924
	} else if (is_pae(vcpu)) {
3925
		context->nx = is_nx(vcpu);
3926
		context->root_level = PT32E_ROOT_LEVEL;
3927 3928
		reset_rsvds_bits_mask(vcpu, context);
		context->gva_to_gpa = paging64_gva_to_gpa;
3929
	} else {
3930
		context->nx = false;
3931
		context->root_level = PT32_ROOT_LEVEL;
3932 3933
		reset_rsvds_bits_mask(vcpu, context);
		context->gva_to_gpa = paging32_gva_to_gpa;
3934 3935
	}

3936
	update_permission_bitmask(vcpu, context, false);
A
Avi Kivity 已提交
3937
	update_last_pte_bitmap(vcpu, context);
3938
	reset_tdp_shadow_zero_bits_mask(vcpu, context);
3939 3940
}

3941
void kvm_init_shadow_mmu(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
3942
{
3943
	bool smep = kvm_read_cr4_bits(vcpu, X86_CR4_SMEP);
3944
	bool smap = kvm_read_cr4_bits(vcpu, X86_CR4_SMAP);
3945 3946
	struct kvm_mmu *context = &vcpu->arch.mmu;

3947
	MMU_WARN_ON(VALID_PAGE(context->root_hpa));
A
Avi Kivity 已提交
3948 3949

	if (!is_paging(vcpu))
3950
		nonpaging_init_context(vcpu, context);
A
Avi Kivity 已提交
3951
	else if (is_long_mode(vcpu))
3952
		paging64_init_context(vcpu, context);
A
Avi Kivity 已提交
3953
	else if (is_pae(vcpu))
3954
		paging32E_init_context(vcpu, context);
A
Avi Kivity 已提交
3955
	else
3956
		paging32_init_context(vcpu, context);
3957

3958 3959 3960 3961
	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
3962
		= smep && !is_write_protection(vcpu);
3963 3964
	context->base_role.smap_andnot_wp
		= smap && !is_write_protection(vcpu);
3965
	context->base_role.smm = is_smm(vcpu);
3966
	reset_shadow_zero_bits_mask(vcpu, context);
3967 3968 3969
}
EXPORT_SYMBOL_GPL(kvm_init_shadow_mmu);

3970
void kvm_init_shadow_ept_mmu(struct kvm_vcpu *vcpu, bool execonly)
N
Nadav Har'El 已提交
3971
{
3972 3973
	struct kvm_mmu *context = &vcpu->arch.mmu;

3974
	MMU_WARN_ON(VALID_PAGE(context->root_hpa));
N
Nadav Har'El 已提交
3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989

	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);
3990
	reset_ept_shadow_zero_bits_mask(vcpu, context, execonly);
N
Nadav Har'El 已提交
3991 3992 3993
}
EXPORT_SYMBOL_GPL(kvm_init_shadow_ept_mmu);

3994
static void init_kvm_softmmu(struct kvm_vcpu *vcpu)
3995
{
3996 3997 3998 3999 4000 4001 4002
	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
4003 4004
}

4005
static void init_kvm_nested_mmu(struct kvm_vcpu *vcpu)
4006 4007 4008 4009
{
	struct kvm_mmu *g_context = &vcpu->arch.nested_mmu;

	g_context->get_cr3           = get_cr3;
4010
	g_context->get_pdptr         = kvm_pdptr_read;
4011 4012 4013
	g_context->inject_page_fault = kvm_inject_page_fault;

