mmu.c 105.7 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"
A
Avi Kivity 已提交
25

26
#include <linux/kvm_host.h>
A
Avi Kivity 已提交
27 28 29 30 31
#include <linux/types.h>
#include <linux/string.h>
#include <linux/mm.h>
#include <linux/highmem.h>
#include <linux/module.h>
32
#include <linux/swap.h>
M
Marcelo Tosatti 已提交
33
#include <linux/hugetlb.h>
34
#include <linux/compiler.h>
35
#include <linux/srcu.h>
36
#include <linux/slab.h>
37
#include <linux/uaccess.h>
A
Avi Kivity 已提交
38

A
Avi Kivity 已提交
39 40
#include <asm/page.h>
#include <asm/cmpxchg.h>
41
#include <asm/io.h>
42
#include <asm/vmx.h>
A
Avi Kivity 已提交
43

44 45 46 47 48 49 50
/*
 * 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.
 */
51
bool tdp_enabled = false;
52

53 54 55 56
enum {
	AUDIT_PRE_PAGE_FAULT,
	AUDIT_POST_PAGE_FAULT,
	AUDIT_PRE_PTE_WRITE,
57 58 59
	AUDIT_POST_PTE_WRITE,
	AUDIT_PRE_SYNC,
	AUDIT_POST_SYNC
60
};
61

62
#undef MMU_DEBUG
63 64 65 66 67 68 69 70 71 72 73 74 75

#ifdef MMU_DEBUG

#define pgprintk(x...) do { if (dbg) printk(x); } while (0)
#define rmap_printk(x...) do { if (dbg) printk(x); } while (0)

#else

#define pgprintk(x...) do { } while (0)
#define rmap_printk(x...) do { } while (0)

#endif

76
#ifdef MMU_DEBUG
77
static bool dbg = 0;
78
module_param(dbg, bool, 0644);
79
#endif
A
Avi Kivity 已提交
80

81 82 83
#ifndef MMU_DEBUG
#define ASSERT(x) do { } while (0)
#else
A
Avi Kivity 已提交
84 85 86 87 88
#define ASSERT(x)							\
	if (!(x)) {							\
		printk(KERN_WARNING "assertion failed %s:%d: %s\n",	\
		       __FILE__, __LINE__, #x);				\
	}
89
#endif
A
Avi Kivity 已提交
90

91 92
#define PTE_PREFETCH_NUM		8

93
#define PT_FIRST_AVAIL_BITS_SHIFT 10
A
Avi Kivity 已提交
94 95 96 97 98
#define PT64_SECOND_AVAIL_BITS_SHIFT 52

#define PT64_LEVEL_BITS 9

#define PT64_LEVEL_SHIFT(level) \
M
Mike Day 已提交
99
		(PAGE_SHIFT + (level - 1) * PT64_LEVEL_BITS)
A
Avi Kivity 已提交
100 101 102 103 104 105 106 107

#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
108
		(PAGE_SHIFT + (level - 1) * PT32_LEVEL_BITS)
A
Avi Kivity 已提交
109

110 111 112
#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
113 114 115 116 117

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


118
#define PT64_BASE_ADDR_MASK (((1ULL << 52) - 1) & ~(u64)(PAGE_SIZE-1))
A
Avi Kivity 已提交
119 120
#define PT64_DIR_BASE_ADDR_MASK \
	(PT64_BASE_ADDR_MASK & ~((1ULL << (PAGE_SHIFT + PT64_LEVEL_BITS)) - 1))
121 122 123 124 125 126
#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
127 128 129 130

#define PT32_BASE_ADDR_MASK PAGE_MASK
#define PT32_DIR_BASE_ADDR_MASK \
	(PAGE_MASK & ~((1ULL << (PAGE_SHIFT + PT32_LEVEL_BITS)) - 1))
131 132 133
#define PT32_LVL_ADDR_MASK(level) \
	(PAGE_MASK & ~((1ULL << (PAGE_SHIFT + (((level) - 1) \
					    * PT32_LEVEL_BITS))) - 1))
A
Avi Kivity 已提交
134

135 136
#define PT64_PERM_MASK (PT_PRESENT_MASK | PT_WRITABLE_MASK | PT_USER_MASK \
			| PT64_NX_MASK)
A
Avi Kivity 已提交
137

138 139 140 141 142
#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)

143 144
#include <trace/events/kvm.h>

145 146 147
#define CREATE_TRACE_POINTS
#include "mmutrace.h"

148 149
#define SPTE_HOST_WRITEABLE	(1ULL << PT_FIRST_AVAIL_BITS_SHIFT)
#define SPTE_MMU_WRITEABLE	(1ULL << (PT_FIRST_AVAIL_BITS_SHIFT + 1))
150

151 152
#define SHADOW_PT_INDEX(addr, level) PT64_INDEX(addr, level)

153 154 155
/* make pte_list_desc fit well in cache line */
#define PTE_LIST_EXT 3

156 157 158
struct pte_list_desc {
	u64 *sptes[PTE_LIST_EXT];
	struct pte_list_desc *more;
159 160
};

161 162 163 164
struct kvm_shadow_walk_iterator {
	u64 addr;
	hpa_t shadow_addr;
	u64 *sptep;
165
	int level;
166 167 168 169 170 171 172 173
	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)))

174 175 176 177 178 179
#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))

180
static struct kmem_cache *pte_list_desc_cache;
181
static struct kmem_cache *mmu_page_header_cache;
182
static struct percpu_counter kvm_total_used_mmu_pages;
183

S
Sheng Yang 已提交
184 185 186 187 188
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;
189 190 191
static u64 __read_mostly shadow_mmio_mask;

static void mmu_spte_set(u64 *sptep, u64 spte);
192
static void mmu_free_roots(struct kvm_vcpu *vcpu);
193 194 195 196 197 198 199 200 201 202 203

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

static void mark_mmio_spte(u64 *sptep, u64 gfn, unsigned access)
{
	access &= ACC_WRITE_MASK | ACC_USER_MASK;

X
Xiao Guangrong 已提交
204
	trace_mark_mmio_spte(sptep, gfn, access);
205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231
	mmu_spte_set(sptep, shadow_mmio_mask | access | gfn << PAGE_SHIFT);
}

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

static gfn_t get_mmio_spte_gfn(u64 spte)
{
	return (spte & ~shadow_mmio_mask) >> PAGE_SHIFT;
}

static unsigned get_mmio_spte_access(u64 spte)
{
	return (spte & ~shadow_mmio_mask) & ~PAGE_MASK;
}

static bool set_mmio_spte(u64 *sptep, gfn_t gfn, pfn_t pfn, unsigned access)
{
	if (unlikely(is_noslot_pfn(pfn))) {
		mark_mmio_spte(sptep, gfn, access);
		return true;
	}

	return false;
}
232

233 234 235 236 237
static inline u64 rsvd_bits(int s, int e)
{
	return ((1ULL << (e - s + 1)) - 1) << s;
}

S
Sheng Yang 已提交
238
void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
239
		u64 dirty_mask, u64 nx_mask, u64 x_mask)
S
Sheng Yang 已提交
240 241 242 243 244 245 246 247 248
{
	shadow_user_mask = user_mask;
	shadow_accessed_mask = accessed_mask;
	shadow_dirty_mask = dirty_mask;
	shadow_nx_mask = nx_mask;
	shadow_x_mask = x_mask;
}
EXPORT_SYMBOL_GPL(kvm_mmu_set_mask_ptes);

A
Avi Kivity 已提交
249 250 251 252 253
static int is_cpuid_PSE36(void)
{
	return 1;
}

254 255
static int is_nx(struct kvm_vcpu *vcpu)
{
256
	return vcpu->arch.efer & EFER_NX;
257 258
}

259 260
static int is_shadow_present_pte(u64 pte)
{
261
	return pte & PT_PRESENT_MASK && !is_mmio_spte(pte);
262 263
}

M
Marcelo Tosatti 已提交
264 265 266 267 268
static int is_large_pte(u64 pte)
{
	return pte & PT_PAGE_SIZE_MASK;
}

269
static int is_dirty_gpte(unsigned long pte)
270
{
A
Avi Kivity 已提交
271
	return pte & PT_DIRTY_MASK;
272 273
}

274
static int is_rmap_spte(u64 pte)
275
{
276
	return is_shadow_present_pte(pte);
277 278
}

279 280 281 282
static int is_last_spte(u64 pte, int level)
{
	if (level == PT_PAGE_TABLE_LEVEL)
		return 1;
283
	if (is_large_pte(pte))
284 285 286 287
		return 1;
	return 0;
}

288
static pfn_t spte_to_pfn(u64 pte)
289
{
290
	return (pte & PT64_BASE_ADDR_MASK) >> PAGE_SHIFT;
291 292
}

293 294 295 296 297 298 299
static gfn_t pse36_gfn_delta(u32 gpte)
{
	int shift = 32 - PT32_DIR_PSE36_SHIFT - PAGE_SHIFT;

	return (gpte & PT32_DIR_PSE36_MASK) << shift;
}

300
#ifdef CONFIG_X86_64
A
Avi Kivity 已提交
301
static void __set_spte(u64 *sptep, u64 spte)
302
{
303
	*sptep = spte;
304 305
}

306
static void __update_clear_spte_fast(u64 *sptep, u64 spte)
307
{
308 309 310 311 312 313 314
	*sptep = spte;
}

static u64 __update_clear_spte_slow(u64 *sptep, u64 spte)
{
	return xchg(sptep, spte);
}
315 316 317 318 319

static u64 __get_spte_lockless(u64 *sptep)
{
	return ACCESS_ONCE(*sptep);
}
320 321 322 323 324 325

static bool __check_direct_spte_mmio_pf(u64 spte)
{
	/* It is valid if the spte is zapped. */
	return spte == 0ull;
}
326
#else
327 328 329 330 331 332 333
union split_spte {
	struct {
		u32 spte_low;
		u32 spte_high;
	};
	u64 spte;
};
334

335 336 337 338 339 340 341 342 343 344 345 346
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++;
}

347 348 349
static void __set_spte(u64 *sptep, u64 spte)
{
	union split_spte *ssptep, sspte;
350

351 352 353 354 355 356 357 358 359 360 361 362 363
	ssptep = (union split_spte *)sptep;
	sspte = (union split_spte)spte;

	ssptep->spte_high = sspte.spte_high;

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

	ssptep->spte_low = sspte.spte_low;
364 365
}

366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381
static void __update_clear_spte_fast(u64 *sptep, u64 spte)
{
	union split_spte *ssptep, sspte;

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

	ssptep->spte_low = sspte.spte_low;

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

	ssptep->spte_high = sspte.spte_high;
382
	count_spte_clear(sptep, spte);
383 384 385 386 387 388 389 390 391 392 393
}

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);
394 395
	orig.spte_high = ssptep->spte_high;
	ssptep->spte_high = sspte.spte_high;
396
	count_spte_clear(sptep, spte);
397 398 399

	return orig.spte;
}
400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429

/*
 * The idea using the light way get the spte on x86_32 guest is from
 * gup_get_pte(arch/x86/mm/gup.c).
 * The difference is we can not catch the spte tlb flush if we leave
 * guest mode, so we emulate it by increase clear_spte_count when spte
 * is cleared.
 */
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;
}
430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446

static bool __check_direct_spte_mmio_pf(u64 spte)
{
	union split_spte sspte = (union split_spte)spte;
	u32 high_mmio_mask = shadow_mmio_mask >> 32;

	/* It is valid if the spte is zapped. */
	if (spte == 0ull)
		return true;

	/* It is valid if the spte is being zapped. */
	if (sspte.spte_low == 0ull &&
	    (sspte.spte_high & high_mmio_mask) == high_mmio_mask)
		return true;

	return false;
}
447 448
#endif

449 450 451 452 453
static bool spte_is_locklessly_modifiable(u64 spte)
{
	return !(~spte & (SPTE_HOST_WRITEABLE | SPTE_MMU_WRITEABLE));
}

454 455
static bool spte_has_volatile_bits(u64 spte)
{
456 457 458 459 460 461 462 463 464
	/*
	 * Always atomicly update spte if it can be updated
	 * out of mmu-lock, it can ensure dirty bit is not lost,
	 * also, it can help us to get a stable is_writable_pte()
	 * to ensure tlb flush is not missed.
	 */
	if (spte_is_locklessly_modifiable(spte))
		return true;

465 466 467 468 469 470
	if (!shadow_accessed_mask)
		return false;

	if (!is_shadow_present_pte(spte))
		return false;

471 472
	if ((spte & shadow_accessed_mask) &&
	      (!is_writable_pte(spte) || (spte & shadow_dirty_mask)))
473 474 475 476 477
		return false;

	return true;
}

478 479 480 481 482
static bool spte_is_bit_cleared(u64 old_spte, u64 new_spte, u64 bit_mask)
{
	return (old_spte & bit_mask) && !(new_spte & bit_mask);
}

483 484 485 486 487 488 489 490 491 492 493 494 495 496
/* Rules for using mmu_spte_set:
 * Set the sptep from nonpresent to present.
 * Note: the sptep being assigned *must* be either not present
 * or in a state where the hardware will not attempt to update
 * the spte.
 */
static void mmu_spte_set(u64 *sptep, u64 new_spte)
{
	WARN_ON(is_shadow_present_pte(*sptep));
	__set_spte(sptep, new_spte);
}

/* Rules for using mmu_spte_update:
 * Update the state bits, it means the mapped pfn is not changged.
497 498 499 500 501 502
 *
 * 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.
503
 */
504
static bool mmu_spte_update(u64 *sptep, u64 new_spte)
505
{
506
	u64 old_spte = *sptep;
507
	bool ret = false;
508 509

	WARN_ON(!is_rmap_spte(new_spte));
510

511 512 513 514
	if (!is_shadow_present_pte(old_spte)) {
		mmu_spte_set(sptep, new_spte);
		return ret;
	}
515

516
	if (!spte_has_volatile_bits(old_spte))
517
		__update_clear_spte_fast(sptep, new_spte);
518
	else
519
		old_spte = __update_clear_spte_slow(sptep, new_spte);
520

521 522 523 524 525
	/*
	 * For the spte updated out of mmu-lock is safe, since
	 * we always atomicly update it, see the comments in
	 * spte_has_volatile_bits().
	 */
526 527 528
	if (is_writable_pte(old_spte) && !is_writable_pte(new_spte))
		ret = true;

529
	if (!shadow_accessed_mask)
530
		return ret;
531 532 533 534 535

	if (spte_is_bit_cleared(old_spte, new_spte, shadow_accessed_mask))
		kvm_set_pfn_accessed(spte_to_pfn(old_spte));
	if (spte_is_bit_cleared(old_spte, new_spte, shadow_dirty_mask))
		kvm_set_pfn_dirty(spte_to_pfn(old_spte));
536 537

	return ret;
538 539
}

540 541 542 543 544 545 546 547 548 549 550
/*
 * Rules for using mmu_spte_clear_track_bits:
 * It sets the sptep from present to nonpresent, and track the
 * state bits, it is used to clear the last level sptep.
 */
static int mmu_spte_clear_track_bits(u64 *sptep)
{
	pfn_t pfn;
	u64 old_spte = *sptep;

	if (!spte_has_volatile_bits(old_spte))
551
		__update_clear_spte_fast(sptep, 0ull);
552
	else
553
		old_spte = __update_clear_spte_slow(sptep, 0ull);
554 555 556 557 558

	if (!is_rmap_spte(old_spte))
		return 0;

	pfn = spte_to_pfn(old_spte);
559 560 561 562 563 564 565 566

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

567 568 569 570 571 572 573 574 575 576 577 578 579 580
	if (!shadow_accessed_mask || old_spte & shadow_accessed_mask)
		kvm_set_pfn_accessed(pfn);
	if (!shadow_dirty_mask || (old_spte & shadow_dirty_mask))
		kvm_set_pfn_dirty(pfn);
	return 1;
}

/*
 * Rules for using mmu_spte_clear_no_track:
 * Directly clear spte without caring the state bits of sptep,
 * it is used to set the upper level spte.
 */
static void mmu_spte_clear_no_track(u64 *sptep)
{
581
	__update_clear_spte_fast(sptep, 0ull);
582 583
}

584 585 586 587 588 589 590
static u64 mmu_spte_get_lockless(u64 *sptep)
{
	return __get_spte_lockless(sptep);
}

static void walk_shadow_page_lockless_begin(struct kvm_vcpu *vcpu)
{
591 592 593 594 595 596 597 598 599 600 601
	/*
	 * Prevent page table teardown by making any free-er wait during
	 * kvm_flush_remote_tlbs() IPI to all active vcpus.
	 */
	local_irq_disable();
	vcpu->mode = READING_SHADOW_PAGE_TABLES;
	/*
	 * Make sure a following spte read is not reordered ahead of the write
	 * to vcpu->mode.
	 */
	smp_mb();
602 603 604 605
}

static void walk_shadow_page_lockless_end(struct kvm_vcpu *vcpu)
{
606 607 608 609 610 611 612 613
	/*
	 * Make sure the write to vcpu->mode is not reordered in front of
	 * reads to sptes.  If it does, kvm_commit_zap_page() can see us
	 * OUTSIDE_GUEST_MODE and proceed to free the shadow page table.
	 */
	smp_mb();
	vcpu->mode = OUTSIDE_GUEST_MODE;
	local_irq_enable();
614 615
}

616
static int mmu_topup_memory_cache(struct kvm_mmu_memory_cache *cache,
617
				  struct kmem_cache *base_cache, int min)
618 619 620 621
{
	void *obj;

	if (cache->nobjs >= min)
622
		return 0;
623
	while (cache->nobjs < ARRAY_SIZE(cache->objects)) {
624
		obj = kmem_cache_zalloc(base_cache, GFP_KERNEL);
625
		if (!obj)
626
			return -ENOMEM;
627 628
		cache->objects[cache->nobjs++] = obj;
	}
629
	return 0;
630 631
}

632 633 634 635 636
static int mmu_memory_cache_free_objects(struct kvm_mmu_memory_cache *cache)
{
	return cache->nobjs;
}

637 638
static void mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc,
				  struct kmem_cache *cache)
639 640
{
	while (mc->nobjs)
641
		kmem_cache_free(cache, mc->objects[--mc->nobjs]);
642 643
}

