xref: /openbmc/linux/arch/x86/include/asm/pgtable.h (revision 781095f903f398148cd0b646d3984234a715f29e)
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 
73 #define pgd_val(x)	native_pgd_val(x)
74 #define __pgd(x)	native_make_pgd(x)
75 
76 #ifndef __PAGETABLE_PUD_FOLDED
77 #define pud_val(x)	native_pud_val(x)
78 #define __pud(x)	native_make_pud(x)
79 #endif
80 
81 #ifndef __PAGETABLE_PMD_FOLDED
82 #define pmd_val(x)	native_pmd_val(x)
83 #define __pmd(x)	native_make_pmd(x)
84 #endif
85 
86 #define pte_val(x)	native_pte_val(x)
87 #define __pte(x)	native_make_pte(x)
88 
89 #define arch_end_context_switch(prev)	do {} while(0)
90 
91 #endif	/* CONFIG_PARAVIRT */
92 
93 /*
94  * The following only work if pte_present() is true.
95  * Undefined behaviour if not..
96  */
97 static inline int pte_dirty(pte_t pte)
98 {
99 	return pte_flags(pte) & _PAGE_DIRTY;
100 }
101 
102 static inline int pte_young(pte_t pte)
103 {
104 	return pte_flags(pte) & _PAGE_ACCESSED;
105 }
106 
107 static inline int pmd_dirty(pmd_t pmd)
108 {
109 	return pmd_flags(pmd) & _PAGE_DIRTY;
110 }
111 
112 static inline int pmd_young(pmd_t pmd)
113 {
114 	return pmd_flags(pmd) & _PAGE_ACCESSED;
115 }
116 
117 static inline int pte_write(pte_t pte)
118 {
119 	return pte_flags(pte) & _PAGE_RW;
120 }
121 
122 static inline int pte_huge(pte_t pte)
123 {
124 	return pte_flags(pte) & _PAGE_PSE;
125 }
126 
127 static inline int pte_global(pte_t pte)
128 {
129 	return pte_flags(pte) & _PAGE_GLOBAL;
130 }
131 
132 static inline int pte_exec(pte_t pte)
133 {
134 	return !(pte_flags(pte) & _PAGE_NX);
135 }
136 
137 static inline int pte_special(pte_t pte)
138 {
139 	return pte_flags(pte) & _PAGE_SPECIAL;
140 }
141 
142 static inline unsigned long pte_pfn(pte_t pte)
143 {
144 	return (pte_val(pte) & PTE_PFN_MASK) >> PAGE_SHIFT;
145 }
146 
147 static inline unsigned long pmd_pfn(pmd_t pmd)
148 {
149 	return (pmd_val(pmd) & pmd_pfn_mask(pmd)) >> PAGE_SHIFT;
150 }
151 
152 static inline unsigned long pud_pfn(pud_t pud)
153 {
154 	return (pud_val(pud) & pud_pfn_mask(pud)) >> PAGE_SHIFT;
155 }
156 
157 #define pte_page(pte)	pfn_to_page(pte_pfn(pte))
158 
159 static inline int pmd_large(pmd_t pte)
160 {
161 	return pmd_flags(pte) & _PAGE_PSE;
162 }
163 
164 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
165 static inline int pmd_trans_huge(pmd_t pmd)
166 {
167 	return (pmd_val(pmd) & (_PAGE_PSE|_PAGE_DEVMAP)) == _PAGE_PSE;
168 }
169 
170 static inline int has_transparent_hugepage(void)
171 {
172 	return cpu_has_pse;
173 }
174 
175 #ifdef __HAVE_ARCH_PTE_DEVMAP
176 static inline int pmd_devmap(pmd_t pmd)
177 {
178 	return !!(pmd_val(pmd) & _PAGE_DEVMAP);
179 }
180 #endif
181 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
182 
183 static inline pte_t pte_set_flags(pte_t pte, pteval_t set)
184 {
185 	pteval_t v = native_pte_val(pte);
186 
187 	return native_make_pte(v | set);
188 }
189 
190 static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
191 {
192 	pteval_t v = native_pte_val(pte);
193 
194 	return native_make_pte(v & ~clear);
195 }
196 
197 static inline pte_t pte_mkclean(pte_t pte)
198 {
199 	return pte_clear_flags(pte, _PAGE_DIRTY);
200 }
201 
202 static inline pte_t pte_mkold(pte_t pte)
203 {
204 	return pte_clear_flags(pte, _PAGE_ACCESSED);
205 }
206 
207 static inline pte_t pte_wrprotect(pte_t pte)
208 {