	/*
4014 4015 4016 4017 4018 4019
	 * Note that arch.mmu.gva_to_gpa translates l2_gpa to l1_gpa using
	 * L1's nested page tables (e.g. EPT12). The nested translation
	 * of l2_gva to l1_gpa is done by arch.nested_mmu.gva_to_gpa using
	 * L2's page tables as the first level of translation and L1's
	 * nested page tables as the second level of translation. Basically
	 * the gva_to_gpa functions between mmu and nested_mmu are swapped.
4020 4021
	 */
	if (!is_paging(vcpu)) {
4022
		g_context->nx = false;
4023 4024 4025
		g_context->root_level = 0;
		g_context->gva_to_gpa = nonpaging_gva_to_gpa_nested;
	} else if (is_long_mode(vcpu)) {
4026
		g_context->nx = is_nx(vcpu);
4027
		g_context->root_level = PT64_ROOT_LEVEL;
4028
		reset_rsvds_bits_mask(vcpu, g_context);
4029 4030
		g_context->gva_to_gpa = paging64_gva_to_gpa_nested;
	} else if (is_pae(vcpu)) {
4031
		g_context->nx = is_nx(vcpu);
4032
		g_context->root_level = PT32E_ROOT_LEVEL;
4033
		reset_rsvds_bits_mask(vcpu, g_context);
4034 4035
		g_context->gva_to_gpa = paging64_gva_to_gpa_nested;
	} else {
4036
		g_context->nx = false;
4037
		g_context->root_level = PT32_ROOT_LEVEL;
4038
		reset_rsvds_bits_mask(vcpu, g_context);
4039 4040 4041
		g_context->gva_to_gpa = paging32_gva_to_gpa_nested;
	}

4042
	update_permission_bitmask(vcpu, g_context, false);
A
Avi Kivity 已提交
4043
	update_last_pte_bitmap(vcpu, g_context);
4044 4045
}

4046
static void init_kvm_mmu(struct kvm_vcpu *vcpu)
4047
{
4048
	if (mmu_is_nested(vcpu))
4049
		init_kvm_nested_mmu(vcpu);
4050
	else if (tdp_enabled)
4051
		init_kvm_tdp_mmu(vcpu);
4052
	else
4053
		init_kvm_softmmu(vcpu);
4054 4055
}

4056
void kvm_mmu_reset_context(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4057
{
4058
	kvm_mmu_unload(vcpu);
4059
	init_kvm_mmu(vcpu);
A
Avi Kivity 已提交
4060
}
4061
EXPORT_SYMBOL_GPL(kvm_mmu_reset_context);
A
Avi Kivity 已提交
4062 4063

int kvm_mmu_load(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4064
{
4065 4066
	int r;

4067
	r = mmu_topup_memory_caches(vcpu);
A
Avi Kivity 已提交
4068 4069
	if (r)
		goto out;
4070
	r = mmu_alloc_roots(vcpu);
4071
	kvm_mmu_sync_roots(vcpu);
4072 4073
	if (r)
		goto out;
4074
	/* set_cr3() should ensure TLB has been flushed */
4075
	vcpu->arch.mmu.set_cr3(vcpu, vcpu->arch.mmu.root_hpa);
4076 4077
out:
	return r;
A
Avi Kivity 已提交
4078
}
A
Avi Kivity 已提交
4079 4080 4081 4082 4083
EXPORT_SYMBOL_GPL(kvm_mmu_load);

void kvm_mmu_unload(struct kvm_vcpu *vcpu)
{
	mmu_free_roots(vcpu);
4084
	WARN_ON(VALID_PAGE(vcpu->arch.mmu.root_hpa));
A
Avi Kivity 已提交
4085
}
4086
EXPORT_SYMBOL_GPL(kvm_mmu_unload);
A
Avi Kivity 已提交
4087

4088
static void mmu_pte_write_new_pte(struct kvm_vcpu *vcpu,
4089 4090
				  struct kvm_mmu_page *sp, u64 *spte,
				  const void *new)
4091
{
4092
	if (sp->role.level != PT_PAGE_TABLE_LEVEL) {
4093 4094
		++vcpu->kvm->stat.mmu_pde_zapped;
		return;
4095
        }
4096

A
Avi Kivity 已提交
4097
	++vcpu->kvm->stat.mmu_pte_updated;
4098
	vcpu->arch.mmu.update_pte(vcpu, sp, spte, new);
4099 4100
}

4101 4102 4103 4104 4105 4106 4107 4108
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;
4109 4110
	old ^= shadow_nx_mask;
	new ^= shadow_nx_mask;
4111 4112 4113
	return (old & ~new & PT64_PERM_MASK) != 0;
}