A
Avi Kivity 已提交
644
static int mmu_topup_memory_cache_page(struct kvm_mmu_memory_cache *cache,
645
				       int min)
A
Avi Kivity 已提交
646
{
647
	void *page;
A
Avi Kivity 已提交
648 649 650 651

	if (cache->nobjs >= min)
		return 0;
	while (cache->nobjs < ARRAY_SIZE(cache->objects)) {
652
		page = (void *)__get_free_page(GFP_KERNEL);
A
Avi Kivity 已提交
653 654
		if (!page)
			return -ENOMEM;
655
		cache->objects[cache->nobjs++] = page;
A
Avi Kivity 已提交
656 657 658 659 660 661 662
	}
	return 0;
}

static void mmu_free_memory_cache_page(struct kvm_mmu_memory_cache *mc)
{
	while (mc->nobjs)
663
		free_page((unsigned long)mc->objects[--mc->nobjs]);
A
Avi Kivity 已提交
664 665
}

666
static int mmu_topup_memory_caches(struct kvm_vcpu *vcpu)
667
{
668 669
	int r;

670
	r = mmu_topup_memory_cache(&vcpu->arch.mmu_pte_list_desc_cache,
671
				   pte_list_desc_cache, 8 + PTE_PREFETCH_NUM);
672 673
	if (r)
		goto out;
674
	r = mmu_topup_memory_cache_page(&vcpu->arch.mmu_page_cache, 8);
675 676
	if (r)
		goto out;
677
	r = mmu_topup_memory_cache(&vcpu->arch.mmu_page_header_cache,
678
				   mmu_page_header_cache, 4);
679 680
out:
	return r;
681 682 683 684
}

static void mmu_free_memory_caches(struct kvm_vcpu *vcpu)
{
685 686
	mmu_free_memory_cache(&vcpu->arch.mmu_pte_list_desc_cache,
				pte_list_desc_cache);
687
	mmu_free_memory_cache_page(&vcpu->arch.mmu_page_cache);
688 689
	mmu_free_memory_cache(&vcpu->arch.mmu_page_header_cache,
				mmu_page_header_cache);
690 691
}

692
static void *mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc)
693 694 695 696 697 698 699 700
{
	void *p;

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

701
static struct pte_list_desc *mmu_alloc_pte_list_desc(struct kvm_vcpu *vcpu)
702
{
703
	return mmu_memory_cache_alloc(&vcpu->arch.mmu_pte_list_desc_cache);
704 705
}

706
static void mmu_free_pte_list_desc(struct pte_list_desc *pte_list_desc)
707
{
708
	kmem_cache_free(pte_list_desc_cache, pte_list_desc);
709 710
}

711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726
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
727
/*
728 729
 * 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
730
 */
731 732 733
static struct kvm_lpage_info *lpage_info_slot(gfn_t gfn,
					      struct kvm_memory_slot *slot,
					      int level)
M
Marcelo Tosatti 已提交
734 735 736
{
	unsigned long idx;

737
	idx = gfn_to_index(gfn, slot->base_gfn, level);
738
	return &slot->arch.lpage_info[level - 2][idx];
M
Marcelo Tosatti 已提交
739 740 741 742
}

static void account_shadowed(struct kvm *kvm, gfn_t gfn)
{
743
	struct kvm_memory_slot *slot;
744
	struct kvm_lpage_info *linfo;
745
	int i;
M
Marcelo Tosatti 已提交
746

A
Avi Kivity 已提交
747
	slot = gfn_to_memslot(kvm, gfn);
748 749
	for (i = PT_DIRECTORY_LEVEL;
	     i < PT_PAGE_TABLE_LEVEL + KVM_NR_PAGE_SIZES; ++i) {
750 751
		linfo = lpage_info_slot(gfn, slot, i);
		linfo->write_count += 1;
752
	}
753
	kvm->arch.indirect_shadow_pages++;
M
Marcelo Tosatti 已提交
754 755 756 757
}

static void unaccount_shadowed(struct kvm *kvm, gfn_t gfn)
{
758
	struct kvm_memory_slot *slot;
759
	struct kvm_lpage_info *linfo;
760
	int i;
M
Marcelo Tosatti 已提交
761

A
Avi Kivity 已提交
762
	slot = gfn_to_memslot(kvm, gfn);
763 764
	for (i = PT_DIRECTORY_LEVEL;
	     i < PT_PAGE_TABLE_LEVEL + KVM_NR_PAGE_SIZES; ++i) {
765 766 767
		linfo = lpage_info_slot(gfn, slot, i);
		linfo->write_count -= 1;
		WARN_ON(linfo->write_count < 0);
768
	}
769
	kvm->arch.indirect_shadow_pages--;
M
Marcelo Tosatti 已提交
770 771
}

772 773 774
static int has_wrprotected_page(struct kvm *kvm,
				gfn_t gfn,
				int level)
M
Marcelo Tosatti 已提交
775
{
776
	struct kvm_memory_slot *slot;
777
	struct kvm_lpage_info *linfo;
M
Marcelo Tosatti 已提交
778

A
Avi Kivity 已提交
779
	slot = gfn_to_memslot(kvm, gfn);
M
Marcelo Tosatti 已提交
780
	if (slot) {
781 782
		linfo = lpage_info_slot(gfn, slot, level);
		return linfo->write_count;
M
Marcelo Tosatti 已提交
783 784 785 786 787
	}

	return 1;
}

788
static int host_mapping_level(struct kvm *kvm, gfn_t gfn)
M
Marcelo Tosatti 已提交
789
{
J
Joerg Roedel 已提交
790
	unsigned long page_size;
791
	int i, ret = 0;
M
Marcelo Tosatti 已提交
792

J
Joerg Roedel 已提交
793
	page_size = kvm_host_page_size(kvm, gfn);
M
Marcelo Tosatti 已提交
794

795 796 797 798 799 800 801 802
	for (i = PT_PAGE_TABLE_LEVEL;
	     i < (PT_PAGE_TABLE_LEVEL + KVM_NR_PAGE_SIZES); ++i) {
		if (page_size >= KVM_HPAGE_SIZE(i))
			ret = i;
		else
			break;
	}

803
	return ret;
M
Marcelo Tosatti 已提交
804 805
}

806 807 808
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
809 810
{
	struct kvm_memory_slot *slot;
811 812 813 814 815 816 817 818 819 820 821

	slot = gfn_to_memslot(vcpu->kvm, gfn);
	if (!slot || slot->flags & KVM_MEMSLOT_INVALID ||
	      (no_dirty_log && slot->dirty_bitmap))
		slot = NULL;

	return slot;
}

static bool mapping_level_dirty_bitmap(struct kvm_vcpu *vcpu, gfn_t large_gfn)
{
822
	return !gfn_to_memslot_dirty_bitmap(vcpu, large_gfn, true);
823 824 825 826 827
}

static int mapping_level(struct kvm_vcpu *vcpu, gfn_t large_gfn)
{
	int host_level, level, max_level;
M
Marcelo Tosatti 已提交
828

829 830 831 832 833
	host_level = host_mapping_level(vcpu->kvm, large_gfn);

	if (host_level == PT_PAGE_TABLE_LEVEL)
		return host_level;

834 835 836 837
	max_level = kvm_x86_ops->get_lpage_level() < host_level ?
		kvm_x86_ops->get_lpage_level() : host_level;

	for (level = PT_DIRECTORY_LEVEL; level <= max_level; ++level)
838 839 840 841
		if (has_wrprotected_page(vcpu->kvm, large_gfn, level))
			break;

	return level - 1;
M
Marcelo Tosatti 已提交
842 843
}

844
/*
845
 * Pte mapping structures:
846
 *
847
 * If pte_list bit zero is zero, then pte_list point to the spte.
848
 *
849 850
 * If pte_list bit zero is one, (then pte_list & ~1) points to a struct
 * pte_list_desc containing more mappings.
851
 *
852
 * Returns the number of pte entries before the spte was added or zero if
853 854
 * the spte was not added.
 *
855
 */
856 857
static int pte_list_add(struct kvm_vcpu *vcpu, u64 *spte,
			unsigned long *pte_list)
858
{
859
	struct pte_list_desc *desc;
860
	int i, count = 0;
861

862 863 864 865 866 867 868
	if (!*pte_list) {
		rmap_printk("pte_list_add: %p %llx 0->1\n", spte, *spte);
		*pte_list = (unsigned long)spte;
	} else if (!(*pte_list & 1)) {
		rmap_printk("pte_list_add: %p %llx 1->many\n", spte, *spte);
		desc = mmu_alloc_pte_list_desc(vcpu);
		desc->sptes[0] = (u64 *)*pte_list;
A
Avi Kivity 已提交
869
		desc->sptes[1] = spte;
870
		*pte_list = (unsigned long)desc | 1;
871
		++count;
872
	} else {
873 874 875
		rmap_printk("pte_list_add: %p %llx many->many\n", spte, *spte);
		desc = (struct pte_list_desc *)(*pte_list & ~1ul);
		while (desc->sptes[PTE_LIST_EXT-1] && desc->more) {
876
			desc = desc->more;
877
			count += PTE_LIST_EXT;
878
		}
879 880
		if (desc->sptes[PTE_LIST_EXT-1]) {
			desc->more = mmu_alloc_pte_list_desc(vcpu);
881 882
			desc = desc->more;
		}
A
Avi Kivity 已提交
883
		for (i = 0; desc->sptes[i]; ++i)
884
			++count;
A
Avi Kivity 已提交
885
		desc->sptes[i] = spte;
886
	}
887
	return count;
888 889
}

890 891 892
static void
pte_list_desc_remove_entry(unsigned long *pte_list, struct pte_list_desc *desc,
			   int i, struct pte_list_desc *prev_desc)
893 894 895
{
	int j;

896
	for (j = PTE_LIST_EXT - 1; !desc->sptes[j] && j > i; --j)
897
		;
A
Avi Kivity 已提交
898 899
	desc->sptes[i] = desc->sptes[j];
	desc->sptes[j] = NULL;
900 901 902
	if (j != 0)
		return;
	if (!prev_desc && !desc->more)
903
		*pte_list = (unsigned long)desc->sptes[0];
904 905 906 907
	else
		if (prev_desc)
			prev_desc->more = desc->more;
		else
908 909
			*pte_list = (unsigned long)desc->more | 1;
	mmu_free_pte_list_desc(desc);
910 911
}

912
static void pte_list_remove(u64 *spte, unsigned long *pte_list)
913
{
914 915
	struct pte_list_desc *desc;
	struct pte_list_desc *prev_desc;
916 917
	int i;

918 919
	if (!*pte_list) {
		printk(KERN_ERR "pte_list_remove: %p 0->BUG\n", spte);
920
		BUG();
921 922 923 924
	} else if (!(*pte_list & 1)) {
		rmap_printk("pte_list_remove:  %p 1->0\n", spte);
		if ((u64 *)*pte_list != spte) {
			printk(KERN_ERR "pte_list_remove:  %p 1->BUG\n", spte);
925 926
			BUG();
		}
927
		*pte_list = 0;
928
	} else {
929 930
		rmap_printk("pte_list_remove:  %p many->many\n", spte);
		desc = (struct pte_list_desc *)(*pte_list & ~1ul);
931 932
		prev_desc = NULL;
		while (desc) {
933
			for (i = 0; i < PTE_LIST_EXT && desc->sptes[i]; ++i)
A
Avi Kivity 已提交
934
				if (desc->sptes[i] == spte) {
935
					pte_list_desc_remove_entry(pte_list,
936
							       desc, i,
937 938 939 940 941 942
							       prev_desc);
					return;
				}
			prev_desc = desc;
			desc = desc->more;
		}
943
		pr_err("pte_list_remove: %p many->many\n", spte);
944 945 946 947
		BUG();
	}
}

948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967
typedef void (*pte_list_walk_fn) (u64 *spte);
static void pte_list_walk(unsigned long *pte_list, pte_list_walk_fn fn)
{
	struct pte_list_desc *desc;
	int i;

	if (!*pte_list)
		return;

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

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

968
static unsigned long *__gfn_to_rmap(gfn_t gfn, int level,
969
				    struct kvm_memory_slot *slot)
970
{
971
	unsigned long idx;
972

973
	idx = gfn_to_index(gfn, slot->base_gfn, level);
974
	return &slot->arch.rmap[level - PT_PAGE_TABLE_LEVEL][idx];
975 976
}

977 978 979 980 981 982 983 984
/*
 * Take gfn and return the reverse mapping to it.
 */
static unsigned long *gfn_to_rmap(struct kvm *kvm, gfn_t gfn, int level)
{
	struct kvm_memory_slot *slot;

	slot = gfn_to_memslot(kvm, gfn);
985
	return __gfn_to_rmap(gfn, level, slot);
986 987
}

988 989 990 991 992 993 994 995
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);
}

996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018
static int rmap_add(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn)
{
	struct kvm_mmu_page *sp;
	unsigned long *rmapp;

	sp = page_header(__pa(spte));
	kvm_mmu_page_set_gfn(sp, spte - sp->spt, gfn);
	rmapp = gfn_to_rmap(vcpu->kvm, gfn, sp->role.level);
	return pte_list_add(vcpu, spte, rmapp);
}

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

	sp = page_header(__pa(spte));
	gfn = kvm_mmu_page_get_gfn(sp, spte - sp->spt);
	rmapp = gfn_to_rmap(kvm, gfn, sp->role.level);
	pte_list_remove(spte, rmapp);
}

1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079
/*
 * Used by the following functions to iterate through the sptes linked by a
 * rmap.  All fields are private and not assumed to be used outside.
 */
struct rmap_iterator {
	/* private fields */
	struct pte_list_desc *desc;	/* holds the sptep if not NULL */
	int pos;			/* index of the sptep */
};

/*
 * Iteration must be started by this function.  This should also be used after
 * removing/dropping sptes from the rmap link because in such cases the
 * information in the itererator may not be valid.
 *
 * Returns sptep if found, NULL otherwise.
 */
static u64 *rmap_get_first(unsigned long rmap, struct rmap_iterator *iter)
{
	if (!rmap)
		return NULL;

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

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

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

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

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

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

	return NULL;
}

1080
static void drop_spte(struct kvm *kvm, u64 *sptep)
1081
{
1082
	if (mmu_spte_clear_track_bits(sptep))
1083
		rmap_remove(kvm, sptep);
A
Avi Kivity 已提交
1084 1085
}

1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106

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);
}

/*
1107 1108 1109 1110 1111 1112 1113 1114 1115 1116
 * Write-protect on the specified @sptep, @pt_protect indicates whether
 * spte writ-protection is caused by protecting shadow page table.
 * @flush indicates whether tlb need be flushed.
 *
 * Note: write protection is difference between drity logging and spte
 * 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.
1117 1118 1119
 *
 * Return true if the spte is dropped.
 */
1120 1121
static bool
spte_write_protect(struct kvm *kvm, u64 *sptep, bool *flush, bool pt_protect)
1122 1123 1124
{
	u64 spte = *sptep;

1125 1126
	if (!is_writable_pte(spte) &&
	      !(pt_protect && spte_is_locklessly_modifiable(spte)))
1127 1128 1129 1130
		return false;

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

1131 1132
	if (__drop_large_spte(kvm, sptep)) {
		*flush |= true;
1133
		return true;
1134
	}
1135

1136 1137
	if (pt_protect)
		spte &= ~SPTE_MMU_WRITEABLE;
1138
	spte = spte & ~PT_WRITABLE_MASK;
1139 1140

	*flush |= mmu_spte_update(sptep, spte);
1141 1142 1143
	return false;
}

1144 1145
static bool __rmap_write_protect(struct kvm *kvm, unsigned long *rmapp,
				 int level, bool pt_protect)
1146
{
1147 1148
	u64 *sptep;
	struct rmap_iterator iter;
1149
	bool flush = false;
1150

1151 1152
	for (sptep = rmap_get_first(*rmapp, &iter); sptep;) {
		BUG_ON(!(*sptep & PT_PRESENT_MASK));
1153
		if (spte_write_protect(kvm, sptep, &flush, pt_protect)) {
1154
			sptep = rmap_get_first(*rmapp, &iter);
1155
			continue;
1156
		}
1157

1158
		sptep = rmap_get_next(&iter);
1159
	}
1160

1161
	return flush;
1162 1163
}

1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176
/**
 * kvm_mmu_write_protect_pt_masked - write protect selected PT level pages
 * @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.
 */
void kvm_mmu_write_protect_pt_masked(struct kvm *kvm,
				     struct kvm_memory_slot *slot,
				     gfn_t gfn_offset, unsigned long mask)
1177 1178 1179
{
	unsigned long *rmapp;

1180
	while (mask) {
1181 1182
		rmapp = __gfn_to_rmap(slot->base_gfn + gfn_offset + __ffs(mask),
				      PT_PAGE_TABLE_LEVEL, slot);
1183
		__rmap_write_protect(kvm, rmapp, PT_PAGE_TABLE_LEVEL, false);
M
Marcelo Tosatti 已提交
1184

1185 1186 1187
		/* clear the first set bit */
		mask &= mask - 1;
	}
1188 1189
}

1190
static bool rmap_write_protect(struct kvm *kvm, u64 gfn)
1191 1192
{
	struct kvm_memory_slot *slot;
1193 1194
	unsigned long *rmapp;
	int i;
1195
	bool write_protected = false;
1196 1197

	slot = gfn_to_memslot(kvm, gfn);
1198 1199 1200 1201

	for (i = PT_PAGE_TABLE_LEVEL;
	     i < PT_PAGE_TABLE_LEVEL + KVM_NR_PAGE_SIZES; ++i) {
		rmapp = __gfn_to_rmap(gfn, i, slot);
1202
		write_protected |= __rmap_write_protect(kvm, rmapp, i, true);
1203 1204 1205
	}

	return write_protected;
1206 1207
}

F
Frederik Deweerdt 已提交
1208
static int kvm_unmap_rmapp(struct kvm *kvm, unsigned long *rmapp,
1209
			   struct kvm_memory_slot *slot, unsigned long data)
1210
{
1211 1212
	u64 *sptep;
	struct rmap_iterator iter;
1213 1214
	int need_tlb_flush = 0;

1215 1216 1217 1218 1219
	while ((sptep = rmap_get_first(*rmapp, &iter))) {
		BUG_ON(!(*sptep & PT_PRESENT_MASK));
		rmap_printk("kvm_rmap_unmap_hva: spte %p %llx\n", sptep, *sptep);

		drop_spte(kvm, sptep);
1220 1221
		need_tlb_flush = 1;
	}
1222

1223 1224 1225
	return need_tlb_flush;
}

F
Frederik Deweerdt 已提交
1226
static int kvm_set_pte_rmapp(struct kvm *kvm, unsigned long *rmapp,
1227
			     struct kvm_memory_slot *slot, unsigned long data)
1228
{
1229 1230
	u64 *sptep;
	struct rmap_iterator iter;
1231
	int need_flush = 0;
1232
	u64 new_spte;
1233 1234 1235 1236 1237
	pte_t *ptep = (pte_t *)data;
	pfn_t new_pfn;

