blob: fa71085f75a3e8989d464daf65bd963e49f47fee [file] [log] [blame]
Avi Kivity6aa8b732006-12-10 02:21:36 -08001/*
2 * Kernel-based Virtual Machine driver for Linux
3 *
4 * This module enables machines with Intel VT-x extensions to run virtual
5 * machines without emulation or binary translation.
6 *
7 * MMU support
8 *
9 * Copyright (C) 2006 Qumranet, Inc.
Nicolas Kaiser9611c182010-10-06 14:23:22 +020010 * Copyright 2010 Red Hat, Inc. and/or its affiliates.
Avi Kivity6aa8b732006-12-10 02:21:36 -080011 *
12 * Authors:
13 * Yaniv Kamay <yaniv@qumranet.com>
14 * Avi Kivity <avi@qumranet.com>
15 *
16 * This work is licensed under the terms of the GNU GPL, version 2. See
17 * the COPYING file in the top-level directory.
18 *
19 */
Avi Kivity6aa8b732006-12-10 02:21:36 -080020
Gleb Natapovaf585b92010-10-14 11:22:46 +020021#include "irq.h"
Zhang Xiantao1d737c82007-12-14 09:35:10 +080022#include "mmu.h"
Avi Kivity836a1b32010-01-21 15:31:49 +020023#include "x86.h"
Avi Kivity6de4f3ad2009-05-31 22:58:47 +030024#include "kvm_cache_regs.h"
Avi Kivity6aa8b732006-12-10 02:21:36 -080025
Avi Kivityedf88412007-12-16 11:02:48 +020026#include <linux/kvm_host.h>
Avi Kivitye4956062007-06-28 14:15:57 -040027#include <linux/types.h>
28#include <linux/string.h>
29#include <linux/mm.h>
30#include <linux/highmem.h>
31#include <linux/module.h>
Izik Eidus448353c2007-11-26 14:08:14 +020032#include <linux/swap.h>
Marcelo Tosatti05da4552008-02-23 11:44:30 -030033#include <linux/hugetlb.h>
Marcelo Tosatti2f333bc2008-02-22 12:21:37 -050034#include <linux/compiler.h>
Marcelo Tosattibc6678a2009-12-23 14:35:21 -020035#include <linux/srcu.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090036#include <linux/slab.h>
Huang Yingbf998152010-05-31 14:28:19 +080037#include <linux/uaccess.h>
Avi Kivitye4956062007-06-28 14:15:57 -040038
39#include <asm/page.h>
40#include <asm/cmpxchg.h>
Avi Kivity4e542372007-11-21 14:08:40 +020041#include <asm/io.h>
Eduardo Habkost13673a92008-11-17 19:03:13 -020042#include <asm/vmx.h>
Avi Kivitye4956062007-06-28 14:15:57 -040043
Joerg Roedel18552672008-02-07 13:47:41 +010044/*
45 * When setting this variable to true it enables Two-Dimensional-Paging
46 * where the hardware walks 2 page tables:
47 * 1. the guest-virtual to guest-physical
48 * 2. while doing 1. it walks guest-physical to host-physical
49 * If the hardware supports that we don't need to do shadow paging.
50 */
Marcelo Tosatti2f333bc2008-02-22 12:21:37 -050051bool tdp_enabled = false;
Joerg Roedel18552672008-02-07 13:47:41 +010052
Xiao Guangrong8b1fe172010-08-30 18:22:53 +080053enum {
54 AUDIT_PRE_PAGE_FAULT,
55 AUDIT_POST_PAGE_FAULT,
56 AUDIT_PRE_PTE_WRITE,
Xiao Guangrong69030742010-09-27 18:09:29 +080057 AUDIT_POST_PTE_WRITE,
58 AUDIT_PRE_SYNC,
59 AUDIT_POST_SYNC
Xiao Guangrong8b1fe172010-08-30 18:22:53 +080060};
61
62char *audit_point_name[] = {
63 "pre page fault",
64 "post page fault",
65 "pre pte write",
Xiao Guangrong69030742010-09-27 18:09:29 +080066 "post pte write",
67 "pre sync",
68 "post sync"
Xiao Guangrong8b1fe172010-08-30 18:22:53 +080069};
70
Avi Kivity37a7d8b2007-01-05 16:36:56 -080071#undef MMU_DEBUG
72
Avi Kivity37a7d8b2007-01-05 16:36:56 -080073#ifdef MMU_DEBUG
74
75#define pgprintk(x...) do { if (dbg) printk(x); } while (0)
76#define rmap_printk(x...) do { if (dbg) printk(x); } while (0)
77
78#else
79
80#define pgprintk(x...) do { } while (0)
81#define rmap_printk(x...) do { } while (0)
82
83#endif
84
Xiao Guangrong8b1fe172010-08-30 18:22:53 +080085#ifdef MMU_DEBUG
Avi Kivity6ada8cc2008-06-22 16:45:24 +030086static int dbg = 0;
87module_param(dbg, bool, 0644);
Avi Kivity37a7d8b2007-01-05 16:36:56 -080088#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -080089
Marcelo Tosatti582801a2008-09-23 13:18:41 -030090static int oos_shadow = 1;
91module_param(oos_shadow, bool, 0644);
92
Yaozu Dongd6c69ee2007-04-25 14:17:25 +080093#ifndef MMU_DEBUG
94#define ASSERT(x) do { } while (0)
95#else
Avi Kivity6aa8b732006-12-10 02:21:36 -080096#define ASSERT(x) \
97 if (!(x)) { \
98 printk(KERN_WARNING "assertion failed %s:%d: %s\n", \
99 __FILE__, __LINE__, #x); \
100 }
Yaozu Dongd6c69ee2007-04-25 14:17:25 +0800101#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -0800102
Xiao Guangrong957ed9e2010-08-22 19:12:48 +0800103#define PTE_PREFETCH_NUM 8
104
Avi Kivity6aa8b732006-12-10 02:21:36 -0800105#define PT_FIRST_AVAIL_BITS_SHIFT 9
106#define PT64_SECOND_AVAIL_BITS_SHIFT 52
107
Avi Kivity6aa8b732006-12-10 02:21:36 -0800108#define PT64_LEVEL_BITS 9
109
110#define PT64_LEVEL_SHIFT(level) \
Mike Dayd77c26f2007-10-08 09:02:08 -0400111 (PAGE_SHIFT + (level - 1) * PT64_LEVEL_BITS)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800112
Avi Kivity6aa8b732006-12-10 02:21:36 -0800113#define PT64_INDEX(address, level)\
114 (((address) >> PT64_LEVEL_SHIFT(level)) & ((1 << PT64_LEVEL_BITS) - 1))
115
116
117#define PT32_LEVEL_BITS 10
118
119#define PT32_LEVEL_SHIFT(level) \
Mike Dayd77c26f2007-10-08 09:02:08 -0400120 (PAGE_SHIFT + (level - 1) * PT32_LEVEL_BITS)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800121
Joerg Roedele04da982009-07-27 16:30:45 +0200122#define PT32_LVL_OFFSET_MASK(level) \
123 (PT32_BASE_ADDR_MASK & ((1ULL << (PAGE_SHIFT + (((level) - 1) \
124 * PT32_LEVEL_BITS))) - 1))
Avi Kivity6aa8b732006-12-10 02:21:36 -0800125
126#define PT32_INDEX(address, level)\
127 (((address) >> PT32_LEVEL_SHIFT(level)) & ((1 << PT32_LEVEL_BITS) - 1))
128
129
Avi Kivity27aba762007-03-09 13:04:31 +0200130#define PT64_BASE_ADDR_MASK (((1ULL << 52) - 1) & ~(u64)(PAGE_SIZE-1))
Avi Kivity6aa8b732006-12-10 02:21:36 -0800131#define PT64_DIR_BASE_ADDR_MASK \
132 (PT64_BASE_ADDR_MASK & ~((1ULL << (PAGE_SHIFT + PT64_LEVEL_BITS)) - 1))
Joerg Roedele04da982009-07-27 16:30:45 +0200133#define PT64_LVL_ADDR_MASK(level) \
134 (PT64_BASE_ADDR_MASK & ~((1ULL << (PAGE_SHIFT + (((level) - 1) \
135 * PT64_LEVEL_BITS))) - 1))
136#define PT64_LVL_OFFSET_MASK(level) \
137 (PT64_BASE_ADDR_MASK & ((1ULL << (PAGE_SHIFT + (((level) - 1) \
138 * PT64_LEVEL_BITS))) - 1))
Avi Kivity6aa8b732006-12-10 02:21:36 -0800139
140#define PT32_BASE_ADDR_MASK PAGE_MASK
141#define PT32_DIR_BASE_ADDR_MASK \
142 (PAGE_MASK & ~((1ULL << (PAGE_SHIFT + PT32_LEVEL_BITS)) - 1))
Joerg Roedele04da982009-07-27 16:30:45 +0200143#define PT32_LVL_ADDR_MASK(level) \
144 (PAGE_MASK & ~((1ULL << (PAGE_SHIFT + (((level) - 1) \
145 * PT32_LEVEL_BITS))) - 1))
Avi Kivity6aa8b732006-12-10 02:21:36 -0800146
Avi Kivity79539ce2007-11-21 02:06:21 +0200147#define PT64_PERM_MASK (PT_PRESENT_MASK | PT_WRITABLE_MASK | PT_USER_MASK \
148 | PT64_NX_MASK)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800149
Xiao Guangrong53c07b12011-05-15 23:26:20 +0800150#define PTE_LIST_EXT 4
Avi Kivitycd4a4e52007-01-05 16:36:38 -0800151
Avi Kivityfe135d22007-12-09 16:15:46 +0200152#define ACC_EXEC_MASK 1
153#define ACC_WRITE_MASK PT_WRITABLE_MASK
154#define ACC_USER_MASK PT_USER_MASK
155#define ACC_ALL (ACC_EXEC_MASK | ACC_WRITE_MASK | ACC_USER_MASK)
156
Avi Kivity90bb6fc2009-12-31 12:10:16 +0200157#include <trace/events/kvm.h>
158
Avi Kivity07420172009-07-06 12:21:32 +0300159#define CREATE_TRACE_POINTS
160#include "mmutrace.h"
161
Izik Eidus14032832009-09-23 21:47:17 +0300162#define SPTE_HOST_WRITEABLE (1ULL << PT_FIRST_AVAIL_BITS_SHIFT)
163
Avi Kivity135f8c22008-08-21 17:49:56 +0300164#define SHADOW_PT_INDEX(addr, level) PT64_INDEX(addr, level)
165
Xiao Guangrong53c07b12011-05-15 23:26:20 +0800166struct pte_list_desc {
167 u64 *sptes[PTE_LIST_EXT];
168 struct pte_list_desc *more;
Avi Kivitycd4a4e52007-01-05 16:36:38 -0800169};
170
Avi Kivity2d111232008-12-25 14:39:47 +0200171struct kvm_shadow_walk_iterator {
172 u64 addr;
173 hpa_t shadow_addr;
Avi Kivity2d111232008-12-25 14:39:47 +0200174 u64 *sptep;
Xiao Guangrongdd3bfd52011-07-12 03:32:54 +0800175 int level;
Avi Kivity2d111232008-12-25 14:39:47 +0200176 unsigned index;
177};
178
179#define for_each_shadow_entry(_vcpu, _addr, _walker) \
180 for (shadow_walk_init(&(_walker), _vcpu, _addr); \
181 shadow_walk_okay(&(_walker)); \
182 shadow_walk_next(&(_walker)))
183
Xiao Guangrongc2a2ac22011-07-12 03:32:13 +0800184#define for_each_shadow_entry_lockless(_vcpu, _addr, _walker, spte) \
185 for (shadow_walk_init(&(_walker), _vcpu, _addr); \
186 shadow_walk_okay(&(_walker)) && \
187 ({ spte = mmu_spte_get_lockless(_walker.sptep); 1; }); \
188 __shadow_walk_next(&(_walker), spte))
189
Xiao Guangrong53c07b12011-05-15 23:26:20 +0800190static struct kmem_cache *pte_list_desc_cache;
Avi Kivityd3d25b02007-05-30 12:34:53 +0300191static struct kmem_cache *mmu_page_header_cache;
Dave Hansen45221ab2010-08-19 18:11:37 -0700192static struct percpu_counter kvm_total_used_mmu_pages;
Avi Kivityb5a33a72007-04-15 16:31:09 +0300193
Sheng Yang7b523452008-04-25 21:13:50 +0800194static u64 __read_mostly shadow_nx_mask;
195static u64 __read_mostly shadow_x_mask; /* mutual exclusive with nx_mask */
196static u64 __read_mostly shadow_user_mask;
197static u64 __read_mostly shadow_accessed_mask;
198static u64 __read_mostly shadow_dirty_mask;
Xiao Guangrongce88dec2011-07-12 03:33:44 +0800199static u64 __read_mostly shadow_mmio_mask;
200
201static void mmu_spte_set(u64 *sptep, u64 spte);
202
203void kvm_mmu_set_mmio_spte_mask(u64 mmio_mask)
204{
205 shadow_mmio_mask = mmio_mask;
206}
207EXPORT_SYMBOL_GPL(kvm_mmu_set_mmio_spte_mask);
208
209static void mark_mmio_spte(u64 *sptep, u64 gfn, unsigned access)
210{
211 access &= ACC_WRITE_MASK | ACC_USER_MASK;
212
Xiao Guangrong4f022642011-07-12 03:34:24 +0800213 trace_mark_mmio_spte(sptep, gfn, access);
Xiao Guangrongce88dec2011-07-12 03:33:44 +0800214 mmu_spte_set(sptep, shadow_mmio_mask | access | gfn << PAGE_SHIFT);
215}
216
217static bool is_mmio_spte(u64 spte)
218{
219 return (spte & shadow_mmio_mask) == shadow_mmio_mask;
220}
221
222static gfn_t get_mmio_spte_gfn(u64 spte)
223{
224 return (spte & ~shadow_mmio_mask) >> PAGE_SHIFT;
225}
226
227static unsigned get_mmio_spte_access(u64 spte)
228{
229 return (spte & ~shadow_mmio_mask) & ~PAGE_MASK;
230}
231
232static bool set_mmio_spte(u64 *sptep, gfn_t gfn, pfn_t pfn, unsigned access)
233{
234 if (unlikely(is_noslot_pfn(pfn))) {
235 mark_mmio_spte(sptep, gfn, access);
236 return true;
237 }
238
239 return false;
240}
Avi Kivityc7addb92007-09-16 18:58:32 +0200241
Dong, Eddie82725b22009-03-30 16:21:08 +0800242static inline u64 rsvd_bits(int s, int e)
243{
244 return ((1ULL << (e - s + 1)) - 1) << s;
245}
246
Sheng Yang7b523452008-04-25 21:13:50 +0800247void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
Sheng Yang4b12f0d2009-04-27 20:35:42 +0800248 u64 dirty_mask, u64 nx_mask, u64 x_mask)
Sheng Yang7b523452008-04-25 21:13:50 +0800249{
250 shadow_user_mask = user_mask;
251 shadow_accessed_mask = accessed_mask;
252 shadow_dirty_mask = dirty_mask;
253 shadow_nx_mask = nx_mask;
254 shadow_x_mask = x_mask;
255}
256EXPORT_SYMBOL_GPL(kvm_mmu_set_mask_ptes);
257
Avi Kivity6aa8b732006-12-10 02:21:36 -0800258static int is_cpuid_PSE36(void)
259{
260 return 1;
261}
262
Avi Kivity73b10872007-01-26 00:56:41 -0800263static int is_nx(struct kvm_vcpu *vcpu)
264{
Avi Kivityf6801df2010-01-21 15:31:50 +0200265 return vcpu->arch.efer & EFER_NX;
Avi Kivity73b10872007-01-26 00:56:41 -0800266}
267
Avi Kivityc7addb92007-09-16 18:58:32 +0200268static int is_shadow_present_pte(u64 pte)
269{
Xiao Guangrongce88dec2011-07-12 03:33:44 +0800270 return pte & PT_PRESENT_MASK && !is_mmio_spte(pte);
Avi Kivityc7addb92007-09-16 18:58:32 +0200271}
272
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300273static int is_large_pte(u64 pte)
274{
275 return pte & PT_PAGE_SIZE_MASK;
276}
277
Avi Kivity43a37952009-06-10 14:12:05 +0300278static int is_dirty_gpte(unsigned long pte)
Avi Kivitye3c5e7ec2007-10-11 12:32:30 +0200279{
Avi Kivity439e2182009-06-10 12:56:54 +0300280 return pte & PT_DIRTY_MASK;
Avi Kivitye3c5e7ec2007-10-11 12:32:30 +0200281}
282
Avi Kivity43a37952009-06-10 14:12:05 +0300283static int is_rmap_spte(u64 pte)
Avi Kivitycd4a4e52007-01-05 16:36:38 -0800284{
Avi Kivity4b1a80f2008-03-23 12:18:19 +0200285 return is_shadow_present_pte(pte);
Avi Kivitycd4a4e52007-01-05 16:36:38 -0800286}
287
Marcelo Tosatti776e6632009-06-10 12:27:03 -0300288static int is_last_spte(u64 pte, int level)
289{
290 if (level == PT_PAGE_TABLE_LEVEL)
291 return 1;
Joerg Roedel852e3c12009-07-27 16:30:44 +0200292 if (is_large_pte(pte))
Marcelo Tosatti776e6632009-06-10 12:27:03 -0300293 return 1;
294 return 0;
295}
296
Anthony Liguori35149e22008-04-02 14:46:56 -0500297static pfn_t spte_to_pfn(u64 pte)
Avi Kivity0b49ea82008-03-23 15:06:23 +0200298{
Anthony Liguori35149e22008-04-02 14:46:56 -0500299 return (pte & PT64_BASE_ADDR_MASK) >> PAGE_SHIFT;
Avi Kivity0b49ea82008-03-23 15:06:23 +0200300}
301
Avi Kivityda928522007-11-21 13:54:47 +0200302static gfn_t pse36_gfn_delta(u32 gpte)
303{
304 int shift = 32 - PT32_DIR_PSE36_SHIFT - PAGE_SHIFT;
305
306 return (gpte & PT32_DIR_PSE36_MASK) << shift;
307}
308
Xiao Guangrong603e0652011-07-12 03:31:28 +0800309#ifdef CONFIG_X86_64
Avi Kivityd555c332009-06-10 14:24:23 +0300310static void __set_spte(u64 *sptep, u64 spte)
Avi Kivitye663ee62007-05-31 15:46:04 +0300311{
Xiao Guangrong603e0652011-07-12 03:31:28 +0800312 *sptep = spte;
Avi Kivitye663ee62007-05-31 15:46:04 +0300313}
314
Xiao Guangrong603e0652011-07-12 03:31:28 +0800315static void __update_clear_spte_fast(u64 *sptep, u64 spte)
Avi Kivitya9221dd2010-06-06 14:48:06 +0300316{
Xiao Guangrong603e0652011-07-12 03:31:28 +0800317 *sptep = spte;
Avi Kivitya9221dd2010-06-06 14:48:06 +0300318}
319
Xiao Guangrong603e0652011-07-12 03:31:28 +0800320static u64 __update_clear_spte_slow(u64 *sptep, u64 spte)
321{
322 return xchg(sptep, spte);
323}
Xiao Guangrongc2a2ac22011-07-12 03:32:13 +0800324
325static u64 __get_spte_lockless(u64 *sptep)
326{
327 return ACCESS_ONCE(*sptep);
328}
Xiao Guangrongce88dec2011-07-12 03:33:44 +0800329
330static bool __check_direct_spte_mmio_pf(u64 spte)
331{
332 /* It is valid if the spte is zapped. */
333 return spte == 0ull;
334}
Xiao Guangrong603e0652011-07-12 03:31:28 +0800335#else
336union split_spte {
337 struct {
338 u32 spte_low;
339 u32 spte_high;
340 };
341 u64 spte;
342};
343
Xiao Guangrongc2a2ac22011-07-12 03:32:13 +0800344static void count_spte_clear(u64 *sptep, u64 spte)
345{
346 struct kvm_mmu_page *sp = page_header(__pa(sptep));
347
348 if (is_shadow_present_pte(spte))
349 return;
350
351 /* Ensure the spte is completely set before we increase the count */
352 smp_wmb();
353 sp->clear_spte_count++;
354}
355
Xiao Guangrong603e0652011-07-12 03:31:28 +0800356static void __set_spte(u64 *sptep, u64 spte)
357{
358 union split_spte *ssptep, sspte;
359
360 ssptep = (union split_spte *)sptep;
361 sspte = (union split_spte)spte;
362
363 ssptep->spte_high = sspte.spte_high;
364
365 /*
366 * If we map the spte from nonpresent to present, We should store
367 * the high bits firstly, then set present bit, so cpu can not
368 * fetch this spte while we are setting the spte.
369 */
370 smp_wmb();
371
372 ssptep->spte_low = sspte.spte_low;
373}
374
375static void __update_clear_spte_fast(u64 *sptep, u64 spte)
376{
377 union split_spte *ssptep, sspte;
378
379 ssptep = (union split_spte *)sptep;
380 sspte = (union split_spte)spte;
381
382 ssptep->spte_low = sspte.spte_low;
383
384 /*
385 * If we map the spte from present to nonpresent, we should clear
386 * present bit firstly to avoid vcpu fetch the old high bits.
387 */
388 smp_wmb();
389
390 ssptep->spte_high = sspte.spte_high;
Xiao Guangrongc2a2ac22011-07-12 03:32:13 +0800391 count_spte_clear(sptep, spte);
Xiao Guangrong603e0652011-07-12 03:31:28 +0800392}
393
394static u64 __update_clear_spte_slow(u64 *sptep, u64 spte)
395{
396 union split_spte *ssptep, sspte, orig;
397
398 ssptep = (union split_spte *)sptep;
399 sspte = (union split_spte)spte;
400
401 /* xchg acts as a barrier before the setting of the high bits */
402 orig.spte_low = xchg(&ssptep->spte_low, sspte.spte_low);
Zhao Jin41bc3182011-09-19 12:19:51 +0800403 orig.spte_high = ssptep->spte_high;
404 ssptep->spte_high = sspte.spte_high;
Xiao Guangrongc2a2ac22011-07-12 03:32:13 +0800405 count_spte_clear(sptep, spte);
Xiao Guangrong603e0652011-07-12 03:31:28 +0800406
407 return orig.spte;
408}
Xiao Guangrongc2a2ac22011-07-12 03:32:13 +0800409
410/*
411 * The idea using the light way get the spte on x86_32 guest is from
412 * gup_get_pte(arch/x86/mm/gup.c).
413 * The difference is we can not catch the spte tlb flush if we leave
414 * guest mode, so we emulate it by increase clear_spte_count when spte
415 * is cleared.
416 */
417static u64 __get_spte_lockless(u64 *sptep)
418{
419 struct kvm_mmu_page *sp = page_header(__pa(sptep));
420 union split_spte spte, *orig = (union split_spte *)sptep;
421 int count;
422
423retry:
424 count = sp->clear_spte_count;
425 smp_rmb();
426
427 spte.spte_low = orig->spte_low;
428 smp_rmb();
429
430 spte.spte_high = orig->spte_high;
431 smp_rmb();
432
433 if (unlikely(spte.spte_low != orig->spte_low ||
434 count != sp->clear_spte_count))
435 goto retry;
436
437 return spte.spte;
438}
Xiao Guangrongce88dec2011-07-12 03:33:44 +0800439
440static bool __check_direct_spte_mmio_pf(u64 spte)
441{
442 union split_spte sspte = (union split_spte)spte;
443 u32 high_mmio_mask = shadow_mmio_mask >> 32;
444
445 /* It is valid if the spte is zapped. */
446 if (spte == 0ull)
447 return true;
448
449 /* It is valid if the spte is being zapped. */
450 if (sspte.spte_low == 0ull &&
451 (sspte.spte_high & high_mmio_mask) == high_mmio_mask)
452 return true;
453
454 return false;
455}
Xiao Guangrong603e0652011-07-12 03:31:28 +0800456#endif
457
Xiao Guangrong8672b722010-08-02 16:14:04 +0800458static bool spte_has_volatile_bits(u64 spte)
459{
460 if (!shadow_accessed_mask)
461 return false;
462
463 if (!is_shadow_present_pte(spte))
464 return false;
465
Xiao Guangrong41327792010-08-02 16:15:08 +0800466 if ((spte & shadow_accessed_mask) &&
467 (!is_writable_pte(spte) || (spte & shadow_dirty_mask)))
Xiao Guangrong8672b722010-08-02 16:14:04 +0800468 return false;
469
470 return true;
471}
472
Xiao Guangrong41327792010-08-02 16:15:08 +0800473static bool spte_is_bit_cleared(u64 old_spte, u64 new_spte, u64 bit_mask)
474{
475 return (old_spte & bit_mask) && !(new_spte & bit_mask);
476}
477
Xiao Guangrong1df9f2d2011-07-12 03:30:35 +0800478/* Rules for using mmu_spte_set:
479 * Set the sptep from nonpresent to present.
480 * Note: the sptep being assigned *must* be either not present
481 * or in a state where the hardware will not attempt to update
482 * the spte.
483 */
484static void mmu_spte_set(u64 *sptep, u64 new_spte)
485{
486 WARN_ON(is_shadow_present_pte(*sptep));
487 __set_spte(sptep, new_spte);
488}
489
490/* Rules for using mmu_spte_update:
491 * Update the state bits, it means the mapped pfn is not changged.
492 */
493static void mmu_spte_update(u64 *sptep, u64 new_spte)
Avi Kivityb79b93f2010-06-06 15:46:44 +0300494{
Xiao Guangrong41327792010-08-02 16:15:08 +0800495 u64 mask, old_spte = *sptep;
Avi Kivityb79b93f2010-06-06 15:46:44 +0300496
Xiao Guangrong41327792010-08-02 16:15:08 +0800497 WARN_ON(!is_rmap_spte(new_spte));
498
Xiao Guangrong1df9f2d2011-07-12 03:30:35 +0800499 if (!is_shadow_present_pte(old_spte))
500 return mmu_spte_set(sptep, new_spte);
501
Xiao Guangrong41327792010-08-02 16:15:08 +0800502 new_spte |= old_spte & shadow_dirty_mask;
503
504 mask = shadow_accessed_mask;
505 if (is_writable_pte(old_spte))
506 mask |= shadow_dirty_mask;
507
508 if (!spte_has_volatile_bits(old_spte) || (new_spte & mask) == mask)
Xiao Guangrong603e0652011-07-12 03:31:28 +0800509 __update_clear_spte_fast(sptep, new_spte);
Xiao Guangrong41327792010-08-02 16:15:08 +0800510 else
Xiao Guangrong603e0652011-07-12 03:31:28 +0800511 old_spte = __update_clear_spte_slow(sptep, new_spte);
Xiao Guangrong41327792010-08-02 16:15:08 +0800512
513 if (!shadow_accessed_mask)
514 return;
515
516 if (spte_is_bit_cleared(old_spte, new_spte, shadow_accessed_mask))
517 kvm_set_pfn_accessed(spte_to_pfn(old_spte));
518 if (spte_is_bit_cleared(old_spte, new_spte, shadow_dirty_mask))
519 kvm_set_pfn_dirty(spte_to_pfn(old_spte));
Avi Kivityb79b93f2010-06-06 15:46:44 +0300520}
521
Xiao Guangrong1df9f2d2011-07-12 03:30:35 +0800522/*
523 * Rules for using mmu_spte_clear_track_bits:
524 * It sets the sptep from present to nonpresent, and track the
525 * state bits, it is used to clear the last level sptep.
526 */
527static int mmu_spte_clear_track_bits(u64 *sptep)
528{
529 pfn_t pfn;
530 u64 old_spte = *sptep;
531
532 if (!spte_has_volatile_bits(old_spte))
Xiao Guangrong603e0652011-07-12 03:31:28 +0800533 __update_clear_spte_fast(sptep, 0ull);
Xiao Guangrong1df9f2d2011-07-12 03:30:35 +0800534 else
Xiao Guangrong603e0652011-07-12 03:31:28 +0800535 old_spte = __update_clear_spte_slow(sptep, 0ull);
Xiao Guangrong1df9f2d2011-07-12 03:30:35 +0800536
537 if (!is_rmap_spte(old_spte))
538 return 0;
539
540 pfn = spte_to_pfn(old_spte);
541 if (!shadow_accessed_mask || old_spte & shadow_accessed_mask)
542 kvm_set_pfn_accessed(pfn);
543 if (!shadow_dirty_mask || (old_spte & shadow_dirty_mask))
544 kvm_set_pfn_dirty(pfn);
545 return 1;
546}
547
548/*
549 * Rules for using mmu_spte_clear_no_track:
550 * Directly clear spte without caring the state bits of sptep,
551 * it is used to set the upper level spte.
