xref: /openbmc/linux/arch/x86/include/asm/pgtable_64.h (revision 9a8f3203)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _ASM_X86_PGTABLE_64_H
3 #define _ASM_X86_PGTABLE_64_H
4 
5 #include <linux/const.h>
6 #include <asm/pgtable_64_types.h>
7 
8 #ifndef __ASSEMBLY__
9 
10 /*
11  * This file contains the functions and defines necessary to modify and use
12  * the x86-64 page table tree.
13  */
14 #include <asm/processor.h>
15 #include <linux/bitops.h>
16 #include <linux/threads.h>
17 #include <asm/fixmap.h>
18 
19 extern p4d_t level4_kernel_pgt[512];
20 extern p4d_t level4_ident_pgt[512];
21 extern pud_t level3_kernel_pgt[512];
22 extern pud_t level3_ident_pgt[512];
23 extern pmd_t level2_kernel_pgt[512];
24 extern pmd_t level2_fixmap_pgt[512];
25 extern pmd_t level2_ident_pgt[512];
26 extern pte_t level1_fixmap_pgt[512 * FIXMAP_PMD_NUM];
27 extern pgd_t init_top_pgt[];
28 
29 #define swapper_pg_dir init_top_pgt
30 
31 extern void paging_init(void);
32 static inline void sync_initial_page_table(void) { }
33 
34 #define pte_ERROR(e)					\
35 	pr_err("%s:%d: bad pte %p(%016lx)\n",		\
36 	       __FILE__, __LINE__, &(e), pte_val(e))
37 #define pmd_ERROR(e)					\
38 	pr_err("%s:%d: bad pmd %p(%016lx)\n",		\
39 	       __FILE__, __LINE__, &(e), pmd_val(e))
40 #define pud_ERROR(e)					\
41 	pr_err("%s:%d: bad pud %p(%016lx)\n",		\
42 	       __FILE__, __LINE__, &(e), pud_val(e))
43 
44 #if CONFIG_PGTABLE_LEVELS >= 5
45 #define p4d_ERROR(e)					\
46 	pr_err("%s:%d: bad p4d %p(%016lx)\n",		\
47 	       __FILE__, __LINE__, &(e), p4d_val(e))
48 #endif
49 
50 #define pgd_ERROR(e)					\
51 	pr_err("%s:%d: bad pgd %p(%016lx)\n",		\
52 	       __FILE__, __LINE__, &(e), pgd_val(e))
53 
54 struct mm_struct;
55 
56 void set_pte_vaddr_p4d(p4d_t *p4d_page, unsigned long vaddr, pte_t new_pte);
57 void set_pte_vaddr_pud(pud_t *pud_page, unsigned long vaddr, pte_t new_pte);
58 
59 static inline void native_set_pte(pte_t *ptep, pte_t pte)
60 {
61 	WRITE_ONCE(*ptep, pte);
62 }
63 
64 static inline void native_pte_clear(struct mm_struct *mm, unsigned long addr,
65 				    pte_t *ptep)
66 {
67 	native_set_pte(ptep, native_make_pte(0));
68 }
69 
70 static inline void native_set_pte_atomic(pte_t *ptep, pte_t pte)
71 {
72 	native_set_pte(ptep, pte);
73 }
74 
75 static inline void native_set_pmd(pmd_t *pmdp, pmd_t pmd)
76 {
77 	WRITE_ONCE(*pmdp, pmd);
78 }
79 
80 static inline void native_pmd_clear(pmd_t *pmd)
81 {
82 	native_set_pmd(pmd, native_make_pmd(0));
83 }
84 
85 static inline pte_t native_ptep_get_and_clear(pte_t *xp)
86 {
87 #ifdef CONFIG_SMP
88 	return native_make_pte(xchg(&xp->pte, 0));
89 #else
90 	/* native_local_ptep_get_and_clear,
91 	   but duplicated because of cyclic dependency */
92 	pte_t ret = *xp;
93 	native_pte_clear(NULL, 0, xp);
94 	return ret;
95 #endif
96 }
97 
