mmu.c 124.2 KB
Newer Older
A
Avi Kivity 已提交
1 2 3 4 5 6 7 8 9
/*
 * Kernel-based Virtual Machine driver for Linux
 *
 * This module enables machines with Intel VT-x extensions to run virtual
 * machines without emulation or binary translation.
 *
 * MMU support
 *
 * Copyright (C) 2006 Qumranet, Inc.
N
Nicolas Kaiser 已提交
10
 * Copyright 2010 Red Hat, Inc. and/or its affiliates.
A
Avi Kivity 已提交
11 12 13 14 15 16 17 18 19
 *
 * Authors:
 *   Yaniv Kamay  <yaniv@qumranet.com>
 *   Avi Kivity   <avi@qumranet.com>
 *
 * This work is licensed under the terms of the GNU GPL, version 2.  See
 * the COPYING file in the top-level directory.
 *
 */
A
Avi Kivity 已提交
20

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

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

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

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

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

63
#undef MMU_DEBUG
64 65

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

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

78 79
#define PTE_PREFETCH_NUM		8

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

#define PT64_LEVEL_BITS 9

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

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


#define PT32_LEVEL_BITS 10

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

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

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


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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

	spte &= ~shadow_mmio_mask;

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

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

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

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

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

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

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

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

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

	return false;
}
271

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

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

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

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

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

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

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

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

314
static int is_rmap_spte(u64 pte)
315
{
316
	return is_shadow_present_pte(pte);
317 318
}

319 320 321 322
static int is_last_spte(u64 pte, int level)
{
	if (level == PT_PAGE_TABLE_LEVEL)
		return 1;
323
	if (is_large_pte(pte))
324 325 326 327
		return 1;
	return 0;
}

328
static pfn_t spte_to_pfn(u64 pte)
329
{
330
	return (pte & PT64_BASE_ADDR_MASK) >> PAGE_SHIFT;
331 332
}

333 334 335 336 337 338 339
static gfn_t pse36_gfn_delta(u32 gpte)
{
	int shift = 32 - PT32_DIR_PSE36_SHIFT - PAGE_SHIFT;

	return (gpte & PT32_DIR_PSE36_MASK) << shift;
}

340
#ifdef CONFIG_X86_64
A
Avi Kivity 已提交
341
static void __set_spte(u64 *sptep, u64 spte)
342
{
343
	*sptep = spte;
344 345
}

346
static void __update_clear_spte_fast(u64 *sptep, u64 spte)
347
{
348 349 350 351 352 353 354
	*sptep = spte;
}

static u64 __update_clear_spte_slow(u64 *sptep, u64 spte)
{
	return xchg(sptep, spte);
}
355 356 357 358 359

static u64 __get_spte_lockless(u64 *sptep)
{
	return ACCESS_ONCE(*sptep);
}
360
#else
361 362 363 364 365 366 367
union split_spte {
	struct {
		u32 spte_low;
		u32 spte_high;
	};
	u64 spte;
};
368

369 370 371 372 373 374 375 376 377 378 379 380
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++;
}

381 382 383
static void __set_spte(u64 *sptep, u64 spte)
{
	union split_spte *ssptep, sspte;
384

385 386 387 388 389 390 391 392 393 394 395 396 397
	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;
398 399
}

400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415
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;
416
	count_spte_clear(sptep, spte);
417 418 419 420 421 422 423 424 425 426 427
}

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);
428 429
	orig.spte_high = ssptep->spte_high;
	ssptep->spte_high = sspte.spte_high;
430
	count_spte_clear(sptep, spte);
431 432 433

	return orig.spte;
}
434 435 436 437

/*
 * The idea using the light way get the spte on x86_32 guest is from
 * gup_get_pte(arch/x86/mm/gup.c).
438 439 440 441 442 443 444 445 446 447 448 449 450 451
 *
 * 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.
452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474
 */
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;
}
475 476
#endif

477 478
static bool spte_is_locklessly_modifiable(u64 spte)
{
479 480
	return (spte & (SPTE_HOST_WRITEABLE | SPTE_MMU_WRITEABLE)) ==
		(SPTE_HOST_WRITEABLE | SPTE_MMU_WRITEABLE);
481 482
}

483 484
static bool spte_has_volatile_bits(u64 spte)
{
485 486 487 488 489 490 491 492 493
	/*
	 * 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;

494 495 496 497 498 499
	if (!shadow_accessed_mask)
		return false;

	if (!is_shadow_present_pte(spte))
		return false;

500 501
	if ((spte & shadow_accessed_mask) &&
	      (!is_writable_pte(spte) || (spte & shadow_dirty_mask)))
502 503 504 505 506
		return false;

	return true;
}

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

512 513 514 515 516
static bool spte_is_bit_changed(u64 old_spte, u64 new_spte, u64 bit_mask)
{
	return (old_spte & bit_mask) != (new_spte & bit_mask);
}

517 518 519 520 521 522 523 524 525 526 527 528 529 530
/* 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.
531 532 533 534 535 536
 *
 * 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.
537
 */
538
static bool mmu_spte_update(u64 *sptep, u64 new_spte)
539
{
540
	u64 old_spte = *sptep;
541
	bool ret = false;
542 543

	WARN_ON(!is_rmap_spte(new_spte));
544

545 546 547 548
	if (!is_shadow_present_pte(old_spte)) {
		mmu_spte_set(sptep, new_spte);
		return ret;
	}
549

550
	if (!spte_has_volatile_bits(old_spte))
551
		__update_clear_spte_fast(sptep, new_spte);
552
	else
553
		old_spte = __update_clear_spte_slow(sptep, new_spte);
554

555 556 557 558 559
	/*
	 * For the spte updated out of mmu-lock is safe, since
	 * we always atomicly update it, see the comments in
	 * spte_has_volatile_bits().
	 */
560 561
	if (spte_is_locklessly_modifiable(old_spte) &&
	      !is_writable_pte(new_spte))
562 563
		ret = true;

564
	if (!shadow_accessed_mask)
565
		return ret;
566

567 568 569 570 571 572 573 574
	/*
	 * 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;

575 576 577 578
	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));
579 580

	return ret;
581 582
}

583 584 585 586 587 588 589 590 591 592 593
/*
 * Rules for using mmu_spte_clear_track_bits:
 * It sets the sptep from present to nonpresent, and track the
 * state bits, it is used to clear the last level sptep.
 */
static int mmu_spte_clear_track_bits(u64 *sptep)
{
	pfn_t pfn;
	u64 old_spte = *sptep;

	if (!spte_has_volatile_bits(old_spte))
594
		__update_clear_spte_fast(sptep, 0ull);
595
	else
596
		old_spte = __update_clear_spte_slow(sptep, 0ull);
597 598 599 600 601

	if (!is_rmap_spte(old_spte))
		return 0;

	pfn = spte_to_pfn(old_spte);
602 603 604 605 606 607

	/*
	 * KVM does not hold the refcount of the page used by
	 * kvm mmu, before reclaiming the page, we should
	 * unmap it from mmu first.
	 */
608
	WARN_ON(!kvm_is_reserved_pfn(pfn) && !page_count(pfn_to_page(pfn)));
609

610 611 612 613 614 615 616 617 618 619 620 621 622 623
	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)
{
624
	__update_clear_spte_fast(sptep, 0ull);
625 626
}

627 628 629 630 631 632 633
static u64 mmu_spte_get_lockless(u64 *sptep)
{
	return __get_spte_lockless(sptep);
}

static void walk_shadow_page_lockless_begin(struct kvm_vcpu *vcpu)
{
634 635 636 637 638 639 640 641 642 643 644
	/*
	 * 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();
645 646 647 648
}

static void walk_shadow_page_lockless_end(struct kvm_vcpu *vcpu)
{
649 650 651 652 653 654 655 656
	/*
	 * 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();
657 658
}

659
static int mmu_topup_memory_cache(struct kvm_mmu_memory_cache *cache,
660
				  struct kmem_cache *base_cache, int min)
661 662 663 664
{
	void *obj;

	if (cache->nobjs >= min)
665
		return 0;
666
	while (cache->nobjs < ARRAY_SIZE(cache->objects)) {
667
		obj = kmem_cache_zalloc(base_cache, GFP_KERNEL);
668
		if (!obj)
669
			return -ENOMEM;
670 671
		cache->objects[cache->nobjs++] = obj;
	}
672
	return 0;
673 674
}

675 676 677 678 679
static int mmu_memory_cache_free_objects(struct kvm_mmu_memory_cache *cache)
{
	return cache->nobjs;
}

680 681
static void mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc,
				  struct kmem_cache *cache)
682 683
{
	while (mc->nobjs)
684
		kmem_cache_free(cache, mc->objects[--mc->nobjs]);
685 686
}

A
Avi Kivity 已提交
687
static int mmu_topup_memory_cache_page(struct kvm_mmu_memory_cache *cache,
688
				       int min)
A
Avi Kivity 已提交
689
{
690
	void *page;
A
Avi Kivity 已提交
691 692 693 694

	if (cache->nobjs >= min)
		return 0;
	while (cache->nobjs < ARRAY_SIZE(cache->objects)) {
695
		page = (void *)__get_free_page(GFP_KERNEL);
A
Avi Kivity 已提交
696 697
		if (!page)
			return -ENOMEM;
698
		cache->objects[cache->nobjs++] = page;
A
Avi Kivity 已提交
699 700 701 702 703 704 705
	}
	return 0;
}

static void mmu_free_memory_cache_page(struct kvm_mmu_memory_cache *mc)
{
	while (mc->nobjs)
706
		free_page((unsigned long)mc->objects[--mc->nobjs]);
A
Avi Kivity 已提交
707 708
}

709
static int mmu_topup_memory_caches(struct kvm_vcpu *vcpu)
710
{
711 712
	int r;

713
	r = mmu_topup_memory_cache(&vcpu->arch.mmu_pte_list_desc_cache,
714
				   pte_list_desc_cache, 8 + PTE_PREFETCH_NUM);
715 716
	if (r)
		goto out;
717
	r = mmu_topup_memory_cache_page(&vcpu->arch.mmu_page_cache, 8);
718 719
	if (r)
		goto out;
720
	r = mmu_topup_memory_cache(&vcpu->arch.mmu_page_header_cache,
721
				   mmu_page_header_cache, 4);
722 723
out:
	return r;
724 725 726 727
}

static void mmu_free_memory_caches(struct kvm_vcpu *vcpu)
{
728 729
	mmu_free_memory_cache(&vcpu->arch.mmu_pte_list_desc_cache,
				pte_list_desc_cache);
730
	mmu_free_memory_cache_page(&vcpu->arch.mmu_page_cache);
731 732
	mmu_free_memory_cache(&vcpu->arch.mmu_page_header_cache,
				mmu_page_header_cache);
733 734
}

735
static void *mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc)
736 737 738 739 740 741 742 743
{
	void *p;

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

744
static struct pte_list_desc *mmu_alloc_pte_list_desc(struct kvm_vcpu *vcpu)
745
{
746
	return mmu_memory_cache_alloc(&vcpu->arch.mmu_pte_list_desc_cache);
747 748
}

749
static void mmu_free_pte_list_desc(struct pte_list_desc *pte_list_desc)
750
{
751
	kmem_cache_free(pte_list_desc_cache, pte_list_desc);
752 753
}

754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769
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
770
/*
771 772
 * 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
773
 */
774 775 776
static struct kvm_lpage_info *lpage_info_slot(gfn_t gfn,
					      struct kvm_memory_slot *slot,
					      int level)
M
Marcelo Tosatti 已提交
777 778 779
{
	unsigned long idx;

780
	idx = gfn_to_index(gfn, slot->base_gfn, level);
781
	return &slot->arch.lpage_info[level - 2][idx];
M
Marcelo Tosatti 已提交
782 783
}

784
static void account_shadowed(struct kvm *kvm, struct kvm_mmu_page *sp)
M
Marcelo Tosatti 已提交
785
{
786
	struct kvm_memslots *slots;
787
	struct kvm_memory_slot *slot;
788
	struct kvm_lpage_info *linfo;
789
	gfn_t gfn;
790
	int i;
M
Marcelo Tosatti 已提交
791

792
	gfn = sp->gfn;
793 794
	slots = kvm_memslots_for_spte_role(kvm, sp->role);
	slot = __gfn_to_memslot(slots, gfn);
795
	for (i = PT_DIRECTORY_LEVEL; i <= PT_MAX_HUGEPAGE_LEVEL; ++i) {
796 797
		linfo = lpage_info_slot(gfn, slot, i);
		linfo->write_count += 1;
798
	}
799
	kvm->arch.indirect_shadow_pages++;
M
Marcelo Tosatti 已提交
800 801
}

802
static void unaccount_shadowed(struct kvm *kvm, struct kvm_mmu_page *sp)
M
Marcelo Tosatti 已提交
803
{
804
	struct kvm_memslots *slots;
805
	struct kvm_memory_slot *slot;
806
	struct kvm_lpage_info *linfo;
807
	gfn_t gfn;
808
	int i;
M
Marcelo Tosatti 已提交
809

810
	gfn = sp->gfn;
811 812
	slots = kvm_memslots_for_spte_role(kvm, sp->role);
	slot = __gfn_to_memslot(slots, gfn);
813
	for (i = PT_DIRECTORY_LEVEL; i <= PT_MAX_HUGEPAGE_LEVEL; ++i) {
814 815 816
		linfo = lpage_info_slot(gfn, slot, i);
		linfo->write_count -= 1;
		WARN_ON(linfo->write_count < 0);
817
	}
818
	kvm->arch.indirect_shadow_pages--;
M
Marcelo Tosatti 已提交
819 820
}

821 822
static int __has_wrprotected_page(gfn_t gfn, int level,
				  struct kvm_memory_slot *slot)
M
Marcelo Tosatti 已提交
823
{
824
	struct kvm_lpage_info *linfo;
M
Marcelo Tosatti 已提交
825 826

	if (slot) {
827 828
		linfo = lpage_info_slot(gfn, slot, level);
		return linfo->write_count;
M
Marcelo Tosatti 已提交
829 830 831 832 833
	}

	return 1;
}

834 835 836 837 838 839 840 841
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);
}

842
static int host_mapping_level(struct kvm *kvm, gfn_t gfn)
M
Marcelo Tosatti 已提交
843
{
J
Joerg Roedel 已提交
844
	unsigned long page_size;
845
	int i, ret = 0;
M
Marcelo Tosatti 已提交
846

J
Joerg Roedel 已提交
847
	page_size = kvm_host_page_size(kvm, gfn);
M
Marcelo Tosatti 已提交
848

849
	for (i = PT_PAGE_TABLE_LEVEL; i <= PT_MAX_HUGEPAGE_LEVEL; ++i) {
850 851 852 853 854 855
		if (page_size >= KVM_HPAGE_SIZE(i))
			ret = i;
		else
			break;
	}

856
	return ret;
M
Marcelo Tosatti 已提交
857 858
}

859 860 861 862 863 864 865 866 867 868 869
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;
}

870 871 872
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
873 874
{
	struct kvm_memory_slot *slot;
875

876
	slot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);
877
	if (!memslot_valid_for_gpte(slot, no_dirty_log))
878 879 880 881 882
		slot = NULL;

	return slot;
}

883 884
static int mapping_level(struct kvm_vcpu *vcpu, gfn_t large_gfn,
			 bool *force_pt_level)
885 886
{
	int host_level, level, max_level;
887 888
	struct kvm_memory_slot *slot;

889 890
	if (unlikely(*force_pt_level))
		return PT_PAGE_TABLE_LEVEL;
M
Marcelo Tosatti 已提交
891

892 893
	slot = kvm_vcpu_gfn_to_memslot(vcpu, large_gfn);
	*force_pt_level = !memslot_valid_for_gpte(slot, true);
894 895 896
	if (unlikely(*force_pt_level))
		return PT_PAGE_TABLE_LEVEL;

897 898 899 900 901
	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
902
	max_level = min(kvm_x86_ops->get_lpage_level(), host_level);
903 904

	for (level = PT_DIRECTORY_LEVEL; level <= max_level; ++level)
905
		if (__has_wrprotected_page(large_gfn, level, slot))
906 907 908
			break;

	return level - 1;
M
Marcelo Tosatti 已提交
909 910
}

911
/*
912
 * Pte mapping structures:
913
 *
914
 * If pte_list bit zero is zero, then pte_list point to the spte.
915
 *
916 917
 * If pte_list bit zero is one, (then pte_list & ~1) points to a struct
 * pte_list_desc containing more mappings.
918
 *
919
 * Returns the number of pte entries before the spte was added or zero if
920 921
 * the spte was not added.
 *
922
 */
923 924
static int pte_list_add(struct kvm_vcpu *vcpu, u64 *spte,
			unsigned long *pte_list)
925
{
926
	struct pte_list_desc *desc;
927
	int i, count = 0;
928