552 */
553static void mmu_spte_clear_no_track(u64 *sptep)
554{
Xiao Guangrong603e0652011-07-12 03:31:28 +0800555 __update_clear_spte_fast(sptep, 0ull);
Xiao Guangrong1df9f2d2011-07-12 03:30:35 +0800556}
557
Xiao Guangrongc2a2ac22011-07-12 03:32:13 +0800558static u64 mmu_spte_get_lockless(u64 *sptep)
559{
560 return __get_spte_lockless(sptep);
561}
562
563static void walk_shadow_page_lockless_begin(struct kvm_vcpu *vcpu)
564{
565 rcu_read_lock();
566 atomic_inc(&vcpu->kvm->arch.reader_counter);
567
568 /* Increase the counter before walking shadow page table */
569 smp_mb__after_atomic_inc();
570}
571
572static void walk_shadow_page_lockless_end(struct kvm_vcpu *vcpu)
573{
574 /* Decrease the counter after walking shadow page table finished */
575 smp_mb__before_atomic_dec();
576 atomic_dec(&vcpu->kvm->arch.reader_counter);
577 rcu_read_unlock();
578}
579
Avi Kivitye2dec932007-01-05 16:36:54 -0800580static int mmu_topup_memory_cache(struct kvm_mmu_memory_cache *cache,
Avi Kivity2e3e5882007-09-10 11:28:17 +0300581 struct kmem_cache *base_cache, int min)
Avi Kivity714b93d2007-01-05 16:36:53 -0800582{
583 void *obj;
584
585 if (cache->nobjs >= min)
Avi Kivitye2dec932007-01-05 16:36:54 -0800586 return 0;
Avi Kivity714b93d2007-01-05 16:36:53 -0800587 while (cache->nobjs < ARRAY_SIZE(cache->objects)) {
Avi Kivity2e3e5882007-09-10 11:28:17 +0300588 obj = kmem_cache_zalloc(base_cache, GFP_KERNEL);
Avi Kivity714b93d2007-01-05 16:36:53 -0800589 if (!obj)
Avi Kivitye2dec932007-01-05 16:36:54 -0800590 return -ENOMEM;
Avi Kivity714b93d2007-01-05 16:36:53 -0800591 cache->objects[cache->nobjs++] = obj;
592 }
Avi Kivitye2dec932007-01-05 16:36:54 -0800593 return 0;
Avi Kivity714b93d2007-01-05 16:36:53 -0800594}
595
Xiao Guangrongf759e2b2011-09-22 16:53:17 +0800596static int mmu_memory_cache_free_objects(struct kvm_mmu_memory_cache *cache)
597{
598 return cache->nobjs;
599}
600
Xiao Guangronge8ad9a72010-05-13 10:06:02 +0800601static void mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc,
602 struct kmem_cache *cache)
Avi Kivity714b93d2007-01-05 16:36:53 -0800603{
604 while (mc->nobjs)
Xiao Guangronge8ad9a72010-05-13 10:06:02 +0800605 kmem_cache_free(cache, mc->objects[--mc->nobjs]);
Avi Kivity714b93d2007-01-05 16:36:53 -0800606}
607
Avi Kivityc1158e62007-07-20 08:18:27 +0300608static int mmu_topup_memory_cache_page(struct kvm_mmu_memory_cache *cache,
Avi Kivity2e3e5882007-09-10 11:28:17 +0300609 int min)
Avi Kivityc1158e62007-07-20 08:18:27 +0300610{
Xiao Guangrong842f22e2011-03-04 19:01:10 +0800611 void *page;
Avi Kivityc1158e62007-07-20 08:18:27 +0300612
613 if (cache->nobjs >= min)
614 return 0;
615 while (cache->nobjs < ARRAY_SIZE(cache->objects)) {
Xiao Guangrong842f22e2011-03-04 19:01:10 +0800616 page = (void *)__get_free_page(GFP_KERNEL);
Avi Kivityc1158e62007-07-20 08:18:27 +0300617 if (!page)
618 return -ENOMEM;
Xiao Guangrong842f22e2011-03-04 19:01:10 +0800619 cache->objects[cache->nobjs++] = page;
Avi Kivityc1158e62007-07-20 08:18:27 +0300620 }
621 return 0;
622}
623
624static void mmu_free_memory_cache_page(struct kvm_mmu_memory_cache *mc)
625{
626 while (mc->nobjs)
Avi Kivityc4d198d2007-07-21 09:06:46 +0300627 free_page((unsigned long)mc->objects[--mc->nobjs]);
Avi Kivityc1158e62007-07-20 08:18:27 +0300628}
629
Avi Kivity8c438502007-04-16 11:53:17 +0300630static int mmu_topup_memory_caches(struct kvm_vcpu *vcpu)
631{
632 int r;
633
Xiao Guangrong53c07b12011-05-15 23:26:20 +0800634 r = mmu_topup_memory_cache(&vcpu->arch.mmu_pte_list_desc_cache,
Xiao Guangrong67052b32011-05-15 23:27:08 +0800635 pte_list_desc_cache, 8 + PTE_PREFETCH_NUM);
Avi Kivity2e3e5882007-09-10 11:28:17 +0300636 if (r)
637 goto out;
Zhang Xiantaoad312c72007-12-13 23:50:52 +0800638 r = mmu_topup_memory_cache_page(&vcpu->arch.mmu_page_cache, 8);
Avi Kivity2e3e5882007-09-10 11:28:17 +0300639 if (r)
640 goto out;
Zhang Xiantaoad312c72007-12-13 23:50:52 +0800641 r = mmu_topup_memory_cache(&vcpu->arch.mmu_page_header_cache,
Avi Kivity2e3e5882007-09-10 11:28:17 +0300642 mmu_page_header_cache, 4);
643out:
Avi Kivity8c438502007-04-16 11:53:17 +0300644 return r;
645}
646
Avi Kivity714b93d2007-01-05 16:36:53 -0800647static void mmu_free_memory_caches(struct kvm_vcpu *vcpu)
648{
Xiao Guangrong53c07b12011-05-15 23:26:20 +0800649 mmu_free_memory_cache(&vcpu->arch.mmu_pte_list_desc_cache,
650 pte_list_desc_cache);
Zhang Xiantaoad312c72007-12-13 23:50:52 +0800651 mmu_free_memory_cache_page(&vcpu->arch.mmu_page_cache);
Xiao Guangronge8ad9a72010-05-13 10:06:02 +0800652 mmu_free_memory_cache(&vcpu->arch.mmu_page_header_cache,
653 mmu_page_header_cache);
Avi Kivity714b93d2007-01-05 16:36:53 -0800654}
655
656static void *mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc,
657 size_t size)
658{
659 void *p;
660
661 BUG_ON(!mc->nobjs);
662 p = mc->objects[--mc->nobjs];
Avi Kivity714b93d2007-01-05 16:36:53 -0800663 return p;
664}
665
Xiao Guangrong53c07b12011-05-15 23:26:20 +0800666static struct pte_list_desc *mmu_alloc_pte_list_desc(struct kvm_vcpu *vcpu)
Avi Kivity714b93d2007-01-05 16:36:53 -0800667{
Xiao Guangrong53c07b12011-05-15 23:26:20 +0800668 return mmu_memory_cache_alloc(&vcpu->arch.mmu_pte_list_desc_cache,
669 sizeof(struct pte_list_desc));
Avi Kivity714b93d2007-01-05 16:36:53 -0800670}
671
Xiao Guangrong53c07b12011-05-15 23:26:20 +0800672static void mmu_free_pte_list_desc(struct pte_list_desc *pte_list_desc)
Avi Kivity714b93d2007-01-05 16:36:53 -0800673{
Xiao Guangrong53c07b12011-05-15 23:26:20 +0800674 kmem_cache_free(pte_list_desc_cache, pte_list_desc);
Avi Kivity714b93d2007-01-05 16:36:53 -0800675}
676
Lai Jiangshan2032a932010-05-26 16:49:59 +0800677static gfn_t kvm_mmu_page_get_gfn(struct kvm_mmu_page *sp, int index)
678{
679 if (!sp->role.direct)
680 return sp->gfns[index];
681
682 return sp->gfn + (index << ((sp->role.level - 1) * PT64_LEVEL_BITS));
683}
684
685static void kvm_mmu_page_set_gfn(struct kvm_mmu_page *sp, int index, gfn_t gfn)
686{
687 if (sp->role.direct)
688 BUG_ON(gfn != kvm_mmu_page_get_gfn(sp, index));
689 else
690 sp->gfns[index] = gfn;
691}
692
Avi Kivitycd4a4e52007-01-05 16:36:38 -0800693/*
Takuya Yoshikawad4dbf472010-12-07 12:59:07 +0900694 * Return the pointer to the large page information for a given gfn,
695 * handling slots that are not large page aligned.
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300696 */
Takuya Yoshikawad4dbf472010-12-07 12:59:07 +0900697static struct kvm_lpage_info *lpage_info_slot(gfn_t gfn,
698 struct kvm_memory_slot *slot,
699 int level)
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300700{
701 unsigned long idx;
702
Joerg Roedel82855412010-07-01 16:00:11 +0200703 idx = (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
704 (slot->base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
Takuya Yoshikawad4dbf472010-12-07 12:59:07 +0900705 return &slot->lpage_info[level - 2][idx];
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300706}
707
708static void account_shadowed(struct kvm *kvm, gfn_t gfn)
709{
Joerg Roedeld25797b2009-07-27 16:30:43 +0200710 struct kvm_memory_slot *slot;
Takuya Yoshikawad4dbf472010-12-07 12:59:07 +0900711 struct kvm_lpage_info *linfo;
Joerg Roedeld25797b2009-07-27 16:30:43 +0200712 int i;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300713
Avi Kivitya1f4d3952010-06-21 11:44:20 +0300714 slot = gfn_to_memslot(kvm, gfn);
Joerg Roedeld25797b2009-07-27 16:30:43 +0200715 for (i = PT_DIRECTORY_LEVEL;
716 i < PT_PAGE_TABLE_LEVEL + KVM_NR_PAGE_SIZES; ++i) {
Takuya Yoshikawad4dbf472010-12-07 12:59:07 +0900717 linfo = lpage_info_slot(gfn, slot, i);
718 linfo->write_count += 1;
Joerg Roedeld25797b2009-07-27 16:30:43 +0200719 }
Xiao Guangrong332b2072011-05-15 23:20:27 +0800720 kvm->arch.indirect_shadow_pages++;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300721}
722
723static void unaccount_shadowed(struct kvm *kvm, gfn_t gfn)
724{
Joerg Roedeld25797b2009-07-27 16:30:43 +0200725 struct kvm_memory_slot *slot;
Takuya Yoshikawad4dbf472010-12-07 12:59:07 +0900726 struct kvm_lpage_info *linfo;
Joerg Roedeld25797b2009-07-27 16:30:43 +0200727 int i;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300728
Avi Kivitya1f4d3952010-06-21 11:44:20 +0300729 slot = gfn_to_memslot(kvm, gfn);
Joerg Roedeld25797b2009-07-27 16:30:43 +0200730 for (i = PT_DIRECTORY_LEVEL;
731 i < PT_PAGE_TABLE_LEVEL + KVM_NR_PAGE_SIZES; ++i) {
Takuya Yoshikawad4dbf472010-12-07 12:59:07 +0900732 linfo = lpage_info_slot(gfn, slot, i);
733 linfo->write_count -= 1;
734 WARN_ON(linfo->write_count < 0);
Joerg Roedeld25797b2009-07-27 16:30:43 +0200735 }
Xiao Guangrong332b2072011-05-15 23:20:27 +0800736 kvm->arch.indirect_shadow_pages--;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300737}
738
Joerg Roedeld25797b2009-07-27 16:30:43 +0200739static int has_wrprotected_page(struct kvm *kvm,
740 gfn_t gfn,
741 int level)
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300742{
Izik Eidus28430992008-10-03 17:40:32 +0300743 struct kvm_memory_slot *slot;
Takuya Yoshikawad4dbf472010-12-07 12:59:07 +0900744 struct kvm_lpage_info *linfo;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300745
Avi Kivitya1f4d3952010-06-21 11:44:20 +0300746 slot = gfn_to_memslot(kvm, gfn);
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300747 if (slot) {
Takuya Yoshikawad4dbf472010-12-07 12:59:07 +0900748 linfo = lpage_info_slot(gfn, slot, level);
749 return linfo->write_count;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300750 }
751
752 return 1;
753}
754
Joerg Roedeld25797b2009-07-27 16:30:43 +0200755static int host_mapping_level(struct kvm *kvm, gfn_t gfn)
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300756{
Joerg Roedel8f0b1ab2010-01-28 12:37:56 +0100757 unsigned long page_size;
Joerg Roedeld25797b2009-07-27 16:30:43 +0200758 int i, ret = 0;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300759
Joerg Roedel8f0b1ab2010-01-28 12:37:56 +0100760 page_size = kvm_host_page_size(kvm, gfn);
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300761
Joerg Roedeld25797b2009-07-27 16:30:43 +0200762 for (i = PT_PAGE_TABLE_LEVEL;
763 i < (PT_PAGE_TABLE_LEVEL + KVM_NR_PAGE_SIZES); ++i) {
764 if (page_size >= KVM_HPAGE_SIZE(i))
765 ret = i;
766 else
767 break;
768 }
769
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -0300770 return ret;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300771}
772
Xiao Guangrong5d163b12011-03-09 15:43:00 +0800773static struct kvm_memory_slot *
774gfn_to_memslot_dirty_bitmap(struct kvm_vcpu *vcpu, gfn_t gfn,
775 bool no_dirty_log)
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300776{
777 struct kvm_memory_slot *slot;
Xiao Guangrong5d163b12011-03-09 15:43:00 +0800778
779 slot = gfn_to_memslot(vcpu->kvm, gfn);
780 if (!slot || slot->flags & KVM_MEMSLOT_INVALID ||
781 (no_dirty_log && slot->dirty_bitmap))
782 slot = NULL;
783
784 return slot;
785}
786
787static bool mapping_level_dirty_bitmap(struct kvm_vcpu *vcpu, gfn_t large_gfn)
788{
Stevea0a8eab2011-06-17 10:25:39 +0800789 return !gfn_to_memslot_dirty_bitmap(vcpu, large_gfn, true);
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800790}
791
792static int mapping_level(struct kvm_vcpu *vcpu, gfn_t large_gfn)
793{
794 int host_level, level, max_level;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300795
Joerg Roedeld25797b2009-07-27 16:30:43 +0200796 host_level = host_mapping_level(vcpu->kvm, large_gfn);
797
798 if (host_level == PT_PAGE_TABLE_LEVEL)
799 return host_level;
800
Sheng Yang878403b2010-01-05 19:02:29 +0800801 max_level = kvm_x86_ops->get_lpage_level() < host_level ?
802 kvm_x86_ops->get_lpage_level() : host_level;
803
804 for (level = PT_DIRECTORY_LEVEL; level <= max_level; ++level)
Joerg Roedeld25797b2009-07-27 16:30:43 +0200805 if (has_wrprotected_page(vcpu->kvm, large_gfn, level))
806 break;
Joerg Roedeld25797b2009-07-27 16:30:43 +0200807
808 return level - 1;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300809}
810
811/*
Xiao Guangrong53c07b12011-05-15 23:26:20 +0800812 * Pte mapping structures:
813 *
814 * If pte_list bit zero is zero, then pte_list point to the spte.
815 *
816 * If pte_list bit zero is one, (then pte_list & ~1) points to a struct
817 * pte_list_desc containing more mappings.
818 *
819 * Returns the number of pte entries before the spte was added or zero if
820 * the spte was not added.
821 *
822 */
823static int pte_list_add(struct kvm_vcpu *vcpu, u64 *spte,
824 unsigned long *pte_list)
825{
826 struct pte_list_desc *desc;
827 int i, count = 0;
828
829 if (!*pte_list) {
830 rmap_printk("pte_list_add: %p %llx 0->1\n", spte, *spte);
831 *pte_list = (unsigned long)spte;
832 } else if (!(*pte_list & 1)) {
833 rmap_printk("pte_list_add: %p %llx 1->many\n", spte, *spte);
834 desc = mmu_alloc_pte_list_desc(vcpu);
835 desc->sptes[0] = (u64 *)*pte_list;
836 desc->sptes[1] = spte;
837 *pte_list = (unsigned long)desc | 1;
838 ++count;
839 } else {
840 rmap_printk("pte_list_add: %p %llx many->many\n", spte, *spte);
841 desc = (struct pte_list_desc *)(*pte_list & ~1ul);
842 while (desc->sptes[PTE_LIST_EXT-1] && desc->more) {
843 desc = desc->more;
844 count += PTE_LIST_EXT;
845 }
846 if (desc->sptes[PTE_LIST_EXT-1]) {
847 desc->more = mmu_alloc_pte_list_desc(vcpu);
848 desc = desc->more;
849 }
850 for (i = 0; desc->sptes[i]; ++i)
851 ++count;
852 desc->sptes[i] = spte;
853 }
854 return count;
855}
856
857static u64 *pte_list_next(unsigned long *pte_list, u64 *spte)
858{
859 struct pte_list_desc *desc;
860 u64 *prev_spte;
861 int i;
862
863 if (!*pte_list)
864 return NULL;
865 else if (!(*pte_list & 1)) {
866 if (!spte)
867 return (u64 *)*pte_list;
868 return NULL;
869 }
870 desc = (struct pte_list_desc *)(*pte_list & ~1ul);
871 prev_spte = NULL;
872 while (desc) {
873 for (i = 0; i < PTE_LIST_EXT && desc->sptes[i]; ++i) {
874 if (prev_spte == spte)
875 return desc->sptes[i];
876 prev_spte = desc->sptes[i];
877 }
878 desc = desc->more;
879 }
880 return NULL;
881}
882
883static void
884pte_list_desc_remove_entry(unsigned long *pte_list, struct pte_list_desc *desc,
885 int i, struct pte_list_desc *prev_desc)
886{
887 int j;
888
889 for (j = PTE_LIST_EXT - 1; !desc->sptes[j] && j > i; --j)
890 ;
891 desc->sptes[i] = desc->sptes[j];
892 desc->sptes[j] = NULL;
893 if (j != 0)
894 return;
895 if (!prev_desc && !desc->more)
896 *pte_list = (unsigned long)desc->sptes[0];
897 else
898 if (prev_desc)
899 prev_desc->more = desc->more;
900 else
901 *pte_list = (unsigned long)desc->more | 1;
902 mmu_free_pte_list_desc(desc);
903}
904
905static void pte_list_remove(u64 *spte, unsigned long *pte_list)
906{
907 struct pte_list_desc *desc;
908 struct pte_list_desc *prev_desc;
909 int i;
910
911 if (!*pte_list) {
912 printk(KERN_ERR "pte_list_remove: %p 0->BUG\n", spte);
913 BUG();
914 } else if (!(*pte_list & 1)) {
915 rmap_printk("pte_list_remove: %p 1->0\n", spte);
916 if ((u64 *)*pte_list != spte) {
917 printk(KERN_ERR "pte_list_remove: %p 1->BUG\n", spte);
918 BUG();
919 }
920 *pte_list = 0;
921 } else {
922 rmap_printk("pte_list_remove: %p many->many\n", spte);
923 desc = (struct pte_list_desc *)(*pte_list & ~1ul);
924 prev_desc = NULL;
925 while (desc) {
926 for (i = 0; i < PTE_LIST_EXT && desc->sptes[i]; ++i)
927 if (desc->sptes[i] == spte) {
928 pte_list_desc_remove_entry(pte_list,
929 desc, i,
930 prev_desc);
931 return;
932 }
933 prev_desc = desc;
934 desc = desc->more;
935 }
936 pr_err("pte_list_remove: %p many->many\n", spte);
937 BUG();
938 }
939}
940
Xiao Guangrong67052b32011-05-15 23:27:08 +0800941typedef void (*pte_list_walk_fn) (u64 *spte);
942static void pte_list_walk(unsigned long *pte_list, pte_list_walk_fn fn)
943{
944 struct pte_list_desc *desc;
945 int i;
946
947 if (!*pte_list)
948 return;
949
950 if (!(*pte_list & 1))
951 return fn((u64 *)*pte_list);
952
953 desc = (struct pte_list_desc *)(*pte_list & ~1ul);
954 while (desc) {
955 for (i = 0; i < PTE_LIST_EXT && desc->sptes[i]; ++i)
956 fn(desc->sptes[i]);
957 desc = desc->more;
958 }
959}
960
Takuya Yoshikawa9b9b1492011-11-14 18:22:28 +0900961static unsigned long *__gfn_to_rmap(struct kvm *kvm, gfn_t gfn, int level,
962 struct kvm_memory_slot *slot)
963{
964 struct kvm_lpage_info *linfo;
965
966 if (likely(level == PT_PAGE_TABLE_LEVEL))
967 return &slot->rmap[gfn - slot->base_gfn];
968
969 linfo = lpage_info_slot(gfn, slot, level);
970 return &linfo->rmap_pde;
971}
972
Xiao Guangrong53c07b12011-05-15 23:26:20 +0800973/*
Izik Eidus290fc382007-09-27 14:11:22 +0200974 * Take gfn and return the reverse mapping to it.
Izik Eidus290fc382007-09-27 14:11:22 +0200975 */
Joerg Roedel44ad9942009-07-27 16:30:42 +0200976static unsigned long *gfn_to_rmap(struct kvm *kvm, gfn_t gfn, int level)
Izik Eidus290fc382007-09-27 14:11:22 +0200977{
978 struct kvm_memory_slot *slot;
979
980 slot = gfn_to_memslot(kvm, gfn);
Takuya Yoshikawa9b9b1492011-11-14 18:22:28 +0900981 return __gfn_to_rmap(kvm, gfn, level, slot);
Izik Eidus290fc382007-09-27 14:11:22 +0200982}
983
Xiao Guangrongf759e2b2011-09-22 16:53:17 +0800984static bool rmap_can_add(struct kvm_vcpu *vcpu)
985{
986 struct kvm_mmu_memory_cache *cache;
987
988 cache = &vcpu->arch.mmu_pte_list_desc_cache;
989 return mmu_memory_cache_free_objects(cache);
990}
991
Joerg Roedel44ad9942009-07-27 16:30:42 +0200992static int rmap_add(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn)
Avi Kivitycd4a4e52007-01-05 16:36:38 -0800993{
Avi Kivity4db35312007-11-21 15:28:32 +0200994 struct kvm_mmu_page *sp;
Izik Eidus290fc382007-09-27 14:11:22 +0200995 unsigned long *rmapp;
Avi Kivitycd4a4e52007-01-05 16:36:38 -0800996
Avi Kivity4db35312007-11-21 15:28:32 +0200997 sp = page_header(__pa(spte));
Lai Jiangshan2032a932010-05-26 16:49:59 +0800998 kvm_mmu_page_set_gfn(sp, spte - sp->spt, gfn);
Joerg Roedel44ad9942009-07-27 16:30:42 +0200999 rmapp = gfn_to_rmap(vcpu->kvm, gfn, sp->role.level);
Xiao Guangrong53c07b12011-05-15 23:26:20 +08001000 return pte_list_add(vcpu, spte, rmapp);
Avi Kivitycd4a4e52007-01-05 16:36:38 -08001001}
1002
Xiao Guangrong53c07b12011-05-15 23:26:20 +08001003static u64 *rmap_next(struct kvm *kvm, unsigned long *rmapp, u64 *spte)
Avi Kivitycd4a4e52007-01-05 16:36:38 -08001004{
Xiao Guangrong53c07b12011-05-15 23:26:20 +08001005 return pte_list_next(rmapp, spte);
Avi Kivitycd4a4e52007-01-05 16:36:38 -08001006}
1007
Izik Eidus290fc382007-09-27 14:11:22 +02001008static void rmap_remove(struct kvm *kvm, u64 *spte)
Avi Kivitycd4a4e52007-01-05 16:36:38 -08001009{
Avi Kivity4db35312007-11-21 15:28:32 +02001010 struct kvm_mmu_page *sp;
Lai Jiangshan2032a932010-05-26 16:49:59 +08001011 gfn_t gfn;
Izik Eidus290fc382007-09-27 14:11:22 +02001012 unsigned long *rmapp;
Avi Kivitycd4a4e52007-01-05 16:36:38 -08001013
Avi Kivity4db35312007-11-21 15:28:32 +02001014 sp = page_header(__pa(spte));
Lai Jiangshan2032a932010-05-26 16:49:59 +08001015 gfn = kvm_mmu_page_get_gfn(sp, spte - sp->spt);
1016 rmapp = gfn_to_rmap(kvm, gfn, sp->role.level);
Xiao Guangrong53c07b12011-05-15 23:26:20 +08001017 pte_list_remove(spte, rmapp);
Avi Kivitycd4a4e52007-01-05 16:36:38 -08001018}
1019
Xiao Guangrongc3707952011-07-12 03:28:04 +08001020static void drop_spte(struct kvm *kvm, u64 *sptep)
Xiao Guangronge4b502e2010-07-16 11:28:09 +08001021{
Xiao Guangrong1df9f2d2011-07-12 03:30:35 +08001022 if (mmu_spte_clear_track_bits(sptep))
Marcelo Tosattieb45fda2010-10-25 11:58:22 -02001023 rmap_remove(kvm, sptep);
Avi Kivitybe38d272010-06-06 14:31:27 +03001024}
1025
Marcelo Tosattib1a36822008-12-01 22:32:03 -02001026static int rmap_write_protect(struct kvm *kvm, u64 gfn)
Izik Eidus98348e92007-10-16 14:42:30 +02001027{
Takuya Yoshikawa9b9b1492011-11-14 18:22:28 +09001028 struct kvm_memory_slot *slot;
Izik Eidus290fc382007-09-27 14:11:22 +02001029 unsigned long *rmapp;
Avi Kivity374cbac2007-01-05 16:36:43 -08001030 u64 *spte;
Joerg Roedel44ad9942009-07-27 16:30:42 +02001031 int i, write_protected = 0;
Avi Kivity374cbac2007-01-05 16:36:43 -08001032
Takuya Yoshikawa9b9b1492011-11-14 18:22:28 +09001033 slot = gfn_to_memslot(kvm, gfn);
Avi Kivity374cbac2007-01-05 16:36:43 -08001034
Takuya Yoshikawa9b9b1492011-11-14 18:22:28 +09001035 rmapp = __gfn_to_rmap(kvm, gfn, PT_PAGE_TABLE_LEVEL, slot);
Izik Eidus98348e92007-10-16 14:42:30 +02001036 spte = rmap_next(kvm, rmapp, NULL);
1037 while (spte) {
Avi Kivity374cbac2007-01-05 16:36:43 -08001038 BUG_ON(!(*spte & PT_PRESENT_MASK));
Avi Kivity374cbac2007-01-05 16:36:43 -08001039 rmap_printk("rmap_write_protect: spte %p %llx\n", spte, *spte);
Takuya Yoshikawa8dae4442010-01-18 18:45:10 +09001040 if (is_writable_pte(*spte)) {
Xiao Guangrong1df9f2d2011-07-12 03:30:35 +08001041 mmu_spte_update(spte, *spte & ~PT_WRITABLE_MASK);
Eddie Dongcaa5b8a2007-12-18 06:08:27 +08001042 write_protected = 1;
1043 }
Izik Eidus9647c142007-10-16 14:43:46 +02001044 spte = rmap_next(kvm, rmapp, spte);
Avi Kivity374cbac2007-01-05 16:36:43 -08001045 }
Izik Eidus855149a2008-03-20 18:17:24 +02001046
Marcelo Tosatti05da4552008-02-23 11:44:30 -03001047 /* check for huge page mappings */
Joerg Roedel44ad9942009-07-27 16:30:42 +02001048 for (i = PT_DIRECTORY_LEVEL;
1049 i < PT_PAGE_TABLE_LEVEL + KVM_NR_PAGE_SIZES; ++i) {
Takuya Yoshikawa9b9b1492011-11-14 18:22:28 +09001050 rmapp = __gfn_to_rmap(kvm, gfn, i, slot);
Joerg Roedel44ad9942009-07-27 16:30:42 +02001051 spte = rmap_next(kvm, rmapp, NULL);
1052 while (spte) {
Joerg Roedel44ad9942009-07-27 16:30:42 +02001053 BUG_ON(!(*spte & PT_PRESENT_MASK));
Takuya Yoshikawad6eebf82011-11-14 18:21:34 +09001054 BUG_ON(!is_large_pte(*spte));
Joerg Roedel44ad9942009-07-27 16:30:42 +02001055 pgprintk("rmap_write_protect(large): spte %p %llx %lld\n", spte, *spte, gfn);
Takuya Yoshikawa8dae4442010-01-18 18:45:10 +09001056 if (is_writable_pte(*spte)) {
Xiao Guangrongc3707952011-07-12 03:28:04 +08001057 drop_spte(kvm, spte);
Joerg Roedel44ad9942009-07-27 16:30:42 +02001058 --kvm->stat.lpages;
Joerg Roedel44ad9942009-07-27 16:30:42 +02001059 spte = NULL;
1060 write_protected = 1;
1061 }
1062 spte = rmap_next(kvm, rmapp, spte);
Marcelo Tosatti05da4552008-02-23 11:44:30 -03001063 }
Marcelo Tosatti05da4552008-02-23 11:44:30 -03001064 }
1065
Marcelo Tosattib1a36822008-12-01 22:32:03 -02001066 return write_protected;
Avi Kivity374cbac2007-01-05 16:36:43 -08001067}
1068
Frederik Deweerdt8a8365c2009-10-09 11:42:56 +00001069static int kvm_unmap_rmapp(struct kvm *kvm, unsigned long *rmapp,
1070 unsigned long data)
Andrea Arcangelie930bff2008-07-25 16:24:52 +02001071{
1072 u64 *spte;
1073 int need_tlb_flush = 0;
1074
1075 while ((spte = rmap_next(kvm, rmapp, NULL))) {
1076 BUG_ON(!(*spte & PT_PRESENT_MASK));
1077 rmap_printk("kvm_rmap_unmap_hva: spte %p %llx\n", spte, *spte);
Xiao Guangrongc3707952011-07-12 03:28:04 +08001078 drop_spte(kvm, spte);
Andrea Arcangelie930bff2008-07-25 16:24:52 +02001079 need_tlb_flush = 1;
1080 }
1081 return need_tlb_flush;
1082}
1083
Frederik Deweerdt8a8365c2009-10-09 11:42:56 +00001084static int kvm_set_pte_rmapp(struct kvm *kvm, unsigned long *rmapp,
1085 unsigned long data)
Izik Eidus3da0dd42009-09-23 21:47:18 +03001086{
1087 int need_flush = 0;
Xiao Guangronge4b502e2010-07-16 11:28:09 +08001088 u64 *spte, new_spte;
Izik Eidus3da0dd42009-09-23 21:47:18 +03001089 pte_t *ptep = (pte_t *)data;
1090 pfn_t new_pfn;
1091
1092 WARN_ON(pte_huge(*ptep));
1093 new_pfn = pte_pfn(*ptep);
1094 spte = rmap_next(kvm, rmapp, NULL);
1095 while (spte) {
1096 BUG_ON(!is_shadow_present_pte(*spte));
1097 rmap_printk("kvm_set_pte_rmapp: spte %p %llx\n", spte, *spte);
1098 need_flush = 1;
1099 if (pte_write(*ptep)) {
Xiao Guangrongc3707952011-07-12 03:28:04 +08001100 drop_spte(kvm, spte);
Izik Eidus3da0dd42009-09-23 21:47:18 +03001101 spte = rmap_next(kvm, rmapp, NULL);
1102 } else {
1103 new_spte = *spte &~ (PT64_BASE_ADDR_MASK);
1104 new_spte |= (u64)new_pfn << PAGE_SHIFT;
1105
1106 new_spte &= ~PT_WRITABLE_MASK;
1107 new_spte &= ~SPTE_HOST_WRITEABLE;
Avi Kivityb79b93f2010-06-06 15:46:44 +03001108 new_spte &= ~shadow_accessed_mask;
Xiao Guangrong1df9f2d2011-07-12 03:30:35 +08001109 mmu_spte_clear_track_bits(spte);
1110 mmu_spte_set(spte, new_spte);
Izik Eidus3da0dd42009-09-23 21:47:18 +03001111 spte = rmap_next(kvm, rmapp, spte);
1112 }
1113 }
1114 if (need_flush)
1115 kvm_flush_remote_tlbs(kvm);
1116
1117 return 0;
1118}
1119
Frederik Deweerdt8a8365c2009-10-09 11:42:56 +00001120static int kvm_handle_hva(struct kvm *kvm, unsigned long hva,
1121 unsigned long data,
Izik Eidus3da0dd42009-09-23 21:47:18 +03001122 int (*handler)(struct kvm *kvm, unsigned long *rmapp,
Frederik Deweerdt8a8365c2009-10-09 11:42:56 +00001123 unsigned long data))
Andrea Arcangelie930bff2008-07-25 16:24:52 +02001124{
Joerg Roedel852e3c12009-07-27 16:30:44 +02001125 int i, j;
Avi Kivity90bb6fc2009-12-31 12:10:16 +02001126 int ret;
Andrea Arcangelie930bff2008-07-25 16:24:52 +02001127 int retval = 0;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -02001128 struct kvm_memslots *slots;
Andrea Arcangelie930bff2008-07-25 16:24:52 +02001129
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08001130 slots = kvm_memslots(kvm);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -02001131
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -02001132 for (i = 0; i < slots->nmemslots; i++) {
1133 struct kvm_memory_slot *memslot = &slots->memslots[i];
Andrea Arcangelie930bff2008-07-25 16:24:52 +02001134 unsigned long start = memslot->userspace_addr;
1135 unsigned long end;
1136
Andrea Arcangelie930bff2008-07-25 16:24:52 +02001137 end = start + (memslot->npages << PAGE_SHIFT);
1138 if (hva >= start && hva < end) {
1139 gfn_t gfn_offset = (hva - start) >> PAGE_SHIFT;
Takuya Yoshikawad4dbf472010-12-07 12:59:07 +09001140 gfn_t gfn = memslot->base_gfn + gfn_offset;
Joerg Roedel852e3c12009-07-27 16:30:44 +02001141
Avi Kivity90bb6fc2009-12-31 12:10:16 +02001142 ret = handler(kvm, &memslot->rmap[gfn_offset], data);
Joerg Roedel852e3c12009-07-27 16:30:44 +02001143
1144 for (j = 0; j < KVM_NR_PAGE_SIZES - 1; ++j) {
Takuya Yoshikawad4dbf472010-12-07 12:59:07 +09001145 struct kvm_lpage_info *linfo;
Andrea Arcangeli6e3e2432010-07-16 11:52:55 +02001146
Takuya Yoshikawad4dbf472010-12-07 12:59:07 +09001147 linfo = lpage_info_slot(gfn, memslot,
1148 PT_DIRECTORY_LEVEL + j);
1149 ret |= handler(kvm, &linfo->rmap_pde, data);
Joerg Roedel852e3c12009-07-27 16:30:44 +02001150 }
Avi Kivity90bb6fc2009-12-31 12:10:16 +02001151 trace_kvm_age_page(hva, memslot, ret);
1152 retval |= ret;
Andrea Arcangelie930bff2008-07-25 16:24:52 +02001153 }
1154 }
1155
1156 return retval;
1157}
1158
1159int kvm_unmap_hva(struct kvm *kvm, unsigned long hva)
1160{
Izik Eidus3da0dd42009-09-23 21:47:18 +03001161 return kvm_handle_hva(kvm, hva, 0, kvm_unmap_rmapp);
Andrea Arcangelie930bff2008-07-25 16:24:52 +02001162}
1163
Izik Eidus3da0dd42009-09-23 21:47:18 +03001164void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte)
1165{
Frederik Deweerdt8a8365c2009-10-09 11:42:56 +00001166 kvm_handle_hva(kvm, hva, (unsigned long)&pte, kvm_set_pte_rmapp);
Izik Eidus3da0dd42009-09-23 21:47:18 +03001167}
1168
Frederik Deweerdt8a8365c2009-10-09 11:42:56 +00001169static int kvm_age_rmapp(struct kvm *kvm, unsigned long *rmapp,
1170 unsigned long data)
Andrea Arcangelie930bff2008-07-25 16:24:52 +02001171{
1172 u64 *spte;
1173 int young = 0;
1174
Rik van Riel6316e1c2010-02-03 16:11:03 -05001175 /*
1176 * Emulate the accessed bit for EPT, by checking if this page has
1177 * an EPT mapping, and clearing it if it does. On the next access,
1178 * a new EPT mapping will be established.