	WARN_ON(pte_huge(*ptep));
	new_pfn = pte_pfn(*ptep);
1238 1239 1240 1241 1242

	for (sptep = rmap_get_first(*rmapp, &iter); sptep;) {
		BUG_ON(!is_shadow_present_pte(*sptep));
		rmap_printk("kvm_set_pte_rmapp: spte %p %llx\n", sptep, *sptep);

1243
		need_flush = 1;
1244

1245
		if (pte_write(*ptep)) {
1246 1247
			drop_spte(kvm, sptep);
			sptep = rmap_get_first(*rmapp, &iter);
1248
		} else {
1249
			new_spte = *sptep & ~PT64_BASE_ADDR_MASK;
1250 1251 1252 1253
			new_spte |= (u64)new_pfn << PAGE_SHIFT;

			new_spte &= ~PT_WRITABLE_MASK;
			new_spte &= ~SPTE_HOST_WRITEABLE;
1254
			new_spte &= ~shadow_accessed_mask;
1255 1256 1257 1258

			mmu_spte_clear_track_bits(sptep);
			mmu_spte_set(sptep, new_spte);
			sptep = rmap_get_next(&iter);
1259 1260
		}
	}
1261

1262 1263 1264 1265 1266 1267
	if (need_flush)
		kvm_flush_remote_tlbs(kvm);

	return 0;
}

1268 1269 1270 1271 1272 1273
static int kvm_handle_hva_range(struct kvm *kvm,
				unsigned long start,
				unsigned long end,
				unsigned long data,
				int (*handler)(struct kvm *kvm,
					       unsigned long *rmapp,
1274
					       struct kvm_memory_slot *slot,
1275
					       unsigned long data))
1276
{
1277
	int j;
1278
	int ret = 0;
1279
	struct kvm_memslots *slots;
1280
	struct kvm_memory_slot *memslot;
1281

1282
	slots = kvm_memslots(kvm);
1283

1284
	kvm_for_each_memslot(memslot, slots) {
1285
		unsigned long hva_start, hva_end;
1286
		gfn_t gfn_start, gfn_end;
1287

1288 1289 1290 1291 1292 1293 1294
		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)} =
1295
		 * {gfn_start, gfn_start+1, ..., gfn_end-1}.
1296
		 */
1297
		gfn_start = hva_to_gfn_memslot(hva_start, memslot);
1298
		gfn_end = hva_to_gfn_memslot(hva_end + PAGE_SIZE - 1, memslot);
1299

1300 1301 1302 1303
		for (j = PT_PAGE_TABLE_LEVEL;
		     j < PT_PAGE_TABLE_LEVEL + KVM_NR_PAGE_SIZES; ++j) {
			unsigned long idx, idx_end;
			unsigned long *rmapp;
1304

1305 1306 1307 1308 1309 1310
			/*
			 * {idx(page_j) | page_j intersects with
			 *  [hva_start, hva_end)} = {idx, idx+1, ..., idx_end}.
			 */
			idx = gfn_to_index(gfn_start, memslot->base_gfn, j);
			idx_end = gfn_to_index(gfn_end - 1, memslot->base_gfn, j);
1311

1312
			rmapp = __gfn_to_rmap(gfn_start, j, memslot);
1313

1314 1315
			for (; idx <= idx_end; ++idx)
				ret |= handler(kvm, rmapp++, memslot, data);
1316 1317 1318
		}
	}

1319
	return ret;
1320 1321
}

1322 1323 1324
static int kvm_handle_hva(struct kvm *kvm, unsigned long hva,
			  unsigned long data,
			  int (*handler)(struct kvm *kvm, unsigned long *rmapp,
1325
					 struct kvm_memory_slot *slot,
1326 1327 1328
					 unsigned long data))
{
	return kvm_handle_hva_range(kvm, hva, hva + 1, data, handler);
1329 1330 1331 1332
}

int kvm_unmap_hva(struct kvm *kvm, unsigned long hva)
{
1333 1334 1335
	return kvm_handle_hva(kvm, hva, 0, kvm_unmap_rmapp);
}

1336 1337 1338 1339 1340
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);
}

1341 1342
void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte)
{
F
Frederik Deweerdt 已提交
1343
	kvm_handle_hva(kvm, hva, (unsigned long)&pte, kvm_set_pte_rmapp);
1344 1345
}

F
Frederik Deweerdt 已提交
1346
static int kvm_age_rmapp(struct kvm *kvm, unsigned long *rmapp,
1347
			 struct kvm_memory_slot *slot, unsigned long data)
1348
{
1349
	u64 *sptep;
1350
	struct rmap_iterator uninitialized_var(iter);
1351 1352
	int young = 0;

1353
	/*
1354 1355
	 * In case of absence of EPT Access and Dirty Bits supports,
	 * emulate the accessed bit for EPT, by checking if this page has
1356 1357 1358 1359 1360
	 * 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.
	 */
1361 1362 1363 1364
	if (!shadow_accessed_mask) {
		young = kvm_unmap_rmapp(kvm, rmapp, slot, data);
		goto out;
	}
1365

1366 1367
	for (sptep = rmap_get_first(*rmapp, &iter); sptep;
	     sptep = rmap_get_next(&iter)) {
1368
		BUG_ON(!is_shadow_present_pte(*sptep));
1369

1370
		if (*sptep & shadow_accessed_mask) {
1371
			young = 1;
1372 1373
			clear_bit((ffs(shadow_accessed_mask) - 1),
				 (unsigned long *)sptep);
1374 1375
		}
	}
1376 1377 1378
out:
	/* @data has hva passed to kvm_age_hva(). */
	trace_kvm_age_page(data, slot, young);
1379 1380 1381
	return young;
}

A
Andrea Arcangeli 已提交
1382
static int kvm_test_age_rmapp(struct kvm *kvm, unsigned long *rmapp,
1383
			      struct kvm_memory_slot *slot, unsigned long data)
A
Andrea Arcangeli 已提交
1384
{
1385 1386
	u64 *sptep;
	struct rmap_iterator iter;
A
Andrea Arcangeli 已提交
1387 1388 1389 1390 1391 1392 1393 1394 1395 1396
	int young = 0;

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

1397 1398
	for (sptep = rmap_get_first(*rmapp, &iter); sptep;
	     sptep = rmap_get_next(&iter)) {
1399
		BUG_ON(!is_shadow_present_pte(*sptep));
1400

1401
		if (*sptep & shadow_accessed_mask) {
A
Andrea Arcangeli 已提交
1402 1403 1404 1405 1406 1407 1408 1409
			young = 1;
			break;
		}
	}
out:
	return young;
}

1410 1411
#define RMAP_RECYCLE_THRESHOLD 1000

1412
static void rmap_recycle(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn)
1413 1414
{
	unsigned long *rmapp;
1415 1416 1417
	struct kvm_mmu_page *sp;

	sp = page_header(__pa(spte));
1418

1419
	rmapp = gfn_to_rmap(vcpu->kvm, gfn, sp->role.level);
1420

1421
	kvm_unmap_rmapp(vcpu->kvm, rmapp, NULL, 0);
1422 1423 1424
	kvm_flush_remote_tlbs(vcpu->kvm);
}

1425 1426
int kvm_age_hva(struct kvm *kvm, unsigned long hva)
{
1427
	return kvm_handle_hva(kvm, hva, hva, kvm_age_rmapp);
1428 1429
}

A
Andrea Arcangeli 已提交
1430 1431 1432 1433 1434
int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
{
	return kvm_handle_hva(kvm, hva, 0, kvm_test_age_rmapp);
}

1435
#ifdef MMU_DEBUG
1436
static int is_empty_shadow_page(u64 *spt)
A
Avi Kivity 已提交
1437
{
1438 1439 1440
	u64 *pos;
	u64 *end;

1441
	for (pos = spt, end = pos + PAGE_SIZE / sizeof(u64); pos != end; pos++)
1442
		if (is_shadow_present_pte(*pos)) {
1443
			printk(KERN_ERR "%s: %p %llx\n", __func__,
1444
			       pos, *pos);
A
Avi Kivity 已提交
1445
			return 0;
1446
		}
A
Avi Kivity 已提交
1447 1448
	return 1;
}
1449
#endif
A
Avi Kivity 已提交
1450

1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462
/*
 * 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);
}

1463 1464 1465 1466 1467 1468 1469
/*
 * Remove the sp from shadow page cache, after call it,
 * we can not find this sp from the cache, and the shadow
 * page table is still valid.
 * It should be under the protection of mmu lock.
 */
static void kvm_mmu_isolate_page(struct kvm_mmu_page *sp)
1470
{
1471
	ASSERT(is_empty_shadow_page(sp->spt));
1472
	hlist_del(&sp->hash_link);
1473
	if (!sp->role.direct)
1474
		free_page((unsigned long)sp->gfns);
1475 1476 1477 1478 1479 1480 1481 1482 1483 1484
}

/*
 * Free the shadow page table and the sp, we can do it
 * out of the protection of mmu lock.
 */
static void kvm_mmu_free_page(struct kvm_mmu_page *sp)
{
	list_del(&sp->link);
	free_page((unsigned long)sp->spt);
1485
	kmem_cache_free(mmu_page_header_cache, sp);
1486 1487
}

1488 1489
static unsigned kvm_page_table_hashfn(gfn_t gfn)
{
1490
	return gfn & ((1 << KVM_MMU_HASH_SHIFT) - 1);
1491 1492
}

1493
static void mmu_page_add_parent_pte(struct kvm_vcpu *vcpu,
1494
				    struct kvm_mmu_page *sp, u64 *parent_pte)
1495 1496 1497 1498
{
	if (!parent_pte)
		return;

1499
	pte_list_add(vcpu, parent_pte, &sp->parent_ptes);
1500 1501
}

1502
static void mmu_page_remove_parent_pte(struct kvm_mmu_page *sp,
1503 1504
				       u64 *parent_pte)
{
1505
	pte_list_remove(parent_pte, &sp->parent_ptes);
1506 1507
}

1508 1509 1510 1511
static void drop_parent_pte(struct kvm_mmu_page *sp,
			    u64 *parent_pte)
{
	mmu_page_remove_parent_pte(sp, parent_pte);
1512
	mmu_spte_clear_no_track(parent_pte);
1513 1514
}

1515 1516
static struct kvm_mmu_page *kvm_mmu_alloc_page(struct kvm_vcpu *vcpu,
					       u64 *parent_pte, int direct)
M
Marcelo Tosatti 已提交
1517
{
1518
	struct kvm_mmu_page *sp;
1519 1520
	sp = mmu_memory_cache_alloc(&vcpu->arch.mmu_page_header_cache);
	sp->spt = mmu_memory_cache_alloc(&vcpu->arch.mmu_page_cache);
1521
	if (!direct)
1522
		sp->gfns = mmu_memory_cache_alloc(&vcpu->arch.mmu_page_cache);
1523 1524
	set_page_private(virt_to_page(sp->spt), (unsigned long)sp);
	list_add(&sp->link, &vcpu->kvm->arch.active_mmu_pages);
1525
	bitmap_zero(sp->slot_bitmap, KVM_MEM_SLOTS_NUM);
1526 1527 1528 1529
	sp->parent_ptes = 0;
	mmu_page_add_parent_pte(vcpu, sp, parent_pte);
	kvm_mod_used_mmu_pages(vcpu->kvm, +1);
	return sp;
M
Marcelo Tosatti 已提交
1530 1531
}

1532
static void mark_unsync(u64 *spte);
1533
static void kvm_mmu_mark_parents_unsync(struct kvm_mmu_page *sp)
1534
{
1535
	pte_list_walk(&sp->parent_ptes, mark_unsync);
1536 1537
}

1538
static void mark_unsync(u64 *spte)
1539
{
1540
	struct kvm_mmu_page *sp;
1541
	unsigned int index;
1542

1543
	sp = page_header(__pa(spte));
1544 1545
	index = spte - sp->spt;
	if (__test_and_set_bit(index, sp->unsync_child_bitmap))
1546
		return;
1547
	if (sp->unsync_children++)
1548
		return;
1549
	kvm_mmu_mark_parents_unsync(sp);
1550 1551
}

1552
static int nonpaging_sync_page(struct kvm_vcpu *vcpu,
1553
			       struct kvm_mmu_page *sp)
1554 1555 1556 1557
{
	return 1;
}

M
Marcelo Tosatti 已提交
1558 1559 1560 1561
static void nonpaging_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
{
}

1562 1563
static void nonpaging_update_pte(struct kvm_vcpu *vcpu,
				 struct kvm_mmu_page *sp, u64 *spte,
1564
				 const void *pte)
1565 1566 1567 1568
{
	WARN_ON(1);
}

1569 1570 1571 1572 1573 1574 1575 1576 1577 1578
#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;
};

1579 1580
static int mmu_pages_add(struct kvm_mmu_pages *pvec, struct kvm_mmu_page *sp,
			 int idx)
1581
{
1582
	int i;
1583

1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598
	if (sp->unsync)
		for (i=0; i < pvec->nr; i++)
			if (pvec->page[i].sp == sp)
				return 0;

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

static int __mmu_unsync_walk(struct kvm_mmu_page *sp,
			   struct kvm_mmu_pages *pvec)
{
	int i, ret, nr_unsync_leaf = 0;
1599

1600
	for_each_set_bit(i, sp->unsync_child_bitmap, 512) {
1601
		struct kvm_mmu_page *child;
1602 1603
		u64 ent = sp->spt[i];

1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632
		if (!is_shadow_present_pte(ent) || is_large_pte(ent))
			goto clear_child_bitmap;

		child = page_header(ent & PT64_BASE_ADDR_MASK);

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

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

		continue;

clear_child_bitmap:
		__clear_bit(i, sp->unsync_child_bitmap);
		sp->unsync_children--;
		WARN_ON((int)sp->unsync_children < 0);
1633 1634 1635
	}


1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646
	return nr_unsync_leaf;
}

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

	mmu_pages_add(pvec, sp, 0);
	return __mmu_unsync_walk(sp, pvec);
1647 1648 1649 1650 1651
}

static void kvm_unlink_unsync_page(struct kvm *kvm, struct kvm_mmu_page *sp)
{
	WARN_ON(!sp->unsync);
1652
	trace_kvm_mmu_sync_page(sp);
1653 1654 1655 1656
	sp->unsync = 0;
	--kvm->stat.mmu_unsync;
}

1657 1658 1659 1660
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);
1661

1662 1663
#define for_each_gfn_sp(kvm, sp, gfn, pos)				\
  hlist_for_each_entry(sp, pos,						\
1664 1665 1666
   &(kvm)->arch.mmu_page_hash[kvm_page_table_hashfn(gfn)], hash_link)	\
	if ((sp)->gfn != (gfn)) {} else

1667 1668
#define for_each_gfn_indirect_valid_sp(kvm, sp, gfn, pos)		\
  hlist_for_each_entry(sp, pos,						\
1669 1670 1671 1672
   &(kvm)->arch.mmu_page_hash[kvm_page_table_hashfn(gfn)], hash_link)	\
		if ((sp)->gfn != (gfn) || (sp)->role.direct ||		\
			(sp)->role.invalid) {} else

1673
/* @sp->gfn should be write-protected at the call site */
1674
static int __kvm_sync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
1675
			   struct list_head *invalid_list, bool clear_unsync)
1676
{
1677
	if (sp->role.cr4_pae != !!is_pae(vcpu)) {
1678
		kvm_mmu_prepare_zap_page(vcpu->kvm, sp, invalid_list);
1679 1680 1681
		return 1;
	}

1682
	if (clear_unsync)
1683 1684
		kvm_unlink_unsync_page(vcpu->kvm, sp);

1685
	if (vcpu->arch.mmu.sync_page(vcpu, sp)) {
1686
		kvm_mmu_prepare_zap_page(vcpu->kvm, sp, invalid_list);
1687 1688 1689 1690 1691 1692 1693
		return 1;
	}

	kvm_mmu_flush_tlb(vcpu);
	return 0;
}

1694 1695 1696
static int kvm_sync_page_transient(struct kvm_vcpu *vcpu,
				   struct kvm_mmu_page *sp)
{
1697
	LIST_HEAD(invalid_list);
1698 1699
	int ret;

1700
	ret = __kvm_sync_page(vcpu, sp, &invalid_list, false);
1701
	if (ret)
1702 1703
		kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);

1704 1705 1706
	return ret;
}

1707 1708 1709 1710 1711 1712 1713
#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

1714 1715
static int kvm_sync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
			 struct list_head *invalid_list)
1716
{
1717
	return __kvm_sync_page(vcpu, sp, invalid_list, true);
1718 1719
}

1720 1721 1722 1723
/* @gfn should be write-protected at the call site */
static void kvm_sync_pages(struct kvm_vcpu *vcpu,  gfn_t gfn)
{
	struct kvm_mmu_page *s;
1724
	struct hlist_node *node;
1725
	LIST_HEAD(invalid_list);
1726 1727
	bool flush = false;

1728
	for_each_gfn_indirect_valid_sp(vcpu->kvm, s, gfn, node) {
1729
		if (!s->unsync)
1730 1731 1732
			continue;

		WARN_ON(s->role.level != PT_PAGE_TABLE_LEVEL);
1733
		kvm_unlink_unsync_page(vcpu->kvm, s);
1734
		if ((s->role.cr4_pae != !!is_pae(vcpu)) ||
1735
			(vcpu->arch.mmu.sync_page(vcpu, s))) {
1736
			kvm_mmu_prepare_zap_page(vcpu->kvm, s, &invalid_list);
1737 1738 1739 1740 1741
			continue;
		}
		flush = true;
	}

1742
	kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
1743 1744 1745 1746
	if (flush)
		kvm_mmu_flush_tlb(vcpu);
}

1747 1748 1749
struct mmu_page_path {
	struct kvm_mmu_page *parent[PT64_ROOT_LEVEL-1];
	unsigned int idx[PT64_ROOT_LEVEL-1];
1750 1751
};

1752 1753 1754 1755 1756 1757
#define for_each_sp(pvec, sp, parents, i)			\
		for (i = mmu_pages_next(&pvec, &parents, -1),	\
			sp = pvec.page[i].sp;			\
			i < pvec.nr && ({ sp = pvec.page[i].sp; 1;});	\
			i = mmu_pages_next(&pvec, &parents, i))

1758 1759 1760
static int mmu_pages_next(struct kvm_mmu_pages *pvec,
			  struct mmu_page_path *parents,
			  int i)
1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778
{
	int n;