929 930 931 932 933 934 935
	if (!*pte_list) {
		rmap_printk("pte_list_add: %p %llx 0->1\n", spte, *spte);
		*pte_list = (unsigned long)spte;
	} else if (!(*pte_list & 1)) {
		rmap_printk("pte_list_add: %p %llx 1->many\n", spte, *spte);
		desc = mmu_alloc_pte_list_desc(vcpu);
		desc->sptes[0] = (u64 *)*pte_list;
A
Avi Kivity 已提交
936
		desc->sptes[1] = spte;
937
		*pte_list = (unsigned long)desc | 1;
938
		++count;
939
	} else {
940 941 942
		rmap_printk("pte_list_add: %p %llx many->many\n", spte, *spte);
		desc = (struct pte_list_desc *)(*pte_list & ~1ul);
		while (desc->sptes[PTE_LIST_EXT-1] && desc->more) {
943
			desc = desc->more;
944
			count += PTE_LIST_EXT;
945
		}
946 947
		if (desc->sptes[PTE_LIST_EXT-1]) {
			desc->more = mmu_alloc_pte_list_desc(vcpu);
948 949
			desc = desc->more;
		}
A
Avi Kivity 已提交
950
		for (i = 0; desc->sptes[i]; ++i)
951
			++count;
A
Avi Kivity 已提交
952
		desc->sptes[i] = spte;
953
	}
954
	return count;
955 956
}

957 958 959
static void
pte_list_desc_remove_entry(unsigned long *pte_list, struct pte_list_desc *desc,
			   int i, struct pte_list_desc *prev_desc)
960 961 962
{
	int j;

963
	for (j = PTE_LIST_EXT - 1; !desc->sptes[j] && j > i; --j)
964
		;
A
Avi Kivity 已提交
965 966
	desc->sptes[i] = desc->sptes[j];
	desc->sptes[j] = NULL;
967 968 969
	if (j != 0)
		return;
	if (!prev_desc && !desc->more)
970
		*pte_list = (unsigned long)desc->sptes[0];
971 972 973 974
	else
		if (prev_desc)
			prev_desc->more = desc->more;
		else
975 976
			*pte_list = (unsigned long)desc->more | 1;
	mmu_free_pte_list_desc(desc);
977 978
}

979
static void pte_list_remove(u64 *spte, unsigned long *pte_list)
980
{
981 982
	struct pte_list_desc *desc;
	struct pte_list_desc *prev_desc;
983 984
	int i;

985 986
	if (!*pte_list) {
		printk(KERN_ERR "pte_list_remove: %p 0->BUG\n", spte);
987
		BUG();
988 989 990 991
	} else if (!(*pte_list & 1)) {
		rmap_printk("pte_list_remove:  %p 1->0\n", spte);
		if ((u64 *)*pte_list != spte) {
			printk(KERN_ERR "pte_list_remove:  %p 1->BUG\n", spte);
992 993
			BUG();
		}
994
		*pte_list = 0;
995
	} else {
996 997
		rmap_printk("pte_list_remove:  %p many->many\n", spte);
		desc = (struct pte_list_desc *)(*pte_list & ~1ul);
998 999
		prev_desc = NULL;
		while (desc) {
1000
			for (i = 0; i < PTE_LIST_EXT && desc->sptes[i]; ++i)
A
Avi Kivity 已提交
1001
				if (desc->sptes[i] == spte) {
1002
					pte_list_desc_remove_entry(pte_list,
1003
							       desc, i,
1004 1005 1006 1007 1008 1009
							       prev_desc);
					return;
				}
			prev_desc = desc;
			desc = desc->more;
		}
1010
		pr_err("pte_list_remove: %p many->many\n", spte);
1011 1012 1013 1014
		BUG();
	}
}

1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034
typedef void (*pte_list_walk_fn) (u64 *spte);
static void pte_list_walk(unsigned long *pte_list, pte_list_walk_fn fn)
{
	struct pte_list_desc *desc;
	int i;

	if (!*pte_list)
		return;

	if (!(*pte_list & 1))
		return fn((u64 *)*pte_list);

	desc = (struct pte_list_desc *)(*pte_list & ~1ul);
	while (desc) {
		for (i = 0; i < PTE_LIST_EXT && desc->sptes[i]; ++i)
			fn(desc->sptes[i]);
		desc = desc->more;
	}
}

1035
static unsigned long *__gfn_to_rmap(gfn_t gfn, int level,
1036
				    struct kvm_memory_slot *slot)
1037
{
1038
	unsigned long idx;
1039

1040
	idx = gfn_to_index(gfn, slot->base_gfn, level);
1041
	return &slot->arch.rmap[level - PT_PAGE_TABLE_LEVEL][idx];
1042 1043
}

1044 1045 1046
/*
 * Take gfn and return the reverse mapping to it.
 */
1047
static unsigned long *gfn_to_rmap(struct kvm *kvm, gfn_t gfn, struct kvm_mmu_page *sp)
1048
{
1049
	struct kvm_memslots *slots;
1050 1051
	struct kvm_memory_slot *slot;

1052 1053
	slots = kvm_memslots_for_spte_role(kvm, sp->role);
	slot = __gfn_to_memslot(slots, gfn);
1054
	return __gfn_to_rmap(gfn, sp->role.level, slot);
1055 1056
}

1057 1058 1059 1060 1061 1062 1063 1064
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);
}

1065 1066 1067 1068 1069 1070 1071
static int rmap_add(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn)
{
	struct kvm_mmu_page *sp;
	unsigned long *rmapp;

	sp = page_header(__pa(spte));
	kvm_mmu_page_set_gfn(sp, spte - sp->spt, gfn);
1072
	rmapp = gfn_to_rmap(vcpu->kvm, gfn, sp);
1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083
	return pte_list_add(vcpu, spte, rmapp);
}

static void rmap_remove(struct kvm *kvm, u64 *spte)
{
	struct kvm_mmu_page *sp;
	gfn_t gfn;
	unsigned long *rmapp;

	sp = page_header(__pa(spte));
	gfn = kvm_mmu_page_get_gfn(sp, spte - sp->spt);
1084
	rmapp = gfn_to_rmap(kvm, gfn, sp);
1085 1086 1087
	pte_list_remove(spte, rmapp);
}

1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148
/*
 * 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.
 */
static u64 *rmap_get_first(unsigned long rmap, struct rmap_iterator *iter)
{
	if (!rmap)
		return NULL;

	if (!(rmap & 1)) {
		iter->desc = NULL;
		return (u64 *)rmap;
	}

	iter->desc = (struct pte_list_desc *)(rmap & ~1ul);
	iter->pos = 0;
	return iter->desc->sptes[iter->pos];
}

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

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

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

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

	return NULL;
}

1149 1150 1151 1152 1153
#define for_each_rmap_spte(_rmap_, _iter_, _spte_)			    \
	   for (_spte_ = rmap_get_first(*_rmap_, _iter_);		    \
		_spte_ && ({BUG_ON(!is_shadow_present_pte(*_spte_)); 1;});  \
			_spte_ = rmap_get_next(_iter_))

1154
static void drop_spte(struct kvm *kvm, u64 *sptep)
1155
{
1156
	if (mmu_spte_clear_track_bits(sptep))
1157
		rmap_remove(kvm, sptep);
A
Avi Kivity 已提交
1158 1159
}

1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180

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);
}

/*
1181
 * Write-protect on the specified @sptep, @pt_protect indicates whether
1182
 * spte write-protection is caused by protecting shadow page table.
1183
 *
T
Tiejun Chen 已提交
1184
 * Note: write protection is difference between dirty logging and spte
1185 1186 1187 1188 1189
 * 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.
1190
 *
1191
 * Return true if tlb need be flushed.
1192
 */
1193
static bool spte_write_protect(struct kvm *kvm, u64 *sptep, bool pt_protect)
1194 1195 1196
{
	u64 spte = *sptep;

1197 1198
	if (!is_writable_pte(spte) &&
	      !(pt_protect && spte_is_locklessly_modifiable(spte)))
1199 1200 1201 1202
		return false;

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

1203 1204
	if (pt_protect)
		spte &= ~SPTE_MMU_WRITEABLE;
1205
	spte = spte & ~PT_WRITABLE_MASK;
1206

1207
	return mmu_spte_update(sptep, spte);
1208 1209
}

1210
static bool __rmap_write_protect(struct kvm *kvm, unsigned long *rmapp,
1211
				 bool pt_protect)
1212
{
1213 1214
	u64 *sptep;
	struct rmap_iterator iter;
1215
	bool flush = false;
1216

1217
	for_each_rmap_spte(rmapp, &iter, sptep)
1218
		flush |= spte_write_protect(kvm, sptep, pt_protect);
1219

1220
	return flush;
1221 1222
}

1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239
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);
}

static bool __rmap_clear_dirty(struct kvm *kvm, unsigned long *rmapp)
{
	u64 *sptep;
	struct rmap_iterator iter;
	bool flush = false;

1240
	for_each_rmap_spte(rmapp, &iter, sptep)
1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262
		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);
}

static bool __rmap_set_dirty(struct kvm *kvm, unsigned long *rmapp)
{
	u64 *sptep;
	struct rmap_iterator iter;
	bool flush = false;

1263
	for_each_rmap_spte(rmapp, &iter, sptep)
1264 1265 1266 1267 1268
		flush |= spte_set_dirty(kvm, sptep);

	return flush;
}

1269
/**
1270
 * kvm_mmu_write_protect_pt_masked - write protect selected PT level pages
1271 1272 1273 1274 1275 1276 1277 1278
 * @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.
 */
1279
static void kvm_mmu_write_protect_pt_masked(struct kvm *kvm,
1280 1281
				     struct kvm_memory_slot *slot,
				     gfn_t gfn_offset, unsigned long mask)
1282 1283 1284
{
	unsigned long *rmapp;

1285
	while (mask) {
1286 1287
		rmapp = __gfn_to_rmap(slot->base_gfn + gfn_offset + __ffs(mask),
				      PT_PAGE_TABLE_LEVEL, slot);
1288
		__rmap_write_protect(kvm, rmapp, false);
M
Marcelo Tosatti 已提交
1289

1290 1291 1292
		/* clear the first set bit */
		mask &= mask - 1;
	}
1293 1294
}

1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320
/**
 * 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)
{
	unsigned long *rmapp;

	while (mask) {
		rmapp = __gfn_to_rmap(slot->base_gfn + gfn_offset + __ffs(mask),
				      PT_PAGE_TABLE_LEVEL, slot);
		__rmap_clear_dirty(kvm, rmapp);

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

1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334
/**
 * 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)
{
1335 1336 1337 1338 1339
	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);
1340 1341
}

1342
static bool rmap_write_protect(struct kvm_vcpu *vcpu, u64 gfn)
1343 1344
{
	struct kvm_memory_slot *slot;
1345 1346
	unsigned long *rmapp;
	int i;
1347
	bool write_protected = false;
1348

1349
	slot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);
1350

1351
	for (i = PT_PAGE_TABLE_LEVEL; i <= PT_MAX_HUGEPAGE_LEVEL; ++i) {
1352
		rmapp = __gfn_to_rmap(gfn, i, slot);
1353
		write_protected |= __rmap_write_protect(vcpu->kvm, rmapp, true);
1354 1355 1356
	}

	return write_protected;
1357 1358
}

1359
static bool kvm_zap_rmapp(struct kvm *kvm, unsigned long *rmapp)
1360
{
1361 1362
	u64 *sptep;
	struct rmap_iterator iter;
1363
	bool flush = false;
1364

1365 1366
	while ((sptep = rmap_get_first(*rmapp, &iter))) {
		BUG_ON(!(*sptep & PT_PRESENT_MASK));
1367
		rmap_printk("%s: spte %p %llx.\n", __func__, sptep, *sptep);
1368 1369

		drop_spte(kvm, sptep);
1370
		flush = true;
1371
	}
1372

1373 1374 1375 1376 1377 1378 1379 1380
	return flush;
}

static int kvm_unmap_rmapp(struct kvm *kvm, unsigned long *rmapp,
			   struct kvm_memory_slot *slot, gfn_t gfn, int level,
			   unsigned long data)
{
	return kvm_zap_rmapp(kvm, rmapp);
1381 1382
}

F
Frederik Deweerdt 已提交
1383
static int kvm_set_pte_rmapp(struct kvm *kvm, unsigned long *rmapp,
1384 1385
			     struct kvm_memory_slot *slot, gfn_t gfn, int level,
			     unsigned long data)
1386
{
1387 1388
	u64 *sptep;
	struct rmap_iterator iter;
1389
	int need_flush = 0;
1390
	u64 new_spte;
1391 1392 1393 1394 1395
	pte_t *ptep = (pte_t *)data;
	pfn_t new_pfn;

	WARN_ON(pte_huge(*ptep));
	new_pfn = pte_pfn(*ptep);
1396

1397 1398
restart:
	for_each_rmap_spte(rmapp, &iter, sptep) {
1399 1400
		rmap_printk("kvm_set_pte_rmapp: spte %p %llx gfn %llx (%d)\n",
			     sptep, *sptep, gfn, level);
1401

1402
		need_flush = 1;
1403

1404
		if (pte_write(*ptep)) {
1405
			drop_spte(kvm, sptep);
1406
			goto restart;
1407
		} else {
1408
			new_spte = *sptep & ~PT64_BASE_ADDR_MASK;
1409 1410 1411 1412
			new_spte |= (u64)new_pfn << PAGE_SHIFT;

			new_spte &= ~PT_WRITABLE_MASK;
			new_spte &= ~SPTE_HOST_WRITEABLE;
1413
			new_spte &= ~shadow_accessed_mask;
1414 1415 1416

			mmu_spte_clear_track_bits(sptep);
			mmu_spte_set(sptep, new_spte);
1417 1418
		}
	}
1419

1420 1421 1422 1423 1424 1425
	if (need_flush)
		kvm_flush_remote_tlbs(kvm);

	return 0;
}

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 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493
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;
	unsigned long *rmap;
	int level;

	/* private field. */
	unsigned long *end_rmap;
};

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_))

1494 1495 1496 1497 1498 1499
static int kvm_handle_hva_range(struct kvm *kvm,
				unsigned long start,
				unsigned long end,
				unsigned long data,
				int (*handler)(struct kvm *kvm,
					       unsigned long *rmapp,
1500
					       struct kvm_memory_slot *slot,
1501 1502
					       gfn_t gfn,
					       int level,
1503
					       unsigned long data))
1504
{
1505
	struct kvm_memslots *slots;
1506
	struct kvm_memory_slot *memslot;
1507 1508
	struct slot_rmap_walk_iterator iterator;
	int ret = 0;
1509
	int i;
1510

1511 1512 1513 1514 1515
	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;
1516

1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535
			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);
		}
1536 1537
	}

1538
	return ret;
1539 1540
}

1541 1542 1543
static int kvm_handle_hva(struct kvm *kvm, unsigned long hva,
			  unsigned long data,
			  int (*handler)(struct kvm *kvm, unsigned long *rmapp,
1544
					 struct kvm_memory_slot *slot,
1545
					 gfn_t gfn, int level,
1546 1547 1548
					 unsigned long data))
{
	return kvm_handle_hva_range(kvm, hva, hva + 1, data, handler);
1549 1550 1551 1552
}

int kvm_unmap_hva(struct kvm *kvm, unsigned long hva)
{
1553 1554 1555
	return kvm_handle_hva(kvm, hva, 0, kvm_unmap_rmapp);
}

1556 1557 1558 1559 1560
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);
}

1561 1562
void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte)
{
F
Frederik Deweerdt 已提交
1563
	kvm_handle_hva(kvm, hva, (unsigned long)&pte, kvm_set_pte_rmapp);
1564 1565
}

F
Frederik Deweerdt 已提交
1566
static int kvm_age_rmapp(struct kvm *kvm, unsigned long *rmapp,
1567 1568
			 struct kvm_memory_slot *slot, gfn_t gfn, int level,
			 unsigned long data)
1569
{
1570
	u64 *sptep;
1571
	struct rmap_iterator uninitialized_var(iter);
1572 1573
	int young = 0;

A
Andres Lagar-Cavilla 已提交
1574
	BUG_ON(!shadow_accessed_mask);
1575

1576
	for_each_rmap_spte(rmapp, &iter, sptep)
1577
		if (*sptep & shadow_accessed_mask) {
1578
			young = 1;
1579 1580
			clear_bit((ffs(shadow_accessed_mask) - 1),
				 (unsigned long *)sptep);
1581
		}
1582

1583
	trace_kvm_age_page(gfn, level, slot, young);
1584 1585 1586
	return young;
}