1179 * This has some overhead, but not as much as the cost of swapping
1180 * out actively used pages or breaking up actively used hugepages.
1181 */
Sheng Yang534e38b2008-09-08 15:12:30 +08001182 if (!shadow_accessed_mask)
Rik van Riel6316e1c2010-02-03 16:11:03 -05001183 return kvm_unmap_rmapp(kvm, rmapp, data);
Sheng Yang534e38b2008-09-08 15:12:30 +08001184
Andrea Arcangelie930bff2008-07-25 16:24:52 +02001185 spte = rmap_next(kvm, rmapp, NULL);
1186 while (spte) {
1187 int _young;
1188 u64 _spte = *spte;
1189 BUG_ON(!(_spte & PT_PRESENT_MASK));
1190 _young = _spte & PT_ACCESSED_MASK;
1191 if (_young) {
1192 young = 1;
1193 clear_bit(PT_ACCESSED_SHIFT, (unsigned long *)spte);
1194 }
1195 spte = rmap_next(kvm, rmapp, spte);
1196 }
1197 return young;
1198}
1199
Andrea Arcangeli8ee53822011-01-13 15:47:10 -08001200static int kvm_test_age_rmapp(struct kvm *kvm, unsigned long *rmapp,
1201 unsigned long data)
1202{
1203 u64 *spte;
1204 int young = 0;
1205
1206 /*
1207 * If there's no access bit in the secondary pte set by the
1208 * hardware it's up to gup-fast/gup to set the access bit in
1209 * the primary pte or in the page structure.
1210 */
1211 if (!shadow_accessed_mask)
1212 goto out;
1213
1214 spte = rmap_next(kvm, rmapp, NULL);
1215 while (spte) {
1216 u64 _spte = *spte;
1217 BUG_ON(!(_spte & PT_PRESENT_MASK));
1218 young = _spte & PT_ACCESSED_MASK;
1219 if (young) {
1220 young = 1;
1221 break;
1222 }
1223 spte = rmap_next(kvm, rmapp, spte);
1224 }
1225out:
1226 return young;
1227}
1228
Marcelo Tosatti53a27b32009-08-05 15:43:58 -03001229#define RMAP_RECYCLE_THRESHOLD 1000
1230
Joerg Roedel852e3c12009-07-27 16:30:44 +02001231static void rmap_recycle(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn)
Marcelo Tosatti53a27b32009-08-05 15:43:58 -03001232{
1233 unsigned long *rmapp;
Joerg Roedel852e3c12009-07-27 16:30:44 +02001234 struct kvm_mmu_page *sp;
1235
1236 sp = page_header(__pa(spte));
Marcelo Tosatti53a27b32009-08-05 15:43:58 -03001237
Joerg Roedel852e3c12009-07-27 16:30:44 +02001238 rmapp = gfn_to_rmap(vcpu->kvm, gfn, sp->role.level);
Marcelo Tosatti53a27b32009-08-05 15:43:58 -03001239
Izik Eidus3da0dd42009-09-23 21:47:18 +03001240 kvm_unmap_rmapp(vcpu->kvm, rmapp, 0);
Marcelo Tosatti53a27b32009-08-05 15:43:58 -03001241 kvm_flush_remote_tlbs(vcpu->kvm);
1242}
1243
Andrea Arcangelie930bff2008-07-25 16:24:52 +02001244int kvm_age_hva(struct kvm *kvm, unsigned long hva)
1245{
Izik Eidus3da0dd42009-09-23 21:47:18 +03001246 return kvm_handle_hva(kvm, hva, 0, kvm_age_rmapp);
Andrea Arcangelie930bff2008-07-25 16:24:52 +02001247}
1248
Andrea Arcangeli8ee53822011-01-13 15:47:10 -08001249int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
1250{
1251 return kvm_handle_hva(kvm, hva, 0, kvm_test_age_rmapp);
1252}
1253
Yaozu Dongd6c69ee2007-04-25 14:17:25 +08001254#ifdef MMU_DEBUG
Avi Kivity47ad8e62007-05-06 15:50:58 +03001255static int is_empty_shadow_page(u64 *spt)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001256{
Avi Kivity139bdb22007-01-05 16:36:50 -08001257 u64 *pos;
1258 u64 *end;
1259
Avi Kivity47ad8e62007-05-06 15:50:58 +03001260 for (pos = spt, end = pos + PAGE_SIZE / sizeof(u64); pos != end; pos++)
Avi Kivity3c915512008-05-20 16:21:13 +03001261 if (is_shadow_present_pte(*pos)) {
Harvey Harrisonb8688d52008-03-03 12:59:56 -08001262 printk(KERN_ERR "%s: %p %llx\n", __func__,
Avi Kivity139bdb22007-01-05 16:36:50 -08001263 pos, *pos);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001264 return 0;
Avi Kivity139bdb22007-01-05 16:36:50 -08001265 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08001266 return 1;
1267}
Yaozu Dongd6c69ee2007-04-25 14:17:25 +08001268#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -08001269
Dave Hansen45221ab2010-08-19 18:11:37 -07001270/*
1271 * This value is the sum of all of the kvm instances's
1272 * kvm->arch.n_used_mmu_pages values. We need a global,
1273 * aggregate version in order to make the slab shrinker
1274 * faster
1275 */
1276static inline void kvm_mod_used_mmu_pages(struct kvm *kvm, int nr)
1277{
1278 kvm->arch.n_used_mmu_pages += nr;
1279 percpu_counter_add(&kvm_total_used_mmu_pages, nr);
1280}
1281
Xiao Guangrongbd4c86e2011-07-12 03:27:14 +08001282/*
1283 * Remove the sp from shadow page cache, after call it,
1284 * we can not find this sp from the cache, and the shadow
1285 * page table is still valid.
1286 * It should be under the protection of mmu lock.
1287 */
1288static void kvm_mmu_isolate_page(struct kvm_mmu_page *sp)
Avi Kivity260746c2007-01-05 16:36:49 -08001289{
Avi Kivity4db35312007-11-21 15:28:32 +02001290 ASSERT(is_empty_shadow_page(sp->spt));
Xiao Guangrong77758342010-06-04 21:53:54 +08001291 hlist_del(&sp->hash_link);
Lai Jiangshan2032a932010-05-26 16:49:59 +08001292 if (!sp->role.direct)
Xiao Guangrong842f22e2011-03-04 19:01:10 +08001293 free_page((unsigned long)sp->gfns);
Xiao Guangrongbd4c86e2011-07-12 03:27:14 +08001294}
1295
1296/*
1297 * Free the shadow page table and the sp, we can do it
1298 * out of the protection of mmu lock.
1299 */
1300static void kvm_mmu_free_page(struct kvm_mmu_page *sp)
1301{
1302 list_del(&sp->link);
1303 free_page((unsigned long)sp->spt);
Xiao Guangronge8ad9a72010-05-13 10:06:02 +08001304 kmem_cache_free(mmu_page_header_cache, sp);
Avi Kivity260746c2007-01-05 16:36:49 -08001305}
1306
Avi Kivitycea0f0e2007-01-05 16:36:43 -08001307static unsigned kvm_page_table_hashfn(gfn_t gfn)
1308{
Dong, Eddie1ae0a132008-01-07 13:20:25 +02001309 return gfn & ((1 << KVM_MMU_HASH_SHIFT) - 1);
Avi Kivitycea0f0e2007-01-05 16:36:43 -08001310}
1311
Xiao Guangrong67052b32011-05-15 23:27:08 +08001312static void mmu_page_add_parent_pte(struct kvm_vcpu *vcpu,
1313 struct kvm_mmu_page *sp, u64 *parent_pte)
1314{
1315 if (!parent_pte)
1316 return;
1317
1318 pte_list_add(vcpu, parent_pte, &sp->parent_ptes);
1319}
1320
1321static void mmu_page_remove_parent_pte(struct kvm_mmu_page *sp,
1322 u64 *parent_pte)
1323{
1324 pte_list_remove(parent_pte, &sp->parent_ptes);
1325}
1326
Xiao Guangrongbcdd9a92011-05-15 23:28:29 +08001327static void drop_parent_pte(struct kvm_mmu_page *sp,
1328 u64 *parent_pte)
1329{
1330 mmu_page_remove_parent_pte(sp, parent_pte);
Xiao Guangrong1df9f2d2011-07-12 03:30:35 +08001331 mmu_spte_clear_no_track(parent_pte);
Xiao Guangrongbcdd9a92011-05-15 23:28:29 +08001332}
1333
Avi Kivity25c0de22007-01-05 16:36:42 -08001334static struct kvm_mmu_page *kvm_mmu_alloc_page(struct kvm_vcpu *vcpu,
Lai Jiangshan2032a932010-05-26 16:49:59 +08001335 u64 *parent_pte, int direct)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001336{
Avi Kivity4db35312007-11-21 15:28:32 +02001337 struct kvm_mmu_page *sp;
Xiao Guangrong67052b32011-05-15 23:27:08 +08001338 sp = mmu_memory_cache_alloc(&vcpu->arch.mmu_page_header_cache,
1339 sizeof *sp);
Zhang Xiantaoad312c72007-12-13 23:50:52 +08001340 sp->spt = mmu_memory_cache_alloc(&vcpu->arch.mmu_page_cache, PAGE_SIZE);
Lai Jiangshan2032a932010-05-26 16:49:59 +08001341 if (!direct)
1342 sp->gfns = mmu_memory_cache_alloc(&vcpu->arch.mmu_page_cache,
1343 PAGE_SIZE);
Avi Kivity4db35312007-11-21 15:28:32 +02001344 set_page_private(virt_to_page(sp->spt), (unsigned long)sp);
Zhang Xiantaof05e70a2007-12-14 10:01:48 +08001345 list_add(&sp->link, &vcpu->kvm->arch.active_mmu_pages);
Sheng Yang291f26b2008-10-16 17:30:57 +08001346 bitmap_zero(sp->slot_bitmap, KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS);
Xiao Guangrong67052b32011-05-15 23:27:08 +08001347 sp->parent_ptes = 0;
1348 mmu_page_add_parent_pte(vcpu, sp, parent_pte);
Dave Hansen45221ab2010-08-19 18:11:37 -07001349 kvm_mod_used_mmu_pages(vcpu->kvm, +1);
Avi Kivity4db35312007-11-21 15:28:32 +02001350 return sp;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001351}
1352
Xiao Guangrong67052b32011-05-15 23:27:08 +08001353static void mark_unsync(u64 *spte);
Xiao Guangrong6b184932010-04-16 21:29:17 +08001354static void kvm_mmu_mark_parents_unsync(struct kvm_mmu_page *sp)
Marcelo Tosatti0074ff62008-09-23 13:18:40 -03001355{
Xiao Guangrong67052b32011-05-15 23:27:08 +08001356 pte_list_walk(&sp->parent_ptes, mark_unsync);
Xiao Guangrong1047df12010-06-11 21:35:15 +08001357}
1358
Xiao Guangrong67052b32011-05-15 23:27:08 +08001359static void mark_unsync(u64 *spte)
Xiao Guangrong1047df12010-06-11 21:35:15 +08001360{
Xiao Guangrong67052b32011-05-15 23:27:08 +08001361 struct kvm_mmu_page *sp;
Xiao Guangrong1047df12010-06-11 21:35:15 +08001362 unsigned int index;
1363
Xiao Guangrong67052b32011-05-15 23:27:08 +08001364 sp = page_header(__pa(spte));
Xiao Guangrong1047df12010-06-11 21:35:15 +08001365 index = spte - sp->spt;
1366 if (__test_and_set_bit(index, sp->unsync_child_bitmap))
1367 return;
1368 if (sp->unsync_children++)
1369 return;
1370 kvm_mmu_mark_parents_unsync(sp);
Marcelo Tosatti0074ff62008-09-23 13:18:40 -03001371}
1372
Marcelo Tosattie8bc2172008-09-23 13:18:33 -03001373static int nonpaging_sync_page(struct kvm_vcpu *vcpu,
Xiao Guangronga4a8e6f2010-11-19 17:04:03 +08001374 struct kvm_mmu_page *sp)
Marcelo Tosattie8bc2172008-09-23 13:18:33 -03001375{
1376 return 1;
1377}
1378
Marcelo Tosattia7052892008-09-23 13:18:35 -03001379static void nonpaging_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
1380{
1381}
1382
Xiao Guangrong0f53b5b2011-03-09 15:43:51 +08001383static void nonpaging_update_pte(struct kvm_vcpu *vcpu,
1384 struct kvm_mmu_page *sp, u64 *spte,
Xiao Guangrong7c562522011-03-28 10:29:27 +08001385 const void *pte)
Xiao Guangrong0f53b5b2011-03-09 15:43:51 +08001386{
1387 WARN_ON(1);
1388}
1389
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001390#define KVM_PAGE_ARRAY_NR 16
1391
1392struct kvm_mmu_pages {
1393 struct mmu_page_and_offset {
1394 struct kvm_mmu_page *sp;
1395 unsigned int idx;
1396 } page[KVM_PAGE_ARRAY_NR];
1397 unsigned int nr;
1398};
1399
Marcelo Tosatti0074ff62008-09-23 13:18:40 -03001400#define for_each_unsync_children(bitmap, idx) \
1401 for (idx = find_first_bit(bitmap, 512); \
1402 idx < 512; \
1403 idx = find_next_bit(bitmap, 512, idx+1))
1404
Hannes Edercded19f2009-02-21 02:19:13 +01001405static int mmu_pages_add(struct kvm_mmu_pages *pvec, struct kvm_mmu_page *sp,
1406 int idx)
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001407{
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001408 int i;
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001409
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001410 if (sp->unsync)
1411 for (i=0; i < pvec->nr; i++)
1412 if (pvec->page[i].sp == sp)
1413 return 0;
1414
1415 pvec->page[pvec->nr].sp = sp;
1416 pvec->page[pvec->nr].idx = idx;
1417 pvec->nr++;
1418 return (pvec->nr == KVM_PAGE_ARRAY_NR);
1419}
1420
1421static int __mmu_unsync_walk(struct kvm_mmu_page *sp,
1422 struct kvm_mmu_pages *pvec)
1423{
1424 int i, ret, nr_unsync_leaf = 0;
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001425
Marcelo Tosatti0074ff62008-09-23 13:18:40 -03001426 for_each_unsync_children(sp->unsync_child_bitmap, i) {
Xiao Guangrong7a8f1a72010-06-11 21:34:04 +08001427 struct kvm_mmu_page *child;
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001428 u64 ent = sp->spt[i];
1429
Xiao Guangrong7a8f1a72010-06-11 21:34:04 +08001430 if (!is_shadow_present_pte(ent) || is_large_pte(ent))
1431 goto clear_child_bitmap;
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001432
Xiao Guangrong7a8f1a72010-06-11 21:34:04 +08001433 child = page_header(ent & PT64_BASE_ADDR_MASK);
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001434
Xiao Guangrong7a8f1a72010-06-11 21:34:04 +08001435 if (child->unsync_children) {
1436 if (mmu_pages_add(pvec, child, i))
1437 return -ENOSPC;
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001438
Xiao Guangrong7a8f1a72010-06-11 21:34:04 +08001439 ret = __mmu_unsync_walk(child, pvec);
1440 if (!ret)
1441 goto clear_child_bitmap;
1442 else if (ret > 0)
1443 nr_unsync_leaf += ret;
1444 else
1445 return ret;
1446 } else if (child->unsync) {
1447 nr_unsync_leaf++;
1448 if (mmu_pages_add(pvec, child, i))
1449 return -ENOSPC;
1450 } else
1451 goto clear_child_bitmap;
1452
1453 continue;
1454
1455clear_child_bitmap:
1456 __clear_bit(i, sp->unsync_child_bitmap);
1457 sp->unsync_children--;
1458 WARN_ON((int)sp->unsync_children < 0);
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001459 }
1460
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001461
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001462 return nr_unsync_leaf;
1463}
1464
1465static int mmu_unsync_walk(struct kvm_mmu_page *sp,
1466 struct kvm_mmu_pages *pvec)
1467{
1468 if (!sp->unsync_children)
1469 return 0;
1470
1471 mmu_pages_add(pvec, sp, 0);
1472 return __mmu_unsync_walk(sp, pvec);
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001473}
1474
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001475static void kvm_unlink_unsync_page(struct kvm *kvm, struct kvm_mmu_page *sp)
1476{
1477 WARN_ON(!sp->unsync);
Xiao Guangrong5e1b3dd2010-04-28 11:55:06 +08001478 trace_kvm_mmu_sync_page(sp);
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001479 sp->unsync = 0;
1480 --kvm->stat.mmu_unsync;
1481}
1482
Xiao Guangrong77758342010-06-04 21:53:54 +08001483static int kvm_mmu_prepare_zap_page(struct kvm *kvm, struct kvm_mmu_page *sp,
1484 struct list_head *invalid_list);
1485static void kvm_mmu_commit_zap_page(struct kvm *kvm,
1486 struct list_head *invalid_list);
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001487
Xiao Guangrongf41d3352010-06-04 21:56:11 +08001488#define for_each_gfn_sp(kvm, sp, gfn, pos) \
1489 hlist_for_each_entry(sp, pos, \
Xiao Guangrong7ae680e2010-06-04 21:53:07 +08001490 &(kvm)->arch.mmu_page_hash[kvm_page_table_hashfn(gfn)], hash_link) \
1491 if ((sp)->gfn != (gfn)) {} else
1492
Xiao Guangrongf41d3352010-06-04 21:56:11 +08001493#define for_each_gfn_indirect_valid_sp(kvm, sp, gfn, pos) \
1494 hlist_for_each_entry(sp, pos, \
Xiao Guangrong7ae680e2010-06-04 21:53:07 +08001495 &(kvm)->arch.mmu_page_hash[kvm_page_table_hashfn(gfn)], hash_link) \
1496 if ((sp)->gfn != (gfn) || (sp)->role.direct || \
1497 (sp)->role.invalid) {} else
1498
Xiao Guangrongf918b442010-06-11 21:30:36 +08001499/* @sp->gfn should be write-protected at the call site */
Xiao Guangrong1d9dc7e2010-05-15 18:51:24 +08001500static int __kvm_sync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
Xiao Guangrongd98ba052010-06-04 21:55:29 +08001501 struct list_head *invalid_list, bool clear_unsync)
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001502{
Avi Kivity5b7e0102010-04-14 19:20:03 +03001503 if (sp->role.cr4_pae != !!is_pae(vcpu)) {
Xiao Guangrongd98ba052010-06-04 21:55:29 +08001504 kvm_mmu_prepare_zap_page(vcpu->kvm, sp, invalid_list);
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001505 return 1;
1506 }
1507
Xiao Guangrongf918b442010-06-11 21:30:36 +08001508 if (clear_unsync)
Xiao Guangrong1d9dc7e2010-05-15 18:51:24 +08001509 kvm_unlink_unsync_page(vcpu->kvm, sp);
Xiao Guangrong1d9dc7e2010-05-15 18:51:24 +08001510
Xiao Guangronga4a8e6f2010-11-19 17:04:03 +08001511 if (vcpu->arch.mmu.sync_page(vcpu, sp)) {
Xiao Guangrongd98ba052010-06-04 21:55:29 +08001512 kvm_mmu_prepare_zap_page(vcpu->kvm, sp, invalid_list);
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001513 return 1;
1514 }
1515
1516 kvm_mmu_flush_tlb(vcpu);
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001517 return 0;
1518}
1519
Xiao Guangrong1d9dc7e2010-05-15 18:51:24 +08001520static int kvm_sync_page_transient(struct kvm_vcpu *vcpu,
1521 struct kvm_mmu_page *sp)
1522{
Xiao Guangrongd98ba052010-06-04 21:55:29 +08001523 LIST_HEAD(invalid_list);
Xiao Guangrong1d9dc7e2010-05-15 18:51:24 +08001524 int ret;
1525
Xiao Guangrongd98ba052010-06-04 21:55:29 +08001526 ret = __kvm_sync_page(vcpu, sp, &invalid_list, false);
Xiao Guangrongbe71e062010-06-11 21:31:38 +08001527 if (ret)
Xiao Guangrongd98ba052010-06-04 21:55:29 +08001528 kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
1529
Xiao Guangrong1d9dc7e2010-05-15 18:51:24 +08001530 return ret;
1531}
1532
Xiao Guangrongd98ba052010-06-04 21:55:29 +08001533static int kvm_sync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
1534 struct list_head *invalid_list)
Xiao Guangrong1d9dc7e2010-05-15 18:51:24 +08001535{
Xiao Guangrongd98ba052010-06-04 21:55:29 +08001536 return __kvm_sync_page(vcpu, sp, invalid_list, true);
Xiao Guangrong1d9dc7e2010-05-15 18:51:24 +08001537}
1538
Xiao Guangrong9f1a1222010-05-24 15:41:33 +08001539/* @gfn should be write-protected at the call site */
1540static void kvm_sync_pages(struct kvm_vcpu *vcpu, gfn_t gfn)
1541{
Xiao Guangrong9f1a1222010-05-24 15:41:33 +08001542 struct kvm_mmu_page *s;
Xiao Guangrongf41d3352010-06-04 21:56:11 +08001543 struct hlist_node *node;
Xiao Guangrongd98ba052010-06-04 21:55:29 +08001544 LIST_HEAD(invalid_list);
Xiao Guangrong9f1a1222010-05-24 15:41:33 +08001545 bool flush = false;
1546
Xiao Guangrongf41d3352010-06-04 21:56:11 +08001547 for_each_gfn_indirect_valid_sp(vcpu->kvm, s, gfn, node) {
Xiao Guangrong7ae680e2010-06-04 21:53:07 +08001548 if (!s->unsync)
Xiao Guangrong9f1a1222010-05-24 15:41:33 +08001549 continue;
1550
1551 WARN_ON(s->role.level != PT_PAGE_TABLE_LEVEL);
Xiao Guangronga4a8e6f2010-11-19 17:04:03 +08001552 kvm_unlink_unsync_page(vcpu->kvm, s);
Xiao Guangrong9f1a1222010-05-24 15:41:33 +08001553 if ((s->role.cr4_pae != !!is_pae(vcpu)) ||
Xiao Guangronga4a8e6f2010-11-19 17:04:03 +08001554 (vcpu->arch.mmu.sync_page(vcpu, s))) {
Xiao Guangrongd98ba052010-06-04 21:55:29 +08001555 kvm_mmu_prepare_zap_page(vcpu->kvm, s, &invalid_list);
Xiao Guangrong9f1a1222010-05-24 15:41:33 +08001556 continue;
1557 }
Xiao Guangrong9f1a1222010-05-24 15:41:33 +08001558 flush = true;
1559 }
1560
Xiao Guangrongd98ba052010-06-04 21:55:29 +08001561 kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
Xiao Guangrong9f1a1222010-05-24 15:41:33 +08001562 if (flush)
1563 kvm_mmu_flush_tlb(vcpu);
1564}
1565
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001566struct mmu_page_path {
1567 struct kvm_mmu_page *parent[PT64_ROOT_LEVEL-1];
1568 unsigned int idx[PT64_ROOT_LEVEL-1];
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001569};
1570
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001571#define for_each_sp(pvec, sp, parents, i) \
1572 for (i = mmu_pages_next(&pvec, &parents, -1), \
1573 sp = pvec.page[i].sp; \
1574 i < pvec.nr && ({ sp = pvec.page[i].sp; 1;}); \
1575 i = mmu_pages_next(&pvec, &parents, i))
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001576
Hannes Edercded19f2009-02-21 02:19:13 +01001577static int mmu_pages_next(struct kvm_mmu_pages *pvec,
1578 struct mmu_page_path *parents,
1579 int i)
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001580{
1581 int n;
1582
1583 for (n = i+1; n < pvec->nr; n++) {
1584 struct kvm_mmu_page *sp = pvec->page[n].sp;
1585
1586 if (sp->role.level == PT_PAGE_TABLE_LEVEL) {
1587 parents->idx[0] = pvec->page[n].idx;
1588 return n;
1589 }
1590
1591 parents->parent[sp->role.level-2] = sp;
1592 parents->idx[sp->role.level-1] = pvec->page[n].idx;
1593 }
1594
1595 return n;
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001596}
1597
Hannes Edercded19f2009-02-21 02:19:13 +01001598static void mmu_pages_clear_parents(struct mmu_page_path *parents)
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001599{
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001600 struct kvm_mmu_page *sp;
1601 unsigned int level = 0;
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001602
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001603 do {
1604 unsigned int idx = parents->idx[level];
1605
1606 sp = parents->parent[level];
1607 if (!sp)
1608 return;
1609
1610 --sp->unsync_children;
1611 WARN_ON((int)sp->unsync_children < 0);
1612 __clear_bit(idx, sp->unsync_child_bitmap);
1613 level++;
1614 } while (level < PT64_ROOT_LEVEL-1 && !sp->unsync_children);
1615}
1616
1617static void kvm_mmu_pages_init(struct kvm_mmu_page *parent,
1618 struct mmu_page_path *parents,
1619 struct kvm_mmu_pages *pvec)
1620{
1621 parents->parent[parent->role.level-1] = NULL;
1622 pvec->nr = 0;
1623}
1624
1625static void mmu_sync_children(struct kvm_vcpu *vcpu,
1626 struct kvm_mmu_page *parent)
1627{
1628 int i;
1629 struct kvm_mmu_page *sp;
1630 struct mmu_page_path parents;
1631 struct kvm_mmu_pages pages;
Xiao Guangrongd98ba052010-06-04 21:55:29 +08001632 LIST_HEAD(invalid_list);
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001633
1634 kvm_mmu_pages_init(parent, &parents, &pages);
1635 while (mmu_unsync_walk(parent, &pages)) {
Marcelo Tosattib1a36822008-12-01 22:32:03 -02001636 int protected = 0;
1637
1638 for_each_sp(pages, sp, parents, i)
1639 protected |= rmap_write_protect(vcpu->kvm, sp->gfn);
1640
1641 if (protected)
1642 kvm_flush_remote_tlbs(vcpu->kvm);
1643
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001644 for_each_sp(pages, sp, parents, i) {
Xiao Guangrongd98ba052010-06-04 21:55:29 +08001645 kvm_sync_page(vcpu, sp, &invalid_list);
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001646 mmu_pages_clear_parents(&parents);
1647 }
Xiao Guangrongd98ba052010-06-04 21:55:29 +08001648 kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001649 cond_resched_lock(&vcpu->kvm->mmu_lock);
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001650 kvm_mmu_pages_init(parent, &parents, &pages);
1651 }
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001652}
1653
Xiao Guangrongc3707952011-07-12 03:28:04 +08001654static void init_shadow_page_table(struct kvm_mmu_page *sp)
1655{
1656 int i;
1657
1658 for (i = 0; i < PT64_ENT_PER_PAGE; ++i)
1659 sp->spt[i] = 0ull;
1660}
1661
Xiao Guangronga30f47c2011-09-22 16:58:36 +08001662static void __clear_sp_write_flooding_count(struct kvm_mmu_page *sp)
1663{
1664 sp->write_flooding_count = 0;
1665}
1666
1667static void clear_sp_write_flooding_count(u64 *spte)
1668{
1669 struct kvm_mmu_page *sp = page_header(__pa(spte));
1670
1671 __clear_sp_write_flooding_count(sp);
1672}
1673
Avi Kivitycea0f0e2007-01-05 16:36:43 -08001674static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
1675 gfn_t gfn,
1676 gva_t gaddr,
1677 unsigned level,
Avi Kivityf6e2c022009-01-11 13:02:10 +02001678 int direct,
Avi Kivity41074d02007-12-09 17:00:02 +02001679 unsigned access,
Avi Kivityf7d9c7b2008-02-26 22:12:10 +02001680 u64 *parent_pte)
Avi Kivitycea0f0e2007-01-05 16:36:43 -08001681{
1682 union kvm_mmu_page_role role;
Avi Kivitycea0f0e2007-01-05 16:36:43 -08001683 unsigned quadrant;
Xiao Guangrong9f1a1222010-05-24 15:41:33 +08001684 struct kvm_mmu_page *sp;
Xiao Guangrongf41d3352010-06-04 21:56:11 +08001685 struct hlist_node *node;
Xiao Guangrong9f1a1222010-05-24 15:41:33 +08001686 bool need_sync = false;
Avi Kivitycea0f0e2007-01-05 16:36:43 -08001687
Avi Kivitya770f6f2008-12-21 19:20:09 +02001688 role = vcpu->arch.mmu.base_role;
Avi Kivitycea0f0e2007-01-05 16:36:43 -08001689 role.level = level;
Avi Kivityf6e2c022009-01-11 13:02:10 +02001690 role.direct = direct;
Avi Kivity84b0c8c2010-03-14 10:16:40 +02001691 if (role.direct)
Avi Kivity5b7e0102010-04-14 19:20:03 +03001692 role.cr4_pae = 0;
Avi Kivity41074d02007-12-09 17:00:02 +02001693 role.access = access;
Joerg Roedelc5a78f2b2010-09-10 17:30:39 +02001694 if (!vcpu->arch.mmu.direct_map
1695 && vcpu->arch.mmu.root_level <= PT32_ROOT_LEVEL) {
Avi Kivitycea0f0e2007-01-05 16:36:43 -08001696 quadrant = gaddr >> (PAGE_SHIFT + (PT64_PT_BITS * level));
1697 quadrant &= (1 << ((PT32_PT_BITS - PT64_PT_BITS) * level)) - 1;
1698 role.quadrant = quadrant;
1699 }
Xiao Guangrongf41d3352010-06-04 21:56:11 +08001700 for_each_gfn_sp(vcpu->kvm, sp, gfn, node) {
Xiao Guangrong7ae680e2010-06-04 21:53:07 +08001701 if (!need_sync && sp->unsync)
1702 need_sync = true;
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001703
Xiao Guangrong7ae680e2010-06-04 21:53:07 +08001704 if (sp->role.word != role.word)
1705 continue;
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001706
Xiao Guangrong7ae680e2010-06-04 21:53:07 +08001707 if (sp->unsync && kvm_sync_page_transient(vcpu, sp))
1708 break;
Xiao Guangronge02aa902010-05-15 18:52:34 +08001709
Xiao Guangrong7ae680e2010-06-04 21:53:07 +08001710 mmu_page_add_parent_pte(vcpu, sp, parent_pte);
1711 if (sp->unsync_children) {
Avi Kivitya8eeb042010-05-10 12:34:53 +03001712 kvm_make_request(KVM_REQ_MMU_SYNC, vcpu);
Xiao Guangrong7ae680e2010-06-04 21:53:07 +08001713 kvm_mmu_mark_parents_unsync(sp);
1714 } else if (sp->unsync)
1715 kvm_mmu_mark_parents_unsync(sp);
Xiao Guangronge02aa902010-05-15 18:52:34 +08001716
Xiao Guangronga30f47c2011-09-22 16:58:36 +08001717 __clear_sp_write_flooding_count(sp);
Xiao Guangrong7ae680e2010-06-04 21:53:07 +08001718 trace_kvm_mmu_get_page(sp, false);
1719 return sp;
1720 }
Avi Kivitydfc5aa02007-12-18 19:47:18 +02001721 ++vcpu->kvm->stat.mmu_cache_miss;
Lai Jiangshan2032a932010-05-26 16:49:59 +08001722 sp = kvm_mmu_alloc_page(vcpu, parent_pte, direct);
Avi Kivity4db35312007-11-21 15:28:32 +02001723 if (!sp)
1724 return sp;
