mmu.c 130.5 KB
Newer Older
A
Avi Kivity 已提交
1 2 3 4 5 6 7 8 9
/*
 * Kernel-based Virtual Machine driver for Linux
 *
 * This module enables machines with Intel VT-x extensions to run virtual
 * machines without emulation or binary translation.
 *
 * MMU support
 *
 * Copyright (C) 2006 Qumranet, Inc.
N
Nicolas Kaiser 已提交
10
 * Copyright 2010 Red Hat, Inc. and/or its affiliates.
A
Avi Kivity 已提交
11 12 13 14 15 16 17 18 19
 *
 * Authors:
 *   Yaniv Kamay  <yaniv@qumranet.com>
 *   Avi Kivity   <avi@qumranet.com>
 *
 * This work is licensed under the terms of the GNU GPL, version 2.  See
 * the COPYING file in the top-level directory.
 *
 */
A
Avi Kivity 已提交
20

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

27
#include <linux/kvm_host.h>
A
Avi Kivity 已提交
28 29 30 31
#include <linux/types.h>
#include <linux/string.h>
#include <linux/mm.h>
#include <linux/highmem.h>
32 33
#include <linux/moduleparam.h>
#include <linux/export.h>
34
#include <linux/swap.h>
M
Marcelo Tosatti 已提交
35
#include <linux/hugetlb.h>
36
#include <linux/compiler.h>
37
#include <linux/srcu.h>
38
#include <linux/slab.h>
39
#include <linux/uaccess.h>
40
#include <linux/hash.h>
A
Avi Kivity 已提交
41

A
Avi Kivity 已提交
42 43
#include <asm/page.h>
#include <asm/cmpxchg.h>
44
#include <asm/io.h>
45
#include <asm/vmx.h>
46
#include <asm/kvm_page_track.h>
A
Avi Kivity 已提交
47

48 49 50 51 52 53 54
/*
 * 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.
 */
55
bool tdp_enabled = false;
56

57 58 59 60
enum {
	AUDIT_PRE_PAGE_FAULT,
	AUDIT_POST_PAGE_FAULT,
	AUDIT_PRE_PTE_WRITE,
61 62 63
	AUDIT_POST_PTE_WRITE,
	AUDIT_PRE_SYNC,
	AUDIT_POST_SYNC
64
};
65

66
#undef MMU_DEBUG
67 68

#ifdef MMU_DEBUG
69 70
static bool dbg = 0;
module_param(dbg, bool, 0644);
71 72 73

#define pgprintk(x...) do { if (dbg) printk(x); } while (0)
#define rmap_printk(x...) do { if (dbg) printk(x); } while (0)
74
#define MMU_WARN_ON(x) WARN_ON(x)
75 76 77
#else
#define pgprintk(x...) do { } while (0)
#define rmap_printk(x...) do { } while (0)
78
#define MMU_WARN_ON(x) do { } while (0)
79
#endif
A
Avi Kivity 已提交
80

81 82
#define PTE_PREFETCH_NUM		8

83
#define PT_FIRST_AVAIL_BITS_SHIFT 10
A
Avi Kivity 已提交
84 85 86 87 88
#define PT64_SECOND_AVAIL_BITS_SHIFT 52

#define PT64_LEVEL_BITS 9

#define PT64_LEVEL_SHIFT(level) \
M
Mike Day 已提交
89
		(PAGE_SHIFT + (level - 1) * PT64_LEVEL_BITS)
A
Avi Kivity 已提交
90 91 92 93 94 95 96 97

#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
98
		(PAGE_SHIFT + (level - 1) * PT32_LEVEL_BITS)
A
Avi Kivity 已提交
99

100 101 102
#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
103 104 105 106 107

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


108
#define PT64_BASE_ADDR_MASK (((1ULL << 52) - 1) & ~(u64)(PAGE_SIZE-1))
A
Avi Kivity 已提交
109 110
#define PT64_DIR_BASE_ADDR_MASK \
	(PT64_BASE_ADDR_MASK & ~((1ULL << (PAGE_SHIFT + PT64_LEVEL_BITS)) - 1))
111 112 113 114 115 116
#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
117 118 119 120

#define PT32_BASE_ADDR_MASK PAGE_MASK
#define PT32_DIR_BASE_ADDR_MASK \
	(PAGE_MASK & ~((1ULL << (PAGE_SHIFT + PT32_LEVEL_BITS)) - 1))
121 122 123
#define PT32_LVL_ADDR_MASK(level) \
	(PAGE_MASK & ~((1ULL << (PAGE_SHIFT + (((level) - 1) \
					    * PT32_LEVEL_BITS))) - 1))
A
Avi Kivity 已提交
124

125 126
#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
127

128 129 130 131 132
#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)

133 134
#include <trace/events/kvm.h>

135 136 137
#define CREATE_TRACE_POINTS
#include "mmutrace.h"

138 139
#define SPTE_HOST_WRITEABLE	(1ULL << PT_FIRST_AVAIL_BITS_SHIFT)
#define SPTE_MMU_WRITEABLE	(1ULL << (PT_FIRST_AVAIL_BITS_SHIFT + 1))
140

141 142
#define SHADOW_PT_INDEX(addr, level) PT64_INDEX(addr, level)

143 144 145
/* make pte_list_desc fit well in cache line */
#define PTE_LIST_EXT 3

146 147 148
struct pte_list_desc {
	u64 *sptes[PTE_LIST_EXT];
	struct pte_list_desc *more;
149 150
};

151 152 153 154
struct kvm_shadow_walk_iterator {
	u64 addr;
	hpa_t shadow_addr;
	u64 *sptep;
155
	int level;
156 157 158 159 160 161 162 163
	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)))

164 165 166 167 168 169
#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))

170
static struct kmem_cache *pte_list_desc_cache;
171
static struct kmem_cache *mmu_page_header_cache;
172
static struct percpu_counter kvm_total_used_mmu_pages;
173

S
Sheng Yang 已提交
174 175 176 177 178
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;
179
static u64 __read_mostly shadow_mmio_mask;
180
static u64 __read_mostly shadow_present_mask;
181 182

static void mmu_spte_set(u64 *sptep, u64 spte);
183
static void mmu_free_roots(struct kvm_vcpu *vcpu);
184 185 186 187 188 189 190

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

191
/*
192 193 194 195 196 197 198
 * 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.
199
 */
200
#define MMIO_SPTE_GEN_LOW_SHIFT		2
201 202
#define MMIO_SPTE_GEN_HIGH_SHIFT	52

203 204 205
#define MMIO_GEN_SHIFT			20
#define MMIO_GEN_LOW_SHIFT		10
#define MMIO_GEN_LOW_MASK		((1 << MMIO_GEN_LOW_SHIFT) - 2)
206
#define MMIO_GEN_MASK			((1 << MMIO_GEN_SHIFT) - 1)
207 208 209 210 211

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

T
Tiejun Chen 已提交
212
	WARN_ON(gen & ~MMIO_GEN_MASK);
213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229

	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;
}

230
static unsigned int kvm_current_mmio_generation(struct kvm_vcpu *vcpu)
231
{
232
	return kvm_vcpu_memslots(vcpu)->generation & MMIO_GEN_MASK;
233 234
}

235
static void mark_mmio_spte(struct kvm_vcpu *vcpu, u64 *sptep, u64 gfn,
236
			   unsigned access)
237
{
238
	unsigned int gen = kvm_current_mmio_generation(vcpu);
239
	u64 mask = generation_mmio_spte_mask(gen);
240

241
	access &= ACC_WRITE_MASK | ACC_USER_MASK;
242 243
	mask |= shadow_mmio_mask | access | gfn << PAGE_SHIFT;

244
	trace_mark_mmio_spte(sptep, gfn, access, gen);
245
	mmu_spte_set(sptep, mask);
246 247 248 249 250 251 252 253 254
}

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
255
	u64 mask = generation_mmio_spte_mask(MMIO_GEN_MASK) | shadow_mmio_mask;
256
	return (spte & ~mask) >> PAGE_SHIFT;
257 258 259 260
}

static unsigned get_mmio_spte_access(u64 spte)
{
T
Tiejun Chen 已提交
261
	u64 mask = generation_mmio_spte_mask(MMIO_GEN_MASK) | shadow_mmio_mask;
262
	return (spte & ~mask) & ~PAGE_MASK;
263 264
}

265
static bool set_mmio_spte(struct kvm_vcpu *vcpu, u64 *sptep, gfn_t gfn,
D
Dan Williams 已提交
266
			  kvm_pfn_t pfn, unsigned access)
267 268
{
	if (unlikely(is_noslot_pfn(pfn))) {
269
		mark_mmio_spte(vcpu, sptep, gfn, access);
270 271 272 273 274
		return true;
	}

	return false;
}
275

276
static bool check_mmio_spte(struct kvm_vcpu *vcpu, u64 spte)
277
{
278 279
	unsigned int kvm_gen, spte_gen;

280
	kvm_gen = kvm_current_mmio_generation(vcpu);
281 282 283 284
	spte_gen = get_mmio_spte_generation(spte);

	trace_check_mmio_spte(spte, kvm_gen, spte_gen);
	return likely(kvm_gen == spte_gen);
285 286
}

S
Sheng Yang 已提交
287
void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
288
		u64 dirty_mask, u64 nx_mask, u64 x_mask, u64 p_mask)
S
Sheng Yang 已提交
289 290 291 292 293 294
{
	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;
295
	shadow_present_mask = p_mask;
S
Sheng Yang 已提交
296 297 298
}
EXPORT_SYMBOL_GPL(kvm_mmu_set_mask_ptes);

A
Avi Kivity 已提交
299 300 301 302 303
static int is_cpuid_PSE36(void)
{
	return 1;
}

304 305
static int is_nx(struct kvm_vcpu *vcpu)
{
306
	return vcpu->arch.efer & EFER_NX;
307 308
}

309 310
static int is_shadow_present_pte(u64 pte)
{
311
	return (pte & 0xFFFFFFFFull) && !is_mmio_spte(pte);
312 313
}

M
Marcelo Tosatti 已提交
314 315 316 317 318
static int is_large_pte(u64 pte)
{
	return pte & PT_PAGE_SIZE_MASK;
}

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;
}

D
Dan Williams 已提交
328
static kvm_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
	WRITE_ONCE(*sptep, spte);
344 345
}

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

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
	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();

397
	WRITE_ONCE(ssptep->spte_low, sspte.spte_low);
398 399
}

400 401 402 403 404 405 406
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;

407
	WRITE_ONCE(ssptep->spte_low, sspte.spte_low);
408 409 410 411 412 413 414 415

	/*
	 * 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
static bool spte_can_locklessly_be_made_writable(u64 spte)
478
{
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
	 * Always atomically update spte if it can be updated
487 488 489 490
	 * 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.
	 */
491
	if (spte_can_locklessly_be_made_writable(spte))
492 493
		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
static bool is_accessed_spte(u64 spte)
508
{
509 510
	return shadow_accessed_mask ? spte & shadow_accessed_mask
				    : true;
511 512
}

513
static bool is_dirty_spte(u64 spte)
514
{
515 516
	return shadow_dirty_mask ? spte & shadow_dirty_mask
				 : spte & PT_WRITABLE_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);
}

531 532 533
/*
 * Update the SPTE (excluding the PFN), but do not track changes in its
 * accessed/dirty status.
534
 */
535
static u64 mmu_spte_update_no_track(u64 *sptep, u64 new_spte)
536
{
537
	u64 old_spte = *sptep;
538

539
	WARN_ON(!is_shadow_present_pte(new_spte));
540

541 542
	if (!is_shadow_present_pte(old_spte)) {
		mmu_spte_set(sptep, new_spte);
543
		return old_spte;
544
	}
545

546
	if (!spte_has_volatile_bits(old_spte))
547
		__update_clear_spte_fast(sptep, new_spte);
548
	else
549
		old_spte = __update_clear_spte_slow(sptep, new_spte);
550

551 552
	WARN_ON(spte_to_pfn(old_spte) != spte_to_pfn(new_spte));

553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574
	return old_spte;
}

/* Rules for using mmu_spte_update:
 * Update the state bits, it means the mapped pfn is not changed.
 *
 * 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.
 *
 * Returns true if the TLB needs to be flushed
 */
static bool mmu_spte_update(u64 *sptep, u64 new_spte)
{
	bool flush = false;
	u64 old_spte = mmu_spte_update_no_track(sptep, new_spte);

	if (!is_shadow_present_pte(old_spte))
		return false;

575 576
	/*
	 * For the spte updated out of mmu-lock is safe, since
577
	 * we always atomically update it, see the comments in
578 579
	 * spte_has_volatile_bits().
	 */
580
	if (spte_can_locklessly_be_made_writable(old_spte) &&
581
	      !is_writable_pte(new_spte))
582
		flush = true;
583

584
	/*
585
	 * Flush TLB when accessed/dirty states are changed in the page tables,
586 587 588
	 * to guarantee consistency between TLB and page tables.
	 */

589 590
	if (is_accessed_spte(old_spte) && !is_accessed_spte(new_spte)) {
		flush = true;
591
		kvm_set_pfn_accessed(spte_to_pfn(old_spte));
592 593 594 595
	}

	if (is_dirty_spte(old_spte) && !is_dirty_spte(new_spte)) {
		flush = true;
596
		kvm_set_pfn_dirty(spte_to_pfn(old_spte));
597
	}
598

599
	return flush;
600 601
}

602 603 604 605
/*
 * 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.
606
 * Returns non-zero if the PTE was previously valid.
607 608 609
 */
static int mmu_spte_clear_track_bits(u64 *sptep)
{
D
Dan Williams 已提交
610
	kvm_pfn_t pfn;
611 612 613
	u64 old_spte = *sptep;

	if (!spte_has_volatile_bits(old_spte))
614
		__update_clear_spte_fast(sptep, 0ull);
615
	else
616
		old_spte = __update_clear_spte_slow(sptep, 0ull);
617

618
	if (!is_shadow_present_pte(old_spte))
619 620 621
		return 0;

	pfn = spte_to_pfn(old_spte);
622 623 624 625 626 627

	/*
	 * KVM does not hold the refcount of the page used by
	 * kvm mmu, before reclaiming the page, we should
	 * unmap it from mmu first.
	 */
628
	WARN_ON(!kvm_is_reserved_pfn(pfn) && !page_count(pfn_to_page(pfn)));
629

630
	if (is_accessed_spte(old_spte))
631
		kvm_set_pfn_accessed(pfn);
632 633

	if (is_dirty_spte(old_spte))
634
		kvm_set_pfn_dirty(pfn);
635

636 637 638 639 640 641 642 643 644 645
	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)
{
646
	__update_clear_spte_fast(sptep, 0ull);
647 648
}

649 650 651 652 653 654 655
static u64 mmu_spte_get_lockless(u64 *sptep)
{
	return __get_spte_lockless(sptep);
}

static void walk_shadow_page_lockless_begin(struct kvm_vcpu *vcpu)
{
656 657 658 659 660
	/*
	 * Prevent page table teardown by making any free-er wait during
	 * kvm_flush_remote_tlbs() IPI to all active vcpus.
	 */
	local_irq_disable();
661

662 663 664 665
	/*
	 * Make sure a following spte read is not reordered ahead of the write
	 * to vcpu->mode.
	 */
666
	smp_store_mb(vcpu->mode, READING_SHADOW_PAGE_TABLES);
667 668 669 670
}

static void walk_shadow_page_lockless_end(struct kvm_vcpu *vcpu)
{
671 672 673 674 675
	/*
	 * 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.
	 */
676
	smp_store_release(&vcpu->mode, OUTSIDE_GUEST_MODE);
677
	local_irq_enable();
678 679
}

680
static int mmu_topup_memory_cache(struct kvm_mmu_memory_cache *cache,
681
				  struct kmem_cache *base_cache, int min)
682 683 684 685
{
	void *obj;

	if (cache->nobjs >= min)
686
		return 0;
687
	while (cache->nobjs < ARRAY_SIZE(cache->objects)) {
688
		obj = kmem_cache_zalloc(base_cache, GFP_KERNEL);
689
		if (!obj)
690
			return -ENOMEM;
691 692
		cache->objects[cache->nobjs++] = obj;
	}
693
	return 0;
694 695
}

696 697 698 699 700
static int mmu_memory_cache_free_objects(struct kvm_mmu_memory_cache *cache)
{
	return cache->nobjs;
}

701 702
static void mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc,
				  struct kmem_cache *cache)
703 704
{
	while (mc->nobjs)
705
		kmem_cache_free(cache, mc->objects[--mc->nobjs]);
706 707
}

A
Avi Kivity 已提交
708
static int mmu_topup_memory_cache_page(struct kvm_mmu_memory_cache *cache,
709
				       int min)
A
Avi Kivity 已提交
710
{
711
	void *page;
A
Avi Kivity 已提交
712 713 714 715

	if (cache->nobjs >= min)
		return 0;
	while (cache->nobjs < ARRAY_SIZE(cache->objects)) {
716
		page = (void *)__get_free_page(GFP_KERNEL);
A
Avi Kivity 已提交
717 718
		if (!page)
			return -ENOMEM;
719
		cache->objects[cache->nobjs++] = page;
A
Avi Kivity 已提交
720 721 722 723 724 725 726
	}
	return 0;
}

static void mmu_free_memory_cache_page(struct kvm_mmu_memory_cache *mc)
{
	while (mc->nobjs)
727
		free_page((unsigned long)mc->objects[--mc->nobjs]);
A
Avi Kivity 已提交
728 729
}

730
static int mmu_topup_memory_caches(struct kvm_vcpu *vcpu)
731
{
732 733
	int r;

734
	r = mmu_topup_memory_cache(&vcpu->arch.mmu_pte_list_desc_cache,
735
				   pte_list_desc_cache, 8 + PTE_PREFETCH_NUM);
736 737
	if (r)
		goto out;
738
	r = mmu_topup_memory_cache_page(&vcpu->arch.mmu_page_cache, 8);
739 740
	if (r)
		goto out;
741
	r = mmu_topup_memory_cache(&vcpu->arch.mmu_page_header_cache,
742
				   mmu_page_header_cache, 4);
743 744
out:
	return r;
745 746 747 748
}

static void mmu_free_memory_caches(struct kvm_vcpu *vcpu)
{
749 750
	mmu_free_memory_cache(&vcpu->arch.mmu_pte_list_desc_cache,
				pte_list_desc_cache);
751
	mmu_free_memory_cache_page(&vcpu->arch.mmu_page_cache);
752 753
	mmu_free_memory_cache(&vcpu->arch.mmu_page_header_cache,
				mmu_page_header_cache);
754 755
}

756
static void *mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc)
757 758 759 760 761 762 763 764
{
	void *p;

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

765
static struct pte_list_desc *mmu_alloc_pte_list_desc(struct kvm_vcpu *vcpu)
766
{
767
	return mmu_memory_cache_alloc(&vcpu->arch.mmu_pte_list_desc_cache);
768 769
}

770
static void mmu_free_pte_list_desc(struct pte_list_desc *pte_list_desc)
771
{
772
	kmem_cache_free(pte_list_desc_cache, pte_list_desc);
773 774
}

775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790
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
791
/*
792 793
 * 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
794
 */
795 796 797
static struct kvm_lpage_info *lpage_info_slot(gfn_t gfn,
					      struct kvm_memory_slot *slot,
					      int level)
M
Marcelo Tosatti 已提交
798 799 800
{
	unsigned long idx;

801
	idx = gfn_to_index(gfn, slot->base_gfn, level);
802
	return &slot->arch.lpage_info[level - 2][idx];
M
Marcelo Tosatti 已提交
803 804
}

805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827
static void update_gfn_disallow_lpage_count(struct kvm_memory_slot *slot,
					    gfn_t gfn, int count)
{
	struct kvm_lpage_info *linfo;
	int i;

	for (i = PT_DIRECTORY_LEVEL; i <= PT_MAX_HUGEPAGE_LEVEL; ++i) {
		linfo = lpage_info_slot(gfn, slot, i);
		linfo->disallow_lpage += count;
		WARN_ON(linfo->disallow_lpage < 0);
	}
}

void kvm_mmu_gfn_disallow_lpage(struct kvm_memory_slot *slot, gfn_t gfn)
{
	update_gfn_disallow_lpage_count(slot, gfn, 1);
}

void kvm_mmu_gfn_allow_lpage(struct kvm_memory_slot *slot, gfn_t gfn)
{
	update_gfn_disallow_lpage_count(slot, gfn, -1);
}

