1 #ifndef _ASM_X86_PGTABLE_DEFS_H
2 #define _ASM_X86_PGTABLE_DEFS_H
3 
4 #include <linux/const.h>
5 #include <asm/page_types.h>
6 
7 #define FIRST_USER_ADDRESS	0UL
8 
9 #define _PAGE_BIT_PRESENT	0	/* is present */
10 #define _PAGE_BIT_RW		1	/* writeable */
11 #define _PAGE_BIT_USER		2	/* userspace addressable */
12 #define _PAGE_BIT_PWT		3	/* page write through */
13 #define _PAGE_BIT_PCD		4	/* page cache disabled */
14 #define _PAGE_BIT_ACCESSED	5	/* was accessed (raised by CPU) */
15 #define _PAGE_BIT_DIRTY		6	/* was written to (raised by CPU) */
16 #define _PAGE_BIT_PSE		7	/* 4 MB (or 2MB) page */
17 #define _PAGE_BIT_PAT		7	/* on 4KB pages */
18 #define _PAGE_BIT_GLOBAL	8	/* Global TLB entry PPro+ */
19 #define _PAGE_BIT_SOFTW1	9	/* available for programmer */
20 #define _PAGE_BIT_SOFTW2	10	/* " */
21 #define _PAGE_BIT_SOFTW3	11	/* " */
22 #define _PAGE_BIT_PAT_LARGE	12	/* On 2MB or 1GB pages */
23 #define _PAGE_BIT_SPECIAL	_PAGE_BIT_SOFTW1
24 #define _PAGE_BIT_CPA_TEST	_PAGE_BIT_SOFTW1
25 #define _PAGE_BIT_SPLITTING	_PAGE_BIT_SOFTW2 /* only valid on a PSE pmd */
26 #define _PAGE_BIT_HIDDEN	_PAGE_BIT_SOFTW3 /* hidden by kmemcheck */
27 #define _PAGE_BIT_SOFT_DIRTY	_PAGE_BIT_SOFTW3 /* software dirty tracking */
28 #define _PAGE_BIT_NX           63       /* No execute: only valid after cpuid check */
29 
30 /* If _PAGE_BIT_PRESENT is clear, we use these: */
31 /* - if the user mapped it with PROT_NONE; pte_present gives true */
32 #define _PAGE_BIT_PROTNONE	_PAGE_BIT_GLOBAL
33 
34 #define _PAGE_PRESENT	(_AT(pteval_t, 1) << _PAGE_BIT_PRESENT)
35 #define _PAGE_RW	(_AT(pteval_t, 1) << _PAGE_BIT_RW)
36 #define _PAGE_USER	(_AT(pteval_t, 1) << _PAGE_BIT_USER)
37 #define _PAGE_PWT	(_AT(pteval_t, 1) << _PAGE_BIT_PWT)
38 #define _PAGE_PCD	(_AT(pteval_t, 1) << _PAGE_BIT_PCD)
39 #define _PAGE_ACCESSED	(_AT(pteval_t, 1) << _PAGE_BIT_ACCESSED)
40 #define _PAGE_DIRTY	(_AT(pteval_t, 1) << _PAGE_BIT_DIRTY)
41 #define _PAGE_PSE	(_AT(pteval_t, 1) << _PAGE_BIT_PSE)
42 #define _PAGE_GLOBAL	(_AT(pteval_t, 1) << _PAGE_BIT_GLOBAL)
43 #define _PAGE_SOFTW1	(_AT(pteval_t, 1) << _PAGE_BIT_SOFTW1)
44 #define _PAGE_SOFTW2	(_AT(pteval_t, 1) << _PAGE_BIT_SOFTW2)
45 #define _PAGE_PAT	(_AT(pteval_t, 1) << _PAGE_BIT_PAT)
46 #define _PAGE_PAT_LARGE (_AT(pteval_t, 1) << _PAGE_BIT_PAT_LARGE)
47 #define _PAGE_SPECIAL	(_AT(pteval_t, 1) << _PAGE_BIT_SPECIAL)
48 #define _PAGE_CPA_TEST	(_AT(pteval_t, 1) << _PAGE_BIT_CPA_TEST)
49 #define _PAGE_SPLITTING	(_AT(pteval_t, 1) << _PAGE_BIT_SPLITTING)
50 #define __HAVE_ARCH_PTE_SPECIAL
51 
52 #ifdef CONFIG_KMEMCHECK
53 #define _PAGE_HIDDEN	(_AT(pteval_t, 1) << _PAGE_BIT_HIDDEN)
54 #else
55 #define _PAGE_HIDDEN	(_AT(pteval_t, 0))
56 #endif
57 
58 /*
59  * The same hidden bit is used by kmemcheck, but since kmemcheck
60  * works on kernel pages while soft-dirty engine on user space,
61  * they do not conflict with each other.
