1 #ifndef _ASM_X86_PGTABLE_H
2 #define _ASM_X86_PGTABLE_H
3
4 #include <asm/page.h>
5 #include <asm/e820.h>
6
7 #include <asm/pgtable_types.h>
8
9 /*
10 * Macro to mark a page protection value as UC-
11 */
12 #define pgprot_noncached(prot) \
13 ((boot_cpu_data.x86 > 3) \
14 ? (__pgprot(pgprot_val(prot) | \
15 cachemode2protval(_PAGE_CACHE_MODE_UC_MINUS))) \
16 : (prot))
17
18 #ifndef __ASSEMBLY__
19 #include <asm/x86_init.h>
20
21 #ifdef CONFIG_PAGE_TABLE_ISOLATION
22 extern int kaiser_enabled;
23 #else
24 #define kaiser_enabled 0
25 #endif
26
27 void ptdump_walk_pgd_level(struct seq_file *m, pgd_t *pgd);
28 void ptdump_walk_pgd_level_checkwx(void);
29
30 #ifdef CONFIG_DEBUG_WX
31 #define debug_checkwx() ptdump_walk_pgd_level_checkwx()
32 #else
33 #define debug_checkwx() do { } while (0)
34 #endif
35
36 /*
37 * ZERO_PAGE is a global shared page that is always zero: used
38 * for zero-mapped memory areas etc..
39 */
40 extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)]
41 __visible;
42 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
43
44 extern spinlock_t pgd_lock;
45 extern struct list_head pgd_list;
46
47 extern struct mm_struct *pgd_page_get_mm(struct page *page);
48
49 #ifdef CONFIG_PARAVIRT
50 #include <asm/paravirt.h>
51 #else /* !CONFIG_PARAVIRT */
52 #define set_pte(ptep, pte) native_set_pte(ptep, pte)
53 #define set_pte_at(mm, addr, ptep, pte) native_set_pte_at(mm, addr, ptep, pte)
54 #define set_pmd_at(mm, addr, pmdp, pmd) native_set_pmd_at(mm, addr, pmdp, pmd)
55
56 #define set_pte_atomic(ptep, pte) \
57 native_set_pte_atomic(ptep, pte)
58
59 #define set_pmd(pmdp, pmd) native_set_pmd(pmdp, pmd)
60
61 #ifndef __PAGETABLE_PUD_FOLDED
62 #define set_pgd(pgdp, pgd) native_set_pgd(pgdp, pgd)
63 #define pgd_clear(pgd) native_pgd_clear(pgd)
64 #endif
65
66 #ifndef set_pud
67 # define set_pud(pudp, pud) native_set_pud(pudp, pud)
68 #endif
69
70 #ifndef __PAGETABLE_PMD_FOLDED
71 #define pud_clear(pud) native_pud_clear(pud)
72 #endif
73
74 #define pte_clear(mm, addr, ptep) native_pte_clear(mm, addr, ptep)
75 #define pmd_clear(pmd) native_pmd_clear(pmd)
76
77 #define pte_update(mm, addr, ptep) do { } while (0)
78 #define pte_update_defer(mm, addr, ptep) do { } while (0)
79 #define pmd_update(mm, addr, ptep) do { } while (0)
80 #define pmd_update_defer(mm, addr, ptep) do { } while (0)
81
82 #define pgd_val(x) native_pgd_val(x)
83 #define __pgd(x) native_make_pgd(x)
84
85 #ifndef __PAGETABLE_PUD_FOLDED
86 #define pud_val(x) native_pud_val(x)
87 #define __pud(x) native_make_pud(x)
88 #endif
89
90 #ifndef __PAGETABLE_PMD_FOLDED
91 #define pmd_val(x) native_pmd_val(x)
92 #define __pmd(x) native_make_pmd(x)
93 #endif
94
95 #define pte_val(x) native_pte_val(x)
96 #define __pte(x) native_make_pte(x)
97
98 #define arch_end_context_switch(prev) do {} while(0)
99
100 #endif /* CONFIG_PARAVIRT */
101
102 /*
103 * The following only work if pte_present() is true.
104 * Undefined behaviour if not..