828
static void account_shadowed(struct kvm *kvm, struct kvm_mmu_page *sp)
M
Marcelo Tosatti 已提交
829
{
830
	struct kvm_memslots *slots;
831
	struct kvm_memory_slot *slot;
832
	gfn_t gfn;
M
Marcelo Tosatti 已提交
833

834
	kvm->arch.indirect_shadow_pages++;
835
	gfn = sp->gfn;
836 837
	slots = kvm_memslots_for_spte_role(kvm, sp->role);
	slot = __gfn_to_memslot(slots, gfn);
838 839 840 841 842 843

	/* the non-leaf shadow pages are keeping readonly. */
	if (sp->role.level > PT_PAGE_TABLE_LEVEL)
		return kvm_slot_page_track_add_page(kvm, slot, gfn,
						    KVM_PAGE_TRACK_WRITE);

844
	kvm_mmu_gfn_disallow_lpage(slot, gfn);
M
Marcelo Tosatti 已提交
845 846
}

847
static void unaccount_shadowed(struct kvm *kvm, struct kvm_mmu_page *sp)
M
Marcelo Tosatti 已提交
848
{
849
	struct kvm_memslots *slots;
850
	struct kvm_memory_slot *slot;
851
	gfn_t gfn;
M
Marcelo Tosatti 已提交
852

853
	kvm->arch.indirect_shadow_pages--;
854
	gfn = sp->gfn;
855 856
	slots = kvm_memslots_for_spte_role(kvm, sp->role);
	slot = __gfn_to_memslot(slots, gfn);
857 858 859 860
	if (sp->role.level > PT_PAGE_TABLE_LEVEL)
		return kvm_slot_page_track_remove_page(kvm, slot, gfn,
						       KVM_PAGE_TRACK_WRITE);

861
	kvm_mmu_gfn_allow_lpage(slot, gfn);
M
Marcelo Tosatti 已提交
862 863
}

864 865
static bool __mmu_gfn_lpage_is_disallowed(gfn_t gfn, int level,
					  struct kvm_memory_slot *slot)
M
Marcelo Tosatti 已提交
866
{
867
	struct kvm_lpage_info *linfo;
M
Marcelo Tosatti 已提交
868 869

	if (slot) {
870
		linfo = lpage_info_slot(gfn, slot, level);
871
		return !!linfo->disallow_lpage;
M
Marcelo Tosatti 已提交
872 873
	}

874
	return true;
M
Marcelo Tosatti 已提交
875 876
}

877 878
static bool mmu_gfn_lpage_is_disallowed(struct kvm_vcpu *vcpu, gfn_t gfn,
					int level)
879 880 881 882
{
	struct kvm_memory_slot *slot;

	slot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);
883
	return __mmu_gfn_lpage_is_disallowed(gfn, level, slot);
884 885
}

886
static int host_mapping_level(struct kvm *kvm, gfn_t gfn)
M
Marcelo Tosatti 已提交
887
{
J
Joerg Roedel 已提交
888
	unsigned long page_size;
889
	int i, ret = 0;
M
Marcelo Tosatti 已提交
890

J
Joerg Roedel 已提交
891
	page_size = kvm_host_page_size(kvm, gfn);
M
Marcelo Tosatti 已提交
892

893
	for (i = PT_PAGE_TABLE_LEVEL; i <= PT_MAX_HUGEPAGE_LEVEL; ++i) {
894 895 896 897 898 899
		if (page_size >= KVM_HPAGE_SIZE(i))
			ret = i;
		else
			break;
	}

900
	return ret;
M
Marcelo Tosatti 已提交
901 902
}

903 904 905 906 907 908 909 910 911 912 913
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;
}

914 915 916
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
917 918
{
	struct kvm_memory_slot *slot;
919

920
	slot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);
921
	if (!memslot_valid_for_gpte(slot, no_dirty_log))
922 923 924 925 926
		slot = NULL;

	return slot;
}

927 928
static int mapping_level(struct kvm_vcpu *vcpu, gfn_t large_gfn,
			 bool *force_pt_level)
929 930
{
	int host_level, level, max_level;
931 932
	struct kvm_memory_slot *slot;

933 934
	if (unlikely(*force_pt_level))
		return PT_PAGE_TABLE_LEVEL;
M
Marcelo Tosatti 已提交
935

936 937
	slot = kvm_vcpu_gfn_to_memslot(vcpu, large_gfn);
	*force_pt_level = !memslot_valid_for_gpte(slot, true);
938 939 940
	if (unlikely(*force_pt_level))
		return PT_PAGE_TABLE_LEVEL;

941 942 943 944 945
	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
946
	max_level = min(kvm_x86_ops->get_lpage_level(), host_level);
947 948

	for (level = PT_DIRECTORY_LEVEL; level <= max_level; ++level)
949
		if (__mmu_gfn_lpage_is_disallowed(large_gfn, level, slot))
950 951 952
			break;

	return level - 1;
M
Marcelo Tosatti 已提交
953 954
}

955
/*
956
 * About rmap_head encoding:
957
 *
958 959
 * If the bit zero of rmap_head->val is clear, then it points to the only spte
 * in this rmap chain. Otherwise, (rmap_head->val & ~1) points to a struct
960
 * pte_list_desc containing more mappings.
961 962 963 964
 */

/*
 * Returns the number of pointers in the rmap chain, not counting the new one.
965
 */
966
static int pte_list_add(struct kvm_vcpu *vcpu, u64 *spte,
967
			struct kvm_rmap_head *rmap_head)
968
{
969
	struct pte_list_desc *desc;
970
	int i, count = 0;
971

972
	if (!rmap_head->val) {
973
		rmap_printk("pte_list_add: %p %llx 0->1\n", spte, *spte);
974 975
		rmap_head->val = (unsigned long)spte;
	} else if (!(rmap_head->val & 1)) {
976 977
		rmap_printk("pte_list_add: %p %llx 1->many\n", spte, *spte);
		desc = mmu_alloc_pte_list_desc(vcpu);
978
		desc->sptes[0] = (u64 *)rmap_head->val;
A
Avi Kivity 已提交
979
		desc->sptes[1] = spte;
980
		rmap_head->val = (unsigned long)desc | 1;
981
		++count;
982
	} else {
983
		rmap_printk("pte_list_add: %p %llx many->many\n", spte, *spte);
984
		desc = (struct pte_list_desc *)(rmap_head->val & ~1ul);
985
		while (desc->sptes[PTE_LIST_EXT-1] && desc->more) {
986
			desc = desc->more;
987
			count += PTE_LIST_EXT;
988
		}
989 990
		if (desc->sptes[PTE_LIST_EXT-1]) {
			desc->more = mmu_alloc_pte_list_desc(vcpu);
991 992
			desc = desc->more;
		}
A
Avi Kivity 已提交
993
		for (i = 0; desc->sptes[i]; ++i)
994
			++count;
A
Avi Kivity 已提交
995
		desc->sptes[i] = spte;
996
	}
997
	return count;
998 999
}

1000
static void
1001 1002 1003
pte_list_desc_remove_entry(struct kvm_rmap_head *rmap_head,
			   struct pte_list_desc *desc, int i,
			   struct pte_list_desc *prev_desc)
1004 1005 1006
{
	int j;

1007
	for (j = PTE_LIST_EXT - 1; !desc->sptes[j] && j > i; --j)
1008
		;
A
Avi Kivity 已提交
1009 1010
	desc->sptes[i] = desc->sptes[j];
	desc->sptes[j] = NULL;
1011 1012 1013
	if (j != 0)
		return;
	if (!prev_desc && !desc->more)
1014
		rmap_head->val = (unsigned long)desc->sptes[0];
1015 1016 1017 1018
	else
		if (prev_desc)
			prev_desc->more = desc->more;
		else
1019
			rmap_head->val = (unsigned long)desc->more | 1;
1020
	mmu_free_pte_list_desc(desc);
1021 1022
}

1023
static void pte_list_remove(u64 *spte, struct kvm_rmap_head *rmap_head)
1024
{
1025 1026
	struct pte_list_desc *desc;
	struct pte_list_desc *prev_desc;
1027 1028
	int i;

1029
	if (!rmap_head->val) {
1030
		printk(KERN_ERR "pte_list_remove: %p 0->BUG\n", spte);
1031
		BUG();
1032
	} else if (!(rmap_head->val & 1)) {
1033
		rmap_printk("pte_list_remove:  %p 1->0\n", spte);
1034
		if ((u64 *)rmap_head->val != spte) {
1035
			printk(KERN_ERR "pte_list_remove:  %p 1->BUG\n", spte);
1036 1037
			BUG();
		}
1038
		rmap_head->val = 0;
1039
	} else {
1040
		rmap_printk("pte_list_remove:  %p many->many\n", spte);
1041
		desc = (struct pte_list_desc *)(rmap_head->val & ~1ul);
1042 1043
		prev_desc = NULL;
		while (desc) {
1044
			for (i = 0; i < PTE_LIST_EXT && desc->sptes[i]; ++i) {
A
Avi Kivity 已提交
1045
				if (desc->sptes[i] == spte) {
1046 1047
					pte_list_desc_remove_entry(rmap_head,
							desc, i, prev_desc);
1048 1049
					return;
				}
1050
			}
1051 1052 1053
			prev_desc = desc;
			desc = desc->more;
		}
1054
		pr_err("pte_list_remove: %p many->many\n", spte);
1055 1056 1057 1058
		BUG();
	}
}

1059 1060
static struct kvm_rmap_head *__gfn_to_rmap(gfn_t gfn, int level,
					   struct kvm_memory_slot *slot)
1061
{
1062
	unsigned long idx;
1063

1064
	idx = gfn_to_index(gfn, slot->base_gfn, level);
1065
	return &slot->arch.rmap[level - PT_PAGE_TABLE_LEVEL][idx];
1066 1067
}

1068 1069
static struct kvm_rmap_head *gfn_to_rmap(struct kvm *kvm, gfn_t gfn,
					 struct kvm_mmu_page *sp)
1070
{
1071
	struct kvm_memslots *slots;
1072 1073
	struct kvm_memory_slot *slot;

1074 1075
	slots = kvm_memslots_for_spte_role(kvm, sp->role);
	slot = __gfn_to_memslot(slots, gfn);
1076
	return __gfn_to_rmap(gfn, sp->role.level, slot);
1077 1078
}

1079 1080 1081 1082 1083 1084 1085 1086
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);
}

1087 1088 1089
static int rmap_add(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn)
{
	struct kvm_mmu_page *sp;
1090
	struct kvm_rmap_head *rmap_head;
1091 1092 1093

	sp = page_header(__pa(spte));
	kvm_mmu_page_set_gfn(sp, spte - sp->spt, gfn);
1094 1095
	rmap_head = gfn_to_rmap(vcpu->kvm, gfn, sp);
	return pte_list_add(vcpu, spte, rmap_head);
1096 1097 1098 1099 1100 1101
}

static void rmap_remove(struct kvm *kvm, u64 *spte)
{
	struct kvm_mmu_page *sp;
	gfn_t gfn;
1102
	struct kvm_rmap_head *rmap_head;
1103 1104 1105

	sp = page_header(__pa(spte));
	gfn = kvm_mmu_page_get_gfn(sp, spte - sp->spt);
1106 1107
	rmap_head = gfn_to_rmap(kvm, gfn, sp);
	pte_list_remove(spte, rmap_head);
1108 1109
}

1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126
/*
 * 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.
 */
1127 1128
static u64 *rmap_get_first(struct kvm_rmap_head *rmap_head,
			   struct rmap_iterator *iter)
1129
{
1130 1131
	u64 *sptep;

1132
	if (!rmap_head->val)
1133 1134
		return NULL;

1135
	if (!(rmap_head->val & 1)) {
1136
		iter->desc = NULL;
1137 1138
		sptep = (u64 *)rmap_head->val;
		goto out;
1139 1140
	}

1141
	iter->desc = (struct pte_list_desc *)(rmap_head->val & ~1ul);
1142
	iter->pos = 0;
1143 1144 1145 1146
	sptep = iter->desc->sptes[iter->pos];
out:
	BUG_ON(!is_shadow_present_pte(*sptep));
	return sptep;
1147 1148 1149 1150 1151 1152 1153 1154 1155
}

/*
 * 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)
{
1156 1157
	u64 *sptep;

1158 1159 1160 1161 1162
	if (iter->desc) {
		if (iter->pos < PTE_LIST_EXT - 1) {
			++iter->pos;
			sptep = iter->desc->sptes[iter->pos];
			if (sptep)
1163
				goto out;
1164 1165 1166 1167 1168 1169 1170
		}

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

		if (iter->desc) {
			iter->pos = 0;
			/* desc->sptes[0] cannot be NULL */
1171 1172
			sptep = iter->desc->sptes[iter->pos];
			goto out;
1173 1174 1175 1176
		}
	}

	return NULL;
1177 1178 1179
out:
	BUG_ON(!is_shadow_present_pte(*sptep));
	return sptep;
1180 1181
}

1182 1183
#define for_each_rmap_spte(_rmap_head_, _iter_, _spte_)			\
	for (_spte_ = rmap_get_first(_rmap_head_, _iter_);		\
1184
	     _spte_; _spte_ = rmap_get_next(_iter_))
1185

1186
static void drop_spte(struct kvm *kvm, u64 *sptep)
1187
{
1188
	if (mmu_spte_clear_track_bits(sptep))
1189
		rmap_remove(kvm, sptep);
A
Avi Kivity 已提交
1190 1191
}

1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212

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);
}

/*
1213
 * Write-protect on the specified @sptep, @pt_protect indicates whether
1214
 * spte write-protection is caused by protecting shadow page table.
1215
 *
T
Tiejun Chen 已提交
1216
 * Note: write protection is difference between dirty logging and spte
1217 1218 1219 1220 1221
 * 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.
1222
 *
1223
 * Return true if tlb need be flushed.
1224
 */
1225
static bool spte_write_protect(u64 *sptep, bool pt_protect)
1226 1227 1228
{
	u64 spte = *sptep;

1229
	if (!is_writable_pte(spte) &&
1230
	      !(pt_protect && spte_can_locklessly_be_made_writable(spte)))
1231 1232 1233 1234
		return false;

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

1235 1236
	if (pt_protect)
		spte &= ~SPTE_MMU_WRITEABLE;
1237
	spte = spte & ~PT_WRITABLE_MASK;
1238

1239
	return mmu_spte_update(sptep, spte);
1240 1241
}

1242 1243
static bool __rmap_write_protect(struct kvm *kvm,
				 struct kvm_rmap_head *rmap_head,
1244
				 bool pt_protect)
1245
{
1246 1247
	u64 *sptep;
	struct rmap_iterator iter;
1248
	bool flush = false;
1249

1250
	for_each_rmap_spte(rmap_head, &iter, sptep)
1251
		flush |= spte_write_protect(sptep, pt_protect);
1252

1253
	return flush;
1254 1255
}

1256
static bool spte_clear_dirty(u64 *sptep)
1257 1258 1259 1260 1261 1262 1263 1264 1265 1266
{
	u64 spte = *sptep;

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

	spte &= ~shadow_dirty_mask;

	return mmu_spte_update(sptep, spte);
}

1267
static bool __rmap_clear_dirty(struct kvm *kvm, struct kvm_rmap_head *rmap_head)
1268 1269 1270 1271 1272
{
	u64 *sptep;
	struct rmap_iterator iter;
	bool flush = false;

1273
	for_each_rmap_spte(rmap_head, &iter, sptep)
1274
		flush |= spte_clear_dirty(sptep);
1275 1276 1277 1278

	return flush;
}

1279
static bool spte_set_dirty(u64 *sptep)
1280 1281 1282 1283 1284 1285 1286 1287 1288 1289
{
	u64 spte = *sptep;

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

	spte |= shadow_dirty_mask;

	return mmu_spte_update(sptep, spte);
}

1290
static bool __rmap_set_dirty(struct kvm *kvm, struct kvm_rmap_head *rmap_head)
1291 1292 1293 1294 1295
{
	u64 *sptep;
	struct rmap_iterator iter;
	bool flush = false;

1296
	for_each_rmap_spte(rmap_head, &iter, sptep)
1297
		flush |= spte_set_dirty(sptep);
1298 1299 1300 1301

	return flush;
}

1302
/**
1303
 * kvm_mmu_write_protect_pt_masked - write protect selected PT level pages
1304 1305 1306 1307 1308 1309 1310 1311
 * @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.
 */
1312
static void kvm_mmu_write_protect_pt_masked(struct kvm *kvm,
1313 1314
				     struct kvm_memory_slot *slot,
				     gfn_t gfn_offset, unsigned long mask)
1315
{
1316
	struct kvm_rmap_head *rmap_head;
1317

1318
	while (mask) {
1319 1320 1321
		rmap_head = __gfn_to_rmap(slot->base_gfn + gfn_offset + __ffs(mask),
					  PT_PAGE_TABLE_LEVEL, slot);
		__rmap_write_protect(kvm, rmap_head, false);
M
Marcelo Tosatti 已提交
1322

1323 1324 1325
		/* clear the first set bit */
		mask &= mask - 1;
	}
1326 1327
}

1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340
/**
 * 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)
{
1341
	struct kvm_rmap_head *rmap_head;
1342 1343

	while (mask) {
1344 1345 1346
		rmap_head = __gfn_to_rmap(slot->base_gfn + gfn_offset + __ffs(mask),
					  PT_PAGE_TABLE_LEVEL, slot);
		__rmap_clear_dirty(kvm, rmap_head);
1347 1348 1349 1350 1351 1352 1353

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

1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367
/**
 * 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)
{
1368 1369 1370 1371 1372
	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);
1373 1374
}

1375 1376
bool kvm_mmu_slot_gfn_write_protect(struct kvm *kvm,
				    struct kvm_memory_slot *slot, u64 gfn)
1377
{
1378
	struct kvm_rmap_head *rmap_head;
1379
	int i;
1380
	bool write_protected = false;
1381

1382
	for (i = PT_PAGE_TABLE_LEVEL; i <= PT_MAX_HUGEPAGE_LEVEL; ++i) {
1383
		rmap_head = __gfn_to_rmap(gfn, i, slot);
1384
		write_protected |= __rmap_write_protect(kvm, rmap_head, true);
1385 1386 1387
	}

	return write_protected;
1388 1389
}

1390 1391 1392 1393 1394 1395 1396 1397
static bool rmap_write_protect(struct kvm_vcpu *vcpu, u64 gfn)
{
	struct kvm_memory_slot *slot;

	slot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);
	return kvm_mmu_slot_gfn_write_protect(vcpu->kvm, slot, gfn);
}

1398
static bool kvm_zap_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head)
1399
{
1400 1401
	u64 *sptep;
	struct rmap_iterator iter;
1402
	bool flush = false;
1403

1404
	while ((sptep = rmap_get_first(rmap_head, &iter))) {
1405
		rmap_printk("%s: spte %p %llx.\n", __func__, sptep, *sptep);
1406 1407

		drop_spte(kvm, sptep);
1408
		flush = true;
1409
	}
1410

1411 1412 1413
	return flush;
}

1414
static int kvm_unmap_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
1415 1416 1417
			   struct kvm_memory_slot *slot, gfn_t gfn, int level,
			   unsigned long data)
{
1418
	return kvm_zap_rmapp(kvm, rmap_head);
1419 1420
}

1421
static int kvm_set_pte_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
1422 1423
			     struct kvm_memory_slot *slot, gfn_t gfn, int level,
			     unsigned long data)
1424
{
1425 1426
	u64 *sptep;
	struct rmap_iterator iter;
1427
	int need_flush = 0;
1428
	u64 new_spte;
1429
	pte_t *ptep = (pte_t *)data;
D
Dan Williams 已提交
1430
	kvm_pfn_t new_pfn;
1431 1432 1433