62  */
63 
64 #ifdef CONFIG_MEM_SOFT_DIRTY
65 #define _PAGE_SOFT_DIRTY	(_AT(pteval_t, 1) << _PAGE_BIT_SOFT_DIRTY)
66 #else
67 #define _PAGE_SOFT_DIRTY	(_AT(pteval_t, 0))
68 #endif
69 
70 /*
71  * Tracking soft dirty bit when a page goes to a swap is tricky.
72  * We need a bit which can be stored in pte _and_ not conflict
73  * with swap entry format. On x86 bits 6 and 7 are *not* involved
74  * into swap entry computation, but bit 6 is used for nonlinear
75  * file mapping, so we borrow bit 7 for soft dirty tracking.
76  *
77  * Please note that this bit must be treated as swap dirty page
78  * mark if and only if the PTE has present bit clear!
79  */
80 #ifdef CONFIG_MEM_SOFT_DIRTY
81 #define _PAGE_SWP_SOFT_DIRTY	_PAGE_PSE
82 #else
83 #define _PAGE_SWP_SOFT_DIRTY	(_AT(pteval_t, 0))
84 #endif
85 
86 #if defined(CONFIG_X86_64) || defined(CONFIG_X86_PAE)
87 #define _PAGE_NX	(_AT(pteval_t, 1) << _PAGE_BIT_NX)
88 #else
89 #define _PAGE_NX	(_AT(pteval_t, 0))
90 #endif
91 
92 #define _PAGE_PROTNONE	(_AT(pteval_t, 1) << _PAGE_BIT_PROTNONE)
93 
94 #define _PAGE_TABLE	(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER |	\
95 			 _PAGE_ACCESSED | _PAGE_DIRTY)
96 #define _KERNPG_TABLE	(_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED |	\
97 			 _PAGE_DIRTY)
98 
99 /* Set of bits not changed in pte_modify */
100 #define _PAGE_CHG_MASK	(PTE_PFN_MASK | _PAGE_PCD | _PAGE_PWT |		\
101 			 _PAGE_SPECIAL | _PAGE_ACCESSED | _PAGE_DIRTY |	\
102 			 _PAGE_SOFT_DIRTY)
103 #define _HPAGE_CHG_MASK (_PAGE_CHG_MASK | _PAGE_PSE)
104 
105 /*
106  * The cache modes defined here are used to translate between pure SW usage
107  * and the HW defined cache mode bits and/or PAT entries.
108  *
109  * The resulting bits for PWT, PCD and PAT should be chosen in a way
110  * to have the WB mode at index 0 (all bits clear). This is the default
111  * right now and likely would break too much if changed.
112  */
113 #ifndef __ASSEMBLY__
114 enum page_cache_mode {
115 	_PAGE_CACHE_MODE_WB = 0,
116 	_PAGE_CACHE_MODE_WC = 1,
117 	_PAGE_CACHE_MODE_UC_MINUS = 2,
118 	_PAGE_CACHE_MODE_UC = 3,
119 	_PAGE_CACHE_MODE_WT = 4,
120 	_PAGE_CACHE_MODE_WP = 5,
121 	_PAGE_CACHE_MODE_NUM = 8
122 };
123 #endif
124 
125 #define _PAGE_CACHE_MASK	(_PAGE_PAT | _PAGE_PCD | _PAGE_PWT)
126 #define _PAGE_NOCACHE		(cachemode2protval(_PAGE_CACHE_MODE_UC))
127 
128 #define PAGE_NONE	__pgprot(_PAGE_PROTNONE | _PAGE_ACCESSED)
129 #define PAGE_SHARED	__pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | \
130 				 _PAGE_ACCESSED | _PAGE_NX)
131 