98 static inline pmd_t native_pmdp_get_and_clear(pmd_t *xp)
99 {
100 #ifdef CONFIG_SMP
101 	return native_make_pmd(xchg(&xp->pmd, 0));
102 #else
103 	/* native_local_pmdp_get_and_clear,
104 	   but duplicated because of cyclic dependency */
105 	pmd_t ret = *xp;
106 	native_pmd_clear(xp);
107 	return ret;
108 #endif
109 }
110 
111 static inline void native_set_pud(pud_t *pudp, pud_t pud)
112 {
113 	WRITE_ONCE(*pudp, pud);
114 }
115 
116 static inline void native_pud_clear(pud_t *pud)
117 {
118 	native_set_pud(pud, native_make_pud(0));
119 }
120 
121 static inline pud_t native_pudp_get_and_clear(pud_t *xp)
122 {
123 #ifdef CONFIG_SMP
124 	return native_make_pud(xchg(&xp->pud, 0));
125 #else
126 	/* native_local_pudp_get_and_clear,
127 	 * but duplicated because of cyclic dependency
128 	 */
129 	pud_t ret = *xp;
130 
131 	native_pud_clear(xp);
132 	return ret;
133 #endif
134 }
135 
136 static inline void native_set_p4d(p4d_t *p4dp, p4d_t p4d)
137 {
138 	pgd_t pgd;
139 
140 	if (pgtable_l5_enabled() || !IS_ENABLED(CONFIG_PAGE_TABLE_ISOLATION)) {
141 		WRITE_ONCE(*p4dp, p4d);
142 		return;
143 	}
144 
145 	pgd = native_make_pgd(native_p4d_val(p4d));
146 	pgd = pti_set_user_pgtbl((pgd_t *)p4dp, pgd);
147 	WRITE_ONCE(*p4dp, native_make_p4d(native_pgd_val(pgd)));
148 }
149 
150 static inline void native_p4d_clear(p4d_t *p4d)
151 {
152 	native_set_p4d(p4d, native_make_p4d(0));
153 }
154 
155 static inline void native_set_pgd(pgd_t *pgdp, pgd_t pgd)
156 {
157 	WRITE_ONCE(*pgdp, pti_set_user_pgtbl(pgdp, pgd));
158 }
159 
160 static inline void native_pgd_clear(pgd_t *pgd)
161 {
162 	native_set_pgd(pgd, native_make_pgd(0));
163 }
164 
165 extern void sync_global_pgds(unsigned long start, unsigned long end);
166 
167 /*
168  * Conversion functions: convert a page and protection to a page entry,
169  * and a page entry and page directory to the page they refer to.
170  */
171 
172 /*
173  * Level 4 access.
174  */
175 #define mk_kernel_pgd(address) __pgd((address) | _KERNPG_TABLE)
176 
177 /* PUD - Level3 access */
178 
179 /* PMD  - Level 2 access */
180 
181 /* PTE - Level 1 access. */
182 
183 /* x86-64 always has all page tables mapped. */
184 #define pte_offset_map(dir, address) pte_offset_kernel((dir), (address))
185 #define pte_unmap(pte) ((void)(pte))/* NOP */
186 
187 /*
188  * Encode and de-code a swap entry
189  *
190  * |     ...            | 11| 10|  9|8|7|6|5| 4| 3|2| 1|0| <- bit number
191  * |     ...            |SW3|SW2|SW1|G|L|D|A|CD|WT|U| W|P| <- bit names
192  * | TYPE (59-63) | ~OFFSET (9-58)  |0|0|X|X| X| X|X|SD|0| <- swp entry
193  *
194  * G (8) is aliased and used as a PROT_NONE indicator for
195  * !present ptes.  We need to start storing swap entries above
196  * there.  We also need to avoid using A and D because of an
197  * erratum where they can be incorrectly set by hardware on
198  * non-present PTEs.
