xref: /openbmc/linux/arch/x86/kvm/mmu/spte.h (revision 5c88af59)
1 // SPDX-License-Identifier: GPL-2.0-only
2 
3 #ifndef KVM_X86_MMU_SPTE_H
4 #define KVM_X86_MMU_SPTE_H
5 
6 #include "mmu_internal.h"
7 
8 #define PT_FIRST_AVAIL_BITS_SHIFT 10
9 #define PT64_SECOND_AVAIL_BITS_SHIFT 54
10 
11 /*
12  * The mask used to denote special SPTEs, which can be either MMIO SPTEs or
13  * Access Tracking SPTEs.
14  */
15 #define SPTE_SPECIAL_MASK (3ULL << 52)
16 #define SPTE_AD_ENABLED_MASK (0ULL << 52)
17 #define SPTE_AD_DISABLED_MASK (1ULL << 52)
18 #define SPTE_AD_WRPROT_ONLY_MASK (2ULL << 52)
19 #define SPTE_MMIO_MASK (3ULL << 52)
20 
21 #ifdef CONFIG_DYNAMIC_PHYSICAL_MASK
22 #define PT64_BASE_ADDR_MASK (physical_mask & ~(u64)(PAGE_SIZE-1))
23 #else
24 #define PT64_BASE_ADDR_MASK (((1ULL << 52) - 1) & ~(u64)(PAGE_SIZE-1))
25 #endif
26 #define PT64_LVL_ADDR_MASK(level) \
27 	(PT64_BASE_ADDR_MASK & ~((1ULL << (PAGE_SHIFT + (((level) - 1) \
28 						* PT64_LEVEL_BITS))) - 1))
29 #define PT64_LVL_OFFSET_MASK(level) \
30 	(PT64_BASE_ADDR_MASK & ((1ULL << (PAGE_SHIFT + (((level) - 1) \
31 						* PT64_LEVEL_BITS))) - 1))
32 
33 #define PT64_PERM_MASK (PT_PRESENT_MASK | PT_WRITABLE_MASK | shadow_user_mask \
34 			| shadow_x_mask | shadow_nx_mask | shadow_me_mask)
35 
36 #define ACC_EXEC_MASK    1
37 #define ACC_WRITE_MASK   PT_WRITABLE_MASK
38 #define ACC_USER_MASK    PT_USER_MASK
39 #define ACC_ALL          (ACC_EXEC_MASK | ACC_WRITE_MASK | ACC_USER_MASK)
40 
41 /* The mask for the R/X bits in EPT PTEs */
42 #define PT64_EPT_READABLE_MASK			0x1ull
43 #define PT64_EPT_EXECUTABLE_MASK		0x4ull
44 
45 #define PT64_LEVEL_BITS 9
46 
47 #define PT64_LEVEL_SHIFT(level) \
48 		(PAGE_SHIFT + (level - 1) * PT64_LEVEL_BITS)
49 
50 #define PT64_INDEX(address, level)\
51 	(((address) >> PT64_LEVEL_SHIFT(level)) & ((1 << PT64_LEVEL_BITS) - 1))
52 #define SHADOW_PT_INDEX(addr, level) PT64_INDEX(addr, level)
53 
54 
55 #define SPTE_HOST_WRITEABLE	(1ULL << PT_FIRST_AVAIL_BITS_SHIFT)
56 #define SPTE_MMU_WRITEABLE	(1ULL << (PT_FIRST_AVAIL_BITS_SHIFT + 1))
57 
58 /*
59  * Due to limited space in PTEs, the MMIO generation is a 18 bit subset of
60  * the memslots generation and is derived as follows:
61  *
62  * Bits 0-8 of the MMIO generation are propagated to spte bits 3-11
63  * Bits 9-17 of the MMIO generation are propagated to spte bits 54-62
64  *
65  * The KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS flag is intentionally not included in
66  * the MMIO generation number, as doing so would require stealing a bit from
67  * the "real" generation number and thus effectively halve the maximum number
68  * of MMIO generations that can be handled before encountering a wrap (which
69  * requires a full MMU zap).  The flag is instead explicitly queried when
70  * checking for MMIO spte cache hits.