132 #define PAGE_SHARED_EXEC	__pgprot(_PAGE_PRESENT | _PAGE_RW |	\
133 					 _PAGE_USER | _PAGE_ACCESSED)
134 #define PAGE_COPY_NOEXEC	__pgprot(_PAGE_PRESENT | _PAGE_USER |	\
135 					 _PAGE_ACCESSED | _PAGE_NX)
136 #define PAGE_COPY_EXEC		__pgprot(_PAGE_PRESENT | _PAGE_USER |	\
137 					 _PAGE_ACCESSED)
138 #define PAGE_COPY		PAGE_COPY_NOEXEC
139 #define PAGE_READONLY		__pgprot(_PAGE_PRESENT | _PAGE_USER |	\
140 					 _PAGE_ACCESSED | _PAGE_NX)
141 #define PAGE_READONLY_EXEC	__pgprot(_PAGE_PRESENT | _PAGE_USER |	\
142 					 _PAGE_ACCESSED)
143 
144 #define __PAGE_KERNEL_EXEC						\
145 	(_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_GLOBAL)
146 #define __PAGE_KERNEL		(__PAGE_KERNEL_EXEC | _PAGE_NX)
147 
148 #define __PAGE_KERNEL_RO		(__PAGE_KERNEL & ~_PAGE_RW)
149 #define __PAGE_KERNEL_RX		(__PAGE_KERNEL_EXEC & ~_PAGE_RW)
150 #define __PAGE_KERNEL_NOCACHE		(__PAGE_KERNEL | _PAGE_NOCACHE)
151 #define __PAGE_KERNEL_VSYSCALL		(__PAGE_KERNEL_RX | _PAGE_USER)
152 #define __PAGE_KERNEL_VVAR		(__PAGE_KERNEL_RO | _PAGE_USER)
153 #define __PAGE_KERNEL_LARGE		(__PAGE_KERNEL | _PAGE_PSE)
154 #define __PAGE_KERNEL_LARGE_EXEC	(__PAGE_KERNEL_EXEC | _PAGE_PSE)
155 
156 #define __PAGE_KERNEL_IO		(__PAGE_KERNEL)
157 #define __PAGE_KERNEL_IO_NOCACHE	(__PAGE_KERNEL_NOCACHE)
158 
159 #define PAGE_KERNEL			__pgprot(__PAGE_KERNEL)
160 #define PAGE_KERNEL_RO			__pgprot(__PAGE_KERNEL_RO)
161 #define PAGE_KERNEL_EXEC		__pgprot(__PAGE_KERNEL_EXEC)
162 #define PAGE_KERNEL_RX			__pgprot(__PAGE_KERNEL_RX)
163 #define PAGE_KERNEL_NOCACHE		__pgprot(__PAGE_KERNEL_NOCACHE)
164 #define PAGE_KERNEL_LARGE		__pgprot(__PAGE_KERNEL_LARGE)
165 #define PAGE_KERNEL_LARGE_EXEC		__pgprot(__PAGE_KERNEL_LARGE_EXEC)
166 #define PAGE_KERNEL_VSYSCALL		__pgprot(__PAGE_KERNEL_VSYSCALL)
167 #define PAGE_KERNEL_VVAR		__pgprot(__PAGE_KERNEL_VVAR)
168 
169 #define PAGE_KERNEL_IO			__pgprot(__PAGE_KERNEL_IO)
170 #define PAGE_KERNEL_IO_NOCACHE		__pgprot(__PAGE_KERNEL_IO_NOCACHE)
171 
172 /*         xwr */
173 #define __P000	PAGE_NONE
174 #define __P001	PAGE_READONLY
175 #define __P010	PAGE_COPY
176 #define __P011	PAGE_COPY
177 #define __P100	PAGE_READONLY_EXEC
178 #define __P101	PAGE_READONLY_EXEC
179 #define __P110	PAGE_COPY_EXEC
180 #define __P111	PAGE_COPY_EXEC
181 
182 #define __S000	PAGE_NONE
183 #define __S001	PAGE_READONLY
184 #define __S010	PAGE_SHARED
185 #define __S011	PAGE_SHARED
186 #define __S100	PAGE_READONLY_EXEC
187 #define __S101	PAGE_READONLY_EXEC
188 #define __S110	PAGE_SHARED_EXEC
189 #define __S111	PAGE_SHARED_EXEC
190 
191 /*
192  * early identity mapping  pte attrib macros.