4114 4115
static void mmu_pte_write_flush_tlb(struct kvm_vcpu *vcpu, bool zap_page,
				    bool remote_flush, bool local_flush)
4116
{
4117 4118 4119 4120
	if (zap_page)
		return;

	if (remote_flush)
4121
		kvm_flush_remote_tlbs(vcpu->kvm);
4122
	else if (local_flush)
4123
		kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
4124 4125
}

4126 4127
static u64 mmu_pte_write_fetch_gpte(struct kvm_vcpu *vcpu, gpa_t *gpa,
				    const u8 *new, int *bytes)
4128
{
4129 4130
	u64 gentry;
	int r;
4131 4132 4133

	/*
	 * Assume that the pte write on a page table of the same type
4134 4135
	 * as the current vcpu paging mode since we update the sptes only
	 * when they have the same mode.
4136
	 */
4137
	if (is_pae(vcpu) && *bytes == 4) {
4138
		/* Handle a 32-bit guest writing two halves of a 64-bit gpte */
4139 4140
		*gpa &= ~(gpa_t)7;
		*bytes = 8;
4141
		r = kvm_vcpu_read_guest(vcpu, *gpa, &gentry, 8);
4142 4143
		if (r)
			gentry = 0;
4144 4145 4146
		new = (const u8 *)&gentry;
	}

4147
	switch (*bytes) {
4148 4149 4150 4151 4152 4153 4154 4155 4156
	case 4:
		gentry = *(const u32 *)new;
		break;
	case 8:
		gentry = *(const u64 *)new;
		break;
	default:
		gentry = 0;
		break;
4157 4158
	}

4159 4160 4161 4162 4163 4164 4165
	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.
 */
4166
static bool detect_write_flooding(struct kvm_mmu_page *sp)
4167
{
4168 4169 4170 4171
	/*
	 * Skip write-flooding detected for the sp whose level is 1, because
	 * it can become unsync, then the guest page is not write-protected.
	 */
4172
	if (sp->role.level == PT_PAGE_TABLE_LEVEL)
4173
		return false;
4174

4175
	return ++sp->write_flooding_count >= 3;
4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191
}

/*
 * 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;
4192 4193 4194 4195 4196 4197 4198 4199

	/*
	 * 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;

4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244
	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;
4245
	bool remote_flush, local_flush, zap_page;
4246 4247 4248 4249 4250 4251 4252
	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;
4253
	mask.smm = 1;
4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276

	/*
	 * 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;
4277
	kvm_mmu_audit(vcpu, AUDIT_PRE_PTE_WRITE);
4278

4279
	for_each_gfn_indirect_valid_sp(vcpu->kvm, sp, gfn) {
4280
		if (detect_write_misaligned(sp, gpa, bytes) ||
4281
		      detect_write_flooding(sp)) {
4282
			zap_page |= !!kvm_mmu_prepare_zap_page(vcpu->kvm, sp,
4283
						     &invalid_list);
A
Avi Kivity 已提交
4284
			++vcpu->kvm->stat.mmu_flooded;
4285 4286
			continue;
		}
4287 4288 4289 4290 4291

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

4292
		local_flush = true;
4293
		while (npte--) {
4294
			entry = *spte;
4295
			mmu_page_zap_pte(vcpu->kvm, sp, spte);
4296 4297
			if (gentry &&
			      !((sp->role.word ^ vcpu->arch.mmu.base_role.word)
4298
			      & mask.word) && rmap_can_add(vcpu))
4299
				mmu_pte_write_new_pte(vcpu, sp, spte, &gentry);
G
Gleb Natapov 已提交
4300
			if (need_remote_flush(entry, *spte))
4301
				remote_flush = true;
4302
			++spte;
4303 4304
		}
	}
4305
	mmu_pte_write_flush_tlb(vcpu, zap_page, remote_flush, local_flush);
4306
	kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
4307
	kvm_mmu_audit(vcpu, AUDIT_POST_PTE_WRITE);
4308
	spin_unlock(&vcpu->kvm->mmu_lock);
4309 4310
}

4311 4312
int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva)
{
4313 4314
	gpa_t gpa;
	int r;
4315