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

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

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

	return n;
}

1779
static void mmu_pages_clear_parents(struct mmu_page_path *parents)
1780
{
1781 1782 1783 1784 1785
	struct kvm_mmu_page *sp;
	unsigned int level = 0;

	do {
		unsigned int idx = parents->idx[level];
1786

1787 1788 1789 1790 1791 1792 1793 1794 1795
		sp = parents->parent[level];
		if (!sp)
			return;

		--sp->unsync_children;
		WARN_ON((int)sp->unsync_children < 0);
		__clear_bit(idx, sp->unsync_child_bitmap);
		level++;
	} while (level < PT64_ROOT_LEVEL-1 && !sp->unsync_children);
1796 1797
}

1798 1799 1800
static void kvm_mmu_pages_init(struct kvm_mmu_page *parent,
			       struct mmu_page_path *parents,
			       struct kvm_mmu_pages *pvec)
1801
{
1802 1803 1804
	parents->parent[parent->role.level-1] = NULL;
	pvec->nr = 0;
}
1805

1806 1807 1808 1809 1810 1811 1812
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;
1813
	LIST_HEAD(invalid_list);
1814 1815 1816

	kvm_mmu_pages_init(parent, &parents, &pages);
	while (mmu_unsync_walk(parent, &pages)) {
1817
		bool protected = false;
1818 1819 1820 1821 1822 1823 1824

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

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

1825
		for_each_sp(pages, sp, parents, i) {
1826
			kvm_sync_page(vcpu, sp, &invalid_list);
1827 1828
			mmu_pages_clear_parents(&parents);
		}
1829
		kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
1830
		cond_resched_lock(&vcpu->kvm->mmu_lock);
1831 1832
		kvm_mmu_pages_init(parent, &parents, &pages);
	}
1833 1834
}

1835 1836 1837 1838 1839 1840 1841 1842
static void init_shadow_page_table(struct kvm_mmu_page *sp)
{
	int i;

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

1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854
static void __clear_sp_write_flooding_count(struct kvm_mmu_page *sp)
{
	sp->write_flooding_count = 0;
}

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

	__clear_sp_write_flooding_count(sp);
}

1855 1856 1857 1858
static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
					     gfn_t gfn,
					     gva_t gaddr,
					     unsigned level,
1859
					     int direct,
1860
					     unsigned access,
1861
					     u64 *parent_pte)
1862 1863 1864
{
	union kvm_mmu_page_role role;
	unsigned quadrant;
1865
	struct kvm_mmu_page *sp;
1866
	struct hlist_node *node;
1867
	bool need_sync = false;
1868

1869
	role = vcpu->arch.mmu.base_role;
1870
	role.level = level;
1871
	role.direct = direct;
1872
	if (role.direct)
1873
		role.cr4_pae = 0;
1874
	role.access = access;
1875 1876
	if (!vcpu->arch.mmu.direct_map
	    && vcpu->arch.mmu.root_level <= PT32_ROOT_LEVEL) {
1877 1878 1879 1880
		quadrant = gaddr >> (PAGE_SHIFT + (PT64_PT_BITS * level));
		quadrant &= (1 << ((PT32_PT_BITS - PT64_PT_BITS) * level)) - 1;
		role.quadrant = quadrant;
	}
1881
	for_each_gfn_sp(vcpu->kvm, sp, gfn, node) {
1882 1883
		if (!need_sync && sp->unsync)
			need_sync = true;
1884

1885 1886
		if (sp->role.word != role.word)
			continue;
1887

1888 1889
		if (sp->unsync && kvm_sync_page_transient(vcpu, sp))
			break;
1890

1891 1892
		mmu_page_add_parent_pte(vcpu, sp, parent_pte);
		if (sp->unsync_children) {
1893
			kvm_make_request(KVM_REQ_MMU_SYNC, vcpu);
1894 1895 1896
			kvm_mmu_mark_parents_unsync(sp);
		} else if (sp->unsync)
			kvm_mmu_mark_parents_unsync(sp);
1897

1898
		__clear_sp_write_flooding_count(sp);
1899 1900 1901
		trace_kvm_mmu_get_page(sp, false);
		return sp;
	}
A
Avi Kivity 已提交
1902
	++vcpu->kvm->stat.mmu_cache_miss;
1903
	sp = kvm_mmu_alloc_page(vcpu, parent_pte, direct);
1904 1905 1906 1907
	if (!sp)
		return sp;
	sp->gfn = gfn;
	sp->role = role;
1908 1909
	hlist_add_head(&sp->hash_link,
		&vcpu->kvm->arch.mmu_page_hash[kvm_page_table_hashfn(gfn)]);
1910
	if (!direct) {
1911 1912
		if (rmap_write_protect(vcpu->kvm, gfn))
			kvm_flush_remote_tlbs(vcpu->kvm);
1913 1914 1915
		if (level > PT_PAGE_TABLE_LEVEL && need_sync)
			kvm_sync_pages(vcpu, gfn);

1916 1917
		account_shadowed(vcpu->kvm, gfn);
	}
1918
	init_shadow_page_table(sp);
A
Avi Kivity 已提交
1919
	trace_kvm_mmu_get_page(sp, true);
1920
	return sp;
1921 1922
}

1923 1924 1925 1926 1927 1928
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;
1929 1930 1931 1932 1933 1934

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

1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948
	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;
1949

1950 1951 1952 1953 1954
	iterator->index = SHADOW_PT_INDEX(iterator->addr, iterator->level);
	iterator->sptep	= ((u64 *)__va(iterator->shadow_addr)) + iterator->index;
	return true;
}

1955 1956
static void __shadow_walk_next(struct kvm_shadow_walk_iterator *iterator,
			       u64 spte)
1957
{
1958
	if (is_last_spte(spte, iterator->level)) {
1959 1960 1961 1962
		iterator->level = 0;
		return;
	}

1963
	iterator->shadow_addr = spte & PT64_BASE_ADDR_MASK;
1964 1965 1966
	--iterator->level;
}

1967 1968 1969 1970 1971
static void shadow_walk_next(struct kvm_shadow_walk_iterator *iterator)
{
	return __shadow_walk_next(iterator, *iterator->sptep);
}

1972 1973 1974 1975 1976 1977 1978
static void link_shadow_page(u64 *sptep, struct kvm_mmu_page *sp)
{
	u64 spte;

	spte = __pa(sp->spt)
		| PT_PRESENT_MASK | PT_ACCESSED_MASK
		| PT_WRITABLE_MASK | PT_USER_MASK;
1979
	mmu_spte_set(sptep, spte);
1980 1981
}

1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998
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;

1999
		drop_parent_pte(child, sptep);
2000 2001 2002 2003
		kvm_flush_remote_tlbs(vcpu->kvm);
	}
}

X
Xiao Guangrong 已提交
2004
static bool mmu_page_zap_pte(struct kvm *kvm, struct kvm_mmu_page *sp,
2005 2006 2007 2008 2009 2010 2011
			     u64 *spte)
{
	u64 pte;
	struct kvm_mmu_page *child;

	pte = *spte;
	if (is_shadow_present_pte(pte)) {
X
Xiao Guangrong 已提交
2012
		if (is_last_spte(pte, sp->role.level)) {
2013
			drop_spte(kvm, spte);
X
Xiao Guangrong 已提交
2014 2015 2016
			if (is_large_pte(pte))
				--kvm->stat.lpages;
		} else {
2017
			child = page_header(pte & PT64_BASE_ADDR_MASK);
2018
			drop_parent_pte(child, spte);
2019
		}
X
Xiao Guangrong 已提交
2020 2021 2022 2023
		return true;
	}

	if (is_mmio_spte(pte))
2024
		mmu_spte_clear_no_track(spte);
2025

X
Xiao Guangrong 已提交
2026
	return false;
2027 2028
}

2029
static void kvm_mmu_page_unlink_children(struct kvm *kvm,
2030
					 struct kvm_mmu_page *sp)
2031
{
2032 2033
	unsigned i;

2034 2035
	for (i = 0; i < PT64_ENT_PER_PAGE; ++i)
		mmu_page_zap_pte(kvm, sp, sp->spt + i);
2036 2037
}

2038
static void kvm_mmu_put_page(struct kvm_mmu_page *sp, u64 *parent_pte)
2039
{
2040
	mmu_page_remove_parent_pte(sp, parent_pte);
2041 2042
}

2043
static void kvm_mmu_unlink_parents(struct kvm *kvm, struct kvm_mmu_page *sp)
2044
{
2045 2046
	u64 *sptep;
	struct rmap_iterator iter;
2047

2048 2049
	while ((sptep = rmap_get_first(sp->parent_ptes, &iter)))
		drop_parent_pte(sp, sptep);
2050 2051
}

2052
static int mmu_zap_unsync_children(struct kvm *kvm,
2053 2054
				   struct kvm_mmu_page *parent,
				   struct list_head *invalid_list)
2055
{
2056 2057 2058
	int i, zapped = 0;
	struct mmu_page_path parents;
	struct kvm_mmu_pages pages;
2059

2060
	if (parent->role.level == PT_PAGE_TABLE_LEVEL)
2061
		return 0;
2062 2063 2064 2065 2066 2067

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

		for_each_sp(pages, sp, parents, i) {
2068
			kvm_mmu_prepare_zap_page(kvm, sp, invalid_list);
2069
			mmu_pages_clear_parents(&parents);
2070
			zapped++;
2071 2072 2073 2074 2075
		}
		kvm_mmu_pages_init(parent, &parents, &pages);
	}

	return zapped;
2076 2077
}

2078 2079
static int kvm_mmu_prepare_zap_page(struct kvm *kvm, struct kvm_mmu_page *sp,
				    struct list_head *invalid_list)
2080
{
2081
	int ret;
A
Avi Kivity 已提交
2082

2083
	trace_kvm_mmu_prepare_zap_page(sp);
2084
	++kvm->stat.mmu_shadow_zapped;
2085
	ret = mmu_zap_unsync_children(kvm, sp, invalid_list);
2086
	kvm_mmu_page_unlink_children(kvm, sp);
2087
	kvm_mmu_unlink_parents(kvm, sp);
2088
	if (!sp->role.invalid && !sp->role.direct)
A
Avi Kivity 已提交
2089
		unaccount_shadowed(kvm, sp->gfn);
2090 2091
	if (sp->unsync)
		kvm_unlink_unsync_page(kvm, sp);
2092
	if (!sp->root_count) {
2093 2094
		/* Count self */
		ret++;
2095
		list_move(&sp->link, invalid_list);
2096
		kvm_mod_used_mmu_pages(kvm, -1);
2097
	} else {
A
Avi Kivity 已提交
2098
		list_move(&sp->link, &kvm->arch.active_mmu_pages);
2099 2100
		kvm_reload_remote_mmus(kvm);
	}
2101 2102

	sp->role.invalid = 1;
2103
	return ret;
2104 2105
}

2106 2107 2108 2109 2110 2111 2112 2113
static void kvm_mmu_commit_zap_page(struct kvm *kvm,
				    struct list_head *invalid_list)
{
	struct kvm_mmu_page *sp;

	if (list_empty(invalid_list))
		return;

2114 2115 2116 2117 2118
	/*
	 * wmb: make sure everyone sees our modifications to the page tables
	 * rmb: make sure we see changes to vcpu->mode
	 */
	smp_mb();
X
Xiao Guangrong 已提交
2119

2120 2121 2122 2123 2124
	/*
	 * Wait for all vcpus to exit guest mode and/or lockless shadow
	 * page table walks.
	 */
	kvm_flush_remote_tlbs(kvm);
2125

2126 2127 2128
	do {
		sp = list_first_entry(invalid_list, struct kvm_mmu_page, link);
		WARN_ON(!sp->role.invalid || sp->root_count);
2129
		kvm_mmu_isolate_page(sp);
2130
		kvm_mmu_free_page(sp);
2131 2132 2133
	} while (!list_empty(invalid_list));
}

2134 2135
/*
 * Changing the number of mmu pages allocated to the vm
2136
 * Note: if goal_nr_mmu_pages is too small, you will get dead lock
2137
 */
2138
void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int goal_nr_mmu_pages)
2139
{
2140
	LIST_HEAD(invalid_list);
2141 2142 2143 2144 2145 2146
	/*
	 * If we set the number of mmu pages to be smaller be than the
	 * number of actived pages , we must to free some mmu pages before we
	 * change the value
	 */

2147 2148
	if (kvm->arch.n_used_mmu_pages > goal_nr_mmu_pages) {
		while (kvm->arch.n_used_mmu_pages > goal_nr_mmu_pages &&
2149
			!list_empty(&kvm->arch.active_mmu_pages)) {
2150 2151
			struct kvm_mmu_page *page;

2152
			page = container_of(kvm->arch.active_mmu_pages.prev,
2153
					    struct kvm_mmu_page, link);
2154
			kvm_mmu_prepare_zap_page(kvm, page, &invalid_list);
2155
		}
2156
		kvm_mmu_commit_zap_page(kvm, &invalid_list);
2157
		goal_nr_mmu_pages = kvm->arch.n_used_mmu_pages;
2158 2159
	}

2160
	kvm->arch.n_max_mmu_pages = goal_nr_mmu_pages;
2161 2162
}

2163
int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn)
2164
{
2165
	struct kvm_mmu_page *sp;
2166
	struct hlist_node *node;
2167
	LIST_HEAD(invalid_list);
2168 2169
	int r;

2170
	pgprintk("%s: looking for gfn %llx\n", __func__, gfn);
2171
	r = 0;
2172
	spin_lock(&kvm->mmu_lock);
2173
	for_each_gfn_indirect_valid_sp(kvm, sp, gfn, node) {
2174
		pgprintk("%s: gfn %llx role %x\n", __func__, gfn,
2175 2176
			 sp->role.word);
		r = 1;
2177
		kvm_mmu_prepare_zap_page(kvm, sp, &invalid_list);
2178
	}
2179
	kvm_mmu_commit_zap_page(kvm, &invalid_list);
2180 2181
	spin_unlock(&kvm->mmu_lock);

2182
	return r;
2183
}
2184
EXPORT_SYMBOL_GPL(kvm_mmu_unprotect_page);
2185

2186
static void page_header_update_slot(struct kvm *kvm, void *pte, gfn_t gfn)
A
Avi Kivity 已提交
2187
{
2188
	int slot = memslot_id(kvm, gfn);
2189
	struct kvm_mmu_page *sp = page_header(__pa(pte));
A
Avi Kivity 已提交
2190

2191
	__set_bit(slot, sp->slot_bitmap);
A
Avi Kivity 已提交
2192 2193
}

2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286
/*
 * The function is based on mtrr_type_lookup() in
 * arch/x86/kernel/cpu/mtrr/generic.c
 */
static int get_mtrr_type(struct mtrr_state_type *mtrr_state,
			 u64 start, u64 end)
{
	int i;
	u64 base, mask;
	u8 prev_match, curr_match;
	int num_var_ranges = KVM_NR_VAR_MTRR;

	if (!mtrr_state->enabled)
		return 0xFF;

	/* Make end inclusive end, instead of exclusive */
	end--;

	/* Look in fixed ranges. Just return the type as per start */
	if (mtrr_state->have_fixed && (start < 0x100000)) {
		int idx;

		if (start < 0x80000) {
			idx = 0;
			idx += (start >> 16);
			return mtrr_state->fixed_ranges[idx];
		} else if (start < 0xC0000) {
			idx = 1 * 8;
			idx += ((start - 0x80000) >> 14);
			return mtrr_state->fixed_ranges[idx];
		} else if (start < 0x1000000) {
			idx = 3 * 8;
			idx += ((start - 0xC0000) >> 12);
			return mtrr_state->fixed_ranges[idx];
		}
	}

	/*
	 * Look in variable ranges
	 * Look of multiple ranges matching this address and pick type
	 * as per MTRR precedence
	 */
	if (!(mtrr_state->enabled & 2))
		return mtrr_state->def_type;

	prev_match = 0xFF;
	for (i = 0; i < num_var_ranges; ++i) {
		unsigned short start_state, end_state;

		if (!(mtrr_state->var_ranges[i].mask_lo & (1 << 11)))
			continue;

		base = (((u64)mtrr_state->var_ranges[i].base_hi) << 32) +
		       (mtrr_state->var_ranges[i].base_lo & PAGE_MASK);
		mask = (((u64)mtrr_state->var_ranges[i].mask_hi) << 32) +
		       (mtrr_state->var_ranges[i].mask_lo & PAGE_MASK);

		start_state = ((start & mask) == (base & mask));
		end_state = ((end & mask) == (base & mask));
		if (start_state != end_state)
			return 0xFE;

		if ((start & mask) != (base & mask))
			continue;

		curr_match = mtrr_state->var_ranges[i].base_lo & 0xff;
		if (prev_match == 0xFF) {
			prev_match = curr_match;
			continue;
		}

		if (prev_match == MTRR_TYPE_UNCACHABLE ||
		    curr_match == MTRR_TYPE_UNCACHABLE)
			return MTRR_TYPE_UNCACHABLE;

		if ((prev_match == MTRR_TYPE_WRBACK &&
		     curr_match == MTRR_TYPE_WRTHROUGH) ||
		    (prev_match == MTRR_TYPE_WRTHROUGH &&
		     curr_match == MTRR_TYPE_WRBACK)) {
			prev_match = MTRR_TYPE_WRTHROUGH;
			curr_match = MTRR_TYPE_WRTHROUGH;
		}

		if (prev_match != curr_match)
			return MTRR_TYPE_UNCACHABLE;
	}

	if (prev_match != 0xFF)
		return prev_match;

	return mtrr_state->def_type;
}

2287
u8 kvm_get_guest_memory_type(struct kvm_vcpu *vcpu, gfn_t gfn)
2288 2289 2290 2291 2292 2293 2294 2295 2296
{
	u8 mtrr;

	mtrr = get_mtrr_type(&vcpu->arch.mtrr_state, gfn << PAGE_SHIFT,
			     (gfn << PAGE_SHIFT) + PAGE_SIZE);
	if (mtrr == 0xfe || mtrr == 0xff)
		mtrr = MTRR_TYPE_WRBACK;
	return mtrr;
}
2297
EXPORT_SYMBOL_GPL(kvm_get_guest_memory_type);
2298

2299 2300 2301 2302 2303 2304 2305 2306 2307 2308
static void __kvm_unsync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp)
{
	trace_kvm_mmu_unsync_page(sp);
	++vcpu->kvm->stat.mmu_unsync;
	sp->unsync = 1;

	kvm_mmu_mark_parents_unsync(sp);
}

static void kvm_unsync_pages(struct kvm_vcpu *vcpu,  gfn_t gfn)
2309 2310
{
	struct kvm_mmu_page *s;
2311
	struct hlist_node *node;
2312