105 */
pte_dirty(pte_t pte)106 static inline int pte_dirty(pte_t pte)
107 {
108 return pte_flags(pte) & _PAGE_DIRTY;
109 }
110
pte_young(pte_t pte)111 static inline int pte_young(pte_t pte)
112 {
113 return pte_flags(pte) & _PAGE_ACCESSED;
114 }
115
pmd_dirty(pmd_t pmd)116 static inline int pmd_dirty(pmd_t pmd)
117 {
118 return pmd_flags(pmd) & _PAGE_DIRTY;
119 }
120
pmd_young(pmd_t pmd)121 static inline int pmd_young(pmd_t pmd)
122 {
123 return pmd_flags(pmd) & _PAGE_ACCESSED;
124 }
125
pte_write(pte_t pte)126 static inline int pte_write(pte_t pte)
127 {
128 return pte_flags(pte) & _PAGE_RW;
129 }
130
pte_huge(pte_t pte)131 static inline int pte_huge(pte_t pte)
132 {
133 return pte_flags(pte) & _PAGE_PSE;
134 }
135
pte_global(pte_t pte)136 static inline int pte_global(pte_t pte)
137 {
138 return pte_flags(pte) & _PAGE_GLOBAL;
139 }
140
pte_exec(pte_t pte)141 static inline int pte_exec(pte_t pte)
142 {
143 return !(pte_flags(pte) & _PAGE_NX);
144 }
145
pte_special(pte_t pte)146 static inline int pte_special(pte_t pte)
147 {
148 return pte_flags(pte) & _PAGE_SPECIAL;
149 }
150
151 /* Entries that were set to PROT_NONE are inverted */
152
153 static inline u64 protnone_mask(u64 val);
154
pte_pfn(pte_t pte)155 static inline unsigned long pte_pfn(pte_t pte)
156 {
157 phys_addr_t pfn = pte_val(pte);
158 pfn ^= protnone_mask(pfn);
159 return (pfn & PTE_PFN_MASK) >> PAGE_SHIFT;
160 }
161
pmd_pfn(pmd_t pmd)162 static inline unsigned long pmd_pfn(pmd_t pmd)
163 {
164 phys_addr_t pfn = pmd_val(pmd);
165 pfn ^= protnone_mask(pfn);
166 return (pfn & pmd_pfn_mask(pmd)) >> PAGE_SHIFT;
167 }
168
pud_pfn(pud_t pud)169 static inline unsigned long pud_pfn(pud_t pud)
170 {
171 phys_addr_t pfn = pud_val(pud);
172 pfn ^= protnone_mask(pfn);
173 return (pfn & pud_pfn_mask(pud)) >> PAGE_SHIFT;
174 }
175
pgd_pfn(pgd_t pgd)176 static inline unsigned long pgd_pfn(pgd_t pgd)
177 {
178 return (pgd_val(pgd) & PTE_PFN_MASK) >> PAGE_SHIFT;
179 }
180
181 #define pte_page(pte) pfn_to_page(pte_pfn(pte))
182
pmd_large(pmd_t pte)183 static inline int pmd_large(pmd_t pte)
184 {
185 return pmd_flags(pte) & _PAGE_PSE;
186 }
187
188 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
pmd_trans_splitting(pmd_t pmd)189 static inline int pmd_trans_splitting(pmd_t pmd)
190 {
191 return pmd_val(pmd) & _PAGE_SPLITTING;
192 }
193
pmd_trans_huge(pmd_t pmd)194 static inline int pmd_trans_huge(pmd_t pmd)
195 {
196 return pmd_val(pmd) & _PAGE_PSE;
197 }
198
has_transparent_hugepage(void)199 static inline int has_transparent_hugepage(void)
200 {
201 return cpu_has_pse;
202 }
203 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
204
pte_set_flags(pte_t pte,pteval_t set)205 static inline pte_t pte_set_flags(pte_t pte, pteval_t set)
206 {
207 pteval_t v = native_pte_val(pte);
208
209 return native_make_pte(v | set);
210 }
211
pte_clear_flags(pte_t pte,pteval_t clear)212 static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
213 {
214 pteval_t v = native_pte_val(pte);
215
216 return native_make_pte(v & ~clear);
217 }
218
pte_mkclean(pte_t pte)219 static inline pte_t pte_mkclean(pte_t pte)
220 {
221 return pte_clear_flags(pte, _PAGE_DIRTY);
222 }
223
pte_mkold(pte_t pte)224 static inline pte_t pte_mkold(pte_t pte)
225 {
226 return pte_clear_flags(pte, _PAGE_ACCESSED);
227 }
228
pte_wrprotect(pte_t pte)229 static inline pte_t pte_wrprotect(pte_t pte)
230 {
231 return pte_clear_flags(pte, _PAGE_RW);
232 }
233
