xref: /openbmc/linux/arch/x86/include/asm/pgtable.h (revision b0f85fa11aefc4f3e03306b4cd47f113bd57dcba)
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 void ptdump_walk_pgd_level(struct seq_file *m, pgd_t *pgd);
22 void ptdump_walk_pgd_level_checkwx(void);
23 
24 #ifdef CONFIG_DEBUG_WX
25 #define debug_checkwx() ptdump_walk_pgd_level_checkwx()
26 #else
27 #define debug_checkwx() do { } while (0)
28 #endif
29 
30 /*
31  * ZERO_PAGE is a global shared page that is always zero: used
32  * for zero-mapped memory areas etc..
33  */
34 extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)]
35 	__visible;
36 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
37 
38 extern spinlock_t pgd_lock;
39 extern struct list_head pgd_list;
40 
41 extern struct mm_struct *pgd_page_get_mm(struct page *page);
42 
43 #ifdef CONFIG_PARAVIRT
44 #include <asm/paravirt.h>
45 #else  /* !CONFIG_PARAVIRT */
46 #define set_pte(ptep, pte)		native_set_pte(ptep, pte)
47 #define set_pte_at(mm, addr, ptep, pte)	native_set_pte_at(mm, addr, ptep, pte)
48 #define set_pmd_at(mm, addr, pmdp, pmd)	native_set_pmd_at(mm, addr, pmdp, pmd)
49 
50 #define set_pte_atomic(ptep, pte)					\
51 	native_set_pte_atomic(ptep, pte)
52 
53 #define set_pmd(pmdp, pmd)		native_set_pmd(pmdp, pmd)
54 
55 #ifndef __PAGETABLE_PUD_FOLDED
56 #define set_pgd(pgdp, pgd)		native_set_pgd(pgdp, pgd)
57 #define pgd_clear(pgd)			native_pgd_clear(pgd)
58 #endif
59 
60 #ifndef set_pud
61 # define set_pud(pudp, pud)		native_set_pud(pudp, pud)
62 #endif
63 
64 #ifndef __PAGETABLE_PMD_FOLDED
65 #define pud_clear(pud)			native_pud_clear(pud)
66 #endif
67 
68 #define pte_clear(mm, addr, ptep)	native_pte_clear(mm, addr, ptep)
69 #define pmd_clear(pmd)			native_pmd_clear(pmd)
70 
71 #define pte_update(mm, addr, ptep)              do { } while (0)
72 #define pte_update_defer(mm, addr, ptep)        do { } while (0)
73 #define pmd_update(mm, addr, ptep)              do { } while (0)
74 #define pmd_update_defer(mm, addr, ptep)        do { } while (0)
75 
76 #define pgd_val(x)	native_pgd_val(x)
77 #define __pgd(x)	native_make_pgd(x)
78 
79 #ifndef __PAGETABLE_PUD_FOLDED
80 #define pud_val(x)	native_pud_val(x)
81 #define __pud(x)	native_make_pud(x)
82 #endif
83 
84 #ifndef __PAGETABLE_PMD_FOLDED
85 #define pmd_val(x)	native_pmd_val(x)
86 #define __pmd(x)	native_make_pmd(x)
87 #endif
88 
89 #define pte_val(x)	native_pte_val(x)
90 #define __pte(x)	native_make_pte(x)
91 
92 #define arch_end_context_switch(prev)	do {} while(0)
93 
94 #endif	/* CONFIG_PARAVIRT */
95 
96 /*
97  * The following only work if pte_present() is true.
98  * Undefined behaviour if not..