4316
	if (vcpu->arch.mmu.direct_map)
4317 4318
		return 0;

4319
	gpa = kvm_mmu_gva_to_gpa_read(vcpu, gva, NULL);
4320 4321

	r = kvm_mmu_unprotect_page(vcpu->kvm, gpa >> PAGE_SHIFT);
4322

4323
	return r;
4324
}
4325
EXPORT_SYMBOL_GPL(kvm_mmu_unprotect_page_virt);
4326

4327
static void make_mmu_pages_available(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4328
{
4329
	LIST_HEAD(invalid_list);
4330

4331 4332 4333
	if (likely(kvm_mmu_available_pages(vcpu->kvm) >= KVM_MIN_FREE_MMU_PAGES))
		return;

4334 4335 4336
	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
4337

A
Avi Kivity 已提交
4338
		++vcpu->kvm->stat.mmu_recycled;
A
Avi Kivity 已提交
4339
	}
4340
	kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
A
Avi Kivity 已提交
4341 4342
}

4343 4344
int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t cr2, u32 error_code,
		       void *insn, int insn_len)
4345
{
4346
	int r, emulation_type = EMULTYPE_RETRY;
4347
	enum emulation_result er;
4348
	bool direct = vcpu->arch.mmu.direct_map || mmu_is_nested(vcpu);
4349

4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361
	if (unlikely(error_code & PFERR_RSVD_MASK)) {
		r = handle_mmio_page_fault(vcpu, cr2, direct);
		if (r == RET_MMIO_PF_EMULATE) {
			emulation_type = 0;
			goto emulate;
		}
		if (r == RET_MMIO_PF_RETRY)
			return 1;
		if (r < 0)
			return r;
	}

G
Gleb Natapov 已提交
4362
	r = vcpu->arch.mmu.page_fault(vcpu, cr2, error_code, false);
4363
	if (r < 0)
4364 4365 4366
		return r;
	if (!r)
		return 1;
4367

4368
	if (mmio_info_in_cache(vcpu, cr2, direct))
4369
		emulation_type = 0;
4370
emulate:
4371
	er = x86_emulate_instruction(vcpu, cr2, emulation_type, insn, insn_len);
4372 4373 4374 4375

	switch (er) {
	case EMULATE_DONE:
		return 1;
P
Paolo Bonzini 已提交
4376
	case EMULATE_USER_EXIT:
4377
		++vcpu->stat.mmio_exits;
4378
		/* fall through */
4379
	case EMULATE_FAIL:
4380
		return 0;
4381 4382 4383 4384 4385 4386
	default:
		BUG();
	}
}
EXPORT_SYMBOL_GPL(kvm_mmu_page_fault);

M
Marcelo Tosatti 已提交
4387 4388 4389
void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
{
	vcpu->arch.mmu.invlpg(vcpu, gva);
4390
	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
M
Marcelo Tosatti 已提交
4391 4392 4393 4394
	++vcpu->stat.invlpg;
}
EXPORT_SYMBOL_GPL(kvm_mmu_invlpg);

4395 4396 4397 4398 4399 4400
void kvm_enable_tdp(void)
{
	tdp_enabled = true;
}
EXPORT_SYMBOL_GPL(kvm_enable_tdp);

4401 4402 4403 4404 4405 4406
void kvm_disable_tdp(void)
{
	tdp_enabled = false;
}
EXPORT_SYMBOL_GPL(kvm_disable_tdp);

A
Avi Kivity 已提交
4407 4408
static void free_mmu_pages(struct kvm_vcpu *vcpu)
{
4409
	free_page((unsigned long)vcpu->arch.mmu.pae_root);
4410 4411
	if (vcpu->arch.mmu.lm_root != NULL)
		free_page((unsigned long)vcpu->arch.mmu.lm_root);
A
Avi Kivity 已提交
4412 4413 4414 4415
}

static int alloc_mmu_pages(struct kvm_vcpu *vcpu)
{
4416
	struct page *page;
A
Avi Kivity 已提交
4417 4418
	int i;

4419 4420 4421 4422 4423 4424 4425
	/*
	 * 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)
4426 4427
		return -ENOMEM;