	WARN_ON(pte_huge(*ptep));
	new_pfn = pte_pfn(*ptep);
1434

1435
restart:
1436
	for_each_rmap_spte(rmap_head, &iter, sptep) {
1437 1438
		rmap_printk("kvm_set_pte_rmapp: spte %p %llx gfn %llx (%d)\n",
			     sptep, *sptep, gfn, level);
1439

1440
		need_flush = 1;
1441

1442
		if (pte_write(*ptep)) {
1443
			drop_spte(kvm, sptep);
1444
			goto restart;
1445
		} else {
1446
			new_spte = *sptep & ~PT64_BASE_ADDR_MASK;
1447 1448 1449 1450
			new_spte |= (u64)new_pfn << PAGE_SHIFT;

			new_spte &= ~PT_WRITABLE_MASK;
			new_spte &= ~SPTE_HOST_WRITEABLE;
1451
			new_spte &= ~shadow_accessed_mask;
1452 1453 1454

			mmu_spte_clear_track_bits(sptep);
			mmu_spte_set(sptep, new_spte);
1455 1456
		}
	}
1457

1458 1459 1460 1461 1462 1463
	if (need_flush)
		kvm_flush_remote_tlbs(kvm);

	return 0;
}

1464 1465 1466 1467 1468 1469 1470 1471 1472 1473
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;
1474
	struct kvm_rmap_head *rmap;
1475 1476 1477
	int level;

	/* private field. */
1478
	struct kvm_rmap_head *end_rmap;
1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531
};

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_))

1532 1533 1534 1535 1536
static int kvm_handle_hva_range(struct kvm *kvm,
				unsigned long start,
				unsigned long end,
				unsigned long data,
				int (*handler)(struct kvm *kvm,
1537
					       struct kvm_rmap_head *rmap_head,
1538
					       struct kvm_memory_slot *slot,
1539 1540
					       gfn_t gfn,
					       int level,
1541
					       unsigned long data))
1542
{
1543
	struct kvm_memslots *slots;
1544
	struct kvm_memory_slot *memslot;
1545 1546
	struct slot_rmap_walk_iterator iterator;
	int ret = 0;
1547
	int i;
1548

1549 1550 1551 1552 1553
	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;
1554

1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573
			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);
		}
1574 1575
	}

1576
	return ret;
1577 1578
}

1579 1580
static int kvm_handle_hva(struct kvm *kvm, unsigned long hva,
			  unsigned long data,
1581 1582
			  int (*handler)(struct kvm *kvm,
					 struct kvm_rmap_head *rmap_head,
1583
					 struct kvm_memory_slot *slot,
1584
					 gfn_t gfn, int level,
1585 1586 1587
					 unsigned long data))
{
	return kvm_handle_hva_range(kvm, hva, hva + 1, data, handler);
1588 1589 1590 1591
}

int kvm_unmap_hva(struct kvm *kvm, unsigned long hva)
{
1592 1593 1594
	return kvm_handle_hva(kvm, hva, 0, kvm_unmap_rmapp);
}

1595 1596 1597 1598 1599
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);
}

1600 1601
void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte)
{
F
Frederik Deweerdt 已提交
1602
	kvm_handle_hva(kvm, hva, (unsigned long)&pte, kvm_set_pte_rmapp);
1603 1604
}

1605
static int kvm_age_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
1606 1607
			 struct kvm_memory_slot *slot, gfn_t gfn, int level,
			 unsigned long data)
1608
{
1609
	u64 *sptep;
1610
	struct rmap_iterator uninitialized_var(iter);
1611 1612
	int young = 0;

A
Andres Lagar-Cavilla 已提交
1613
	BUG_ON(!shadow_accessed_mask);
1614

1615
	for_each_rmap_spte(rmap_head, &iter, sptep) {
1616
		if (*sptep & shadow_accessed_mask) {
1617
			young = 1;
1618 1619
			clear_bit((ffs(shadow_accessed_mask) - 1),
				 (unsigned long *)sptep);
1620
		}
1621
	}
1622

1623
	trace_kvm_age_page(gfn, level, slot, young);
1624 1625 1626
	return young;
}

1627
static int kvm_test_age_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
1628 1629
			      struct kvm_memory_slot *slot, gfn_t gfn,
			      int level, unsigned long data)
A
Andrea Arcangeli 已提交
1630
{
1631 1632
	u64 *sptep;
	struct rmap_iterator iter;
A
Andrea Arcangeli 已提交
1633 1634 1635 1636 1637 1638 1639 1640 1641

	/*
	 * 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;

1642 1643 1644
	for_each_rmap_spte(rmap_head, &iter, sptep)
		if (is_accessed_spte(*sptep))
			return 1;
A
Andrea Arcangeli 已提交
1645
out:
1646
	return 0;
A
Andrea Arcangeli 已提交
1647 1648
}

1649 1650
#define RMAP_RECYCLE_THRESHOLD 1000

1651
static void rmap_recycle(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn)
1652
{
1653
	struct kvm_rmap_head *rmap_head;
1654 1655 1656
	struct kvm_mmu_page *sp;

	sp = page_header(__pa(spte));
1657

1658
	rmap_head = gfn_to_rmap(vcpu->kvm, gfn, sp);
1659

1660
	kvm_unmap_rmapp(vcpu->kvm, rmap_head, NULL, gfn, sp->role.level, 0);
1661 1662 1663
	kvm_flush_remote_tlbs(vcpu->kvm);
}

A
Andres Lagar-Cavilla 已提交
1664
int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
1665
{
A
Andres Lagar-Cavilla 已提交
1666 1667 1668 1669 1670 1671 1672 1673
	/*
	 * 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.
	 */
1674
	if (!shadow_accessed_mask)
A
Andres Lagar-Cavilla 已提交
1675 1676 1677 1678
		return kvm_handle_hva_range(kvm, start, end, 0,
					    kvm_unmap_rmapp);

	return kvm_handle_hva_range(kvm, start, end, 0, kvm_age_rmapp);
1679 1680
}

A
Andrea Arcangeli 已提交
1681 1682 1683 1684 1685
int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
{
	return kvm_handle_hva(kvm, hva, 0, kvm_test_age_rmapp);
}

1686
#ifdef MMU_DEBUG
1687
static int is_empty_shadow_page(u64 *spt)
A
Avi Kivity 已提交
1688
{
1689 1690 1691
	u64 *pos;
	u64 *end;

1692
	for (pos = spt, end = pos + PAGE_SIZE / sizeof(u64); pos != end; pos++)
1693
		if (is_shadow_present_pte(*pos)) {
1694
			printk(KERN_ERR "%s: %p %llx\n", __func__,
1695
			       pos, *pos);
A
Avi Kivity 已提交
1696
			return 0;
1697
		}
A
Avi Kivity 已提交
1698 1699
	return 1;
}
1700
#endif
A
Avi Kivity 已提交
1701

1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713
/*
 * 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);
}

1714
static void kvm_mmu_free_page(struct kvm_mmu_page *sp)
1715
{
1716
	MMU_WARN_ON(!is_empty_shadow_page(sp->spt));
1717
	hlist_del(&sp->hash_link);
1718 1719
	list_del(&sp->link);
	free_page((unsigned long)sp->spt);
1720 1721
	if (!sp->role.direct)
		free_page((unsigned long)sp->gfns);
1722
	kmem_cache_free(mmu_page_header_cache, sp);
1723 1724
}

1725 1726
static unsigned kvm_page_table_hashfn(gfn_t gfn)
{
1727
	return hash_64(gfn, KVM_MMU_HASH_SHIFT);
1728 1729
}

1730
static void mmu_page_add_parent_pte(struct kvm_vcpu *vcpu,
1731
				    struct kvm_mmu_page *sp, u64 *parent_pte)
1732 1733 1734 1735
{
	if (!parent_pte)
		return;

1736
	pte_list_add(vcpu, parent_pte, &sp->parent_ptes);
1737 1738
}

1739
static void mmu_page_remove_parent_pte(struct kvm_mmu_page *sp,
1740 1741
				       u64 *parent_pte)
{
1742
	pte_list_remove(parent_pte, &sp->parent_ptes);
1743 1744
}

1745 1746 1747 1748
static void drop_parent_pte(struct kvm_mmu_page *sp,
			    u64 *parent_pte)
{
	mmu_page_remove_parent_pte(sp, parent_pte);
1749
	mmu_spte_clear_no_track(parent_pte);
1750 1751
}

1752
static struct kvm_mmu_page *kvm_mmu_alloc_page(struct kvm_vcpu *vcpu, int direct)
M
Marcelo Tosatti 已提交
1753
{
1754
	struct kvm_mmu_page *sp;
1755

1756 1757
	sp = mmu_memory_cache_alloc(&vcpu->arch.mmu_page_header_cache);
	sp->spt = mmu_memory_cache_alloc(&vcpu->arch.mmu_page_cache);
1758
	if (!direct)
1759
		sp->gfns = mmu_memory_cache_alloc(&vcpu->arch.mmu_page_cache);
1760
	set_page_private(virt_to_page(sp->spt), (unsigned long)sp);
1761 1762 1763 1764 1765 1766

	/*
	 * 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().
	 */
1767 1768 1769
	list_add(&sp->link, &vcpu->kvm->arch.active_mmu_pages);
	kvm_mod_used_mmu_pages(vcpu->kvm, +1);
	return sp;
M
Marcelo Tosatti 已提交
1770 1771
}

1772
static void mark_unsync(u64 *spte);
1773
static void kvm_mmu_mark_parents_unsync(struct kvm_mmu_page *sp)
1774
{
1775 1776 1777 1778 1779 1780
	u64 *sptep;
	struct rmap_iterator iter;

	for_each_rmap_spte(&sp->parent_ptes, &iter, sptep) {
		mark_unsync(sptep);
	}
1781 1782
}

1783
static void mark_unsync(u64 *spte)
1784
{
1785
	struct kvm_mmu_page *sp;
1786
	unsigned int index;
1787

1788
	sp = page_header(__pa(spte));
1789 1790
	index = spte - sp->spt;
	if (__test_and_set_bit(index, sp->unsync_child_bitmap))
1791
		return;
1792
	if (sp->unsync_children++)
1793
		return;
1794
	kvm_mmu_mark_parents_unsync(sp);
1795 1796
}

1797
static int nonpaging_sync_page(struct kvm_vcpu *vcpu,
1798
			       struct kvm_mmu_page *sp)
1799
{
1800
	return 0;
1801 1802
}

M
Marcelo Tosatti 已提交
1803 1804 1805 1806
static void nonpaging_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
{
}

1807 1808
static void nonpaging_update_pte(struct kvm_vcpu *vcpu,
				 struct kvm_mmu_page *sp, u64 *spte,
1809
				 const void *pte)
1810 1811 1812 1813
{
	WARN_ON(1);
}

1814 1815 1816 1817 1818 1819 1820 1821 1822 1823
#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;
};

1824 1825
static int mmu_pages_add(struct kvm_mmu_pages *pvec, struct kvm_mmu_page *sp,
			 int idx)
1826
{
1827
	int i;
1828

1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839
	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);
}

1840 1841 1842 1843 1844 1845 1846
static inline void clear_unsync_child_bit(struct kvm_mmu_page *sp, int idx)
{
	--sp->unsync_children;
	WARN_ON((int)sp->unsync_children < 0);
	__clear_bit(idx, sp->unsync_child_bitmap);
}

1847 1848 1849 1850
static int __mmu_unsync_walk(struct kvm_mmu_page *sp,
			   struct kvm_mmu_pages *pvec)
{
	int i, ret, nr_unsync_leaf = 0;
1851

1852
	for_each_set_bit(i, sp->unsync_child_bitmap, 512) {
1853
		struct kvm_mmu_page *child;
1854 1855
		u64 ent = sp->spt[i];

1856 1857 1858 1859
		if (!is_shadow_present_pte(ent) || is_large_pte(ent)) {
			clear_unsync_child_bit(sp, i);
			continue;
		}
1860 1861 1862 1863 1864 1865 1866 1867

		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);
1868 1869 1870 1871
			if (!ret) {
				clear_unsync_child_bit(sp, i);
				continue;
			} else if (ret > 0) {
1872
				nr_unsync_leaf += ret;
1873
			} else
1874 1875 1876 1877 1878 1879
				return ret;
		} else if (child->unsync) {
			nr_unsync_leaf++;
			if (mmu_pages_add(pvec, child, i))
				return -ENOSPC;
		} else
1880
			clear_unsync_child_bit(sp, i);
1881 1882
	}

1883 1884 1885
	return nr_unsync_leaf;
}

1886 1887
#define INVALID_INDEX (-1)

1888 1889 1890
static int mmu_unsync_walk(struct kvm_mmu_page *sp,
			   struct kvm_mmu_pages *pvec)
{
P
Paolo Bonzini 已提交
1891
	pvec->nr = 0;
1892 1893 1894
	if (!sp->unsync_children)
		return 0;

1895
	mmu_pages_add(pvec, sp, INVALID_INDEX);
1896
	return __mmu_unsync_walk(sp, pvec);
1897 1898 1899 1900 1901
}

static void kvm_unlink_unsync_page(struct kvm *kvm, struct kvm_mmu_page *sp)
{
	WARN_ON(!sp->unsync);
1902
	trace_kvm_mmu_sync_page(sp);
1903 1904 1905 1906
	sp->unsync = 0;
	--kvm->stat.mmu_unsync;
}

1907 1908 1909 1910
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);
1911

1912 1913 1914 1915 1916 1917
/*
 * 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.
 *
1918
 * for_each_valid_sp() has skipped that kind of pages.
1919
 */
1920
#define for_each_valid_sp(_kvm, _sp, _gfn)				\
1921 1922
	hlist_for_each_entry(_sp,					\
	  &(_kvm)->arch.mmu_page_hash[kvm_page_table_hashfn(_gfn)], hash_link) \
1923 1924
		if (is_obsolete_sp((_kvm), (_sp)) || (_sp)->role.invalid) {    \
		} else
1925 1926

#define for_each_gfn_indirect_valid_sp(_kvm, _sp, _gfn)			\
1927 1928
	for_each_valid_sp(_kvm, _sp, _gfn)				\
		if ((_sp)->gfn != (_gfn) || (_sp)->role.direct) {} else
1929

1930
/* @sp->gfn should be write-protected at the call site */
1931 1932
static bool __kvm_sync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
			    struct list_head *invalid_list)
1933
{
1934
	if (sp->role.cr4_pae != !!is_pae(vcpu)) {
1935
		kvm_mmu_prepare_zap_page(vcpu->kvm, sp, invalid_list);
1936
		return false;
1937 1938
	}

1939
	if (vcpu->arch.mmu.sync_page(vcpu, sp) == 0) {
1940
		kvm_mmu_prepare_zap_page(vcpu->kvm, sp, invalid_list);
1941
		return false;
1942 1943
	}

1944
	return true;
1945 1946
}

1947 1948 1949
static void kvm_mmu_flush_or_zap(struct kvm_vcpu *vcpu,
				 struct list_head *invalid_list,
				 bool remote_flush, bool local_flush)
1950
{
1951 1952 1953 1954
	if (!list_empty(invalid_list)) {
		kvm_mmu_commit_zap_page(vcpu->kvm, invalid_list);
		return;
	}
1955

1956 1957 1958 1959
	if (remote_flush)
		kvm_flush_remote_tlbs(vcpu->kvm);
	else if (local_flush)
		kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1960 1961
}

1962 1963 1964 1965 1966 1967 1968
#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

1969 1970 1971 1972 1973
static bool is_obsolete_sp(struct kvm *kvm, struct kvm_mmu_page *sp)
{
	return unlikely(sp->mmu_valid_gen != kvm->arch.mmu_valid_gen);
}

1974
static bool kvm_sync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
1975
			 struct list_head *invalid_list)
1976
{
1977 1978
	kvm_unlink_unsync_page(vcpu->kvm, sp);
	return __kvm_sync_page(vcpu, sp, invalid_list);
1979 1980
}

1981
/* @gfn should be write-protected at the call site */
1982 1983
static bool kvm_sync_pages(struct kvm_vcpu *vcpu, gfn_t gfn,
			   struct list_head *invalid_list)
1984 1985
{
	struct kvm_mmu_page *s;
1986
	bool ret = false;
1987

1988
	for_each_gfn_indirect_valid_sp(vcpu->kvm, s, gfn) {
1989
		if (!s->unsync)
1990 1991 1992
			continue;

		WARN_ON(s->role.level != PT_PAGE_TABLE_LEVEL);
1993
		ret |= kvm_sync_page(vcpu, s, invalid_list);
1994 1995
	}

1996
	return ret;
1997 1998
}

1999
struct mmu_page_path {
P
Paolo Bonzini 已提交
2000 2001
	struct kvm_mmu_page *parent[PT64_ROOT_LEVEL];
	unsigned int idx[PT64_ROOT_LEVEL];
2002 2003
};

2004
#define for_each_sp(pvec, sp, parents, i)			\
P
Paolo Bonzini 已提交
2005
		for (i = mmu_pages_first(&pvec, &parents);	\
2006 2007 2008
			i < pvec.nr && ({ sp = pvec.page[i].sp; 1;});	\
			i = mmu_pages_next(&pvec, &parents, i))

2009 2010 2011
static int mmu_pages_next(struct kvm_mmu_pages *pvec,
			  struct mmu_page_path *parents,
			  int i)
2012 2013 2014 2015 2016
{
	int n;

	for (n = i+1; n < pvec->nr; n++) {
		struct kvm_mmu_page *sp = pvec->page[n].sp;
P
Paolo Bonzini 已提交
2017 2018
		unsigned idx = pvec->page[n].idx;
		int level = sp->role.level;
2019

P
Paolo Bonzini 已提交
2020 2021 2022
		parents->idx[level-1] = idx;
		if (level == PT_PAGE_TABLE_LEVEL)
			break;
2023

P
Paolo Bonzini 已提交
2024
		parents->parent[level-2] = sp;
2025 2026 2027 2028 2029
	}

	return n;
}

P
Paolo Bonzini 已提交
2030 2031 2032 2033 2034 2035 2036 2037 2038
static int mmu_pages_first(struct kvm_mmu_pages *pvec,
			   struct mmu_page_path *parents)
{
	struct kvm_mmu_page *sp;
	int level;

	if (pvec->nr == 0)
		return 0;

2039 2040
	WARN_ON(pvec->page[0].idx != INVALID_INDEX);

P
Paolo Bonzini 已提交
2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053
	sp = pvec->page[0].sp;
	level = sp->role.level;
	WARN_ON(level == PT_PAGE_TABLE_LEVEL);

	parents->parent[level-2] = sp;

	/* Also set up a sentinel.  Further entries in pvec are all
	 * children of sp, so this element is never overwritten.
	 */
	parents->parent[level-1] = NULL;
	return mmu_pages_next(pvec, parents, 0);
}

2054
static void mmu_pages_clear_parents(struct mmu_page_path *parents)
2055
{
2056 2057 2058 2059 2060 2061 2062 2063 2064
	struct kvm_mmu_page *sp;
	unsigned int level = 0;

	do {
		unsigned int idx = parents->idx[level];
		sp = parents->parent[level];
		if (!sp)
			return;

2065
		WARN_ON(idx == INVALID_INDEX);
2066
		clear_unsync_child_bit(sp, idx);
2067
		level++;
P
Paolo Bonzini 已提交
2068
	} while (!sp->unsync_children);
2069
}
2070

2071 2072 2073 2074 2075 2076 2077
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;
2078
	LIST_HEAD(invalid_list);
2079
	bool flush = false;
2080 2081

	while (mmu_unsync_walk(parent, &pages)) {
2082
		bool protected = false;
2083 2084

		for_each_sp(pages, sp, parents, i)
2085
			protected |= rmap_write_protect(vcpu, sp->gfn);
2086

2087
		if (protected) {
2088
			kvm_flush_remote_tlbs(vcpu->kvm);
2089 2090
			flush = false;
		}
2091