99  */
100 static inline int pte_dirty(pte_t pte)
101 {
102 	return pte_flags(pte) & _PAGE_DIRTY;
103 }
104 
105 static inline int pte_young(pte_t pte)
106 {
107 	return pte_flags(pte) & _PAGE_ACCESSED;
108 }
109 
110 static inline int pmd_dirty(pmd_t pmd)
111 {
112 	return pmd_flags(pmd) & _PAGE_DIRTY;
113 }
114 
115 static inline int pmd_young(pmd_t pmd)
116 {
117 	return pmd_flags(pmd) & _PAGE_ACCESSED;
118 }
119 
120 static inline int pte_write(pte_t pte)
121 {
122 	return pte_flags(pte) & _PAGE_RW;
123 }
124 
125 static inline int pte_huge(pte_t pte)
126 {
127 	return pte_flags(pte) & _PAGE_PSE;
128 }
129 
130 static inline int pte_global(pte_t pte)
131 {
132 	return pte_flags(pte) & _PAGE_GLOBAL;
133 }
134 
135 static inline int pte_exec(pte_t pte)
136 {
137 	return !(pte_flags(pte) & _PAGE_NX);
138 }
139 
140 static inline int pte_special(pte_t pte)
141 {
142 	return pte_flags(pte) & _PAGE_SPECIAL;
143 }
144 
145 static inline unsigned long pte_pfn(pte_t pte)
146 {
147 	return (pte_val(pte) & PTE_PFN_MASK) >> PAGE_SHIFT;
148 }
149 
150 static inline unsigned long pmd_pfn(pmd_t pmd)
151 {
152 	return (pmd_val(pmd) & pmd_pfn_mask(pmd)) >> PAGE_SHIFT;
153 }
154 
155 static inline unsigned long pud_pfn(pud_t pud)
156 {
157 	return (pud_val(pud) & pud_pfn_mask(pud)) >> PAGE_SHIFT;
158 }
159 
160 #define pte_page(pte)	pfn_to_page(pte_pfn(pte))
161 
162 static inline int pmd_large(pmd_t pte)
163 {
164 	return pmd_flags(pte) & _PAGE_PSE;
165 }
166 
167 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
168 static inline int pmd_trans_splitting(pmd_t pmd)
169 {
170 	return pmd_val(pmd) & _PAGE_SPLITTING;
171 }
172 
173 static inline int pmd_trans_huge(pmd_t pmd)
174 {
175 	return pmd_val(pmd) & _PAGE_PSE;
176 }
177 
178 static inline int has_transparent_hugepage(void)
179 {
180 	return cpu_has_pse;
181 }
182 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
183 
184 static inline pte_t pte_set_flags(pte_t pte, pteval_t set)
185 {
186 	pteval_t v = native_pte_val(pte);
187 
188 	return native_make_pte(v | set);
189 }
190 
191 static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
192 {
193 	pteval_t v = native_pte_val(pte);
194 
195 	return native_make_pte(v & ~clear);
196 }
197 
198 static inline pte_t pte_mkclean(pte_t pte)
199 {
200 	return pte_clear_flags(pte, _PAGE_DIRTY);
201 }
202 
203 static inline pte_t pte_mkold(pte_t pte)
204 {
205 	return pte_clear_flags(pte, _PAGE_ACCESSED);
206 }
207 
208 static inline pte_t pte_wrprotect(pte_t pte)
209 {
210 	return pte_clear_flags(pte, _PAGE_RW);
211 }
212 
213 static inline pte_t pte_mkexec(pte_t pte)
214 {
215 	return pte_clear_flags(pte, _PAGE_NX);
216 }
217 
218 static inline pte_t pte_mkdirty(pte_t pte)
219 {
220 	return pte_set_flags(pte, _PAGE_DIRTY | _PAGE_SOFT_DIRTY);
221 }
222 
223 static inline pte_t pte_mkyoung(pte_t pte)
224 {
225 	return pte_set_flags(pte, _PAGE_ACCESSED);
226 }
227 
228 static inline pte_t pte_mkwrite(pte_t pte)
229 {
230 	return pte_set_flags(pte, _PAGE_RW);
231 }
232 
233 static inline pte_t pte_mkhuge(pte_t pte)
234 {
235 	return pte_set_flags(pte, _PAGE_PSE);
236 }
237 
238 static inline pte_t pte_clrhuge(pte_t pte)
239 {
240 	return pte_clear_flags(pte, _PAGE_PSE);
241 }
242 
243 static inline pte_t pte_mkglobal(pte_t pte)
244 {
245 	return pte_set_flags(pte, _PAGE_GLOBAL);
246 }
247 
248 static inline pte_t pte_clrglobal(pte_t pte)
249 {
250 	return pte_clear_flags(pte, _PAGE_GLOBAL);
251 }
252 
253 static inline pte_t pte_mkspecial(pte_t pte)
254 {
255 	return pte_set_flags(pte, _PAGE_SPECIAL);
256 }
257 
258 static inline pmd_t pmd_set_flags(pmd_t pmd, pmdval_t set)
259 {
260 	pmdval_t v = native_pmd_val(pmd);