Avi Kivity4db35312007-11-21 15:28:32 +02001725 sp->gfn = gfn;
1726 sp->role = role;
Xiao Guangrong7ae680e2010-06-04 21:53:07 +08001727 hlist_add_head(&sp->hash_link,
1728 &vcpu->kvm->arch.mmu_page_hash[kvm_page_table_hashfn(gfn)]);
Avi Kivityf6e2c022009-01-11 13:02:10 +02001729 if (!direct) {
Marcelo Tosattib1a36822008-12-01 22:32:03 -02001730 if (rmap_write_protect(vcpu->kvm, gfn))
1731 kvm_flush_remote_tlbs(vcpu->kvm);
Xiao Guangrong9f1a1222010-05-24 15:41:33 +08001732 if (level > PT_PAGE_TABLE_LEVEL && need_sync)
1733 kvm_sync_pages(vcpu, gfn);
1734
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001735 account_shadowed(vcpu->kvm, gfn);
1736 }
Xiao Guangrongc3707952011-07-12 03:28:04 +08001737 init_shadow_page_table(sp);
Avi Kivityf691fe12009-07-06 15:58:14 +03001738 trace_kvm_mmu_get_page(sp, true);
Avi Kivity4db35312007-11-21 15:28:32 +02001739 return sp;
Avi Kivitycea0f0e2007-01-05 16:36:43 -08001740}
1741
Avi Kivity2d111232008-12-25 14:39:47 +02001742static void shadow_walk_init(struct kvm_shadow_walk_iterator *iterator,
1743 struct kvm_vcpu *vcpu, u64 addr)
1744{
1745 iterator->addr = addr;
1746 iterator->shadow_addr = vcpu->arch.mmu.root_hpa;
1747 iterator->level = vcpu->arch.mmu.shadow_root_level;
Joerg Roedel81407ca2010-09-10 17:31:00 +02001748
1749 if (iterator->level == PT64_ROOT_LEVEL &&
1750 vcpu->arch.mmu.root_level < PT64_ROOT_LEVEL &&
1751 !vcpu->arch.mmu.direct_map)
1752 --iterator->level;
1753
Avi Kivity2d111232008-12-25 14:39:47 +02001754 if (iterator->level == PT32E_ROOT_LEVEL) {
1755 iterator->shadow_addr
1756 = vcpu->arch.mmu.pae_root[(addr >> 30) & 3];
1757 iterator->shadow_addr &= PT64_BASE_ADDR_MASK;
1758 --iterator->level;
1759 if (!iterator->shadow_addr)
1760 iterator->level = 0;
1761 }
1762}
1763
1764static bool shadow_walk_okay(struct kvm_shadow_walk_iterator *iterator)
1765{
1766 if (iterator->level < PT_PAGE_TABLE_LEVEL)
1767 return false;
Marcelo Tosatti4d889542009-06-11 12:07:41 -03001768
Avi Kivity2d111232008-12-25 14:39:47 +02001769 iterator->index = SHADOW_PT_INDEX(iterator->addr, iterator->level);
1770 iterator->sptep = ((u64 *)__va(iterator->shadow_addr)) + iterator->index;
1771 return true;
1772}
1773
Xiao Guangrongc2a2ac22011-07-12 03:32:13 +08001774static void __shadow_walk_next(struct kvm_shadow_walk_iterator *iterator,
1775 u64 spte)
Avi Kivity2d111232008-12-25 14:39:47 +02001776{
Xiao Guangrongc2a2ac22011-07-12 03:32:13 +08001777 if (is_last_spte(spte, iterator->level)) {
Xiao Guangrong052331b2011-07-12 03:21:17 +08001778 iterator->level = 0;
1779 return;
1780 }
1781
Xiao Guangrongc2a2ac22011-07-12 03:32:13 +08001782 iterator->shadow_addr = spte & PT64_BASE_ADDR_MASK;
Avi Kivity2d111232008-12-25 14:39:47 +02001783 --iterator->level;
1784}
1785
Xiao Guangrongc2a2ac22011-07-12 03:32:13 +08001786static void shadow_walk_next(struct kvm_shadow_walk_iterator *iterator)
1787{
1788 return __shadow_walk_next(iterator, *iterator->sptep);
1789}
1790
Avi Kivity32ef26a2010-07-13 14:27:04 +03001791static void link_shadow_page(u64 *sptep, struct kvm_mmu_page *sp)
1792{
1793 u64 spte;
1794
1795 spte = __pa(sp->spt)
1796 | PT_PRESENT_MASK | PT_ACCESSED_MASK
1797 | PT_WRITABLE_MASK | PT_USER_MASK;
Xiao Guangrong1df9f2d2011-07-12 03:30:35 +08001798 mmu_spte_set(sptep, spte);
Avi Kivity32ef26a2010-07-13 14:27:04 +03001799}
1800
Avi Kivitya3aa51c2010-07-13 14:27:06 +03001801static void drop_large_spte(struct kvm_vcpu *vcpu, u64 *sptep)
1802{
1803 if (is_large_pte(*sptep)) {
Xiao Guangrongc3707952011-07-12 03:28:04 +08001804 drop_spte(vcpu->kvm, sptep);
Avi Kivitya3aa51c2010-07-13 14:27:06 +03001805 kvm_flush_remote_tlbs(vcpu->kvm);
1806 }
1807}
1808
Avi Kivitya357bd22010-07-13 14:27:07 +03001809static void validate_direct_spte(struct kvm_vcpu *vcpu, u64 *sptep,
1810 unsigned direct_access)
1811{
1812 if (is_shadow_present_pte(*sptep) && !is_large_pte(*sptep)) {
1813 struct kvm_mmu_page *child;
1814
1815 /*
1816 * For the direct sp, if the guest pte's dirty bit
1817 * changed form clean to dirty, it will corrupt the
1818 * sp's access: allow writable in the read-only sp,
1819 * so we should update the spte at this point to get
1820 * a new sp with the correct access.
1821 */
1822 child = page_header(*sptep & PT64_BASE_ADDR_MASK);
1823 if (child->role.access == direct_access)
1824 return;
1825
Xiao Guangrongbcdd9a92011-05-15 23:28:29 +08001826 drop_parent_pte(child, sptep);
Avi Kivitya357bd22010-07-13 14:27:07 +03001827 kvm_flush_remote_tlbs(vcpu->kvm);
1828 }
1829}
1830
Xiao Guangrong505aef82011-09-22 16:56:06 +08001831static bool mmu_page_zap_pte(struct kvm *kvm, struct kvm_mmu_page *sp,
Xiao Guangrong38e3b2b2011-05-15 23:27:52 +08001832 u64 *spte)
1833{
1834 u64 pte;
1835 struct kvm_mmu_page *child;
1836
1837 pte = *spte;
1838 if (is_shadow_present_pte(pte)) {
Xiao Guangrong505aef82011-09-22 16:56:06 +08001839 if (is_last_spte(pte, sp->role.level)) {
Xiao Guangrongc3707952011-07-12 03:28:04 +08001840 drop_spte(kvm, spte);
Xiao Guangrong505aef82011-09-22 16:56:06 +08001841 if (is_large_pte(pte))
1842 --kvm->stat.lpages;
1843 } else {
Xiao Guangrong38e3b2b2011-05-15 23:27:52 +08001844 child = page_header(pte & PT64_BASE_ADDR_MASK);
Xiao Guangrongbcdd9a92011-05-15 23:28:29 +08001845 drop_parent_pte(child, spte);
Xiao Guangrong38e3b2b2011-05-15 23:27:52 +08001846 }
Xiao Guangrong505aef82011-09-22 16:56:06 +08001847 return true;
1848 }
1849
1850 if (is_mmio_spte(pte))
Xiao Guangrongce88dec2011-07-12 03:33:44 +08001851 mmu_spte_clear_no_track(spte);
Xiao Guangrongc3707952011-07-12 03:28:04 +08001852
Xiao Guangrong505aef82011-09-22 16:56:06 +08001853 return false;
Xiao Guangrong38e3b2b2011-05-15 23:27:52 +08001854}
1855
Avi Kivity90cb0522007-07-17 13:04:56 +03001856static void kvm_mmu_page_unlink_children(struct kvm *kvm,
Avi Kivity4db35312007-11-21 15:28:32 +02001857 struct kvm_mmu_page *sp)
Avi Kivitya4360362007-01-05 16:36:45 -08001858{
Avi Kivity697fe2e2007-01-05 16:36:46 -08001859 unsigned i;
Avi Kivity697fe2e2007-01-05 16:36:46 -08001860
Xiao Guangrong38e3b2b2011-05-15 23:27:52 +08001861 for (i = 0; i < PT64_ENT_PER_PAGE; ++i)
1862 mmu_page_zap_pte(kvm, sp, sp->spt + i);
Avi Kivitya4360362007-01-05 16:36:45 -08001863}
1864
Avi Kivity4db35312007-11-21 15:28:32 +02001865static void kvm_mmu_put_page(struct kvm_mmu_page *sp, u64 *parent_pte)
Avi Kivitycea0f0e2007-01-05 16:36:43 -08001866{
Avi Kivity4db35312007-11-21 15:28:32 +02001867 mmu_page_remove_parent_pte(sp, parent_pte);
Avi Kivitya4360362007-01-05 16:36:45 -08001868}
1869
Avi Kivity31aa2b42008-07-11 17:59:46 +03001870static void kvm_mmu_unlink_parents(struct kvm *kvm, struct kvm_mmu_page *sp)
Avi Kivitya4360362007-01-05 16:36:45 -08001871{
1872 u64 *parent_pte;
1873
Xiao Guangrongbcdd9a92011-05-15 23:28:29 +08001874 while ((parent_pte = pte_list_next(&sp->parent_ptes, NULL)))
1875 drop_parent_pte(sp, parent_pte);
Avi Kivity31aa2b42008-07-11 17:59:46 +03001876}
1877
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001878static int mmu_zap_unsync_children(struct kvm *kvm,
Xiao Guangrong77758342010-06-04 21:53:54 +08001879 struct kvm_mmu_page *parent,
1880 struct list_head *invalid_list)
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001881{
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001882 int i, zapped = 0;
1883 struct mmu_page_path parents;
1884 struct kvm_mmu_pages pages;
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001885
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001886 if (parent->role.level == PT_PAGE_TABLE_LEVEL)
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001887 return 0;
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001888
1889 kvm_mmu_pages_init(parent, &parents, &pages);
1890 while (mmu_unsync_walk(parent, &pages)) {
1891 struct kvm_mmu_page *sp;
1892
1893 for_each_sp(pages, sp, parents, i) {
Xiao Guangrong77758342010-06-04 21:53:54 +08001894 kvm_mmu_prepare_zap_page(kvm, sp, invalid_list);
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001895 mmu_pages_clear_parents(&parents);
Xiao Guangrong77662e02010-04-16 16:34:42 +08001896 zapped++;
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001897 }
Marcelo Tosatti60c8aec2008-12-01 22:32:02 -02001898 kvm_mmu_pages_init(parent, &parents, &pages);
1899 }
1900
1901 return zapped;
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001902}
1903
Xiao Guangrong77758342010-06-04 21:53:54 +08001904static int kvm_mmu_prepare_zap_page(struct kvm *kvm, struct kvm_mmu_page *sp,
1905 struct list_head *invalid_list)
Avi Kivity31aa2b42008-07-11 17:59:46 +03001906{
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001907 int ret;
Avi Kivityf691fe12009-07-06 15:58:14 +03001908
Xiao Guangrong77758342010-06-04 21:53:54 +08001909 trace_kvm_mmu_prepare_zap_page(sp);
Avi Kivity31aa2b42008-07-11 17:59:46 +03001910 ++kvm->stat.mmu_shadow_zapped;
Xiao Guangrong77758342010-06-04 21:53:54 +08001911 ret = mmu_zap_unsync_children(kvm, sp, invalid_list);
Avi Kivity4db35312007-11-21 15:28:32 +02001912 kvm_mmu_page_unlink_children(kvm, sp);
Avi Kivity31aa2b42008-07-11 17:59:46 +03001913 kvm_mmu_unlink_parents(kvm, sp);
Avi Kivityf6e2c022009-01-11 13:02:10 +02001914 if (!sp->role.invalid && !sp->role.direct)
Avi Kivity5b5c6a52008-07-11 18:07:26 +03001915 unaccount_shadowed(kvm, sp->gfn);
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001916 if (sp->unsync)
1917 kvm_unlink_unsync_page(kvm, sp);
Avi Kivity4db35312007-11-21 15:28:32 +02001918 if (!sp->root_count) {
Gui Jianfeng54a4f022010-05-05 09:03:49 +08001919 /* Count self */
1920 ret++;
Xiao Guangrong77758342010-06-04 21:53:54 +08001921 list_move(&sp->link, invalid_list);
Xiao Guangrongaa6bd182011-07-12 03:26:40 +08001922 kvm_mod_used_mmu_pages(kvm, -1);
Marcelo Tosatti2e53d632008-02-20 14:47:24 -05001923 } else {
Avi Kivity5b5c6a52008-07-11 18:07:26 +03001924 list_move(&sp->link, &kvm->arch.active_mmu_pages);
Marcelo Tosatti2e53d632008-02-20 14:47:24 -05001925 kvm_reload_remote_mmus(kvm);
1926 }
Xiao Guangrong77758342010-06-04 21:53:54 +08001927
1928 sp->role.invalid = 1;
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03001929 return ret;
Avi Kivitya4360362007-01-05 16:36:45 -08001930}
1931
Xiao Guangrongc2a2ac22011-07-12 03:32:13 +08001932static void kvm_mmu_isolate_pages(struct list_head *invalid_list)
1933{
1934 struct kvm_mmu_page *sp;
1935
1936 list_for_each_entry(sp, invalid_list, link)
1937 kvm_mmu_isolate_page(sp);
1938}
1939
1940static void free_pages_rcu(struct rcu_head *head)
1941{
1942 struct kvm_mmu_page *next, *sp;
1943
1944 sp = container_of(head, struct kvm_mmu_page, rcu);
1945 while (sp) {
1946 if (!list_empty(&sp->link))
1947 next = list_first_entry(&sp->link,
1948 struct kvm_mmu_page, link);
1949 else
1950 next = NULL;
1951 kvm_mmu_free_page(sp);
1952 sp = next;
1953 }
1954}
1955
Xiao Guangrong77758342010-06-04 21:53:54 +08001956static void kvm_mmu_commit_zap_page(struct kvm *kvm,
1957 struct list_head *invalid_list)
1958{
1959 struct kvm_mmu_page *sp;
1960
1961 if (list_empty(invalid_list))
1962 return;
1963
1964 kvm_flush_remote_tlbs(kvm);
1965
Xiao Guangrongc2a2ac22011-07-12 03:32:13 +08001966 if (atomic_read(&kvm->arch.reader_counter)) {
1967 kvm_mmu_isolate_pages(invalid_list);
1968 sp = list_first_entry(invalid_list, struct kvm_mmu_page, link);
1969 list_del_init(invalid_list);
Xiao Guangrong4f022642011-07-12 03:34:24 +08001970
1971 trace_kvm_mmu_delay_free_pages(sp);
Xiao Guangrongc2a2ac22011-07-12 03:32:13 +08001972 call_rcu(&sp->rcu, free_pages_rcu);
1973 return;
1974 }
1975
Xiao Guangrong77758342010-06-04 21:53:54 +08001976 do {
1977 sp = list_first_entry(invalid_list, struct kvm_mmu_page, link);
1978 WARN_ON(!sp->role.invalid || sp->root_count);
Xiao Guangrongbd4c86e2011-07-12 03:27:14 +08001979 kvm_mmu_isolate_page(sp);
Xiao Guangrongaa6bd182011-07-12 03:26:40 +08001980 kvm_mmu_free_page(sp);
Xiao Guangrong77758342010-06-04 21:53:54 +08001981 } while (!list_empty(invalid_list));
1982
1983}
1984
Izik Eidus82ce2c92007-10-02 18:52:55 +02001985/*
1986 * Changing the number of mmu pages allocated to the vm
Dave Hansen49d5ca22010-08-19 18:11:28 -07001987 * Note: if goal_nr_mmu_pages is too small, you will get dead lock
Izik Eidus82ce2c92007-10-02 18:52:55 +02001988 */
Dave Hansen49d5ca22010-08-19 18:11:28 -07001989void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int goal_nr_mmu_pages)
Izik Eidus82ce2c92007-10-02 18:52:55 +02001990{
Xiao Guangrongd98ba052010-06-04 21:55:29 +08001991 LIST_HEAD(invalid_list);
Izik Eidus82ce2c92007-10-02 18:52:55 +02001992 /*
1993 * If we set the number of mmu pages to be smaller be than the
1994 * number of actived pages , we must to free some mmu pages before we
1995 * change the value
1996 */
1997
Dave Hansen49d5ca22010-08-19 18:11:28 -07001998 if (kvm->arch.n_used_mmu_pages > goal_nr_mmu_pages) {
1999 while (kvm->arch.n_used_mmu_pages > goal_nr_mmu_pages &&
Xiao Guangrong77662e02010-04-16 16:34:42 +08002000 !list_empty(&kvm->arch.active_mmu_pages)) {
Izik Eidus82ce2c92007-10-02 18:52:55 +02002001 struct kvm_mmu_page *page;
2002
Zhang Xiantaof05e70a2007-12-14 10:01:48 +08002003 page = container_of(kvm->arch.active_mmu_pages.prev,
Izik Eidus82ce2c92007-10-02 18:52:55 +02002004 struct kvm_mmu_page, link);
Xiaotian Feng80b63fa2010-08-24 10:31:07 +08002005 kvm_mmu_prepare_zap_page(kvm, page, &invalid_list);
Izik Eidus82ce2c92007-10-02 18:52:55 +02002006 }
Xiao Guangrongaa6bd182011-07-12 03:26:40 +08002007 kvm_mmu_commit_zap_page(kvm, &invalid_list);
Dave Hansen49d5ca22010-08-19 18:11:28 -07002008 goal_nr_mmu_pages = kvm->arch.n_used_mmu_pages;
Izik Eidus82ce2c92007-10-02 18:52:55 +02002009 }
Izik Eidus82ce2c92007-10-02 18:52:55 +02002010
Dave Hansen49d5ca22010-08-19 18:11:28 -07002011 kvm->arch.n_max_mmu_pages = goal_nr_mmu_pages;
Izik Eidus82ce2c92007-10-02 18:52:55 +02002012}
2013
Xiao Guangrong1cb3f3a2011-09-22 17:02:48 +08002014int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn)
Avi Kivitya4360362007-01-05 16:36:45 -08002015{
Avi Kivity4db35312007-11-21 15:28:32 +02002016 struct kvm_mmu_page *sp;
Xiao Guangrongf41d3352010-06-04 21:56:11 +08002017 struct hlist_node *node;
Xiao Guangrongd98ba052010-06-04 21:55:29 +08002018 LIST_HEAD(invalid_list);
Avi Kivitya4360362007-01-05 16:36:45 -08002019 int r;
2020
Xiao Guangrong9ad17b102010-08-28 19:19:42 +08002021 pgprintk("%s: looking for gfn %llx\n", __func__, gfn);
Avi Kivitya4360362007-01-05 16:36:45 -08002022 r = 0;
Xiao Guangrong1cb3f3a2011-09-22 17:02:48 +08002023 spin_lock(&kvm->mmu_lock);
Xiao Guangrongf41d3352010-06-04 21:56:11 +08002024 for_each_gfn_indirect_valid_sp(kvm, sp, gfn, node) {
Xiao Guangrong9ad17b102010-08-28 19:19:42 +08002025 pgprintk("%s: gfn %llx role %x\n", __func__, gfn,
Xiao Guangrong7ae680e2010-06-04 21:53:07 +08002026 sp->role.word);
2027 r = 1;
Xiao Guangrongf41d3352010-06-04 21:56:11 +08002028 kvm_mmu_prepare_zap_page(kvm, sp, &invalid_list);
Xiao Guangrong7ae680e2010-06-04 21:53:07 +08002029 }
Xiao Guangrongd98ba052010-06-04 21:55:29 +08002030 kvm_mmu_commit_zap_page(kvm, &invalid_list);
Xiao Guangrong1cb3f3a2011-09-22 17:02:48 +08002031 spin_unlock(&kvm->mmu_lock);
2032
Avi Kivitya4360362007-01-05 16:36:45 -08002033 return r;
Avi Kivitycea0f0e2007-01-05 16:36:43 -08002034}
Xiao Guangrong1cb3f3a2011-09-22 17:02:48 +08002035EXPORT_SYMBOL_GPL(kvm_mmu_unprotect_page);
Avi Kivitycea0f0e2007-01-05 16:36:43 -08002036
Avi Kivity38c335f2007-11-21 14:20:22 +02002037static void page_header_update_slot(struct kvm *kvm, void *pte, gfn_t gfn)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002038{
Marcelo Tosattibc6678a2009-12-23 14:35:21 -02002039 int slot = memslot_id(kvm, gfn);
Avi Kivity4db35312007-11-21 15:28:32 +02002040 struct kvm_mmu_page *sp = page_header(__pa(pte));
Avi Kivity6aa8b732006-12-10 02:21:36 -08002041
Sheng Yang291f26b2008-10-16 17:30:57 +08002042 __set_bit(slot, sp->slot_bitmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002043}
2044
Sheng Yang74be52e2008-10-09 16:01:56 +08002045/*
2046 * The function is based on mtrr_type_lookup() in
2047 * arch/x86/kernel/cpu/mtrr/generic.c
2048 */
2049static int get_mtrr_type(struct mtrr_state_type *mtrr_state,
2050 u64 start, u64 end)
2051{
2052 int i;
2053 u64 base, mask;
2054 u8 prev_match, curr_match;
2055 int num_var_ranges = KVM_NR_VAR_MTRR;
2056
2057 if (!mtrr_state->enabled)
2058 return 0xFF;
2059
2060 /* Make end inclusive end, instead of exclusive */
2061 end--;
2062
2063 /* Look in fixed ranges. Just return the type as per start */
2064 if (mtrr_state->have_fixed && (start < 0x100000)) {
2065 int idx;
2066
2067 if (start < 0x80000) {
2068 idx = 0;
2069 idx += (start >> 16);
2070 return mtrr_state->fixed_ranges[idx];
2071 } else if (start < 0xC0000) {
2072 idx = 1 * 8;
2073 idx += ((start - 0x80000) >> 14);
2074 return mtrr_state->fixed_ranges[idx];
2075 } else if (start < 0x1000000) {
2076 idx = 3 * 8;
2077 idx += ((start - 0xC0000) >> 12);
2078 return mtrr_state->fixed_ranges[idx];
2079 }
2080 }
2081
2082 /*
2083 * Look in variable ranges
2084 * Look of multiple ranges matching this address and pick type
2085 * as per MTRR precedence
2086 */
2087 if (!(mtrr_state->enabled & 2))
2088 return mtrr_state->def_type;
2089
2090 prev_match = 0xFF;
2091 for (i = 0; i < num_var_ranges; ++i) {
2092 unsigned short start_state, end_state;
2093
2094 if (!(mtrr_state->var_ranges[i].mask_lo & (1 << 11)))
2095 continue;
2096
2097 base = (((u64)mtrr_state->var_ranges[i].base_hi) << 32) +
2098 (mtrr_state->var_ranges[i].base_lo & PAGE_MASK);
2099 mask = (((u64)mtrr_state->var_ranges[i].mask_hi) << 32) +
2100 (mtrr_state->var_ranges[i].mask_lo & PAGE_MASK);
2101
2102 start_state = ((start & mask) == (base & mask));
2103 end_state = ((end & mask) == (base & mask));
2104 if (start_state != end_state)
2105 return 0xFE;
2106
2107 if ((start & mask) != (base & mask))
2108 continue;
2109
2110 curr_match = mtrr_state->var_ranges[i].base_lo & 0xff;
2111 if (prev_match == 0xFF) {
2112 prev_match = curr_match;
2113 continue;
2114 }
2115
2116 if (prev_match == MTRR_TYPE_UNCACHABLE ||
2117 curr_match == MTRR_TYPE_UNCACHABLE)
2118 return MTRR_TYPE_UNCACHABLE;
2119
2120 if ((prev_match == MTRR_TYPE_WRBACK &&
2121 curr_match == MTRR_TYPE_WRTHROUGH) ||
2122 (prev_match == MTRR_TYPE_WRTHROUGH &&
2123 curr_match == MTRR_TYPE_WRBACK)) {
2124 prev_match = MTRR_TYPE_WRTHROUGH;
2125 curr_match = MTRR_TYPE_WRTHROUGH;
2126 }
2127
2128 if (prev_match != curr_match)
2129 return MTRR_TYPE_UNCACHABLE;
2130 }
2131
2132 if (prev_match != 0xFF)
2133 return prev_match;
2134
2135 return mtrr_state->def_type;
2136}
2137
Sheng Yang4b12f0d2009-04-27 20:35:42 +08002138u8 kvm_get_guest_memory_type(struct kvm_vcpu *vcpu, gfn_t gfn)
Sheng Yang74be52e2008-10-09 16:01:56 +08002139{
2140 u8 mtrr;
2141
2142 mtrr = get_mtrr_type(&vcpu->arch.mtrr_state, gfn << PAGE_SHIFT,
2143 (gfn << PAGE_SHIFT) + PAGE_SIZE);
2144 if (mtrr == 0xfe || mtrr == 0xff)
2145 mtrr = MTRR_TYPE_WRBACK;
2146 return mtrr;
2147}
Sheng Yang4b12f0d2009-04-27 20:35:42 +08002148EXPORT_SYMBOL_GPL(kvm_get_guest_memory_type);
Sheng Yang74be52e2008-10-09 16:01:56 +08002149
Xiao Guangrong9cf5cf52010-05-24 15:40:07 +08002150static void __kvm_unsync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp)
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03002151{
Xiao Guangrong5e1b3dd2010-04-28 11:55:06 +08002152 trace_kvm_mmu_unsync_page(sp);
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03002153 ++vcpu->kvm->stat.mmu_unsync;
2154 sp->unsync = 1;
Marcelo Tosatti6cffe8c2008-12-01 22:32:04 -02002155
Xiao Guangrong6b184932010-04-16 21:29:17 +08002156 kvm_mmu_mark_parents_unsync(sp);
Xiao Guangrong9cf5cf52010-05-24 15:40:07 +08002157}
2158
2159static void kvm_unsync_pages(struct kvm_vcpu *vcpu, gfn_t gfn)
2160{
Xiao Guangrong9cf5cf52010-05-24 15:40:07 +08002161 struct kvm_mmu_page *s;
Xiao Guangrongf41d3352010-06-04 21:56:11 +08002162 struct hlist_node *node;
Xiao Guangrong9cf5cf52010-05-24 15:40:07 +08002163
Xiao Guangrongf41d3352010-06-04 21:56:11 +08002164 for_each_gfn_indirect_valid_sp(vcpu->kvm, s, gfn, node) {
Xiao Guangrong7ae680e2010-06-04 21:53:07 +08002165 if (s->unsync)
Xiao Guangrong9cf5cf52010-05-24 15:40:07 +08002166 continue;
2167 WARN_ON(s->role.level != PT_PAGE_TABLE_LEVEL);
2168 __kvm_unsync_page(vcpu, s);
2169 }
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03002170}
2171
2172static int mmu_need_write_protect(struct kvm_vcpu *vcpu, gfn_t gfn,
2173 bool can_unsync)
2174{
Xiao Guangrong9cf5cf52010-05-24 15:40:07 +08002175 struct kvm_mmu_page *s;
Xiao Guangrongf41d3352010-06-04 21:56:11 +08002176 struct hlist_node *node;
Xiao Guangrong9cf5cf52010-05-24 15:40:07 +08002177 bool need_unsync = false;
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03002178
Xiao Guangrongf41d3352010-06-04 21:56:11 +08002179 for_each_gfn_indirect_valid_sp(vcpu->kvm, s, gfn, node) {
Xiao Guangrong36a2e672010-06-30 16:02:02 +08002180 if (!can_unsync)
2181 return 1;
2182
Xiao Guangrong9cf5cf52010-05-24 15:40:07 +08002183 if (s->role.level != PT_PAGE_TABLE_LEVEL)
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03002184 return 1;
Xiao Guangrong9cf5cf52010-05-24 15:40:07 +08002185
2186 if (!need_unsync && !s->unsync) {
Xiao Guangrong36a2e672010-06-30 16:02:02 +08002187 if (!oos_shadow)
Xiao Guangrong9cf5cf52010-05-24 15:40:07 +08002188 return 1;
2189 need_unsync = true;
2190 }
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03002191 }
Xiao Guangrong9cf5cf52010-05-24 15:40:07 +08002192 if (need_unsync)
2193 kvm_unsync_pages(vcpu, gfn);
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03002194 return 0;
2195}
2196
Avi Kivityd555c332009-06-10 14:24:23 +03002197static int set_spte(struct kvm_vcpu *vcpu, u64 *sptep,
Marcelo Tosatti1e73f9d2008-09-23 13:18:30 -03002198 unsigned pte_access, int user_fault,
Xiao Guangrong640d9b02011-07-12 03:24:39 +08002199 int write_fault, int level,
Marcelo Tosattic2d0ee42009-04-05 14:54:47 -03002200 gfn_t gfn, pfn_t pfn, bool speculative,
Lai Jiangshan9bdbba12010-11-19 17:03:22 +08002201 bool can_unsync, bool host_writable)
Avi Kivity1c4f1fd2007-12-09 17:40:31 +02002202{
Xiao Guangrongb330aa02010-11-19 17:02:35 +08002203 u64 spte, entry = *sptep;
Marcelo Tosatti1e73f9d2008-09-23 13:18:30 -03002204 int ret = 0;
Sheng Yang64d4d522008-10-09 16:01:57 +08002205
Xiao Guangrongce88dec2011-07-12 03:33:44 +08002206 if (set_mmio_spte(sptep, gfn, pfn, pte_access))
2207 return 0;
2208
Marcelo Tosatti982c2562010-10-22 14:18:16 -02002209 spte = PT_PRESENT_MASK;
Avi Kivity947da532008-03-18 11:05:52 +02002210 if (!speculative)
Avi Kivity3201b5d2008-08-27 20:01:04 +03002211 spte |= shadow_accessed_mask;
Xiao Guangrong640d9b02011-07-12 03:24:39 +08002212
Sheng Yang7b523452008-04-25 21:13:50 +08002213 if (pte_access & ACC_EXEC_MASK)
2214 spte |= shadow_x_mask;
2215 else
2216 spte |= shadow_nx_mask;
Avi Kivity1c4f1fd2007-12-09 17:40:31 +02002217 if (pte_access & ACC_USER_MASK)
Sheng Yang7b523452008-04-25 21:13:50 +08002218 spte |= shadow_user_mask;
Joerg Roedel852e3c12009-07-27 16:30:44 +02002219 if (level > PT_PAGE_TABLE_LEVEL)
Marcelo Tosatti05da4552008-02-23 11:44:30 -03002220 spte |= PT_PAGE_SIZE_MASK;
Avi Kivityb0bc3ee2010-09-13 16:45:28 +02002221 if (tdp_enabled)
Sheng Yang4b12f0d2009-04-27 20:35:42 +08002222 spte |= kvm_x86_ops->get_mt_mask(vcpu, gfn,
2223 kvm_is_mmio_pfn(pfn));
Avi Kivity1c4f1fd2007-12-09 17:40:31 +02002224
Lai Jiangshan9bdbba12010-11-19 17:03:22 +08002225 if (host_writable)
Izik Eidus14032832009-09-23 21:47:17 +03002226 spte |= SPTE_HOST_WRITEABLE;
Xiao Guangrongf8e453b2010-12-23 16:09:29 +08002227 else
2228 pte_access &= ~ACC_WRITE_MASK;
Izik Eidus14032832009-09-23 21:47:17 +03002229
Anthony Liguori35149e22008-04-02 14:46:56 -05002230 spte |= (u64)pfn << PAGE_SHIFT;
Avi Kivity1c4f1fd2007-12-09 17:40:31 +02002231
2232 if ((pte_access & ACC_WRITE_MASK)
Joerg Roedelc5a78f2b2010-09-10 17:30:39 +02002233 || (!vcpu->arch.mmu.direct_map && write_fault
2234 && !is_write_protection(vcpu) && !user_fault)) {
Avi Kivity1c4f1fd2007-12-09 17:40:31 +02002235
Joerg Roedel852e3c12009-07-27 16:30:44 +02002236 if (level > PT_PAGE_TABLE_LEVEL &&
2237 has_wrprotected_page(vcpu->kvm, gfn, level)) {
Marcelo Tosatti38187c82008-09-23 13:18:32 -03002238 ret = 1;
Xiao Guangrongc3707952011-07-12 03:28:04 +08002239 drop_spte(vcpu->kvm, sptep);
Avi Kivitybe38d272010-06-06 14:31:27 +03002240 goto done;
Marcelo Tosatti38187c82008-09-23 13:18:32 -03002241 }
2242
Avi Kivity1c4f1fd2007-12-09 17:40:31 +02002243 spte |= PT_WRITABLE_MASK;
Avi Kivity1c4f1fd2007-12-09 17:40:31 +02002244
Joerg Roedelc5a78f2b2010-09-10 17:30:39 +02002245 if (!vcpu->arch.mmu.direct_map
Avi Kivity411c5882011-06-06 16:11:54 +03002246 && !(pte_access & ACC_WRITE_MASK)) {
Avi Kivity69325a12010-05-27 14:35:58 +03002247 spte &= ~PT_USER_MASK;
Avi Kivity411c5882011-06-06 16:11:54 +03002248 /*
2249 * If we converted a user page to a kernel page,
2250 * so that the kernel can write to it when cr0.wp=0,
2251 * then we should prevent the kernel from executing it
2252 * if SMEP is enabled.