pte_mkexec(pte_t pte)234 static inline pte_t pte_mkexec(pte_t pte)
235 {
236 return pte_clear_flags(pte, _PAGE_NX);
237 }
238
pte_mkdirty(pte_t pte)239 static inline pte_t pte_mkdirty(pte_t pte)
240 {
241 return pte_set_flags(pte, _PAGE_DIRTY | _PAGE_SOFT_DIRTY);
242 }
243
pte_mkyoung(pte_t pte)244 static inline pte_t pte_mkyoung(pte_t pte)
245 {
246 return pte_set_flags(pte, _PAGE_ACCESSED);
247 }
248
pte_mkwrite(pte_t pte)249 static inline pte_t pte_mkwrite(pte_t pte)
250 {
251 return pte_set_flags(pte, _PAGE_RW);
252 }
253
pte_mkhuge(pte_t pte)254 static inline pte_t pte_mkhuge(pte_t pte)
255 {
256 return pte_set_flags(pte, _PAGE_PSE);
257 }
258
pte_clrhuge(pte_t pte)259 static inline pte_t pte_clrhuge(pte_t pte)
260 {
261 return pte_clear_flags(pte, _PAGE_PSE);
262 }
263
pte_mkglobal(pte_t pte)264 static inline pte_t pte_mkglobal(pte_t pte)
265 {
266 return pte_set_flags(pte, _PAGE_GLOBAL);
267 }
268
pte_clrglobal(pte_t pte)269 static inline pte_t pte_clrglobal(pte_t pte)
270 {
271 return pte_clear_flags(pte, _PAGE_GLOBAL);
272 }
273
pte_mkspecial(pte_t pte)274 static inline pte_t pte_mkspecial(pte_t pte)
275 {
276 return pte_set_flags(pte, _PAGE_SPECIAL);
277 }
278
pmd_set_flags(pmd_t pmd,pmdval_t set)279 static inline pmd_t pmd_set_flags(pmd_t pmd, pmdval_t set)
280 {
281 pmdval_t v = native_pmd_val(pmd);
282
283 return __pmd(v | set);
284 }
285
pmd_clear_flags(pmd_t pmd,pmdval_t clear)286 static inline pmd_t pmd_clear_flags(pmd_t pmd, pmdval_t clear)
287 {
288 pmdval_t v = native_pmd_val(pmd);
289
290 return __pmd(v & ~clear);
291 }
292
pmd_mkold(pmd_t pmd)293 static inline pmd_t pmd_mkold(pmd_t pmd)
294 {
295 return pmd_clear_flags(pmd, _PAGE_ACCESSED);
296 }
297
pmd_wrprotect(pmd_t pmd)298 static inline pmd_t pmd_wrprotect(pmd_t pmd)
299 {
300 return pmd_clear_flags(pmd, _PAGE_RW);
301 }
302
pmd_mkdirty(pmd_t pmd)303 static inline pmd_t pmd_mkdirty(pmd_t pmd)
304 {
305 return pmd_set_flags(pmd, _PAGE_DIRTY | _PAGE_SOFT_DIRTY);
306 }
307
pmd_mkhuge(pmd_t pmd)308 static inline pmd_t pmd_mkhuge(pmd_t pmd)
309 {
310 return pmd_set_flags(pmd, _PAGE_PSE);
311 }
312
pmd_mkyoung(pmd_t pmd)313 static inline pmd_t pmd_mkyoung(pmd_t pmd)
314 {
315 return pmd_set_flags(pmd, _PAGE_ACCESSED);
316 }
317
pmd_mkwrite(pmd_t pmd)318 static inline pmd_t pmd_mkwrite(pmd_t pmd)
319 {
320 return pmd_set_flags(pmd, _PAGE_RW);
321 }
322
323 #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY
pte_soft_dirty(pte_t pte)324 static inline int pte_soft_dirty(pte_t pte)
325 {
326 return pte_flags(pte) & _PAGE_SOFT_DIRTY;
327 }
328
pmd_soft_dirty(pmd_t pmd)329 static inline int pmd_soft_dirty(pmd_t pmd)
330 {
331 return pmd_flags(pmd) & _PAGE_SOFT_DIRTY;
332 }
333
pte_mksoft_dirty(pte_t pte)334 static inline pte_t pte_mksoft_dirty(pte_t pte)
335 {
336 return pte_set_flags(pte, _PAGE_SOFT_DIRTY);
337 }
338
pmd_mksoft_dirty(pmd_t pmd)339 static inline pmd_t pmd_mksoft_dirty(pmd_t pmd)
340 {
341 return pmd_set_flags(pmd, _PAGE_SOFT_DIRTY);
342 }
343
pte_clear_soft_dirty(pte_t pte)344 static inline pte_t pte_clear_soft_dirty(pte_t pte)
345 {
346 return pte_clear_flags(pte, _PAGE_SOFT_DIRTY);
347 }
348
pmd_clear_soft_dirty(pmd_t pmd)349 static inline pmd_t pmd_clear_soft_dirty(pmd_t pmd)
350 {
351 return pmd_clear_flags(pmd, _PAGE_SOFT_DIRTY);
352 }
353
354 #endif /* CONFIG_HAVE_ARCH_SOFT_DIRTY */
355
356 /*
357 * Mask out unsupported bits in a present pgprot. Non-present pgprots
358 * can use those bits for other purposes, so leave them be.