71  */
72 
73 #define MMIO_SPTE_GEN_LOW_START		3
74 #define MMIO_SPTE_GEN_LOW_END		11
75 
76 #define MMIO_SPTE_GEN_HIGH_START	PT64_SECOND_AVAIL_BITS_SHIFT
77 #define MMIO_SPTE_GEN_HIGH_END		62
78 
79 #define MMIO_SPTE_GEN_LOW_MASK		GENMASK_ULL(MMIO_SPTE_GEN_LOW_END, \
80 						    MMIO_SPTE_GEN_LOW_START)
81 #define MMIO_SPTE_GEN_HIGH_MASK		GENMASK_ULL(MMIO_SPTE_GEN_HIGH_END, \
82 						    MMIO_SPTE_GEN_HIGH_START)
83 
84 #define MMIO_SPTE_GEN_LOW_BITS		(MMIO_SPTE_GEN_LOW_END - MMIO_SPTE_GEN_LOW_START + 1)
85 #define MMIO_SPTE_GEN_HIGH_BITS		(MMIO_SPTE_GEN_HIGH_END - MMIO_SPTE_GEN_HIGH_START + 1)
86 
87 /* remember to adjust the comment above as well if you change these */
88 static_assert(MMIO_SPTE_GEN_LOW_BITS == 9 && MMIO_SPTE_GEN_HIGH_BITS == 9);
89 
90 #define MMIO_SPTE_GEN_LOW_SHIFT		(MMIO_SPTE_GEN_LOW_START - 0)
91 #define MMIO_SPTE_GEN_HIGH_SHIFT	(MMIO_SPTE_GEN_HIGH_START - MMIO_SPTE_GEN_LOW_BITS)
92 
93 #define MMIO_SPTE_GEN_MASK		GENMASK_ULL(MMIO_SPTE_GEN_LOW_BITS + MMIO_SPTE_GEN_HIGH_BITS - 1, 0)
94 
95 extern u64 __read_mostly shadow_nx_mask;
96 extern u64 __read_mostly shadow_x_mask; /* mutual exclusive with nx_mask */
97 extern u64 __read_mostly shadow_user_mask;
98 extern u64 __read_mostly shadow_accessed_mask;
99 extern u64 __read_mostly shadow_dirty_mask;
100 extern u64 __read_mostly shadow_mmio_value;
101 extern u64 __read_mostly shadow_mmio_access_mask;
102 extern u64 __read_mostly shadow_present_mask;
103 extern u64 __read_mostly shadow_me_mask;
104 
105 /*
106  * SPTEs used by MMUs without A/D bits are marked with SPTE_AD_DISABLED_MASK;
107  * shadow_acc_track_mask is the set of bits to be cleared in non-accessed
108  * pages.
109  */
110 extern u64 __read_mostly shadow_acc_track_mask;
111 
112 /*
113  * This mask must be set on all non-zero Non-Present or Reserved SPTEs in order
114  * to guard against L1TF attacks.
115  */
116 extern u64 __read_mostly shadow_nonpresent_or_rsvd_mask;
117 
118 /*
119  * The number of high-order 1 bits to use in the mask above.
120  */
121 #define SHADOW_NONPRESENT_OR_RSVD_MASK_LEN 5
122 
123 /*
124  * The mask/shift to use for saving the original R/X bits when marking the PTE
125  * as not-present for access tracking purposes. We do not save the W bit as the
126  * PTEs being access tracked also need to be dirty tracked, so the W bit will be
127  * restored only when a write is attempted to the page.
128  */
129 #define SHADOW_ACC_TRACK_SAVED_BITS_MASK (PT64_EPT_READABLE_MASK | \
130 					  PT64_EPT_EXECUTABLE_MASK)
131 #define SHADOW_ACC_TRACK_SAVED_BITS_SHIFT PT64_SECOND_AVAIL_BITS_SHIFT
132 
133 /*
134  * If a thread running without exclusive control of the MMU lock must perform a
135  * multi-part operation on an SPTE, it can set the SPTE to REMOVED_SPTE as a
136  * non-present intermediate value. Other threads which encounter this value
137  * should not modify the SPTE.
138  *
139  * This constant works because it is considered non-present on both AMD and
140  * Intel CPUs and does not create a L1TF vulnerability because the pfn section
141  * is zeroed out.
142  *
143  * Only used by the TDP MMU.
144  */
145 #define REMOVED_SPTE (1ull << 59)
146 
147 static inline bool is_removed_spte(u64 spte)
148 {
149 	return spte == REMOVED_SPTE;
150 }
151 
152 /*
153  * In some cases, we need to preserve the GFN of a non-present or reserved
154  * SPTE when we usurp the upper five bits of the physical address space to
155  * defend against L1TF, e.g. for MMIO SPTEs.  To preserve the GFN, we'll
156  * shift bits of the GFN that overlap with shadow_nonpresent_or_rsvd_mask
157  * left into the reserved bits, i.e. the GFN in the SPTE will be split into
158  * high and low parts.  This mask covers the lower bits of the GFN.
159  */
160 extern u64 __read_mostly shadow_nonpresent_or_rsvd_lower_gfn_mask;
161 
162 /*
163  * The number of non-reserved physical address bits irrespective of features
164  * that repurpose legal bits, e.g. MKTME.