A
Andrea Arcangeli 已提交
1587
static int kvm_test_age_rmapp(struct kvm *kvm, unsigned long *rmapp,
1588 1589
			      struct kvm_memory_slot *slot, gfn_t gfn,
			      int level, unsigned long data)
A
Andrea Arcangeli 已提交
1590
{
1591 1592
	u64 *sptep;
	struct rmap_iterator iter;
A
Andrea Arcangeli 已提交
1593 1594 1595 1596 1597 1598 1599 1600 1601 1602
	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;

1603
	for_each_rmap_spte(rmapp, &iter, sptep)
1604
		if (*sptep & shadow_accessed_mask) {
A
Andrea Arcangeli 已提交
1605 1606 1607 1608 1609 1610 1611
			young = 1;
			break;
		}
out:
	return young;
}

1612 1613
#define RMAP_RECYCLE_THRESHOLD 1000

1614
static void rmap_recycle(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn)
1615 1616
{
	unsigned long *rmapp;
1617 1618 1619
	struct kvm_mmu_page *sp;

	sp = page_header(__pa(spte));
1620

1621
	rmapp = gfn_to_rmap(vcpu->kvm, gfn, sp);
1622

1623
	kvm_unmap_rmapp(vcpu->kvm, rmapp, NULL, gfn, sp->role.level, 0);
1624 1625 1626
	kvm_flush_remote_tlbs(vcpu->kvm);
}

A
Andres Lagar-Cavilla 已提交
1627
int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
1628
{
A
Andres Lagar-Cavilla 已提交
1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649
	/*
	 * 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);
1650 1651
}

A
Andrea Arcangeli 已提交
1652 1653 1654 1655 1656
int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
{
	return kvm_handle_hva(kvm, hva, 0, kvm_test_age_rmapp);
}

1657
#ifdef MMU_DEBUG
1658
static int is_empty_shadow_page(u64 *spt)
A
Avi Kivity 已提交
1659
{
1660 1661 1662
	u64 *pos;
	u64 *end;

1663
	for (pos = spt, end = pos + PAGE_SIZE / sizeof(u64); pos != end; pos++)
1664
		if (is_shadow_present_pte(*pos)) {
1665
			printk(KERN_ERR "%s: %p %llx\n", __func__,
1666
			       pos, *pos);
A
Avi Kivity 已提交
1667
			return 0;
1668
		}
A
Avi Kivity 已提交
1669 1670
	return 1;
}
1671
#endif
A
Avi Kivity 已提交
1672

1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684
/*
 * 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);
}

1685
static void kvm_mmu_free_page(struct kvm_mmu_page *sp)
1686
{
1687
	MMU_WARN_ON(!is_empty_shadow_page(sp->spt));
1688
	hlist_del(&sp->hash_link);
1689 1690
	list_del(&sp->link);
	free_page((unsigned long)sp->spt);
1691 1692
	if (!sp->role.direct)
		free_page((unsigned long)sp->gfns);
1693
	kmem_cache_free(mmu_page_header_cache, sp);
1694 1695
}

1696 1697
static unsigned kvm_page_table_hashfn(gfn_t gfn)
{
1698
	return gfn & ((1 << KVM_MMU_HASH_SHIFT) - 1);
1699 1700
}

1701
static void mmu_page_add_parent_pte(struct kvm_vcpu *vcpu,
1702
				    struct kvm_mmu_page *sp, u64 *parent_pte)
1703 1704 1705 1706
{
	if (!parent_pte)
		return;

1707
	pte_list_add(vcpu, parent_pte, &sp->parent_ptes);
1708 1709
}

1710
static void mmu_page_remove_parent_pte(struct kvm_mmu_page *sp,
1711 1712
				       u64 *parent_pte)
{
1713
	pte_list_remove(parent_pte, &sp->parent_ptes);
1714 1715
}

1716 1717 1718 1719
static void drop_parent_pte(struct kvm_mmu_page *sp,
			    u64 *parent_pte)
{
	mmu_page_remove_parent_pte(sp, parent_pte);
1720
	mmu_spte_clear_no_track(parent_pte);
1721 1722
}

1723 1724
static struct kvm_mmu_page *kvm_mmu_alloc_page(struct kvm_vcpu *vcpu,
					       u64 *parent_pte, int direct)
M
Marcelo Tosatti 已提交
1725
{
1726
	struct kvm_mmu_page *sp;
1727

1728 1729
	sp = mmu_memory_cache_alloc(&vcpu->arch.mmu_page_header_cache);
	sp->spt = mmu_memory_cache_alloc(&vcpu->arch.mmu_page_cache);
1730
	if (!direct)
1731
		sp->gfns = mmu_memory_cache_alloc(&vcpu->arch.mmu_page_cache);
1732
	set_page_private(virt_to_page(sp->spt), (unsigned long)sp);
1733 1734 1735 1736 1737 1738

	/*
	 * 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().
	 */
1739 1740 1741 1742 1743
	list_add(&sp->link, &vcpu->kvm->arch.active_mmu_pages);
	sp->parent_ptes = 0;
	mmu_page_add_parent_pte(vcpu, sp, parent_pte);
	kvm_mod_used_mmu_pages(vcpu->kvm, +1);
	return sp;
M
Marcelo Tosatti 已提交
1744 1745
}

1746
static void mark_unsync(u64 *spte);
1747
static void kvm_mmu_mark_parents_unsync(struct kvm_mmu_page *sp)
1748
{
1749
	pte_list_walk(&sp->parent_ptes, mark_unsync);
1750 1751
}

1752
static void mark_unsync(u64 *spte)
1753
{
1754
	struct kvm_mmu_page *sp;
1755
	unsigned int index;
1756

1757
	sp = page_header(__pa(spte));
1758 1759
	index = spte - sp->spt;
	if (__test_and_set_bit(index, sp->unsync_child_bitmap))
1760
		return;
1761
	if (sp->unsync_children++)
1762
		return;
1763
	kvm_mmu_mark_parents_unsync(sp);
1764 1765
}

1766
static int nonpaging_sync_page(struct kvm_vcpu *vcpu,
1767
			       struct kvm_mmu_page *sp)
1768 1769 1770 1771
{
	return 1;
}

M
Marcelo Tosatti 已提交
1772 1773 1774 1775
static void nonpaging_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
{
}

1776 1777
static void nonpaging_update_pte(struct kvm_vcpu *vcpu,
				 struct kvm_mmu_page *sp, u64 *spte,
1778
				 const void *pte)
1779 1780 1781 1782
{
	WARN_ON(1);
}

1783 1784 1785 1786 1787 1788 1789 1790 1791 1792
#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;
};

1793 1794
static int mmu_pages_add(struct kvm_mmu_pages *pvec, struct kvm_mmu_page *sp,
			 int idx)
1795
{
1796
	int i;
1797

1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812
	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);
}

static int __mmu_unsync_walk(struct kvm_mmu_page *sp,
			   struct kvm_mmu_pages *pvec)
{
	int i, ret, nr_unsync_leaf = 0;
1813

1814
	for_each_set_bit(i, sp->unsync_child_bitmap, 512) {
1815
		struct kvm_mmu_page *child;
1816 1817
		u64 ent = sp->spt[i];

1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846
		if (!is_shadow_present_pte(ent) || is_large_pte(ent))
			goto clear_child_bitmap;

		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);
			if (!ret)
				goto clear_child_bitmap;
			else if (ret > 0)
				nr_unsync_leaf += ret;
			else
				return ret;
		} else if (child->unsync) {
			nr_unsync_leaf++;
			if (mmu_pages_add(pvec, child, i))
				return -ENOSPC;
		} else
			 goto clear_child_bitmap;

		continue;

clear_child_bitmap:
		__clear_bit(i, sp->unsync_child_bitmap);
		sp->unsync_children--;
		WARN_ON((int)sp->unsync_children < 0);
1847 1848 1849
	}


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

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

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

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

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

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

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

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

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

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

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

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

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

1926 1927 1928
	return ret;
}

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

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

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

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

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

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

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

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

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

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

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

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

	return n;
}

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

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

2008 2009 2010 2011 2012 2013 2014 2015 2016
		sp = parents->parent[level];
		if (!sp)
			return;

		--sp->unsync_children;
		WARN_ON((int)sp->unsync_children < 0);
		__clear_bit(idx, sp->unsync_child_bitmap);
		level++;
	} while (level < PT64_ROOT_LEVEL-1 && !sp->unsync_children);
2017 2018
}

2019 2020 2021
static void kvm_mmu_pages_init(struct kvm_mmu_page *parent,
			       struct mmu_page_path *parents,
			       struct kvm_mmu_pages *pvec)
2022
{
2023 2024 2025
	parents->parent[parent->role.level-1] = NULL;
	pvec->nr = 0;
}
2026

2027 2028 2029 2030 2031 2032 2033
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;
2034
	LIST_HEAD(invalid_list);
2035 2036 2037

	kvm_mmu_pages_init(parent, &parents, &pages);
	while (mmu_unsync_walk(parent, &pages)) {
2038
		bool protected = false;
2039 2040

		for_each_sp(pages, sp, parents, i)
2041
			protected |= rmap_write_protect(vcpu, sp->gfn);
2042 2043 2044 2045

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

2046
		for_each_sp(pages, sp, parents, i) {
2047
			kvm_sync_page(vcpu, sp, &invalid_list);
2048 2049
			mmu_pages_clear_parents(&parents);
		}
2050
		kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
2051
		cond_resched_lock(&vcpu->kvm->mmu_lock);
2052 2053
		kvm_mmu_pages_init(parent, &parents, &pages);
	}
2054 2055
}

2056 2057 2058 2059 2060 2061 2062 2063
static void init_shadow_page_table(struct kvm_mmu_page *sp)
{
	int i;

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

2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075
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);
}

2076 2077 2078 2079 2080
static bool is_obsolete_sp(struct kvm *kvm, struct kvm_mmu_page *sp)
{
	return unlikely(sp->mmu_valid_gen != kvm->arch.mmu_valid_gen);
}

2081 2082 2083 2084
static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
					     gfn_t gfn,
					     gva_t gaddr,
					     unsigned level,
2085
					     int direct,
2086
					     unsigned access,
2087
					     u64 *parent_pte)
2088 2089 2090
{
	union kvm_mmu_page_role role;
	unsigned quadrant;
2091 2092
	struct kvm_mmu_page *sp;
	bool need_sync = false;
2093

2094
	role = vcpu->arch.mmu.base_role;
2095
	role.level = level;
2096
	role.direct = direct;
2097
	if (role.direct)
2098
		role.cr4_pae = 0;
2099
	role.access = access;
2100 2101
	if (!vcpu->arch.mmu.direct_map
	    && vcpu->arch.mmu.root_level <= PT32_ROOT_LEVEL) {
2102 2103 2104 2105
		quadrant = gaddr >> (PAGE_SHIFT + (PT64_PT_BITS * level));
		quadrant &= (1 << ((PT32_PT_BITS - PT64_PT_BITS) * level)) - 1;
		role.quadrant = quadrant;
	}
2106
	for_each_gfn_sp(vcpu->kvm, sp, gfn) {
2107 2108 2109
		if (is_obsolete_sp(vcpu->kvm, sp))
			continue;

2110 2111
		if (!need_sync && sp->unsync)
			need_sync = true;
2112

2113 2114
		if (sp->role.word != role.word)
			continue;
2115

2116 2117
		if (sp->unsync && kvm_sync_page_transient(vcpu, sp))
			break;
2118

2119 2120
		mmu_page_add_parent_pte(vcpu, sp, parent_pte);
		if (sp->unsync_children) {
2121
			kvm_make_request(KVM_REQ_MMU_SYNC, vcpu);
2122 2123 2124
			kvm_mmu_mark_parents_unsync(sp);
		} else if (sp->unsync)
			kvm_mmu_mark_parents_unsync(sp);
2125

2126
		__clear_sp_write_flooding_count(sp);
2127 2128 2129
		trace_kvm_mmu_get_page(sp, false);
		return sp;
	}
A
Avi Kivity 已提交
2130
	++vcpu->kvm->stat.mmu_cache_miss;
2131
	sp = kvm_mmu_alloc_page(vcpu, parent_pte, direct);
2132 2133 2134 2135
	if (!sp)
		return sp;
	sp->gfn = gfn;
	sp->role = role;
2136 2137
	hlist_add_head(&sp->hash_link,
		&vcpu->kvm->arch.mmu_page_hash[kvm_page_table_hashfn(gfn)]);
2138
	if (!direct) {
2139
		if (rmap_write_protect(vcpu, gfn))
2140
			kvm_flush_remote_tlbs(vcpu->kvm);
2141 2142 2143
		if (level > PT_PAGE_TABLE_LEVEL && need_sync)
			kvm_sync_pages(vcpu, gfn);

2144
		account_shadowed(vcpu->kvm, sp);
2145
	}
2146
	sp->mmu_valid_gen = vcpu->kvm->arch.mmu_valid_gen;
2147
	init_shadow_page_table(sp);
A
Avi Kivity 已提交
2148
	trace_kvm_mmu_get_page(sp, true);
2149
	return sp;
2150 2151
}

2152 2153 2154 2155 2156 2157
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;
2158 2159 2160 2161 2162 2163

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

2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177
	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;
2178

2179 2180 2181 2182 2183
	iterator->index = SHADOW_PT_INDEX(iterator->addr, iterator->level);
	iterator->sptep	= ((u64 *)__va(iterator->shadow_addr)) + iterator->index;
	return true;
}

2184 2185
static void __shadow_walk_next(struct kvm_shadow_walk_iterator *iterator,
			       u64 spte)
2186
{
2187
	if (is_last_spte(spte, iterator->level)) {
2188 2189 2190 2191
		iterator->level = 0;
		return;
	}

2192
	iterator->shadow_addr = spte & PT64_BASE_ADDR_MASK;
2193 2194 2195
	--iterator->level;
}

2196 2197 2198 2199 2200
static void shadow_walk_next(struct kvm_shadow_walk_iterator *iterator)
{
	return __shadow_walk_next(iterator, *iterator->sptep);
}

2201
static void link_shadow_page(u64 *sptep, struct kvm_mmu_page *sp, bool accessed)
2202 2203 2204
{
	u64 spte;

2205 2206 2207
	BUILD_BUG_ON(VMX_EPT_READABLE_MASK != PT_PRESENT_MASK ||
			VMX_EPT_WRITABLE_MASK != PT_WRITABLE_MASK);

X
Xiao Guangrong 已提交
2208
	spte = __pa(sp->spt) | PT_PRESENT_MASK | PT_WRITABLE_MASK |
2209 2210 2211 2212
	       shadow_user_mask | shadow_x_mask;

	if (accessed)
		spte |= shadow_accessed_mask;
X
Xiao Guangrong 已提交
2213

2214
	mmu_spte_set(sptep, spte);
2215 2216
}

2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233
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;

2234
		drop_parent_pte(child, sptep);
2235 2236 2237 2238
		kvm_flush_remote_tlbs(vcpu->kvm);
	}
}

X
Xiao Guangrong 已提交
2239
static bool mmu_page_zap_pte(struct kvm *kvm, struct kvm_mmu_page *sp,
2240 2241 2242 2243 2244 2245 2246
			     u64 *spte)
{
	u64 pte;
	struct kvm_mmu_page *child;

	pte = *spte;
	if (is_shadow_present_pte(pte)) {
X
Xiao Guangrong 已提交
2247
		if (is_last_spte(pte, sp->role.level)) {
2248
			drop_spte(kvm, spte);
X
Xiao Guangrong 已提交
2249 2250 2251
			if (is_large_pte(pte))
				--kvm->stat.lpages;
		} else {
2252
			child = page_header(pte & PT64_BASE_ADDR_MASK);
2253
			drop_parent_pte(child, spte);
2254
		}
X
Xiao Guangrong 已提交
2255 2256 2257 2258
		return true;
	}

	if (is_mmio_spte(pte))
2259
		mmu_spte_clear_no_track(spte);
2260

X
Xiao Guangrong 已提交
2261
	return false;
2262 2263
}

2264
static void kvm_mmu_page_unlink_children(struct kvm *kvm,
2265
					 struct kvm_mmu_page *sp)
2266
{
2267 2268
	unsigned i;

2269 2270
	for (i = 0; i < PT64_ENT_PER_PAGE; ++i)
		mmu_page_zap_pte(kvm, sp, sp->spt + i);
2271 2272
}

2273
static void kvm_mmu_put_page(struct kvm_mmu_page *sp, u64 *parent_pte)
2274
{
2275
	mmu_page_remove_parent_pte(sp, parent_pte);
2276 2277
}

2278
static void kvm_mmu_unlink_parents(struct kvm *kvm, struct kvm_mmu_page *sp)
2279
{
2280 2281
	u64 *sptep;
	struct rmap_iterator iter;
2282

2283 2284
	while ((sptep = rmap_get_first(sp->parent_ptes, &iter)))
		drop_parent_pte(sp, sptep);
2285 2286
}