359 */
massage_pgprot(pgprot_t pgprot)360 static inline pgprotval_t massage_pgprot(pgprot_t pgprot)
361 {
362 pgprotval_t protval = pgprot_val(pgprot);
363
364 if (protval & _PAGE_PRESENT)
365 protval &= __supported_pte_mask;
366
367 return protval;
368 }
369
pfn_pte(unsigned long page_nr,pgprot_t pgprot)370 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
371 {
372 phys_addr_t pfn = (phys_addr_t)page_nr << PAGE_SHIFT;
373 pfn ^= protnone_mask(pgprot_val(pgprot));
374 pfn &= PTE_PFN_MASK;
375 return __pte(pfn | massage_pgprot(pgprot));
376 }
377
pfn_pmd(unsigned long page_nr,pgprot_t pgprot)378 static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
379 {
380 phys_addr_t pfn = (phys_addr_t)page_nr << PAGE_SHIFT;
381 pfn ^= protnone_mask(pgprot_val(pgprot));
382 pfn &= PHYSICAL_PMD_PAGE_MASK;
383 return __pmd(pfn | massage_pgprot(pgprot));
384 }
385
pfn_pud(unsigned long page_nr,pgprot_t pgprot)386 static inline pud_t pfn_pud(unsigned long page_nr, pgprot_t pgprot)
387 {
388 phys_addr_t pfn = (phys_addr_t)page_nr << PAGE_SHIFT;
389 pfn ^= protnone_mask(pgprot_val(pgprot));
390 pfn &= PHYSICAL_PUD_PAGE_MASK;
391 return __pud(pfn | massage_pgprot(pgprot));
392 }
393
pmd_mknotpresent(pmd_t pmd)394 static inline pmd_t pmd_mknotpresent(pmd_t pmd)
395 {
396 return pfn_pmd(pmd_pfn(pmd),
397 __pgprot(pmd_flags(pmd) & ~(_PAGE_PRESENT|_PAGE_PROTNONE)));
398 }
399
pud_set_flags(pud_t pud,pudval_t set)400 static inline pud_t pud_set_flags(pud_t pud, pudval_t set)
401 {
402 pudval_t v = native_pud_val(pud);
403
404 return __pud(v | set);
405 }
406
pud_clear_flags(pud_t pud,pudval_t clear)407 static inline pud_t pud_clear_flags(pud_t pud, pudval_t clear)
408 {
409 pudval_t v = native_pud_val(pud);
410
411 return __pud(v & ~clear);
412 }
413
pud_mkhuge(pud_t pud)414 static inline pud_t pud_mkhuge(pud_t pud)
415 {
416 return pud_set_flags(pud, _PAGE_PSE);
417 }
418
419 static inline u64 flip_protnone_guard(u64 oldval, u64 val, u64 mask);
420
pte_modify(pte_t pte,pgprot_t newprot)421 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
422 {
423 pteval_t val = pte_val(pte), oldval = val;
424
425 /*
426 * Chop off the NX bit (if present), and add the NX portion of
427 * the newprot (if present):
428 */
429 val &= _PAGE_CHG_MASK;
430 val |= massage_pgprot(newprot) & ~_PAGE_CHG_MASK;
431 val = flip_protnone_guard(oldval, val, PTE_PFN_MASK);
432 return __pte(val);
433 }
434
pmd_modify(pmd_t pmd,pgprot_t newprot)435 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
436 {
437 pmdval_t val = pmd_val(pmd), oldval = val;
438
439 val &= _HPAGE_CHG_MASK;
440 val |= massage_pgprot(newprot) & ~_HPAGE_CHG_MASK;
441 val = flip_protnone_guard(oldval, val, PHYSICAL_PMD_PAGE_MASK);
442 return __pmd(val);
443 }
444
445 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
446 #define pgprot_modify pgprot_modify
pgprot_modify(pgprot_t oldprot,pgprot_t newprot)447 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
448 {
449 pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
450 pgprotval_t addbits = pgprot_val(newprot);
451 return __pgprot(preservebits | addbits);
452 }
453
454 #define pte_pgprot(x) __pgprot(pte_flags(x))
455 #define pmd_pgprot(x) __pgprot(pmd_flags(x))
456 #define pud_pgprot(x) __pgprot(pud_flags(x))
457
458 #define canon_pgprot(p) __pgprot(massage_pgprot(p))
459
is_new_memtype_allowed(u64 paddr,unsigned long size,enum page_cache_mode pcm,enum page_cache_mode new_pcm)460 static inline int is_new_memtype_allowed(u64 paddr, unsigned long size,
461 enum page_cache_mode pcm,
462 enum page_cache_mode new_pcm)
463 {
464 /*
465 * PAT type is always WB for untracked ranges, so no need to check.