261 
262 	return __pmd(v | set);
263 }
264 
265 static inline pmd_t pmd_clear_flags(pmd_t pmd, pmdval_t clear)
266 {
267 	pmdval_t v = native_pmd_val(pmd);
268 
269 	return __pmd(v & ~clear);
270 }
271 
272 static inline pmd_t pmd_mkold(pmd_t pmd)
273 {
274 	return pmd_clear_flags(pmd, _PAGE_ACCESSED);
275 }
276 
277 static inline pmd_t pmd_wrprotect(pmd_t pmd)
278 {
279 	return pmd_clear_flags(pmd, _PAGE_RW);
280 }
281 
282 static inline pmd_t pmd_mkdirty(pmd_t pmd)
283 {
284 	return pmd_set_flags(pmd, _PAGE_DIRTY | _PAGE_SOFT_DIRTY);
285 }
286 
287 static inline pmd_t pmd_mkhuge(pmd_t pmd)
288 {
289 	return pmd_set_flags(pmd, _PAGE_PSE);
290 }
291 
292 static inline pmd_t pmd_mkyoung(pmd_t pmd)
293 {
294 	return pmd_set_flags(pmd, _PAGE_ACCESSED);
295 }
296 
297 static inline pmd_t pmd_mkwrite(pmd_t pmd)
298 {
299 	return pmd_set_flags(pmd, _PAGE_RW);
300 }
301 
302 static inline pmd_t pmd_mknotpresent(pmd_t pmd)
303 {
304 	return pmd_clear_flags(pmd, _PAGE_PRESENT | _PAGE_PROTNONE);
305 }
306 
307 #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY
308 static inline int pte_soft_dirty(pte_t pte)
309 {
310 	return pte_flags(pte) & _PAGE_SOFT_DIRTY;
311 }
312 
313 static inline int pmd_soft_dirty(pmd_t pmd)
314 {
315 	return pmd_flags(pmd) & _PAGE_SOFT_DIRTY;
316 }
317 
318 static inline pte_t pte_mksoft_dirty(pte_t pte)
319 {
320 	return pte_set_flags(pte, _PAGE_SOFT_DIRTY);
321 }
322 
323 static inline pmd_t pmd_mksoft_dirty(pmd_t pmd)
324 {
325 	return pmd_set_flags(pmd, _PAGE_SOFT_DIRTY);
326 }
327 
328 #endif /* CONFIG_HAVE_ARCH_SOFT_DIRTY */
329 
330 /*
331  * Mask out unsupported bits in a present pgprot.  Non-present pgprots
332  * can use those bits for other purposes, so leave them be.
333  */
334 static inline pgprotval_t massage_pgprot(pgprot_t pgprot)
335 {
336 	pgprotval_t protval = pgprot_val(pgprot);
337 
338 	if (protval & _PAGE_PRESENT)
339 		protval &= __supported_pte_mask;
340 
341 	return protval;
342 }
343 
344 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
345 {
346 	return __pte(((phys_addr_t)page_nr << PAGE_SHIFT) |
347 		     massage_pgprot(pgprot));
348 }
349 
350 static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
351 {
352 	return __pmd(((phys_addr_t)page_nr << PAGE_SHIFT) |
353 		     massage_pgprot(pgprot));
354 }
355 
356 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
357 {
358 	pteval_t val = pte_val(pte);
359 
360 	/*
361 	 * Chop off the NX bit (if present), and add the NX portion of
362 	 * the newprot (if present):
363 	 */
364 	val &= _PAGE_CHG_MASK;
365 	val |= massage_pgprot(newprot) & ~_PAGE_CHG_MASK;
366 
367 	return __pte(val);
368 }
369 
370 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
371 {
372 	pmdval_t val = pmd_val(pmd);
373 
374 	val &= _HPAGE_CHG_MASK;
375 	val |= massage_pgprot(newprot) & ~_HPAGE_CHG_MASK;
376 
377 	return __pmd(val);
378 }
379 
380 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
381 #define pgprot_modify pgprot_modify
382 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
383 {
384 	pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
385 	pgprotval_t addbits = pgprot_val(newprot);
386 	return __pgprot(preservebits | addbits);
387 }
388 
389 #define pte_pgprot(x) __pgprot(pte_flags(x))
390 #define pmd_pgprot(x) __pgprot(pmd_flags(x))
391 #define pud_pgprot(x) __pgprot(pud_flags(x))
392 
393 #define canon_pgprot(p) __pgprot(massage_pgprot(p))
394 
395 static inline int is_new_memtype_allowed(u64 paddr, unsigned long size,
396 					 enum page_cache_mode pcm,
397 					 enum page_cache_mode new_pcm)
398 {
399 	/*
400 	 * PAT type is always WB for untracked ranges, so no need to check.