4428
	vcpu->arch.mmu.pae_root = page_address(page);
4429
	for (i = 0; i < 4; ++i)
4430
		vcpu->arch.mmu.pae_root[i] = INVALID_PAGE;
4431

A
Avi Kivity 已提交
4432 4433 4434
	return 0;
}

4435
int kvm_mmu_create(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4436
{
4437 4438 4439 4440
	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
4441

4442 4443
	return alloc_mmu_pages(vcpu);
}
A
Avi Kivity 已提交
4444

4445
void kvm_mmu_setup(struct kvm_vcpu *vcpu)
4446
{
4447
	MMU_WARN_ON(VALID_PAGE(vcpu->arch.mmu.root_hpa));
4448

4449
	init_kvm_mmu(vcpu);
A
Avi Kivity 已提交
4450 4451
}

4452
/* The return value indicates if tlb flush on all vcpus is needed. */
4453
typedef bool (*slot_level_handler) (struct kvm *kvm, struct kvm_rmap_head *rmap_head);
4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520

/* 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
4521 4522 4523 4524
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;
4525
	int i;
X
Xiao Guangrong 已提交
4526 4527

	spin_lock(&kvm->mmu_lock);
4528 4529 4530 4531 4532 4533 4534 4535 4536
	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
4537

4538 4539 4540 4541
			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
4542 4543 4544 4545 4546
	}

	spin_unlock(&kvm->mmu_lock);
}

4547 4548
static bool slot_rmap_write_protect(struct kvm *kvm,
				    struct kvm_rmap_head *rmap_head)
4549
{
4550
	return __rmap_write_protect(kvm, rmap_head, false);
4551 4552
}

4553 4554
void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
				      struct kvm_memory_slot *memslot)
A
Avi Kivity 已提交
4555
{
4556
	bool flush;
A
Avi Kivity 已提交
4557

4558
	spin_lock(&kvm->mmu_lock);
4559 4560
	flush = slot_handle_all_level(kvm, memslot, slot_rmap_write_protect,
				      false);
4561
	spin_unlock(&kvm->mmu_lock);
4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580

	/*
	 * 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.
	 */
4581 4582
	if (flush)
		kvm_flush_remote_tlbs(kvm);
A
Avi Kivity 已提交
4583
}
4584

4585
static bool kvm_mmu_zap_collapsible_spte(struct kvm *kvm,
4586
					 struct kvm_rmap_head *rmap_head)
4587 4588 4589 4590
{
	u64 *sptep;
	struct rmap_iterator iter;
	int need_tlb_flush = 0;
D
Dan Williams 已提交
4591
	kvm_pfn_t pfn;
4592 4593
	struct kvm_mmu_page *sp;

4594
restart:
4595
	for_each_rmap_spte(rmap_head, &iter, sptep) {
4596 4597 4598 4599
		sp = page_header(__pa(sptep));
		pfn = spte_to_pfn(*sptep);

		/*
4600 4601 4602 4603 4604
		 * 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.
4605 4606 4607 4608 4609 4610
		 */
		if (sp->role.direct &&
			!kvm_is_reserved_pfn(pfn) &&
			PageTransCompound(pfn_to_page(pfn))) {
			drop_spte(kvm, sptep);
			need_tlb_flush = 1;
4611 4612
			goto restart;
		}
4613 4614 4615 4616 4617 4618
	}

	return need_tlb_flush;
}

void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
4619
				   const struct kvm_memory_slot *memslot)
4620
{
4621
	/* FIXME: const-ify all uses of struct kvm_memory_slot.  */
4622
	spin_lock(&kvm->mmu_lock);
4623 4624
	slot_handle_leaf(kvm, (struct kvm_memory_slot *)memslot,
			 kvm_mmu_zap_collapsible_spte, true);
4625 4626 4627
	spin_unlock(&kvm->mmu_lock);
}