2313
	for_each_gfn_indirect_valid_sp(vcpu->kvm, s, gfn, node) {
2314
		if (s->unsync)
2315
			continue;
2316 2317
		WARN_ON(s->role.level != PT_PAGE_TABLE_LEVEL);
		__kvm_unsync_page(vcpu, s);
2318 2319 2320 2321 2322 2323
	}
}

static int mmu_need_write_protect(struct kvm_vcpu *vcpu, gfn_t gfn,
				  bool can_unsync)
{
2324
	struct kvm_mmu_page *s;
2325
	struct hlist_node *node;
2326 2327
	bool need_unsync = false;

2328
	for_each_gfn_indirect_valid_sp(vcpu->kvm, s, gfn, node) {
2329 2330 2331
		if (!can_unsync)
			return 1;

2332
		if (s->role.level != PT_PAGE_TABLE_LEVEL)
2333
			return 1;
2334 2335 2336 2337

		if (!need_unsync && !s->unsync) {
			need_unsync = true;
		}
2338
	}
2339 2340
	if (need_unsync)
		kvm_unsync_pages(vcpu, gfn);
2341 2342 2343
	return 0;
}

A
Avi Kivity 已提交
2344
static int set_spte(struct kvm_vcpu *vcpu, u64 *sptep,
M
Marcelo Tosatti 已提交
2345
		    unsigned pte_access, int user_fault,
2346
		    int write_fault, int level,
2347
		    gfn_t gfn, pfn_t pfn, bool speculative,
2348
		    bool can_unsync, bool host_writable)
2349
{
2350
	u64 spte;
M
Marcelo Tosatti 已提交
2351
	int ret = 0;
S
Sheng Yang 已提交
2352

2353 2354 2355
	if (set_mmio_spte(sptep, gfn, pfn, pte_access))
		return 0;

2356
	spte = PT_PRESENT_MASK;
2357
	if (!speculative)
2358
		spte |= shadow_accessed_mask;
2359

S
Sheng Yang 已提交
2360 2361 2362 2363
	if (pte_access & ACC_EXEC_MASK)
		spte |= shadow_x_mask;
	else
		spte |= shadow_nx_mask;
2364

2365
	if (pte_access & ACC_USER_MASK)
S
Sheng Yang 已提交
2366
		spte |= shadow_user_mask;
2367

2368
	if (level > PT_PAGE_TABLE_LEVEL)
M
Marcelo Tosatti 已提交
2369
		spte |= PT_PAGE_SIZE_MASK;
2370
	if (tdp_enabled)
2371 2372
		spte |= kvm_x86_ops->get_mt_mask(vcpu, gfn,
			kvm_is_mmio_pfn(pfn));
2373

2374
	if (host_writable)
2375
		spte |= SPTE_HOST_WRITEABLE;
2376 2377
	else
		pte_access &= ~ACC_WRITE_MASK;
2378

2379
	spte |= (u64)pfn << PAGE_SHIFT;
2380 2381

	if ((pte_access & ACC_WRITE_MASK)
2382 2383
	    || (!vcpu->arch.mmu.direct_map && write_fault
		&& !is_write_protection(vcpu) && !user_fault)) {
2384

2385 2386
		if (level > PT_PAGE_TABLE_LEVEL &&
		    has_wrprotected_page(vcpu->kvm, gfn, level)) {
2387
			ret = 1;
2388
			drop_spte(vcpu->kvm, sptep);
A
Avi Kivity 已提交
2389
			goto done;
2390 2391
		}

2392
		spte |= PT_WRITABLE_MASK | SPTE_MMU_WRITEABLE;
2393

2394
		if (!vcpu->arch.mmu.direct_map
2395
		    && !(pte_access & ACC_WRITE_MASK)) {
2396
			spte &= ~PT_USER_MASK;
2397 2398 2399 2400 2401 2402 2403 2404 2405
			/*
			 * If we converted a user page to a kernel page,
			 * so that the kernel can write to it when cr0.wp=0,
			 * then we should prevent the kernel from executing it
			 * if SMEP is enabled.
			 */
			if (kvm_read_cr4_bits(vcpu, X86_CR4_SMEP))
				spte |= PT64_NX_MASK;
		}
2406

2407 2408 2409 2410 2411 2412
		/*
		 * 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.
		 */
2413
		if (!can_unsync && is_writable_pte(*sptep))
2414 2415
			goto set_pte;

2416
		if (mmu_need_write_protect(vcpu, gfn, can_unsync)) {
2417
			pgprintk("%s: found shadow page for %llx, marking ro\n",
2418
				 __func__, gfn);
M
Marcelo Tosatti 已提交
2419
			ret = 1;
2420
			pte_access &= ~ACC_WRITE_MASK;
2421
			spte &= ~(PT_WRITABLE_MASK | SPTE_MMU_WRITEABLE);
2422 2423 2424 2425 2426 2427
		}
	}

	if (pte_access & ACC_WRITE_MASK)
		mark_page_dirty(vcpu->kvm, gfn);

2428
set_pte:
2429
	if (mmu_spte_update(sptep, spte))
2430
		kvm_flush_remote_tlbs(vcpu->kvm);
A
Avi Kivity 已提交
2431
done:
M
Marcelo Tosatti 已提交
2432 2433 2434
	return ret;
}

A
Avi Kivity 已提交
2435
static void mmu_set_spte(struct kvm_vcpu *vcpu, u64 *sptep,
M
Marcelo Tosatti 已提交
2436
			 unsigned pt_access, unsigned pte_access,
2437
			 int user_fault, int write_fault,
2438
			 int *emulate, int level, gfn_t gfn,
2439
			 pfn_t pfn, bool speculative,
2440
			 bool host_writable)
M
Marcelo Tosatti 已提交
2441 2442
{
	int was_rmapped = 0;
2443
	int rmap_count;
M
Marcelo Tosatti 已提交
2444 2445

	pgprintk("%s: spte %llx access %x write_fault %d"
2446
		 " user_fault %d gfn %llx\n",
A
Avi Kivity 已提交
2447
		 __func__, *sptep, pt_access,
M
Marcelo Tosatti 已提交
2448 2449
		 write_fault, user_fault, gfn);

A
Avi Kivity 已提交
2450
	if (is_rmap_spte(*sptep)) {
M
Marcelo Tosatti 已提交
2451 2452 2453 2454
		/*
		 * If we overwrite a PTE page pointer with a 2MB PMD, unlink
		 * the parent of the now unreachable PTE.
		 */
2455 2456
		if (level > PT_PAGE_TABLE_LEVEL &&
		    !is_large_pte(*sptep)) {
M
Marcelo Tosatti 已提交
2457
			struct kvm_mmu_page *child;
A
Avi Kivity 已提交
2458
			u64 pte = *sptep;
M
Marcelo Tosatti 已提交
2459 2460

			child = page_header(pte & PT64_BASE_ADDR_MASK);
2461
			drop_parent_pte(child, sptep);
2462
			kvm_flush_remote_tlbs(vcpu->kvm);
A
Avi Kivity 已提交
2463
		} else if (pfn != spte_to_pfn(*sptep)) {
2464
			pgprintk("hfn old %llx new %llx\n",
A
Avi Kivity 已提交
2465
				 spte_to_pfn(*sptep), pfn);
2466
			drop_spte(vcpu->kvm, sptep);
2467
			kvm_flush_remote_tlbs(vcpu->kvm);
2468 2469
		} else
			was_rmapped = 1;
M
Marcelo Tosatti 已提交
2470
	}
2471

A
Avi Kivity 已提交
2472
	if (set_spte(vcpu, sptep, pte_access, user_fault, write_fault,
2473
		      level, gfn, pfn, speculative, true,
2474
		      host_writable)) {
M
Marcelo Tosatti 已提交
2475
		if (write_fault)
2476
			*emulate = 1;
2477
		kvm_mmu_flush_tlb(vcpu);
2478
	}
M
Marcelo Tosatti 已提交
2479

2480 2481 2482
	if (unlikely(is_mmio_spte(*sptep) && emulate))
		*emulate = 1;

A
Avi Kivity 已提交
2483
	pgprintk("%s: setting spte %llx\n", __func__, *sptep);
2484
	pgprintk("instantiating %s PTE (%s) at %llx (%llx) addr %p\n",
A
Avi Kivity 已提交
2485
		 is_large_pte(*sptep)? "2MB" : "4kB",
2486 2487
		 *sptep & PT_PRESENT_MASK ?"RW":"R", gfn,
		 *sptep, sptep);
A
Avi Kivity 已提交
2488
	if (!was_rmapped && is_large_pte(*sptep))
M
Marcelo Tosatti 已提交
2489 2490
		++vcpu->kvm->stat.lpages;

2491 2492 2493 2494 2495 2496 2497
	if (is_shadow_present_pte(*sptep)) {
		page_header_update_slot(vcpu->kvm, sptep, gfn);
		if (!was_rmapped) {
			rmap_count = rmap_add(vcpu, sptep, gfn);
			if (rmap_count > RMAP_RECYCLE_THRESHOLD)
				rmap_recycle(vcpu, sptep, gfn);
		}
2498
	}
2499 2500 2501

	if (!is_error_pfn(pfn))
		kvm_release_pfn_clean(pfn);
2502 2503
}

A
Avi Kivity 已提交
2504 2505
static void nonpaging_new_cr3(struct kvm_vcpu *vcpu)
{
2506
	mmu_free_roots(vcpu);
A
Avi Kivity 已提交
2507 2508
}

2509 2510 2511 2512 2513 2514 2515 2516
static bool is_rsvd_bits_set(struct kvm_mmu *mmu, u64 gpte, int level)
{
	int bit7;

	bit7 = (gpte >> 7) & 1;
	return (gpte & mmu->rsvd_bits_mask[bit7][level-1]) != 0;
}

2517 2518 2519 2520 2521
static pfn_t pte_prefetch_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn,
				     bool no_dirty_log)
{
	struct kvm_memory_slot *slot;

2522
	slot = gfn_to_memslot_dirty_bitmap(vcpu, gfn, no_dirty_log);
2523
	if (!slot)
2524
		return KVM_PFN_ERR_FAULT;
2525

2526
	return gfn_to_pfn_memslot_atomic(slot, gfn);
2527 2528
}

2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548
static bool prefetch_invalid_gpte(struct kvm_vcpu *vcpu,
				  struct kvm_mmu_page *sp, u64 *spte,
				  u64 gpte)
{
	if (is_rsvd_bits_set(&vcpu->arch.mmu, gpte, PT_PAGE_TABLE_LEVEL))
		goto no_present;

	if (!is_present_gpte(gpte))
		goto no_present;

	if (!(gpte & PT_ACCESSED_MASK))
		goto no_present;

	return false;

no_present:
	drop_spte(vcpu->kvm, spte);
	return true;
}

2549 2550 2551 2552 2553 2554 2555 2556 2557 2558
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];
	unsigned access = sp->role.access;
	int i, ret;
	gfn_t gfn;

	gfn = kvm_mmu_page_get_gfn(sp, start - sp->spt);
2559
	if (!gfn_to_memslot_dirty_bitmap(vcpu, gfn, access & ACC_WRITE_MASK))
2560 2561 2562 2563 2564 2565 2566 2567
		return -1;

	ret = gfn_to_page_many_atomic(vcpu->kvm, gfn, pages, end - start);
	if (ret <= 0)
		return -1;

	for (i = 0; i < ret; i++, gfn++, start++)
		mmu_set_spte(vcpu, start, ACC_ALL,
2568
			     access, 0, 0, NULL,
2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586
			     sp->role.level, gfn,
			     page_to_pfn(pages[i]), true, true);

	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++) {
2587
		if (is_shadow_present_pte(*spte) || spte == sptep) {
2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617
			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);
}

2618
static int __direct_map(struct kvm_vcpu *vcpu, gpa_t v, int write,
2619 2620
			int map_writable, int level, gfn_t gfn, pfn_t pfn,
			bool prefault)
2621
{
2622
	struct kvm_shadow_walk_iterator iterator;
2623
	struct kvm_mmu_page *sp;
2624
	int emulate = 0;
2625
	gfn_t pseudo_gfn;
A
Avi Kivity 已提交
2626

2627
	for_each_shadow_entry(vcpu, (u64)gfn << PAGE_SHIFT, iterator) {
2628
		if (iterator.level == level) {
2629 2630 2631
			unsigned pte_access = ACC_ALL;

			mmu_set_spte(vcpu, iterator.sptep, ACC_ALL, pte_access,
2632
				     0, write, &emulate,
2633
				     level, gfn, pfn, prefault, map_writable);
2634
			direct_pte_prefetch(vcpu, iterator.sptep);
2635 2636
			++vcpu->stat.pf_fixed;
			break;
A
Avi Kivity 已提交
2637 2638
		}

2639
		if (!is_shadow_present_pte(*iterator.sptep)) {
2640 2641 2642 2643
			u64 base_addr = iterator.addr;

			base_addr &= PT64_LVL_ADDR_MASK(iterator.level);
			pseudo_gfn = base_addr >> PAGE_SHIFT;
2644 2645 2646
			sp = kvm_mmu_get_page(vcpu, pseudo_gfn, iterator.addr,
					      iterator.level - 1,
					      1, ACC_ALL, iterator.sptep);
2647

2648 2649 2650 2651 2652
			mmu_spte_set(iterator.sptep,
				     __pa(sp->spt)
				     | PT_PRESENT_MASK | PT_WRITABLE_MASK
				     | shadow_user_mask | shadow_x_mask
				     | shadow_accessed_mask);
2653 2654
		}
	}
2655
	return emulate;
A
Avi Kivity 已提交
2656 2657
}

H
Huang Ying 已提交
2658
static void kvm_send_hwpoison_signal(unsigned long address, struct task_struct *tsk)
2659
{
H
Huang Ying 已提交
2660 2661 2662 2663 2664 2665 2666
	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;
2667

H
Huang Ying 已提交
2668
	send_sig_info(SIGBUS, &info, tsk);
2669 2670
}

2671
static int kvm_handle_bad_page(struct kvm_vcpu *vcpu, gfn_t gfn, pfn_t pfn)
2672
{
X
Xiao Guangrong 已提交
2673 2674 2675 2676 2677 2678 2679 2680 2681
	/*
	 * 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;

2682
	if (pfn == KVM_PFN_ERR_HWPOISON) {
2683
		kvm_send_hwpoison_signal(gfn_to_hva(vcpu->kvm, gfn), current);
2684
		return 0;
2685
	}
2686

2687
	return -EFAULT;
2688 2689
}

2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724
static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
					gfn_t *gfnp, pfn_t *pfnp, int *levelp)
{
	pfn_t pfn = *pfnp;
	gfn_t gfn = *gfnp;
	int level = *levelp;

	/*
	 * Check if it's a transparent hugepage. If this would be an
	 * hugetlbfs page, level wouldn't be set to
	 * PT_PAGE_TABLE_LEVEL and there would be no adjustment done
	 * here.
	 */
	if (!is_error_pfn(pfn) && !kvm_is_mmio_pfn(pfn) &&
	    level == PT_PAGE_TABLE_LEVEL &&
	    PageTransCompound(pfn_to_page(pfn)) &&
	    !has_wrprotected_page(vcpu->kvm, gfn, PT_DIRECTORY_LEVEL)) {
		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;
2725
			kvm_get_pfn(pfn);
2726 2727 2728 2729 2730
			*pfnp = pfn;
		}
	}
}

2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741
static bool handle_abnormal_pfn(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
				pfn_t pfn, unsigned access, int *ret_val)
{
	bool ret = true;

	/* The pfn is invalid, report the error! */
	if (unlikely(is_invalid_pfn(pfn))) {
		*ret_val = kvm_handle_bad_page(vcpu, gfn, pfn);
		goto exit;
	}

2742
	if (unlikely(is_noslot_pfn(pfn)))
2743 2744 2745 2746 2747 2748 2749
		vcpu_cache_mmio_info(vcpu, gva, gfn, access);

	ret = false;
exit:
	return ret;
}

2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840
static bool page_fault_can_be_fast(struct kvm_vcpu *vcpu, u32 error_code)
{
	/*
	 * #PF can be fast only if the shadow page table is present and it
	 * is caused by write-protect, that means we just need change the
	 * W bit of the spte which can be done out of mmu-lock.
	 */
	if (!(error_code & PFERR_PRESENT_MASK) ||
	      !(error_code & PFERR_WRITE_MASK))
		return false;

	return true;
}

static bool
fast_pf_fix_direct_spte(struct kvm_vcpu *vcpu, u64 *sptep, u64 spte)
{
	struct kvm_mmu_page *sp = page_header(__pa(sptep));
	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);

	if (cmpxchg64(sptep, spte, spte | PT_WRITABLE_MASK) == spte)
		mark_page_dirty(vcpu->kvm, gfn);

	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;
	bool ret = false;
	u64 spte = 0ull;

	if (!page_fault_can_be_fast(vcpu, error_code))
		return false;

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

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

	if (!is_last_spte(spte, level))
		goto exit;

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

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

	/*
	 * 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.
	 */
	ret = fast_pf_fix_direct_spte(vcpu, iterator.sptep, spte);
exit:
X
Xiao Guangrong 已提交
2841 2842
	trace_fast_page_fault(vcpu, gva, error_code, iterator.sptep,
			      spte, ret);
2843 2844 2845 2846 2847
	walk_shadow_page_lockless_end(vcpu);

	return ret;
}

2848
static bool try_async_pf(struct kvm_vcpu *vcpu, bool prefault, gfn_t gfn,
2849 2850
			 gva_t gva, pfn_t *pfn, bool write, bool *writable);

2851 2852
static int nonpaging_map(struct kvm_vcpu *vcpu, gva_t v, u32 error_code,
			 gfn_t gfn, bool prefault)
2853 2854
{
	int r;
2855
	int level;
2856
	int force_pt_level;
2857
	pfn_t pfn;
2858
	unsigned long mmu_seq;
2859
	bool map_writable, write = error_code & PFERR_WRITE_MASK;
2860

2861 2862 2863 2864 2865 2866 2867 2868 2869 2870
	force_pt_level = mapping_level_dirty_bitmap(vcpu, gfn);
	if (likely(!force_pt_level)) {
		level = mapping_level(vcpu, gfn);
		/*
		 * 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;
2871

2872 2873 2874
		gfn &= ~(KVM_PAGES_PER_HPAGE(level) - 1);
	} else
		level = PT_PAGE_TABLE_LEVEL;
M
Marcelo Tosatti 已提交
2875

2876 2877 2878
	if (fast_page_fault(vcpu, v, level, error_code))
		return 0;

2879
	mmu_seq = vcpu->kvm->mmu_notifier_seq;
2880
	smp_rmb();
2881

2882
	if (try_async_pf(vcpu, prefault, gfn, v, &pfn, write, &map_writable))
2883
		return 0;
2884