401 	 */
402 	if (x86_platform.is_untracked_pat_range(paddr, paddr + size))
403 		return 1;
404 
405 	/*
406 	 * Certain new memtypes are not allowed with certain
407 	 * requested memtype:
408 	 * - request is uncached, return cannot be write-back
409 	 * - request is write-combine, return cannot be write-back
410 	 * - request is write-through, return cannot be write-back
411 	 * - request is write-through, return cannot be write-combine
412 	 */
413 	if ((pcm == _PAGE_CACHE_MODE_UC_MINUS &&
414 	     new_pcm == _PAGE_CACHE_MODE_WB) ||
415 	    (pcm == _PAGE_CACHE_MODE_WC &&
416 	     new_pcm == _PAGE_CACHE_MODE_WB) ||
417 	    (pcm == _PAGE_CACHE_MODE_WT &&
418 	     new_pcm == _PAGE_CACHE_MODE_WB) ||
419 	    (pcm == _PAGE_CACHE_MODE_WT &&
420 	     new_pcm == _PAGE_CACHE_MODE_WC)) {
421 		return 0;
422 	}
423 
424 	return 1;
425 }
426 
427 pmd_t *populate_extra_pmd(unsigned long vaddr);
428 pte_t *populate_extra_pte(unsigned long vaddr);
429 #endif	/* __ASSEMBLY__ */
430 
431 #ifdef CONFIG_X86_32
432 # include <asm/pgtable_32.h>
433 #else
434 # include <asm/pgtable_64.h>
435 #endif
436 
437 #ifndef __ASSEMBLY__
438 #include <linux/mm_types.h>
439 #include <linux/mmdebug.h>
440 #include <linux/log2.h>
441 
442 static inline int pte_none(pte_t pte)
443 {
444 	return !pte.pte;
445 }
446 
447 #define __HAVE_ARCH_PTE_SAME
448 static inline int pte_same(pte_t a, pte_t b)
449 {
450 	return a.pte == b.pte;
451 }
452 
453 static inline int pte_present(pte_t a)
454 {
455 	return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
456 }
457 
458 #define pte_accessible pte_accessible
459 static inline bool pte_accessible(struct mm_struct *mm, pte_t a)
460 {
461 	if (pte_flags(a) & _PAGE_PRESENT)
462 		return true;
463 
464 	if ((pte_flags(a) & _PAGE_PROTNONE) &&
465 			mm_tlb_flush_pending(mm))
466 		return true;
467 
468 	return false;
469 }
470 
471 static inline int pte_hidden(pte_t pte)
472 {
473 	return pte_flags(pte) & _PAGE_HIDDEN;
474 }
475 
476 static inline int pmd_present(pmd_t pmd)
477 {
478 	/*
479 	 * Checking for _PAGE_PSE is needed too because
480 	 * split_huge_page will temporarily clear the present bit (but
481 	 * the _PAGE_PSE flag will remain set at all times while the
482 	 * _PAGE_PRESENT bit is clear).