2092
		for_each_sp(pages, sp, parents, i) {
2093
			flush |= kvm_sync_page(vcpu, sp, &invalid_list);
2094 2095
			mmu_pages_clear_parents(&parents);
		}
2096 2097 2098 2099 2100
		if (need_resched() || spin_needbreak(&vcpu->kvm->mmu_lock)) {
			kvm_mmu_flush_or_zap(vcpu, &invalid_list, false, flush);
			cond_resched_lock(&vcpu->kvm->mmu_lock);
			flush = false;
		}
2101
	}
2102 2103

	kvm_mmu_flush_or_zap(vcpu, &invalid_list, false, flush);
2104 2105
}

2106 2107
static void __clear_sp_write_flooding_count(struct kvm_mmu_page *sp)
{
2108
	atomic_set(&sp->write_flooding_count,  0);
2109 2110 2111 2112 2113 2114 2115 2116 2117
}

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

	__clear_sp_write_flooding_count(sp);
}

2118 2119 2120 2121
static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
					     gfn_t gfn,
					     gva_t gaddr,
					     unsigned level,
2122
					     int direct,
2123
					     unsigned access)
2124 2125 2126
{
	union kvm_mmu_page_role role;
	unsigned quadrant;
2127 2128
	struct kvm_mmu_page *sp;
	bool need_sync = false;
2129
	bool flush = false;
2130
	int collisions = 0;
2131
	LIST_HEAD(invalid_list);
2132

2133
	role = vcpu->arch.mmu.base_role;
2134
	role.level = level;
2135
	role.direct = direct;
2136
	if (role.direct)
2137
		role.cr4_pae = 0;
2138
	role.access = access;
2139 2140
	if (!vcpu->arch.mmu.direct_map
	    && vcpu->arch.mmu.root_level <= PT32_ROOT_LEVEL) {
2141 2142 2143 2144
		quadrant = gaddr >> (PAGE_SHIFT + (PT64_PT_BITS * level));
		quadrant &= (1 << ((PT32_PT_BITS - PT64_PT_BITS) * level)) - 1;
		role.quadrant = quadrant;
	}
2145 2146 2147 2148 2149 2150
	for_each_valid_sp(vcpu->kvm, sp, gfn) {
		if (sp->gfn != gfn) {
			collisions++;
			continue;
		}

2151 2152
		if (!need_sync && sp->unsync)
			need_sync = true;
2153

2154 2155
		if (sp->role.word != role.word)
			continue;
2156

2157 2158 2159 2160 2161 2162 2163 2164 2165 2166
		if (sp->unsync) {
			/* The page is good, but __kvm_sync_page might still end
			 * up zapping it.  If so, break in order to rebuild it.
			 */
			if (!__kvm_sync_page(vcpu, sp, &invalid_list))
				break;

			WARN_ON(!list_empty(&invalid_list));
			kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
		}
2167

2168
		if (sp->unsync_children)
2169
			kvm_make_request(KVM_REQ_MMU_SYNC, vcpu);
2170

2171
		__clear_sp_write_flooding_count(sp);
2172
		trace_kvm_mmu_get_page(sp, false);
2173
		goto out;
2174
	}
2175

A
Avi Kivity 已提交
2176
	++vcpu->kvm->stat.mmu_cache_miss;
2177 2178 2179

	sp = kvm_mmu_alloc_page(vcpu, direct);

2180 2181
	sp->gfn = gfn;
	sp->role = role;
2182 2183
	hlist_add_head(&sp->hash_link,
		&vcpu->kvm->arch.mmu_page_hash[kvm_page_table_hashfn(gfn)]);
2184
	if (!direct) {
2185 2186 2187 2188 2189 2190 2191 2192
		/*
		 * we should do write protection before syncing pages
		 * otherwise the content of the synced shadow page may
		 * be inconsistent with guest page table.
		 */
		account_shadowed(vcpu->kvm, sp);
		if (level == PT_PAGE_TABLE_LEVEL &&
		      rmap_write_protect(vcpu, gfn))
2193
			kvm_flush_remote_tlbs(vcpu->kvm);
2194 2195

		if (level > PT_PAGE_TABLE_LEVEL && need_sync)
2196
			flush |= kvm_sync_pages(vcpu, gfn, &invalid_list);
2197
	}
2198
	sp->mmu_valid_gen = vcpu->kvm->arch.mmu_valid_gen;
2199
	clear_page(sp->spt);
A
Avi Kivity 已提交
2200
	trace_kvm_mmu_get_page(sp, true);
2201 2202

	kvm_mmu_flush_or_zap(vcpu, &invalid_list, false, flush);
2203 2204 2205
out:
	if (collisions > vcpu->kvm->stat.max_mmu_page_hash_collisions)
		vcpu->kvm->stat.max_mmu_page_hash_collisions = collisions;
2206
	return sp;
2207 2208
}

2209 2210 2211 2212 2213 2214
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;
2215 2216 2217 2218 2219 2220

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

2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234
	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;
2235

2236 2237 2238 2239 2240
	iterator->index = SHADOW_PT_INDEX(iterator->addr, iterator->level);
	iterator->sptep	= ((u64 *)__va(iterator->shadow_addr)) + iterator->index;
	return true;
}

2241 2242
static void __shadow_walk_next(struct kvm_shadow_walk_iterator *iterator,
			       u64 spte)
2243
{
2244
	if (is_last_spte(spte, iterator->level)) {
2245 2246 2247 2248
		iterator->level = 0;
		return;
	}

2249
	iterator->shadow_addr = spte & PT64_BASE_ADDR_MASK;
2250 2251 2252
	--iterator->level;
}

2253 2254 2255 2256 2257
static void shadow_walk_next(struct kvm_shadow_walk_iterator *iterator)
{
	return __shadow_walk_next(iterator, *iterator->sptep);
}

2258 2259
static void link_shadow_page(struct kvm_vcpu *vcpu, u64 *sptep,
			     struct kvm_mmu_page *sp)
2260 2261 2262
{
	u64 spte;

2263
	BUILD_BUG_ON(VMX_EPT_WRITABLE_MASK != PT_WRITABLE_MASK);
2264

2265
	spte = __pa(sp->spt) | shadow_present_mask | PT_WRITABLE_MASK |
2266
	       shadow_user_mask | shadow_x_mask | shadow_accessed_mask;
X
Xiao Guangrong 已提交
2267

2268
	mmu_spte_set(sptep, spte);
2269 2270 2271 2272 2273

	mmu_page_add_parent_pte(vcpu, sp, sptep);

	if (sp->unsync_children || sp->unsync)
		mark_unsync(sptep);
2274 2275
}

2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292
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;

2293
		drop_parent_pte(child, sptep);
2294 2295 2296 2297
		kvm_flush_remote_tlbs(vcpu->kvm);
	}
}

X
Xiao Guangrong 已提交
2298
static bool mmu_page_zap_pte(struct kvm *kvm, struct kvm_mmu_page *sp,
2299 2300 2301 2302 2303 2304 2305
			     u64 *spte)
{
	u64 pte;
	struct kvm_mmu_page *child;

	pte = *spte;
	if (is_shadow_present_pte(pte)) {
X
Xiao Guangrong 已提交
2306
		if (is_last_spte(pte, sp->role.level)) {
2307
			drop_spte(kvm, spte);
X
Xiao Guangrong 已提交
2308 2309 2310
			if (is_large_pte(pte))
				--kvm->stat.lpages;
		} else {
2311
			child = page_header(pte & PT64_BASE_ADDR_MASK);
2312
			drop_parent_pte(child, spte);
2313
		}
X
Xiao Guangrong 已提交
2314 2315 2316 2317
		return true;
	}

	if (is_mmio_spte(pte))
2318
		mmu_spte_clear_no_track(spte);
2319

X
Xiao Guangrong 已提交
2320
	return false;
2321 2322
}

2323
static void kvm_mmu_page_unlink_children(struct kvm *kvm,
2324
					 struct kvm_mmu_page *sp)
2325
{
2326 2327
	unsigned i;

2328 2329
	for (i = 0; i < PT64_ENT_PER_PAGE; ++i)
		mmu_page_zap_pte(kvm, sp, sp->spt + i);
2330 2331
}

2332
static void kvm_mmu_unlink_parents(struct kvm *kvm, struct kvm_mmu_page *sp)
2333
{
2334 2335
	u64 *sptep;
	struct rmap_iterator iter;
2336

2337
	while ((sptep = rmap_get_first(&sp->parent_ptes, &iter)))
2338
		drop_parent_pte(sp, sptep);
2339 2340
}

2341
static int mmu_zap_unsync_children(struct kvm *kvm,
2342 2343
				   struct kvm_mmu_page *parent,
				   struct list_head *invalid_list)
2344
{
2345 2346 2347
	int i, zapped = 0;
	struct mmu_page_path parents;
	struct kvm_mmu_pages pages;
2348

2349
	if (parent->role.level == PT_PAGE_TABLE_LEVEL)
2350
		return 0;
2351 2352 2353 2354 2355

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

		for_each_sp(pages, sp, parents, i) {
2356
			kvm_mmu_prepare_zap_page(kvm, sp, invalid_list);
2357
			mmu_pages_clear_parents(&parents);
2358
			zapped++;
2359 2360 2361 2362
		}
	}

	return zapped;
2363 2364
}

2365 2366
static int kvm_mmu_prepare_zap_page(struct kvm *kvm, struct kvm_mmu_page *sp,
				    struct list_head *invalid_list)
2367
{
2368
	int ret;
A
Avi Kivity 已提交
2369

2370
	trace_kvm_mmu_prepare_zap_page(sp);
2371
	++kvm->stat.mmu_shadow_zapped;
2372
	ret = mmu_zap_unsync_children(kvm, sp, invalid_list);
2373
	kvm_mmu_page_unlink_children(kvm, sp);
2374
	kvm_mmu_unlink_parents(kvm, sp);
2375

2376
	if (!sp->role.invalid && !sp->role.direct)
2377
		unaccount_shadowed(kvm, sp);
2378

2379 2380
	if (sp->unsync)
		kvm_unlink_unsync_page(kvm, sp);
2381
	if (!sp->root_count) {
2382 2383
		/* Count self */
		ret++;
2384
		list_move(&sp->link, invalid_list);
2385
		kvm_mod_used_mmu_pages(kvm, -1);
2386
	} else {
A
Avi Kivity 已提交
2387
		list_move(&sp->link, &kvm->arch.active_mmu_pages);
2388 2389 2390 2391 2392 2393 2394

		/*
		 * 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);
2395
	}
2396 2397

	sp->role.invalid = 1;
2398
	return ret;
2399 2400
}

2401 2402 2403
static void kvm_mmu_commit_zap_page(struct kvm *kvm,
				    struct list_head *invalid_list)
{
2404
	struct kvm_mmu_page *sp, *nsp;
2405 2406 2407 2408

	if (list_empty(invalid_list))
		return;

2409
	/*
2410 2411 2412 2413 2414 2415 2416
	 * We need to make sure everyone sees our modifications to
	 * the page tables and see changes to vcpu->mode here. The barrier
	 * in the kvm_flush_remote_tlbs() achieves this. This pairs
	 * with vcpu_enter_guest and walk_shadow_page_lockless_begin/end.
	 *
	 * In addition, kvm_flush_remote_tlbs waits for all vcpus to exit
	 * guest mode and/or lockless shadow page table walks.
2417 2418
	 */
	kvm_flush_remote_tlbs(kvm);
2419

2420
	list_for_each_entry_safe(sp, nsp, invalid_list, link) {
2421
		WARN_ON(!sp->role.invalid || sp->root_count);
2422
		kvm_mmu_free_page(sp);
2423
	}
2424 2425
}

2426 2427 2428 2429 2430 2431 2432 2433
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;

G
Geliang Tang 已提交
2434 2435
	sp = list_last_entry(&kvm->arch.active_mmu_pages,
			     struct kvm_mmu_page, link);
2436 2437 2438 2439 2440
	kvm_mmu_prepare_zap_page(kvm, sp, invalid_list);

	return true;
}

2441 2442
/*
 * Changing the number of mmu pages allocated to the vm
2443
 * Note: if goal_nr_mmu_pages is too small, you will get dead lock
2444
 */
2445
void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int goal_nr_mmu_pages)
2446
{
2447
	LIST_HEAD(invalid_list);
2448

2449 2450
	spin_lock(&kvm->mmu_lock);

2451
	if (kvm->arch.n_used_mmu_pages > goal_nr_mmu_pages) {
2452 2453 2454 2455
		/* 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;
2456

2457
		kvm_mmu_commit_zap_page(kvm, &invalid_list);
2458
		goal_nr_mmu_pages = kvm->arch.n_used_mmu_pages;
2459 2460
	}

2461
	kvm->arch.n_max_mmu_pages = goal_nr_mmu_pages;
2462 2463

	spin_unlock(&kvm->mmu_lock);
2464 2465
}

2466
int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn)
2467
{
2468
	struct kvm_mmu_page *sp;
2469
	LIST_HEAD(invalid_list);
2470 2471
	int r;

2472
	pgprintk("%s: looking for gfn %llx\n", __func__, gfn);
2473
	r = 0;
2474
	spin_lock(&kvm->mmu_lock);
2475
	for_each_gfn_indirect_valid_sp(kvm, sp, gfn) {
2476
		pgprintk("%s: gfn %llx role %x\n", __func__, gfn,
2477 2478
			 sp->role.word);
		r = 1;
2479
		kvm_mmu_prepare_zap_page(kvm, sp, &invalid_list);
2480
	}
2481
	kvm_mmu_commit_zap_page(kvm, &invalid_list);
2482 2483
	spin_unlock(&kvm->mmu_lock);

2484
	return r;
2485
}
2486
EXPORT_SYMBOL_GPL(kvm_mmu_unprotect_page);
2487

2488
static void kvm_unsync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp)
2489 2490 2491 2492 2493 2494 2495 2496
{
	trace_kvm_mmu_unsync_page(sp);
	++vcpu->kvm->stat.mmu_unsync;
	sp->unsync = 1;

	kvm_mmu_mark_parents_unsync(sp);
}

2497 2498
static bool mmu_need_write_protect(struct kvm_vcpu *vcpu, gfn_t gfn,
				   bool can_unsync)
2499
{
2500
	struct kvm_mmu_page *sp;
2501

2502 2503
	if (kvm_page_track_is_active(vcpu, gfn, KVM_PAGE_TRACK_WRITE))
		return true;
2504

2505
	for_each_gfn_indirect_valid_sp(vcpu->kvm, sp, gfn) {
2506
		if (!can_unsync)
2507
			return true;
2508

2509 2510
		if (sp->unsync)
			continue;
2511

2512 2513
		WARN_ON(sp->role.level != PT_PAGE_TABLE_LEVEL);
		kvm_unsync_page(vcpu, sp);
2514
	}
2515 2516

	return false;
2517 2518
}

D
Dan Williams 已提交
2519
static bool kvm_is_mmio_pfn(kvm_pfn_t pfn)
2520 2521 2522 2523 2524 2525 2526
{
	if (pfn_valid(pfn))
		return !is_zero_pfn(pfn) && PageReserved(pfn_to_page(pfn));

	return true;
}

A
Avi Kivity 已提交
2527
static int set_spte(struct kvm_vcpu *vcpu, u64 *sptep,
2528
		    unsigned pte_access, int level,
D
Dan Williams 已提交
2529
		    gfn_t gfn, kvm_pfn_t pfn, bool speculative,
2530
		    bool can_unsync, bool host_writable)
2531
{
2532
	u64 spte = 0;
M
Marcelo Tosatti 已提交
2533
	int ret = 0;
S
Sheng Yang 已提交
2534

2535
	if (set_mmio_spte(vcpu, sptep, gfn, pfn, pte_access))
2536 2537
		return 0;

2538 2539 2540 2541 2542 2543
	/*
	 * For the EPT case, shadow_present_mask is 0 if hardware
	 * supports exec-only page table entries.  In that case,
	 * ACC_USER_MASK and shadow_user_mask are used to represent
	 * read access.  See FNAME(gpte_access) in paging_tmpl.h.
	 */
2544
	spte |= shadow_present_mask;
2545
	if (!speculative)
2546
		spte |= shadow_accessed_mask;
2547

S
Sheng Yang 已提交
2548 2549 2550 2551
	if (pte_access & ACC_EXEC_MASK)
		spte |= shadow_x_mask;
	else
		spte |= shadow_nx_mask;
2552

2553
	if (pte_access & ACC_USER_MASK)
S
Sheng Yang 已提交
2554
		spte |= shadow_user_mask;
2555

2556
	if (level > PT_PAGE_TABLE_LEVEL)
M
Marcelo Tosatti 已提交
2557
		spte |= PT_PAGE_SIZE_MASK;
2558
	if (tdp_enabled)
2559
		spte |= kvm_x86_ops->get_mt_mask(vcpu, gfn,
2560
			kvm_is_mmio_pfn(pfn));
2561

2562
	if (host_writable)
2563
		spte |= SPTE_HOST_WRITEABLE;
2564 2565
	else
		pte_access &= ~ACC_WRITE_MASK;
2566

2567
	spte |= (u64)pfn << PAGE_SHIFT;
2568

2569
	if (pte_access & ACC_WRITE_MASK) {
2570

X
Xiao Guangrong 已提交
2571
		/*
2572 2573 2574 2575
		 * 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
2576
		 */
2577
		if (level > PT_PAGE_TABLE_LEVEL &&
2578
		    mmu_gfn_lpage_is_disallowed(vcpu, gfn, level))
A
Avi Kivity 已提交
2579
			goto done;
2580

2581
		spte |= PT_WRITABLE_MASK | SPTE_MMU_WRITEABLE;
2582

2583 2584 2585 2586 2587 2588
		/*
		 * 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.
		 */
2589
		if (!can_unsync && is_writable_pte(*sptep))
2590 2591
			goto set_pte;

2592
		if (mmu_need_write_protect(vcpu, gfn, can_unsync)) {
2593
			pgprintk("%s: found shadow page for %llx, marking ro\n",
2594
				 __func__, gfn);
M
Marcelo Tosatti 已提交
2595
			ret = 1;
2596
			pte_access &= ~ACC_WRITE_MASK;
2597
			spte &= ~(PT_WRITABLE_MASK | SPTE_MMU_WRITEABLE);
2598 2599 2600
		}
	}

2601
	if (pte_access & ACC_WRITE_MASK) {
2602
		kvm_vcpu_mark_page_dirty(vcpu, gfn);
2603 2604
		spte |= shadow_dirty_mask;
	}
2605

2606
set_pte:
2607
	if (mmu_spte_update(sptep, spte))
2608
		kvm_flush_remote_tlbs(vcpu->kvm);
A
Avi Kivity 已提交
2609
done:
M
Marcelo Tosatti 已提交
2610 2611 2612
	return ret;
}

2613
static bool mmu_set_spte(struct kvm_vcpu *vcpu, u64 *sptep, unsigned pte_access,
D
Dan Williams 已提交
2614
			 int write_fault, int level, gfn_t gfn, kvm_pfn_t pfn,
2615
			 bool speculative, bool host_writable)
M
Marcelo Tosatti 已提交
2616 2617
{
	int was_rmapped = 0;
2618
	int rmap_count;
2619
	bool emulate = false;
M
Marcelo Tosatti 已提交
2620

2621 2622
	pgprintk("%s: spte %llx write_fault %d gfn %llx\n", __func__,
		 *sptep, write_fault, gfn);
M
Marcelo Tosatti 已提交
2623

2624
	if (is_shadow_present_pte(*sptep)) {
M
Marcelo Tosatti 已提交
2625 2626 2627 2628
		/*
		 * If we overwrite a PTE page pointer with a 2MB PMD, unlink
		 * the parent of the now unreachable PTE.
		 */
2629 2630
		if (level > PT_PAGE_TABLE_LEVEL &&
		    !is_large_pte(*sptep)) {
M
Marcelo Tosatti 已提交
2631
			struct kvm_mmu_page *child;
A
Avi Kivity 已提交
2632
			u64 pte = *sptep;
M
Marcelo Tosatti 已提交
2633 2634

			child = page_header(pte & PT64_BASE_ADDR_MASK);
2635
			drop_parent_pte(child, sptep);
2636
			kvm_flush_remote_tlbs(vcpu->kvm);
A
Avi Kivity 已提交
2637
		} else if (pfn != spte_to_pfn(*sptep)) {
2638
			pgprintk("hfn old %llx new %llx\n",
A
Avi Kivity 已提交
2639
				 spte_to_pfn(*sptep), pfn);
2640
			drop_spte(vcpu->kvm, sptep);
2641
			kvm_flush_remote_tlbs(vcpu->kvm);
2642 2643
		} else
			was_rmapped = 1;
M
Marcelo Tosatti 已提交
2644
	}
2645