165  */
166 extern u8 __read_mostly shadow_phys_bits;
167 
168 static inline bool is_mmio_spte(u64 spte)
169 {
170 	return (spte & SPTE_SPECIAL_MASK) == SPTE_MMIO_MASK;
171 }
172 
173 static inline bool sp_ad_disabled(struct kvm_mmu_page *sp)
174 {
175 	return sp->role.ad_disabled;
176 }
177 
178 static inline bool spte_ad_enabled(u64 spte)
179 {
180 	MMU_WARN_ON(is_mmio_spte(spte));
181 	return (spte & SPTE_SPECIAL_MASK) != SPTE_AD_DISABLED_MASK;
182 }
183 
184 static inline bool spte_ad_need_write_protect(u64 spte)
185 {
186 	MMU_WARN_ON(is_mmio_spte(spte));
187 	return (spte & SPTE_SPECIAL_MASK) != SPTE_AD_ENABLED_MASK;
188 }
189 
190 static inline u64 spte_shadow_accessed_mask(u64 spte)
191 {
192 	MMU_WARN_ON(is_mmio_spte(spte));
193 	return spte_ad_enabled(spte) ? shadow_accessed_mask : 0;
194 }
195 
196 static inline u64 spte_shadow_dirty_mask(u64 spte)
197 {
198 	MMU_WARN_ON(is_mmio_spte(spte));
199 	return spte_ad_enabled(spte) ? shadow_dirty_mask : 0;
200 }
201 
202 static inline bool is_access_track_spte(u64 spte)
203 {
204 	return !spte_ad_enabled(spte) && (spte & shadow_acc_track_mask) == 0;
205 }
206 
207 static inline bool is_shadow_present_pte(u64 pte)
208 {
209 	return (pte != 0) && !is_mmio_spte(pte) && !is_removed_spte(pte);
210 }
211 
212 static inline bool is_large_pte(u64 pte)
213 {
214 	return pte & PT_PAGE_SIZE_MASK;
215 }
216 
217 static inline bool is_last_spte(u64 pte, int level)
218 {
219 	return (level == PG_LEVEL_4K) || is_large_pte(pte);
220 }
221 
222 static inline bool is_executable_pte(u64 spte)
223 {
224 	return (spte & (shadow_x_mask | shadow_nx_mask)) == shadow_x_mask;
225 }
226 
227 static inline kvm_pfn_t spte_to_pfn(u64 pte)
228 {
229 	return (pte & PT64_BASE_ADDR_MASK) >> PAGE_SHIFT;
230 }
231 
232 static inline bool is_accessed_spte(u64 spte)
233 {
234 	u64 accessed_mask = spte_shadow_accessed_mask(spte);
235 
236 	return accessed_mask ? spte & accessed_mask
237 			     : !is_access_track_spte(spte);
238 }
239 
240 static inline bool is_dirty_spte(u64 spte)
241 {
242 	u64 dirty_mask = spte_shadow_dirty_mask(spte);
243 
244 	return dirty_mask ? spte & dirty_mask : spte & PT_WRITABLE_MASK;
245 }
246 
247 static inline bool spte_can_locklessly_be_made_writable(u64 spte)
248 {
249 	return (spte & (SPTE_HOST_WRITEABLE | SPTE_MMU_WRITEABLE)) ==
250 		(SPTE_HOST_WRITEABLE | SPTE_MMU_WRITEABLE);
251 }
252 
253 static inline u64 get_mmio_spte_generation(u64 spte)
254 {
255 	u64 gen;
256 
257 	gen = (spte & MMIO_SPTE_GEN_LOW_MASK) >> MMIO_SPTE_GEN_LOW_SHIFT;
258 	gen |= (spte & MMIO_SPTE_GEN_HIGH_MASK) >> MMIO_SPTE_GEN_HIGH_SHIFT;
259 	return gen;
260 }
261 
262 /* Bits which may be returned by set_spte() */
263 #define SET_SPTE_WRITE_PROTECTED_PT    BIT(0)
264 #define SET_SPTE_NEED_REMOTE_TLB_FLUSH BIT(1)
265 #define SET_SPTE_SPURIOUS              BIT(2)
266 
267 int make_spte(struct kvm_vcpu *vcpu, unsigned int pte_access, int level,
268 		     gfn_t gfn, kvm_pfn_t pfn, u64 old_spte, bool speculative,
269 		     bool can_unsync, bool host_writable, bool ad_disabled,
270 		     u64 *new_spte);
271 u64 make_nonleaf_spte(u64 *child_pt, bool ad_disabled);
272 u64 make_mmio_spte(struct kvm_vcpu *vcpu, u64 gfn, unsigned int access);
273 u64 mark_spte_for_access_track(u64 spte);
274 u64 kvm_mmu_changed_pte_notifier_make_spte(u64 old_spte, kvm_pfn_t new_pfn);
275 
276 void kvm_mmu_reset_all_pte_masks(void);
277 
278 #endif
279