483 	 */
484 	return pmd_flags(pmd) & (_PAGE_PRESENT | _PAGE_PROTNONE | _PAGE_PSE);
485 }
486 
487 #ifdef CONFIG_NUMA_BALANCING
488 /*
489  * These work without NUMA balancing but the kernel does not care. See the
490  * comment in include/asm-generic/pgtable.h
491  */
492 static inline int pte_protnone(pte_t pte)
493 {
494 	return (pte_flags(pte) & (_PAGE_PROTNONE | _PAGE_PRESENT))
495 		== _PAGE_PROTNONE;
496 }
497 
498 static inline int pmd_protnone(pmd_t pmd)
499 {
500 	return (pmd_flags(pmd) & (_PAGE_PROTNONE | _PAGE_PRESENT))
501 		== _PAGE_PROTNONE;
502 }
503 #endif /* CONFIG_NUMA_BALANCING */
504 
505 static inline int pmd_none(pmd_t pmd)
506 {
507 	/* Only check low word on 32-bit platforms, since it might be
508 	   out of sync with upper half. */
509 	return (unsigned long)native_pmd_val(pmd) == 0;
510 }
511 
512 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
513 {
514 	return (unsigned long)__va(pmd_val(pmd) & pmd_pfn_mask(pmd));
515 }
516 
517 /*
518  * Currently stuck as a macro due to indirect forward reference to
519  * linux/mmzone.h's __section_mem_map_addr() definition:
520  */
521 #define pmd_page(pmd)		\
522 	pfn_to_page((pmd_val(pmd) & pmd_pfn_mask(pmd)) >> PAGE_SHIFT)
523 
524 /*
525  * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
526  *
527  * this macro returns the index of the entry in the pmd page which would
528  * control the given virtual address
529  */
530 static inline unsigned long pmd_index(unsigned long address)
531 {
532 	return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
533 }
534 
535 /*
536  * Conversion functions: convert a page and protection to a page entry,
537  * and a page entry and page directory to the page they refer to.
538  *
539  * (Currently stuck as a macro because of indirect forward reference
540  * to linux/mm.h:page_to_nid())
541  */
542 #define mk_pte(page, pgprot)   pfn_pte(page_to_pfn(page), (pgprot))
543 
544 /*
545  * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
546  *
547  * this function returns the index of the entry in the pte page which would
548  * control the given virtual address
549  */
550 static inline unsigned long pte_index(unsigned long address)
551 {
552 	return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
553 }
554 
555 static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
556 {
557 	return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
558 }
559 
560 static inline int pmd_bad(pmd_t pmd)
561 {
562 	return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
563 }
564 
565 static inline unsigned long pages_to_mb(unsigned long npg)
566 {
567 	return npg >> (20 - PAGE_SHIFT);
568 }
569 
570 #if CONFIG_PGTABLE_LEVELS > 2
571 static inline int pud_none(pud_t pud)
572 {
573 	return native_pud_val(pud) == 0;
574 }
575 
576 static inline int pud_present(pud_t pud)
577 {
578 	return pud_flags(pud) & _PAGE_PRESENT;
579 }
580 
581 static inline unsigned long pud_page_vaddr(pud_t pud)
582 {
583 	return (unsigned long)__va(pud_val(pud) & pud_pfn_mask(pud));
584 }
585 
586 /*
587  * Currently stuck as a macro due to indirect forward reference to
588  * linux/mmzone.h's __section_mem_map_addr() definition:
589  */
590 #define pud_page(pud)		\
591 	pfn_to_page((pud_val(pud) & pud_pfn_mask(pud)) >> PAGE_SHIFT)
592 
593 /* Find an entry in the second-level page table.. */
594 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
595 {
596 	return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
597 }
598 
599 static inline int pud_large(pud_t pud)
600 {
601 	return (pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
602 		(_PAGE_PSE | _PAGE_PRESENT);
603 }
604 
605 static inline int pud_bad(pud_t pud)
606 {
607 	return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
608 }
609 #else
610 static inline int pud_large(pud_t pud)
611 {
612 	return 0;
613 }
614 #endif	/* CONFIG_PGTABLE_LEVELS > 2 */
615 
616 #if CONFIG_PGTABLE_LEVELS > 3
617 static inline int pgd_present(pgd_t pgd)
618 {