2646 2647
	if (set_spte(vcpu, sptep, pte_access, level, gfn, pfn, speculative,
	      true, host_writable)) {
M
Marcelo Tosatti 已提交
2648
		if (write_fault)
2649
			emulate = true;
2650
		kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
2651
	}
M
Marcelo Tosatti 已提交
2652

2653 2654
	if (unlikely(is_mmio_spte(*sptep)))
		emulate = true;
2655

A
Avi Kivity 已提交
2656
	pgprintk("%s: setting spte %llx\n", __func__, *sptep);
2657
	pgprintk("instantiating %s PTE (%s) at %llx (%llx) addr %p\n",
A
Avi Kivity 已提交
2658
		 is_large_pte(*sptep)? "2MB" : "4kB",
2659 2660
		 *sptep & PT_PRESENT_MASK ?"RW":"R", gfn,
		 *sptep, sptep);
A
Avi Kivity 已提交
2661
	if (!was_rmapped && is_large_pte(*sptep))
M
Marcelo Tosatti 已提交
2662 2663
		++vcpu->kvm->stat.lpages;

2664 2665 2666 2667 2668 2669
	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);
		}
2670
	}
2671

X
Xiao Guangrong 已提交
2672
	kvm_release_pfn_clean(pfn);
2673 2674

	return emulate;
2675 2676
}

D
Dan Williams 已提交
2677
static kvm_pfn_t pte_prefetch_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn,
2678 2679 2680 2681
				     bool no_dirty_log)
{
	struct kvm_memory_slot *slot;

2682
	slot = gfn_to_memslot_dirty_bitmap(vcpu, gfn, no_dirty_log);
2683
	if (!slot)
2684
		return KVM_PFN_ERR_FAULT;
2685

2686
	return gfn_to_pfn_memslot_atomic(slot, gfn);
2687 2688 2689 2690 2691 2692 2693
}

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];
2694
	struct kvm_memory_slot *slot;
2695 2696 2697 2698 2699
	unsigned access = sp->role.access;
	int i, ret;
	gfn_t gfn;

	gfn = kvm_mmu_page_get_gfn(sp, start - sp->spt);
2700 2701
	slot = gfn_to_memslot_dirty_bitmap(vcpu, gfn, access & ACC_WRITE_MASK);
	if (!slot)
2702 2703
		return -1;

2704
	ret = gfn_to_page_many_atomic(slot, gfn, pages, end - start);
2705 2706 2707 2708
	if (ret <= 0)
		return -1;

	for (i = 0; i < ret; i++, gfn++, start++)
2709 2710
		mmu_set_spte(vcpu, start, access, 0, sp->role.level, gfn,
			     page_to_pfn(pages[i]), true, true);
2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726

	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++) {
2727
		if (is_shadow_present_pte(*spte) || spte == sptep) {
2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757
			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);
}

2758
static int __direct_map(struct kvm_vcpu *vcpu, int write, int map_writable,
D
Dan Williams 已提交
2759
			int level, gfn_t gfn, kvm_pfn_t pfn, bool prefault)
2760
{
2761
	struct kvm_shadow_walk_iterator iterator;
2762
	struct kvm_mmu_page *sp;
2763
	int emulate = 0;
2764
	gfn_t pseudo_gfn;
A
Avi Kivity 已提交
2765

2766 2767 2768
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
		return 0;

2769
	for_each_shadow_entry(vcpu, (u64)gfn << PAGE_SHIFT, iterator) {
2770
		if (iterator.level == level) {
2771 2772 2773
			emulate = mmu_set_spte(vcpu, iterator.sptep, ACC_ALL,
					       write, level, gfn, pfn, prefault,
					       map_writable);
2774
			direct_pte_prefetch(vcpu, iterator.sptep);
2775 2776
			++vcpu->stat.pf_fixed;
			break;
A
Avi Kivity 已提交
2777 2778
		}

2779
		drop_large_spte(vcpu, iterator.sptep);
2780
		if (!is_shadow_present_pte(*iterator.sptep)) {
2781 2782 2783 2784
			u64 base_addr = iterator.addr;

			base_addr &= PT64_LVL_ADDR_MASK(iterator.level);
			pseudo_gfn = base_addr >> PAGE_SHIFT;
2785
			sp = kvm_mmu_get_page(vcpu, pseudo_gfn, iterator.addr,
2786
					      iterator.level - 1, 1, ACC_ALL);
2787

2788
			link_shadow_page(vcpu, iterator.sptep, sp);
2789 2790
		}
	}
2791
	return emulate;
A
Avi Kivity 已提交
2792 2793
}

H
Huang Ying 已提交
2794
static void kvm_send_hwpoison_signal(unsigned long address, struct task_struct *tsk)
2795
{
H
Huang Ying 已提交
2796 2797 2798 2799 2800 2801 2802
	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;
2803

H
Huang Ying 已提交
2804
	send_sig_info(SIGBUS, &info, tsk);
2805 2806
}

D
Dan Williams 已提交
2807
static int kvm_handle_bad_page(struct kvm_vcpu *vcpu, gfn_t gfn, kvm_pfn_t pfn)
2808
{
X
Xiao Guangrong 已提交
2809 2810 2811 2812 2813 2814 2815 2816 2817
	/*
	 * 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;

2818
	if (pfn == KVM_PFN_ERR_HWPOISON) {
2819
		kvm_send_hwpoison_signal(kvm_vcpu_gfn_to_hva(vcpu, gfn), current);
2820
		return 0;
2821
	}
2822

2823
	return -EFAULT;
2824 2825
}

2826
static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
D
Dan Williams 已提交
2827 2828
					gfn_t *gfnp, kvm_pfn_t *pfnp,
					int *levelp)
2829
{
D
Dan Williams 已提交
2830
	kvm_pfn_t pfn = *pfnp;
2831 2832 2833 2834 2835 2836 2837 2838 2839
	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.
	 */
2840
	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn) &&
2841
	    level == PT_PAGE_TABLE_LEVEL &&
2842
	    PageTransCompoundMap(pfn_to_page(pfn)) &&
2843
	    !mmu_gfn_lpage_is_disallowed(vcpu, gfn, PT_DIRECTORY_LEVEL)) {
2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861
		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;
2862
			kvm_get_pfn(pfn);
2863 2864 2865 2866 2867
			*pfnp = pfn;
		}
	}
}

2868
static bool handle_abnormal_pfn(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
D
Dan Williams 已提交
2869
				kvm_pfn_t pfn, unsigned access, int *ret_val)
2870 2871
{
	/* The pfn is invalid, report the error! */
2872
	if (unlikely(is_error_pfn(pfn))) {
2873
		*ret_val = kvm_handle_bad_page(vcpu, gfn, pfn);
2874
		return true;
2875 2876
	}

2877
	if (unlikely(is_noslot_pfn(pfn)))
2878 2879
		vcpu_cache_mmio_info(vcpu, gva, gfn, access);

2880
	return false;
2881 2882
}

2883
static bool page_fault_can_be_fast(u32 error_code)
2884
{
2885 2886 2887 2888 2889 2890 2891
	/*
	 * 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;

2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903
	/*
	 * #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;
}

2904 2905 2906 2907
/*
 * Returns true if the SPTE was fixed successfully. Otherwise,
 * someone else modified the SPTE from its original value.
 */
2908
static bool
2909 2910
fast_pf_fix_direct_spte(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
			u64 *sptep, u64 spte)
2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921
{
	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);

2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933
	/*
	 * 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.
	 */
2934 2935 2936 2937
	if (cmpxchg64(sptep, spte, spte | PT_WRITABLE_MASK) != spte)
		return false;

	kvm_vcpu_mark_page_dirty(vcpu, gfn);
2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950

	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;
2951
	struct kvm_mmu_page *sp;
2952
	bool fault_handled = false;
2953
	u64 spte = 0ull;
2954
	uint retry_count = 0;
2955

2956 2957 2958
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
		return false;

2959
	if (!page_fault_can_be_fast(error_code))
2960 2961 2962 2963 2964 2965 2966
		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;

2967 2968 2969 2970 2971 2972 2973 2974 2975
	do {
		/*
		 * If the mapping has been changed, let the vcpu fault on the
		 * same address again.
		 */
		if (!is_shadow_present_pte(spte)) {
			fault_handled = true;
			break;
		}
2976

2977 2978 2979
		sp = page_header(__pa(iterator.sptep));
		if (!is_last_spte(spte, sp->role.level))
			break;
2980

2981 2982 2983 2984 2985 2986 2987 2988 2989 2990
		/*
		 * 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)) {
			fault_handled = true;
			break;
		}
2991

2992 2993 2994 2995 2996 2997
		/*
		 * Currently, to simplify the code, only the spte
		 * write-protected by dirty-log can be fast fixed.
		 */
		if (!spte_can_locklessly_be_made_writable(spte))
			break;
2998

2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031
		/*
		 * 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)
			break;

		/*
		 * 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.
		 */
		fault_handled = fast_pf_fix_direct_spte(vcpu, sp,
							iterator.sptep, spte);
		if (fault_handled)
			break;

		if (++retry_count > 4) {
			printk_once(KERN_WARNING
				"kvm: Fast #PF retrying more than 4 times.\n");
			break;
		}

		spte = mmu_spte_get_lockless(iterator.sptep);

	} while (true);
3032

X
Xiao Guangrong 已提交
3033
	trace_fast_page_fault(vcpu, gva, error_code, iterator.sptep,
3034
			      spte, fault_handled);
3035 3036
	walk_shadow_page_lockless_end(vcpu);

3037
	return fault_handled;
3038 3039
}

3040
static bool try_async_pf(struct kvm_vcpu *vcpu, bool prefault, gfn_t gfn,
D
Dan Williams 已提交
3041
			 gva_t gva, kvm_pfn_t *pfn, bool write, bool *writable);
3042
static void make_mmu_pages_available(struct kvm_vcpu *vcpu);
3043

3044 3045
static int nonpaging_map(struct kvm_vcpu *vcpu, gva_t v, u32 error_code,
			 gfn_t gfn, bool prefault)
3046 3047
{
	int r;
3048
	int level;
3049
	bool force_pt_level = false;
D
Dan Williams 已提交
3050
	kvm_pfn_t pfn;
3051
	unsigned long mmu_seq;
3052
	bool map_writable, write = error_code & PFERR_WRITE_MASK;
3053

3054
	level = mapping_level(vcpu, gfn, &force_pt_level);
3055 3056 3057 3058 3059 3060 3061 3062
	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;
3063

3064
		gfn &= ~(KVM_PAGES_PER_HPAGE(level) - 1);
3065
	}
M
Marcelo Tosatti 已提交
3066

3067 3068 3069
	if (fast_page_fault(vcpu, v, level, error_code))
		return 0;

3070
	mmu_seq = vcpu->kvm->mmu_notifier_seq;
3071
	smp_rmb();
3072

3073
	if (try_async_pf(vcpu, prefault, gfn, v, &pfn, write, &map_writable))
3074
		return 0;
3075

3076 3077
	if (handle_abnormal_pfn(vcpu, v, gfn, pfn, ACC_ALL, &r))
		return r;
3078

3079
	spin_lock(&vcpu->kvm->mmu_lock);
3080
	if (mmu_notifier_retry(vcpu->kvm, mmu_seq))
3081
		goto out_unlock;
3082
	make_mmu_pages_available(vcpu);
3083 3084
	if (likely(!force_pt_level))
		transparent_hugepage_adjust(vcpu, &gfn, &pfn, &level);
3085
	r = __direct_map(vcpu, write, map_writable, level, gfn, pfn, prefault);
3086 3087
	spin_unlock(&vcpu->kvm->mmu_lock);

3088
	return r;
3089 3090 3091 3092 3093

out_unlock:
	spin_unlock(&vcpu->kvm->mmu_lock);
	kvm_release_pfn_clean(pfn);
	return 0;
3094 3095 3096
}


3097 3098 3099
static void mmu_free_roots(struct kvm_vcpu *vcpu)
{
	int i;
3100
	struct kvm_mmu_page *sp;
3101
	LIST_HEAD(invalid_list);
3102

3103
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
A
Avi Kivity 已提交
3104
		return;
3105

3106 3107 3108
	if (vcpu->arch.mmu.shadow_root_level == PT64_ROOT_LEVEL &&
	    (vcpu->arch.mmu.root_level == PT64_ROOT_LEVEL ||
	     vcpu->arch.mmu.direct_map)) {
3109
		hpa_t root = vcpu->arch.mmu.root_hpa;
3110

3111
		spin_lock(&vcpu->kvm->mmu_lock);
3112 3113
		sp = page_header(root);
		--sp->root_count;
3114 3115 3116 3117
		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);
		}
3118
		spin_unlock(&vcpu->kvm->mmu_lock);
3119
		vcpu->arch.mmu.root_hpa = INVALID_PAGE;
3120 3121
		return;
	}
3122 3123

	spin_lock(&vcpu->kvm->mmu_lock);
3124
	for (i = 0; i < 4; ++i) {
3125
		hpa_t root = vcpu->arch.mmu.pae_root[i];
3126

A
Avi Kivity 已提交
3127 3128
		if (root) {
			root &= PT64_BASE_ADDR_MASK;
3129 3130
			sp = page_header(root);
			--sp->root_count;
3131
			if (!sp->root_count && sp->role.invalid)
3132 3133
				kvm_mmu_prepare_zap_page(vcpu->kvm, sp,
							 &invalid_list);
A
Avi Kivity 已提交
3134
		}
3135
		vcpu->arch.mmu.pae_root[i] = INVALID_PAGE;
3136
	}
3137
	kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
3138
	spin_unlock(&vcpu->kvm->mmu_lock);
3139
	vcpu->arch.mmu.root_hpa = INVALID_PAGE;
3140 3141
}

3142 3143 3144 3145 3146
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)) {
3147
		kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu);
3148 3149 3150 3151 3152 3153
		ret = 1;
	}

	return ret;
}

3154 3155 3156
static int mmu_alloc_direct_roots(struct kvm_vcpu *vcpu)
{
	struct kvm_mmu_page *sp;
3157
	unsigned i;
3158 3159 3160

	if (vcpu->arch.mmu.shadow_root_level == PT64_ROOT_LEVEL) {
		spin_lock(&vcpu->kvm->mmu_lock);
3161
		make_mmu_pages_available(vcpu);
3162
		sp = kvm_mmu_get_page(vcpu, 0, 0, PT64_ROOT_LEVEL, 1, ACC_ALL);
3163 3164 3165 3166 3167 3168 3169
		++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];

3170
			MMU_WARN_ON(VALID_PAGE(root));
3171
			spin_lock(&vcpu->kvm->mmu_lock);
3172
			make_mmu_pages_available(vcpu);
3173
			sp = kvm_mmu_get_page(vcpu, i << (30 - PAGE_SHIFT),
3174
					i << 30, PT32_ROOT_LEVEL, 1, ACC_ALL);
3175 3176 3177 3178 3179
			root = __pa(sp->spt);
			++sp->root_count;
			spin_unlock(&vcpu->kvm->mmu_lock);
			vcpu->arch.mmu.pae_root[i] = root | PT_PRESENT_MASK;
		}
3180
		vcpu->arch.mmu.root_hpa = __pa(vcpu->arch.mmu.pae_root);
3181 3182 3183 3184 3185 3186 3187
	} else
		BUG();

	return 0;
}

static int mmu_alloc_shadow_roots(struct kvm_vcpu *vcpu)
3188
{
3189
	struct kvm_mmu_page *sp;
3190 3191 3192
	u64 pdptr, pm_mask;
	gfn_t root_gfn;
	int i;
3193

3194
	root_gfn = vcpu->arch.mmu.get_cr3(vcpu) >> PAGE_SHIFT;
3195

3196 3197 3198 3199 3200 3201 3202 3203
	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) {
3204
		hpa_t root = vcpu->arch.mmu.root_hpa;
3205

3206
		MMU_WARN_ON(VALID_PAGE(root));
3207

3208
		spin_lock(&vcpu->kvm->mmu_lock);
3209
		make_mmu_pages_available(vcpu);
3210
		sp = kvm_mmu_get_page(vcpu, root_gfn, 0, PT64_ROOT_LEVEL,
3211
				      0, ACC_ALL);
3212 3213
		root = __pa(sp->spt);
		++sp->root_count;
3214
		spin_unlock(&vcpu->kvm->mmu_lock);
3215
		vcpu->arch.mmu.root_hpa = root;
3216
		return 0;
3217
	}
3218

3219 3220
	/*
	 * We shadow a 32 bit page table. This may be a legacy 2-level
3221 3222
	 * 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.
3223
	 */
3224 3225 3226 3227
	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;

3228
	for (i = 0; i < 4; ++i) {
3229
		hpa_t root = vcpu->arch.mmu.pae_root[i];
3230

3231
		MMU_WARN_ON(VALID_PAGE(root));
3232
		if (vcpu->arch.mmu.root_level == PT32E_ROOT_LEVEL) {
3233
			pdptr = vcpu->arch.mmu.get_pdptr(vcpu, i);
B
Bandan Das 已提交
3234
			if (!(pdptr & PT_PRESENT_MASK)) {
3235
				vcpu->arch.mmu.pae_root[i] = 0;
A
Avi Kivity 已提交
3236 3237
				continue;
			}
A
Avi Kivity 已提交
3238
			root_gfn = pdptr >> PAGE_SHIFT;
3239 3240
			if (mmu_check_root(vcpu, root_gfn))
				return 1;
3241
		}
3242
		spin_lock(&vcpu->kvm->mmu_lock);
3243
		make_mmu_pages_available(vcpu);
3244 3245
		sp = kvm_mmu_get_page(vcpu, root_gfn, i << 30, PT32_ROOT_LEVEL,
				      0, ACC_ALL);
3246 3247
		root = __pa(sp->spt);
		++sp->root_count;
3248 3249
		spin_unlock(&vcpu->kvm->mmu_lock);

3250
		vcpu->arch.mmu.pae_root[i] = root | pm_mask;
3251
	}
3252
	vcpu->arch.mmu.root_hpa = __pa(vcpu->arch.mmu.pae_root);
3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278

	/*
	 * 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);
	}

3279
	return 0;
3280 3281
}

3282 3283 3284 3285 3286 3287 3288 3289
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);
}

3290 3291 3292 3293 3294
static void mmu_sync_roots(struct kvm_vcpu *vcpu)
{
	int i;
	struct kvm_mmu_page *sp;

3295 3296 3297
	if (vcpu->arch.mmu.direct_map)
		return;

3298 3299
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
		return;
3300

3301
	vcpu_clear_mmio_info(vcpu, MMIO_GVA_ANY);
3302
	kvm_mmu_audit(vcpu, AUDIT_PRE_SYNC);
3303
	if (vcpu->arch.mmu.root_level == PT64_ROOT_LEVEL) {
3304 3305 3306
		hpa_t root = vcpu->arch.mmu.root_hpa;
		sp = page_header(root);
		mmu_sync_children(vcpu, sp);
3307
		kvm_mmu_audit(vcpu, AUDIT_POST_SYNC);
3308 3309 3310 3311 3312
		return;
	}
	for (i = 0; i < 4; ++i) {
		hpa_t root = vcpu->arch.mmu.pae_root[i];

3313
		if (root && VALID_PAGE(root)) {
3314 3315 3316 3317 3318
			root &= PT64_BASE_ADDR_MASK;
			sp = page_header(root);
			mmu_sync_children(vcpu, sp);
		}
	}
3319
	kvm_mmu_audit(vcpu, AUDIT_POST_SYNC);
3320 3321 3322 3323 3324 3325
}

void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu)
{
	spin_lock(&vcpu->kvm->mmu_lock);
	mmu_sync_roots(vcpu);
3326
	spin_unlock(&vcpu->kvm->mmu_lock);
3327
}
N
Nadav Har'El 已提交
3328
EXPORT_SYMBOL_GPL(kvm_mmu_sync_roots);
3329