2253 */
2254 if (kvm_read_cr4_bits(vcpu, X86_CR4_SMEP))
2255 spte |= PT64_NX_MASK;
2256 }
Avi Kivity69325a12010-05-27 14:35:58 +03002257
Marcelo Tosattiecc55892008-11-25 15:58:07 +01002258 /*
2259 * Optimization: for pte sync, if spte was writable the hash
2260 * lookup is unnecessary (and expensive). Write protection
2261 * is responsibility of mmu_get_page / kvm_sync_page.
2262 * Same reasoning can be applied to dirty page accounting.
2263 */
Takuya Yoshikawa8dae4442010-01-18 18:45:10 +09002264 if (!can_unsync && is_writable_pte(*sptep))
Marcelo Tosattiecc55892008-11-25 15:58:07 +01002265 goto set_pte;
2266
Marcelo Tosatti4731d4c2008-09-23 13:18:39 -03002267 if (mmu_need_write_protect(vcpu, gfn, can_unsync)) {
Xiao Guangrong9ad17b102010-08-28 19:19:42 +08002268 pgprintk("%s: found shadow page for %llx, marking ro\n",
Harvey Harrisonb8688d52008-03-03 12:59:56 -08002269 __func__, gfn);
Marcelo Tosatti1e73f9d2008-09-23 13:18:30 -03002270 ret = 1;
Avi Kivity1c4f1fd2007-12-09 17:40:31 +02002271 pte_access &= ~ACC_WRITE_MASK;
Takuya Yoshikawa8dae4442010-01-18 18:45:10 +09002272 if (is_writable_pte(spte))
Avi Kivity1c4f1fd2007-12-09 17:40:31 +02002273 spte &= ~PT_WRITABLE_MASK;
Avi Kivity1c4f1fd2007-12-09 17:40:31 +02002274 }
2275 }
2276
Avi Kivity1c4f1fd2007-12-09 17:40:31 +02002277 if (pte_access & ACC_WRITE_MASK)
2278 mark_page_dirty(vcpu->kvm, gfn);
2279
Marcelo Tosatti38187c82008-09-23 13:18:32 -03002280set_pte:
Xiao Guangrong1df9f2d2011-07-12 03:30:35 +08002281 mmu_spte_update(sptep, spte);
Xiao Guangrongb330aa02010-11-19 17:02:35 +08002282 /*
2283 * If we overwrite a writable spte with a read-only one we
2284 * should flush remote TLBs. Otherwise rmap_write_protect
2285 * will find a read-only spte, even though the writable spte
2286 * might be cached on a CPU's TLB.
2287 */
2288 if (is_writable_pte(entry) && !is_writable_pte(*sptep))
2289 kvm_flush_remote_tlbs(vcpu->kvm);
Avi Kivitybe38d272010-06-06 14:31:27 +03002290done:
Marcelo Tosatti1e73f9d2008-09-23 13:18:30 -03002291 return ret;
2292}
2293
Avi Kivityd555c332009-06-10 14:24:23 +03002294static void mmu_set_spte(struct kvm_vcpu *vcpu, u64 *sptep,
Marcelo Tosatti1e73f9d2008-09-23 13:18:30 -03002295 unsigned pt_access, unsigned pte_access,
Xiao Guangrong640d9b02011-07-12 03:24:39 +08002296 int user_fault, int write_fault,
Xiao Guangrongb90a0e62011-07-12 03:25:56 +08002297 int *emulate, int level, gfn_t gfn,
Izik Eidus14032832009-09-23 21:47:17 +03002298 pfn_t pfn, bool speculative,
Lai Jiangshan9bdbba12010-11-19 17:03:22 +08002299 bool host_writable)
Marcelo Tosatti1e73f9d2008-09-23 13:18:30 -03002300{
2301 int was_rmapped = 0;
Marcelo Tosatti53a27b32009-08-05 15:43:58 -03002302 int rmap_count;
Marcelo Tosatti1e73f9d2008-09-23 13:18:30 -03002303
2304 pgprintk("%s: spte %llx access %x write_fault %d"
Xiao Guangrong9ad17b102010-08-28 19:19:42 +08002305 " user_fault %d gfn %llx\n",
Avi Kivityd555c332009-06-10 14:24:23 +03002306 __func__, *sptep, pt_access,
Marcelo Tosatti1e73f9d2008-09-23 13:18:30 -03002307 write_fault, user_fault, gfn);
2308
Avi Kivityd555c332009-06-10 14:24:23 +03002309 if (is_rmap_spte(*sptep)) {
Marcelo Tosatti1e73f9d2008-09-23 13:18:30 -03002310 /*
2311 * If we overwrite a PTE page pointer with a 2MB PMD, unlink
2312 * the parent of the now unreachable PTE.
2313 */
Joerg Roedel852e3c12009-07-27 16:30:44 +02002314 if (level > PT_PAGE_TABLE_LEVEL &&
2315 !is_large_pte(*sptep)) {
Marcelo Tosatti1e73f9d2008-09-23 13:18:30 -03002316 struct kvm_mmu_page *child;
Avi Kivityd555c332009-06-10 14:24:23 +03002317 u64 pte = *sptep;
Marcelo Tosatti1e73f9d2008-09-23 13:18:30 -03002318
2319 child = page_header(pte & PT64_BASE_ADDR_MASK);
Xiao Guangrongbcdd9a92011-05-15 23:28:29 +08002320 drop_parent_pte(child, sptep);
Marcelo Tosatti3be22642010-05-28 09:44:59 -03002321 kvm_flush_remote_tlbs(vcpu->kvm);
Avi Kivityd555c332009-06-10 14:24:23 +03002322 } else if (pfn != spte_to_pfn(*sptep)) {
Xiao Guangrong9ad17b102010-08-28 19:19:42 +08002323 pgprintk("hfn old %llx new %llx\n",
Avi Kivityd555c332009-06-10 14:24:23 +03002324 spte_to_pfn(*sptep), pfn);
Xiao Guangrongc3707952011-07-12 03:28:04 +08002325 drop_spte(vcpu->kvm, sptep);
Xiao Guangrong91546352010-06-30 16:04:06 +08002326 kvm_flush_remote_tlbs(vcpu->kvm);
Joerg Roedel6bed6b92009-02-18 14:08:59 +01002327 } else
2328 was_rmapped = 1;
Marcelo Tosatti1e73f9d2008-09-23 13:18:30 -03002329 }
Joerg Roedel852e3c12009-07-27 16:30:44 +02002330
Avi Kivityd555c332009-06-10 14:24:23 +03002331 if (set_spte(vcpu, sptep, pte_access, user_fault, write_fault,
Xiao Guangrong640d9b02011-07-12 03:24:39 +08002332 level, gfn, pfn, speculative, true,
Lai Jiangshan9bdbba12010-11-19 17:03:22 +08002333 host_writable)) {
Marcelo Tosatti1e73f9d2008-09-23 13:18:30 -03002334 if (write_fault)
Xiao Guangrongb90a0e62011-07-12 03:25:56 +08002335 *emulate = 1;
Xiao Guangrong5304efd2010-06-08 20:05:57 +08002336 kvm_mmu_flush_tlb(vcpu);
Marcelo Tosattia378b4e2008-09-23 13:18:31 -03002337 }
Marcelo Tosatti1e73f9d2008-09-23 13:18:30 -03002338
Xiao Guangrongce88dec2011-07-12 03:33:44 +08002339 if (unlikely(is_mmio_spte(*sptep) && emulate))
2340 *emulate = 1;
2341
Avi Kivityd555c332009-06-10 14:24:23 +03002342 pgprintk("%s: setting spte %llx\n", __func__, *sptep);
Xiao Guangrong9ad17b102010-08-28 19:19:42 +08002343 pgprintk("instantiating %s PTE (%s) at %llx (%llx) addr %p\n",
Avi Kivityd555c332009-06-10 14:24:23 +03002344 is_large_pte(*sptep)? "2MB" : "4kB",
Joerg Roedela205bc12009-07-09 16:36:01 +02002345 *sptep & PT_PRESENT_MASK ?"RW":"R", gfn,
2346 *sptep, sptep);
Avi Kivityd555c332009-06-10 14:24:23 +03002347 if (!was_rmapped && is_large_pte(*sptep))
Marcelo Tosatti05da4552008-02-23 11:44:30 -03002348 ++vcpu->kvm->stat.lpages;
2349
Xiao Guangrongffb61bb2011-07-12 03:22:01 +08002350 if (is_shadow_present_pte(*sptep)) {
2351 page_header_update_slot(vcpu->kvm, sptep, gfn);
2352 if (!was_rmapped) {
2353 rmap_count = rmap_add(vcpu, sptep, gfn);
2354 if (rmap_count > RMAP_RECYCLE_THRESHOLD)
2355 rmap_recycle(vcpu, sptep, gfn);
2356 }
Avi Kivity1c4f1fd2007-12-09 17:40:31 +02002357 }
Xiao Guangrong9ed55202010-07-16 11:25:17 +08002358 kvm_release_pfn_clean(pfn);
Avi Kivity1c4f1fd2007-12-09 17:40:31 +02002359}
2360
Avi Kivity6aa8b732006-12-10 02:21:36 -08002361static void nonpaging_new_cr3(struct kvm_vcpu *vcpu)
2362{
2363}
2364
Xiao Guangrong957ed9e2010-08-22 19:12:48 +08002365static pfn_t pte_prefetch_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn,
2366 bool no_dirty_log)
2367{
2368 struct kvm_memory_slot *slot;
2369 unsigned long hva;
2370
Xiao Guangrong5d163b12011-03-09 15:43:00 +08002371 slot = gfn_to_memslot_dirty_bitmap(vcpu, gfn, no_dirty_log);
Xiao Guangrong957ed9e2010-08-22 19:12:48 +08002372 if (!slot) {
Xiao Guangrongfce92dc2011-07-12 03:28:54 +08002373 get_page(fault_page);
2374 return page_to_pfn(fault_page);
Xiao Guangrong957ed9e2010-08-22 19:12:48 +08002375 }
2376
2377 hva = gfn_to_hva_memslot(slot, gfn);
2378
2379 return hva_to_pfn_atomic(vcpu->kvm, hva);
2380}
2381
2382static int direct_pte_prefetch_many(struct kvm_vcpu *vcpu,
2383 struct kvm_mmu_page *sp,
2384 u64 *start, u64 *end)
2385{
2386 struct page *pages[PTE_PREFETCH_NUM];
2387 unsigned access = sp->role.access;
2388 int i, ret;
2389 gfn_t gfn;
2390
2391 gfn = kvm_mmu_page_get_gfn(sp, start - sp->spt);
Xiao Guangrong5d163b12011-03-09 15:43:00 +08002392 if (!gfn_to_memslot_dirty_bitmap(vcpu, gfn, access & ACC_WRITE_MASK))
Xiao Guangrong957ed9e2010-08-22 19:12:48 +08002393 return -1;
2394
2395 ret = gfn_to_page_many_atomic(vcpu->kvm, gfn, pages, end - start);
2396 if (ret <= 0)
2397 return -1;
2398
2399 for (i = 0; i < ret; i++, gfn++, start++)
2400 mmu_set_spte(vcpu, start, ACC_ALL,
Xiao Guangrong640d9b02011-07-12 03:24:39 +08002401 access, 0, 0, NULL,
Xiao Guangrong957ed9e2010-08-22 19:12:48 +08002402 sp->role.level, gfn,
2403 page_to_pfn(pages[i]), true, true);
2404
2405 return 0;
2406}
2407
2408static void __direct_pte_prefetch(struct kvm_vcpu *vcpu,
2409 struct kvm_mmu_page *sp, u64 *sptep)
2410{
2411 u64 *spte, *start = NULL;
2412 int i;
2413
2414 WARN_ON(!sp->role.direct);
2415
2416 i = (sptep - sp->spt) & ~(PTE_PREFETCH_NUM - 1);
2417 spte = sp->spt + i;
2418
2419 for (i = 0; i < PTE_PREFETCH_NUM; i++, spte++) {
Xiao Guangrongc3707952011-07-12 03:28:04 +08002420 if (is_shadow_present_pte(*spte) || spte == sptep) {
Xiao Guangrong957ed9e2010-08-22 19:12:48 +08002421 if (!start)
2422 continue;
2423 if (direct_pte_prefetch_many(vcpu, sp, start, spte) < 0)
2424 break;
2425 start = NULL;
2426 } else if (!start)
2427 start = spte;
2428 }
2429}
2430
2431static void direct_pte_prefetch(struct kvm_vcpu *vcpu, u64 *sptep)
2432{
2433 struct kvm_mmu_page *sp;
2434
2435 /*
2436 * Since it's no accessed bit on EPT, it's no way to
2437 * distinguish between actually accessed translations
2438 * and prefetched, so disable pte prefetch if EPT is
2439 * enabled.
2440 */
2441 if (!shadow_accessed_mask)
2442 return;
2443
2444 sp = page_header(__pa(sptep));
2445 if (sp->role.level > PT_PAGE_TABLE_LEVEL)
2446 return;
2447
2448 __direct_pte_prefetch(vcpu, sp, sptep);
2449}
2450
Joerg Roedel4d9976b2008-02-07 13:47:42 +01002451static int __direct_map(struct kvm_vcpu *vcpu, gpa_t v, int write,
Xiao Guangrong2ec47392010-12-07 10:34:42 +08002452 int map_writable, int level, gfn_t gfn, pfn_t pfn,
2453 bool prefault)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002454{
Avi Kivity9f652d22008-12-25 14:54:25 +02002455 struct kvm_shadow_walk_iterator iterator;
2456 struct kvm_mmu_page *sp;
Xiao Guangrongb90a0e62011-07-12 03:25:56 +08002457 int emulate = 0;
Avi Kivity9f652d22008-12-25 14:54:25 +02002458 gfn_t pseudo_gfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002459
Avi Kivity9f652d22008-12-25 14:54:25 +02002460 for_each_shadow_entry(vcpu, (u64)gfn << PAGE_SHIFT, iterator) {
Joerg Roedel852e3c12009-07-27 16:30:44 +02002461 if (iterator.level == level) {
Marcelo Tosatti612819c2010-10-22 14:18:18 -02002462 unsigned pte_access = ACC_ALL;
2463
Marcelo Tosatti612819c2010-10-22 14:18:18 -02002464 mmu_set_spte(vcpu, iterator.sptep, ACC_ALL, pte_access,
Xiao Guangrongb90a0e62011-07-12 03:25:56 +08002465 0, write, &emulate,
Xiao Guangrong2ec47392010-12-07 10:34:42 +08002466 level, gfn, pfn, prefault, map_writable);
Xiao Guangrong957ed9e2010-08-22 19:12:48 +08002467 direct_pte_prefetch(vcpu, iterator.sptep);
Avi Kivity9f652d22008-12-25 14:54:25 +02002468 ++vcpu->stat.pf_fixed;
2469 break;
2470 }
2471
Xiao Guangrongc3707952011-07-12 03:28:04 +08002472 if (!is_shadow_present_pte(*iterator.sptep)) {
Lai Jiangshanc9fa0b32010-05-26 16:48:25 +08002473 u64 base_addr = iterator.addr;
2474
2475 base_addr &= PT64_LVL_ADDR_MASK(iterator.level);
2476 pseudo_gfn = base_addr >> PAGE_SHIFT;
Avi Kivity9f652d22008-12-25 14:54:25 +02002477 sp = kvm_mmu_get_page(vcpu, pseudo_gfn, iterator.addr,
2478 iterator.level - 1,
2479 1, ACC_ALL, iterator.sptep);
2480 if (!sp) {
2481 pgprintk("nonpaging_map: ENOMEM\n");
2482 kvm_release_pfn_clean(pfn);
2483 return -ENOMEM;
2484 }
2485
Xiao Guangrong1df9f2d2011-07-12 03:30:35 +08002486 mmu_spte_set(iterator.sptep,
2487 __pa(sp->spt)
2488 | PT_PRESENT_MASK | PT_WRITABLE_MASK
2489 | shadow_user_mask | shadow_x_mask
2490 | shadow_accessed_mask);
Avi Kivity9f652d22008-12-25 14:54:25 +02002491 }
2492 }
Xiao Guangrongb90a0e62011-07-12 03:25:56 +08002493 return emulate;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002494}
2495
Huang Ying77db5cb2010-10-08 16:24:15 +08002496static void kvm_send_hwpoison_signal(unsigned long address, struct task_struct *tsk)
Huang Yingbf998152010-05-31 14:28:19 +08002497{
Huang Ying77db5cb2010-10-08 16:24:15 +08002498 siginfo_t info;
Huang Yingbf998152010-05-31 14:28:19 +08002499
Huang Ying77db5cb2010-10-08 16:24:15 +08002500 info.si_signo = SIGBUS;
2501 info.si_errno = 0;
2502 info.si_code = BUS_MCEERR_AR;
2503 info.si_addr = (void __user *)address;
2504 info.si_addr_lsb = PAGE_SHIFT;
2505
2506 send_sig_info(SIGBUS, &info, tsk);
Huang Yingbf998152010-05-31 14:28:19 +08002507}
2508
Xiao Guangrongd7c55202011-07-12 03:29:38 +08002509static int kvm_handle_bad_page(struct kvm_vcpu *vcpu, gfn_t gfn, pfn_t pfn)
Huang Yingbf998152010-05-31 14:28:19 +08002510{
2511 kvm_release_pfn_clean(pfn);
2512 if (is_hwpoison_pfn(pfn)) {
Xiao Guangrongbebb1062011-07-12 03:23:20 +08002513 kvm_send_hwpoison_signal(gfn_to_hva(vcpu->kvm, gfn), current);
Huang Yingbf998152010-05-31 14:28:19 +08002514 return 0;
Xiao Guangrongd7c55202011-07-12 03:29:38 +08002515 }
Gleb Natapovedba23e2010-07-07 20:16:45 +03002516
Xiao Guangrongd7c55202011-07-12 03:29:38 +08002517 return -EFAULT;
Huang Yingbf998152010-05-31 14:28:19 +08002518}
2519
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -08002520static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
2521 gfn_t *gfnp, pfn_t *pfnp, int *levelp)
2522{
2523 pfn_t pfn = *pfnp;
2524 gfn_t gfn = *gfnp;
2525 int level = *levelp;
2526
2527 /*
2528 * Check if it's a transparent hugepage. If this would be an
2529 * hugetlbfs page, level wouldn't be set to
2530 * PT_PAGE_TABLE_LEVEL and there would be no adjustment done
2531 * here.
2532 */
2533 if (!is_error_pfn(pfn) && !kvm_is_mmio_pfn(pfn) &&
2534 level == PT_PAGE_TABLE_LEVEL &&
2535 PageTransCompound(pfn_to_page(pfn)) &&
2536 !has_wrprotected_page(vcpu->kvm, gfn, PT_DIRECTORY_LEVEL)) {
2537 unsigned long mask;
2538 /*
2539 * mmu_notifier_retry was successful and we hold the
2540 * mmu_lock here, so the pmd can't become splitting
2541 * from under us, and in turn
2542 * __split_huge_page_refcount() can't run from under
2543 * us and we can safely transfer the refcount from
2544 * PG_tail to PG_head as we switch the pfn to tail to
2545 * head.
2546 */
2547 *levelp = level = PT_DIRECTORY_LEVEL;
2548 mask = KVM_PAGES_PER_HPAGE(level) - 1;
2549 VM_BUG_ON((gfn & mask) != (pfn & mask));
2550 if (pfn & mask) {
2551 gfn &= ~mask;
2552 *gfnp = gfn;
2553 kvm_release_pfn_clean(pfn);
2554 pfn &= ~mask;
2555 if (!get_page_unless_zero(pfn_to_page(pfn)))
2556 BUG();
2557 *pfnp = pfn;
2558 }
2559 }
2560}
2561
Xiao Guangrongd7c55202011-07-12 03:29:38 +08002562static bool mmu_invalid_pfn(pfn_t pfn)
2563{
Xiao Guangrongce88dec2011-07-12 03:33:44 +08002564 return unlikely(is_invalid_pfn(pfn));
Xiao Guangrongd7c55202011-07-12 03:29:38 +08002565}
2566
2567static bool handle_abnormal_pfn(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
2568 pfn_t pfn, unsigned access, int *ret_val)
2569{
2570 bool ret = true;
2571
2572 /* The pfn is invalid, report the error! */
2573 if (unlikely(is_invalid_pfn(pfn))) {
2574 *ret_val = kvm_handle_bad_page(vcpu, gfn, pfn);
2575 goto exit;
2576 }
2577
Xiao Guangrongce88dec2011-07-12 03:33:44 +08002578 if (unlikely(is_noslot_pfn(pfn)))
Xiao Guangrongd7c55202011-07-12 03:29:38 +08002579 vcpu_cache_mmio_info(vcpu, gva, gfn, access);
Xiao Guangrongd7c55202011-07-12 03:29:38 +08002580
2581 ret = false;
2582exit:
2583 return ret;
2584}
2585
Xiao Guangrong78b2c542010-12-07 10:48:06 +08002586static bool try_async_pf(struct kvm_vcpu *vcpu, bool prefault, gfn_t gfn,
Xiao Guangrong060c2ab2010-11-12 14:49:11 +08002587 gva_t gva, pfn_t *pfn, bool write, bool *writable);
2588
2589static int nonpaging_map(struct kvm_vcpu *vcpu, gva_t v, int write, gfn_t gfn,
Xiao Guangrong78b2c542010-12-07 10:48:06 +08002590 bool prefault)
Marcelo Tosatti10589a42007-12-20 19:18:22 -05002591{
2592 int r;
Joerg Roedel852e3c12009-07-27 16:30:44 +02002593 int level;
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -08002594 int force_pt_level;
Anthony Liguori35149e22008-04-02 14:46:56 -05002595 pfn_t pfn;
Andrea Arcangelie930bff2008-07-25 16:24:52 +02002596 unsigned long mmu_seq;
Marcelo Tosatti612819c2010-10-22 14:18:18 -02002597 bool map_writable;
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -05002598
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -08002599 force_pt_level = mapping_level_dirty_bitmap(vcpu, gfn);
2600 if (likely(!force_pt_level)) {
2601 level = mapping_level(vcpu, gfn);
2602 /*
2603 * This path builds a PAE pagetable - so we can map
2604 * 2mb pages at maximum. Therefore check if the level
2605 * is larger than that.