619 	return pgd_flags(pgd) & _PAGE_PRESENT;
620 }
621 
622 static inline unsigned long pgd_page_vaddr(pgd_t pgd)
623 {
624 	return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
625 }
626 
627 /*
628  * Currently stuck as a macro due to indirect forward reference to
629  * linux/mmzone.h's __section_mem_map_addr() definition:
630  */
631 #define pgd_page(pgd)		pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
632 
633 /* to find an entry in a page-table-directory. */
634 static inline unsigned long pud_index(unsigned long address)
635 {
636 	return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
637 }
638 
639 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
640 {
641 	return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
642 }
643 
644 static inline int pgd_bad(pgd_t pgd)
645 {
646 	return (pgd_flags(pgd) & ~_PAGE_USER) != _KERNPG_TABLE;
647 }
648 
649 static inline int pgd_none(pgd_t pgd)
650 {
651 	return !native_pgd_val(pgd);
652 }
653 #endif	/* CONFIG_PGTABLE_LEVELS > 3 */
654 
655 #endif	/* __ASSEMBLY__ */
656 
657 /*
658  * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
659  *
660  * this macro returns the index of the entry in the pgd page which would
661  * control the given virtual address
662  */
663 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
664 
665 /*
666  * pgd_offset() returns a (pgd_t *)
667  * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
668  */
669 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
670 /*
671  * a shortcut which implies the use of the kernel's pgd, instead
672  * of a process's
673  */
674 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
675 
676 
677 #define KERNEL_PGD_BOUNDARY	pgd_index(PAGE_OFFSET)
678 #define KERNEL_PGD_PTRS		(PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
679 
680 #ifndef __ASSEMBLY__
681 
682 extern int direct_gbpages;
683 void init_mem_mapping(void);
684 void early_alloc_pgt_buf(void);
685 
686 /* local pte updates need not use xchg for locking */
687 static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
688 {
689 	pte_t res = *ptep;
690 
691 	/* Pure native function needs no input for mm, addr */
692 	native_pte_clear(NULL, 0, ptep);
693 	return res;
694 }
695 
696 static inline pmd_t native_local_pmdp_get_and_clear(pmd_t *pmdp)
697 {
698 	pmd_t res = *pmdp;
699 
700 	native_pmd_clear(pmdp);
701 	return res;
702 }
703 
704 static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
705 				     pte_t *ptep , pte_t pte)
706 {
707 	native_set_pte(ptep, pte);
708 }
709 
710 static inline void native_set_pmd_at(struct mm_struct *mm, unsigned long addr,
711 				     pmd_t *pmdp , pmd_t pmd)
712 {
713 	native_set_pmd(pmdp, pmd);
714 }
715 
716 #ifndef CONFIG_PARAVIRT
717 /*
718  * Rules for using pte_update - it must be called after any PTE update which
719  * has not been done using the set_pte / clear_pte interfaces.  It is used by
720  * shadow mode hypervisors to resynchronize the shadow page tables.  Kernel PTE
721  * updates should either be sets, clears, or set_pte_atomic for P->P
722  * transitions, which means this hook should only be called for user PTEs.
723  * This hook implies a P->P protection or access change has taken place, which
724  * requires a subsequent TLB flush.  The notification can optionally be delayed
725  * until the TLB flush event by using the pte_update_defer form of the
726  * interface, but care must be taken to assure that the flush happens while
727  * still holding the same page table lock so that the shadow and primary pages
728  * do not become out of sync on SMP.
729  */
730 #define pte_update(mm, addr, ptep)		do { } while (0)
731 #define pte_update_defer(mm, addr, ptep)	do { } while (0)
732 #endif
733 
734 /*
735  * We only update the dirty/accessed state if we set
736  * the dirty bit by hand in the kernel, since the hardware
737  * will do the accessed bit for us, and we don't want to
738  * race with other CPU's that might be updating the dirty
739  * bit at the same time.