2885 2886
	if (handle_abnormal_pfn(vcpu, v, gfn, pfn, ACC_ALL, &r))
		return r;
2887

2888
	spin_lock(&vcpu->kvm->mmu_lock);
2889 2890
	if (mmu_notifier_retry(vcpu, mmu_seq))
		goto out_unlock;
2891
	kvm_mmu_free_some_pages(vcpu);
2892 2893
	if (likely(!force_pt_level))
		transparent_hugepage_adjust(vcpu, &gfn, &pfn, &level);
2894 2895
	r = __direct_map(vcpu, v, write, map_writable, level, gfn, pfn,
			 prefault);
2896 2897 2898
	spin_unlock(&vcpu->kvm->mmu_lock);


2899
	return r;
2900 2901 2902 2903 2904

out_unlock:
	spin_unlock(&vcpu->kvm->mmu_lock);
	kvm_release_pfn_clean(pfn);
	return 0;
2905 2906 2907
}


2908 2909 2910
static void mmu_free_roots(struct kvm_vcpu *vcpu)
{
	int i;
2911
	struct kvm_mmu_page *sp;
2912
	LIST_HEAD(invalid_list);
2913

2914
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
A
Avi Kivity 已提交
2915
		return;
2916
	spin_lock(&vcpu->kvm->mmu_lock);
2917 2918 2919
	if (vcpu->arch.mmu.shadow_root_level == PT64_ROOT_LEVEL &&
	    (vcpu->arch.mmu.root_level == PT64_ROOT_LEVEL ||
	     vcpu->arch.mmu.direct_map)) {
2920
		hpa_t root = vcpu->arch.mmu.root_hpa;
2921

2922 2923
		sp = page_header(root);
		--sp->root_count;
2924 2925 2926 2927
		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);
		}
2928
		vcpu->arch.mmu.root_hpa = INVALID_PAGE;
2929
		spin_unlock(&vcpu->kvm->mmu_lock);
2930 2931 2932
		return;
	}
	for (i = 0; i < 4; ++i) {
2933
		hpa_t root = vcpu->arch.mmu.pae_root[i];
2934

A
Avi Kivity 已提交
2935 2936
		if (root) {
			root &= PT64_BASE_ADDR_MASK;
2937 2938
			sp = page_header(root);
			--sp->root_count;
2939
			if (!sp->root_count && sp->role.invalid)
2940 2941
				kvm_mmu_prepare_zap_page(vcpu->kvm, sp,
							 &invalid_list);
A
Avi Kivity 已提交
2942
		}
2943
		vcpu->arch.mmu.pae_root[i] = INVALID_PAGE;
2944
	}
2945
	kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
2946
	spin_unlock(&vcpu->kvm->mmu_lock);
2947
	vcpu->arch.mmu.root_hpa = INVALID_PAGE;
2948 2949
}

2950 2951 2952 2953 2954
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)) {
2955
		kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu);
2956 2957 2958 2959 2960 2961
		ret = 1;
	}

	return ret;
}

2962 2963 2964
static int mmu_alloc_direct_roots(struct kvm_vcpu *vcpu)
{
	struct kvm_mmu_page *sp;
2965
	unsigned i;
2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981

	if (vcpu->arch.mmu.shadow_root_level == PT64_ROOT_LEVEL) {
		spin_lock(&vcpu->kvm->mmu_lock);
		kvm_mmu_free_some_pages(vcpu);
		sp = kvm_mmu_get_page(vcpu, 0, 0, PT64_ROOT_LEVEL,
				      1, ACC_ALL, NULL);
		++sp->root_count;
		spin_unlock(&vcpu->kvm->mmu_lock);
		vcpu->arch.mmu.root_hpa = __pa(sp->spt);
	} else if (vcpu->arch.mmu.shadow_root_level == PT32E_ROOT_LEVEL) {
		for (i = 0; i < 4; ++i) {
			hpa_t root = vcpu->arch.mmu.pae_root[i];

			ASSERT(!VALID_PAGE(root));
			spin_lock(&vcpu->kvm->mmu_lock);
			kvm_mmu_free_some_pages(vcpu);
2982 2983
			sp = kvm_mmu_get_page(vcpu, i << (30 - PAGE_SHIFT),
					      i << 30,
2984 2985 2986 2987 2988 2989 2990
					      PT32_ROOT_LEVEL, 1, ACC_ALL,
					      NULL);
			root = __pa(sp->spt);
			++sp->root_count;
			spin_unlock(&vcpu->kvm->mmu_lock);
			vcpu->arch.mmu.pae_root[i] = root | PT_PRESENT_MASK;
		}
2991
		vcpu->arch.mmu.root_hpa = __pa(vcpu->arch.mmu.pae_root);
2992 2993 2994 2995 2996 2997 2998
	} else
		BUG();

	return 0;
}

static int mmu_alloc_shadow_roots(struct kvm_vcpu *vcpu)
2999
{
3000
	struct kvm_mmu_page *sp;
3001 3002 3003
	u64 pdptr, pm_mask;
	gfn_t root_gfn;
	int i;
3004

3005
	root_gfn = vcpu->arch.mmu.get_cr3(vcpu) >> PAGE_SHIFT;
3006

3007 3008 3009 3010 3011 3012 3013 3014
	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) {
3015
		hpa_t root = vcpu->arch.mmu.root_hpa;
3016 3017

		ASSERT(!VALID_PAGE(root));
3018

3019
		spin_lock(&vcpu->kvm->mmu_lock);
3020
		kvm_mmu_free_some_pages(vcpu);
3021 3022
		sp = kvm_mmu_get_page(vcpu, root_gfn, 0, PT64_ROOT_LEVEL,
				      0, ACC_ALL, NULL);
3023 3024
		root = __pa(sp->spt);
		++sp->root_count;
3025
		spin_unlock(&vcpu->kvm->mmu_lock);
3026
		vcpu->arch.mmu.root_hpa = root;
3027
		return 0;
3028
	}
3029

3030 3031
	/*
	 * We shadow a 32 bit page table. This may be a legacy 2-level
3032 3033
	 * 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.
3034
	 */
3035 3036 3037 3038
	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;

3039
	for (i = 0; i < 4; ++i) {
3040
		hpa_t root = vcpu->arch.mmu.pae_root[i];
3041 3042

		ASSERT(!VALID_PAGE(root));
3043
		if (vcpu->arch.mmu.root_level == PT32E_ROOT_LEVEL) {
3044
			pdptr = vcpu->arch.mmu.get_pdptr(vcpu, i);
3045
			if (!is_present_gpte(pdptr)) {
3046
				vcpu->arch.mmu.pae_root[i] = 0;
A
Avi Kivity 已提交
3047 3048
				continue;
			}
A
Avi Kivity 已提交
3049
			root_gfn = pdptr >> PAGE_SHIFT;
3050 3051
			if (mmu_check_root(vcpu, root_gfn))
				return 1;
3052
		}
3053
		spin_lock(&vcpu->kvm->mmu_lock);
3054
		kvm_mmu_free_some_pages(vcpu);
3055
		sp = kvm_mmu_get_page(vcpu, root_gfn, i << 30,
3056
				      PT32_ROOT_LEVEL, 0,
3057
				      ACC_ALL, NULL);
3058 3059
		root = __pa(sp->spt);
		++sp->root_count;
3060 3061
		spin_unlock(&vcpu->kvm->mmu_lock);

3062
		vcpu->arch.mmu.pae_root[i] = root | pm_mask;
3063
	}
3064
	vcpu->arch.mmu.root_hpa = __pa(vcpu->arch.mmu.pae_root);
3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090

	/*
	 * 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);
	}

3091
	return 0;
3092 3093
}

3094 3095 3096 3097 3098 3099 3100 3101
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);
}

3102 3103 3104 3105 3106
static void mmu_sync_roots(struct kvm_vcpu *vcpu)
{
	int i;
	struct kvm_mmu_page *sp;

3107 3108 3109
	if (vcpu->arch.mmu.direct_map)
		return;

3110 3111
	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
		return;
3112

3113
	vcpu_clear_mmio_info(vcpu, ~0ul);
3114
	kvm_mmu_audit(vcpu, AUDIT_PRE_SYNC);
3115
	if (vcpu->arch.mmu.root_level == PT64_ROOT_LEVEL) {
3116 3117 3118
		hpa_t root = vcpu->arch.mmu.root_hpa;
		sp = page_header(root);
		mmu_sync_children(vcpu, sp);
3119
		kvm_mmu_audit(vcpu, AUDIT_POST_SYNC);
3120 3121 3122 3123 3124
		return;
	}
	for (i = 0; i < 4; ++i) {
		hpa_t root = vcpu->arch.mmu.pae_root[i];

3125
		if (root && VALID_PAGE(root)) {
3126 3127 3128 3129 3130
			root &= PT64_BASE_ADDR_MASK;
			sp = page_header(root);
			mmu_sync_children(vcpu, sp);
		}
	}
3131
	kvm_mmu_audit(vcpu, AUDIT_POST_SYNC);
3132 3133 3134 3135 3136 3137
}

void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu)
{
	spin_lock(&vcpu->kvm->mmu_lock);
	mmu_sync_roots(vcpu);
3138
	spin_unlock(&vcpu->kvm->mmu_lock);
3139 3140
}

3141
static gpa_t nonpaging_gva_to_gpa(struct kvm_vcpu *vcpu, gva_t vaddr,
3142
				  u32 access, struct x86_exception *exception)
A
Avi Kivity 已提交
3143
{
3144 3145
	if (exception)
		exception->error_code = 0;
A
Avi Kivity 已提交
3146 3147 3148
	return vaddr;
}

3149
static gpa_t nonpaging_gva_to_gpa_nested(struct kvm_vcpu *vcpu, gva_t vaddr,
3150 3151
					 u32 access,
					 struct x86_exception *exception)
3152
{
3153 3154
	if (exception)
		exception->error_code = 0;
3155 3156 3157
	return vcpu->arch.nested_mmu.translate_gpa(vcpu, vaddr, access);
}

3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214
static bool quickly_check_mmio_pf(struct kvm_vcpu *vcpu, u64 addr, bool direct)
{
	if (direct)
		return vcpu_match_mmio_gpa(vcpu, addr);

	return vcpu_match_mmio_gva(vcpu, addr);
}


/*
 * On direct hosts, the last spte is only allows two states
 * for mmio page fault:
 *   - It is the mmio spte
 *   - It is zapped or it is being zapped.
 *
 * This function completely checks the spte when the last spte
 * is not the mmio spte.
 */
static bool check_direct_spte_mmio_pf(u64 spte)
{
	return __check_direct_spte_mmio_pf(spte);
}

static u64 walk_shadow_page_get_mmio_spte(struct kvm_vcpu *vcpu, u64 addr)
{
	struct kvm_shadow_walk_iterator iterator;
	u64 spte = 0ull;

	walk_shadow_page_lockless_begin(vcpu);
	for_each_shadow_entry_lockless(vcpu, addr, iterator, spte)
		if (!is_shadow_present_pte(spte))
			break;
	walk_shadow_page_lockless_end(vcpu);

	return spte;
}

/*
 * If it is a real mmio page fault, return 1 and emulat the instruction
 * directly, return 0 to let CPU fault again on the address, -1 is
 * returned if bug is detected.
 */
int handle_mmio_page_fault_common(struct kvm_vcpu *vcpu, u64 addr, bool direct)
{
	u64 spte;

	if (quickly_check_mmio_pf(vcpu, addr, direct))
		return 1;

	spte = walk_shadow_page_get_mmio_spte(vcpu, addr);

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

		if (direct)
			addr = 0;
X
Xiao Guangrong 已提交
3215 3216

		trace_handle_mmio_page_fault(addr, gfn, access);
3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245
		vcpu_cache_mmio_info(vcpu, addr, gfn, access);
		return 1;
	}

	/*
	 * It's ok if the gva is remapped by other cpus on shadow guest,
	 * it's a BUG if the gfn is not a mmio page.
	 */
	if (direct && !check_direct_spte_mmio_pf(spte))
		return -1;

	/*
	 * If the page table is zapped by other cpus, let CPU fault again on
	 * the address.
	 */
	return 0;
}
EXPORT_SYMBOL_GPL(handle_mmio_page_fault_common);

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

	ret = handle_mmio_page_fault_common(vcpu, addr, direct);
	WARN_ON(ret < 0);
	return ret;
}

A
Avi Kivity 已提交
3246
static int nonpaging_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
3247
				u32 error_code, bool prefault)
A
Avi Kivity 已提交
3248
{
3249
	gfn_t gfn;
3250
	int r;
A
Avi Kivity 已提交
3251

3252
	pgprintk("%s: gva %lx error %x\n", __func__, gva, error_code);
3253 3254 3255 3256

	if (unlikely(error_code & PFERR_RSVD_MASK))
		return handle_mmio_page_fault(vcpu, gva, error_code, true);

3257 3258 3259
	r = mmu_topup_memory_caches(vcpu);
	if (r)
		return r;
3260

A
Avi Kivity 已提交
3261
	ASSERT(vcpu);
3262
	ASSERT(VALID_PAGE(vcpu->arch.mmu.root_hpa));
A
Avi Kivity 已提交
3263

3264
	gfn = gva >> PAGE_SHIFT;
A
Avi Kivity 已提交
3265

3266
	return nonpaging_map(vcpu, gva & PAGE_MASK,
3267
			     error_code, gfn, prefault);
A
Avi Kivity 已提交
3268 3269
}

3270
static int kvm_arch_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn)
3271 3272
{
	struct kvm_arch_async_pf arch;
X
Xiao Guangrong 已提交
3273

3274
	arch.token = (vcpu->arch.apf.id++ << 12) | vcpu->vcpu_id;
3275
	arch.gfn = gfn;
3276
	arch.direct_map = vcpu->arch.mmu.direct_map;
X
Xiao Guangrong 已提交
3277
	arch.cr3 = vcpu->arch.mmu.get_cr3(vcpu);
3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290

	return kvm_setup_async_pf(vcpu, gva, gfn, &arch);
}

static bool can_do_async_pf(struct kvm_vcpu *vcpu)
{
	if (unlikely(!irqchip_in_kernel(vcpu->kvm) ||
		     kvm_event_needs_reinjection(vcpu)))
		return false;

	return kvm_x86_ops->interrupt_allowed(vcpu);
}

3291
static bool try_async_pf(struct kvm_vcpu *vcpu, bool prefault, gfn_t gfn,
3292
			 gva_t gva, pfn_t *pfn, bool write, bool *writable)
3293 3294 3295
{
	bool async;

3296
	*pfn = gfn_to_pfn_async(vcpu->kvm, gfn, &async, write, writable);
3297 3298 3299 3300

	if (!async)
		return false; /* *pfn has correct page already */

3301
	if (!prefault && can_do_async_pf(vcpu)) {
3302
		trace_kvm_try_async_get_page(gva, gfn);
3303 3304 3305 3306 3307 3308 3309 3310
		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;
	}

3311
	*pfn = gfn_to_pfn_prot(vcpu->kvm, gfn, write, writable);
3312 3313 3314 3315

	return false;
}

G
Gleb Natapov 已提交
3316
static int tdp_page_fault(struct kvm_vcpu *vcpu, gva_t gpa, u32 error_code,
3317
			  bool prefault)
3318
{
3319
	pfn_t pfn;
3320
	int r;
3321
	int level;
3322
	int force_pt_level;
M
Marcelo Tosatti 已提交
3323
	gfn_t gfn = gpa >> PAGE_SHIFT;
3324
	unsigned long mmu_seq;
3325 3326
	int write = error_code & PFERR_WRITE_MASK;
	bool map_writable;
3327 3328 3329 3330

	ASSERT(vcpu);
	ASSERT(VALID_PAGE(vcpu->arch.mmu.root_hpa));

3331 3332 3333
	if (unlikely(error_code & PFERR_RSVD_MASK))
		return handle_mmio_page_fault(vcpu, gpa, error_code, true);

3334 3335 3336 3337
	r = mmu_topup_memory_caches(vcpu);
	if (r)
		return r;

3338 3339 3340 3341 3342 3343
	force_pt_level = mapping_level_dirty_bitmap(vcpu, gfn);
	if (likely(!force_pt_level)) {
		level = mapping_level(vcpu, gfn);
		gfn &= ~(KVM_PAGES_PER_HPAGE(level) - 1);
	} else
		level = PT_PAGE_TABLE_LEVEL;
3344

3345 3346 3347
	if (fast_page_fault(vcpu, gpa, level, error_code))
		return 0;

3348
	mmu_seq = vcpu->kvm->mmu_notifier_seq;
3349
	smp_rmb();
3350

3351
	if (try_async_pf(vcpu, prefault, gfn, gpa, &pfn, write, &map_writable))
3352 3353
		return 0;

3354 3355 3356
	if (handle_abnormal_pfn(vcpu, 0, gfn, pfn, ACC_ALL, &r))
		return r;

3357
	spin_lock(&vcpu->kvm->mmu_lock);
3358 3359
	if (mmu_notifier_retry(vcpu, mmu_seq))
		goto out_unlock;
3360
	kvm_mmu_free_some_pages(vcpu);
3361 3362
	if (likely(!force_pt_level))
		transparent_hugepage_adjust(vcpu, &gfn, &pfn, &level);
3363
	r = __direct_map(vcpu, gpa, write, map_writable,
3364
			 level, gfn, pfn, prefault);
3365 3366 3367
	spin_unlock(&vcpu->kvm->mmu_lock);

	return r;
3368 3369 3370 3371 3372

out_unlock:
	spin_unlock(&vcpu->kvm->mmu_lock);
	kvm_release_pfn_clean(pfn);
	return 0;
3373 3374
}

A
Avi Kivity 已提交
3375 3376
static void nonpaging_free(struct kvm_vcpu *vcpu)
{
3377
	mmu_free_roots(vcpu);
A
Avi Kivity 已提交
3378 3379
}

3380 3381
static int nonpaging_init_context(struct kvm_vcpu *vcpu,
				  struct kvm_mmu *context)
A
Avi Kivity 已提交
3382 3383 3384 3385 3386
{
	context->new_cr3 = nonpaging_new_cr3;
	context->page_fault = nonpaging_page_fault;
	context->gva_to_gpa = nonpaging_gva_to_gpa;
	context->free = nonpaging_free;
3387
	context->sync_page = nonpaging_sync_page;
M
Marcelo Tosatti 已提交
3388
	context->invlpg = nonpaging_invlpg;
3389
	context->update_pte = nonpaging_update_pte;
3390
	context->root_level = 0;
A
Avi Kivity 已提交
3391
	context->shadow_root_level = PT32E_ROOT_LEVEL;
A
Avi Kivity 已提交
3392
	context->root_hpa = INVALID_PAGE;
3393
	context->direct_map = true;
3394
	context->nx = false;
A
Avi Kivity 已提交
3395 3396 3397
	return 0;
}