4628 4629 4630
void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
				   struct kvm_memory_slot *memslot)
{
4631
	bool flush;
4632 4633

	spin_lock(&kvm->mmu_lock);
4634
	flush = slot_handle_leaf(kvm, memslot, __rmap_clear_dirty, false);
4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652
	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)
{
4653
	bool flush;
4654 4655

	spin_lock(&kvm->mmu_lock);
4656 4657
	flush = slot_handle_large_level(kvm, memslot, slot_rmap_write_protect,
					false);
4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670
	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)
{
4671
	bool flush;
4672 4673

	spin_lock(&kvm->mmu_lock);
4674
	flush = slot_handle_all_level(kvm, memslot, __rmap_set_dirty, false);
4675 4676 4677 4678 4679 4680 4681 4682 4683 4684
	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
4685
#define BATCH_ZAP_PAGES	10
4686 4687 4688
static void kvm_zap_obsolete_pages(struct kvm *kvm)
{
	struct kvm_mmu_page *sp, *node;
X
Xiao Guangrong 已提交
4689
	int batch = 0;
4690 4691 4692 4693

restart:
	list_for_each_entry_safe_reverse(sp, node,
	      &kvm->arch.active_mmu_pages, link) {
X
Xiao Guangrong 已提交
4694 4695
		int ret;

4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710
		/*
		 * 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;

4711 4712 4713 4714
		/*
		 * Need not flush tlb since we only zap the sp with invalid
		 * generation number.
		 */
X
Xiao Guangrong 已提交
4715
		if (batch >= BATCH_ZAP_PAGES &&
4716
		      cond_resched_lock(&kvm->mmu_lock)) {
X
Xiao Guangrong 已提交
4717
			batch = 0;
4718 4719 4720
			goto restart;
		}

4721 4722
		ret = kvm_mmu_prepare_zap_page(kvm, sp,
				&kvm->arch.zapped_obsolete_pages);
X
Xiao Guangrong 已提交
4723 4724 4725
		batch += ret;

		if (ret)
4726 4727 4728
			goto restart;
	}

4729 4730 4731 4732
	/*
	 * Should flush tlb before free page tables since lockless-walking
	 * may use the pages.
	 */
4733
	kvm_mmu_commit_zap_page(kvm, &kvm->arch.zapped_obsolete_pages);
4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747
}

/*
 * 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);
4748
	trace_kvm_mmu_invalidate_zap_all_pages(kvm);
4749 4750
	kvm->arch.mmu_valid_gen++;

4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761
	/*
	 * 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);

4762 4763 4764 4765
	kvm_zap_obsolete_pages(kvm);
	spin_unlock(&kvm->mmu_lock);
}

4766 4767 4768 4769 4770
static bool kvm_has_zapped_obsolete_pages(struct kvm *kvm)
{
	return unlikely(!list_empty_careful(&kvm->arch.zapped_obsolete_pages));
}

4771
void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, struct kvm_memslots *slots)
4772 4773 4774 4775 4776
{
	/*
	 * The very rare case: if the generation-number is round,
	 * zap all shadow pages.
	 */
4777
	if (unlikely((slots->generation & MMIO_GEN_MASK) == 0)) {
4778
		printk_ratelimited(KERN_DEBUG "kvm: zapping shadow pages for mmio generation wraparound\n");
4779
		kvm_mmu_invalidate_zap_all_pages(kvm);
4780
	}
4781 4782
}

4783 4784
static unsigned long
mmu_shrink_scan(struct shrinker *shrink, struct shrink_control *sc)
4785 4786
{
	struct kvm *kvm;
4787
	int nr_to_scan = sc->nr_to_scan;
4788
	unsigned long freed = 0;
4789

4790
	spin_lock(&kvm_lock);
4791 4792

	list_for_each_entry(kvm, &vm_list, vm_list) {
4793
		int idx;
4794
		LIST_HEAD(invalid_list);
4795

4796 4797 4798 4799 4800 4801 4802 4803
		/*
		 * 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;
4804 4805 4806 4807 4808 4809
		/*
		 * 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.
		 */
4810 4811
		if (!kvm->arch.n_used_mmu_pages &&
		      !kvm_has_zapped_obsolete_pages(kvm))
4812 4813
			continue;

4814
		idx = srcu_read_lock(&kvm->srcu);
4815 4816
		spin_lock(&kvm->mmu_lock);