3330
static gpa_t nonpaging_gva_to_gpa(struct kvm_vcpu *vcpu, gva_t vaddr,
3331
				  u32 access, struct x86_exception *exception)
A
Avi Kivity 已提交
3332
{
3333 3334
	if (exception)
		exception->error_code = 0;
A
Avi Kivity 已提交
3335 3336 3337
	return vaddr;
}

3338
static gpa_t nonpaging_gva_to_gpa_nested(struct kvm_vcpu *vcpu, gva_t vaddr,
3339 3340
					 u32 access,
					 struct x86_exception *exception)
3341
{
3342 3343
	if (exception)
		exception->error_code = 0;
3344
	return vcpu->arch.nested_mmu.translate_gpa(vcpu, vaddr, access, exception);
3345 3346
}

3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365
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);
}

3366
static bool mmio_info_in_cache(struct kvm_vcpu *vcpu, u64 addr, bool direct)
3367 3368 3369 3370 3371 3372 3373
{
	if (direct)
		return vcpu_match_mmio_gpa(vcpu, addr);

	return vcpu_match_mmio_gva(vcpu, addr);
}

3374 3375 3376
/* 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)
3377 3378
{
	struct kvm_shadow_walk_iterator iterator;
3379 3380 3381
	u64 sptes[PT64_ROOT_LEVEL], spte = 0ull;
	int root, leaf;
	bool reserved = false;
3382

3383
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
3384
		goto exit;
3385

3386
	walk_shadow_page_lockless_begin(vcpu);
3387

3388 3389
	for (shadow_walk_init(&iterator, vcpu, addr),
		 leaf = root = iterator.level;
3390 3391 3392 3393 3394
	     shadow_walk_okay(&iterator);
	     __shadow_walk_next(&iterator, spte)) {
		spte = mmu_spte_get_lockless(iterator.sptep);

		sptes[leaf - 1] = spte;
3395
		leaf--;
3396

3397 3398
		if (!is_shadow_present_pte(spte))
			break;
3399 3400

		reserved |= is_shadow_zero_bits_set(&vcpu->arch.mmu, spte,
3401
						    iterator.level);
3402 3403
	}

3404 3405
	walk_shadow_page_lockless_end(vcpu);

3406 3407 3408
	if (reserved) {
		pr_err("%s: detect reserved bits on spte, addr 0x%llx, dump hierarchy:\n",
		       __func__, addr);
3409
		while (root > leaf) {
3410 3411 3412 3413 3414 3415 3416 3417
			pr_err("------ spte 0x%llx level %d.\n",
			       sptes[root - 1], root);
			root--;
		}
	}
exit:
	*sptep = spte;
	return reserved;
3418 3419
}

3420
int handle_mmio_page_fault(struct kvm_vcpu *vcpu, u64 addr, bool direct)
3421 3422
{
	u64 spte;
3423
	bool reserved;
3424

3425
	if (mmio_info_in_cache(vcpu, addr, direct))
3426
		return RET_MMIO_PF_EMULATE;
3427

3428
	reserved = walk_shadow_page_get_mmio_spte(vcpu, addr, &spte);
3429
	if (WARN_ON(reserved))
3430
		return RET_MMIO_PF_BUG;
3431 3432 3433 3434 3435

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

3436
		if (!check_mmio_spte(vcpu, spte))
3437 3438
			return RET_MMIO_PF_INVALID;

3439 3440
		if (direct)
			addr = 0;
X
Xiao Guangrong 已提交
3441 3442

		trace_handle_mmio_page_fault(addr, gfn, access);
3443
		vcpu_cache_mmio_info(vcpu, addr, gfn, access);
3444
		return RET_MMIO_PF_EMULATE;
3445 3446 3447 3448 3449 3450
	}

	/*
	 * If the page table is zapped by other cpus, let CPU fault again on
	 * the address.
	 */
3451
	return RET_MMIO_PF_RETRY;
3452
}
3453
EXPORT_SYMBOL_GPL(handle_mmio_page_fault);
3454

3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474
static bool page_fault_handle_page_track(struct kvm_vcpu *vcpu,
					 u32 error_code, gfn_t gfn)
{
	if (unlikely(error_code & PFERR_RSVD_MASK))
		return false;

	if (!(error_code & PFERR_PRESENT_MASK) ||
	      !(error_code & PFERR_WRITE_MASK))
		return false;

	/*
	 * guest is writing the page which is write tracked which can
	 * not be fixed by page fault handler.
	 */
	if (kvm_page_track_is_active(vcpu, gfn, KVM_PAGE_TRACK_WRITE))
		return true;

	return false;
}

3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491
static void shadow_page_table_clear_flood(struct kvm_vcpu *vcpu, gva_t addr)
{
	struct kvm_shadow_walk_iterator iterator;
	u64 spte;

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

	walk_shadow_page_lockless_begin(vcpu);
	for_each_shadow_entry_lockless(vcpu, addr, iterator, spte) {
		clear_sp_write_flooding_count(iterator.sptep);
		if (!is_shadow_present_pte(spte))
			break;
	}
	walk_shadow_page_lockless_end(vcpu);
}

A
Avi Kivity 已提交
3492
static int nonpaging_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
3493
				u32 error_code, bool prefault)
A
Avi Kivity 已提交
3494
{
3495
	gfn_t gfn = gva >> PAGE_SHIFT;
3496
	int r;
A
Avi Kivity 已提交
3497

3498
	pgprintk("%s: gva %lx error %x\n", __func__, gva, error_code);
3499

3500 3501
	if (page_fault_handle_page_track(vcpu, error_code, gfn))
		return 1;
3502

3503 3504 3505
	r = mmu_topup_memory_caches(vcpu);
	if (r)
		return r;
3506

3507
	MMU_WARN_ON(!VALID_PAGE(vcpu->arch.mmu.root_hpa));
A
Avi Kivity 已提交
3508 3509


3510
	return nonpaging_map(vcpu, gva & PAGE_MASK,
3511
			     error_code, gfn, prefault);
A
Avi Kivity 已提交
3512 3513
}

3514
static int kvm_arch_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn)
3515 3516
{
	struct kvm_arch_async_pf arch;
X
Xiao Guangrong 已提交
3517

3518
	arch.token = (vcpu->arch.apf.id++ << 12) | vcpu->vcpu_id;
3519
	arch.gfn = gfn;
3520
	arch.direct_map = vcpu->arch.mmu.direct_map;
X
Xiao Guangrong 已提交
3521
	arch.cr3 = vcpu->arch.mmu.get_cr3(vcpu);
3522

3523
	return kvm_setup_async_pf(vcpu, gva, kvm_vcpu_gfn_to_hva(vcpu, gfn), &arch);
3524 3525 3526 3527
}

static bool can_do_async_pf(struct kvm_vcpu *vcpu)
{
3528
	if (unlikely(!lapic_in_kernel(vcpu) ||
3529 3530 3531 3532 3533 3534
		     kvm_event_needs_reinjection(vcpu)))
		return false;

	return kvm_x86_ops->interrupt_allowed(vcpu);
}

3535
static bool try_async_pf(struct kvm_vcpu *vcpu, bool prefault, gfn_t gfn,
D
Dan Williams 已提交
3536
			 gva_t gva, kvm_pfn_t *pfn, bool write, bool *writable)
3537
{
3538
	struct kvm_memory_slot *slot;
3539 3540
	bool async;

3541
	slot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);
3542 3543
	async = false;
	*pfn = __gfn_to_pfn_memslot(slot, gfn, false, &async, write, writable);
3544 3545 3546
	if (!async)
		return false; /* *pfn has correct page already */

3547
	if (!prefault && can_do_async_pf(vcpu)) {
3548
		trace_kvm_try_async_get_page(gva, gfn);
3549 3550 3551 3552 3553 3554 3555 3556
		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;
	}

3557
	*pfn = __gfn_to_pfn_memslot(slot, gfn, false, NULL, write, writable);
3558 3559 3560
	return false;
}

3561 3562 3563 3564 3565 3566 3567 3568 3569 3570
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
3571
static int tdp_page_fault(struct kvm_vcpu *vcpu, gva_t gpa, u32 error_code,
3572
			  bool prefault)
3573
{
D
Dan Williams 已提交
3574
	kvm_pfn_t pfn;
3575
	int r;
3576
	int level;
3577
	bool force_pt_level;
M
Marcelo Tosatti 已提交
3578
	gfn_t gfn = gpa >> PAGE_SHIFT;
3579
	unsigned long mmu_seq;
3580 3581
	int write = error_code & PFERR_WRITE_MASK;
	bool map_writable;
3582

3583
	MMU_WARN_ON(!VALID_PAGE(vcpu->arch.mmu.root_hpa));
3584

3585 3586
	if (page_fault_handle_page_track(vcpu, error_code, gfn))
		return 1;
3587

3588 3589 3590 3591
	r = mmu_topup_memory_caches(vcpu);
	if (r)
		return r;

3592 3593 3594
	force_pt_level = !check_hugepage_cache_consistency(vcpu, gfn,
							   PT_DIRECTORY_LEVEL);
	level = mapping_level(vcpu, gfn, &force_pt_level);
3595
	if (likely(!force_pt_level)) {
3596 3597 3598
		if (level > PT_DIRECTORY_LEVEL &&
		    !check_hugepage_cache_consistency(vcpu, gfn, level))
			level = PT_DIRECTORY_LEVEL;
3599
		gfn &= ~(KVM_PAGES_PER_HPAGE(level) - 1);
3600
	}
3601

3602 3603 3604
	if (fast_page_fault(vcpu, gpa, level, error_code))
		return 0;

3605
	mmu_seq = vcpu->kvm->mmu_notifier_seq;
3606
	smp_rmb();
3607

3608
	if (try_async_pf(vcpu, prefault, gfn, gpa, &pfn, write, &map_writable))
3609 3610
		return 0;

3611 3612 3613
	if (handle_abnormal_pfn(vcpu, 0, gfn, pfn, ACC_ALL, &r))
		return r;

3614
	spin_lock(&vcpu->kvm->mmu_lock);
3615
	if (mmu_notifier_retry(vcpu->kvm, mmu_seq))
3616
		goto out_unlock;
3617
	make_mmu_pages_available(vcpu);
3618 3619
	if (likely(!force_pt_level))
		transparent_hugepage_adjust(vcpu, &gfn, &pfn, &level);
3620
	r = __direct_map(vcpu, write, map_writable, level, gfn, pfn, prefault);
3621 3622 3623
	spin_unlock(&vcpu->kvm->mmu_lock);

	return r;
3624 3625 3626 3627 3628

out_unlock:
	spin_unlock(&vcpu->kvm->mmu_lock);
	kvm_release_pfn_clean(pfn);
	return 0;
3629 3630
}

3631 3632
static void nonpaging_init_context(struct kvm_vcpu *vcpu,
				   struct kvm_mmu *context)
A
Avi Kivity 已提交
3633 3634 3635
{
	context->page_fault = nonpaging_page_fault;
	context->gva_to_gpa = nonpaging_gva_to_gpa;
3636
	context->sync_page = nonpaging_sync_page;
M
Marcelo Tosatti 已提交
3637
	context->invlpg = nonpaging_invlpg;
3638
	context->update_pte = nonpaging_update_pte;
3639
	context->root_level = 0;
A
Avi Kivity 已提交
3640
	context->shadow_root_level = PT32E_ROOT_LEVEL;
A
Avi Kivity 已提交
3641
	context->root_hpa = INVALID_PAGE;
3642
	context->direct_map = true;
3643
	context->nx = false;
A
Avi Kivity 已提交
3644 3645
}

3646
void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
3647
{
3648
	mmu_free_roots(vcpu);
A
Avi Kivity 已提交
3649 3650
}

3651 3652
static unsigned long get_cr3(struct kvm_vcpu *vcpu)
{
3653
	return kvm_read_cr3(vcpu);
3654 3655
}

3656 3657
static void inject_page_fault(struct kvm_vcpu *vcpu,
			      struct x86_exception *fault)
A
Avi Kivity 已提交
3658
{
3659
	vcpu->arch.mmu.inject_page_fault(vcpu, fault);
A
Avi Kivity 已提交
3660 3661
}

3662
static bool sync_mmio_spte(struct kvm_vcpu *vcpu, u64 *sptep, gfn_t gfn,
3663
			   unsigned access, int *nr_present)
3664 3665 3666 3667 3668 3669 3670 3671
{
	if (unlikely(is_mmio_spte(*sptep))) {
		if (gfn != get_mmio_spte_gfn(*sptep)) {
			mmu_spte_clear_no_track(sptep);
			return true;
		}

		(*nr_present)++;
3672
		mark_mmio_spte(vcpu, sptep, gfn, access);
3673 3674 3675 3676 3677 3678
		return true;
	}

	return false;
}

3679 3680
static inline bool is_last_gpte(struct kvm_mmu *mmu,
				unsigned level, unsigned gpte)
A
Avi Kivity 已提交
3681
{
3682 3683 3684 3685 3686 3687
	/*
	 * PT_PAGE_TABLE_LEVEL always terminates.  The RHS has bit 7 set
	 * iff level <= PT_PAGE_TABLE_LEVEL, which for our purpose means
	 * level == PT_PAGE_TABLE_LEVEL; set PT_PAGE_SIZE_MASK in gpte then.
	 */
	gpte |= level - PT_PAGE_TABLE_LEVEL - 1;
A
Avi Kivity 已提交
3688

3689 3690 3691 3692 3693 3694 3695 3696
	/*
	 * The RHS has bit 7 set iff level < mmu->last_nonleaf_level.
	 * If it is clear, there are no large pages at this level, so clear
	 * PT_PAGE_SIZE_MASK in gpte if that is the case.
	 */
	gpte &= level - mmu->last_nonleaf_level;

	return gpte & PT_PAGE_SIZE_MASK;
A
Avi Kivity 已提交
3697 3698
}

3699 3700 3701 3702 3703
#define PTTYPE_EPT 18 /* arbitrary */
#define PTTYPE PTTYPE_EPT
#include "paging_tmpl.h"
#undef PTTYPE

A
Avi Kivity 已提交
3704 3705 3706 3707 3708 3709 3710 3711
#define PTTYPE 64
#include "paging_tmpl.h"
#undef PTTYPE

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

3712 3713 3714 3715
static void
__reset_rsvds_bits_mask(struct kvm_vcpu *vcpu,
			struct rsvd_bits_validate *rsvd_check,
			int maxphyaddr, int level, bool nx, bool gbpages,
3716
			bool pse, bool amd)
3717 3718
{
	u64 exb_bit_rsvd = 0;
3719
	u64 gbpages_bit_rsvd = 0;
3720
	u64 nonleaf_bit8_rsvd = 0;
3721

3722
	rsvd_check->bad_mt_xwr = 0;
3723

3724
	if (!nx)
3725
		exb_bit_rsvd = rsvd_bits(63, 63);
3726
	if (!gbpages)
3727
		gbpages_bit_rsvd = rsvd_bits(7, 7);
3728 3729 3730 3731 3732

	/*
	 * Non-leaf PML4Es and PDPEs reserve bit 8 (which would be the G bit for
	 * leaf entries) on AMD CPUs only.
	 */
3733
	if (amd)
3734 3735
		nonleaf_bit8_rsvd = rsvd_bits(8, 8);

3736
	switch (level) {
3737 3738
	case PT32_ROOT_LEVEL:
		/* no rsvd bits for 2 level 4K page table entries */
3739 3740 3741 3742
		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];
3743

3744
		if (!pse) {
3745
			rsvd_check->rsvd_bits_mask[1][1] = 0;
3746 3747 3748
			break;
		}

3749 3750
		if (is_cpuid_PSE36())
			/* 36bits PSE 4MB page */
3751
			rsvd_check->rsvd_bits_mask[1][1] = rsvd_bits(17, 21);
3752 3753
		else
			/* 32 bits PSE 4MB page */
3754
			rsvd_check->rsvd_bits_mask[1][1] = rsvd_bits(13, 21);
3755 3756
		break;
	case PT32E_ROOT_LEVEL:
3757
		rsvd_check->rsvd_bits_mask[0][2] =
3758
			rsvd_bits(maxphyaddr, 63) |
3759
			rsvd_bits(5, 8) | rsvd_bits(1, 2);	/* PDPTE */
3760
		rsvd_check->rsvd_bits_mask[0][1] = exb_bit_rsvd |
3761
			rsvd_bits(maxphyaddr, 62);	/* PDE */
3762
		rsvd_check->rsvd_bits_mask[0][0] = exb_bit_rsvd |
3763
			rsvd_bits(maxphyaddr, 62); 	/* PTE */
3764
		rsvd_check->rsvd_bits_mask[1][1] = exb_bit_rsvd |
3765 3766
			rsvd_bits(maxphyaddr, 62) |
			rsvd_bits(13, 20);		/* large page */
3767 3768
		rsvd_check->rsvd_bits_mask[1][0] =
			rsvd_check->rsvd_bits_mask[0][0];
3769 3770
		break;
	case PT64_ROOT_LEVEL:
3771 3772
		rsvd_check->rsvd_bits_mask[0][3] = exb_bit_rsvd |
			nonleaf_bit8_rsvd | rsvd_bits(7, 7) |
3773
			rsvd_bits(maxphyaddr, 51);
3774 3775
		rsvd_check->rsvd_bits_mask[0][2] = exb_bit_rsvd |
			nonleaf_bit8_rsvd | gbpages_bit_rsvd |
3776
			rsvd_bits(maxphyaddr, 51);
3777 3778 3779 3780 3781 3782 3783
		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 |
3784
			gbpages_bit_rsvd | rsvd_bits(maxphyaddr, 51) |
3785
			rsvd_bits(13, 29);
3786
		rsvd_check->rsvd_bits_mask[1][1] = exb_bit_rsvd |
3787 3788
			rsvd_bits(maxphyaddr, 51) |
			rsvd_bits(13, 20);		/* large page */
3789 3790
		rsvd_check->rsvd_bits_mask[1][0] =
			rsvd_check->rsvd_bits_mask[0][0];
3791 3792 3793 3794
		break;
	}
}

3795 3796 3797 3798 3799 3800
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),
3801
				is_pse(vcpu), guest_cpuid_is_amd(vcpu));
3802 3803
}

3804 3805 3806
static void
__reset_rsvds_bits_mask_ept(struct rsvd_bits_validate *rsvd_check,
			    int maxphyaddr, bool execonly)
3807
{
3808
	u64 bad_mt_xwr;
3809

3810
	rsvd_check->rsvd_bits_mask[0][3] =
3811
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(3, 7);
3812
	rsvd_check->rsvd_bits_mask[0][2] =
3813
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(3, 6);
3814
	rsvd_check->rsvd_bits_mask[0][1] =
3815
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(3, 6);
3816
	rsvd_check->rsvd_bits_mask[0][0] = rsvd_bits(maxphyaddr, 51);
3817 3818

	/* large page */
3819 3820
	rsvd_check->rsvd_bits_mask[1][3] = rsvd_check->rsvd_bits_mask[0][3];
	rsvd_check->rsvd_bits_mask[1][2] =
3821
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(12, 29);
3822
	rsvd_check->rsvd_bits_mask[1][1] =
3823
		rsvd_bits(maxphyaddr, 51) | rsvd_bits(12, 20);
3824
	rsvd_check->rsvd_bits_mask[1][0] = rsvd_check->rsvd_bits_mask[0][0];
3825

3826 3827 3828 3829 3830 3831 3832 3833
	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);
3834
	}
3835
	rsvd_check->bad_mt_xwr = bad_mt_xwr;
3836 3837
}

3838 3839 3840 3841 3842 3843 3844
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);
}

3845 3846 3847 3848 3849 3850 3851 3852
/*
 * 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)
{
3853 3854
	bool uses_nx = context->nx || context->base_role.smep_andnot_wp;