2287
static int mmu_zap_unsync_children(struct kvm *kvm,
2288 2289
				   struct kvm_mmu_page *parent,
				   struct list_head *invalid_list)
2290
{
2291 2292 2293
	int i, zapped = 0;
	struct mmu_page_path parents;
	struct kvm_mmu_pages pages;
2294

2295
	if (parent->role.level == PT_PAGE_TABLE_LEVEL)
2296
		return 0;
2297 2298 2299 2300 2301 2302

	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) {
2303
			kvm_mmu_prepare_zap_page(kvm, sp, invalid_list);
2304
			mmu_pages_clear_parents(&parents);
2305
			zapped++;
2306 2307 2308 2309 2310
		}
		kvm_mmu_pages_init(parent, &parents, &pages);
	}

	return zapped;
2311 2312
}

2313 2314
static int kvm_mmu_prepare_zap_page(struct kvm *kvm, struct kvm_mmu_page *sp,
				    struct list_head *invalid_list)
2315
{
2316
	int ret;
A
Avi Kivity 已提交
2317

2318
	trace_kvm_mmu_prepare_zap_page(sp);
2319
	++kvm->stat.mmu_shadow_zapped;
2320
	ret = mmu_zap_unsync_children(kvm, sp, invalid_list);
2321
	kvm_mmu_page_unlink_children(kvm, sp);
2322
	kvm_mmu_unlink_parents(kvm, sp);
2323

2324
	if (!sp->role.invalid && !sp->role.direct)
2325
		unaccount_shadowed(kvm, sp);
2326

2327 2328
	if (sp->unsync)
		kvm_unlink_unsync_page(kvm, sp);
2329
	if (!sp->root_count) {
2330 2331
		/* Count self */
		ret++;
2332
		list_move(&sp->link, invalid_list);
2333
		kvm_mod_used_mmu_pages(kvm, -1);
2334
	} else {
A
Avi Kivity 已提交
2335
		list_move(&sp->link, &kvm->arch.active_mmu_pages);
2336 2337 2338 2339 2340 2341 2342

		/*
		 * 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);
2343
	}
2344 2345

	sp->role.invalid = 1;
2346
	return ret;
2347 2348
}

2349 2350 2351
static void kvm_mmu_commit_zap_page(struct kvm *kvm,
				    struct list_head *invalid_list)
{
2352
	struct kvm_mmu_page *sp, *nsp;
2353 2354 2355 2356

	if (list_empty(invalid_list))
		return;

2357 2358 2359 2360 2361
	/*
	 * 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
2362

2363 2364 2365 2366 2367
	/*
	 * Wait for all vcpus to exit guest mode and/or lockless shadow
	 * page table walks.
	 */
	kvm_flush_remote_tlbs(kvm);
2368

2369
	list_for_each_entry_safe(sp, nsp, invalid_list, link) {
2370
		WARN_ON(!sp->role.invalid || sp->root_count);
2371
		kvm_mmu_free_page(sp);
2372
	}
2373 2374
}

2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389
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;
}

2390 2391
/*
 * Changing the number of mmu pages allocated to the vm
2392
 * Note: if goal_nr_mmu_pages is too small, you will get dead lock
2393
 */
2394
void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int goal_nr_mmu_pages)
2395
{
2396
	LIST_HEAD(invalid_list);
2397

2398 2399
	spin_lock(&kvm->mmu_lock);

2400
	if (kvm->arch.n_used_mmu_pages > goal_nr_mmu_pages) {
2401 2402 2403 2404
		/* 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;
2405

2406
		kvm_mmu_commit_zap_page(kvm, &invalid_list);
2407
		goal_nr_mmu_pages = kvm->arch.n_used_mmu_pages;
2408 2409
	}

2410
	kvm->arch.n_max_mmu_pages = goal_nr_mmu_pages;
2411 2412

	spin_unlock(&kvm->mmu_lock);
2413 2414
}

2415
int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn)
2416
{
2417
	struct kvm_mmu_page *sp;
2418
	LIST_HEAD(invalid_list);
2419 2420
	int r;

2421
	pgprintk("%s: looking for gfn %llx\n", __func__, gfn);
2422
	r = 0;
2423
	spin_lock(&kvm->mmu_lock);
2424
	for_each_gfn_indirect_valid_sp(kvm, sp, gfn) {
2425
		pgprintk("%s: gfn %llx role %x\n", __func__, gfn,
2426 2427
			 sp->role.word);
		r = 1;
2428
		kvm_mmu_prepare_zap_page(kvm, sp, &invalid_list);
2429
	}
2430
	kvm_mmu_commit_zap_page(kvm, &invalid_list);
2431 2432
	spin_unlock(&kvm->mmu_lock);

2433
	return r;
2434
}
2435
EXPORT_SYMBOL_GPL(kvm_mmu_unprotect_page);
2436

2437 2438 2439 2440 2441 2442 2443 2444 2445 2446
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)
2447 2448
{
	struct kvm_mmu_page *s;
2449

2450
	for_each_gfn_indirect_valid_sp(vcpu->kvm, s, gfn) {
2451
		if (s->unsync)
2452
			continue;
2453 2454
		WARN_ON(s->role.level != PT_PAGE_TABLE_LEVEL);
		__kvm_unsync_page(vcpu, s);
2455 2456 2457 2458 2459 2460
	}
}

static int mmu_need_write_protect(struct kvm_vcpu *vcpu, gfn_t gfn,
				  bool can_unsync)
{
2461 2462 2463
	struct kvm_mmu_page *s;
	bool need_unsync = false;

2464
	for_each_gfn_indirect_valid_sp(vcpu->kvm, s, gfn) {
2465 2466 2467
		if (!can_unsync)
			return 1;

2468
		if (s->role.level != PT_PAGE_TABLE_LEVEL)
2469
			return 1;
2470

G
Gleb Natapov 已提交
2471
		if (!s->unsync)
2472
			need_unsync = true;
2473
	}
2474 2475
	if (need_unsync)
		kvm_unsync_pages(vcpu, gfn);
2476 2477 2478
	return 0;
}

2479 2480 2481 2482 2483 2484 2485 2486
static bool kvm_is_mmio_pfn(pfn_t pfn)
{
	if (pfn_valid(pfn))
		return !is_zero_pfn(pfn) && PageReserved(pfn_to_page(pfn));

	return true;
}

A
Avi Kivity 已提交
2487
static int set_spte(struct kvm_vcpu *vcpu, u64 *sptep,
2488
		    unsigned pte_access, int level,
2489
		    gfn_t gfn, pfn_t pfn, bool speculative,
2490
		    bool can_unsync, bool host_writable)
2491
{
2492
	u64 spte;
M
Marcelo Tosatti 已提交
2493
	int ret = 0;
S
Sheng Yang 已提交
2494

2495
	if (set_mmio_spte(vcpu, sptep, gfn, pfn, pte_access))
2496 2497
		return 0;

2498
	spte = PT_PRESENT_MASK;
2499
	if (!speculative)
2500
		spte |= shadow_accessed_mask;
2501

S
Sheng Yang 已提交
2502 2503 2504 2505
	if (pte_access & ACC_EXEC_MASK)
		spte |= shadow_x_mask;
	else
		spte |= shadow_nx_mask;
2506

2507
	if (pte_access & ACC_USER_MASK)
S
Sheng Yang 已提交
2508
		spte |= shadow_user_mask;
2509

2510
	if (level > PT_PAGE_TABLE_LEVEL)
M
Marcelo Tosatti 已提交
2511
		spte |= PT_PAGE_SIZE_MASK;
2512
	if (tdp_enabled)
2513
		spte |= kvm_x86_ops->get_mt_mask(vcpu, gfn,
2514
			kvm_is_mmio_pfn(pfn));
2515

2516
	if (host_writable)
2517
		spte |= SPTE_HOST_WRITEABLE;
2518 2519
	else
		pte_access &= ~ACC_WRITE_MASK;
2520

2521
	spte |= (u64)pfn << PAGE_SHIFT;
2522

2523
	if (pte_access & ACC_WRITE_MASK) {
2524

X
Xiao Guangrong 已提交
2525
		/*
2526 2527 2528 2529
		 * 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
2530
		 */
2531
		if (level > PT_PAGE_TABLE_LEVEL &&
2532
		    has_wrprotected_page(vcpu, gfn, level))
A
Avi Kivity 已提交
2533
			goto done;
2534

2535
		spte |= PT_WRITABLE_MASK | SPTE_MMU_WRITEABLE;
2536

2537 2538 2539 2540 2541 2542
		/*
		 * 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.
		 */
2543
		if (!can_unsync && is_writable_pte(*sptep))
2544 2545
			goto set_pte;

2546
		if (mmu_need_write_protect(vcpu, gfn, can_unsync)) {
2547
			pgprintk("%s: found shadow page for %llx, marking ro\n",
2548
				 __func__, gfn);
M
Marcelo Tosatti 已提交
2549
			ret = 1;
2550
			pte_access &= ~ACC_WRITE_MASK;
2551
			spte &= ~(PT_WRITABLE_MASK | SPTE_MMU_WRITEABLE);
2552 2553 2554
		}
	}

2555
	if (pte_access & ACC_WRITE_MASK) {
2556
		kvm_vcpu_mark_page_dirty(vcpu, gfn);
2557 2558
		spte |= shadow_dirty_mask;
	}
2559

2560
set_pte:
2561
	if (mmu_spte_update(sptep, spte))
2562
		kvm_flush_remote_tlbs(vcpu->kvm);
A
Avi Kivity 已提交
2563
done:
M
Marcelo Tosatti 已提交
2564 2565 2566
	return ret;
}

A
Avi Kivity 已提交
2567
static void mmu_set_spte(struct kvm_vcpu *vcpu, u64 *sptep,
2568 2569 2570
			 unsigned pte_access, int write_fault, int *emulate,
			 int level, gfn_t gfn, pfn_t pfn, bool speculative,
			 bool host_writable)
M
Marcelo Tosatti 已提交
2571 2572
{
	int was_rmapped = 0;
2573
	int rmap_count;
M
Marcelo Tosatti 已提交
2574

2575 2576
	pgprintk("%s: spte %llx write_fault %d gfn %llx\n", __func__,
		 *sptep, write_fault, gfn);
M
Marcelo Tosatti 已提交
2577

A
Avi Kivity 已提交
2578
	if (is_rmap_spte(*sptep)) {
M
Marcelo Tosatti 已提交
2579 2580 2581 2582
		/*
		 * If we overwrite a PTE page pointer with a 2MB PMD, unlink
		 * the parent of the now unreachable PTE.
		 */
2583 2584
		if (level > PT_PAGE_TABLE_LEVEL &&
		    !is_large_pte(*sptep)) {
M
Marcelo Tosatti 已提交
2585
			struct kvm_mmu_page *child;
A
Avi Kivity 已提交
2586
			u64 pte = *sptep;
M
Marcelo Tosatti 已提交
2587 2588

			child = page_header(pte & PT64_BASE_ADDR_MASK);
2589
			drop_parent_pte(child, sptep);
2590
			kvm_flush_remote_tlbs(vcpu->kvm);
A
Avi Kivity 已提交
2591
		} else if (pfn != spte_to_pfn(*sptep)) {
2592
			pgprintk("hfn old %llx new %llx\n",
A
Avi Kivity 已提交
2593
				 spte_to_pfn(*sptep), pfn);
2594
			drop_spte(vcpu->kvm, sptep);
2595
			kvm_flush_remote_tlbs(vcpu->kvm);
2596 2597
		} else
			was_rmapped = 1;
M
Marcelo Tosatti 已提交
2598
	}
2599

2600 2601
	if (set_spte(vcpu, sptep, pte_access, level, gfn, pfn, speculative,
	      true, host_writable)) {
M
Marcelo Tosatti 已提交
2602
		if (write_fault)
2603
			*emulate = 1;
2604
		kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
2605
	}
M
Marcelo Tosatti 已提交
2606

2607 2608 2609
	if (unlikely(is_mmio_spte(*sptep) && emulate))
		*emulate = 1;

A
Avi Kivity 已提交
2610
	pgprintk("%s: setting spte %llx\n", __func__, *sptep);
2611
	pgprintk("instantiating %s PTE (%s) at %llx (%llx) addr %p\n",
A
Avi Kivity 已提交
2612
		 is_large_pte(*sptep)? "2MB" : "4kB",
2613 2614
		 *sptep & PT_PRESENT_MASK ?"RW":"R", gfn,
		 *sptep, sptep);
A
Avi Kivity 已提交
2615
	if (!was_rmapped && is_large_pte(*sptep))
M
Marcelo Tosatti 已提交
2616 2617
		++vcpu->kvm->stat.lpages;

2618 2619 2620 2621 2622 2623
	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);
		}
2624
	}
2625

X
Xiao Guangrong 已提交
2626
	kvm_release_pfn_clean(pfn);
2627 2628
}

2629 2630 2631 2632 2633
static pfn_t pte_prefetch_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn,
				     bool no_dirty_log)
{
	struct kvm_memory_slot *slot;

2634
	slot = gfn_to_memslot_dirty_bitmap(vcpu, gfn, no_dirty_log);
2635
	if (!slot)
2636
		return KVM_PFN_ERR_FAULT;
2637

2638
	return gfn_to_pfn_memslot_atomic(slot, gfn);
2639 2640 2641 2642 2643 2644 2645
}

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];
2646
	struct kvm_memory_slot *slot;
2647 2648 2649 2650 2651
	unsigned access = sp->role.access;
	int i, ret;
	gfn_t gfn;

	gfn = kvm_mmu_page_get_gfn(sp, start - sp->spt);
2652 2653
	slot = gfn_to_memslot_dirty_bitmap(vcpu, gfn, access & ACC_WRITE_MASK);
	if (!slot)
2654 2655
		return -1;

2656
	ret = gfn_to_page_many_atomic(slot, gfn, pages, end - start);
2657 2658 2659 2660
	if (ret <= 0)
		return -1;

	for (i = 0; i < ret; i++, gfn++, start++)
2661
		mmu_set_spte(vcpu, start, access, 0, NULL,
2662 2663
			     sp->role.level, gfn, page_to_pfn(pages[i]),
			     true, true);
2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679

	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++) {
2680
		if (is_shadow_present_pte(*spte) || spte == sptep) {
2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710
			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);
}

2711
static int __direct_map(struct kvm_vcpu *vcpu, gpa_t v, int write,
2712 2713
			int map_writable, int level, gfn_t gfn, pfn_t pfn,
			bool prefault)
2714
{
2715
	struct kvm_shadow_walk_iterator iterator;
2716
	struct kvm_mmu_page *sp;
2717
	int emulate = 0;
2718
	gfn_t pseudo_gfn;
A
Avi Kivity 已提交
2719

2720 2721 2722
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
		return 0;

2723
	for_each_shadow_entry(vcpu, (u64)gfn << PAGE_SHIFT, iterator) {
2724
		if (iterator.level == level) {
2725
			mmu_set_spte(vcpu, iterator.sptep, ACC_ALL,
2726 2727
				     write, &emulate, level, gfn, pfn,
				     prefault, map_writable);
2728
			direct_pte_prefetch(vcpu, iterator.sptep);
2729 2730
			++vcpu->stat.pf_fixed;
			break;
A
Avi Kivity 已提交
2731 2732
		}

2733
		drop_large_spte(vcpu, iterator.sptep);
2734
		if (!is_shadow_present_pte(*iterator.sptep)) {
2735 2736 2737 2738
			u64 base_addr = iterator.addr;

			base_addr &= PT64_LVL_ADDR_MASK(iterator.level);
			pseudo_gfn = base_addr >> PAGE_SHIFT;
2739 2740 2741
			sp = kvm_mmu_get_page(vcpu, pseudo_gfn, iterator.addr,
					      iterator.level - 1,
					      1, ACC_ALL, iterator.sptep);
2742

2743
			link_shadow_page(iterator.sptep, sp, true);
2744 2745
		}
	}
2746
	return emulate;
A
Avi Kivity 已提交
2747 2748
}

H
Huang Ying 已提交
2749
static void kvm_send_hwpoison_signal(unsigned long address, struct task_struct *tsk)
2750
{
H
Huang Ying 已提交
2751 2752 2753 2754 2755 2756 2757
	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;
2758

H
Huang Ying 已提交
2759
	send_sig_info(SIGBUS, &info, tsk);
2760 2761
}

2762
static int kvm_handle_bad_page(struct kvm_vcpu *vcpu, gfn_t gfn, pfn_t pfn)
2763
{
X
Xiao Guangrong 已提交
2764 2765 2766 2767 2768 2769 2770 2771 2772
	/*
	 * 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;