2606 */
2607 if (level > PT_DIRECTORY_LEVEL)
2608 level = PT_DIRECTORY_LEVEL;
Joerg Roedel852e3c12009-07-27 16:30:44 +02002609
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -08002610 gfn &= ~(KVM_PAGES_PER_HPAGE(level) - 1);
2611 } else
2612 level = PT_PAGE_TABLE_LEVEL;
Marcelo Tosatti05da4552008-02-23 11:44:30 -03002613
Andrea Arcangelie930bff2008-07-25 16:24:52 +02002614 mmu_seq = vcpu->kvm->mmu_notifier_seq;
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03002615 smp_rmb();
Xiao Guangrong060c2ab2010-11-12 14:49:11 +08002616
Xiao Guangrong78b2c542010-12-07 10:48:06 +08002617 if (try_async_pf(vcpu, prefault, gfn, v, &pfn, write, &map_writable))
Xiao Guangrong060c2ab2010-11-12 14:49:11 +08002618 return 0;
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -05002619
Xiao Guangrongd7c55202011-07-12 03:29:38 +08002620 if (handle_abnormal_pfn(vcpu, v, gfn, pfn, ACC_ALL, &r))
2621 return r;
Avi Kivityd196e342008-01-24 11:44:11 +02002622
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -05002623 spin_lock(&vcpu->kvm->mmu_lock);
Andrea Arcangelie930bff2008-07-25 16:24:52 +02002624 if (mmu_notifier_retry(vcpu, mmu_seq))
2625 goto out_unlock;
Avi Kivityeb787d12007-12-31 15:27:49 +02002626 kvm_mmu_free_some_pages(vcpu);
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -08002627 if (likely(!force_pt_level))
2628 transparent_hugepage_adjust(vcpu, &gfn, &pfn, &level);
Xiao Guangrong2ec47392010-12-07 10:34:42 +08002629 r = __direct_map(vcpu, v, write, map_writable, level, gfn, pfn,
2630 prefault);
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -05002631 spin_unlock(&vcpu->kvm->mmu_lock);
2632
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -05002633
Marcelo Tosatti10589a42007-12-20 19:18:22 -05002634 return r;
Andrea Arcangelie930bff2008-07-25 16:24:52 +02002635
2636out_unlock:
2637 spin_unlock(&vcpu->kvm->mmu_lock);
2638 kvm_release_pfn_clean(pfn);
2639 return 0;
Marcelo Tosatti10589a42007-12-20 19:18:22 -05002640}
2641
2642
Avi Kivity17ac10a2007-01-05 16:36:40 -08002643static void mmu_free_roots(struct kvm_vcpu *vcpu)
2644{
2645 int i;
Avi Kivity4db35312007-11-21 15:28:32 +02002646 struct kvm_mmu_page *sp;
Xiao Guangrongd98ba052010-06-04 21:55:29 +08002647 LIST_HEAD(invalid_list);
Avi Kivity17ac10a2007-01-05 16:36:40 -08002648
Zhang Xiantaoad312c72007-12-13 23:50:52 +08002649 if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
Avi Kivity7b53aa52007-06-05 12:17:03 +03002650 return;
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -05002651 spin_lock(&vcpu->kvm->mmu_lock);
Joerg Roedel81407ca2010-09-10 17:31:00 +02002652 if (vcpu->arch.mmu.shadow_root_level == PT64_ROOT_LEVEL &&
2653 (vcpu->arch.mmu.root_level == PT64_ROOT_LEVEL ||
2654 vcpu->arch.mmu.direct_map)) {
Zhang Xiantaoad312c72007-12-13 23:50:52 +08002655 hpa_t root = vcpu->arch.mmu.root_hpa;
Avi Kivity17ac10a2007-01-05 16:36:40 -08002656
Avi Kivity4db35312007-11-21 15:28:32 +02002657 sp = page_header(root);
2658 --sp->root_count;
Xiao Guangrongd98ba052010-06-04 21:55:29 +08002659 if (!sp->root_count && sp->role.invalid) {
2660 kvm_mmu_prepare_zap_page(vcpu->kvm, sp, &invalid_list);
2661 kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
2662 }
Zhang Xiantaoad312c72007-12-13 23:50:52 +08002663 vcpu->arch.mmu.root_hpa = INVALID_PAGE;
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -05002664 spin_unlock(&vcpu->kvm->mmu_lock);
Avi Kivity17ac10a2007-01-05 16:36:40 -08002665 return;
2666 }
Avi Kivity17ac10a2007-01-05 16:36:40 -08002667 for (i = 0; i < 4; ++i) {
Zhang Xiantaoad312c72007-12-13 23:50:52 +08002668 hpa_t root = vcpu->arch.mmu.pae_root[i];
Avi Kivity17ac10a2007-01-05 16:36:40 -08002669
Avi Kivity417726a2007-04-12 17:35:58 +03002670 if (root) {
Avi Kivity417726a2007-04-12 17:35:58 +03002671 root &= PT64_BASE_ADDR_MASK;
Avi Kivity4db35312007-11-21 15:28:32 +02002672 sp = page_header(root);
2673 --sp->root_count;
Marcelo Tosatti2e53d632008-02-20 14:47:24 -05002674 if (!sp->root_count && sp->role.invalid)
Xiao Guangrongd98ba052010-06-04 21:55:29 +08002675 kvm_mmu_prepare_zap_page(vcpu->kvm, sp,
2676 &invalid_list);
Avi Kivity417726a2007-04-12 17:35:58 +03002677 }
Zhang Xiantaoad312c72007-12-13 23:50:52 +08002678 vcpu->arch.mmu.pae_root[i] = INVALID_PAGE;
Avi Kivity17ac10a2007-01-05 16:36:40 -08002679 }
Xiao Guangrongd98ba052010-06-04 21:55:29 +08002680 kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -05002681 spin_unlock(&vcpu->kvm->mmu_lock);
Zhang Xiantaoad312c72007-12-13 23:50:52 +08002682 vcpu->arch.mmu.root_hpa = INVALID_PAGE;
Avi Kivity17ac10a2007-01-05 16:36:40 -08002683}
2684
Marcelo Tosatti8986ecc2009-05-12 18:55:45 -03002685static int mmu_check_root(struct kvm_vcpu *vcpu, gfn_t root_gfn)
2686{
2687 int ret = 0;
2688
2689 if (!kvm_is_visible_gfn(vcpu->kvm, root_gfn)) {
Avi Kivitya8eeb042010-05-10 12:34:53 +03002690 kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu);
Marcelo Tosatti8986ecc2009-05-12 18:55:45 -03002691 ret = 1;
2692 }
2693
2694 return ret;
2695}
2696
Joerg Roedel651dd372010-09-10 17:30:59 +02002697static int mmu_alloc_direct_roots(struct kvm_vcpu *vcpu)
2698{
2699 struct kvm_mmu_page *sp;
Avi Kivity7ebaf152010-10-03 18:51:39 +02002700 unsigned i;
Joerg Roedel651dd372010-09-10 17:30:59 +02002701
2702 if (vcpu->arch.mmu.shadow_root_level == PT64_ROOT_LEVEL) {
2703 spin_lock(&vcpu->kvm->mmu_lock);
2704 kvm_mmu_free_some_pages(vcpu);
2705 sp = kvm_mmu_get_page(vcpu, 0, 0, PT64_ROOT_LEVEL,
2706 1, ACC_ALL, NULL);
2707 ++sp->root_count;
2708 spin_unlock(&vcpu->kvm->mmu_lock);
2709 vcpu->arch.mmu.root_hpa = __pa(sp->spt);
2710 } else if (vcpu->arch.mmu.shadow_root_level == PT32E_ROOT_LEVEL) {
2711 for (i = 0; i < 4; ++i) {
2712 hpa_t root = vcpu->arch.mmu.pae_root[i];
2713
2714 ASSERT(!VALID_PAGE(root));
2715 spin_lock(&vcpu->kvm->mmu_lock);
2716 kvm_mmu_free_some_pages(vcpu);
Avi Kivity649497d2010-12-28 12:09:07 +02002717 sp = kvm_mmu_get_page(vcpu, i << (30 - PAGE_SHIFT),
2718 i << 30,
Joerg Roedel651dd372010-09-10 17:30:59 +02002719 PT32_ROOT_LEVEL, 1, ACC_ALL,
2720 NULL);
2721 root = __pa(sp->spt);
2722 ++sp->root_count;
2723 spin_unlock(&vcpu->kvm->mmu_lock);
2724 vcpu->arch.mmu.pae_root[i] = root | PT_PRESENT_MASK;
Joerg Roedel651dd372010-09-10 17:30:59 +02002725 }
Xiao Guangrong62927572010-09-27 18:02:12 +08002726 vcpu->arch.mmu.root_hpa = __pa(vcpu->arch.mmu.pae_root);
Joerg Roedel651dd372010-09-10 17:30:59 +02002727 } else
2728 BUG();
2729
2730 return 0;
2731}
2732
2733static int mmu_alloc_shadow_roots(struct kvm_vcpu *vcpu)
Avi Kivity17ac10a2007-01-05 16:36:40 -08002734{
Avi Kivity4db35312007-11-21 15:28:32 +02002735 struct kvm_mmu_page *sp;
Joerg Roedel81407ca2010-09-10 17:31:00 +02002736 u64 pdptr, pm_mask;
2737 gfn_t root_gfn;
2738 int i;
Avi Kivity3bb65a22007-01-05 16:36:51 -08002739
Joerg Roedel5777ed32010-09-10 17:30:42 +02002740 root_gfn = vcpu->arch.mmu.get_cr3(vcpu) >> PAGE_SHIFT;
Avi Kivity17ac10a2007-01-05 16:36:40 -08002741
Joerg Roedel651dd372010-09-10 17:30:59 +02002742 if (mmu_check_root(vcpu, root_gfn))
2743 return 1;
2744
2745 /*
2746 * Do we shadow a long mode page table? If so we need to
2747 * write-protect the guests page table root.
2748 */
2749 if (vcpu->arch.mmu.root_level == PT64_ROOT_LEVEL) {
Zhang Xiantaoad312c72007-12-13 23:50:52 +08002750 hpa_t root = vcpu->arch.mmu.root_hpa;
Avi Kivity17ac10a2007-01-05 16:36:40 -08002751
2752 ASSERT(!VALID_PAGE(root));
Joerg Roedel651dd372010-09-10 17:30:59 +02002753
Avi Kivity8facbbf2010-05-04 12:58:32 +03002754 spin_lock(&vcpu->kvm->mmu_lock);
Marcelo Tosatti24955b62010-05-12 21:00:35 -03002755 kvm_mmu_free_some_pages(vcpu);
Joerg Roedel651dd372010-09-10 17:30:59 +02002756 sp = kvm_mmu_get_page(vcpu, root_gfn, 0, PT64_ROOT_LEVEL,
2757 0, ACC_ALL, NULL);
Avi Kivity4db35312007-11-21 15:28:32 +02002758 root = __pa(sp->spt);
2759 ++sp->root_count;
Avi Kivity8facbbf2010-05-04 12:58:32 +03002760 spin_unlock(&vcpu->kvm->mmu_lock);
Zhang Xiantaoad312c72007-12-13 23:50:52 +08002761 vcpu->arch.mmu.root_hpa = root;
Marcelo Tosatti8986ecc2009-05-12 18:55:45 -03002762 return 0;
Avi Kivity17ac10a2007-01-05 16:36:40 -08002763 }
Joerg Roedelf87f9282010-09-02 17:29:45 +02002764
Joerg Roedel651dd372010-09-10 17:30:59 +02002765 /*
2766 * We shadow a 32 bit page table. This may be a legacy 2-level
Joerg Roedel81407ca2010-09-10 17:31:00 +02002767 * or a PAE 3-level page table. In either case we need to be aware that
2768 * the shadow page table may be a PAE or a long mode page table.
Joerg Roedel651dd372010-09-10 17:30:59 +02002769 */
Joerg Roedel81407ca2010-09-10 17:31:00 +02002770 pm_mask = PT_PRESENT_MASK;
2771 if (vcpu->arch.mmu.shadow_root_level == PT64_ROOT_LEVEL)
2772 pm_mask |= PT_ACCESSED_MASK | PT_WRITABLE_MASK | PT_USER_MASK;
2773
Avi Kivity17ac10a2007-01-05 16:36:40 -08002774 for (i = 0; i < 4; ++i) {
Zhang Xiantaoad312c72007-12-13 23:50:52 +08002775 hpa_t root = vcpu->arch.mmu.pae_root[i];
Avi Kivity17ac10a2007-01-05 16:36:40 -08002776
2777 ASSERT(!VALID_PAGE(root));
Zhang Xiantaoad312c72007-12-13 23:50:52 +08002778 if (vcpu->arch.mmu.root_level == PT32E_ROOT_LEVEL) {
Avi Kivitye4e517b2011-07-28 11:36:17 +03002779 pdptr = vcpu->arch.mmu.get_pdptr(vcpu, i);
Avi Kivity43a37952009-06-10 14:12:05 +03002780 if (!is_present_gpte(pdptr)) {
Zhang Xiantaoad312c72007-12-13 23:50:52 +08002781 vcpu->arch.mmu.pae_root[i] = 0;
Avi Kivity417726a2007-04-12 17:35:58 +03002782 continue;
2783 }
Avi Kivity6de4f3ad2009-05-31 22:58:47 +03002784 root_gfn = pdptr >> PAGE_SHIFT;
Joerg Roedelf87f9282010-09-02 17:29:45 +02002785 if (mmu_check_root(vcpu, root_gfn))
2786 return 1;
Eric Northup5a7388c2010-04-26 17:00:05 -07002787 }
Avi Kivity8facbbf2010-05-04 12:58:32 +03002788 spin_lock(&vcpu->kvm->mmu_lock);
Marcelo Tosatti24955b62010-05-12 21:00:35 -03002789 kvm_mmu_free_some_pages(vcpu);
Avi Kivity4db35312007-11-21 15:28:32 +02002790 sp = kvm_mmu_get_page(vcpu, root_gfn, i << 30,
Joerg Roedel651dd372010-09-10 17:30:59 +02002791 PT32_ROOT_LEVEL, 0,
Avi Kivityf7d9c7b2008-02-26 22:12:10 +02002792 ACC_ALL, NULL);
Avi Kivity4db35312007-11-21 15:28:32 +02002793 root = __pa(sp->spt);
2794 ++sp->root_count;
Avi Kivity8facbbf2010-05-04 12:58:32 +03002795 spin_unlock(&vcpu->kvm->mmu_lock);
2796
Joerg Roedel81407ca2010-09-10 17:31:00 +02002797 vcpu->arch.mmu.pae_root[i] = root | pm_mask;
Avi Kivity17ac10a2007-01-05 16:36:40 -08002798 }
Xiao Guangrong62927572010-09-27 18:02:12 +08002799 vcpu->arch.mmu.root_hpa = __pa(vcpu->arch.mmu.pae_root);
Joerg Roedel81407ca2010-09-10 17:31:00 +02002800
2801 /*
2802 * If we shadow a 32 bit page table with a long mode page
2803 * table we enter this path.
2804 */
2805 if (vcpu->arch.mmu.shadow_root_level == PT64_ROOT_LEVEL) {
2806 if (vcpu->arch.mmu.lm_root == NULL) {
2807 /*
2808 * The additional page necessary for this is only
2809 * allocated on demand.
2810 */
2811
2812 u64 *lm_root;
2813
2814 lm_root = (void*)get_zeroed_page(GFP_KERNEL);
2815 if (lm_root == NULL)
2816 return 1;
2817
2818 lm_root[0] = __pa(vcpu->arch.mmu.pae_root) | pm_mask;
2819
2820 vcpu->arch.mmu.lm_root = lm_root;
2821 }
2822
2823 vcpu->arch.mmu.root_hpa = __pa(vcpu->arch.mmu.lm_root);
2824 }
2825
Marcelo Tosatti8986ecc2009-05-12 18:55:45 -03002826 return 0;
Avi Kivity17ac10a2007-01-05 16:36:40 -08002827}
2828
Joerg Roedel651dd372010-09-10 17:30:59 +02002829static int mmu_alloc_roots(struct kvm_vcpu *vcpu)
2830{
2831 if (vcpu->arch.mmu.direct_map)
2832 return mmu_alloc_direct_roots(vcpu);
2833 else
2834 return mmu_alloc_shadow_roots(vcpu);
2835}
2836
Marcelo Tosatti0ba73cd2008-09-23 13:18:34 -03002837static void mmu_sync_roots(struct kvm_vcpu *vcpu)
2838{
2839 int i;
2840 struct kvm_mmu_page *sp;
2841
Joerg Roedel81407ca2010-09-10 17:31:00 +02002842 if (vcpu->arch.mmu.direct_map)
2843 return;
2844
Marcelo Tosatti0ba73cd2008-09-23 13:18:34 -03002845 if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
2846 return;
Xiao Guangrong69030742010-09-27 18:09:29 +08002847
Xiao Guangrongbebb1062011-07-12 03:23:20 +08002848 vcpu_clear_mmio_info(vcpu, ~0ul);
Xiao Guangrong69030742010-09-27 18:09:29 +08002849 trace_kvm_mmu_audit(vcpu, AUDIT_PRE_SYNC);
Joerg Roedel81407ca2010-09-10 17:31:00 +02002850 if (vcpu->arch.mmu.root_level == PT64_ROOT_LEVEL) {
Marcelo Tosatti0ba73cd2008-09-23 13:18:34 -03002851 hpa_t root = vcpu->arch.mmu.root_hpa;
2852 sp = page_header(root);
2853 mmu_sync_children(vcpu, sp);
Xiao Guangrong5054c0d2010-11-12 14:46:08 +08002854 trace_kvm_mmu_audit(vcpu, AUDIT_POST_SYNC);
Marcelo Tosatti0ba73cd2008-09-23 13:18:34 -03002855 return;
2856 }
2857 for (i = 0; i < 4; ++i) {
2858 hpa_t root = vcpu->arch.mmu.pae_root[i];
2859
Marcelo Tosatti8986ecc2009-05-12 18:55:45 -03002860 if (root && VALID_PAGE(root)) {
Marcelo Tosatti0ba73cd2008-09-23 13:18:34 -03002861 root &= PT64_BASE_ADDR_MASK;
2862 sp = page_header(root);
2863 mmu_sync_children(vcpu, sp);
2864 }
2865 }
Xiao Guangrong69030742010-09-27 18:09:29 +08002866 trace_kvm_mmu_audit(vcpu, AUDIT_POST_SYNC);
Marcelo Tosatti0ba73cd2008-09-23 13:18:34 -03002867}
2868
2869void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu)
2870{
2871 spin_lock(&vcpu->kvm->mmu_lock);
2872 mmu_sync_roots(vcpu);
2873 spin_unlock(&vcpu->kvm->mmu_lock);
2874}
2875
Gleb Natapov1871c602010-02-10 14:21:32 +02002876static gpa_t nonpaging_gva_to_gpa(struct kvm_vcpu *vcpu, gva_t vaddr,
Avi Kivityab9ae312010-11-22 17:53:26 +02002877 u32 access, struct x86_exception *exception)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002878{
Avi Kivityab9ae312010-11-22 17:53:26 +02002879 if (exception)
2880 exception->error_code = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002881 return vaddr;
2882}
2883
Joerg Roedel6539e732010-09-10 17:30:50 +02002884static gpa_t nonpaging_gva_to_gpa_nested(struct kvm_vcpu *vcpu, gva_t vaddr,
Avi Kivityab9ae312010-11-22 17:53:26 +02002885 u32 access,
2886 struct x86_exception *exception)
Joerg Roedel6539e732010-09-10 17:30:50 +02002887{
Avi Kivityab9ae312010-11-22 17:53:26 +02002888 if (exception)
2889 exception->error_code = 0;
Joerg Roedel6539e732010-09-10 17:30:50 +02002890 return vcpu->arch.nested_mmu.translate_gpa(vcpu, vaddr, access);
2891}
2892
Xiao Guangrongce88dec2011-07-12 03:33:44 +08002893static bool quickly_check_mmio_pf(struct kvm_vcpu *vcpu, u64 addr, bool direct)
2894{
2895 if (direct)
2896 return vcpu_match_mmio_gpa(vcpu, addr);
2897
2898 return vcpu_match_mmio_gva(vcpu, addr);
2899}
2900
2901
2902/*
2903 * On direct hosts, the last spte is only allows two states
2904 * for mmio page fault:
2905 * - It is the mmio spte
2906 * - It is zapped or it is being zapped.
2907 *
2908 * This function completely checks the spte when the last spte
2909 * is not the mmio spte.
2910 */
2911static bool check_direct_spte_mmio_pf(u64 spte)
2912{
2913 return __check_direct_spte_mmio_pf(spte);
2914}
2915
2916static u64 walk_shadow_page_get_mmio_spte(struct kvm_vcpu *vcpu, u64 addr)
2917{
2918 struct kvm_shadow_walk_iterator iterator;
2919 u64 spte = 0ull;
2920
2921 walk_shadow_page_lockless_begin(vcpu);
2922 for_each_shadow_entry_lockless(vcpu, addr, iterator, spte)
2923 if (!is_shadow_present_pte(spte))
2924 break;
2925 walk_shadow_page_lockless_end(vcpu);
2926
2927 return spte;
2928}
2929
2930/*
2931 * If it is a real mmio page fault, return 1 and emulat the instruction
2932 * directly, return 0 to let CPU fault again on the address, -1 is
2933 * returned if bug is detected.
2934 */
2935int handle_mmio_page_fault_common(struct kvm_vcpu *vcpu, u64 addr, bool direct)
2936{
2937 u64 spte;
2938
2939 if (quickly_check_mmio_pf(vcpu, addr, direct))
2940 return 1;
2941
2942 spte = walk_shadow_page_get_mmio_spte(vcpu, addr);
2943
2944 if (is_mmio_spte(spte)) {
2945 gfn_t gfn = get_mmio_spte_gfn(spte);
2946 unsigned access = get_mmio_spte_access(spte);
2947
2948 if (direct)
2949 addr = 0;
Xiao Guangrong4f022642011-07-12 03:34:24 +08002950
2951 trace_handle_mmio_page_fault(addr, gfn, access);
Xiao Guangrongce88dec2011-07-12 03:33:44 +08002952 vcpu_cache_mmio_info(vcpu, addr, gfn, access);
2953 return 1;
2954 }
2955
2956 /*
2957 * It's ok if the gva is remapped by other cpus on shadow guest,
2958 * it's a BUG if the gfn is not a mmio page.
2959 */
2960 if (direct && !check_direct_spte_mmio_pf(spte))
2961 return -1;
2962
2963 /*
2964 * If the page table is zapped by other cpus, let CPU fault again on
2965 * the address.
2966 */
2967 return 0;
2968}
2969EXPORT_SYMBOL_GPL(handle_mmio_page_fault_common);
2970
2971static int handle_mmio_page_fault(struct kvm_vcpu *vcpu, u64 addr,
2972 u32 error_code, bool direct)
2973{
2974 int ret;
2975
2976 ret = handle_mmio_page_fault_common(vcpu, addr, direct);
2977 WARN_ON(ret < 0);
2978 return ret;
2979}
2980
Avi Kivity6aa8b732006-12-10 02:21:36 -08002981static int nonpaging_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
Xiao Guangrong78b2c542010-12-07 10:48:06 +08002982 u32 error_code, bool prefault)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002983{
Avi Kivitye8332402007-12-09 18:43:00 +02002984 gfn_t gfn;
Avi Kivitye2dec932007-01-05 16:36:54 -08002985 int r;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002986
Harvey Harrisonb8688d52008-03-03 12:59:56 -08002987 pgprintk("%s: gva %lx error %x\n", __func__, gva, error_code);
Xiao Guangrongce88dec2011-07-12 03:33:44 +08002988
2989 if (unlikely(error_code & PFERR_RSVD_MASK))
2990 return handle_mmio_page_fault(vcpu, gva, error_code, true);
2991
Avi Kivitye2dec932007-01-05 16:36:54 -08002992 r = mmu_topup_memory_caches(vcpu);
2993 if (r)
2994 return r;
Avi Kivity714b93d2007-01-05 16:36:53 -08002995
Avi Kivity6aa8b732006-12-10 02:21:36 -08002996 ASSERT(vcpu);
Zhang Xiantaoad312c72007-12-13 23:50:52 +08002997 ASSERT(VALID_PAGE(vcpu->arch.mmu.root_hpa));
Avi Kivity6aa8b732006-12-10 02:21:36 -08002998
Avi Kivitye8332402007-12-09 18:43:00 +02002999 gfn = gva >> PAGE_SHIFT;
Avi Kivity6aa8b732006-12-10 02:21:36 -08003000
Avi Kivitye8332402007-12-09 18:43:00 +02003001 return nonpaging_map(vcpu, gva & PAGE_MASK,
Xiao Guangrong78b2c542010-12-07 10:48:06 +08003002 error_code & PFERR_WRITE_MASK, gfn, prefault);
Avi Kivity6aa8b732006-12-10 02:21:36 -08003003}
3004
Jan Kiszka7e1fbea2010-10-20 15:18:02 +02003005static int kvm_arch_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn)
Gleb Natapovaf585b92010-10-14 11:22:46 +02003006{
3007 struct kvm_arch_async_pf arch;
Xiao Guangrongfb67e142010-12-07 10:35:25 +08003008
Gleb Natapov7c907052010-10-14 11:22:53 +02003009 arch.token = (vcpu->arch.apf.id++ << 12) | vcpu->vcpu_id;
Gleb Natapovaf585b92010-10-14 11:22:46 +02003010 arch.gfn = gfn;
Xiao Guangrongc4806ac2010-11-12 14:49:55 +08003011 arch.direct_map = vcpu->arch.mmu.direct_map;
Xiao Guangrongfb67e142010-12-07 10:35:25 +08003012 arch.cr3 = vcpu->arch.mmu.get_cr3(vcpu);
Gleb Natapovaf585b92010-10-14 11:22:46 +02003013
3014 return kvm_setup_async_pf(vcpu, gva, gfn, &arch);
3015}
3016
3017static bool can_do_async_pf(struct kvm_vcpu *vcpu)
3018{
3019 if (unlikely(!irqchip_in_kernel(vcpu->kvm) ||
3020 kvm_event_needs_reinjection(vcpu)))
3021 return false;
3022
3023 return kvm_x86_ops->interrupt_allowed(vcpu);
3024}
3025
Xiao Guangrong78b2c542010-12-07 10:48:06 +08003026static bool try_async_pf(struct kvm_vcpu *vcpu, bool prefault, gfn_t gfn,
Marcelo Tosatti612819c2010-10-22 14:18:18 -02003027 gva_t gva, pfn_t *pfn, bool write, bool *writable)
Gleb Natapovaf585b92010-10-14 11:22:46 +02003028{
3029 bool async;
3030
Marcelo Tosatti612819c2010-10-22 14:18:18 -02003031 *pfn = gfn_to_pfn_async(vcpu->kvm, gfn, &async, write, writable);
Gleb Natapovaf585b92010-10-14 11:22:46 +02003032
3033 if (!async)
3034 return false; /* *pfn has correct page already */
3035
3036 put_page(pfn_to_page(*pfn));
3037
Xiao Guangrong78b2c542010-12-07 10:48:06 +08003038 if (!prefault && can_do_async_pf(vcpu)) {
Xiao Guangrongc9b263d2010-11-01 16:58:43 +08003039 trace_kvm_try_async_get_page(gva, gfn);
Gleb Natapovaf585b92010-10-14 11:22:46 +02003040 if (kvm_find_async_pf_gfn(vcpu, gfn)) {
3041 trace_kvm_async_pf_doublefault(gva, gfn);
3042 kvm_make_request(KVM_REQ_APF_HALT, vcpu);
3043 return true;
3044 } else if (kvm_arch_setup_async_pf(vcpu, gva, gfn))
3045 return true;
3046 }
3047
Marcelo Tosatti612819c2010-10-22 14:18:18 -02003048 *pfn = gfn_to_pfn_prot(vcpu->kvm, gfn, write, writable);
Gleb Natapovaf585b92010-10-14 11:22:46 +02003049
3050 return false;
3051}
3052
Gleb Natapov56028d02010-10-17 18:13:42 +02003053static int tdp_page_fault(struct kvm_vcpu *vcpu, gva_t gpa, u32 error_code,
Xiao Guangrong78b2c542010-12-07 10:48:06 +08003054 bool prefault)
Joerg Roedelfb72d162008-02-07 13:47:44 +01003055{
Anthony Liguori35149e22008-04-02 14:46:56 -05003056 pfn_t pfn;
Joerg Roedelfb72d162008-02-07 13:47:44 +01003057 int r;
Joerg Roedel852e3c12009-07-27 16:30:44 +02003058 int level;
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -08003059 int force_pt_level;
Marcelo Tosatti05da4552008-02-23 11:44:30 -03003060 gfn_t gfn = gpa >> PAGE_SHIFT;
Andrea Arcangelie930bff2008-07-25 16:24:52 +02003061 unsigned long mmu_seq;
Marcelo Tosatti612819c2010-10-22 14:18:18 -02003062 int write = error_code & PFERR_WRITE_MASK;
3063 bool map_writable;
Joerg Roedelfb72d162008-02-07 13:47:44 +01003064
3065 ASSERT(vcpu);
3066 ASSERT(VALID_PAGE(vcpu->arch.mmu.root_hpa));
3067
Xiao Guangrongce88dec2011-07-12 03:33:44 +08003068 if (unlikely(error_code & PFERR_RSVD_MASK))
3069 return handle_mmio_page_fault(vcpu, gpa, error_code, true);
3070
Joerg Roedelfb72d162008-02-07 13:47:44 +01003071 r = mmu_topup_memory_caches(vcpu);
3072 if (r)
3073 return r;
3074
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -08003075 force_pt_level = mapping_level_dirty_bitmap(vcpu, gfn);
3076 if (likely(!force_pt_level)) {
3077 level = mapping_level(vcpu, gfn);
3078 gfn &= ~(KVM_PAGES_PER_HPAGE(level) - 1);
3079 } else
3080 level = PT_PAGE_TABLE_LEVEL;
Joerg Roedel852e3c12009-07-27 16:30:44 +02003081
Andrea Arcangelie930bff2008-07-25 16:24:52 +02003082 mmu_seq = vcpu->kvm->mmu_notifier_seq;
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03003083 smp_rmb();
Gleb Natapovaf585b92010-10-14 11:22:46 +02003084
Xiao Guangrong78b2c542010-12-07 10:48:06 +08003085 if (try_async_pf(vcpu, prefault, gfn, gpa, &pfn, write, &map_writable))
Gleb Natapovaf585b92010-10-14 11:22:46 +02003086 return 0;
3087
Xiao Guangrongd7c55202011-07-12 03:29:38 +08003088 if (handle_abnormal_pfn(vcpu, 0, gfn, pfn, ACC_ALL, &r))
3089 return r;
3090
Joerg Roedelfb72d162008-02-07 13:47:44 +01003091 spin_lock(&vcpu->kvm->mmu_lock);
Andrea Arcangelie930bff2008-07-25 16:24:52 +02003092 if (mmu_notifier_retry(vcpu, mmu_seq))
3093 goto out_unlock;
Joerg Roedelfb72d162008-02-07 13:47:44 +01003094 kvm_mmu_free_some_pages(vcpu);
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -08003095 if (likely(!force_pt_level))
3096 transparent_hugepage_adjust(vcpu, &gfn, &pfn, &level);
Marcelo Tosatti612819c2010-10-22 14:18:18 -02003097 r = __direct_map(vcpu, gpa, write, map_writable,
Xiao Guangrong2ec47392010-12-07 10:34:42 +08003098 level, gfn, pfn, prefault);
Joerg Roedelfb72d162008-02-07 13:47:44 +01003099 spin_unlock(&vcpu->kvm->mmu_lock);
Joerg Roedelfb72d162008-02-07 13:47:44 +01003100
3101 return r;
Andrea Arcangelie930bff2008-07-25 16:24:52 +02003102
3103out_unlock:
3104 spin_unlock(&vcpu->kvm->mmu_lock);
3105 kvm_release_pfn_clean(pfn);
3106 return 0;