3398
void kvm_mmu_flush_tlb(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
3399
{
A
Avi Kivity 已提交
3400
	++vcpu->stat.tlb_flush;
3401
	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
A
Avi Kivity 已提交
3402 3403 3404 3405
}

static void paging_new_cr3(struct kvm_vcpu *vcpu)
{
3406
	pgprintk("%s: cr3 %lx\n", __func__, kvm_read_cr3(vcpu));
3407
	mmu_free_roots(vcpu);
A
Avi Kivity 已提交
3408 3409
}

3410 3411
static unsigned long get_cr3(struct kvm_vcpu *vcpu)
{
3412
	return kvm_read_cr3(vcpu);
3413 3414
}

3415 3416
static void inject_page_fault(struct kvm_vcpu *vcpu,
			      struct x86_exception *fault)
A
Avi Kivity 已提交
3417
{
3418
	vcpu->arch.mmu.inject_page_fault(vcpu, fault);
A
Avi Kivity 已提交
3419 3420 3421 3422 3423 3424 3425
}

static void paging_free(struct kvm_vcpu *vcpu)
{
	nonpaging_free(vcpu);
}

3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437
static inline void protect_clean_gpte(unsigned *access, unsigned gpte)
{
	unsigned mask;

	BUILD_BUG_ON(PT_WRITABLE_MASK != ACC_WRITE_MASK);

	mask = (unsigned)~ACC_WRITE_MASK;
	/* Allow write access to dirty gptes */
	mask |= (gpte >> (PT_DIRTY_SHIFT - PT_WRITABLE_SHIFT)) & PT_WRITABLE_MASK;
	*access &= mask;
}

3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454
static bool sync_mmio_spte(u64 *sptep, gfn_t gfn, unsigned access,
			   int *nr_present)
{
	if (unlikely(is_mmio_spte(*sptep))) {
		if (gfn != get_mmio_spte_gfn(*sptep)) {
			mmu_spte_clear_no_track(sptep);
			return true;
		}

		(*nr_present)++;
		mark_mmio_spte(sptep, gfn, access);
		return true;
	}

	return false;
}

3455 3456 3457 3458 3459 3460 3461 3462 3463 3464
static inline unsigned gpte_access(struct kvm_vcpu *vcpu, u64 gpte)
{
	unsigned access;

	access = (gpte & (PT_WRITABLE_MASK | PT_USER_MASK)) | ACC_EXEC_MASK;
	access &= ~(gpte >> PT64_NX_SHIFT);

	return access;
}

A
Avi Kivity 已提交
3465 3466 3467 3468 3469 3470 3471 3472 3473
static inline bool is_last_gpte(struct kvm_mmu *mmu, unsigned level, unsigned gpte)
{
	unsigned index;

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

A
Avi Kivity 已提交
3474 3475 3476 3477 3478 3479 3480 3481
#define PTTYPE 64
#include "paging_tmpl.h"
#undef PTTYPE

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

3482
static void reset_rsvds_bits_mask(struct kvm_vcpu *vcpu,
3483
				  struct kvm_mmu *context)
3484 3485 3486 3487
{
	int maxphyaddr = cpuid_maxphyaddr(vcpu);
	u64 exb_bit_rsvd = 0;

3488
	if (!context->nx)
3489
		exb_bit_rsvd = rsvd_bits(63, 63);
3490
	switch (context->root_level) {
3491 3492 3493 3494
	case PT32_ROOT_LEVEL:
		/* no rsvd bits for 2 level 4K page table entries */
		context->rsvd_bits_mask[0][1] = 0;
		context->rsvd_bits_mask[0][0] = 0;
3495 3496 3497 3498 3499 3500 3501
		context->rsvd_bits_mask[1][0] = context->rsvd_bits_mask[0][0];

		if (!is_pse(vcpu)) {
			context->rsvd_bits_mask[1][1] = 0;
			break;
		}

3502 3503 3504 3505 3506 3507 3508 3509
		if (is_cpuid_PSE36())
			/* 36bits PSE 4MB page */
			context->rsvd_bits_mask[1][1] = rsvd_bits(17, 21);
		else
			/* 32 bits PSE 4MB page */
			context->rsvd_bits_mask[1][1] = rsvd_bits(13, 21);
		break;
	case PT32E_ROOT_LEVEL:
3510 3511 3512
		context->rsvd_bits_mask[0][2] =
			rsvd_bits(maxphyaddr, 63) |
			rsvd_bits(7, 8) | rsvd_bits(1, 2);	/* PDPTE */
3513
		context->rsvd_bits_mask[0][1] = exb_bit_rsvd |
3514
			rsvd_bits(maxphyaddr, 62);	/* PDE */
3515 3516 3517 3518 3519
		context->rsvd_bits_mask[0][0] = exb_bit_rsvd |
			rsvd_bits(maxphyaddr, 62); 	/* PTE */
		context->rsvd_bits_mask[1][1] = exb_bit_rsvd |
			rsvd_bits(maxphyaddr, 62) |
			rsvd_bits(13, 20);		/* large page */
3520
		context->rsvd_bits_mask[1][0] = context->rsvd_bits_mask[0][0];
3521 3522 3523 3524 3525 3526 3527
		break;
	case PT64_ROOT_LEVEL:
		context->rsvd_bits_mask[0][3] = exb_bit_rsvd |
			rsvd_bits(maxphyaddr, 51) | rsvd_bits(7, 8);
		context->rsvd_bits_mask[0][2] = exb_bit_rsvd |
			rsvd_bits(maxphyaddr, 51) | rsvd_bits(7, 8);
		context->rsvd_bits_mask[0][1] = exb_bit_rsvd |
3528
			rsvd_bits(maxphyaddr, 51);
3529 3530 3531
		context->rsvd_bits_mask[0][0] = exb_bit_rsvd |
			rsvd_bits(maxphyaddr, 51);
		context->rsvd_bits_mask[1][3] = context->rsvd_bits_mask[0][3];
3532 3533 3534
		context->rsvd_bits_mask[1][2] = exb_bit_rsvd |
			rsvd_bits(maxphyaddr, 51) |
			rsvd_bits(13, 29);
3535
		context->rsvd_bits_mask[1][1] = exb_bit_rsvd |
3536 3537
			rsvd_bits(maxphyaddr, 51) |
			rsvd_bits(13, 20);		/* large page */
3538
		context->rsvd_bits_mask[1][0] = context->rsvd_bits_mask[0][0];
3539 3540 3541 3542
		break;
	}
}

3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574
static void update_permission_bitmask(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu)
{
	unsigned bit, byte, pfec;
	u8 map;
	bool fault, x, w, u, wf, uf, ff, smep;

	smep = kvm_read_cr4_bits(vcpu, X86_CR4_SMEP);
	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;
		for (bit = 0; bit < 8; ++bit) {
			x = bit & ACC_EXEC_MASK;
			w = bit & ACC_WRITE_MASK;
			u = bit & ACC_USER_MASK;

			/* 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 */
			x &= !(smep && u && !uf);

			fault = (ff && !x) || (uf && !u) || (wf && !w);
			map |= fault << bit;
		}
		mmu->permissions[byte] = map;
	}
}

A
Avi Kivity 已提交
3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592
static void update_last_pte_bitmap(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu)
{
	u8 map;
	unsigned level, root_level = mmu->root_level;
	const unsigned ps_set_index = 1 << 2;  /* bit 2 of index: ps */

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

3593 3594 3595
static int paging64_init_context_common(struct kvm_vcpu *vcpu,
					struct kvm_mmu *context,
					int level)
A
Avi Kivity 已提交
3596
{
3597
	context->nx = is_nx(vcpu);
3598
	context->root_level = level;
3599

3600
	reset_rsvds_bits_mask(vcpu, context);
3601
	update_permission_bitmask(vcpu, context);
A
Avi Kivity 已提交
3602
	update_last_pte_bitmap(vcpu, context);
A
Avi Kivity 已提交
3603 3604 3605 3606 3607

	ASSERT(is_pae(vcpu));
	context->new_cr3 = paging_new_cr3;
	context->page_fault = paging64_page_fault;
	context->gva_to_gpa = paging64_gva_to_gpa;
3608
	context->sync_page = paging64_sync_page;
M
Marcelo Tosatti 已提交
3609
	context->invlpg = paging64_invlpg;
3610
	context->update_pte = paging64_update_pte;
A
Avi Kivity 已提交
3611
	context->free = paging_free;
3612
	context->shadow_root_level = level;
A
Avi Kivity 已提交
3613
	context->root_hpa = INVALID_PAGE;
3614
	context->direct_map = false;
A
Avi Kivity 已提交
3615 3616 3617
	return 0;
}

3618 3619
static int paging64_init_context(struct kvm_vcpu *vcpu,
				 struct kvm_mmu *context)
3620
{
3621
	return paging64_init_context_common(vcpu, context, PT64_ROOT_LEVEL);
3622 3623
}

3624 3625
static int paging32_init_context(struct kvm_vcpu *vcpu,
				 struct kvm_mmu *context)
A
Avi Kivity 已提交
3626
{
3627
	context->nx = false;
3628
	context->root_level = PT32_ROOT_LEVEL;
3629

3630
	reset_rsvds_bits_mask(vcpu, context);
3631
	update_permission_bitmask(vcpu, context);
A
Avi Kivity 已提交
3632
	update_last_pte_bitmap(vcpu, context);
A
Avi Kivity 已提交
3633 3634 3635 3636 3637

	context->new_cr3 = paging_new_cr3;
	context->page_fault = paging32_page_fault;
	context->gva_to_gpa = paging32_gva_to_gpa;
	context->free = paging_free;
3638
	context->sync_page = paging32_sync_page;
M
Marcelo Tosatti 已提交
3639
	context->invlpg = paging32_invlpg;
3640
	context->update_pte = paging32_update_pte;
A
Avi Kivity 已提交
3641
	context->shadow_root_level = PT32E_ROOT_LEVEL;
A
Avi Kivity 已提交
3642
	context->root_hpa = INVALID_PAGE;
3643
	context->direct_map = false;
A
Avi Kivity 已提交
3644 3645 3646
	return 0;
}

3647 3648
static int paging32E_init_context(struct kvm_vcpu *vcpu,
				  struct kvm_mmu *context)
A
Avi Kivity 已提交
3649
{
3650
	return paging64_init_context_common(vcpu, context, PT32E_ROOT_LEVEL);
A
Avi Kivity 已提交
3651 3652
}

3653 3654
static int init_kvm_tdp_mmu(struct kvm_vcpu *vcpu)
{
3655
	struct kvm_mmu *context = vcpu->arch.walk_mmu;
3656

3657
	context->base_role.word = 0;
3658 3659 3660
	context->new_cr3 = nonpaging_new_cr3;
	context->page_fault = tdp_page_fault;
	context->free = nonpaging_free;
3661
	context->sync_page = nonpaging_sync_page;
M
Marcelo Tosatti 已提交
3662
	context->invlpg = nonpaging_invlpg;
3663
	context->update_pte = nonpaging_update_pte;
3664
	context->shadow_root_level = kvm_x86_ops->get_tdp_level();
3665
	context->root_hpa = INVALID_PAGE;
3666
	context->direct_map = true;
3667
	context->set_cr3 = kvm_x86_ops->set_tdp_cr3;
3668
	context->get_cr3 = get_cr3;
3669
	context->get_pdptr = kvm_pdptr_read;
3670
	context->inject_page_fault = kvm_inject_page_fault;
3671 3672

	if (!is_paging(vcpu)) {
3673
		context->nx = false;
3674 3675 3676
		context->gva_to_gpa = nonpaging_gva_to_gpa;
		context->root_level = 0;
	} else if (is_long_mode(vcpu)) {
3677
		context->nx = is_nx(vcpu);
3678
		context->root_level = PT64_ROOT_LEVEL;
3679 3680
		reset_rsvds_bits_mask(vcpu, context);
		context->gva_to_gpa = paging64_gva_to_gpa;
3681
	} else if (is_pae(vcpu)) {
3682
		context->nx = is_nx(vcpu);
3683
		context->root_level = PT32E_ROOT_LEVEL;
3684 3685
		reset_rsvds_bits_mask(vcpu, context);
		context->gva_to_gpa = paging64_gva_to_gpa;
3686
	} else {
3687
		context->nx = false;
3688
		context->root_level = PT32_ROOT_LEVEL;
3689 3690
		reset_rsvds_bits_mask(vcpu, context);
		context->gva_to_gpa = paging32_gva_to_gpa;
3691 3692
	}

3693
	update_permission_bitmask(vcpu, context);
A
Avi Kivity 已提交
3694
	update_last_pte_bitmap(vcpu, context);
3695

3696 3697 3698
	return 0;
}

3699
int kvm_init_shadow_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *context)
A
Avi Kivity 已提交
3700
{
3701
	int r;
3702
	bool smep = kvm_read_cr4_bits(vcpu, X86_CR4_SMEP);
A
Avi Kivity 已提交
3703
	ASSERT(vcpu);
3704
	ASSERT(!VALID_PAGE(vcpu->arch.mmu.root_hpa));
A
Avi Kivity 已提交
3705 3706

	if (!is_paging(vcpu))
3707
		r = nonpaging_init_context(vcpu, context);
A
Avi Kivity 已提交
3708
	else if (is_long_mode(vcpu))
3709
		r = paging64_init_context(vcpu, context);
A
Avi Kivity 已提交
3710
	else if (is_pae(vcpu))
3711
		r = paging32E_init_context(vcpu, context);
A
Avi Kivity 已提交
3712
	else
3713
		r = paging32_init_context(vcpu, context);
3714

3715
	vcpu->arch.mmu.base_role.cr4_pae = !!is_pae(vcpu);
3716
	vcpu->arch.mmu.base_role.cr0_wp  = is_write_protection(vcpu);
3717 3718
	vcpu->arch.mmu.base_role.smep_andnot_wp
		= smep && !is_write_protection(vcpu);
3719 3720 3721 3722 3723 3724 3725

	return r;
}
EXPORT_SYMBOL_GPL(kvm_init_shadow_mmu);

static int init_kvm_softmmu(struct kvm_vcpu *vcpu)
{
3726
	int r = kvm_init_shadow_mmu(vcpu, vcpu->arch.walk_mmu);
3727

3728 3729
	vcpu->arch.walk_mmu->set_cr3           = kvm_x86_ops->set_cr3;
	vcpu->arch.walk_mmu->get_cr3           = get_cr3;
3730
	vcpu->arch.walk_mmu->get_pdptr         = kvm_pdptr_read;
3731
	vcpu->arch.walk_mmu->inject_page_fault = kvm_inject_page_fault;
3732 3733

	return r;
A
Avi Kivity 已提交
3734 3735
}

3736 3737 3738 3739 3740
static int init_kvm_nested_mmu(struct kvm_vcpu *vcpu)
{
	struct kvm_mmu *g_context = &vcpu->arch.nested_mmu;

	g_context->get_cr3           = get_cr3;
3741
	g_context->get_pdptr         = kvm_pdptr_read;
3742 3743 3744 3745 3746 3747 3748 3749 3750
	g_context->inject_page_fault = kvm_inject_page_fault;

	/*
	 * Note that arch.mmu.gva_to_gpa translates l2_gva to l1_gpa. The
	 * translation of l2_gpa to l1_gpa addresses is done using the
	 * arch.nested_mmu.gva_to_gpa function. Basically the gva_to_gpa
	 * functions between mmu and nested_mmu are swapped.
	 */
	if (!is_paging(vcpu)) {
3751
		g_context->nx = false;
3752 3753 3754
		g_context->root_level = 0;
		g_context->gva_to_gpa = nonpaging_gva_to_gpa_nested;
	} else if (is_long_mode(vcpu)) {
3755
		g_context->nx = is_nx(vcpu);
3756
		g_context->root_level = PT64_ROOT_LEVEL;
3757
		reset_rsvds_bits_mask(vcpu, g_context);
3758 3759
		g_context->gva_to_gpa = paging64_gva_to_gpa_nested;
	} else if (is_pae(vcpu)) {
3760
		g_context->nx = is_nx(vcpu);
3761
		g_context->root_level = PT32E_ROOT_LEVEL;
3762
		reset_rsvds_bits_mask(vcpu, g_context);
3763 3764
		g_context->gva_to_gpa = paging64_gva_to_gpa_nested;
	} else {
3765
		g_context->nx = false;
3766
		g_context->root_level = PT32_ROOT_LEVEL;
3767
		reset_rsvds_bits_mask(vcpu, g_context);
3768 3769 3770
		g_context->gva_to_gpa = paging32_gva_to_gpa_nested;
	}

3771
	update_permission_bitmask(vcpu, g_context);
A
Avi Kivity 已提交
3772
	update_last_pte_bitmap(vcpu, g_context);
3773

3774 3775 3776
	return 0;
}

3777 3778
static int init_kvm_mmu(struct kvm_vcpu *vcpu)
{
3779 3780 3781
	if (mmu_is_nested(vcpu))
		return init_kvm_nested_mmu(vcpu);
	else if (tdp_enabled)
3782 3783 3784 3785 3786
		return init_kvm_tdp_mmu(vcpu);
	else
		return init_kvm_softmmu(vcpu);
}

A
Avi Kivity 已提交
3787 3788 3789
static void destroy_kvm_mmu(struct kvm_vcpu *vcpu)
{
	ASSERT(vcpu);
3790 3791
	if (VALID_PAGE(vcpu->arch.mmu.root_hpa))
		/* mmu.free() should set root_hpa = INVALID_PAGE */
3792
		vcpu->arch.mmu.free(vcpu);
A
Avi Kivity 已提交
3793 3794 3795
}

int kvm_mmu_reset_context(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
3796 3797
{
	destroy_kvm_mmu(vcpu);
3798
	return init_kvm_mmu(vcpu);
A
Avi Kivity 已提交
3799
}
3800
EXPORT_SYMBOL_GPL(kvm_mmu_reset_context);
A
Avi Kivity 已提交
3801 3802

int kvm_mmu_load(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
3803
{
3804 3805
	int r;