2773
	if (pfn == KVM_PFN_ERR_HWPOISON) {
2774
		kvm_send_hwpoison_signal(kvm_vcpu_gfn_to_hva(vcpu, gfn), current);
2775
		return 0;
2776
	}
2777

2778
	return -EFAULT;
2779 2780
}

2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793
static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
					gfn_t *gfnp, pfn_t *pfnp, int *levelp)
{
	pfn_t pfn = *pfnp;
	gfn_t gfn = *gfnp;
	int level = *levelp;

	/*
	 * Check if it's a transparent hugepage. If this would be an
	 * hugetlbfs page, level wouldn't be set to
	 * PT_PAGE_TABLE_LEVEL and there would be no adjustment done
	 * here.
	 */
2794
	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn) &&
2795 2796
	    level == PT_PAGE_TABLE_LEVEL &&
	    PageTransCompound(pfn_to_page(pfn)) &&
2797
	    !has_wrprotected_page(vcpu, gfn, PT_DIRECTORY_LEVEL)) {
2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815
		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;
2816
			kvm_get_pfn(pfn);
2817 2818 2819 2820 2821
			*pfnp = pfn;
		}
	}
}

2822 2823 2824 2825 2826 2827
static bool handle_abnormal_pfn(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
				pfn_t pfn, unsigned access, int *ret_val)
{
	bool ret = true;

	/* The pfn is invalid, report the error! */
2828
	if (unlikely(is_error_pfn(pfn))) {
2829 2830 2831 2832
		*ret_val = kvm_handle_bad_page(vcpu, gfn, pfn);
		goto exit;
	}

2833
	if (unlikely(is_noslot_pfn(pfn)))
2834 2835 2836 2837 2838 2839 2840
		vcpu_cache_mmio_info(vcpu, gva, gfn, access);

	ret = false;
exit:
	return ret;
}

2841
static bool page_fault_can_be_fast(u32 error_code)
2842
{
2843 2844 2845 2846 2847 2848 2849
	/*
	 * 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;

2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862
	/*
	 * #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
2863 2864
fast_pf_fix_direct_spte(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
			u64 *sptep, u64 spte)
2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875
{
	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);

2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887
	/*
	 * 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.
	 */
2888
	if (cmpxchg64(sptep, spte, spte | PT_WRITABLE_MASK) == spte)
2889
		kvm_vcpu_mark_page_dirty(vcpu, gfn);
2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902

	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;
2903
	struct kvm_mmu_page *sp;
2904 2905 2906
	bool ret = false;
	u64 spte = 0ull;

2907 2908 2909
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
		return false;

2910
	if (!page_fault_can_be_fast(error_code))
2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926
		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.
	 */
	if (!is_rmap_spte(spte)) {
		ret = true;
		goto exit;
	}

2927 2928
	sp = page_header(__pa(iterator.sptep));
	if (!is_last_spte(spte, sp->role.level))
2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948
		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;

2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961
	/*
	 * 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;

2962 2963 2964 2965 2966
	/*
	 * 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.
	 */
2967
	ret = fast_pf_fix_direct_spte(vcpu, sp, iterator.sptep, spte);
2968
exit:
X
Xiao Guangrong 已提交
2969 2970
	trace_fast_page_fault(vcpu, gva, error_code, iterator.sptep,
			      spte, ret);
2971 2972 2973 2974 2975
	walk_shadow_page_lockless_end(vcpu);

	return ret;
}

2976
static bool try_async_pf(struct kvm_vcpu *vcpu, bool prefault, gfn_t gfn,
2977
			 gva_t gva, pfn_t *pfn, bool write, bool *writable);
2978
static void make_mmu_pages_available(struct kvm_vcpu *vcpu);
2979

2980 2981
static int nonpaging_map(struct kvm_vcpu *vcpu, gva_t v, u32 error_code,
			 gfn_t gfn, bool prefault)
2982 2983
{
	int r;
2984
	int level;
2985
	bool force_pt_level = false;
2986
	pfn_t pfn;
2987
	unsigned long mmu_seq;
2988
	bool map_writable, write = error_code & PFERR_WRITE_MASK;
2989

2990
	level = mapping_level(vcpu, gfn, &force_pt_level);
2991 2992 2993 2994 2995 2996 2997 2998
	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;
2999

3000
		gfn &= ~(KVM_PAGES_PER_HPAGE(level) - 1);
3001
	}
M
Marcelo Tosatti 已提交
3002

3003 3004 3005
	if (fast_page_fault(vcpu, v, level, error_code))
		return 0;

3006
	mmu_seq = vcpu->kvm->mmu_notifier_seq;
3007
	smp_rmb();
3008

3009
	if (try_async_pf(vcpu, prefault, gfn, v, &pfn, write, &map_writable))
3010
		return 0;
3011

3012 3013
	if (handle_abnormal_pfn(vcpu, v, gfn, pfn, ACC_ALL, &r))
		return r;
3014

3015
	spin_lock(&vcpu->kvm->mmu_lock);
3016
	if (mmu_notifier_retry(vcpu->kvm, mmu_seq))
3017
		goto out_unlock;
3018
	make_mmu_pages_available(vcpu);
3019 3020
	if (likely(!force_pt_level))
		transparent_hugepage_adjust(vcpu, &gfn, &pfn, &level);
3021 3022
	r = __direct_map(vcpu, v, write, map_writable, level, gfn, pfn,
			 prefault);
3023 3024 3025
	spin_unlock(&vcpu->kvm->mmu_lock);


3026
	return r;
3027 3028 3029 3030 3031

out_unlock:
	spin_unlock(&vcpu->kvm->mmu_lock);
	kvm_release_pfn_clean(pfn);
	return 0;
3032 3033 3034
}


3035 3036 3037
static void mmu_free_roots(struct kvm_vcpu *vcpu)
{
	int i;
3038
	struct kvm_mmu_page *sp;
3039
	LIST_HEAD(invalid_list);
3040

3041
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
A
Avi Kivity 已提交
3042
		return;
3043

3044 3045 3046
	if (vcpu->arch.mmu.shadow_root_level == PT64_ROOT_LEVEL &&
	    (vcpu->arch.mmu.root_level == PT64_ROOT_LEVEL ||
	     vcpu->arch.mmu.direct_map)) {
3047
		hpa_t root = vcpu->arch.mmu.root_hpa;
3048

3049
		spin_lock(&vcpu->kvm->mmu_lock);
3050 3051
		sp = page_header(root);
		--sp->root_count;
3052 3053 3054 3055
		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);
		}
3056
		spin_unlock(&vcpu->kvm->mmu_lock);
3057
		vcpu->arch.mmu.root_hpa = INVALID_PAGE;
3058 3059
		return;
	}
3060 3061

	spin_lock(&vcpu->kvm->mmu_lock);
3062
	for (i = 0; i < 4; ++i) {
3063
		hpa_t root = vcpu->arch.mmu.pae_root[i];
3064

A
Avi Kivity 已提交
3065 3066
		if (root) {
			root &= PT64_BASE_ADDR_MASK;
3067 3068
			sp = page_header(root);
			--sp->root_count;
3069
			if (!sp->root_count && sp->role.invalid)
3070 3071
				kvm_mmu_prepare_zap_page(vcpu->kvm, sp,
							 &invalid_list);
A
Avi Kivity 已提交
3072
		}
3073
		vcpu->arch.mmu.pae_root[i] = INVALID_PAGE;
3074
	}
3075
	kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
3076
	spin_unlock(&vcpu->kvm->mmu_lock);
3077
	vcpu->arch.mmu.root_hpa = INVALID_PAGE;
3078 3079
}

3080 3081 3082 3083 3084
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)) {
3085
		kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu);
3086 3087 3088 3089 3090 3091
		ret = 1;
	}

	return ret;
}

3092 3093 3094
static int mmu_alloc_direct_roots(struct kvm_vcpu *vcpu)
{
	struct kvm_mmu_page *sp;
3095
	unsigned i;
3096 3097 3098

	if (vcpu->arch.mmu.shadow_root_level == PT64_ROOT_LEVEL) {
		spin_lock(&vcpu->kvm->mmu_lock);
3099
		make_mmu_pages_available(vcpu);
3100 3101 3102 3103 3104 3105 3106 3107 3108
		sp = kvm_mmu_get_page(vcpu, 0, 0, PT64_ROOT_LEVEL,
				      1, ACC_ALL, NULL);
		++sp->root_count;
		spin_unlock(&vcpu->kvm->mmu_lock);
		vcpu->arch.mmu.root_hpa = __pa(sp->spt);
	} else if (vcpu->arch.mmu.shadow_root_level == PT32E_ROOT_LEVEL) {
		for (i = 0; i < 4; ++i) {
			hpa_t root = vcpu->arch.mmu.pae_root[i];

3109
			MMU_WARN_ON(VALID_PAGE(root));
3110
			spin_lock(&vcpu->kvm->mmu_lock);
3111
			make_mmu_pages_available(vcpu);
3112 3113
			sp = kvm_mmu_get_page(vcpu, i << (30 - PAGE_SHIFT),
					      i << 30,
3114 3115 3116 3117 3118 3119 3120
					      PT32_ROOT_LEVEL, 1, ACC_ALL,
					      NULL);
			root = __pa(sp->spt);
			++sp->root_count;
			spin_unlock(&vcpu->kvm->mmu_lock);
			vcpu->arch.mmu.pae_root[i] = root | PT_PRESENT_MASK;
		}
3121
		vcpu->arch.mmu.root_hpa = __pa(vcpu->arch.mmu.pae_root);
3122 3123 3124 3125 3126 3127 3128
	} else
		BUG();

	return 0;
}

static int mmu_alloc_shadow_roots(struct kvm_vcpu *vcpu)
3129
{
3130
	struct kvm_mmu_page *sp;
3131 3132 3133
	u64 pdptr, pm_mask;
	gfn_t root_gfn;
	int i;
3134

3135
	root_gfn = vcpu->arch.mmu.get_cr3(vcpu) >> PAGE_SHIFT;
3136

3137 3138 3139 3140 3141 3142 3143 3144
	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) {
3145
		hpa_t root = vcpu->arch.mmu.root_hpa;
3146

3147
		MMU_WARN_ON(VALID_PAGE(root));
3148

3149
		spin_lock(&vcpu->kvm->mmu_lock);
3150
		make_mmu_pages_available(vcpu);
3151 3152
		sp = kvm_mmu_get_page(vcpu, root_gfn, 0, PT64_ROOT_LEVEL,
				      0, ACC_ALL, NULL);
3153 3154
		root = __pa(sp->spt);
		++sp->root_count;
3155
		spin_unlock(&vcpu->kvm->mmu_lock);
3156
		vcpu->arch.mmu.root_hpa = root;
3157
		return 0;
3158
	}
3159

3160 3161
	/*
	 * We shadow a 32 bit page table. This may be a legacy 2-level
3162 3163
	 * 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.
3164
	 */
3165 3166 3167 3168
	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;

3169
	for (i = 0; i < 4; ++i) {
3170
		hpa_t root = vcpu->arch.mmu.pae_root[i];
3171

3172
		MMU_WARN_ON(VALID_PAGE(root));
3173
		if (vcpu->arch.mmu.root_level == PT32E_ROOT_LEVEL) {
3174
			pdptr = vcpu->arch.mmu.get_pdptr(vcpu, i);
3175
			if (!is_present_gpte(pdptr)) {
3176
				vcpu->arch.mmu.pae_root[i] = 0;
A
Avi Kivity 已提交
3177 3178
				continue;
			}
A
Avi Kivity 已提交
3179
			root_gfn = pdptr >> PAGE_SHIFT;
3180 3181
			if (mmu_check_root(vcpu, root_gfn))
				return 1;
3182
		}
3183
		spin_lock(&vcpu->kvm->mmu_lock);
3184
		make_mmu_pages_available(vcpu);
3185
		sp = kvm_mmu_get_page(vcpu, root_gfn, i << 30,
3186
				      PT32_ROOT_LEVEL, 0,
3187
				      ACC_ALL, NULL);
3188 3189
		root = __pa(sp->spt);
		++sp->root_count;
3190 3191
		spin_unlock(&vcpu->kvm->mmu_lock);

3192
		vcpu->arch.mmu.pae_root[i] = root | pm_mask;
3193
	}
3194
	vcpu->arch.mmu.root_hpa = __pa(vcpu->arch.mmu.pae_root);
3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220

	/*
	 * 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);
	}

3221
	return 0;
3222 3223
}

3224 3225 3226 3227 3228 3229 3230 3231
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);
}

3232 3233 3234 3235 3236
static void mmu_sync_roots(struct kvm_vcpu *vcpu)
{
	int i;
	struct kvm_mmu_page *sp;

3237 3238 3239
	if (vcpu->arch.mmu.direct_map)
		return;

3240 3241
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
		return;
3242

3243
	vcpu_clear_mmio_info(vcpu, MMIO_GVA_ANY);
3244
	kvm_mmu_audit(vcpu, AUDIT_PRE_SYNC);
3245
	if (vcpu->arch.mmu.root_level == PT64_ROOT_LEVEL) {
3246 3247 3248
		hpa_t root = vcpu->arch.mmu.root_hpa;
		sp = page_header(root);
		mmu_sync_children(vcpu, sp);
3249
		kvm_mmu_audit(vcpu, AUDIT_POST_SYNC);
3250 3251 3252 3253 3254
		return;
	}
	for (i = 0; i < 4; ++i) {
		hpa_t root = vcpu->arch.mmu.pae_root[i];

3255
		if (root && VALID_PAGE(root)) {
3256 3257 3258 3259 3260
			root &= PT64_BASE_ADDR_MASK;
			sp = page_header(root);
			mmu_sync_children(vcpu, sp);
		}
	}
3261
	kvm_mmu_audit(vcpu, AUDIT_POST_SYNC);
3262 3263 3264 3265 3266 3267
}

void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu)
{
	spin_lock(&vcpu->kvm->mmu_lock);
	mmu_sync_roots(vcpu);
3268
	spin_unlock(&vcpu->kvm->mmu_lock);
3269
}
N
Nadav Har'El 已提交
3270
EXPORT_SYMBOL_GPL(kvm_mmu_sync_roots);
3271

3272
static gpa_t nonpaging_gva_to_gpa(struct kvm_vcpu *vcpu, gva_t vaddr,
3273
				  u32 access, struct x86_exception *exception)
A
Avi Kivity 已提交
3274
{
3275 3276
	if (exception)
		exception->error_code = 0;
A
Avi Kivity 已提交
3277 3278 3279
	return vaddr;
}

3280
static gpa_t nonpaging_gva_to_gpa_nested(struct kvm_vcpu *vcpu, gva_t vaddr,
3281 3282
					 u32 access,
					 struct x86_exception *exception)
3283
{
3284 3285
	if (exception)
		exception->error_code = 0;
3286
	return vcpu->arch.nested_mmu.translate_gpa(vcpu, vaddr, access, exception);
3287 3288
}

3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307
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);
}

3308 3309 3310 3311 3312 3313 3314 3315
static bool quickly_check_mmio_pf(struct kvm_vcpu *vcpu, u64 addr, bool direct)
{
	if (direct)
		return vcpu_match_mmio_gpa(vcpu, addr);

	return vcpu_match_mmio_gva(vcpu, addr);
}

3316 3317 3318
/* 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)
3319 3320
{
	struct kvm_shadow_walk_iterator iterator;
3321 3322 3323
	u64 sptes[PT64_ROOT_LEVEL], spte = 0ull;
	int root, leaf;
	bool reserved = false;
3324

3325
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
3326
		goto exit;
3327

3328
	walk_shadow_page_lockless_begin(vcpu);
3329

3330 3331
	for (shadow_walk_init(&iterator, vcpu, addr),
		 leaf = root = iterator.level;
3332 3333 3334 3335 3336
	     shadow_walk_okay(&iterator);
	     __shadow_walk_next(&iterator, spte)) {
		spte = mmu_spte_get_lockless(iterator.sptep);

		sptes[leaf - 1] = spte;
3337
		leaf--;
3338

3339 3340
		if (!is_shadow_present_pte(spte))
			break;
3341 3342

		reserved |= is_shadow_zero_bits_set(&vcpu->arch.mmu, spte,
3343
						    iterator.level);
3344 3345
	}

3346 3347
	walk_shadow_page_lockless_end(vcpu);

3348 3349 3350
	if (reserved) {
		pr_err("%s: detect reserved bits on spte, addr 0x%llx, dump hierarchy:\n",
		       __func__, addr);
3351
		while (root > leaf) {
3352 3353 3354 3355 3356 3357 3358 3359
			pr_err("------ spte 0x%llx level %d.\n",
			       sptes[root - 1], root);
			root--;
		}
	}
exit:
	*sptep = spte;
	return reserved;
3360 3361 3362 3363 3364
}

int handle_mmio_page_fault_common(struct kvm_vcpu *vcpu, u64 addr, bool direct)
{
	u64 spte;
3365
	bool reserved;
3366 3367

	if (quickly_check_mmio_pf(vcpu, addr, direct))
3368
		return RET_MMIO_PF_EMULATE;
3369