Joerg Roedelfb72d162008-02-07 13:47:44 +01003107}
3108
Avi Kivity6aa8b732006-12-10 02:21:36 -08003109static void nonpaging_free(struct kvm_vcpu *vcpu)
3110{
Avi Kivity17ac10a2007-01-05 16:36:40 -08003111 mmu_free_roots(vcpu);
Avi Kivity6aa8b732006-12-10 02:21:36 -08003112}
3113
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003114static int nonpaging_init_context(struct kvm_vcpu *vcpu,
3115 struct kvm_mmu *context)
Avi Kivity6aa8b732006-12-10 02:21:36 -08003116{
Avi Kivity6aa8b732006-12-10 02:21:36 -08003117 context->new_cr3 = nonpaging_new_cr3;
3118 context->page_fault = nonpaging_page_fault;
Avi Kivity6aa8b732006-12-10 02:21:36 -08003119 context->gva_to_gpa = nonpaging_gva_to_gpa;
3120 context->free = nonpaging_free;
Marcelo Tosattie8bc2172008-09-23 13:18:33 -03003121 context->sync_page = nonpaging_sync_page;
Marcelo Tosattia7052892008-09-23 13:18:35 -03003122 context->invlpg = nonpaging_invlpg;
Xiao Guangrong0f53b5b2011-03-09 15:43:51 +08003123 context->update_pte = nonpaging_update_pte;
Avi Kivitycea0f0e2007-01-05 16:36:43 -08003124 context->root_level = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08003125 context->shadow_root_level = PT32E_ROOT_LEVEL;
Avi Kivity17c3ba92007-06-04 15:58:30 +03003126 context->root_hpa = INVALID_PAGE;
Joerg Roedelc5a78f2b2010-09-10 17:30:39 +02003127 context->direct_map = true;
Joerg Roedel2d48a982010-09-10 17:31:01 +02003128 context->nx = false;
Avi Kivity6aa8b732006-12-10 02:21:36 -08003129 return 0;
3130}
3131
Avi Kivityd835dfe2007-11-21 02:57:59 +02003132void kvm_mmu_flush_tlb(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -08003133{
Avi Kivity1165f5f2007-04-19 17:27:43 +03003134 ++vcpu->stat.tlb_flush;
Avi Kivitya8eeb042010-05-10 12:34:53 +03003135 kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
Avi Kivity6aa8b732006-12-10 02:21:36 -08003136}
3137
3138static void paging_new_cr3(struct kvm_vcpu *vcpu)
3139{
Avi Kivity9f8fe502010-12-05 17:30:00 +02003140 pgprintk("%s: cr3 %lx\n", __func__, kvm_read_cr3(vcpu));
Avi Kivitycea0f0e2007-01-05 16:36:43 -08003141 mmu_free_roots(vcpu);
Avi Kivity6aa8b732006-12-10 02:21:36 -08003142}
3143
Joerg Roedel5777ed32010-09-10 17:30:42 +02003144static unsigned long get_cr3(struct kvm_vcpu *vcpu)
3145{
Avi Kivity9f8fe502010-12-05 17:30:00 +02003146 return kvm_read_cr3(vcpu);
Joerg Roedel5777ed32010-09-10 17:30:42 +02003147}
3148
Avi Kivity6389ee92010-11-29 16:12:30 +02003149static void inject_page_fault(struct kvm_vcpu *vcpu,
3150 struct x86_exception *fault)
Avi Kivity6aa8b732006-12-10 02:21:36 -08003151{
Avi Kivity6389ee92010-11-29 16:12:30 +02003152 vcpu->arch.mmu.inject_page_fault(vcpu, fault);
Avi Kivity6aa8b732006-12-10 02:21:36 -08003153}
3154
Avi Kivity6aa8b732006-12-10 02:21:36 -08003155static void paging_free(struct kvm_vcpu *vcpu)
3156{
3157 nonpaging_free(vcpu);
3158}
3159
Joerg Roedel3241f222010-09-10 17:30:45 +02003160static bool is_rsvd_bits_set(struct kvm_mmu *mmu, u64 gpte, int level)
Dong, Eddie82725b22009-03-30 16:21:08 +08003161{
3162 int bit7;
3163
3164 bit7 = (gpte >> 7) & 1;
Joerg Roedel3241f222010-09-10 17:30:45 +02003165 return (gpte & mmu->rsvd_bits_mask[bit7][level-1]) != 0;
Dong, Eddie82725b22009-03-30 16:21:08 +08003166}
3167
Xiao Guangrongce88dec2011-07-12 03:33:44 +08003168static bool sync_mmio_spte(u64 *sptep, gfn_t gfn, unsigned access,
3169 int *nr_present)
3170{
3171 if (unlikely(is_mmio_spte(*sptep))) {
3172 if (gfn != get_mmio_spte_gfn(*sptep)) {
3173 mmu_spte_clear_no_track(sptep);
3174 return true;
3175 }
3176
3177 (*nr_present)++;
3178 mark_mmio_spte(sptep, gfn, access);
3179 return true;
3180 }
3181
3182 return false;
3183}
3184
Avi Kivity6aa8b732006-12-10 02:21:36 -08003185#define PTTYPE 64
3186#include "paging_tmpl.h"
3187#undef PTTYPE
3188
3189#define PTTYPE 32
3190#include "paging_tmpl.h"
3191#undef PTTYPE
3192
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003193static void reset_rsvds_bits_mask(struct kvm_vcpu *vcpu,
3194 struct kvm_mmu *context,
3195 int level)
Dong, Eddie82725b22009-03-30 16:21:08 +08003196{
Dong, Eddie82725b22009-03-30 16:21:08 +08003197 int maxphyaddr = cpuid_maxphyaddr(vcpu);
3198 u64 exb_bit_rsvd = 0;
3199
Joerg Roedel2d48a982010-09-10 17:31:01 +02003200 if (!context->nx)
Dong, Eddie82725b22009-03-30 16:21:08 +08003201 exb_bit_rsvd = rsvd_bits(63, 63);
3202 switch (level) {
3203 case PT32_ROOT_LEVEL:
3204 /* no rsvd bits for 2 level 4K page table entries */
3205 context->rsvd_bits_mask[0][1] = 0;
3206 context->rsvd_bits_mask[0][0] = 0;
Xiao Guangrongf815bce2010-03-19 17:58:53 +08003207 context->rsvd_bits_mask[1][0] = context->rsvd_bits_mask[0][0];
3208
3209 if (!is_pse(vcpu)) {
3210 context->rsvd_bits_mask[1][1] = 0;
3211 break;
3212 }
3213
Dong, Eddie82725b22009-03-30 16:21:08 +08003214 if (is_cpuid_PSE36())
3215 /* 36bits PSE 4MB page */
3216 context->rsvd_bits_mask[1][1] = rsvd_bits(17, 21);
3217 else
3218 /* 32 bits PSE 4MB page */
3219 context->rsvd_bits_mask[1][1] = rsvd_bits(13, 21);
Dong, Eddie82725b22009-03-30 16:21:08 +08003220 break;
3221 case PT32E_ROOT_LEVEL:
Dong, Eddie20c466b2009-03-31 23:03:45 +08003222 context->rsvd_bits_mask[0][2] =
3223 rsvd_bits(maxphyaddr, 63) |
3224 rsvd_bits(7, 8) | rsvd_bits(1, 2); /* PDPTE */
Dong, Eddie82725b22009-03-30 16:21:08 +08003225 context->rsvd_bits_mask[0][1] = exb_bit_rsvd |
Sheng Yang4c26b4c2009-04-02 10:28:37 +08003226 rsvd_bits(maxphyaddr, 62); /* PDE */
Dong, Eddie82725b22009-03-30 16:21:08 +08003227 context->rsvd_bits_mask[0][0] = exb_bit_rsvd |
3228 rsvd_bits(maxphyaddr, 62); /* PTE */
3229 context->rsvd_bits_mask[1][1] = exb_bit_rsvd |
3230 rsvd_bits(maxphyaddr, 62) |
3231 rsvd_bits(13, 20); /* large page */
Xiao Guangrongf815bce2010-03-19 17:58:53 +08003232 context->rsvd_bits_mask[1][0] = context->rsvd_bits_mask[0][0];
Dong, Eddie82725b22009-03-30 16:21:08 +08003233 break;
3234 case PT64_ROOT_LEVEL:
3235 context->rsvd_bits_mask[0][3] = exb_bit_rsvd |
3236 rsvd_bits(maxphyaddr, 51) | rsvd_bits(7, 8);
3237 context->rsvd_bits_mask[0][2] = exb_bit_rsvd |
3238 rsvd_bits(maxphyaddr, 51) | rsvd_bits(7, 8);
3239 context->rsvd_bits_mask[0][1] = exb_bit_rsvd |
Sheng Yang4c26b4c2009-04-02 10:28:37 +08003240 rsvd_bits(maxphyaddr, 51);
Dong, Eddie82725b22009-03-30 16:21:08 +08003241 context->rsvd_bits_mask[0][0] = exb_bit_rsvd |
3242 rsvd_bits(maxphyaddr, 51);
3243 context->rsvd_bits_mask[1][3] = context->rsvd_bits_mask[0][3];
Joerg Roedele04da982009-07-27 16:30:45 +02003244 context->rsvd_bits_mask[1][2] = exb_bit_rsvd |
3245 rsvd_bits(maxphyaddr, 51) |
3246 rsvd_bits(13, 29);
Dong, Eddie82725b22009-03-30 16:21:08 +08003247 context->rsvd_bits_mask[1][1] = exb_bit_rsvd |
Sheng Yang4c26b4c2009-04-02 10:28:37 +08003248 rsvd_bits(maxphyaddr, 51) |
3249 rsvd_bits(13, 20); /* large page */
Xiao Guangrongf815bce2010-03-19 17:58:53 +08003250 context->rsvd_bits_mask[1][0] = context->rsvd_bits_mask[0][0];
Dong, Eddie82725b22009-03-30 16:21:08 +08003251 break;
3252 }
3253}
3254
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003255static int paging64_init_context_common(struct kvm_vcpu *vcpu,
3256 struct kvm_mmu *context,
3257 int level)
Avi Kivity6aa8b732006-12-10 02:21:36 -08003258{
Joerg Roedel2d48a982010-09-10 17:31:01 +02003259 context->nx = is_nx(vcpu);
3260
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003261 reset_rsvds_bits_mask(vcpu, context, level);
Avi Kivity6aa8b732006-12-10 02:21:36 -08003262
3263 ASSERT(is_pae(vcpu));
3264 context->new_cr3 = paging_new_cr3;
3265 context->page_fault = paging64_page_fault;
Avi Kivity6aa8b732006-12-10 02:21:36 -08003266 context->gva_to_gpa = paging64_gva_to_gpa;
Marcelo Tosattie8bc2172008-09-23 13:18:33 -03003267 context->sync_page = paging64_sync_page;
Marcelo Tosattia7052892008-09-23 13:18:35 -03003268 context->invlpg = paging64_invlpg;
Xiao Guangrong0f53b5b2011-03-09 15:43:51 +08003269 context->update_pte = paging64_update_pte;
Avi Kivity6aa8b732006-12-10 02:21:36 -08003270 context->free = paging_free;
Avi Kivity17ac10a2007-01-05 16:36:40 -08003271 context->root_level = level;
3272 context->shadow_root_level = level;
Avi Kivity17c3ba92007-06-04 15:58:30 +03003273 context->root_hpa = INVALID_PAGE;
Joerg Roedelc5a78f2b2010-09-10 17:30:39 +02003274 context->direct_map = false;
Avi Kivity6aa8b732006-12-10 02:21:36 -08003275 return 0;
3276}
3277
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003278static int paging64_init_context(struct kvm_vcpu *vcpu,
3279 struct kvm_mmu *context)
Avi Kivity17ac10a2007-01-05 16:36:40 -08003280{
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003281 return paging64_init_context_common(vcpu, context, PT64_ROOT_LEVEL);
Avi Kivity17ac10a2007-01-05 16:36:40 -08003282}
3283
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003284static int paging32_init_context(struct kvm_vcpu *vcpu,
3285 struct kvm_mmu *context)
Avi Kivity6aa8b732006-12-10 02:21:36 -08003286{
Joerg Roedel2d48a982010-09-10 17:31:01 +02003287 context->nx = false;
3288
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003289 reset_rsvds_bits_mask(vcpu, context, PT32_ROOT_LEVEL);
Avi Kivity6aa8b732006-12-10 02:21:36 -08003290
3291 context->new_cr3 = paging_new_cr3;
3292 context->page_fault = paging32_page_fault;
Avi Kivity6aa8b732006-12-10 02:21:36 -08003293 context->gva_to_gpa = paging32_gva_to_gpa;
3294 context->free = paging_free;
Marcelo Tosattie8bc2172008-09-23 13:18:33 -03003295 context->sync_page = paging32_sync_page;
Marcelo Tosattia7052892008-09-23 13:18:35 -03003296 context->invlpg = paging32_invlpg;
Xiao Guangrong0f53b5b2011-03-09 15:43:51 +08003297 context->update_pte = paging32_update_pte;
Avi Kivity6aa8b732006-12-10 02:21:36 -08003298 context->root_level = PT32_ROOT_LEVEL;
3299 context->shadow_root_level = PT32E_ROOT_LEVEL;
Avi Kivity17c3ba92007-06-04 15:58:30 +03003300 context->root_hpa = INVALID_PAGE;
Joerg Roedelc5a78f2b2010-09-10 17:30:39 +02003301 context->direct_map = false;
Avi Kivity6aa8b732006-12-10 02:21:36 -08003302 return 0;
3303}
3304
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003305static int paging32E_init_context(struct kvm_vcpu *vcpu,
3306 struct kvm_mmu *context)
Avi Kivity6aa8b732006-12-10 02:21:36 -08003307{
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003308 return paging64_init_context_common(vcpu, context, PT32E_ROOT_LEVEL);
Avi Kivity6aa8b732006-12-10 02:21:36 -08003309}
3310
Joerg Roedelfb72d162008-02-07 13:47:44 +01003311static int init_kvm_tdp_mmu(struct kvm_vcpu *vcpu)
3312{
Joerg Roedel14dfe852010-09-10 17:30:49 +02003313 struct kvm_mmu *context = vcpu->arch.walk_mmu;
Joerg Roedelfb72d162008-02-07 13:47:44 +01003314
Avi Kivityc445f8e2010-12-21 16:26:01 +02003315 context->base_role.word = 0;
Joerg Roedelfb72d162008-02-07 13:47:44 +01003316 context->new_cr3 = nonpaging_new_cr3;
3317 context->page_fault = tdp_page_fault;
3318 context->free = nonpaging_free;
Marcelo Tosattie8bc2172008-09-23 13:18:33 -03003319 context->sync_page = nonpaging_sync_page;
Marcelo Tosattia7052892008-09-23 13:18:35 -03003320 context->invlpg = nonpaging_invlpg;
Xiao Guangrong0f53b5b2011-03-09 15:43:51 +08003321 context->update_pte = nonpaging_update_pte;
Sheng Yang67253af2008-04-25 10:20:22 +08003322 context->shadow_root_level = kvm_x86_ops->get_tdp_level();
Joerg Roedelfb72d162008-02-07 13:47:44 +01003323 context->root_hpa = INVALID_PAGE;
Joerg Roedelc5a78f2b2010-09-10 17:30:39 +02003324 context->direct_map = true;
Joerg Roedel1c97f0a2010-09-10 17:30:41 +02003325 context->set_cr3 = kvm_x86_ops->set_tdp_cr3;
Joerg Roedel5777ed32010-09-10 17:30:42 +02003326 context->get_cr3 = get_cr3;
Avi Kivitye4e517b2011-07-28 11:36:17 +03003327 context->get_pdptr = kvm_pdptr_read;
Joerg Roedelcb659db2010-09-10 17:30:43 +02003328 context->inject_page_fault = kvm_inject_page_fault;
Joerg Roedel2d48a982010-09-10 17:31:01 +02003329 context->nx = is_nx(vcpu);
Joerg Roedelfb72d162008-02-07 13:47:44 +01003330
3331 if (!is_paging(vcpu)) {
Joerg Roedel2d48a982010-09-10 17:31:01 +02003332 context->nx = false;
Joerg Roedelfb72d162008-02-07 13:47:44 +01003333 context->gva_to_gpa = nonpaging_gva_to_gpa;
3334 context->root_level = 0;
3335 } else if (is_long_mode(vcpu)) {
Joerg Roedel2d48a982010-09-10 17:31:01 +02003336 context->nx = is_nx(vcpu);
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003337 reset_rsvds_bits_mask(vcpu, context, PT64_ROOT_LEVEL);
Joerg Roedelfb72d162008-02-07 13:47:44 +01003338 context->gva_to_gpa = paging64_gva_to_gpa;
3339 context->root_level = PT64_ROOT_LEVEL;
3340 } else if (is_pae(vcpu)) {
Joerg Roedel2d48a982010-09-10 17:31:01 +02003341 context->nx = is_nx(vcpu);
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003342 reset_rsvds_bits_mask(vcpu, context, PT32E_ROOT_LEVEL);
Joerg Roedelfb72d162008-02-07 13:47:44 +01003343 context->gva_to_gpa = paging64_gva_to_gpa;
3344 context->root_level = PT32E_ROOT_LEVEL;
3345 } else {
Joerg Roedel2d48a982010-09-10 17:31:01 +02003346 context->nx = false;
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003347 reset_rsvds_bits_mask(vcpu, context, PT32_ROOT_LEVEL);
Joerg Roedelfb72d162008-02-07 13:47:44 +01003348 context->gva_to_gpa = paging32_gva_to_gpa;
3349 context->root_level = PT32_ROOT_LEVEL;
3350 }
3351
3352 return 0;
3353}
3354
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003355int kvm_init_shadow_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *context)
Avi Kivity6aa8b732006-12-10 02:21:36 -08003356{
Avi Kivitya770f6f2008-12-21 19:20:09 +02003357 int r;
Avi Kivity411c5882011-06-06 16:11:54 +03003358 bool smep = kvm_read_cr4_bits(vcpu, X86_CR4_SMEP);
Avi Kivity6aa8b732006-12-10 02:21:36 -08003359 ASSERT(vcpu);
Zhang Xiantaoad312c72007-12-13 23:50:52 +08003360 ASSERT(!VALID_PAGE(vcpu->arch.mmu.root_hpa));
Avi Kivity6aa8b732006-12-10 02:21:36 -08003361
3362 if (!is_paging(vcpu))
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003363 r = nonpaging_init_context(vcpu, context);
Avi Kivitya9058ec2006-12-29 16:49:37 -08003364 else if (is_long_mode(vcpu))
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003365 r = paging64_init_context(vcpu, context);
Avi Kivity6aa8b732006-12-10 02:21:36 -08003366 else if (is_pae(vcpu))
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003367 r = paging32E_init_context(vcpu, context);
Avi Kivity6aa8b732006-12-10 02:21:36 -08003368 else
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003369 r = paging32_init_context(vcpu, context);
Avi Kivitya770f6f2008-12-21 19:20:09 +02003370
Avi Kivity5b7e0102010-04-14 19:20:03 +03003371 vcpu->arch.mmu.base_role.cr4_pae = !!is_pae(vcpu);
Joerg Roedelf43addd2010-09-10 17:30:40 +02003372 vcpu->arch.mmu.base_role.cr0_wp = is_write_protection(vcpu);
Avi Kivity411c5882011-06-06 16:11:54 +03003373 vcpu->arch.mmu.base_role.smep_andnot_wp
3374 = smep && !is_write_protection(vcpu);
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003375
3376 return r;
3377}
3378EXPORT_SYMBOL_GPL(kvm_init_shadow_mmu);
3379
3380static int init_kvm_softmmu(struct kvm_vcpu *vcpu)
3381{
Joerg Roedel14dfe852010-09-10 17:30:49 +02003382 int r = kvm_init_shadow_mmu(vcpu, vcpu->arch.walk_mmu);
Joerg Roedel52fde8d2010-09-10 17:30:44 +02003383
Joerg Roedel14dfe852010-09-10 17:30:49 +02003384 vcpu->arch.walk_mmu->set_cr3 = kvm_x86_ops->set_cr3;
3385 vcpu->arch.walk_mmu->get_cr3 = get_cr3;
Avi Kivitye4e517b2011-07-28 11:36:17 +03003386 vcpu->arch.walk_mmu->get_pdptr = kvm_pdptr_read;
Joerg Roedel14dfe852010-09-10 17:30:49 +02003387 vcpu->arch.walk_mmu->inject_page_fault = kvm_inject_page_fault;
Avi Kivitya770f6f2008-12-21 19:20:09 +02003388
3389 return r;
Avi Kivity6aa8b732006-12-10 02:21:36 -08003390}
3391
Joerg Roedel02f59dc2010-09-10 17:30:54 +02003392static int init_kvm_nested_mmu(struct kvm_vcpu *vcpu)
3393{
3394 struct kvm_mmu *g_context = &vcpu->arch.nested_mmu;
3395
3396 g_context->get_cr3 = get_cr3;
Avi Kivitye4e517b2011-07-28 11:36:17 +03003397 g_context->get_pdptr = kvm_pdptr_read;
Joerg Roedel02f59dc2010-09-10 17:30:54 +02003398 g_context->inject_page_fault = kvm_inject_page_fault;
3399
3400 /*
3401 * Note that arch.mmu.gva_to_gpa translates l2_gva to l1_gpa. The
3402 * translation of l2_gpa to l1_gpa addresses is done using the
3403 * arch.nested_mmu.gva_to_gpa function. Basically the gva_to_gpa
3404 * functions between mmu and nested_mmu are swapped.
3405 */
3406 if (!is_paging(vcpu)) {
Joerg Roedel2d48a982010-09-10 17:31:01 +02003407 g_context->nx = false;
Joerg Roedel02f59dc2010-09-10 17:30:54 +02003408 g_context->root_level = 0;
3409 g_context->gva_to_gpa = nonpaging_gva_to_gpa_nested;
3410 } else if (is_long_mode(vcpu)) {
Joerg Roedel2d48a982010-09-10 17:31:01 +02003411 g_context->nx = is_nx(vcpu);
Joerg Roedel02f59dc2010-09-10 17:30:54 +02003412 reset_rsvds_bits_mask(vcpu, g_context, PT64_ROOT_LEVEL);
3413 g_context->root_level = PT64_ROOT_LEVEL;
3414 g_context->gva_to_gpa = paging64_gva_to_gpa_nested;
3415 } else if (is_pae(vcpu)) {
Joerg Roedel2d48a982010-09-10 17:31:01 +02003416 g_context->nx = is_nx(vcpu);
Joerg Roedel02f59dc2010-09-10 17:30:54 +02003417 reset_rsvds_bits_mask(vcpu, g_context, PT32E_ROOT_LEVEL);
3418 g_context->root_level = PT32E_ROOT_LEVEL;
3419 g_context->gva_to_gpa = paging64_gva_to_gpa_nested;
3420 } else {
Joerg Roedel2d48a982010-09-10 17:31:01 +02003421 g_context->nx = false;
Joerg Roedel02f59dc2010-09-10 17:30:54 +02003422 reset_rsvds_bits_mask(vcpu, g_context, PT32_ROOT_LEVEL);
3423 g_context->root_level = PT32_ROOT_LEVEL;
3424 g_context->gva_to_gpa = paging32_gva_to_gpa_nested;
3425 }
3426
3427 return 0;
3428}
3429
Joerg Roedelfb72d162008-02-07 13:47:44 +01003430static int init_kvm_mmu(struct kvm_vcpu *vcpu)
3431{
Joerg Roedel02f59dc2010-09-10 17:30:54 +02003432 if (mmu_is_nested(vcpu))
3433 return init_kvm_nested_mmu(vcpu);
3434 else if (tdp_enabled)
Joerg Roedelfb72d162008-02-07 13:47:44 +01003435 return init_kvm_tdp_mmu(vcpu);
3436 else
3437 return init_kvm_softmmu(vcpu);
3438}
3439
Avi Kivity6aa8b732006-12-10 02:21:36 -08003440static void destroy_kvm_mmu(struct kvm_vcpu *vcpu)
3441{
3442 ASSERT(vcpu);
Sheng Yang62ad0752010-05-12 16:40:41 +08003443 if (VALID_PAGE(vcpu->arch.mmu.root_hpa))
3444 /* mmu.free() should set root_hpa = INVALID_PAGE */
Zhang Xiantaoad312c72007-12-13 23:50:52 +08003445 vcpu->arch.mmu.free(vcpu);
Avi Kivity6aa8b732006-12-10 02:21:36 -08003446}
3447
3448int kvm_mmu_reset_context(struct kvm_vcpu *vcpu)
3449{
Avi Kivity17c3ba92007-06-04 15:58:30 +03003450 destroy_kvm_mmu(vcpu);
Marcelo Tosattif8f7e5e2011-06-21 14:00:10 -03003451 return init_kvm_mmu(vcpu);
Avi Kivity17c3ba92007-06-04 15:58:30 +03003452}
Eddie Dong8668a3c2007-10-10 14:26:45 +08003453EXPORT_SYMBOL_GPL(kvm_mmu_reset_context);
Avi Kivity17c3ba92007-06-04 15:58:30 +03003454
3455int kvm_mmu_load(struct kvm_vcpu *vcpu)
3456{
Avi Kivity714b93d2007-01-05 16:36:53 -08003457 int r;
3458
Avi Kivitye2dec932007-01-05 16:36:54 -08003459 r = mmu_topup_memory_caches(vcpu);
Avi Kivity17c3ba92007-06-04 15:58:30 +03003460 if (r)
3461 goto out;
Marcelo Tosatti8986ecc2009-05-12 18:55:45 -03003462 r = mmu_alloc_roots(vcpu);
Avi Kivity8facbbf2010-05-04 12:58:32 +03003463 spin_lock(&vcpu->kvm->mmu_lock);
Marcelo Tosatti0ba73cd2008-09-23 13:18:34 -03003464 mmu_sync_roots(vcpu);
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -05003465 spin_unlock(&vcpu->kvm->mmu_lock);
Marcelo Tosatti8986ecc2009-05-12 18:55:45 -03003466 if (r)
3467 goto out;
Sheng Yang3662cb12009-07-09 17:00:42 +08003468 /* set_cr3() should ensure TLB has been flushed */
Joerg Roedelf43addd2010-09-10 17:30:40 +02003469 vcpu->arch.mmu.set_cr3(vcpu, vcpu->arch.mmu.root_hpa);
Avi Kivity714b93d2007-01-05 16:36:53 -08003470out:
3471 return r;
Avi Kivity6aa8b732006-12-10 02:21:36 -08003472}
Avi Kivity17c3ba92007-06-04 15:58:30 +03003473EXPORT_SYMBOL_GPL(kvm_mmu_load);
3474
3475void kvm_mmu_unload(struct kvm_vcpu *vcpu)
3476{
3477 mmu_free_roots(vcpu);
3478}
Joerg Roedel4b161842010-09-10 17:31:03 +02003479EXPORT_SYMBOL_GPL(kvm_mmu_unload);
Avi Kivity6aa8b732006-12-10 02:21:36 -08003480
Avi Kivity00284252007-05-01 16:53:31 +03003481static void mmu_pte_write_new_pte(struct kvm_vcpu *vcpu,
Xiao Guangrong7c562522011-03-28 10:29:27 +08003482 struct kvm_mmu_page *sp, u64 *spte,
3483 const void *new)
Avi Kivity00284252007-05-01 16:53:31 +03003484{
Marcelo Tosatti30945382008-06-11 20:32:40 -03003485 if (sp->role.level != PT_PAGE_TABLE_LEVEL) {
Joerg Roedel7e4e4052009-07-27 16:30:46 +02003486 ++vcpu->kvm->stat.mmu_pde_zapped;
3487 return;
Marcelo Tosatti30945382008-06-11 20:32:40 -03003488 }
Avi Kivity00284252007-05-01 16:53:31 +03003489
Avi Kivity4cee5762007-11-18 16:37:07 +02003490 ++vcpu->kvm->stat.mmu_pte_updated;
Xiao Guangrong7c562522011-03-28 10:29:27 +08003491 vcpu->arch.mmu.update_pte(vcpu, sp, spte, new);
Avi Kivity00284252007-05-01 16:53:31 +03003492}
3493
Avi Kivity79539ce2007-11-21 02:06:21 +02003494static bool need_remote_flush(u64 old, u64 new)
3495{
3496 if (!is_shadow_present_pte(old))
3497 return false;
3498 if (!is_shadow_present_pte(new))
3499 return true;
3500 if ((old ^ new) & PT64_BASE_ADDR_MASK)
3501 return true;
3502 old ^= PT64_NX_MASK;
3503 new ^= PT64_NX_MASK;
3504 return (old & ~new & PT64_PERM_MASK) != 0;
3505}
3506
Xiao Guangrong0671a8e2010-06-04 21:56:59 +08003507static void mmu_pte_write_flush_tlb(struct kvm_vcpu *vcpu, bool zap_page,
3508 bool remote_flush, bool local_flush)
Avi Kivity79539ce2007-11-21 02:06:21 +02003509{
Xiao Guangrong0671a8e2010-06-04 21:56:59 +08003510 if (zap_page)
3511 return;
3512
3513 if (remote_flush)
Avi Kivity79539ce2007-11-21 02:06:21 +02003514 kvm_flush_remote_tlbs(vcpu->kvm);
Xiao Guangrong0671a8e2010-06-04 21:56:59 +08003515 else if (local_flush)
Avi Kivity79539ce2007-11-21 02:06:21 +02003516 kvm_mmu_flush_tlb(vcpu);
3517}
3518
Xiao Guangrong889e5cb2011-09-22 16:57:23 +08003519static u64 mmu_pte_write_fetch_gpte(struct kvm_vcpu *vcpu, gpa_t *gpa,
3520 const u8 *new, int *bytes)
Avi Kivityda4a00f2007-01-05 16:36:44 -08003521{
Xiao Guangrong889e5cb2011-09-22 16:57:23 +08003522 u64 gentry;
3523 int r;
Avi Kivity72016f32010-03-15 13:59:53 +02003524
Avi Kivity08e850c2010-03-15 13:59:57 +02003525 /*
3526 * Assume that the pte write on a page table of the same type
Xiao Guangrong49b26e22011-03-04 19:00:00 +08003527 * as the current vcpu paging mode since we update the sptes only
3528 * when they have the same mode.
Avi Kivity08e850c2010-03-15 13:59:57 +02003529 */
Xiao Guangrong889e5cb2011-09-22 16:57:23 +08003530 if (is_pae(vcpu) && *bytes == 4) {
Avi Kivity08e850c2010-03-15 13:59:57 +02003531 /* Handle a 32-bit guest writing two halves of a 64-bit gpte */
Xiao Guangrong889e5cb2011-09-22 16:57:23 +08003532 *gpa &= ~(gpa_t)7;
3533 *bytes = 8;
3534 r = kvm_read_guest(vcpu->kvm, *gpa, &gentry, min(*bytes, 8));
Avi Kivity08e850c2010-03-15 13:59:57 +02003535 if (r)
3536 gentry = 0;
3537 new = (const u8 *)&gentry;
3538 }
3539
Xiao Guangrong889e5cb2011-09-22 16:57:23 +08003540 switch (*bytes) {
Avi Kivity72016f32010-03-15 13:59:53 +02003541 case 4:
3542 gentry = *(const u32 *)new;
3543 break;
3544 case 8:
3545 gentry = *(const u64 *)new;
3546 break;
3547 default:
3548 gentry = 0;
3549 break;
3550 }
3551
Xiao Guangrong889e5cb2011-09-22 16:57:23 +08003552 return gentry;
3553}
3554
3555/*
3556 * If we're seeing too many writes to a page, it may no longer be a page table,
3557 * or we may be forking, in which case it is better to unmap the page.