199  *
200  * SD (1) in swp entry is used to store soft dirty bit, which helps us
201  * remember soft dirty over page migration
202  *
203  * Bit 7 in swp entry should be 0 because pmd_present checks not only P,
204  * but also L and G.
205  *
206  * The offset is inverted by a binary not operation to make the high
207  * physical bits set.
208  */
209 #define SWP_TYPE_BITS		5
210 
211 #define SWP_OFFSET_FIRST_BIT	(_PAGE_BIT_PROTNONE + 1)
212 
213 /* We always extract/encode the offset by shifting it all the way up, and then down again */
214 #define SWP_OFFSET_SHIFT	(SWP_OFFSET_FIRST_BIT+SWP_TYPE_BITS)
215 
216 #define MAX_SWAPFILES_CHECK() BUILD_BUG_ON(MAX_SWAPFILES_SHIFT > SWP_TYPE_BITS)
217 
218 /* Extract the high bits for type */
219 #define __swp_type(x) ((x).val >> (64 - SWP_TYPE_BITS))
220 
221 /* Shift up (to get rid of type), then down to get value */
222 #define __swp_offset(x) (~(x).val << SWP_TYPE_BITS >> SWP_OFFSET_SHIFT)
223 
224 /*
225  * Shift the offset up "too far" by TYPE bits, then down again
226  * The offset is inverted by a binary not operation to make the high
227  * physical bits set.
228  */
229 #define __swp_entry(type, offset) ((swp_entry_t) { \
230 	(~(unsigned long)(offset) << SWP_OFFSET_SHIFT >> SWP_TYPE_BITS) \
231 	| ((unsigned long)(type) << (64-SWP_TYPE_BITS)) })
232 
233 #define __pte_to_swp_entry(pte)		((swp_entry_t) { pte_val((pte)) })
234 #define __pmd_to_swp_entry(pmd)		((swp_entry_t) { pmd_val((pmd)) })
235 #define __swp_entry_to_pte(x)		((pte_t) { .pte = (x).val })
236 #define __swp_entry_to_pmd(x)		((pmd_t) { .pmd = (x).val })
237 
238 extern int kern_addr_valid(unsigned long addr);
239 extern void cleanup_highmap(void);
240 
241 #define HAVE_ARCH_UNMAPPED_AREA
242 #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
243 
244 #define pgtable_cache_init()   do { } while (0)
245 #define check_pgt_cache()      do { } while (0)
246 
247 #define PAGE_AGP    PAGE_KERNEL_NOCACHE
248 #define HAVE_PAGE_AGP 1
249 
250 /* fs/proc/kcore.c */
251 #define	kc_vaddr_to_offset(v) ((v) & __VIRTUAL_MASK)
252 #define	kc_offset_to_vaddr(o) ((o) | ~__VIRTUAL_MASK)
253 
254 #define __HAVE_ARCH_PTE_SAME
255 
256 #define vmemmap ((struct page *)VMEMMAP_START)
257 
258 extern void init_extra_mapping_uc(unsigned long phys, unsigned long size);
259 extern void init_extra_mapping_wb(unsigned long phys, unsigned long size);
260 
261 #define gup_fast_permitted gup_fast_permitted
262 static inline bool gup_fast_permitted(unsigned long start, int nr_pages)
263 {
264 	unsigned long len, end;
265 
266 	len = (unsigned long)nr_pages << PAGE_SHIFT;
267 	end = start + len;
268 	if (end < start)
269 		return false;
270 	if (end >> __VIRTUAL_MASK_SHIFT)
271 		return false;
272 	return true;
273 }
274 
275 #include <asm/pgtable-invert.h>
276 
277 #endif /* !__ASSEMBLY__ */
278 #endif /* _ASM_X86_PGTABLE_64_H */
279