3370 3371 3372
	reserved = walk_shadow_page_get_mmio_spte(vcpu, addr, &spte);
	if (unlikely(reserved))
		return RET_MMIO_PF_BUG;
3373 3374 3375 3376 3377

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

3378
		if (!check_mmio_spte(vcpu, spte))
3379 3380
			return RET_MMIO_PF_INVALID;

3381 3382
		if (direct)
			addr = 0;
X
Xiao Guangrong 已提交
3383 3384

		trace_handle_mmio_page_fault(addr, gfn, access);
3385
		vcpu_cache_mmio_info(vcpu, addr, gfn, access);
3386
		return RET_MMIO_PF_EMULATE;
3387 3388 3389 3390 3391 3392
	}

	/*
	 * If the page table is zapped by other cpus, let CPU fault again on
	 * the address.
	 */
3393
	return RET_MMIO_PF_RETRY;
3394 3395 3396 3397 3398 3399 3400 3401 3402
}
EXPORT_SYMBOL_GPL(handle_mmio_page_fault_common);

static int handle_mmio_page_fault(struct kvm_vcpu *vcpu, u64 addr,
				  u32 error_code, bool direct)
{
	int ret;

	ret = handle_mmio_page_fault_common(vcpu, addr, direct);
3403
	WARN_ON(ret == RET_MMIO_PF_BUG);
3404 3405 3406
	return ret;
}

A
Avi Kivity 已提交
3407
static int nonpaging_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
3408
				u32 error_code, bool prefault)
A
Avi Kivity 已提交
3409
{
3410
	gfn_t gfn;
3411
	int r;
A
Avi Kivity 已提交
3412

3413
	pgprintk("%s: gva %lx error %x\n", __func__, gva, error_code);
3414

3415 3416 3417 3418 3419 3420
	if (unlikely(error_code & PFERR_RSVD_MASK)) {
		r = handle_mmio_page_fault(vcpu, gva, error_code, true);

		if (likely(r != RET_MMIO_PF_INVALID))
			return r;
	}
3421

3422 3423 3424
	r = mmu_topup_memory_caches(vcpu);
	if (r)
		return r;
3425

3426
	MMU_WARN_ON(!VALID_PAGE(vcpu->arch.mmu.root_hpa));
A
Avi Kivity 已提交
3427

3428
	gfn = gva >> PAGE_SHIFT;
A
Avi Kivity 已提交
3429

3430
	return nonpaging_map(vcpu, gva & PAGE_MASK,
3431
			     error_code, gfn, prefault);
A
Avi Kivity 已提交
3432 3433
}

3434
static int kvm_arch_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn)
3435 3436
{
	struct kvm_arch_async_pf arch;
X
Xiao Guangrong 已提交
3437

3438
	arch.token = (vcpu->arch.apf.id++ << 12) | vcpu->vcpu_id;
3439
	arch.gfn = gfn;
3440
	arch.direct_map = vcpu->arch.mmu.direct_map;
X
Xiao Guangrong 已提交
3441
	arch.cr3 = vcpu->arch.mmu.get_cr3(vcpu);
3442

3443
	return kvm_setup_async_pf(vcpu, gva, kvm_vcpu_gfn_to_hva(vcpu, gfn), &arch);
3444 3445 3446 3447
}

static bool can_do_async_pf(struct kvm_vcpu *vcpu)
{
3448
	if (unlikely(!lapic_in_kernel(vcpu) ||
3449 3450 3451 3452 3453 3454
		     kvm_event_needs_reinjection(vcpu)))
		return false;

	return kvm_x86_ops->interrupt_allowed(vcpu);
}

3455
static bool try_async_pf(struct kvm_vcpu *vcpu, bool prefault, gfn_t gfn,
3456
			 gva_t gva, pfn_t *pfn, bool write, bool *writable)
3457
{
3458
	struct kvm_memory_slot *slot;
3459 3460
	bool async;

3461
	slot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);
3462 3463
	async = false;
	*pfn = __gfn_to_pfn_memslot(slot, gfn, false, &async, write, writable);
3464 3465 3466
	if (!async)
		return false; /* *pfn has correct page already */

3467
	if (!prefault && can_do_async_pf(vcpu)) {
3468
		trace_kvm_try_async_get_page(gva, gfn);
3469 3470 3471 3472 3473 3474 3475 3476
		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;
	}

3477
	*pfn = __gfn_to_pfn_memslot(slot, gfn, false, NULL, write, writable);
3478 3479 3480
	return false;
}

3481 3482 3483 3484 3485 3486 3487 3488 3489 3490
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
3491
static int tdp_page_fault(struct kvm_vcpu *vcpu, gva_t gpa, u32 error_code,
3492
			  bool prefault)
3493
{
3494
	pfn_t pfn;
3495
	int r;
3496
	int level;
3497
	bool force_pt_level;
M
Marcelo Tosatti 已提交
3498
	gfn_t gfn = gpa >> PAGE_SHIFT;
3499
	unsigned long mmu_seq;
3500 3501
	int write = error_code & PFERR_WRITE_MASK;
	bool map_writable;
3502

3503
	MMU_WARN_ON(!VALID_PAGE(vcpu->arch.mmu.root_hpa));
3504

3505 3506 3507 3508 3509 3510
	if (unlikely(error_code & PFERR_RSVD_MASK)) {
		r = handle_mmio_page_fault(vcpu, gpa, error_code, true);

		if (likely(r != RET_MMIO_PF_INVALID))
			return r;
	}
3511

3512 3513 3514 3515
	r = mmu_topup_memory_caches(vcpu);
	if (r)
		return r;

3516 3517 3518
	force_pt_level = !check_hugepage_cache_consistency(vcpu, gfn,
							   PT_DIRECTORY_LEVEL);
	level = mapping_level(vcpu, gfn, &force_pt_level);
3519
	if (likely(!force_pt_level)) {
3520 3521 3522
		if (level > PT_DIRECTORY_LEVEL &&
		    !check_hugepage_cache_consistency(vcpu, gfn, level))
			level = PT_DIRECTORY_LEVEL;
3523
		gfn &= ~(KVM_PAGES_PER_HPAGE(level) - 1);
3524
	}
3525

3526 3527 3528
	if (fast_page_fault(vcpu, gpa, level, error_code))
		return 0;

3529
	mmu_seq = vcpu->kvm->mmu_notifier_seq;
3530
	smp_rmb();
3531

3532
	if (try_async_pf(vcpu, prefault, gfn, gpa, &pfn, write, &map_writable))
3533 3534
		return 0;

3535 3536 3537
	if (handle_abnormal_pfn(vcpu, 0, gfn, pfn, ACC_ALL, &r))
		return r;

3538
	spin_lock(&vcpu->kvm->mmu_lock);
3539
	if (mmu_notifier_retry(vcpu->kvm, mmu_seq))
3540
		goto out_unlock;
3541
	make_mmu_pages_available(vcpu);
3542 3543
	if (likely(!force_pt_level))
		transparent_hugepage_adjust(vcpu, &gfn, &pfn, &level);
3544
	r = __direct_map(vcpu, gpa, write, map_writable,
3545
			 level, gfn, pfn, prefault);
3546 3547 3548
	spin_unlock(&vcpu->kvm->mmu_lock);

	return r;
3549 3550 3551 3552 3553

out_unlock:
	spin_unlock(&vcpu->kvm->mmu_lock);
	kvm_release_pfn_clean(pfn);
	return 0;
3554 3555
}

3556 3557
static void nonpaging_init_context(struct kvm_vcpu *vcpu,
				   struct kvm_mmu *context)
A
Avi Kivity 已提交
3558 3559 3560
{
	context->page_fault = nonpaging_page_fault;
	context->gva_to_gpa = nonpaging_gva_to_gpa;
3561
	context->sync_page = nonpaging_sync_page;
M
Marcelo Tosatti 已提交
3562
	context->invlpg = nonpaging_invlpg;
3563
	context->update_pte = nonpaging_update_pte;
3564
	context->root_level = 0;
A
Avi Kivity 已提交
3565
	context->shadow_root_level = PT32E_ROOT_LEVEL;
A
Avi Kivity 已提交
3566
	context->root_hpa = INVALID_PAGE;
3567
	context->direct_map = true;
3568
	context->nx = false;
A
Avi Kivity 已提交
3569 3570
}

3571
void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
3572
{
3573
	mmu_free_roots(vcpu);
A
Avi Kivity 已提交
3574 3575
}

3576 3577
static unsigned long get_cr3(struct kvm_vcpu *vcpu)
{
3578
	return kvm_read_cr3(vcpu);
3579 3580
}

3581 3582
static void inject_page_fault(struct kvm_vcpu *vcpu,
			      struct x86_exception *fault)
A
Avi Kivity 已提交
3583
{
3584
	vcpu->arch.mmu.inject_page_fault(vcpu, fault);
A
Avi Kivity 已提交
3585 3586
}

3587
static bool sync_mmio_spte(struct kvm_vcpu *vcpu, u64 *sptep, gfn_t gfn,
3588
			   unsigned access, int *nr_present)
3589 3590 3591 3592 3593 3594 3595 3596
{
	if (unlikely(is_mmio_spte(*sptep))) {
		if (gfn != get_mmio_spte_gfn(*sptep)) {
			mmu_spte_clear_no_track(sptep);
			return true;
		}

		(*nr_present)++;
3597
		mark_mmio_spte(vcpu, sptep, gfn, access);
3598 3599 3600 3601 3602 3603
		return true;
	}

	return false;
}

A
Avi Kivity 已提交
3604 3605 3606 3607 3608 3609 3610 3611 3612
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);
}

3613 3614 3615 3616 3617
#define PTTYPE_EPT 18 /* arbitrary */
#define PTTYPE PTTYPE_EPT
#include "paging_tmpl.h"
#undef PTTYPE

A
Avi Kivity 已提交
3618 3619 3620 3621 3622 3623 3624 3625
#define PTTYPE 64
#include "paging_tmpl.h"
#undef PTTYPE

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

3626 3627 3628 3629
static void
__reset_rsvds_bits_mask(struct kvm_vcpu *vcpu,
			struct rsvd_bits_validate *rsvd_check,
			int maxphyaddr, int level, bool nx, bool gbpages,
3630
			bool pse, bool amd)
3631 3632
{
	u64 exb_bit_rsvd = 0;
3633
	u64 gbpages_bit_rsvd = 0;
3634
	u64 nonleaf_bit8_rsvd = 0;
3635

3636
	rsvd_check->bad_mt_xwr = 0;
3637

3638
	if (!nx)
3639
		exb_bit_rsvd = rsvd_bits(63, 63);
3640
	if (!gbpages)
3641
		gbpages_bit_rsvd = rsvd_bits(7, 7);
3642 3643 3644 3645 3646

	/*
	 * Non-leaf PML4Es and PDPEs reserve bit 8 (which would be the G bit for
	 * leaf entries) on AMD CPUs only.
	 */
3647
	if (amd)
3648 3649
		nonleaf_bit8_rsvd = rsvd_bits(8, 8);

3650
	switch (level) {
3651 3652
	case PT32_ROOT_LEVEL:
		/* no rsvd bits for 2 level 4K page table entries */
3653 3654 3655 3656
		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];
3657

3658
		if (!pse) {
3659
			rsvd_check->rsvd_bits_mask[1][1] = 0;
3660 3661 3662
			break;
		}

3663 3664
		if (is_cpuid_PSE36())
			/* 36bits PSE 4MB page */
3665
			rsvd_check->rsvd_bits_mask[1][1] = rsvd_bits(17, 21);
3666 3667
		else
			/* 32 bits PSE 4MB page */
3668
			rsvd_check->rsvd_bits_mask[1][1] = rsvd_bits(13, 21);
3669 3670
		break;
	case PT32E_ROOT_LEVEL:
3671
		rsvd_check->rsvd_bits_mask[0][2] =
3672
			rsvd_bits(maxphyaddr, 63) |
3673
			rsvd_bits(5, 8) | rsvd_bits(1, 2);	/* PDPTE */
3674
		rsvd_check->rsvd_bits_mask[0][1] = exb_bit_rsvd |
3675
			rsvd_bits(maxphyaddr, 62);	/* PDE */
3676
		rsvd_check->rsvd_bits_mask[0][0] = exb_bit_rsvd |
3677
			rsvd_bits(maxphyaddr, 62); 	/* PTE */
3678
		rsvd_check->rsvd_bits_mask[1][1] = exb_bit_rsvd |
3679 3680
			rsvd_bits(maxphyaddr, 62) |
			rsvd_bits(13, 20);		/* large page */
3681 3682
		rsvd_check->rsvd_bits_mask[1][0] =
			rsvd_check->rsvd_bits_mask[0][0];
3683 3684
		break;
	case PT64_ROOT_LEVEL:
3685 3686
		rsvd_check->rsvd_bits_mask[0][3] = exb_bit_rsvd |
			nonleaf_bit8_rsvd | rsvd_bits(7, 7) |
3687
			rsvd_bits(maxphyaddr, 51);
3688 3689
		rsvd_check->rsvd_bits_mask[0][2] = exb_bit_rsvd |
			nonleaf_bit8_rsvd | gbpages_bit_rsvd |
3690
			rsvd_bits(maxphyaddr, 51);
3691 3692 3693 3694 3695 3696 3697
		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 |
3698
			gbpages_bit_rsvd | rsvd_bits(maxphyaddr, 51) |
3699
			rsvd_bits(13, 29);
3700
		rsvd_check->rsvd_bits_mask[1][1] = exb_bit_rsvd |
3701 3702
			rsvd_bits(maxphyaddr, 51) |
			rsvd_bits(13, 20);		/* large page */
3703 3704
		rsvd_check->rsvd_bits_mask[1][0] =
			rsvd_check->rsvd_bits_mask[0][0];
3705 3706 3707 3708
		break;
	}
}

3709 3710 3711 3712 3713 3714
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),
3715
				is_pse(vcpu), guest_cpuid_is_amd(vcpu));
3716 3717
}

3718 3719 3720
static void
__reset_rsvds_bits_mask_ept(struct rsvd_bits_validate *rsvd_check,
			    int maxphyaddr, bool execonly)
3721
{
3722
	u64 bad_mt_xwr;
3723

3724
	rsvd_check->rsvd_bits_mask[0][3] =
3725
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(3, 7);
3726
	rsvd_check->rsvd_bits_mask[0][2] =
3727
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(3, 6);
3728
	rsvd_check->rsvd_bits_mask[0][1] =
3729
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(3, 6);
3730
	rsvd_check->rsvd_bits_mask[0][0] = rsvd_bits(maxphyaddr, 51);
3731 3732

	/* large page */
3733 3734
	rsvd_check->rsvd_bits_mask[1][3] = rsvd_check->rsvd_bits_mask[0][3];
	rsvd_check->rsvd_bits_mask[1][2] =
3735
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(12, 29);
3736
	rsvd_check->rsvd_bits_mask[1][1] =
3737
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(12, 20);
3738
	rsvd_check->rsvd_bits_mask[1][0] = rsvd_check->rsvd_bits_mask[0][0];
3739

3740 3741 3742 3743 3744 3745 3746 3747
	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);
3748
	}
3749
	rsvd_check->bad_mt_xwr = bad_mt_xwr;
3750 3751
}

3752 3753 3754 3755 3756 3757 3758
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);
}

3759 3760 3761 3762 3763 3764 3765 3766
/*
 * 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)
{
3767 3768 3769 3770
	/*
	 * Passing "true" to the last argument is okay; it adds a check
	 * on bit 8 of the SPTEs which KVM doesn't use anyway.
	 */
3771 3772 3773
	__reset_rsvds_bits_mask(vcpu, &context->shadow_zero_check,
				boot_cpu_data.x86_phys_bits,
				context->shadow_root_level, context->nx,
3774 3775
				guest_cpuid_has_gbpages(vcpu), is_pse(vcpu),
				true);
3776 3777 3778
}
EXPORT_SYMBOL_GPL(reset_shadow_zero_bits_mask);

3779 3780 3781 3782 3783 3784
static inline bool boot_cpu_is_amd(void)
{
	WARN_ON_ONCE(!tdp_enabled);
	return shadow_x_mask == 0;
}

3785 3786 3787 3788 3789 3790 3791 3792
/*
 * 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)
{
3793
	if (boot_cpu_is_amd())
3794 3795 3796
		__reset_rsvds_bits_mask(vcpu, &context->shadow_zero_check,
					boot_cpu_data.x86_phys_bits,
					context->shadow_root_level, false,
3797
					cpu_has_gbpages, true, true);
3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816
	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);
}