466 */
467 if (x86_platform.is_untracked_pat_range(paddr, paddr + size))
468 return 1;
469
470 /*
471 * Certain new memtypes are not allowed with certain
472 * requested memtype:
473 * - request is uncached, return cannot be write-back
474 * - request is write-combine, return cannot be write-back
475 * - request is write-through, return cannot be write-back
476 * - request is write-through, return cannot be write-combine
477 */
478 if ((pcm == _PAGE_CACHE_MODE_UC_MINUS &&
479 new_pcm == _PAGE_CACHE_MODE_WB) ||
480 (pcm == _PAGE_CACHE_MODE_WC &&
481 new_pcm == _PAGE_CACHE_MODE_WB) ||
482 (pcm == _PAGE_CACHE_MODE_WT &&
483 new_pcm == _PAGE_CACHE_MODE_WB) ||
484 (pcm == _PAGE_CACHE_MODE_WT &&
485 new_pcm == _PAGE_CACHE_MODE_WC)) {
486 return 0;
487 }
488
489 return 1;
490 }
491
492 pmd_t *populate_extra_pmd(unsigned long vaddr);
493 pte_t *populate_extra_pte(unsigned long vaddr);
494 #endif /* __ASSEMBLY__ */
495
496 #ifdef CONFIG_X86_32
497 # include <asm/pgtable_32.h>
498 #else
499 # include <asm/pgtable_64.h>
500 #endif
501
502 #ifndef __ASSEMBLY__
503 #include <linux/mm_types.h>
504 #include <linux/mmdebug.h>
505 #include <linux/log2.h>
506
pte_none(pte_t pte)507 static inline int pte_none(pte_t pte)
508 {
509 return !pte.pte;
510 }
511
512 #define __HAVE_ARCH_PTE_SAME
pte_same(pte_t a,pte_t b)513 static inline int pte_same(pte_t a, pte_t b)
514 {
515 return a.pte == b.pte;
516 }
517
pte_present(pte_t a)518 static inline int pte_present(pte_t a)
519 {
520 return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
521 }
522
523 #define pte_accessible pte_accessible
pte_accessible(struct mm_struct * mm,pte_t a)524 static inline bool pte_accessible(struct mm_struct *mm, pte_t a)
525 {
526 if (pte_flags(a) & _PAGE_PRESENT)
527 return true;
528
529 if ((pte_flags(a) & _PAGE_PROTNONE) &&
530 mm_tlb_flush_pending(mm))
531 return true;
532
533 return false;
534 }
535
pte_hidden(pte_t pte)536 static inline int pte_hidden(pte_t pte)
537 {
538 return pte_flags(pte) & _PAGE_HIDDEN;
539 }
540
pmd_present(pmd_t pmd)541 static inline int pmd_present(pmd_t pmd)
542 {
543 /*
544 * Checking for _PAGE_PSE is needed too because
545 * split_huge_page will temporarily clear the present bit (but
546 * the _PAGE_PSE flag will remain set at all times while the
547 * _PAGE_PRESENT bit is clear).
548 */
549 return pmd_flags(pmd) & (_PAGE_PRESENT | _PAGE_PROTNONE | _PAGE_PSE);
550 }
551
552 #ifdef CONFIG_NUMA_BALANCING
553 /*
554 * These work without NUMA balancing but the kernel does not care. See the
555 * comment in include/asm-generic/pgtable.h
556 */
pte_protnone(pte_t pte)557 static inline int pte_protnone(pte_t pte)
558 {
559 return (pte_flags(pte) & (_PAGE_PROTNONE | _PAGE_PRESENT))
560 == _PAGE_PROTNONE;
561 }
562
pmd_protnone(pmd_t pmd)563 static inline int pmd_protnone(pmd_t pmd)
564 {
565 return (pmd_flags(pmd) & (_PAGE_PROTNONE | _PAGE_PRESENT))
566 == _PAGE_PROTNONE;
567 }
568 #endif /* CONFIG_NUMA_BALANCING */
569
pmd_none(pmd_t pmd)570 static inline int pmd_none(pmd_t pmd)
571 {
572 /* Only check low word on 32-bit platforms, since it might be
573 out of sync with upper half. */
574 return (unsigned long)native_pmd_val(pmd) == 0;
575 }
576
pmd_page_vaddr(pmd_t pmd)577 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
578 {
579 return (unsigned long)__va(pmd_val(pmd) & pmd_pfn_mask(pmd));
580 }
581
582 /*
583 * Currently stuck as a macro due to indirect forward reference to
584 * linux/mmzone.h's __section_mem_map_addr() definition:
585 */
586 #define pmd_page(pmd) pfn_to_page(pmd_pfn(pmd))
587
588 /*
589 * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
590 *
591 * this macro returns the index of the entry in the pmd page which would
592 * control the given virtual address
593 */
pmd_index(unsigned long address)594 static inline unsigned long pmd_index(unsigned long address)
595 {
596 return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
597 }
598
599 /*
600 * Conversion functions: convert a page and protection to a page entry,
601 * and a page entry and page directory to the page they refer to.