193  */
194 #ifdef CONFIG_X86_64
195 #define __PAGE_KERNEL_IDENT_LARGE_EXEC	__PAGE_KERNEL_LARGE_EXEC
196 #else
197 #define PTE_IDENT_ATTR	 0x003		/* PRESENT+RW */
198 #define PDE_IDENT_ATTR	 0x063		/* PRESENT+RW+DIRTY+ACCESSED */
199 #define PGD_IDENT_ATTR	 0x001		/* PRESENT (no other attributes) */
200 #endif
201 
202 #ifdef CONFIG_X86_32
203 # include <asm/pgtable_32_types.h>
204 #else
205 # include <asm/pgtable_64_types.h>
206 #endif
207 
208 #ifndef __ASSEMBLY__
209 
210 #include <linux/types.h>
211 
212 /* PTE_PFN_MASK extracts the PFN from a (pte|pmd|pud|pgd)val_t */
213 #define PTE_PFN_MASK		((pteval_t)PHYSICAL_PAGE_MASK)
214 
215 /* PTE_FLAGS_MASK extracts the flags from a (pte|pmd|pud|pgd)val_t */
216 #define PTE_FLAGS_MASK		(~PTE_PFN_MASK)
217 
218 typedef struct pgprot { pgprotval_t pgprot; } pgprot_t;
219 
220 typedef struct { pgdval_t pgd; } pgd_t;
221 
native_make_pgd(pgdval_t val)222 static inline pgd_t native_make_pgd(pgdval_t val)
223 {
224 	return (pgd_t) { val };
225 }
226 
native_pgd_val(pgd_t pgd)227 static inline pgdval_t native_pgd_val(pgd_t pgd)
228 {
229 	return pgd.pgd;
230 }
231 
pgd_flags(pgd_t pgd)232 static inline pgdval_t pgd_flags(pgd_t pgd)
233 {
234 	return native_pgd_val(pgd) & PTE_FLAGS_MASK;
235 }
236 
237 #if CONFIG_PGTABLE_LEVELS > 3
238 typedef struct { pudval_t pud; } pud_t;
239 
native_make_pud(pmdval_t val)240 static inline pud_t native_make_pud(pmdval_t val)
241 {
242 	return (pud_t) { val };
243 }
244 
native_pud_val(pud_t pud)245 static inline pudval_t native_pud_val(pud_t pud)
246 {
247 	return pud.pud;
248 }
249 #else
250 #include <asm-generic/pgtable-nopud.h>
251 
native_pud_val(pud_t pud)252 static inline pudval_t native_pud_val(pud_t pud)
253 {
254 	return native_pgd_val(pud.pgd);
255 }
256 #endif
257 
258 #if CONFIG_PGTABLE_LEVELS > 2
259 typedef struct { pmdval_t pmd; } pmd_t;
260 
native_make_pmd(pmdval_t val)261 static inline pmd_t native_make_pmd(pmdval_t val)
262 {
263 	return (pmd_t) { val };
264 }
265 
native_pmd_val(pmd_t pmd)266 static inline pmdval_t native_pmd_val(pmd_t pmd)
267 {
268 	return pmd.pmd;
269 }
270 #else
271 #include <asm-generic/pgtable-nopmd.h>
272 
native_pmd_val(pmd_t pmd)273 static inline pmdval_t native_pmd_val(pmd_t pmd)
274 {
275 	return native_pgd_val(pmd.pud.pgd);
276 }
277 #endif
278 
pud_flags(pud_t pud)279 static inline pudval_t pud_flags(pud_t pud)
280 {
281 	return native_pud_val(pud) & PTE_FLAGS_MASK;
282 }
283 
pmd_flags(pmd_t pmd)284 static inline pmdval_t pmd_flags(pmd_t pmd)
285 {
286 	return native_pmd_val(pmd) & PTE_FLAGS_MASK;
287 }
288 
native_make_pte(pteval_t val)289 static inline pte_t native_make_pte(pteval_t val)
290 {
291 	return (pte_t) { .pte = val };
292 }
293 
native_pte_val(pte_t pte)294 static inline pteval_t native_pte_val(pte_t pte)
295 {
296 	return pte.pte;
297 }
298 
pte_flags(pte_t pte)299 static inline pteval_t pte_flags(pte_t pte)
300 {
301 	return native_pte_val(pte) & PTE_FLAGS_MASK;
302 }
303 
304 #define pgprot_val(x)	((x).pgprot)
305 #define __pgprot(x)	((pgprot_t) { (x) } )
306 
307 extern uint16_t __cachemode2pte_tbl[_PAGE_CACHE_MODE_NUM];
308 extern uint8_t __pte2cachemode_tbl[8];
309 
310 #define __pte2cm_idx(cb)				\
311 	((((cb) >> (_PAGE_BIT_PAT - 2)) & 4) |		\
312 	 (((cb) >> (_PAGE_BIT_PCD - 1)) & 2) |		\
313 	 (((cb) >> _PAGE_BIT_PWT) & 1))
314 #define __cm_idx2pte(i)					\
315 	((((i) & 4) << (_PAGE_BIT_PAT - 2)) |		\
316 	 (((i) & 2) << (_PAGE_BIT_PCD - 1)) |		\
317 	 (((i) & 1) << _PAGE_BIT_PWT))
318 
cachemode2protval(enum page_cache_mode pcm)319 static inline unsigned long cachemode2protval(enum page_cache_mode pcm)