3855 3856 3857 3858
	/*
	 * Passing "true" to the last argument is okay; it adds a check
	 * on bit 8 of the SPTEs which KVM doesn't use anyway.
	 */
3859 3860
	__reset_rsvds_bits_mask(vcpu, &context->shadow_zero_check,
				boot_cpu_data.x86_phys_bits,
3861
				context->shadow_root_level, uses_nx,
3862 3863
				guest_cpuid_has_gbpages(vcpu), is_pse(vcpu),
				true);
3864 3865 3866
}
EXPORT_SYMBOL_GPL(reset_shadow_zero_bits_mask);

3867 3868 3869 3870 3871 3872
static inline bool boot_cpu_is_amd(void)
{
	WARN_ON_ONCE(!tdp_enabled);
	return shadow_x_mask == 0;
}

3873 3874 3875 3876 3877 3878 3879 3880
/*
 * 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)
{
3881
	if (boot_cpu_is_amd())
3882 3883 3884
		__reset_rsvds_bits_mask(vcpu, &context->shadow_zero_check,
					boot_cpu_data.x86_phys_bits,
					context->shadow_root_level, false,
3885 3886
					boot_cpu_has(X86_FEATURE_GBPAGES),
					true, true);
3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905
	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);
}

3906 3907
static void update_permission_bitmask(struct kvm_vcpu *vcpu,
				      struct kvm_mmu *mmu, bool ept)
3908 3909 3910
{
	unsigned bit, byte, pfec;
	u8 map;
F
Feng Wu 已提交
3911
	bool fault, x, w, u, wf, uf, ff, smapf, cr4_smap, cr4_smep, smap = 0;
3912

F
Feng Wu 已提交
3913
	cr4_smep = kvm_read_cr4_bits(vcpu, X86_CR4_SMEP);
F
Feng Wu 已提交
3914
	cr4_smap = kvm_read_cr4_bits(vcpu, X86_CR4_SMAP);
3915 3916 3917 3918 3919 3920
	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
3921 3922 3923 3924 3925 3926
		/*
		 * 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);
3927 3928 3929 3930 3931
		for (bit = 0; bit < 8; ++bit) {
			x = bit & ACC_EXEC_MASK;
			w = bit & ACC_WRITE_MASK;
			u = bit & ACC_USER_MASK;

3932 3933 3934 3935 3936 3937
			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
3938
				x &= !(cr4_smep && u && !uf);
F
Feng Wu 已提交
3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958

				/*
				 * 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;
3959
			}
3960

F
Feng Wu 已提交
3961 3962
			fault = (ff && !x) || (uf && !u) || (wf && !w) ||
				(smapf && smap);
3963 3964 3965 3966 3967 3968
			map |= fault << bit;
		}
		mmu->permissions[byte] = map;
	}
}

3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043
/*
* PKU is an additional mechanism by which the paging controls access to
* user-mode addresses based on the value in the PKRU register.  Protection
* key violations are reported through a bit in the page fault error code.
* Unlike other bits of the error code, the PK bit is not known at the
* call site of e.g. gva_to_gpa; it must be computed directly in
* permission_fault based on two bits of PKRU, on some machine state (CR4,
* CR0, EFER, CPL), and on other bits of the error code and the page tables.
*
* In particular the following conditions come from the error code, the
* page tables and the machine state:
* - PK is always zero unless CR4.PKE=1 and EFER.LMA=1
* - PK is always zero if RSVD=1 (reserved bit set) or F=1 (instruction fetch)
* - PK is always zero if U=0 in the page tables
* - PKRU.WD is ignored if CR0.WP=0 and the access is a supervisor access.
*
* The PKRU bitmask caches the result of these four conditions.  The error
* code (minus the P bit) and the page table's U bit form an index into the
* PKRU bitmask.  Two bits of the PKRU bitmask are then extracted and ANDed
* with the two bits of the PKRU register corresponding to the protection key.
* For the first three conditions above the bits will be 00, thus masking
* away both AD and WD.  For all reads or if the last condition holds, WD
* only will be masked away.
*/
static void update_pkru_bitmask(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
				bool ept)
{
	unsigned bit;
	bool wp;

	if (ept) {
		mmu->pkru_mask = 0;
		return;
	}

	/* PKEY is enabled only if CR4.PKE and EFER.LMA are both set. */
	if (!kvm_read_cr4_bits(vcpu, X86_CR4_PKE) || !is_long_mode(vcpu)) {
		mmu->pkru_mask = 0;
		return;
	}

	wp = is_write_protection(vcpu);

	for (bit = 0; bit < ARRAY_SIZE(mmu->permissions); ++bit) {
		unsigned pfec, pkey_bits;
		bool check_pkey, check_write, ff, uf, wf, pte_user;

		pfec = bit << 1;
		ff = pfec & PFERR_FETCH_MASK;
		uf = pfec & PFERR_USER_MASK;
		wf = pfec & PFERR_WRITE_MASK;

		/* PFEC.RSVD is replaced by ACC_USER_MASK. */
		pte_user = pfec & PFERR_RSVD_MASK;

		/*
		 * Only need to check the access which is not an
		 * instruction fetch and is to a user page.
		 */
		check_pkey = (!ff && pte_user);
		/*
		 * write access is controlled by PKRU if it is a
		 * user access or CR0.WP = 1.
		 */
		check_write = check_pkey && wf && (uf || wp);

		/* PKRU.AD stops both read and write access. */
		pkey_bits = !!check_pkey;
		/* PKRU.WD stops write access. */
		pkey_bits |= (!!check_write) << 1;

		mmu->pkru_mask |= (pkey_bits & 3) << pfec;
	}
}

4044
static void update_last_nonleaf_level(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu)
A
Avi Kivity 已提交
4045
{
4046 4047 4048 4049 4050
	unsigned root_level = mmu->root_level;

	mmu->last_nonleaf_level = root_level;
	if (root_level == PT32_ROOT_LEVEL && is_pse(vcpu))
		mmu->last_nonleaf_level++;
A
Avi Kivity 已提交
4051 4052
}

4053 4054 4055
static void paging64_init_context_common(struct kvm_vcpu *vcpu,
					 struct kvm_mmu *context,
					 int level)
A
Avi Kivity 已提交
4056
{
4057
	context->nx = is_nx(vcpu);
4058
	context->root_level = level;
4059

4060
	reset_rsvds_bits_mask(vcpu, context);
4061
	update_permission_bitmask(vcpu, context, false);
4062
	update_pkru_bitmask(vcpu, context, false);
4063
	update_last_nonleaf_level(vcpu, context);
A
Avi Kivity 已提交
4064

4065
	MMU_WARN_ON(!is_pae(vcpu));
A
Avi Kivity 已提交
4066 4067
	context->page_fault = paging64_page_fault;
	context->gva_to_gpa = paging64_gva_to_gpa;
4068
	context->sync_page = paging64_sync_page;
M
Marcelo Tosatti 已提交
4069
	context->invlpg = paging64_invlpg;
4070
	context->update_pte = paging64_update_pte;
4071
	context->shadow_root_level = level;
A
Avi Kivity 已提交
4072
	context->root_hpa = INVALID_PAGE;
4073
	context->direct_map = false;
A
Avi Kivity 已提交
4074 4075
}

4076 4077
static void paging64_init_context(struct kvm_vcpu *vcpu,
				  struct kvm_mmu *context)
4078
{
4079
	paging64_init_context_common(vcpu, context, PT64_ROOT_LEVEL);
4080 4081
}

4082 4083
static void paging32_init_context(struct kvm_vcpu *vcpu,
				  struct kvm_mmu *context)
A
Avi Kivity 已提交
4084
{
4085
	context->nx = false;
4086
	context->root_level = PT32_ROOT_LEVEL;
4087

4088
	reset_rsvds_bits_mask(vcpu, context);
4089
	update_permission_bitmask(vcpu, context, false);
4090
	update_pkru_bitmask(vcpu, context, false);
4091
	update_last_nonleaf_level(vcpu, context);
A
Avi Kivity 已提交
4092 4093 4094

	context->page_fault = paging32_page_fault;
	context->gva_to_gpa = paging32_gva_to_gpa;
4095
	context->sync_page = paging32_sync_page;
M
Marcelo Tosatti 已提交
4096
	context->invlpg = paging32_invlpg;
4097
	context->update_pte = paging32_update_pte;
A
Avi Kivity 已提交
4098
	context->shadow_root_level = PT32E_ROOT_LEVEL;
A
Avi Kivity 已提交
4099
	context->root_hpa = INVALID_PAGE;
4100
	context->direct_map = false;
A
Avi Kivity 已提交
4101 4102
}

4103 4104
static void paging32E_init_context(struct kvm_vcpu *vcpu,
				   struct kvm_mmu *context)
A
Avi Kivity 已提交
4105
{
4106
	paging64_init_context_common(vcpu, context, PT32E_ROOT_LEVEL);
A
Avi Kivity 已提交
4107 4108
}

4109
static void init_kvm_tdp_mmu(struct kvm_vcpu *vcpu)
4110
{
4111
	struct kvm_mmu *context = &vcpu->arch.mmu;
4112

4113
	context->base_role.word = 0;
4114
	context->base_role.smm = is_smm(vcpu);
4115
	context->page_fault = tdp_page_fault;
4116
	context->sync_page = nonpaging_sync_page;
M
Marcelo Tosatti 已提交
4117
	context->invlpg = nonpaging_invlpg;
4118
	context->update_pte = nonpaging_update_pte;
4119
	context->shadow_root_level = kvm_x86_ops->get_tdp_level();
4120
	context->root_hpa = INVALID_PAGE;
4121
	context->direct_map = true;
4122
	context->set_cr3 = kvm_x86_ops->set_tdp_cr3;
4123
	context->get_cr3 = get_cr3;
4124
	context->get_pdptr = kvm_pdptr_read;
4125
	context->inject_page_fault = kvm_inject_page_fault;
4126 4127

	if (!is_paging(vcpu)) {
4128
		context->nx = false;
4129 4130 4131
		context->gva_to_gpa = nonpaging_gva_to_gpa;
		context->root_level = 0;
	} else if (is_long_mode(vcpu)) {
4132
		context->nx = is_nx(vcpu);
4133
		context->root_level = PT64_ROOT_LEVEL;
4134 4135
		reset_rsvds_bits_mask(vcpu, context);
		context->gva_to_gpa = paging64_gva_to_gpa;
4136
	} else if (is_pae(vcpu)) {
4137
		context->nx = is_nx(vcpu);
4138
		context->root_level = PT32E_ROOT_LEVEL;
4139 4140
		reset_rsvds_bits_mask(vcpu, context);
		context->gva_to_gpa = paging64_gva_to_gpa;
4141
	} else {
4142
		context->nx = false;
4143
		context->root_level = PT32_ROOT_LEVEL;
4144 4145
		reset_rsvds_bits_mask(vcpu, context);
		context->gva_to_gpa = paging32_gva_to_gpa;
4146 4147
	}

4148
	update_permission_bitmask(vcpu, context, false);
4149
	update_pkru_bitmask(vcpu, context, false);
4150
	update_last_nonleaf_level(vcpu, context);
4151
	reset_tdp_shadow_zero_bits_mask(vcpu, context);
4152 4153
}

4154
void kvm_init_shadow_mmu(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4155
{
4156
	bool smep = kvm_read_cr4_bits(vcpu, X86_CR4_SMEP);
4157
	bool smap = kvm_read_cr4_bits(vcpu, X86_CR4_SMAP);
4158 4159
	struct kvm_mmu *context = &vcpu->arch.mmu;

4160
	MMU_WARN_ON(VALID_PAGE(context->root_hpa));
A
Avi Kivity 已提交
4161 4162

	if (!is_paging(vcpu))
4163
		nonpaging_init_context(vcpu, context);
A
Avi Kivity 已提交
4164
	else if (is_long_mode(vcpu))
4165
		paging64_init_context(vcpu, context);
A
Avi Kivity 已提交
4166
	else if (is_pae(vcpu))
4167
		paging32E_init_context(vcpu, context);
A
Avi Kivity 已提交
4168
	else
4169
		paging32_init_context(vcpu, context);
4170

4171 4172 4173 4174
	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
4175
		= smep && !is_write_protection(vcpu);
4176 4177
	context->base_role.smap_andnot_wp
		= smap && !is_write_protection(vcpu);
4178
	context->base_role.smm = is_smm(vcpu);
4179
	reset_shadow_zero_bits_mask(vcpu, context);
4180 4181 4182
}
EXPORT_SYMBOL_GPL(kvm_init_shadow_mmu);

4183
void kvm_init_shadow_ept_mmu(struct kvm_vcpu *vcpu, bool execonly)
N
Nadav Har'El 已提交
4184
{
4185 4186
	struct kvm_mmu *context = &vcpu->arch.mmu;

4187
	MMU_WARN_ON(VALID_PAGE(context->root_hpa));
N
Nadav Har'El 已提交
4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201

	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);
4202
	update_pkru_bitmask(vcpu, context, true);
N
Nadav Har'El 已提交
4203
	reset_rsvds_bits_mask_ept(vcpu, context, execonly);
4204
	reset_ept_shadow_zero_bits_mask(vcpu, context, execonly);
N
Nadav Har'El 已提交
4205 4206 4207
}
EXPORT_SYMBOL_GPL(kvm_init_shadow_ept_mmu);

4208
static void init_kvm_softmmu(struct kvm_vcpu *vcpu)
4209
{
4210 4211 4212 4213 4214 4215 4216
	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
4217 4218
}

4219
static void init_kvm_nested_mmu(struct kvm_vcpu *vcpu)
4220 4221 4222 4223
{
	struct kvm_mmu *g_context = &vcpu->arch.nested_mmu;

	g_context->get_cr3           = get_cr3;
4224
	g_context->get_pdptr         = kvm_pdptr_read;
4225 4226 4227
	g_context->inject_page_fault = kvm_inject_page_fault;

	/*
4228 4229 4230 4231 4232 4233
	 * Note that arch.mmu.gva_to_gpa translates l2_gpa to l1_gpa using
	 * L1's nested page tables (e.g. EPT12). The nested translation
	 * of l2_gva to l1_gpa is done by arch.nested_mmu.gva_to_gpa using
	 * L2's page tables as the first level of translation and L1's
	 * nested page tables as the second level of translation. Basically
	 * the gva_to_gpa functions between mmu and nested_mmu are swapped.
4234 4235
	 */
	if (!is_paging(vcpu)) {
4236
		g_context->nx = false;
4237 4238 4239
		g_context->root_level = 0;
		g_context->gva_to_gpa = nonpaging_gva_to_gpa_nested;
	} else if (is_long_mode(vcpu)) {
4240
		g_context->nx = is_nx(vcpu);
4241
		g_context->root_level = PT64_ROOT_LEVEL;
4242
		reset_rsvds_bits_mask(vcpu, g_context);
4243 4244
		g_context->gva_to_gpa = paging64_gva_to_gpa_nested;
	} else if (is_pae(vcpu)) {
4245
		g_context->nx = is_nx(vcpu);
4246
		g_context->root_level = PT32E_ROOT_LEVEL;
4247
		reset_rsvds_bits_mask(vcpu, g_context);
4248 4249
		g_context->gva_to_gpa = paging64_gva_to_gpa_nested;
	} else {
4250
		g_context->nx = false;
4251
		g_context->root_level = PT32_ROOT_LEVEL;
4252
		reset_rsvds_bits_mask(vcpu, g_context);
4253 4254 4255
		g_context->gva_to_gpa = paging32_gva_to_gpa_nested;
	}

4256
	update_permission_bitmask(vcpu, g_context, false);
4257
	update_pkru_bitmask(vcpu, g_context, false);
4258
	update_last_nonleaf_level(vcpu, g_context);
4259 4260
}

4261
static void init_kvm_mmu(struct kvm_vcpu *vcpu)
4262
{
4263
	if (mmu_is_nested(vcpu))
4264
		init_kvm_nested_mmu(vcpu);
4265
	else if (tdp_enabled)
4266
		init_kvm_tdp_mmu(vcpu);
4267
	else
4268
		init_kvm_softmmu(vcpu);
4269 4270
}

4271
void kvm_mmu_reset_context(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4272
{
4273
	kvm_mmu_unload(vcpu);
4274
	init_kvm_mmu(vcpu);
A
Avi Kivity 已提交
4275
}
4276
EXPORT_SYMBOL_GPL(kvm_mmu_reset_context);
A
Avi Kivity 已提交
4277 4278

int kvm_mmu_load(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4279
{
4280 4281
	int r;

4282
	r = mmu_topup_memory_caches(vcpu);
A
Avi Kivity 已提交
4283 4284
	if (r)
		goto out;
4285
	r = mmu_alloc_roots(vcpu);
4286
	kvm_mmu_sync_roots(vcpu);
4287 4288
	if (r)
		goto out;
4289
	/* set_cr3() should ensure TLB has been flushed */
4290
	vcpu->arch.mmu.set_cr3(vcpu, vcpu->arch.mmu.root_hpa);
4291 4292
out:
	return r;
A
Avi Kivity 已提交
4293
}
A
Avi Kivity 已提交
4294 4295 4296 4297 4298
EXPORT_SYMBOL_GPL(kvm_mmu_load);

void kvm_mmu_unload(struct kvm_vcpu *vcpu)
{
	mmu_free_roots(vcpu);
4299
	WARN_ON(VALID_PAGE(vcpu->arch.mmu.root_hpa));
A
Avi Kivity 已提交
4300
}
4301
EXPORT_SYMBOL_GPL(kvm_mmu_unload);
A
Avi Kivity 已提交
4302

4303
static void mmu_pte_write_new_pte(struct kvm_vcpu *vcpu,
4304 4305
				  struct kvm_mmu_page *sp, u64 *spte,
				  const void *new)
4306
{
4307
	if (sp->role.level != PT_PAGE_TABLE_LEVEL) {
4308 4309
		++vcpu->kvm->stat.mmu_pde_zapped;
		return;
4310
        }
4311

A
Avi Kivity 已提交
4312
	++vcpu->kvm->stat.mmu_pte_updated;
4313
	vcpu->arch.mmu.update_pte(vcpu, sp, spte, new);
4314 4315
}

4316 4317 4318 4319 4320 4321 4322 4323
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;
4324 4325
	old ^= shadow_nx_mask;
	new ^= shadow_nx_mask;
4326 4327 4328
	return (old & ~new & PT64_PERM_MASK) != 0;
}

4329 4330
static u64 mmu_pte_write_fetch_gpte(struct kvm_vcpu *vcpu, gpa_t *gpa,
				    const u8 *new, int *bytes)
4331
{
4332 4333
	u64 gentry;
	int r;
4334 4335 4336

	/*
	 * Assume that the pte write on a page table of the same type
4337 4338
	 * as the current vcpu paging mode since we update the sptes only
	 * when they have the same mode.
4339
	 */
4340
	if (is_pae(vcpu) && *bytes == 4) {
4341
		/* Handle a 32-bit guest writing two halves of a 64-bit gpte */
4342 4343
		*gpa &= ~(gpa_t)7;
		*bytes = 8;
4344
		r = kvm_vcpu_read_guest(vcpu, *gpa, &gentry, 8);
4345 4346
		if (r)
			gentry = 0;
4347 4348 4349
		new = (const u8 *)&gentry;
	}

4350
	switch (*bytes) {
4351 4352 4353 4354 4355 4356 4357 4358 4359
	case 4:
		gentry = *(const u32 *)new;
		break;
	case 8:
		gentry = *(const u64 *)new;
		break;
	default:
		gentry = 0;
		break;
4360 4361
	}

4362 4363 4364 4365 4366 4367 4368
	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.
 */
4369
static bool detect_write_flooding(struct kvm_mmu_page *sp)
4370
{
4371 4372 4373 4374
	/*
	 * Skip write-flooding detected for the sp whose level is 1, because
	 * it can become unsync, then the guest page is not write-protected.
	 */
4375
	if (sp->role.level == PT_PAGE_TABLE_LEVEL)
4376
		return false;
4377

4378 4379
	atomic_inc(&sp->write_flooding_count);
	return atomic_read(&sp->write_flooding_count) >= 3;
4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395
}