602 *
603 * (Currently stuck as a macro because of indirect forward reference
604 * to linux/mm.h:page_to_nid())
605 */
606 #define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
607
608 /*
609 * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
610 *
611 * this function returns the index of the entry in the pte page which would
612 * control the given virtual address
613 */
pte_index(unsigned long address)614 static inline unsigned long pte_index(unsigned long address)
615 {
616 return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
617 }
618
pte_offset_kernel(pmd_t * pmd,unsigned long address)619 static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
620 {
621 return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
622 }
623
pmd_bad(pmd_t pmd)624 static inline int pmd_bad(pmd_t pmd)
625 {
626 return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
627 }
628
pages_to_mb(unsigned long npg)629 static inline unsigned long pages_to_mb(unsigned long npg)
630 {
631 return npg >> (20 - PAGE_SHIFT);
632 }
633
634 #if CONFIG_PGTABLE_LEVELS > 2
pud_none(pud_t pud)635 static inline int pud_none(pud_t pud)
636 {
637 return native_pud_val(pud) == 0;
638 }
639
pud_present(pud_t pud)640 static inline int pud_present(pud_t pud)
641 {
642 return pud_flags(pud) & _PAGE_PRESENT;
643 }
644
pud_page_vaddr(pud_t pud)645 static inline unsigned long pud_page_vaddr(pud_t pud)
646 {
647 return (unsigned long)__va(pud_val(pud) & pud_pfn_mask(pud));
648 }
649
650 /*
651 * Currently stuck as a macro due to indirect forward reference to
652 * linux/mmzone.h's __section_mem_map_addr() definition:
653 */
654 #define pud_page(pud) pfn_to_page(pud_pfn(pud))
655
656 /* Find an entry in the second-level page table.. */
pmd_offset(pud_t * pud,unsigned long address)657 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
658 {
659 return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
660 }
661
pud_large(pud_t pud)662 static inline int pud_large(pud_t pud)
663 {
664 return (pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
665 (_PAGE_PSE | _PAGE_PRESENT);
666 }
667
pud_bad(pud_t pud)668 static inline int pud_bad(pud_t pud)
669 {
670 return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
671 }
672 #else
pud_large(pud_t pud)673 static inline int pud_large(pud_t pud)
674 {
675 return 0;
676 }
677 #endif /* CONFIG_PGTABLE_LEVELS > 2 */
678
679 #if CONFIG_PGTABLE_LEVELS > 3
pgd_present(pgd_t pgd)680 static inline int pgd_present(pgd_t pgd)
681 {
682 return pgd_flags(pgd) & _PAGE_PRESENT;
683 }
684
pgd_page_vaddr(pgd_t pgd)685 static inline unsigned long pgd_page_vaddr(pgd_t pgd)
686 {
687 return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
688 }
689
690 /*
691 * Currently stuck as a macro due to indirect forward reference to
692 * linux/mmzone.h's __section_mem_map_addr() definition:
693 */
694 #define pgd_page(pgd) pfn_to_page(pgd_pfn(pgd))
695
696 /* to find an entry in a page-table-directory. */
pud_index(unsigned long address)697 static inline unsigned long pud_index(unsigned long address)
698 {
699 return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
700 }
701
pud_offset(pgd_t * pgd,unsigned long address)702 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
703 {
704 return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
705 }
706
pgd_bad(pgd_t pgd)707 static inline int pgd_bad(pgd_t pgd)
708 {
709 pgdval_t ignore_flags = _PAGE_USER;
710 /*
711 * We set NX on KAISER pgds that map userspace memory so
712 * that userspace can not meaningfully use the kernel
713 * page table by accident; it will fault on the first
714 * instruction it tries to run. See native_set_pgd().