3817 3818
static void update_permission_bitmask(struct kvm_vcpu *vcpu,
				      struct kvm_mmu *mmu, bool ept)
3819 3820 3821
{
	unsigned bit, byte, pfec;
	u8 map;
F
Feng Wu 已提交
3822
	bool fault, x, w, u, wf, uf, ff, smapf, cr4_smap, cr4_smep, smap = 0;
3823

F
Feng Wu 已提交
3824
	cr4_smep = kvm_read_cr4_bits(vcpu, X86_CR4_SMEP);
F
Feng Wu 已提交
3825
	cr4_smap = kvm_read_cr4_bits(vcpu, X86_CR4_SMAP);
3826 3827 3828 3829 3830 3831
	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
3832 3833 3834 3835 3836 3837
		/*
		 * 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);
3838 3839 3840 3841 3842
		for (bit = 0; bit < 8; ++bit) {
			x = bit & ACC_EXEC_MASK;
			w = bit & ACC_WRITE_MASK;
			u = bit & ACC_USER_MASK;

3843 3844 3845 3846 3847 3848
			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
3849
				x &= !(cr4_smep && u && !uf);
F
Feng Wu 已提交
3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869

				/*
				 * 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;
3870 3871 3872
			} else
				/* Not really needed: no U/S accesses on ept  */
				u = 1;
3873

F
Feng Wu 已提交
3874 3875
			fault = (ff && !x) || (uf && !u) || (wf && !w) ||
				(smapf && smap);
3876 3877 3878 3879 3880 3881
			map |= fault << bit;
		}
		mmu->permissions[byte] = map;
	}
}

A
Avi Kivity 已提交
3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899
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;
}

3900 3901 3902
static void paging64_init_context_common(struct kvm_vcpu *vcpu,
					 struct kvm_mmu *context,
					 int level)
A
Avi Kivity 已提交
3903
{
3904
	context->nx = is_nx(vcpu);
3905
	context->root_level = level;
3906

3907
	reset_rsvds_bits_mask(vcpu, context);
3908
	update_permission_bitmask(vcpu, context, false);
A
Avi Kivity 已提交
3909
	update_last_pte_bitmap(vcpu, context);
A
Avi Kivity 已提交
3910

3911
	MMU_WARN_ON(!is_pae(vcpu));
A
Avi Kivity 已提交
3912 3913
	context->page_fault = paging64_page_fault;
	context->gva_to_gpa = paging64_gva_to_gpa;
3914
	context->sync_page = paging64_sync_page;
M
Marcelo Tosatti 已提交
3915
	context->invlpg = paging64_invlpg;
3916
	context->update_pte = paging64_update_pte;
3917
	context->shadow_root_level = level;
A
Avi Kivity 已提交
3918
	context->root_hpa = INVALID_PAGE;
3919
	context->direct_map = false;
A
Avi Kivity 已提交
3920 3921
}

3922 3923
static void paging64_init_context(struct kvm_vcpu *vcpu,
				  struct kvm_mmu *context)
3924
{
3925
	paging64_init_context_common(vcpu, context, PT64_ROOT_LEVEL);
3926 3927
}

3928 3929
static void paging32_init_context(struct kvm_vcpu *vcpu,
				  struct kvm_mmu *context)
A
Avi Kivity 已提交
3930
{
3931
	context->nx = false;
3932
	context->root_level = PT32_ROOT_LEVEL;
3933

3934
	reset_rsvds_bits_mask(vcpu, context);
3935
	update_permission_bitmask(vcpu, context, false);
A
Avi Kivity 已提交
3936
	update_last_pte_bitmap(vcpu, context);
A
Avi Kivity 已提交
3937 3938 3939

	context->page_fault = paging32_page_fault;
	context->gva_to_gpa = paging32_gva_to_gpa;
3940
	context->sync_page = paging32_sync_page;
M
Marcelo Tosatti 已提交
3941
	context->invlpg = paging32_invlpg;
3942
	context->update_pte = paging32_update_pte;
A
Avi Kivity 已提交
3943
	context->shadow_root_level = PT32E_ROOT_LEVEL;
A
Avi Kivity 已提交
3944
	context->root_hpa = INVALID_PAGE;
3945
	context->direct_map = false;
A
Avi Kivity 已提交
3946 3947
}

3948 3949
static void paging32E_init_context(struct kvm_vcpu *vcpu,
				   struct kvm_mmu *context)
A
Avi Kivity 已提交
3950
{
3951
	paging64_init_context_common(vcpu, context, PT32E_ROOT_LEVEL);
A
Avi Kivity 已提交
3952 3953
}

3954
static void init_kvm_tdp_mmu(struct kvm_vcpu *vcpu)
3955
{
3956
	struct kvm_mmu *context = &vcpu->arch.mmu;
3957

3958
	context->base_role.word = 0;
3959
	context->base_role.smm = is_smm(vcpu);
3960
	context->page_fault = tdp_page_fault;
3961
	context->sync_page = nonpaging_sync_page;
M
Marcelo Tosatti 已提交
3962
	context->invlpg = nonpaging_invlpg;
3963
	context->update_pte = nonpaging_update_pte;
3964
	context->shadow_root_level = kvm_x86_ops->get_tdp_level();
3965
	context->root_hpa = INVALID_PAGE;
3966
	context->direct_map = true;
3967
	context->set_cr3 = kvm_x86_ops->set_tdp_cr3;
3968
	context->get_cr3 = get_cr3;
3969
	context->get_pdptr = kvm_pdptr_read;
3970
	context->inject_page_fault = kvm_inject_page_fault;
3971 3972

	if (!is_paging(vcpu)) {
3973
		context->nx = false;
3974 3975 3976
		context->gva_to_gpa = nonpaging_gva_to_gpa;
		context->root_level = 0;
	} else if (is_long_mode(vcpu)) {
3977
		context->nx = is_nx(vcpu);
3978
		context->root_level = PT64_ROOT_LEVEL;
3979 3980
		reset_rsvds_bits_mask(vcpu, context);
		context->gva_to_gpa = paging64_gva_to_gpa;
3981
	} else if (is_pae(vcpu)) {
3982
		context->nx = is_nx(vcpu);
3983
		context->root_level = PT32E_ROOT_LEVEL;
3984 3985
		reset_rsvds_bits_mask(vcpu, context);
		context->gva_to_gpa = paging64_gva_to_gpa;
3986
	} else {
3987
		context->nx = false;
3988
		context->root_level = PT32_ROOT_LEVEL;
3989 3990
		reset_rsvds_bits_mask(vcpu, context);
		context->gva_to_gpa = paging32_gva_to_gpa;
3991 3992
	}

3993
	update_permission_bitmask(vcpu, context, false);
A
Avi Kivity 已提交
3994
	update_last_pte_bitmap(vcpu, context);
3995
	reset_tdp_shadow_zero_bits_mask(vcpu, context);
3996 3997
}

3998
void kvm_init_shadow_mmu(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
3999
{
4000
	bool smep = kvm_read_cr4_bits(vcpu, X86_CR4_SMEP);
4001
	bool smap = kvm_read_cr4_bits(vcpu, X86_CR4_SMAP);
4002 4003
	struct kvm_mmu *context = &vcpu->arch.mmu;

4004
	MMU_WARN_ON(VALID_PAGE(context->root_hpa));
A
Avi Kivity 已提交
4005 4006

	if (!is_paging(vcpu))
4007
		nonpaging_init_context(vcpu, context);
A
Avi Kivity 已提交
4008
	else if (is_long_mode(vcpu))
4009
		paging64_init_context(vcpu, context);
A
Avi Kivity 已提交
4010
	else if (is_pae(vcpu))
4011
		paging32E_init_context(vcpu, context);
A
Avi Kivity 已提交
4012
	else
4013
		paging32_init_context(vcpu, context);
4014

4015 4016 4017 4018
	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
4019
		= smep && !is_write_protection(vcpu);
4020 4021
	context->base_role.smap_andnot_wp
		= smap && !is_write_protection(vcpu);
4022
	context->base_role.smm = is_smm(vcpu);
4023
	reset_shadow_zero_bits_mask(vcpu, context);
4024 4025 4026
}
EXPORT_SYMBOL_GPL(kvm_init_shadow_mmu);

4027
void kvm_init_shadow_ept_mmu(struct kvm_vcpu *vcpu, bool execonly)
N
Nadav Har'El 已提交
4028
{
4029 4030
	struct kvm_mmu *context = &vcpu->arch.mmu;

4031
	MMU_WARN_ON(VALID_PAGE(context->root_hpa));
N
Nadav Har'El 已提交
4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046

	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);
4047
	reset_ept_shadow_zero_bits_mask(vcpu, context, execonly);
N
Nadav Har'El 已提交
4048 4049 4050
}
EXPORT_SYMBOL_GPL(kvm_init_shadow_ept_mmu);

4051
static void init_kvm_softmmu(struct kvm_vcpu *vcpu)
4052
{
4053 4054 4055 4056 4057 4058 4059
	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
4060 4061
}

4062
static void init_kvm_nested_mmu(struct kvm_vcpu *vcpu)
4063 4064 4065 4066
{
	struct kvm_mmu *g_context = &vcpu->arch.nested_mmu;

	g_context->get_cr3           = get_cr3;
4067
	g_context->get_pdptr         = kvm_pdptr_read;
4068 4069 4070 4071 4072 4073 4074 4075 4076
	g_context->inject_page_fault = kvm_inject_page_fault;

	/*
	 * Note that arch.mmu.gva_to_gpa translates l2_gva to l1_gpa. The
	 * translation of l2_gpa to l1_gpa addresses is done using the
	 * arch.nested_mmu.gva_to_gpa function. Basically the gva_to_gpa
	 * functions between mmu and nested_mmu are swapped.
	 */
	if (!is_paging(vcpu)) {
4077
		g_context->nx = false;
4078 4079 4080
		g_context->root_level = 0;
		g_context->gva_to_gpa = nonpaging_gva_to_gpa_nested;
	} else if (is_long_mode(vcpu)) {
4081
		g_context->nx = is_nx(vcpu);
4082
		g_context->root_level = PT64_ROOT_LEVEL;
4083
		reset_rsvds_bits_mask(vcpu, g_context);
4084 4085
		g_context->gva_to_gpa = paging64_gva_to_gpa_nested;
	} else if (is_pae(vcpu)) {
4086
		g_context->nx = is_nx(vcpu);
4087
		g_context->root_level = PT32E_ROOT_LEVEL;
4088
		reset_rsvds_bits_mask(vcpu, g_context);
4089 4090
		g_context->gva_to_gpa = paging64_gva_to_gpa_nested;
	} else {
4091
		g_context->nx = false;
4092
		g_context->root_level = PT32_ROOT_LEVEL;
4093
		reset_rsvds_bits_mask(vcpu, g_context);
4094 4095 4096
		g_context->gva_to_gpa = paging32_gva_to_gpa_nested;
	}

4097
	update_permission_bitmask(vcpu, g_context, false);
A
Avi Kivity 已提交
4098
	update_last_pte_bitmap(vcpu, g_context);
4099 4100
}

4101
static void init_kvm_mmu(struct kvm_vcpu *vcpu)
4102
{
4103
	if (mmu_is_nested(vcpu))
4104
		init_kvm_nested_mmu(vcpu);
4105
	else if (tdp_enabled)
4106
		init_kvm_tdp_mmu(vcpu);
4107
	else
4108
		init_kvm_softmmu(vcpu);
4109 4110
}

4111
void kvm_mmu_reset_context(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4112
{
4113
	kvm_mmu_unload(vcpu);
4114
	init_kvm_mmu(vcpu);
A
Avi Kivity 已提交
4115
}
4116
EXPORT_SYMBOL_GPL(kvm_mmu_reset_context);
A
Avi Kivity 已提交
4117 4118

int kvm_mmu_load(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4119
{
4120 4121
	int r;

4122
	r = mmu_topup_memory_caches(vcpu);
A
Avi Kivity 已提交
4123 4124
	if (r)
		goto out;
4125
	r = mmu_alloc_roots(vcpu);
4126
	kvm_mmu_sync_roots(vcpu);
4127 4128
	if (r)
		goto out;
4129
	/* set_cr3() should ensure TLB has been flushed */
4130
	vcpu->arch.mmu.set_cr3(vcpu, vcpu->arch.mmu.root_hpa);
4131 4132
out:
	return r;
A
Avi Kivity 已提交
4133
}
A
Avi Kivity 已提交
4134 4135 4136 4137 4138
EXPORT_SYMBOL_GPL(kvm_mmu_load);

void kvm_mmu_unload(struct kvm_vcpu *vcpu)
{
	mmu_free_roots(vcpu);
4139
	WARN_ON(VALID_PAGE(vcpu->arch.mmu.root_hpa));
A
Avi Kivity 已提交
4140
}
4141
EXPORT_SYMBOL_GPL(kvm_mmu_unload);
A
Avi Kivity 已提交
4142

4143
static void mmu_pte_write_new_pte(struct kvm_vcpu *vcpu,
4144 4145
				  struct kvm_mmu_page *sp, u64 *spte,
				  const void *new)
4146
{
4147
	if (sp->role.level != PT_PAGE_TABLE_LEVEL) {
4148 4149
		++vcpu->kvm->stat.mmu_pde_zapped;
		return;
4150
        }
4151

A
Avi Kivity 已提交
4152
	++vcpu->kvm->stat.mmu_pte_updated;
4153
	vcpu->arch.mmu.update_pte(vcpu, sp, spte, new);
4154 4155
}

4156 4157 4158 4159 4160 4161 4162 4163
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;
4164 4165
	old ^= shadow_nx_mask;
	new ^= shadow_nx_mask;
4166 4167 4168
	return (old & ~new & PT64_PERM_MASK) != 0;
}

4169 4170
static void mmu_pte_write_flush_tlb(struct kvm_vcpu *vcpu, bool zap_page,
				    bool remote_flush, bool local_flush)
4171
{
4172 4173 4174 4175
	if (zap_page)
		return;

	if (remote_flush)
4176
		kvm_flush_remote_tlbs(vcpu->kvm);
4177
	else if (local_flush)
4178
		kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
4179 4180
}

4181 4182
static u64 mmu_pte_write_fetch_gpte(struct kvm_vcpu *vcpu, gpa_t *gpa,
				    const u8 *new, int *bytes)
4183
{
4184 4185
	u64 gentry;
	int r;
4186 4187 4188

	/*
	 * Assume that the pte write on a page table of the same type
4189 4190
	 * as the current vcpu paging mode since we update the sptes only
	 * when they have the same mode.
4191
	 */
4192
	if (is_pae(vcpu) && *bytes == 4) {
4193
		/* Handle a 32-bit guest writing two halves of a 64-bit gpte */
4194 4195
		*gpa &= ~(gpa_t)7;
		*bytes = 8;
4196
		r = kvm_vcpu_read_guest(vcpu, *gpa, &gentry, 8);
4197 4198
		if (r)
			gentry = 0;
4199 4200 4201
		new = (const u8 *)&gentry;
	}

4202
	switch (*bytes) {
4203 4204 4205 4206 4207 4208 4209 4210 4211
	case 4:
		gentry = *(const u32 *)new;
		break;
	case 8:
		gentry = *(const u64 *)new;
		break;
	default:
		gentry = 0;
		break;
4212 4213
	}

4214 4215 4216 4217 4218 4219 4220
	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.
 */
4221
static bool detect_write_flooding(struct kvm_mmu_page *sp)
4222
{
4223 4224 4225 4226
	/*
	 * Skip write-flooding detected for the sp whose level is 1, because
	 * it can become unsync, then the guest page is not write-protected.
	 */
4227
	if (sp->role.level == PT_PAGE_TABLE_LEVEL)
4228
		return false;
4229

4230
	return ++sp->write_flooding_count >= 3;
4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246
}

/*
 * 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;
4247 4248 4249 4250 4251 4252 4253 4254

	/*
	 * 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;

4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299
	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;
4300
	bool remote_flush, local_flush, zap_page;
4301 4302 4303 4304 4305 4306 4307
	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;
4308
	mask.smm = 1;
4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331

	/*
	 * 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;
4332
	kvm_mmu_audit(vcpu, AUDIT_PRE_PTE_WRITE);
4333

4334
	for_each_gfn_indirect_valid_sp(vcpu->kvm, sp, gfn) {
4335
		if (detect_write_misaligned(sp, gpa, bytes) ||
4336
		      detect_write_flooding(sp)) {
4337
			zap_page |= !!kvm_mmu_prepare_zap_page(vcpu->kvm, sp,
4338
						     &invalid_list);
A
Avi Kivity 已提交
4339
			++vcpu->kvm->stat.mmu_flooded;
4340 4341
			continue;
		}
4342 4343 4344 4345 4346