740  */
741 struct vm_area_struct;
742 
743 #define  __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
744 extern int ptep_set_access_flags(struct vm_area_struct *vma,
745 				 unsigned long address, pte_t *ptep,
746 				 pte_t entry, int dirty);
747 
748 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
749 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
750 				     unsigned long addr, pte_t *ptep);
751 
752 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
753 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
754 				  unsigned long address, pte_t *ptep);
755 
756 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
757 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
758 				       pte_t *ptep)
759 {
760 	pte_t pte = native_ptep_get_and_clear(ptep);
761 	pte_update(mm, addr, ptep);
762 	return pte;
763 }
764 
765 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
766 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
767 					    unsigned long addr, pte_t *ptep,
768 					    int full)
769 {
770 	pte_t pte;
771 	if (full) {
772 		/*
773 		 * Full address destruction in progress; paravirt does not
774 		 * care about updates and native needs no locking
775 		 */
776 		pte = native_local_ptep_get_and_clear(ptep);
777 	} else {
778 		pte = ptep_get_and_clear(mm, addr, ptep);
779 	}
780 	return pte;
781 }
782 
783 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
784 static inline void ptep_set_wrprotect(struct mm_struct *mm,
785 				      unsigned long addr, pte_t *ptep)
786 {
787 	clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
788 	pte_update(mm, addr, ptep);
789 }
790 
791 #define flush_tlb_fix_spurious_fault(vma, address) do { } while (0)
792 
793 #define mk_pmd(page, pgprot)   pfn_pmd(page_to_pfn(page), (pgprot))
794 
795 #define  __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
796 extern int pmdp_set_access_flags(struct vm_area_struct *vma,
797 				 unsigned long address, pmd_t *pmdp,
798 				 pmd_t entry, int dirty);
799 
800 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
801 extern int pmdp_test_and_clear_young(struct vm_area_struct *vma,
802 				     unsigned long addr, pmd_t *pmdp);
803 
804 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
805 extern int pmdp_clear_flush_young(struct vm_area_struct *vma,
806 				  unsigned long address, pmd_t *pmdp);
807 
808 
809 #define __HAVE_ARCH_PMDP_SPLITTING_FLUSH
810 extern void pmdp_splitting_flush(struct vm_area_struct *vma,
811 				 unsigned long addr, pmd_t *pmdp);
812 
813 #define __HAVE_ARCH_PMD_WRITE
814 static inline int pmd_write(pmd_t pmd)
815 {
816 	return pmd_flags(pmd) & _PAGE_RW;
817 }
818 
819 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
820 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm, unsigned long addr,
821 				       pmd_t *pmdp)
822 {
823 	pmd_t pmd = native_pmdp_get_and_clear(pmdp);
824 	pmd_update(mm, addr, pmdp);
825 	return pmd;
826 }
827 
828 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
829 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
830 				      unsigned long addr, pmd_t *pmdp)
831 {
832 	clear_bit(_PAGE_BIT_RW, (unsigned long *)pmdp);
833 	pmd_update(mm, addr, pmdp);
834 }
835 
836 /*
837  * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
838  *
839  *  dst - pointer to pgd range anwhere on a pgd page
840  *  src - ""
841  *  count - the number of pgds to copy.
842  *
843  * dst and src can be on the same page, but the range must not overlap,
844  * and must not cross a page boundary.
845  */
846 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
847 {
848        memcpy(dst, src, count * sizeof(pgd_t));
849 }
850 
851 #define PTE_SHIFT ilog2(PTRS_PER_PTE)
852 static inline int page_level_shift(enum pg_level level)
853 {
854 	return (PAGE_SHIFT - PTE_SHIFT) + level * PTE_SHIFT;
855 }
856 static inline unsigned long page_level_size(enum pg_level level)
857 {
858 	return 1UL << page_level_shift(level);
859 }
860 static inline unsigned long page_level_mask(enum pg_level level)
861 {
862 	return ~(page_level_size(level) - 1);
863 }
864 
865 /*
866  * The x86 doesn't have any external MMU info: the kernel page
867  * tables contain all the necessary information.
868  */
869 static inline void update_mmu_cache(struct vm_area_struct *vma,
870 		unsigned long addr, pte_t *ptep)
871 {
872 }
873 static inline void update_mmu_cache_pmd(struct vm_area_struct *vma,
874 		unsigned long addr, pmd_t *pmd)
875 {
876 }
877 
878 #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY
879 static inline pte_t pte_swp_mksoft_dirty(pte_t pte)
880 {
881 	return pte_set_flags(pte, _PAGE_SWP_SOFT_DIRTY);
882 }
883 
884 static inline int pte_swp_soft_dirty(pte_t pte)
885 {
886 	return pte_flags(pte) & _PAGE_SWP_SOFT_DIRTY;
887 }
888 
889 static inline pte_t pte_swp_clear_soft_dirty(pte_t pte)
890 {
891 	return pte_clear_flags(pte, _PAGE_SWP_SOFT_DIRTY);
892 }
893 #endif
894 
895 #include <asm-generic/pgtable.h>
896 #endif	/* __ASSEMBLY__ */
897 
898 #endif /* _ASM_X86_PGTABLE_H */
899