/*
 * 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;
4396 4397 4398 4399 4400 4401 4402 4403

	/*
	 * 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;

4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440
	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;
}

4441
static void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
4442 4443
			      const u8 *new, int bytes,
			      struct kvm_page_track_notifier_node *node)
4444 4445 4446 4447 4448 4449
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
	struct kvm_mmu_page *sp;
	LIST_HEAD(invalid_list);
	u64 entry, gentry, *spte;
	int npte;
4450
	bool remote_flush, local_flush;
4451 4452 4453 4454 4455 4456 4457
	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;
4458
	mask.smm = 1;
4459 4460 4461 4462 4463 4464 4465 4466

	/*
	 * 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;

4467
	remote_flush = local_flush = false;
4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481

	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;
4482
	kvm_mmu_audit(vcpu, AUDIT_PRE_PTE_WRITE);
4483

4484
	for_each_gfn_indirect_valid_sp(vcpu->kvm, sp, gfn) {
4485
		if (detect_write_misaligned(sp, gpa, bytes) ||
4486
		      detect_write_flooding(sp)) {
4487
			kvm_mmu_prepare_zap_page(vcpu->kvm, sp, &invalid_list);
A
Avi Kivity 已提交
4488
			++vcpu->kvm->stat.mmu_flooded;
4489 4490
			continue;
		}
4491 4492 4493 4494 4495

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

4496
		local_flush = true;
4497
		while (npte--) {
4498
			entry = *spte;
4499
			mmu_page_zap_pte(vcpu->kvm, sp, spte);
4500 4501
			if (gentry &&
			      !((sp->role.word ^ vcpu->arch.mmu.base_role.word)
4502
			      & mask.word) && rmap_can_add(vcpu))
4503
				mmu_pte_write_new_pte(vcpu, sp, spte, &gentry);
G
Gleb Natapov 已提交
4504
			if (need_remote_flush(entry, *spte))
4505
				remote_flush = true;
4506
			++spte;
4507 4508
		}
	}
4509
	kvm_mmu_flush_or_zap(vcpu, &invalid_list, remote_flush, local_flush);
4510
	kvm_mmu_audit(vcpu, AUDIT_POST_PTE_WRITE);
4511
	spin_unlock(&vcpu->kvm->mmu_lock);
4512 4513
}

4514 4515
int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva)
{
4516 4517
	gpa_t gpa;
	int r;
4518

4519
	if (vcpu->arch.mmu.direct_map)
4520 4521
		return 0;

4522
	gpa = kvm_mmu_gva_to_gpa_read(vcpu, gva, NULL);
4523 4524

	r = kvm_mmu_unprotect_page(vcpu->kvm, gpa >> PAGE_SHIFT);
4525

4526
	return r;
4527
}
4528
EXPORT_SYMBOL_GPL(kvm_mmu_unprotect_page_virt);
4529

4530
static void make_mmu_pages_available(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4531
{
4532
	LIST_HEAD(invalid_list);
4533

4534 4535 4536
	if (likely(kvm_mmu_available_pages(vcpu->kvm) >= KVM_MIN_FREE_MMU_PAGES))
		return;

4537 4538 4539
	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
4540

A
Avi Kivity 已提交
4541
		++vcpu->kvm->stat.mmu_recycled;
A
Avi Kivity 已提交
4542
	}
4543
	kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
A
Avi Kivity 已提交
4544 4545
}

4546
int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t cr2, u64 error_code,
4547
		       void *insn, int insn_len)
4548
{
4549
	int r, emulation_type = EMULTYPE_RETRY;
4550
	enum emulation_result er;
4551
	bool direct = vcpu->arch.mmu.direct_map || mmu_is_nested(vcpu);
4552

4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563
	if (unlikely(error_code & PFERR_RSVD_MASK)) {
		r = handle_mmio_page_fault(vcpu, cr2, direct);
		if (r == RET_MMIO_PF_EMULATE) {
			emulation_type = 0;
			goto emulate;
		}
		if (r == RET_MMIO_PF_RETRY)
			return 1;
		if (r < 0)
			return r;
	}
4564

4565 4566
	r = vcpu->arch.mmu.page_fault(vcpu, cr2, lower_32_bits(error_code),
				      false);
4567
	if (r < 0)
4568 4569 4570
		return r;
	if (!r)
		return 1;
4571

4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586
	/*
	 * Before emulating the instruction, check if the error code
	 * was due to a RO violation while translating the guest page.
	 * This can occur when using nested virtualization with nested
	 * paging in both guests. If true, we simply unprotect the page
	 * and resume the guest.
	 *
	 * Note: AMD only (since it supports the PFERR_GUEST_PAGE_MASK used
	 *       in PFERR_NEXT_GUEST_PAGE)
	 */
	if (error_code == PFERR_NESTED_GUEST_PAGE) {
		kvm_mmu_unprotect_page(vcpu->kvm, gpa_to_gfn(cr2));
		return 1;
	}

4587
	if (mmio_info_in_cache(vcpu, cr2, direct))
4588
		emulation_type = 0;
4589
emulate:
4590
	er = x86_emulate_instruction(vcpu, cr2, emulation_type, insn, insn_len);
4591 4592 4593 4594

	switch (er) {
	case EMULATE_DONE:
		return 1;
P
Paolo Bonzini 已提交
4595
	case EMULATE_USER_EXIT:
4596
		++vcpu->stat.mmio_exits;
4597
		/* fall through */
4598
	case EMULATE_FAIL:
4599
		return 0;
4600 4601 4602 4603 4604 4605
	default:
		BUG();
	}
}
EXPORT_SYMBOL_GPL(kvm_mmu_page_fault);

M
Marcelo Tosatti 已提交
4606 4607 4608
void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
{
	vcpu->arch.mmu.invlpg(vcpu, gva);
4609
	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
M
Marcelo Tosatti 已提交
4610 4611 4612 4613
	++vcpu->stat.invlpg;
}
EXPORT_SYMBOL_GPL(kvm_mmu_invlpg);

4614 4615 4616 4617 4618 4619
void kvm_enable_tdp(void)
{
	tdp_enabled = true;
}
EXPORT_SYMBOL_GPL(kvm_enable_tdp);

4620 4621 4622 4623 4624 4625
void kvm_disable_tdp(void)
{
	tdp_enabled = false;
}
EXPORT_SYMBOL_GPL(kvm_disable_tdp);

A
Avi Kivity 已提交
4626 4627
static void free_mmu_pages(struct kvm_vcpu *vcpu)
{
4628
	free_page((unsigned long)vcpu->arch.mmu.pae_root);
4629 4630
	if (vcpu->arch.mmu.lm_root != NULL)
		free_page((unsigned long)vcpu->arch.mmu.lm_root);
A
Avi Kivity 已提交
4631 4632 4633 4634
}

static int alloc_mmu_pages(struct kvm_vcpu *vcpu)
{
4635
	struct page *page;
A
Avi Kivity 已提交
4636 4637
	int i;

4638 4639 4640 4641 4642 4643 4644
	/*
	 * 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)
4645 4646
		return -ENOMEM;

4647
	vcpu->arch.mmu.pae_root = page_address(page);
4648
	for (i = 0; i < 4; ++i)
4649
		vcpu->arch.mmu.pae_root[i] = INVALID_PAGE;
4650

A
Avi Kivity 已提交
4651 4652 4653
	return 0;
}

4654
int kvm_mmu_create(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4655
{
4656 4657 4658 4659
	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
4660

4661 4662
	return alloc_mmu_pages(vcpu);
}
A
Avi Kivity 已提交
4663

4664
void kvm_mmu_setup(struct kvm_vcpu *vcpu)
4665
{
4666
	MMU_WARN_ON(VALID_PAGE(vcpu->arch.mmu.root_hpa));
4667

4668
	init_kvm_mmu(vcpu);
A
Avi Kivity 已提交
4669 4670
}

4671
static void kvm_mmu_invalidate_zap_pages_in_memslot(struct kvm *kvm,
4672 4673
			struct kvm_memory_slot *slot,
			struct kvm_page_track_notifier_node *node)
4674 4675 4676 4677
{
	kvm_mmu_invalidate_zap_all_pages(kvm);
}

4678 4679 4680 4681 4682
void kvm_mmu_init_vm(struct kvm *kvm)
{
	struct kvm_page_track_notifier_node *node = &kvm->arch.mmu_sp_tracker;

	node->track_write = kvm_mmu_pte_write;
4683
	node->track_flush_slot = kvm_mmu_invalidate_zap_pages_in_memslot;
4684 4685 4686 4687 4688 4689 4690 4691 4692 4693
	kvm_page_track_register_notifier(kvm, node);
}

void kvm_mmu_uninit_vm(struct kvm *kvm)
{
	struct kvm_page_track_notifier_node *node = &kvm->arch.mmu_sp_tracker;

	kvm_page_track_unregister_notifier(kvm, node);
}

4694
/* The return value indicates if tlb flush on all vcpus is needed. */
4695
typedef bool (*slot_level_handler) (struct kvm *kvm, struct kvm_rmap_head *rmap_head);
4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716 4717 4718 4719 4720 4721 4722 4723 4724 4725 4726 4727 4728 4729 4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747 4748 4749 4750 4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762

/* 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
4763 4764 4765 4766
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;
4767
	int i;
X
Xiao Guangrong 已提交
4768 4769

	spin_lock(&kvm->mmu_lock);
4770 4771 4772 4773 4774 4775 4776 4777 4778
	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
4779

4780 4781 4782 4783
			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
4784 4785 4786 4787 4788
	}

	spin_unlock(&kvm->mmu_lock);
}

4789 4790
static bool slot_rmap_write_protect(struct kvm *kvm,
				    struct kvm_rmap_head *rmap_head)
4791
{
4792
	return __rmap_write_protect(kvm, rmap_head, false);
4793 4794
}

4795 4796
void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
				      struct kvm_memory_slot *memslot)
A
Avi Kivity 已提交
4797
{
4798
	bool flush;
A
Avi Kivity 已提交
4799

4800
	spin_lock(&kvm->mmu_lock);
4801 4802
	flush = slot_handle_all_level(kvm, memslot, slot_rmap_write_protect,
				      false);
4803
	spin_unlock(&kvm->mmu_lock);
4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822

	/*
	 * 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.
	 */
4823 4824
	if (flush)
		kvm_flush_remote_tlbs(kvm);
A
Avi Kivity 已提交
4825
}
4826

4827
static bool kvm_mmu_zap_collapsible_spte(struct kvm *kvm,
4828
					 struct kvm_rmap_head *rmap_head)
4829 4830 4831 4832
{
	u64 *sptep;
	struct rmap_iterator iter;
	int need_tlb_flush = 0;
D
Dan Williams 已提交
4833
	kvm_pfn_t pfn;
4834 4835
	struct kvm_mmu_page *sp;

4836
restart:
4837
	for_each_rmap_spte(rmap_head, &iter, sptep) {
4838 4839 4840 4841
		sp = page_header(__pa(sptep));
		pfn = spte_to_pfn(*sptep);

		/*
4842 4843 4844 4845 4846
		 * 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.
4847 4848 4849
		 */
		if (sp->role.direct &&
			!kvm_is_reserved_pfn(pfn) &&
4850
			PageTransCompoundMap(pfn_to_page(pfn))) {
4851 4852
			drop_spte(kvm, sptep);
			need_tlb_flush = 1;
4853 4854
			goto restart;
		}
4855 4856 4857 4858 4859 4860
	}

	return need_tlb_flush;
}

void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
4861
				   const struct kvm_memory_slot *memslot)
4862
{
4863
	/* FIXME: const-ify all uses of struct kvm_memory_slot.  */
4864
	spin_lock(&kvm->mmu_lock);
4865 4866
	slot_handle_leaf(kvm, (struct kvm_memory_slot *)memslot,
			 kvm_mmu_zap_collapsible_spte, true);
4867 4868 4869
	spin_unlock(&kvm->mmu_lock);
}

4870 4871 4872
void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
				   struct kvm_memory_slot *memslot)
{
4873
	bool flush;
4874 4875

	spin_lock(&kvm->mmu_lock);
4876
	flush = slot_handle_leaf(kvm, memslot, __rmap_clear_dirty, false);
4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893 4894
	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)
{
4895
	bool flush;
4896 4897

	spin_lock(&kvm->mmu_lock);
4898 4899
	flush = slot_handle_large_level(kvm, memslot, slot_rmap_write_protect,
					false);
4900 4901 4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912
	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)
{
4913
	bool flush;
4914 4915

	spin_lock(&kvm->mmu_lock);
4916
	flush = slot_handle_all_level(kvm, memslot, __rmap_set_dirty, false);
4917 4918 4919 4920 4921 4922 4923 4924 4925 4926
	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
4927
#define BATCH_ZAP_PAGES	10
4928 4929 4930
static void kvm_zap_obsolete_pages(struct kvm *kvm)
{
	struct kvm_mmu_page *sp, *node;
X
Xiao Guangrong 已提交
4931
	int batch = 0;
4932 4933 4934 4935

restart:
	list_for_each_entry_safe_reverse(sp, node,
	      &kvm->arch.active_mmu_pages, link) {
X
Xiao Guangrong 已提交
4936 4937
		int ret;

4938 4939 4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952
		/*
		 * 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;

4953 4954 4955 4956
		/*
		 * Need not flush tlb since we only zap the sp with invalid
		 * generation number.
		 */
X
Xiao Guangrong 已提交
4957
		if (batch >= BATCH_ZAP_PAGES &&
4958
		      cond_resched_lock(&kvm->mmu_lock)) {
X
Xiao Guangrong 已提交
4959
			batch = 0;
4960 4961 4962
			goto restart;
		}

4963 4964
		ret = kvm_mmu_prepare_zap_page(kvm, sp,
				&kvm->arch.zapped_obsolete_pages);
X
Xiao Guangrong 已提交
4965 4966 4967
		batch += ret;

		if (ret)
4968 4969 4970
			goto restart;
	}

4971 4972 4973 4974
	/*
	 * Should flush tlb before free page tables since lockless-walking
	 * may use the pages.
	 */
4975
	kvm_mmu_commit_zap_page(kvm, &kvm->arch.zapped_obsolete_pages);
4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989
}

/*
 * 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);
4990
	trace_kvm_mmu_invalidate_zap_all_pages(kvm);
4991 4992
	kvm->arch.mmu_valid_gen++;

4993 4994 4995 4996 4997 4998 4999 5000 5001 5002 5003
	/*
	 * 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);

5004 5005 5006 5007
	kvm_zap_obsolete_pages(kvm);
	spin_unlock(&kvm->mmu_lock);
}

5008 5009 5010 5011 5012
static bool kvm_has_zapped_obsolete_pages(struct kvm *kvm)
{
	return unlikely(!list_empty_careful(&kvm->arch.zapped_obsolete_pages));
}

5013
void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, struct kvm_memslots *slots)
5014 5015 5016 5017 5018
{
	/*
	 * The very rare case: if the generation-number is round,
	 * zap all shadow pages.
	 */
5019
	if (unlikely((slots->generation & MMIO_GEN_MASK) == 0)) {
5020
		kvm_debug_ratelimited("kvm: zapping shadow pages for mmio generation wraparound\n");
5021
		kvm_mmu_invalidate_zap_all_pages(kvm);
5022
	}
5023 5024
}

5025 5026
static unsigned long
mmu_shrink_scan(struct shrinker *shrink, struct shrink_control *sc)
5027 5028
{
	struct kvm *kvm;
5029
	int nr_to_scan = sc->nr_to_scan;
5030
	unsigned long freed = 0;
5031

5032
	spin_lock(&kvm_lock);
5033 5034

	list_for_each_entry(kvm, &vm_list, vm_list) {
5035
		int idx;
5036
		LIST_HEAD(invalid_list);
5037

5038 5039 5040 5041 5042 5043 5044 5045
		/*
		 * 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;
5046 5047 5048 5049 5050 5051
		/*
		 * 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.
		 */
5052 5053
		if (!kvm->arch.n_used_mmu_pages &&
		      !kvm_has_zapped_obsolete_pages(kvm))
5054 5055
			continue;

5056
		idx = srcu_read_lock(&kvm->srcu);
5057 5058
		spin_lock(&kvm->mmu_lock);

5059 5060 5061 5062 5063 5064
		if (kvm_has_zapped_obsolete_pages(kvm)) {
			kvm_mmu_commit_zap_page(kvm,
			      &kvm->arch.zapped_obsolete_pages);
			goto unlock;
		}

5065 5066
		if (prepare_zap_oldest_mmu_page(kvm, &invalid_list))
			freed++;
5067
		kvm_mmu_commit_zap_page(kvm, &invalid_list);
5068

5069
unlock:
5070
		spin_unlock(&kvm->mmu_lock);
5071
		srcu_read_unlock(&kvm->srcu, idx);
5072

5073 5074 5075 5076 5077
		/*
		 * unfair on small ones
		 * per-vm shrinkers cry out
		 * sadness comes quickly
		 */
5078 5079
		list_move_tail(&kvm->vm_list, &vm_list);
		break;
5080 5081
	}

5082
	spin_unlock(&kvm_lock);
5083 5084 5085 5086 5087 5088
	return freed;
}

static unsigned long
mmu_shrink_count(struct shrinker *shrink, struct shrink_control *sc)
{
5089
	return percpu_counter_read_positive(&kvm_total_used_mmu_pages);
5090 5091 5092
}

static struct shrinker mmu_shrinker = {
5093 5094
	.count_objects = mmu_shrink_count,
	.scan_objects = mmu_shrink_scan,
5095 5096 5097
	.seeks = DEFAULT_SEEKS * 10,
};

I
Ingo Molnar 已提交
5098
static void mmu_destroy_caches(void)
5099
{
5100 5101
	if (pte_list_desc_cache)
		kmem_cache_destroy(pte_list_desc_cache);
5102 5103
	if (mmu_page_header_cache)
		kmem_cache_destroy(mmu_page_header_cache);
5104 5105 5106 5107
}

int kvm_mmu_module_init(void)
{
5108 5109
	pte_list_desc_cache = kmem_cache_create("pte_list_desc",
					    sizeof(struct pte_list_desc),
5110
					    0, 0, NULL);
5111
	if (!pte_list_desc_cache)
5112 5113
		goto nomem;

5114 5115
	mmu_page_header_cache = kmem_cache_create("kvm_mmu_page_header",
						  sizeof(struct kvm_mmu_page),
5116
						  0, 0, NULL);
5117 5118 5119
	if (!mmu_page_header_cache)
		goto nomem;

5120
	if (percpu_counter_init(&kvm_total_used_mmu_pages, 0, GFP_KERNEL))
5121 5122
		goto nomem;

5123 5124
	register_shrinker(&mmu_shrinker);

5125 5126 5127
	return 0;

nomem:
5128
	mmu_destroy_caches();
5129 5130 5131
	return -ENOMEM;
}

5132 5133 5134 5135 5136 5137 5138
/*
 * 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;
5139
	struct kvm_memslots *slots;
5140
	struct kvm_memory_slot *memslot;
5141
	int i;
5142

5143 5144
	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
		slots = __kvm_memslots(kvm, i);
5145

5146 5147 5148
		kvm_for_each_memslot(memslot, slots)
			nr_pages += memslot->npages;
	}
5149 5150 5151

	nr_mmu_pages = nr_pages * KVM_PERMILLE_MMU_PAGES / 1000;
	nr_mmu_pages = max(nr_mmu_pages,
5152
			   (unsigned int) KVM_MIN_ALLOC_MMU_PAGES);
5153 5154 5155 5156

	return nr_mmu_pages;
}

5157 5158
void kvm_mmu_destroy(struct kvm_vcpu *vcpu)
{
5159
	kvm_mmu_unload(vcpu);
5160 5161
	free_mmu_pages(vcpu);
	mmu_free_memory_caches(vcpu);
5162 5163 5164 5165 5166 5167 5168
}

void kvm_mmu_module_exit(void)
{
	mmu_destroy_caches();
	percpu_counter_destroy(&kvm_total_used_mmu_pages);
	unregister_shrinker(&mmu_shrinker);
5169 5170
	mmu_audit_disable();
}