3558 */
Xiao Guangronga30f47c2011-09-22 16:58:36 +08003559static bool detect_write_flooding(struct kvm_mmu_page *sp, u64 *spte)
Xiao Guangrong889e5cb2011-09-22 16:57:23 +08003560{
Xiao Guangronga30f47c2011-09-22 16:58:36 +08003561 /*
3562 * Skip write-flooding detected for the sp whose level is 1, because
3563 * it can become unsync, then the guest page is not write-protected.
3564 */
3565 if (sp->role.level == 1)
3566 return false;
Xiao Guangrong889e5cb2011-09-22 16:57:23 +08003567
Xiao Guangronga30f47c2011-09-22 16:58:36 +08003568 return ++sp->write_flooding_count >= 3;
Xiao Guangrong889e5cb2011-09-22 16:57:23 +08003569}
3570
3571/*
3572 * Misaligned accesses are too much trouble to fix up; also, they usually
3573 * indicate a page is not used as a page table.
3574 */
3575static bool detect_write_misaligned(struct kvm_mmu_page *sp, gpa_t gpa,
3576 int bytes)
3577{
3578 unsigned offset, pte_size, misaligned;
3579
3580 pgprintk("misaligned: gpa %llx bytes %d role %x\n",
3581 gpa, bytes, sp->role.word);
3582
3583 offset = offset_in_page(gpa);
3584 pte_size = sp->role.cr4_pae ? 8 : 4;
Xiao Guangrong5d9ca302011-09-22 16:57:55 +08003585
3586 /*
3587 * Sometimes, the OS only writes the last one bytes to update status
3588 * bits, for example, in linux, andb instruction is used in clear_bit().
3589 */
3590 if (!(offset & (pte_size - 1)) && bytes == 1)
3591 return false;
3592
Xiao Guangrong889e5cb2011-09-22 16:57:23 +08003593 misaligned = (offset ^ (offset + bytes - 1)) & ~(pte_size - 1);
3594 misaligned |= bytes < 4;
3595
3596 return misaligned;
3597}
3598
3599static u64 *get_written_sptes(struct kvm_mmu_page *sp, gpa_t gpa, int *nspte)
3600{
3601 unsigned page_offset, quadrant;
3602 u64 *spte;
3603 int level;
3604
3605 page_offset = offset_in_page(gpa);
3606 level = sp->role.level;
3607 *nspte = 1;
3608 if (!sp->role.cr4_pae) {
3609 page_offset <<= 1; /* 32->64 */
3610 /*
3611 * A 32-bit pde maps 4MB while the shadow pdes map
3612 * only 2MB. So we need to double the offset again
3613 * and zap two pdes instead of one.
3614 */
3615 if (level == PT32_ROOT_LEVEL) {
3616 page_offset &= ~7; /* kill rounding error */
3617 page_offset <<= 1;
3618 *nspte = 2;
3619 }
3620 quadrant = page_offset >> PAGE_SHIFT;
3621 page_offset &= ~PAGE_MASK;
3622 if (quadrant != sp->role.quadrant)
3623 return NULL;
3624 }
3625
3626 spte = &sp->spt[page_offset / sizeof(*spte)];
3627 return spte;
3628}
3629
3630void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
3631 const u8 *new, int bytes)
3632{
3633 gfn_t gfn = gpa >> PAGE_SHIFT;
3634 union kvm_mmu_page_role mask = { .word = 0 };
3635 struct kvm_mmu_page *sp;
3636 struct hlist_node *node;
3637 LIST_HEAD(invalid_list);
3638 u64 entry, gentry, *spte;
3639 int npte;
Xiao Guangronga30f47c2011-09-22 16:58:36 +08003640 bool remote_flush, local_flush, zap_page;
Xiao Guangrong889e5cb2011-09-22 16:57:23 +08003641
3642 /*
3643 * If we don't have indirect shadow pages, it means no page is
3644 * write-protected, so we can exit simply.
3645 */
3646 if (!ACCESS_ONCE(vcpu->kvm->arch.indirect_shadow_pages))
3647 return;
3648
3649 zap_page = remote_flush = local_flush = false;
3650
3651 pgprintk("%s: gpa %llx bytes %d\n", __func__, gpa, bytes);
3652
3653 gentry = mmu_pte_write_fetch_gpte(vcpu, &gpa, new, &bytes);
3654
3655 /*
3656 * No need to care whether allocation memory is successful
3657 * or not since pte prefetch is skiped if it does not have
3658 * enough objects in the cache.
3659 */
3660 mmu_topup_memory_caches(vcpu);
3661
3662 spin_lock(&vcpu->kvm->mmu_lock);
3663 ++vcpu->kvm->stat.mmu_pte_write;
3664 trace_kvm_mmu_audit(vcpu, AUDIT_PRE_PTE_WRITE);
3665
Xiao Guangrongfa1de2b2010-07-16 11:19:51 +08003666 mask.cr0_wp = mask.cr4_pae = mask.nxe = 1;
Xiao Guangrongf41d3352010-06-04 21:56:11 +08003667 for_each_gfn_indirect_valid_sp(vcpu->kvm, sp, gfn, node) {
Xiao Guangronga30f47c2011-09-22 16:58:36 +08003668 spte = get_written_sptes(sp, gpa, &npte);
Xiao Guangrong889e5cb2011-09-22 16:57:23 +08003669
Xiao Guangronga30f47c2011-09-22 16:58:36 +08003670 if (detect_write_misaligned(sp, gpa, bytes) ||
3671 detect_write_flooding(sp, spte)) {
Xiao Guangrong0671a8e2010-06-04 21:56:59 +08003672 zap_page |= !!kvm_mmu_prepare_zap_page(vcpu->kvm, sp,
Xiao Guangrongf41d3352010-06-04 21:56:11 +08003673 &invalid_list);
Avi Kivity4cee5762007-11-18 16:37:07 +02003674 ++vcpu->kvm->stat.mmu_flooded;
Avi Kivity0e7bc4b2007-01-05 16:36:48 -08003675 continue;
3676 }
Xiao Guangrong889e5cb2011-09-22 16:57:23 +08003677
3678 spte = get_written_sptes(sp, gpa, &npte);
3679 if (!spte)
3680 continue;
3681
Xiao Guangrong0671a8e2010-06-04 21:56:59 +08003682 local_flush = true;
Avi Kivityac1b7142007-03-08 17:13:32 +02003683 while (npte--) {
Avi Kivity79539ce2007-11-21 02:06:21 +02003684 entry = *spte;
Xiao Guangrong38e3b2b2011-05-15 23:27:52 +08003685 mmu_page_zap_pte(vcpu->kvm, sp, spte);
Xiao Guangrongfa1de2b2010-07-16 11:19:51 +08003686 if (gentry &&
3687 !((sp->role.word ^ vcpu->arch.mmu.base_role.word)
Xiao Guangrongf759e2b2011-09-22 16:53:17 +08003688 & mask.word) && rmap_can_add(vcpu))
Xiao Guangrong7c562522011-03-28 10:29:27 +08003689 mmu_pte_write_new_pte(vcpu, sp, spte, &gentry);
Xiao Guangrong0671a8e2010-06-04 21:56:59 +08003690 if (!remote_flush && need_remote_flush(entry, *spte))
3691 remote_flush = true;
Avi Kivityac1b7142007-03-08 17:13:32 +02003692 ++spte;
Avi Kivity9b7a0322007-01-05 16:36:45 -08003693 }
Avi Kivity9b7a0322007-01-05 16:36:45 -08003694 }
Xiao Guangrong0671a8e2010-06-04 21:56:59 +08003695 mmu_pte_write_flush_tlb(vcpu, zap_page, remote_flush, local_flush);
Xiao Guangrongd98ba052010-06-04 21:55:29 +08003696 kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
Xiao Guangrong8b1fe172010-08-30 18:22:53 +08003697 trace_kvm_mmu_audit(vcpu, AUDIT_POST_PTE_WRITE);
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -05003698 spin_unlock(&vcpu->kvm->mmu_lock);
Avi Kivityda4a00f2007-01-05 16:36:44 -08003699}
3700
Avi Kivitya4360362007-01-05 16:36:45 -08003701int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva)
3702{
Marcelo Tosatti10589a42007-12-20 19:18:22 -05003703 gpa_t gpa;
3704 int r;
Avi Kivitya4360362007-01-05 16:36:45 -08003705
Joerg Roedelc5a78f2b2010-09-10 17:30:39 +02003706 if (vcpu->arch.mmu.direct_map)
Avi Kivity60f24782009-08-27 13:37:06 +03003707 return 0;
3708
Gleb Natapov1871c602010-02-10 14:21:32 +02003709 gpa = kvm_mmu_gva_to_gpa_read(vcpu, gva, NULL);
Marcelo Tosatti10589a42007-12-20 19:18:22 -05003710
Marcelo Tosatti10589a42007-12-20 19:18:22 -05003711 r = kvm_mmu_unprotect_page(vcpu->kvm, gpa >> PAGE_SHIFT);
Xiao Guangrong1cb3f3a2011-09-22 17:02:48 +08003712
Marcelo Tosatti10589a42007-12-20 19:18:22 -05003713 return r;
Avi Kivitya4360362007-01-05 16:36:45 -08003714}
Avi Kivity577bdc42008-07-19 08:57:05 +03003715EXPORT_SYMBOL_GPL(kvm_mmu_unprotect_page_virt);
Avi Kivitya4360362007-01-05 16:36:45 -08003716
Avi Kivity22d95b12007-09-14 20:26:06 +03003717void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu)
Avi Kivityebeace82007-01-05 16:36:47 -08003718{
Xiao Guangrongd98ba052010-06-04 21:55:29 +08003719 LIST_HEAD(invalid_list);
Xiao Guangrong103ad252010-06-04 21:54:38 +08003720
Dave Hansene0df7b92010-08-19 18:11:05 -07003721 while (kvm_mmu_available_pages(vcpu->kvm) < KVM_REFILL_PAGES &&
Izik Eidus3b80fff2009-07-28 15:26:58 -03003722 !list_empty(&vcpu->kvm->arch.active_mmu_pages)) {
Avi Kivity4db35312007-11-21 15:28:32 +02003723 struct kvm_mmu_page *sp;
Avi Kivityebeace82007-01-05 16:36:47 -08003724
Zhang Xiantaof05e70a2007-12-14 10:01:48 +08003725 sp = container_of(vcpu->kvm->arch.active_mmu_pages.prev,
Avi Kivity4db35312007-11-21 15:28:32 +02003726 struct kvm_mmu_page, link);
Dave Hansene0df7b92010-08-19 18:11:05 -07003727 kvm_mmu_prepare_zap_page(vcpu->kvm, sp, &invalid_list);
Avi Kivity4cee5762007-11-18 16:37:07 +02003728 ++vcpu->kvm->stat.mmu_recycled;
Avi Kivityebeace82007-01-05 16:36:47 -08003729 }
Xiao Guangrongaa6bd182011-07-12 03:26:40 +08003730 kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
Avi Kivityebeace82007-01-05 16:36:47 -08003731}
Avi Kivityebeace82007-01-05 16:36:47 -08003732
Xiao Guangrong1cb3f3a2011-09-22 17:02:48 +08003733static bool is_mmio_page_fault(struct kvm_vcpu *vcpu, gva_t addr)
3734{
3735 if (vcpu->arch.mmu.direct_map || mmu_is_nested(vcpu))
3736 return vcpu_match_mmio_gpa(vcpu, addr);
3737
3738 return vcpu_match_mmio_gva(vcpu, addr);
3739}
3740
Andre Przywaradc25e892010-12-21 11:12:07 +01003741int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t cr2, u32 error_code,
3742 void *insn, int insn_len)
Avi Kivity30677142007-10-28 18:48:59 +02003743{
Xiao Guangrong1cb3f3a2011-09-22 17:02:48 +08003744 int r, emulation_type = EMULTYPE_RETRY;
Avi Kivity30677142007-10-28 18:48:59 +02003745 enum emulation_result er;
3746
Gleb Natapov56028d02010-10-17 18:13:42 +02003747 r = vcpu->arch.mmu.page_fault(vcpu, cr2, error_code, false);
Avi Kivity30677142007-10-28 18:48:59 +02003748 if (r < 0)
3749 goto out;
3750
3751 if (!r) {
3752 r = 1;
3753 goto out;
3754 }
3755
Xiao Guangrong1cb3f3a2011-09-22 17:02:48 +08003756 if (is_mmio_page_fault(vcpu, cr2))
3757 emulation_type = 0;
3758
3759 er = x86_emulate_instruction(vcpu, cr2, emulation_type, insn, insn_len);
Avi Kivity30677142007-10-28 18:48:59 +02003760
3761 switch (er) {
3762 case EMULATE_DONE:
3763 return 1;
3764 case EMULATE_DO_MMIO:
3765 ++vcpu->stat.mmio_exits;
Gleb Natapov6d77dbf2010-05-10 11:16:56 +03003766 /* fall through */
Avi Kivity30677142007-10-28 18:48:59 +02003767 case EMULATE_FAIL:
Avi Kivity3f5d18a2009-06-11 15:43:28 +03003768 return 0;
Avi Kivity30677142007-10-28 18:48:59 +02003769 default:
3770 BUG();
3771 }
3772out:
Avi Kivity30677142007-10-28 18:48:59 +02003773 return r;
3774}
3775EXPORT_SYMBOL_GPL(kvm_mmu_page_fault);
3776
Marcelo Tosattia7052892008-09-23 13:18:35 -03003777void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
3778{
Marcelo Tosattia7052892008-09-23 13:18:35 -03003779 vcpu->arch.mmu.invlpg(vcpu, gva);
Marcelo Tosattia7052892008-09-23 13:18:35 -03003780 kvm_mmu_flush_tlb(vcpu);
3781 ++vcpu->stat.invlpg;
3782}
3783EXPORT_SYMBOL_GPL(kvm_mmu_invlpg);
3784
Joerg Roedel18552672008-02-07 13:47:41 +01003785void kvm_enable_tdp(void)
3786{
3787 tdp_enabled = true;
3788}
3789EXPORT_SYMBOL_GPL(kvm_enable_tdp);
3790
Joerg Roedel5f4cb662008-07-14 20:36:36 +02003791void kvm_disable_tdp(void)
3792{
3793 tdp_enabled = false;
3794}
3795EXPORT_SYMBOL_GPL(kvm_disable_tdp);
3796
Avi Kivity6aa8b732006-12-10 02:21:36 -08003797static void free_mmu_pages(struct kvm_vcpu *vcpu)
3798{
Zhang Xiantaoad312c72007-12-13 23:50:52 +08003799 free_page((unsigned long)vcpu->arch.mmu.pae_root);
Joerg Roedel81407ca2010-09-10 17:31:00 +02003800 if (vcpu->arch.mmu.lm_root != NULL)
3801 free_page((unsigned long)vcpu->arch.mmu.lm_root);
Avi Kivity6aa8b732006-12-10 02:21:36 -08003802}
3803
3804static int alloc_mmu_pages(struct kvm_vcpu *vcpu)
3805{
Avi Kivity17ac10a2007-01-05 16:36:40 -08003806 struct page *page;
Avi Kivity6aa8b732006-12-10 02:21:36 -08003807 int i;
3808
3809 ASSERT(vcpu);
3810
Avi Kivity17ac10a2007-01-05 16:36:40 -08003811 /*
3812 * When emulating 32-bit mode, cr3 is only 32 bits even on x86_64.
3813 * Therefore we need to allocate shadow page tables in the first
3814 * 4GB of memory, which happens to fit the DMA32 zone.
3815 */
3816 page = alloc_page(GFP_KERNEL | __GFP_DMA32);
3817 if (!page)
Wei Yongjund7fa6ab2010-01-22 16:55:05 +08003818 return -ENOMEM;
3819
Zhang Xiantaoad312c72007-12-13 23:50:52 +08003820 vcpu->arch.mmu.pae_root = page_address(page);
Avi Kivity17ac10a2007-01-05 16:36:40 -08003821 for (i = 0; i < 4; ++i)
Zhang Xiantaoad312c72007-12-13 23:50:52 +08003822 vcpu->arch.mmu.pae_root[i] = INVALID_PAGE;
Avi Kivity17ac10a2007-01-05 16:36:40 -08003823
Avi Kivity6aa8b732006-12-10 02:21:36 -08003824 return 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08003825}
3826
Ingo Molnar8018c272006-12-29 16:50:01 -08003827int kvm_mmu_create(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -08003828{
Avi Kivity6aa8b732006-12-10 02:21:36 -08003829 ASSERT(vcpu);
Zhang Xiantaoad312c72007-12-13 23:50:52 +08003830 ASSERT(!VALID_PAGE(vcpu->arch.mmu.root_hpa));
Avi Kivity6aa8b732006-12-10 02:21:36 -08003831
Ingo Molnar8018c272006-12-29 16:50:01 -08003832 return alloc_mmu_pages(vcpu);
3833}
Avi Kivity6aa8b732006-12-10 02:21:36 -08003834
Ingo Molnar8018c272006-12-29 16:50:01 -08003835int kvm_mmu_setup(struct kvm_vcpu *vcpu)
3836{
3837 ASSERT(vcpu);
Zhang Xiantaoad312c72007-12-13 23:50:52 +08003838 ASSERT(!VALID_PAGE(vcpu->arch.mmu.root_hpa));
Avi Kivity2c264952006-12-22 01:05:28 -08003839
Ingo Molnar8018c272006-12-29 16:50:01 -08003840 return init_kvm_mmu(vcpu);
Avi Kivity6aa8b732006-12-10 02:21:36 -08003841}
3842
Avi Kivity90cb0522007-07-17 13:04:56 +03003843void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot)
Avi Kivity6aa8b732006-12-10 02:21:36 -08003844{
Avi Kivity4db35312007-11-21 15:28:32 +02003845 struct kvm_mmu_page *sp;
Avi Kivity6aa8b732006-12-10 02:21:36 -08003846
Zhang Xiantaof05e70a2007-12-14 10:01:48 +08003847 list_for_each_entry(sp, &kvm->arch.active_mmu_pages, link) {
Avi Kivity6aa8b732006-12-10 02:21:36 -08003848 int i;
3849 u64 *pt;
3850
Sheng Yang291f26b2008-10-16 17:30:57 +08003851 if (!test_bit(slot, sp->slot_bitmap))
Avi Kivity6aa8b732006-12-10 02:21:36 -08003852 continue;
3853
Avi Kivity4db35312007-11-21 15:28:32 +02003854 pt = sp->spt;
Avi Kivity8234b222010-12-27 12:08:45 +02003855 for (i = 0; i < PT64_ENT_PER_PAGE; ++i) {
Xiao Guangrongda8dc752011-03-04 18:56:41 +08003856 if (!is_shadow_present_pte(pt[i]) ||
3857 !is_last_spte(pt[i], sp->role.level))
3858 continue;
3859
3860 if (is_large_pte(pt[i])) {
Xiao Guangrongc3707952011-07-12 03:28:04 +08003861 drop_spte(kvm, &pt[i]);
Avi Kivity8234b222010-12-27 12:08:45 +02003862 --kvm->stat.lpages;
Xiao Guangrongda8dc752011-03-04 18:56:41 +08003863 continue;
Avi Kivity8234b222010-12-27 12:08:45 +02003864 }
Xiao Guangrongda8dc752011-03-04 18:56:41 +08003865
Avi Kivity6aa8b732006-12-10 02:21:36 -08003866 /* avoid RMW */
Gui Jianfeng01c168a2010-05-27 16:09:48 +08003867 if (is_writable_pte(pt[i]))
Xiao Guangrong1df9f2d2011-07-12 03:30:35 +08003868 mmu_spte_update(&pt[i],
3869 pt[i] & ~PT_WRITABLE_MASK);
Avi Kivity8234b222010-12-27 12:08:45 +02003870 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08003871 }
Avi Kivity171d5952008-08-27 16:40:51 +03003872 kvm_flush_remote_tlbs(kvm);
Avi Kivity6aa8b732006-12-10 02:21:36 -08003873}
Avi Kivity37a7d8b2007-01-05 16:36:56 -08003874
Avi Kivity90cb0522007-07-17 13:04:56 +03003875void kvm_mmu_zap_all(struct kvm *kvm)
Dor Laore0fa8262007-03-30 13:06:33 +03003876{
Avi Kivity4db35312007-11-21 15:28:32 +02003877 struct kvm_mmu_page *sp, *node;
Xiao Guangrongd98ba052010-06-04 21:55:29 +08003878 LIST_HEAD(invalid_list);
Dor Laore0fa8262007-03-30 13:06:33 +03003879
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -05003880 spin_lock(&kvm->mmu_lock);
Xiao Guangrong3246af02010-04-16 16:35:54 +08003881restart:
Zhang Xiantaof05e70a2007-12-14 10:01:48 +08003882 list_for_each_entry_safe(sp, node, &kvm->arch.active_mmu_pages, link)
Xiao Guangrongd98ba052010-06-04 21:55:29 +08003883 if (kvm_mmu_prepare_zap_page(kvm, sp, &invalid_list))
Xiao Guangrong3246af02010-04-16 16:35:54 +08003884 goto restart;
3885
Xiao Guangrongd98ba052010-06-04 21:55:29 +08003886 kvm_mmu_commit_zap_page(kvm, &invalid_list);
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -05003887 spin_unlock(&kvm->mmu_lock);
Dor Laore0fa8262007-03-30 13:06:33 +03003888}
3889
Xiao Guangrongd98ba052010-06-04 21:55:29 +08003890static int kvm_mmu_remove_some_alloc_mmu_pages(struct kvm *kvm,
3891 struct list_head *invalid_list)
Izik Eidus3ee16c82008-03-30 15:17:21 +03003892{
3893 struct kvm_mmu_page *page;
3894
3895 page = container_of(kvm->arch.active_mmu_pages.prev,
3896 struct kvm_mmu_page, link);
Xiao Guangrongd98ba052010-06-04 21:55:29 +08003897 return kvm_mmu_prepare_zap_page(kvm, page, invalid_list);
Izik Eidus3ee16c82008-03-30 15:17:21 +03003898}
3899
Ying Han1495f232011-05-24 17:12:27 -07003900static int mmu_shrink(struct shrinker *shrink, struct shrink_control *sc)
Izik Eidus3ee16c82008-03-30 15:17:21 +03003901{
3902 struct kvm *kvm;
3903 struct kvm *kvm_freed = NULL;
Ying Han1495f232011-05-24 17:12:27 -07003904 int nr_to_scan = sc->nr_to_scan;
Dave Hansen45221ab2010-08-19 18:11:37 -07003905
3906 if (nr_to_scan == 0)
3907 goto out;
Izik Eidus3ee16c82008-03-30 15:17:21 +03003908
Jan Kiszkae935b832011-02-08 12:55:33 +01003909 raw_spin_lock(&kvm_lock);
Izik Eidus3ee16c82008-03-30 15:17:21 +03003910
3911 list_for_each_entry(kvm, &vm_list, vm_list) {
Dave Hansen45221ab2010-08-19 18:11:37 -07003912 int idx, freed_pages;
Xiao Guangrongd98ba052010-06-04 21:55:29 +08003913 LIST_HEAD(invalid_list);
Izik Eidus3ee16c82008-03-30 15:17:21 +03003914
Marcelo Tosattif656ce02009-12-23 14:35:25 -02003915 idx = srcu_read_lock(&kvm->srcu);
Izik Eidus3ee16c82008-03-30 15:17:21 +03003916 spin_lock(&kvm->mmu_lock);
Dave Hansen45221ab2010-08-19 18:11:37 -07003917 if (!kvm_freed && nr_to_scan > 0 &&
3918 kvm->arch.n_used_mmu_pages > 0) {
Xiao Guangrongd98ba052010-06-04 21:55:29 +08003919 freed_pages = kvm_mmu_remove_some_alloc_mmu_pages(kvm,
3920 &invalid_list);
Izik Eidus3ee16c82008-03-30 15:17:21 +03003921 kvm_freed = kvm;
3922 }
3923 nr_to_scan--;
3924
Xiao Guangrongd98ba052010-06-04 21:55:29 +08003925 kvm_mmu_commit_zap_page(kvm, &invalid_list);
Izik Eidus3ee16c82008-03-30 15:17:21 +03003926 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattif656ce02009-12-23 14:35:25 -02003927 srcu_read_unlock(&kvm->srcu, idx);
Izik Eidus3ee16c82008-03-30 15:17:21 +03003928 }
3929 if (kvm_freed)
3930 list_move_tail(&kvm_freed->vm_list, &vm_list);
3931
Jan Kiszkae935b832011-02-08 12:55:33 +01003932 raw_spin_unlock(&kvm_lock);
Izik Eidus3ee16c82008-03-30 15:17:21 +03003933
Dave Hansen45221ab2010-08-19 18:11:37 -07003934out:
3935 return percpu_counter_read_positive(&kvm_total_used_mmu_pages);
Izik Eidus3ee16c82008-03-30 15:17:21 +03003936}
3937
3938static struct shrinker mmu_shrinker = {
3939 .shrink = mmu_shrink,
3940 .seeks = DEFAULT_SEEKS * 10,
3941};
3942
Ingo Molnar2ddfd202008-05-22 10:37:48 +02003943static void mmu_destroy_caches(void)
Avi Kivityb5a33a72007-04-15 16:31:09 +03003944{
Xiao Guangrong53c07b12011-05-15 23:26:20 +08003945 if (pte_list_desc_cache)
3946 kmem_cache_destroy(pte_list_desc_cache);
Avi Kivityd3d25b02007-05-30 12:34:53 +03003947 if (mmu_page_header_cache)
3948 kmem_cache_destroy(mmu_page_header_cache);
Avi Kivityb5a33a72007-04-15 16:31:09 +03003949}
3950
3951int kvm_mmu_module_init(void)
3952{
Xiao Guangrong53c07b12011-05-15 23:26:20 +08003953 pte_list_desc_cache = kmem_cache_create("pte_list_desc",
3954 sizeof(struct pte_list_desc),
Paul Mundt20c2df82007-07-20 10:11:58 +09003955 0, 0, NULL);
Xiao Guangrong53c07b12011-05-15 23:26:20 +08003956 if (!pte_list_desc_cache)
Avi Kivityb5a33a72007-04-15 16:31:09 +03003957 goto nomem;
3958
Avi Kivityd3d25b02007-05-30 12:34:53 +03003959 mmu_page_header_cache = kmem_cache_create("kvm_mmu_page_header",
3960 sizeof(struct kvm_mmu_page),
Paul Mundt20c2df82007-07-20 10:11:58 +09003961 0, 0, NULL);
Avi Kivityd3d25b02007-05-30 12:34:53 +03003962 if (!mmu_page_header_cache)
3963 goto nomem;
3964
Wei Yongjun45bf21a2010-08-23 16:13:15 +08003965 if (percpu_counter_init(&kvm_total_used_mmu_pages, 0))
3966 goto nomem;
3967
Izik Eidus3ee16c82008-03-30 15:17:21 +03003968 register_shrinker(&mmu_shrinker);
3969
Avi Kivityb5a33a72007-04-15 16:31:09 +03003970 return 0;
3971
3972nomem:
Izik Eidus3ee16c82008-03-30 15:17:21 +03003973 mmu_destroy_caches();
Avi Kivityb5a33a72007-04-15 16:31:09 +03003974 return -ENOMEM;
3975}
3976
Zhang Xiantao3ad82a72007-11-20 13:11:38 +08003977/*
3978 * Caculate mmu pages needed for kvm.
3979 */
3980unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm)
3981{
3982 int i;
3983 unsigned int nr_mmu_pages;
3984 unsigned int nr_pages = 0;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -02003985 struct kvm_memslots *slots;
Zhang Xiantao3ad82a72007-11-20 13:11:38 +08003986
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08003987 slots = kvm_memslots(kvm);
3988
Marcelo Tosattibc6678a2009-12-23 14:35:21 -02003989 for (i = 0; i < slots->nmemslots; i++)
3990 nr_pages += slots->memslots[i].npages;
Zhang Xiantao3ad82a72007-11-20 13:11:38 +08003991
3992 nr_mmu_pages = nr_pages * KVM_PERMILLE_MMU_PAGES / 1000;
3993 nr_mmu_pages = max(nr_mmu_pages,
3994 (unsigned int) KVM_MIN_ALLOC_MMU_PAGES);
3995
3996 return nr_mmu_pages;
3997}
3998
Marcelo Tosatti94d8b052009-06-11 12:07:42 -03003999int kvm_mmu_get_spte_hierarchy(struct kvm_vcpu *vcpu, u64 addr, u64 sptes[4])
4000{
4001 struct kvm_shadow_walk_iterator iterator;
Xiao Guangrongc2a2ac22011-07-12 03:32:13 +08004002 u64 spte;
Marcelo Tosatti94d8b052009-06-11 12:07:42 -03004003 int nr_sptes = 0;
4004
Xiao Guangrongc2a2ac22011-07-12 03:32:13 +08004005 walk_shadow_page_lockless_begin(vcpu);
4006 for_each_shadow_entry_lockless(vcpu, addr, iterator, spte) {
4007 sptes[iterator.level-1] = spte;
Marcelo Tosatti94d8b052009-06-11 12:07:42 -03004008 nr_sptes++;
Xiao Guangrongc2a2ac22011-07-12 03:32:13 +08004009 if (!is_shadow_present_pte(spte))
Marcelo Tosatti94d8b052009-06-11 12:07:42 -03004010 break;
4011 }
Xiao Guangrongc2a2ac22011-07-12 03:32:13 +08004012 walk_shadow_page_lockless_end(vcpu);
Marcelo Tosatti94d8b052009-06-11 12:07:42 -03004013
4014 return nr_sptes;
4015}
4016EXPORT_SYMBOL_GPL(kvm_mmu_get_spte_hierarchy);
4017
Xiao Guangrongc42fffe2010-09-27 18:07:07 +08004018void kvm_mmu_destroy(struct kvm_vcpu *vcpu)
4019{
4020 ASSERT(vcpu);
4021
4022 destroy_kvm_mmu(vcpu);
4023 free_mmu_pages(vcpu);
4024 mmu_free_memory_caches(vcpu);
Xiao Guangrongb034cf02010-12-23 16:08:35 +08004025}
4026
4027#ifdef CONFIG_KVM_MMU_AUDIT
4028#include "mmu_audit.c"
4029#else
4030static void mmu_audit_disable(void) { }
4031#endif
4032
4033void kvm_mmu_module_exit(void)
4034{
4035 mmu_destroy_caches();
4036 percpu_counter_destroy(&kvm_total_used_mmu_pages);
4037 unregister_shrinker(&mmu_shrinker);
Xiao Guangrongc42fffe2010-09-27 18:07:07 +08004038 mmu_audit_disable();
4039}