4817 4818 4819 4820 4821 4822
		if (kvm_has_zapped_obsolete_pages(kvm)) {
			kvm_mmu_commit_zap_page(kvm,
			      &kvm->arch.zapped_obsolete_pages);
			goto unlock;
		}

4823 4824
		if (prepare_zap_oldest_mmu_page(kvm, &invalid_list))
			freed++;
4825
		kvm_mmu_commit_zap_page(kvm, &invalid_list);
4826

4827
unlock:
4828
		spin_unlock(&kvm->mmu_lock);
4829
		srcu_read_unlock(&kvm->srcu, idx);
4830

4831 4832 4833 4834 4835
		/*
		 * unfair on small ones
		 * per-vm shrinkers cry out
		 * sadness comes quickly
		 */
4836 4837
		list_move_tail(&kvm->vm_list, &vm_list);
		break;
4838 4839
	}

4840
	spin_unlock(&kvm_lock);
4841 4842 4843 4844 4845 4846
	return freed;
}

static unsigned long
mmu_shrink_count(struct shrinker *shrink, struct shrink_control *sc)
{
4847
	return percpu_counter_read_positive(&kvm_total_used_mmu_pages);
4848 4849 4850
}

static struct shrinker mmu_shrinker = {
4851 4852
	.count_objects = mmu_shrink_count,
	.scan_objects = mmu_shrink_scan,
4853 4854 4855
	.seeks = DEFAULT_SEEKS * 10,
};

I
Ingo Molnar 已提交
4856
static void mmu_destroy_caches(void)
4857
{
4858 4859
	if (pte_list_desc_cache)
		kmem_cache_destroy(pte_list_desc_cache);
4860 4861
	if (mmu_page_header_cache)
		kmem_cache_destroy(mmu_page_header_cache);
4862 4863 4864 4865
}

int kvm_mmu_module_init(void)
{
4866 4867
	pte_list_desc_cache = kmem_cache_create("pte_list_desc",
					    sizeof(struct pte_list_desc),
4868
					    0, 0, NULL);
4869
	if (!pte_list_desc_cache)
4870 4871
		goto nomem;

4872 4873
	mmu_page_header_cache = kmem_cache_create("kvm_mmu_page_header",
						  sizeof(struct kvm_mmu_page),
4874
						  0, 0, NULL);
4875 4876 4877
	if (!mmu_page_header_cache)
		goto nomem;

4878
	if (percpu_counter_init(&kvm_total_used_mmu_pages, 0, GFP_KERNEL))
4879 4880
		goto nomem;

4881 4882
	register_shrinker(&mmu_shrinker);

4883 4884 4885
	return 0;

nomem:
4886
	mmu_destroy_caches();
4887 4888 4889
	return -ENOMEM;
}

4890 4891 4892 4893 4894 4895 4896
/*
 * 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;
4897
	struct kvm_memslots *slots;
4898
	struct kvm_memory_slot *memslot;
4899
	int i;
4900

4901 4902
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
		slots = __kvm_memslots(kvm, i);
4903

4904 4905 4906
		kvm_for_each_memslot(memslot, slots)
			nr_pages += memslot->npages;
	}
4907 4908 4909

	nr_mmu_pages = nr_pages * KVM_PERMILLE_MMU_PAGES / 1000;
	nr_mmu_pages = max(nr_mmu_pages,
4910
			   (unsigned int) KVM_MIN_ALLOC_MMU_PAGES);
4911 4912 4913 4914

	return nr_mmu_pages;
}

4915 4916
void kvm_mmu_destroy(struct kvm_vcpu *vcpu)
{
4917
	kvm_mmu_unload(vcpu);
4918 4919
	free_mmu_pages(vcpu);
	mmu_free_memory_caches(vcpu);
4920 4921 4922 4923 4924 4925 4926
}

void kvm_mmu_module_exit(void)
{
	mmu_destroy_caches();
	percpu_counter_destroy(&kvm_total_used_mmu_pages);
	unregister_shrinker(&mmu_shrinker);
4927 4928
	mmu_audit_disable();
}