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

4347
		local_flush = true;
4348
		while (npte--) {
4349
			entry = *spte;
4350
			mmu_page_zap_pte(vcpu->kvm, sp, spte);
4351 4352
			if (gentry &&
			      !((sp->role.word ^ vcpu->arch.mmu.base_role.word)
4353
			      & mask.word) && rmap_can_add(vcpu))
4354
				mmu_pte_write_new_pte(vcpu, sp, spte, &gentry);
G
Gleb Natapov 已提交
4355
			if (need_remote_flush(entry, *spte))
4356
				remote_flush = true;
4357
			++spte;
4358 4359
		}
	}
4360
	mmu_pte_write_flush_tlb(vcpu, zap_page, remote_flush, local_flush);
4361
	kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
4362
	kvm_mmu_audit(vcpu, AUDIT_POST_PTE_WRITE);
4363
	spin_unlock(&vcpu->kvm->mmu_lock);
4364 4365
}

4366 4367
int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva)
{
4368 4369
	gpa_t gpa;
	int r;
4370

4371
	if (vcpu->arch.mmu.direct_map)
4372 4373
		return 0;

4374
	gpa = kvm_mmu_gva_to_gpa_read(vcpu, gva, NULL);
4375 4376

	r = kvm_mmu_unprotect_page(vcpu->kvm, gpa >> PAGE_SHIFT);
4377

4378
	return r;
4379
}
4380
EXPORT_SYMBOL_GPL(kvm_mmu_unprotect_page_virt);
4381

4382
static void make_mmu_pages_available(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4383
{
4384
	LIST_HEAD(invalid_list);
4385

4386 4387 4388
	if (likely(kvm_mmu_available_pages(vcpu->kvm) >= KVM_MIN_FREE_MMU_PAGES))
		return;

4389 4390 4391
	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
4392

A
Avi Kivity 已提交
4393
		++vcpu->kvm->stat.mmu_recycled;
A
Avi Kivity 已提交
4394
	}
4395
	kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
A
Avi Kivity 已提交
4396 4397
}

4398 4399 4400 4401 4402 4403 4404 4405
static bool is_mmio_page_fault(struct kvm_vcpu *vcpu, gva_t addr)
{
	if (vcpu->arch.mmu.direct_map || mmu_is_nested(vcpu))
		return vcpu_match_mmio_gpa(vcpu, addr);

	return vcpu_match_mmio_gva(vcpu, addr);
}

4406 4407
int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t cr2, u32 error_code,
		       void *insn, int insn_len)
4408
{
4409
	int r, emulation_type = EMULTYPE_RETRY;
4410 4411
	enum emulation_result er;

G
Gleb Natapov 已提交
4412
	r = vcpu->arch.mmu.page_fault(vcpu, cr2, error_code, false);
4413 4414 4415 4416 4417 4418 4419 4420
	if (r < 0)
		goto out;

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

4421 4422 4423 4424
	if (is_mmio_page_fault(vcpu, cr2))
		emulation_type = 0;

	er = x86_emulate_instruction(vcpu, cr2, emulation_type, insn, insn_len);
4425 4426 4427 4428

	switch (er) {
	case EMULATE_DONE:
		return 1;
P
Paolo Bonzini 已提交
4429
	case EMULATE_USER_EXIT:
4430
		++vcpu->stat.mmio_exits;
4431
		/* fall through */
4432
	case EMULATE_FAIL:
4433
		return 0;
4434 4435 4436 4437 4438 4439 4440 4441
	default:
		BUG();
	}
out:
	return r;
}
EXPORT_SYMBOL_GPL(kvm_mmu_page_fault);

M
Marcelo Tosatti 已提交
4442 4443 4444
void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
{
	vcpu->arch.mmu.invlpg(vcpu, gva);
4445
	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
M
Marcelo Tosatti 已提交
4446 4447 4448 4449
	++vcpu->stat.invlpg;
}
EXPORT_SYMBOL_GPL(kvm_mmu_invlpg);

4450 4451 4452 4453 4454 4455
void kvm_enable_tdp(void)
{
	tdp_enabled = true;
}
EXPORT_SYMBOL_GPL(kvm_enable_tdp);

4456 4457 4458 4459 4460 4461
void kvm_disable_tdp(void)
{
	tdp_enabled = false;
}
EXPORT_SYMBOL_GPL(kvm_disable_tdp);

A
Avi Kivity 已提交
4462 4463
static void free_mmu_pages(struct kvm_vcpu *vcpu)
{
4464
	free_page((unsigned long)vcpu->arch.mmu.pae_root);
4465 4466
	if (vcpu->arch.mmu.lm_root != NULL)
		free_page((unsigned long)vcpu->arch.mmu.lm_root);
A
Avi Kivity 已提交
4467 4468 4469 4470
}

static int alloc_mmu_pages(struct kvm_vcpu *vcpu)
{
4471
	struct page *page;
A
Avi Kivity 已提交
4472 4473
	int i;

4474 4475 4476 4477 4478 4479 4480
	/*
	 * 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)
4481 4482
		return -ENOMEM;

4483
	vcpu->arch.mmu.pae_root = page_address(page);
4484
	for (i = 0; i < 4; ++i)
4485
		vcpu->arch.mmu.pae_root[i] = INVALID_PAGE;
4486

A
Avi Kivity 已提交
4487 4488 4489
	return 0;
}

4490
int kvm_mmu_create(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4491
{
4492 4493 4494 4495
	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
4496

4497 4498
	return alloc_mmu_pages(vcpu);
}
A
Avi Kivity 已提交
4499

4500
void kvm_mmu_setup(struct kvm_vcpu *vcpu)
4501
{
4502
	MMU_WARN_ON(VALID_PAGE(vcpu->arch.mmu.root_hpa));
4503

4504
	init_kvm_mmu(vcpu);
A
Avi Kivity 已提交
4505 4506
}

4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575
/* The return value indicates if tlb flush on all vcpus is needed. */
typedef bool (*slot_level_handler) (struct kvm *kvm, unsigned long *rmap);

/* 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
4576 4577 4578 4579
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;
4580
	int i;
X
Xiao Guangrong 已提交
4581 4582

	spin_lock(&kvm->mmu_lock);
4583 4584 4585 4586 4587 4588 4589 4590 4591
	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
4592

4593 4594 4595 4596
			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
4597 4598 4599 4600 4601
	}

	spin_unlock(&kvm->mmu_lock);
}

4602 4603 4604 4605 4606
static bool slot_rmap_write_protect(struct kvm *kvm, unsigned long *rmapp)
{
	return __rmap_write_protect(kvm, rmapp, false);
}

4607 4608
void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
				      struct kvm_memory_slot *memslot)
A
Avi Kivity 已提交
4609
{
4610
	bool flush;
A
Avi Kivity 已提交
4611

4612
	spin_lock(&kvm->mmu_lock);
4613 4614
	flush = slot_handle_all_level(kvm, memslot, slot_rmap_write_protect,
				      false);
4615
	spin_unlock(&kvm->mmu_lock);
4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634

	/*
	 * 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.
	 */
4635 4636
	if (flush)
		kvm_flush_remote_tlbs(kvm);
A
Avi Kivity 已提交
4637
}
4638

4639 4640 4641 4642 4643 4644 4645 4646 4647
static bool kvm_mmu_zap_collapsible_spte(struct kvm *kvm,
		unsigned long *rmapp)
{
	u64 *sptep;
	struct rmap_iterator iter;
	int need_tlb_flush = 0;
	pfn_t pfn;
	struct kvm_mmu_page *sp;

4648 4649
restart:
	for_each_rmap_spte(rmapp, &iter, sptep) {
4650 4651 4652 4653
		sp = page_header(__pa(sptep));
		pfn = spte_to_pfn(*sptep);

		/*
4654 4655 4656 4657 4658
		 * 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.
4659 4660 4661 4662 4663 4664
		 */
		if (sp->role.direct &&
			!kvm_is_reserved_pfn(pfn) &&
			PageTransCompound(pfn_to_page(pfn))) {
			drop_spte(kvm, sptep);
			need_tlb_flush = 1;
4665 4666
			goto restart;
		}
4667 4668 4669 4670 4671 4672
	}

	return need_tlb_flush;
}

void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
4673
				   const struct kvm_memory_slot *memslot)
4674
{
4675
	/* FIXME: const-ify all uses of struct kvm_memory_slot.  */
4676
	spin_lock(&kvm->mmu_lock);
4677 4678
	slot_handle_leaf(kvm, (struct kvm_memory_slot *)memslot,
			 kvm_mmu_zap_collapsible_spte, true);
4679 4680 4681
	spin_unlock(&kvm->mmu_lock);
}

4682 4683 4684
void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
				   struct kvm_memory_slot *memslot)
{
4685
	bool flush;
4686 4687

	spin_lock(&kvm->mmu_lock);
4688
	flush = slot_handle_leaf(kvm, memslot, __rmap_clear_dirty, false);
4689 4690 4691 4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706
	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)
{
4707
	bool flush;
4708 4709

	spin_lock(&kvm->mmu_lock);
4710 4711
	flush = slot_handle_large_level(kvm, memslot, slot_rmap_write_protect,
					false);
4712 4713 4714 4715 4716 4717 4718 4719 4720 4721 4722 4723 4724
	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)
{
4725
	bool flush;
4726 4727

	spin_lock(&kvm->mmu_lock);
4728
	flush = slot_handle_all_level(kvm, memslot, __rmap_set_dirty, false);
4729 4730 4731 4732 4733 4734 4735 4736 4737 4738
	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
4739
#define BATCH_ZAP_PAGES	10
4740 4741 4742
static void kvm_zap_obsolete_pages(struct kvm *kvm)
{
	struct kvm_mmu_page *sp, *node;
X
Xiao Guangrong 已提交
4743
	int batch = 0;
4744 4745 4746 4747

restart:
	list_for_each_entry_safe_reverse(sp, node,
	      &kvm->arch.active_mmu_pages, link) {
X
Xiao Guangrong 已提交
4748 4749
		int ret;

4750 4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764
		/*
		 * 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;

4765 4766 4767 4768
		/*
		 * Need not flush tlb since we only zap the sp with invalid
		 * generation number.
		 */
X
Xiao Guangrong 已提交
4769
		if (batch >= BATCH_ZAP_PAGES &&
4770
		      cond_resched_lock(&kvm->mmu_lock)) {
X
Xiao Guangrong 已提交
4771
			batch = 0;
4772 4773 4774
			goto restart;
		}

4775 4776
		ret = kvm_mmu_prepare_zap_page(kvm, sp,
				&kvm->arch.zapped_obsolete_pages);
X
Xiao Guangrong 已提交
4777 4778 4779
		batch += ret;

		if (ret)
4780 4781 4782
			goto restart;
	}

4783 4784 4785 4786
	/*
	 * Should flush tlb before free page tables since lockless-walking
	 * may use the pages.
	 */
4787
	kvm_mmu_commit_zap_page(kvm, &kvm->arch.zapped_obsolete_pages);
4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801
}

/*
 * 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);
4802
	trace_kvm_mmu_invalidate_zap_all_pages(kvm);
4803 4804
	kvm->arch.mmu_valid_gen++;

4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815
	/*
	 * 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);

4816 4817 4818 4819
	kvm_zap_obsolete_pages(kvm);
	spin_unlock(&kvm->mmu_lock);
}

4820 4821 4822 4823 4824
static bool kvm_has_zapped_obsolete_pages(struct kvm *kvm)
{
	return unlikely(!list_empty_careful(&kvm->arch.zapped_obsolete_pages));
}

4825
void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, struct kvm_memslots *slots)
4826 4827 4828 4829 4830
{
	/*
	 * The very rare case: if the generation-number is round,
	 * zap all shadow pages.
	 */
4831
	if (unlikely((slots->generation & MMIO_GEN_MASK) == 0)) {
4832
		printk_ratelimited(KERN_DEBUG "kvm: zapping shadow pages for mmio generation wraparound\n");
4833
		kvm_mmu_invalidate_zap_all_pages(kvm);
4834
	}
4835 4836
}

4837 4838
static unsigned long
mmu_shrink_scan(struct shrinker *shrink, struct shrink_control *sc)
4839 4840
{
	struct kvm *kvm;
4841
	int nr_to_scan = sc->nr_to_scan;
4842
	unsigned long freed = 0;
4843

4844
	spin_lock(&kvm_lock);
4845 4846

	list_for_each_entry(kvm, &vm_list, vm_list) {
4847
		int idx;
4848
		LIST_HEAD(invalid_list);
4849

4850 4851 4852 4853 4854 4855 4856 4857
		/*
		 * 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;
4858 4859 4860 4861 4862 4863
		/*
		 * 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.
		 */
4864 4865
		if (!kvm->arch.n_used_mmu_pages &&
		      !kvm_has_zapped_obsolete_pages(kvm))
4866 4867
			continue;

4868
		idx = srcu_read_lock(&kvm->srcu);
4869 4870
		spin_lock(&kvm->mmu_lock);

4871 4872 4873 4874 4875 4876
		if (kvm_has_zapped_obsolete_pages(kvm)) {
			kvm_mmu_commit_zap_page(kvm,
			      &kvm->arch.zapped_obsolete_pages);
			goto unlock;
		}

4877 4878
		if (prepare_zap_oldest_mmu_page(kvm, &invalid_list))
			freed++;
4879
		kvm_mmu_commit_zap_page(kvm, &invalid_list);
4880

4881
unlock:
4882
		spin_unlock(&kvm->mmu_lock);
4883
		srcu_read_unlock(&kvm->srcu, idx);
4884

4885 4886 4887 4888 4889
		/*
		 * unfair on small ones
		 * per-vm shrinkers cry out
		 * sadness comes quickly
		 */
4890 4891
		list_move_tail(&kvm->vm_list, &vm_list);
		break;
4892 4893
	}

4894
	spin_unlock(&kvm_lock);
4895 4896 4897 4898 4899 4900
	return freed;
}

static unsigned long
mmu_shrink_count(struct shrinker *shrink, struct shrink_control *sc)
{
4901
	return percpu_counter_read_positive(&kvm_total_used_mmu_pages);
4902 4903 4904
}

static struct shrinker mmu_shrinker = {
4905 4906
	.count_objects = mmu_shrink_count,
	.scan_objects = mmu_shrink_scan,
4907 4908 4909
	.seeks = DEFAULT_SEEKS * 10,
};

I
Ingo Molnar 已提交
4910
static void mmu_destroy_caches(void)
4911
{
4912 4913
	if (pte_list_desc_cache)
		kmem_cache_destroy(pte_list_desc_cache);
4914 4915
	if (mmu_page_header_cache)
		kmem_cache_destroy(mmu_page_header_cache);
4916 4917 4918 4919
}

int kvm_mmu_module_init(void)
{
4920 4921
	pte_list_desc_cache = kmem_cache_create("pte_list_desc",
					    sizeof(struct pte_list_desc),
4922
					    0, 0, NULL);
4923
	if (!pte_list_desc_cache)
4924 4925
		goto nomem;

4926 4927
	mmu_page_header_cache = kmem_cache_create("kvm_mmu_page_header",
						  sizeof(struct kvm_mmu_page),
4928
						  0, 0, NULL);
4929 4930 4931
	if (!mmu_page_header_cache)
		goto nomem;

4932
	if (percpu_counter_init(&kvm_total_used_mmu_pages, 0, GFP_KERNEL))
4933 4934
		goto nomem;

4935 4936
	register_shrinker(&mmu_shrinker);

4937 4938 4939
	return 0;

nomem:
4940
	mmu_destroy_caches();
4941 4942 4943
	return -ENOMEM;
}

4944 4945 4946 4947 4948 4949 4950
/*
 * 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;
4951
	struct kvm_memslots *slots;
4952
	struct kvm_memory_slot *memslot;
4953
	int i;
4954

4955 4956
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
		slots = __kvm_memslots(kvm, i);
4957

4958 4959 4960
		kvm_for_each_memslot(memslot, slots)
			nr_pages += memslot->npages;
	}
4961 4962 4963

	nr_mmu_pages = nr_pages * KVM_PERMILLE_MMU_PAGES / 1000;
	nr_mmu_pages = max(nr_mmu_pages,
4964
			   (unsigned int) KVM_MIN_ALLOC_MMU_PAGES);
4965 4966 4967 4968

	return nr_mmu_pages;
}

4969 4970
void kvm_mmu_destroy(struct kvm_vcpu *vcpu)
{
4971
	kvm_mmu_unload(vcpu);
4972 4973
	free_mmu_pages(vcpu);
	mmu_free_memory_caches(vcpu);
4974 4975 4976 4977 4978 4979 4980
}

void kvm_mmu_module_exit(void)
{
	mmu_destroy_caches();
	percpu_counter_destroy(&kvm_total_used_mmu_pages);
	unregister_shrinker(&mmu_shrinker);
4981 4982
	mmu_audit_disable();
}