3806
	r = mmu_topup_memory_caches(vcpu);
A
Avi Kivity 已提交
3807 3808
	if (r)
		goto out;
3809
	r = mmu_alloc_roots(vcpu);
3810
	spin_lock(&vcpu->kvm->mmu_lock);
3811
	mmu_sync_roots(vcpu);
3812
	spin_unlock(&vcpu->kvm->mmu_lock);
3813 3814
	if (r)
		goto out;
3815
	/* set_cr3() should ensure TLB has been flushed */
3816
	vcpu->arch.mmu.set_cr3(vcpu, vcpu->arch.mmu.root_hpa);
3817 3818
out:
	return r;
A
Avi Kivity 已提交
3819
}
A
Avi Kivity 已提交
3820 3821 3822 3823 3824 3825
EXPORT_SYMBOL_GPL(kvm_mmu_load);

void kvm_mmu_unload(struct kvm_vcpu *vcpu)
{
	mmu_free_roots(vcpu);
}
3826
EXPORT_SYMBOL_GPL(kvm_mmu_unload);
A
Avi Kivity 已提交
3827

3828
static void mmu_pte_write_new_pte(struct kvm_vcpu *vcpu,
3829 3830
				  struct kvm_mmu_page *sp, u64 *spte,
				  const void *new)
3831
{
3832
	if (sp->role.level != PT_PAGE_TABLE_LEVEL) {
3833 3834
		++vcpu->kvm->stat.mmu_pde_zapped;
		return;
3835
        }
3836

A
Avi Kivity 已提交
3837
	++vcpu->kvm->stat.mmu_pte_updated;
3838
	vcpu->arch.mmu.update_pte(vcpu, sp, spte, new);
3839 3840
}

3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853
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;
	old ^= PT64_NX_MASK;
	new ^= PT64_NX_MASK;
	return (old & ~new & PT64_PERM_MASK) != 0;
}

3854 3855
static void mmu_pte_write_flush_tlb(struct kvm_vcpu *vcpu, bool zap_page,
				    bool remote_flush, bool local_flush)
3856
{
3857 3858 3859 3860
	if (zap_page)
		return;

	if (remote_flush)
3861
		kvm_flush_remote_tlbs(vcpu->kvm);
3862
	else if (local_flush)
3863 3864 3865
		kvm_mmu_flush_tlb(vcpu);
}

3866 3867
static u64 mmu_pte_write_fetch_gpte(struct kvm_vcpu *vcpu, gpa_t *gpa,
				    const u8 *new, int *bytes)
3868
{
3869 3870
	u64 gentry;
	int r;
3871 3872 3873

	/*
	 * Assume that the pte write on a page table of the same type
3874 3875
	 * as the current vcpu paging mode since we update the sptes only
	 * when they have the same mode.
3876
	 */
3877
	if (is_pae(vcpu) && *bytes == 4) {
3878
		/* Handle a 32-bit guest writing two halves of a 64-bit gpte */
3879 3880 3881
		*gpa &= ~(gpa_t)7;
		*bytes = 8;
		r = kvm_read_guest(vcpu->kvm, *gpa, &gentry, min(*bytes, 8));
3882 3883
		if (r)
			gentry = 0;
3884 3885 3886
		new = (const u8 *)&gentry;
	}

3887
	switch (*bytes) {
3888 3889 3890 3891 3892 3893 3894 3895 3896
	case 4:
		gentry = *(const u32 *)new;
		break;
	case 8:
		gentry = *(const u64 *)new;
		break;
	default:
		gentry = 0;
		break;
3897 3898
	}

3899 3900 3901 3902 3903 3904 3905
	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.
 */
3906
static bool detect_write_flooding(struct kvm_mmu_page *sp)
3907
{
3908 3909 3910 3911
	/*
	 * Skip write-flooding detected for the sp whose level is 1, because
	 * it can become unsync, then the guest page is not write-protected.
	 */
3912
	if (sp->role.level == PT_PAGE_TABLE_LEVEL)
3913
		return false;
3914

3915
	return ++sp->write_flooding_count >= 3;
3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931
}

/*
 * 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;
3932 3933 3934 3935 3936 3937 3938 3939

	/*
	 * 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;

3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986
	misaligned = (offset ^ (offset + bytes - 1)) & ~(pte_size - 1);
	misaligned |= bytes < 4;

	return misaligned;
}

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

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

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

void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
		       const u8 *new, int bytes)
{
	gfn_t gfn = gpa >> PAGE_SHIFT;
	union kvm_mmu_page_role mask = { .word = 0 };
	struct kvm_mmu_page *sp;
	struct hlist_node *node;
	LIST_HEAD(invalid_list);
	u64 entry, gentry, *spte;
	int npte;
3987
	bool remote_flush, local_flush, zap_page;
3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010

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

	zap_page = remote_flush = local_flush = false;

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

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

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

	spin_lock(&vcpu->kvm->mmu_lock);
	++vcpu->kvm->stat.mmu_pte_write;
4011
	kvm_mmu_audit(vcpu, AUDIT_PRE_PTE_WRITE);
4012

4013
	mask.cr0_wp = mask.cr4_pae = mask.nxe = 1;
4014
	for_each_gfn_indirect_valid_sp(vcpu->kvm, sp, gfn, node) {
4015
		if (detect_write_misaligned(sp, gpa, bytes) ||
4016
		      detect_write_flooding(sp)) {
4017
			zap_page |= !!kvm_mmu_prepare_zap_page(vcpu->kvm, sp,
4018
						     &invalid_list);
A
Avi Kivity 已提交
4019
			++vcpu->kvm->stat.mmu_flooded;
4020 4021
			continue;
		}
4022 4023 4024 4025 4026

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

4027
		local_flush = true;
4028
		while (npte--) {
4029
			entry = *spte;
4030
			mmu_page_zap_pte(vcpu->kvm, sp, spte);
4031 4032
			if (gentry &&
			      !((sp->role.word ^ vcpu->arch.mmu.base_role.word)
4033
			      & mask.word) && rmap_can_add(vcpu))
4034
				mmu_pte_write_new_pte(vcpu, sp, spte, &gentry);
4035 4036
			if (!remote_flush && need_remote_flush(entry, *spte))
				remote_flush = true;
4037
			++spte;
4038 4039
		}
	}
4040
	mmu_pte_write_flush_tlb(vcpu, zap_page, remote_flush, local_flush);
4041
	kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
4042
	kvm_mmu_audit(vcpu, AUDIT_POST_PTE_WRITE);
4043
	spin_unlock(&vcpu->kvm->mmu_lock);
4044 4045
}

4046 4047
int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva)
{
4048 4049
	gpa_t gpa;
	int r;
4050

4051
	if (vcpu->arch.mmu.direct_map)
4052 4053
		return 0;

4054
	gpa = kvm_mmu_gva_to_gpa_read(vcpu, gva, NULL);
4055 4056

	r = kvm_mmu_unprotect_page(vcpu->kvm, gpa >> PAGE_SHIFT);
4057

4058
	return r;
4059
}
4060
EXPORT_SYMBOL_GPL(kvm_mmu_unprotect_page_virt);
4061

4062
void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4063
{
4064
	LIST_HEAD(invalid_list);
4065

4066
	while (kvm_mmu_available_pages(vcpu->kvm) < KVM_REFILL_PAGES &&
4067
	       !list_empty(&vcpu->kvm->arch.active_mmu_pages)) {
4068
		struct kvm_mmu_page *sp;
A
Avi Kivity 已提交
4069

4070
		sp = container_of(vcpu->kvm->arch.active_mmu_pages.prev,
4071
				  struct kvm_mmu_page, link);
4072
		kvm_mmu_prepare_zap_page(vcpu->kvm, sp, &invalid_list);
A
Avi Kivity 已提交
4073
		++vcpu->kvm->stat.mmu_recycled;
A
Avi Kivity 已提交
4074
	}
4075
	kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
A
Avi Kivity 已提交
4076 4077
}

4078 4079 4080 4081 4082 4083 4084 4085
static bool is_mmio_page_fault(struct kvm_vcpu *vcpu, gva_t addr)
{
	if (vcpu->arch.mmu.direct_map || mmu_is_nested(vcpu))
		return vcpu_match_mmio_gpa(vcpu, addr);

	return vcpu_match_mmio_gva(vcpu, addr);
}

4086 4087
int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t cr2, u32 error_code,
		       void *insn, int insn_len)
4088
{
4089
	int r, emulation_type = EMULTYPE_RETRY;
4090 4091
	enum emulation_result er;

G
Gleb Natapov 已提交
4092
	r = vcpu->arch.mmu.page_fault(vcpu, cr2, error_code, false);
4093 4094 4095 4096 4097 4098 4099 4100
	if (r < 0)
		goto out;

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

4101 4102 4103 4104
	if (is_mmio_page_fault(vcpu, cr2))
		emulation_type = 0;

	er = x86_emulate_instruction(vcpu, cr2, emulation_type, insn, insn_len);
4105 4106 4107 4108 4109 4110

	switch (er) {
	case EMULATE_DONE:
		return 1;
	case EMULATE_DO_MMIO:
		++vcpu->stat.mmio_exits;
4111
		/* fall through */
4112
	case EMULATE_FAIL:
4113
		return 0;
4114 4115 4116 4117 4118 4119 4120 4121
	default:
		BUG();
	}
out:
	return r;
}
EXPORT_SYMBOL_GPL(kvm_mmu_page_fault);

M
Marcelo Tosatti 已提交
4122 4123 4124 4125 4126 4127 4128 4129
void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
{
	vcpu->arch.mmu.invlpg(vcpu, gva);
	kvm_mmu_flush_tlb(vcpu);
	++vcpu->stat.invlpg;
}
EXPORT_SYMBOL_GPL(kvm_mmu_invlpg);

4130 4131 4132 4133 4134 4135
void kvm_enable_tdp(void)
{
	tdp_enabled = true;
}
EXPORT_SYMBOL_GPL(kvm_enable_tdp);

4136 4137 4138 4139 4140 4141
void kvm_disable_tdp(void)
{
	tdp_enabled = false;
}
EXPORT_SYMBOL_GPL(kvm_disable_tdp);

A
Avi Kivity 已提交
4142 4143
static void free_mmu_pages(struct kvm_vcpu *vcpu)
{
4144
	free_page((unsigned long)vcpu->arch.mmu.pae_root);
4145 4146
	if (vcpu->arch.mmu.lm_root != NULL)
		free_page((unsigned long)vcpu->arch.mmu.lm_root);
A
Avi Kivity 已提交
4147 4148 4149 4150
}

static int alloc_mmu_pages(struct kvm_vcpu *vcpu)
{
4151
	struct page *page;
A
Avi Kivity 已提交
4152 4153 4154 4155
	int i;

	ASSERT(vcpu);

4156 4157 4158 4159 4160 4161 4162
	/*
	 * 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)
4163 4164
		return -ENOMEM;

4165
	vcpu->arch.mmu.pae_root = page_address(page);
4166
	for (i = 0; i < 4; ++i)
4167
		vcpu->arch.mmu.pae_root[i] = INVALID_PAGE;
4168

A
Avi Kivity 已提交
4169 4170 4171
	return 0;
}

4172
int kvm_mmu_create(struct kvm_vcpu *vcpu)
A
Avi Kivity 已提交
4173 4174
{
	ASSERT(vcpu);
4175 4176 4177 4178 4179

	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
4180

4181 4182
	return alloc_mmu_pages(vcpu);
}
A
Avi Kivity 已提交
4183

4184 4185 4186
int kvm_mmu_setup(struct kvm_vcpu *vcpu)
{
	ASSERT(vcpu);
4187
	ASSERT(!VALID_PAGE(vcpu->arch.mmu.root_hpa));
4188

4189
	return init_kvm_mmu(vcpu);
A
Avi Kivity 已提交
4190 4191
}

4192
void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot)
A
Avi Kivity 已提交
4193
{
4194
	struct kvm_mmu_page *sp;
4195
	bool flush = false;
A
Avi Kivity 已提交
4196

4197
	list_for_each_entry(sp, &kvm->arch.active_mmu_pages, link) {
A
Avi Kivity 已提交
4198 4199 4200
		int i;
		u64 *pt;

4201
		if (!test_bit(slot, sp->slot_bitmap))
A
Avi Kivity 已提交
4202 4203
			continue;

4204
		pt = sp->spt;
4205
		for (i = 0; i < PT64_ENT_PER_PAGE; ++i) {
4206 4207 4208 4209
			if (!is_shadow_present_pte(pt[i]) ||
			      !is_last_spte(pt[i], sp->role.level))
				continue;

4210
			spte_write_protect(kvm, &pt[i], &flush, false);
4211
		}
A
Avi Kivity 已提交
4212
	}
4213
	kvm_flush_remote_tlbs(kvm);
A
Avi Kivity 已提交
4214
}
4215

4216
void kvm_mmu_zap_all(struct kvm *kvm)
D
Dor Laor 已提交
4217
{
4218
	struct kvm_mmu_page *sp, *node;
4219
	LIST_HEAD(invalid_list);
D
Dor Laor 已提交
4220

4221
	spin_lock(&kvm->mmu_lock);
4222
restart:
4223
	list_for_each_entry_safe(sp, node, &kvm->arch.active_mmu_pages, link)
4224
		if (kvm_mmu_prepare_zap_page(kvm, sp, &invalid_list))
4225 4226
			goto restart;

4227
	kvm_mmu_commit_zap_page(kvm, &invalid_list);
4228
	spin_unlock(&kvm->mmu_lock);
D
Dor Laor 已提交
4229 4230
}

4231 4232
static void kvm_mmu_remove_some_alloc_mmu_pages(struct kvm *kvm,
						struct list_head *invalid_list)
4233 4234 4235
{
	struct kvm_mmu_page *page;

4236 4237 4238
	if (list_empty(&kvm->arch.active_mmu_pages))
		return;

4239 4240
	page = container_of(kvm->arch.active_mmu_pages.prev,
			    struct kvm_mmu_page, link);
4241
	kvm_mmu_prepare_zap_page(kvm, page, invalid_list);
4242 4243
}

4244
static int mmu_shrink(struct shrinker *shrink, struct shrink_control *sc)
4245 4246
{
	struct kvm *kvm;
4247
	int nr_to_scan = sc->nr_to_scan;
4248 4249 4250

	if (nr_to_scan == 0)
		goto out;
4251

4252
	raw_spin_lock(&kvm_lock);
4253 4254

	list_for_each_entry(kvm, &vm_list, vm_list) {
4255
		int idx;
4256
		LIST_HEAD(invalid_list);
4257

4258 4259 4260 4261 4262 4263 4264 4265
		/*
		 * 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;
4266 4267 4268 4269 4270 4271
		/*
		 * 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.
		 */
4272
		if (!kvm->arch.n_used_mmu_pages)
4273 4274
			continue;

4275
		idx = srcu_read_lock(&kvm->srcu);
4276 4277
		spin_lock(&kvm->mmu_lock);

4278
		kvm_mmu_remove_some_alloc_mmu_pages(kvm, &invalid_list);
4279
		kvm_mmu_commit_zap_page(kvm, &invalid_list);
4280

4281
		spin_unlock(&kvm->mmu_lock);
4282
		srcu_read_unlock(&kvm->srcu, idx);
4283 4284 4285

		list_move_tail(&kvm->vm_list, &vm_list);
		break;
4286 4287
	}

4288
	raw_spin_unlock(&kvm_lock);
4289

4290 4291
out:
	return percpu_counter_read_positive(&kvm_total_used_mmu_pages);
4292 4293 4294 4295 4296 4297 4298
}

static struct shrinker mmu_shrinker = {
	.shrink = mmu_shrink,
	.seeks = DEFAULT_SEEKS * 10,
};

I
Ingo Molnar 已提交
4299
static void mmu_destroy_caches(void)
4300
{
4301 4302
	if (pte_list_desc_cache)
		kmem_cache_destroy(pte_list_desc_cache);
4303 4304
	if (mmu_page_header_cache)
		kmem_cache_destroy(mmu_page_header_cache);
4305 4306 4307 4308
}

int kvm_mmu_module_init(void)
{
4309 4310
	pte_list_desc_cache = kmem_cache_create("pte_list_desc",
					    sizeof(struct pte_list_desc),
4311
					    0, 0, NULL);
4312
	if (!pte_list_desc_cache)
4313 4314
		goto nomem;

4315 4316
	mmu_page_header_cache = kmem_cache_create("kvm_mmu_page_header",
						  sizeof(struct kvm_mmu_page),
4317
						  0, 0, NULL);
4318 4319 4320
	if (!mmu_page_header_cache)
		goto nomem;

4321 4322 4323
	if (percpu_counter_init(&kvm_total_used_mmu_pages, 0))
		goto nomem;

4324 4325
	register_shrinker(&mmu_shrinker);

4326 4327 4328
	return 0;

nomem:
4329
	mmu_destroy_caches();
4330 4331 4332
	return -ENOMEM;
}

4333 4334 4335 4336 4337 4338 4339
/*
 * 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;
4340
	struct kvm_memslots *slots;
4341
	struct kvm_memory_slot *memslot;
4342

4343 4344
	slots = kvm_memslots(kvm);

4345 4346
	kvm_for_each_memslot(memslot, slots)
		nr_pages += memslot->npages;
4347 4348 4349 4350 4351 4352 4353 4354

	nr_mmu_pages = nr_pages * KVM_PERMILLE_MMU_PAGES / 1000;
	nr_mmu_pages = max(nr_mmu_pages,
			(unsigned int) KVM_MIN_ALLOC_MMU_PAGES);

	return nr_mmu_pages;
}

4355 4356 4357
int kvm_mmu_get_spte_hierarchy(struct kvm_vcpu *vcpu, u64 addr, u64 sptes[4])
{
	struct kvm_shadow_walk_iterator iterator;
4358
	u64 spte;
4359 4360
	int nr_sptes = 0;

4361 4362 4363
	walk_shadow_page_lockless_begin(vcpu);
	for_each_shadow_entry_lockless(vcpu, addr, iterator, spte) {
		sptes[iterator.level-1] = spte;
4364
		nr_sptes++;
4365
		if (!is_shadow_present_pte(spte))
4366 4367
			break;
	}
4368
	walk_shadow_page_lockless_end(vcpu);
4369 4370 4371 4372 4373

	return nr_sptes;
}
EXPORT_SYMBOL_GPL(kvm_mmu_get_spte_hierarchy);

4374 4375 4376 4377 4378 4379 4380
void kvm_mmu_destroy(struct kvm_vcpu *vcpu)
{
	ASSERT(vcpu);

	destroy_kvm_mmu(vcpu);
	free_mmu_pages(vcpu);
	mmu_free_memory_caches(vcpu);
4381 4382 4383 4384 4385 4386 4387
}

void kvm_mmu_module_exit(void)
{
	mmu_destroy_caches();
	percpu_counter_destroy(&kvm_total_used_mmu_pages);
	unregister_shrinker(&mmu_shrinker);
4388 4389
	mmu_audit_disable();
}