320 {
321 	if (likely(pcm == 0))
322 		return 0;
323 	return __cachemode2pte_tbl[pcm];
324 }
cachemode2pgprot(enum page_cache_mode pcm)325 static inline pgprot_t cachemode2pgprot(enum page_cache_mode pcm)
326 {
327 	return __pgprot(cachemode2protval(pcm));
328 }
pgprot2cachemode(pgprot_t pgprot)329 static inline enum page_cache_mode pgprot2cachemode(pgprot_t pgprot)
330 {
331 	unsigned long masked;
332 
333 	masked = pgprot_val(pgprot) & _PAGE_CACHE_MASK;
334 	if (likely(masked == 0))
335 		return 0;
336 	return __pte2cachemode_tbl[__pte2cm_idx(masked)];
337 }
pgprot_4k_2_large(pgprot_t pgprot)338 static inline pgprot_t pgprot_4k_2_large(pgprot_t pgprot)
339 {
340 	pgprotval_t val = pgprot_val(pgprot);
341 	pgprot_t new;
342 
343 	pgprot_val(new) = (val & ~(_PAGE_PAT | _PAGE_PAT_LARGE)) |
344 		((val & _PAGE_PAT) << (_PAGE_BIT_PAT_LARGE - _PAGE_BIT_PAT));
345 	return new;
346 }
pgprot_large_2_4k(pgprot_t pgprot)347 static inline pgprot_t pgprot_large_2_4k(pgprot_t pgprot)
348 {
349 	pgprotval_t val = pgprot_val(pgprot);
350 	pgprot_t new;
351 
352 	pgprot_val(new) = (val & ~(_PAGE_PAT | _PAGE_PAT_LARGE)) |
353 			  ((val & _PAGE_PAT_LARGE) >>
354 			   (_PAGE_BIT_PAT_LARGE - _PAGE_BIT_PAT));
355 	return new;
356 }
357 
358 
359 typedef struct page *pgtable_t;
360 
361 extern pteval_t __supported_pte_mask;
362 extern void set_nx(void);
363 extern int nx_enabled;
364 
365 #define pgprot_writecombine	pgprot_writecombine
366 extern pgprot_t pgprot_writecombine(pgprot_t prot);
367 
368 /* Indicate that x86 has its own track and untrack pfn vma functions */
369 #define __HAVE_PFNMAP_TRACKING
370 
371 #define __HAVE_PHYS_MEM_ACCESS_PROT
372 struct file;
373 pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
374                               unsigned long size, pgprot_t vma_prot);
375 int phys_mem_access_prot_allowed(struct file *file, unsigned long pfn,
376                               unsigned long size, pgprot_t *vma_prot);
377 
378 /* Install a pte for a particular vaddr in kernel space. */
379 void set_pte_vaddr(unsigned long vaddr, pte_t pte);
380 
381 #ifdef CONFIG_X86_32
382 extern void native_pagetable_init(void);
383 #else
384 #define native_pagetable_init        paging_init
385 #endif
386 
387 struct seq_file;
388 extern void arch_report_meminfo(struct seq_file *m);
389 
390 enum pg_level {
391 	PG_LEVEL_NONE,
392 	PG_LEVEL_4K,
393 	PG_LEVEL_2M,
394 	PG_LEVEL_1G,
395 	PG_LEVEL_NUM
396 };
397 
398 #ifdef CONFIG_PROC_FS
399 extern void update_page_count(int level, unsigned long pages);
400 #else
update_page_count(int level,unsigned long pages)401 static inline void update_page_count(int level, unsigned long pages) { }
402 #endif
403 
404 /*
405  * Helper function that returns the kernel pagetable entry controlling
406  * the virtual address 'address'. NULL means no pagetable entry present.
407  * NOTE: the return type is pte_t but if the pmd is PSE then we return it
408  * as a pte too.
409  */
410 extern pte_t *lookup_address(unsigned long address, unsigned int *level);
411 extern pte_t *lookup_address_in_pgd(pgd_t *pgd, unsigned long address,
412 				    unsigned int *level);
413 extern pmd_t *lookup_pmd_address(unsigned long address);
414 extern phys_addr_t slow_virt_to_phys(void *__address);
415 extern int kernel_map_pages_in_pgd(pgd_t *pgd, u64 pfn, unsigned long address,
416 				   unsigned numpages, unsigned long page_flags);
417 void kernel_unmap_pages_in_pgd(pgd_t *root, unsigned long address,
418 			       unsigned numpages);
419 #endif	/* !__ASSEMBLY__ */
420 
421 #endif /* _ASM_X86_PGTABLE_DEFS_H */
422