715 */
716 if (kaiser_enabled)
717 ignore_flags |= _PAGE_NX;
718
719 return (pgd_flags(pgd) & ~ignore_flags) != _KERNPG_TABLE;
720 }
721
pgd_none(pgd_t pgd)722 static inline int pgd_none(pgd_t pgd)
723 {
724 return !native_pgd_val(pgd);
725 }
726 #endif /* CONFIG_PGTABLE_LEVELS > 3 */
727
728 #endif /* __ASSEMBLY__ */
729
730 /*
731 * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
732 *
733 * this macro returns the index of the entry in the pgd page which would
734 * control the given virtual address
735 */
736 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
737
738 /*
739 * pgd_offset() returns a (pgd_t *)
740 * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
741 */
742 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
743 /*
744 * a shortcut which implies the use of the kernel's pgd, instead
745 * of a process's
746 */
747 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
748
749
750 #define KERNEL_PGD_BOUNDARY pgd_index(PAGE_OFFSET)
751 #define KERNEL_PGD_PTRS (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
752
753 #ifndef __ASSEMBLY__
754
755 extern int direct_gbpages;
756 void init_mem_mapping(void);
757 void early_alloc_pgt_buf(void);
758
759 /* local pte updates need not use xchg for locking */
native_local_ptep_get_and_clear(pte_t * ptep)760 static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
761 {
762 pte_t res = *ptep;
763
764 /* Pure native function needs no input for mm, addr */
765 native_pte_clear(NULL, 0, ptep);
766 return res;
767 }
768
native_local_pmdp_get_and_clear(pmd_t * pmdp)769 static inline pmd_t native_local_pmdp_get_and_clear(pmd_t *pmdp)
770 {
771 pmd_t res = *pmdp;
772
773 native_pmd_clear(pmdp);
774 return res;
775 }
776
native_set_pte_at(struct mm_struct * mm,unsigned long addr,pte_t * ptep,pte_t pte)777 static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
778 pte_t *ptep , pte_t pte)
779 {
780 native_set_pte(ptep, pte);
781 }
782
native_set_pmd_at(struct mm_struct * mm,unsigned long addr,pmd_t * pmdp,pmd_t pmd)783 static inline void native_set_pmd_at(struct mm_struct *mm, unsigned long addr,
784 pmd_t *pmdp , pmd_t pmd)
785 {
786 native_set_pmd(pmdp, pmd);
787 }
788
789 #ifndef CONFIG_PARAVIRT
790 /*
791 * Rules for using pte_update - it must be called after any PTE update which
792 * has not been done using the set_pte / clear_pte interfaces. It is used by
793 * shadow mode hypervisors to resynchronize the shadow page tables. Kernel PTE
794 * updates should either be sets, clears, or set_pte_atomic for P->P
795 * transitions, which means this hook should only be called for user PTEs.
796 * This hook implies a P->P protection or access change has taken place, which
797 * requires a subsequent TLB flush. The notification can optionally be delayed
798 * until the TLB flush event by using the pte_update_defer form of the
799 * interface, but care must be taken to assure that the flush happens while
800 * still holding the same page table lock so that the shadow and primary pages
801 * do not become out of sync on SMP.
802 */
803 #define pte_update(mm, addr, ptep) do { } while (0)
804 #define pte_update_defer(mm, addr, ptep) do { } while (0)
805 #endif
806
807 /*
808 * We only update the dirty/accessed state if we set
809 * the dirty bit by hand in the kernel, since the hardware
810 * will do the accessed bit for us, and we don't want to
811 * race with other CPU's that might be updating the dirty
812 * bit at the same time.
813 */
814 struct vm_area_struct;
815
816 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
817 extern int ptep_set_access_flags(struct vm_area_struct *vma,
818 unsigned long address, pte_t *ptep,
819 pte_t entry, int dirty);
820
821 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
822 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
823 unsigned long addr, pte_t *ptep);
824
825 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
826 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
827 unsigned long address, pte_t *ptep);
828
829 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
ptep_get_and_clear(struct mm_struct * mm,unsigned long addr,pte_t * ptep)830 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
831 pte_t *ptep)
832 {
833 pte_t pte = native_ptep_get_and_clear(ptep);
834 pte_update(mm, addr, ptep);
835 return pte;
836 }
837
838 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
ptep_get_and_clear_full(struct mm_struct * mm,unsigned long addr,pte_t * ptep,int full)839 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
840 unsigned long addr, pte_t *ptep,
841 int full)
842 {
843 pte_t pte;
844 if (full) {
845 /*
846 * Full address destruction in progress; paravirt does not
847 * care about updates and native needs no locking
848 */
849 pte = native_local_ptep_get_and_clear(ptep);
850 } else {
851 pte = ptep_get_and_clear(mm, addr, ptep);
852 }
853 return pte;
854 }
855
856 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
ptep_set_wrprotect(struct mm_struct * mm,unsigned long addr,pte_t * ptep)857 static inline void ptep_set_wrprotect(struct mm_struct *mm,
858 unsigned long addr, pte_t *ptep)
859 {
860 clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
861 pte_update(mm, addr, ptep);
862 }
863
864 #define flush_tlb_fix_spurious_fault(vma, address) do { } while (0)
865
866 #define mk_pmd(page, pgprot) pfn_pmd(page_to_pfn(page), (pgprot))
867
868 #define __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
869 extern int pmdp_set_access_flags(struct vm_area_struct *vma,
870 unsigned long address, pmd_t *pmdp,
871 pmd_t entry, int dirty);
872
873 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
874 extern int pmdp_test_and_clear_young(struct vm_area_struct *vma,
875 unsigned long addr, pmd_t *pmdp);
876
877 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
878 extern int pmdp_clear_flush_young(struct vm_area_struct *vma,
879 unsigned long address, pmd_t *pmdp);
880
881
882 #define __HAVE_ARCH_PMDP_SPLITTING_FLUSH
883 extern void pmdp_splitting_flush(struct vm_area_struct *vma,
884 unsigned long addr, pmd_t *pmdp);
885
886 #define __HAVE_ARCH_PMD_WRITE
pmd_write(pmd_t pmd)887 static inline int pmd_write(pmd_t pmd)
888 {
889 return pmd_flags(pmd) & _PAGE_RW;
890 }
891
892 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
pmdp_huge_get_and_clear(struct mm_struct * mm,unsigned long addr,pmd_t * pmdp)893 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm, unsigned long addr,
894 pmd_t *pmdp)
895 {
896 pmd_t pmd = native_pmdp_get_and_clear(pmdp);
897 pmd_update(mm, addr, pmdp);
898 return pmd;
899 }
900
901 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
pmdp_set_wrprotect(struct mm_struct * mm,unsigned long addr,pmd_t * pmdp)902 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
903 unsigned long addr, pmd_t *pmdp)
904 {
905 clear_bit(_PAGE_BIT_RW, (unsigned long *)pmdp);
906 pmd_update(mm, addr, pmdp);
907 }
908
909 /*
910 * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
911 *
912 * dst - pointer to pgd range anwhere on a pgd page
913 * src - ""
914 * count - the number of pgds to copy.
915 *
916 * dst and src can be on the same page, but the range must not overlap,
917 * and must not cross a page boundary.
918 */
clone_pgd_range(pgd_t * dst,pgd_t * src,int count)919 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
920 {
921 memcpy(dst, src, count * sizeof(pgd_t));
922 #ifdef CONFIG_PAGE_TABLE_ISOLATION
923 if (kaiser_enabled) {
924 /* Clone the shadow pgd part as well */
925 memcpy(native_get_shadow_pgd(dst),
926 native_get_shadow_pgd(src),
927 count * sizeof(pgd_t));
928 }
929 #endif
930 }
931
932 #define PTE_SHIFT ilog2(PTRS_PER_PTE)
page_level_shift(enum pg_level level)933 static inline int page_level_shift(enum pg_level level)
934 {
935 return (PAGE_SHIFT - PTE_SHIFT) + level * PTE_SHIFT;
936 }
page_level_size(enum pg_level level)937 static inline unsigned long page_level_size(enum pg_level level)
938 {
939 return 1UL << page_level_shift(level);
940 }
page_level_mask(enum pg_level level)941 static inline unsigned long page_level_mask(enum pg_level level)
942 {
943 return ~(page_level_size(level) - 1);
944 }
945
946 /*
947 * The x86 doesn't have any external MMU info: the kernel page
948 * tables contain all the necessary information.
949 */
update_mmu_cache(struct vm_area_struct * vma,unsigned long addr,pte_t * ptep)950 static inline void update_mmu_cache(struct vm_area_struct *vma,
951 unsigned long addr, pte_t *ptep)
952 {
953 }
update_mmu_cache_pmd(struct vm_area_struct * vma,unsigned long addr,pmd_t * pmd)954 static inline void update_mmu_cache_pmd(struct vm_area_struct *vma,
955 unsigned long addr, pmd_t *pmd)
956 {
957 }
958
959 #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY
pte_swp_mksoft_dirty(pte_t pte)960 static inline pte_t pte_swp_mksoft_dirty(pte_t pte)
961 {
962 return pte_set_flags(pte, _PAGE_SWP_SOFT_DIRTY);
963 }
964
pte_swp_soft_dirty(pte_t pte)965 static inline int pte_swp_soft_dirty(pte_t pte)
966 {
967 return pte_flags(pte) & _PAGE_SWP_SOFT_DIRTY;
968 }
969
pte_swp_clear_soft_dirty(pte_t pte)970 static inline pte_t pte_swp_clear_soft_dirty(pte_t pte)
971 {
972 return pte_clear_flags(pte, _PAGE_SWP_SOFT_DIRTY);
973 }
974 #endif
975
976 #define __HAVE_ARCH_PFN_MODIFY_ALLOWED 1
977 extern bool pfn_modify_allowed(unsigned long pfn, pgprot_t prot);
978
arch_has_pfn_modify_check(void)979 static inline bool arch_has_pfn_modify_check(void)
980 {
981 return boot_cpu_has_bug(X86_BUG_L1TF);
982 }
983
984 #include <asm-generic/pgtable.h>
985 #endif /* __ASSEMBLY__ */
986
987 #endif /* _ASM_X86_PGTABLE_H */
988