209 	return pte_clear_flags(pte, _PAGE_RW);
210 }
211 
212 static inline pte_t pte_mkexec(pte_t pte)
213 {
214 	return pte_clear_flags(pte, _PAGE_NX);
215 }
216 
217 static inline pte_t pte_mkdirty(pte_t pte)
218 {
219 	return pte_set_flags(pte, _PAGE_DIRTY | _PAGE_SOFT_DIRTY);
220 }
221 
222 static inline pte_t pte_mkyoung(pte_t pte)
223 {
224 	return pte_set_flags(pte, _PAGE_ACCESSED);
225 }
226 
227 static inline pte_t pte_mkwrite(pte_t pte)
228 {
229 	return pte_set_flags(pte, _PAGE_RW);
230 }
231 
232 static inline pte_t pte_mkhuge(pte_t pte)
233 {
234 	return pte_set_flags(pte, _PAGE_PSE);
235 }
236 
237 static inline pte_t pte_clrhuge(pte_t pte)
238 {
239 	return pte_clear_flags(pte, _PAGE_PSE);
240 }
241 
242 static inline pte_t pte_mkglobal(pte_t pte)
243 {
244 	return pte_set_flags(pte, _PAGE_GLOBAL);
245 }
246 
247 static inline pte_t pte_clrglobal(pte_t pte)
248 {
249 	return pte_clear_flags(pte, _PAGE_GLOBAL);
250 }
251 
252 static inline pte_t pte_mkspecial(pte_t pte)
253 {
254 	return pte_set_flags(pte, _PAGE_SPECIAL);
255 }
256 
257 static inline pte_t pte_mkdevmap(pte_t pte)
258 {
259 	return pte_set_flags(pte, _PAGE_SPECIAL|_PAGE_DEVMAP);
260 }
261 
262 static inline pmd_t pmd_set_flags(pmd_t pmd, pmdval_t set)
263 {
264 	pmdval_t v = native_pmd_val(pmd);
265 
266 	return __pmd(v | set);
267 }
268 
269 static inline pmd_t pmd_clear_flags(pmd_t pmd, pmdval_t clear)
270 {
271 	pmdval_t v = native_pmd_val(pmd);
272 
273 	return __pmd(v & ~clear);
274 }
275 
276 static inline pmd_t pmd_mkold(pmd_t pmd)
277 {
278 	return pmd_clear_flags(pmd, _PAGE_ACCESSED);
279 }
280 
281 static inline pmd_t pmd_mkclean(pmd_t pmd)
282 {
283 	return pmd_clear_flags(pmd, _PAGE_DIRTY);
284 }
285 
286 static inline pmd_t pmd_wrprotect(pmd_t pmd)
287 {
288 	return pmd_clear_flags(pmd, _PAGE_RW);
289 }
290 
291 static inline pmd_t pmd_mkdirty(pmd_t pmd)
292 {
293 	return pmd_set_flags(pmd, _PAGE_DIRTY | _PAGE_SOFT_DIRTY);
294 }
295 
296 static inline pmd_t pmd_mkdevmap(pmd_t pmd)
297 {
298 	return pmd_set_flags(pmd, _PAGE_DEVMAP);
299 }
300 
301 static inline pmd_t pmd_mkhuge(pmd_t pmd)
302 {
303 	return pmd_set_flags(pmd, _PAGE_PSE);
304 }
305 
306 static inline pmd_t pmd_mkyoung(pmd_t pmd)
307 {
308 	return pmd_set_flags(pmd, _PAGE_ACCESSED);
309 }
310 
311 static inline pmd_t pmd_mkwrite(pmd_t pmd)
312 {
313 	return pmd_set_flags(pmd, _PAGE_RW);
314 }
315 
316 static inline pmd_t pmd_mknotpresent(pmd_t pmd)
317 {
318 	return pmd_clear_flags(pmd, _PAGE_PRESENT | _PAGE_PROTNONE);
319 }
320 
321 #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY
322 static inline int pte_soft_dirty(pte_t pte)
323 {
324 	return pte_flags(pte) & _PAGE_SOFT_DIRTY;
325 }
326 
327 static inline int pmd_soft_dirty(pmd_t pmd)
328 {
329 	return pmd_flags(pmd) & _PAGE_SOFT_DIRTY;
330 }
331 
332 static inline pte_t pte_mksoft_dirty(pte_t pte)
333 {
334 	return pte_set_flags(pte, _PAGE_SOFT_DIRTY);
335 }
336 
337 static inline pmd_t pmd_mksoft_dirty(pmd_t pmd)
338 {
339 	return pmd_set_flags(pmd, _PAGE_SOFT_DIRTY);
340 }
341 
342 static inline pte_t pte_clear_soft_dirty(pte_t pte)
343 {
344 	return pte_clear_flags(pte, _PAGE_SOFT_DIRTY);
345 }
346 
347 static inline pmd_t pmd_clear_soft_dirty(pmd_t pmd)
348 {
349 	return pmd_clear_flags(pmd, _PAGE_SOFT_DIRTY);
350 }
351 
352 #endif /* CONFIG_HAVE_ARCH_SOFT_DIRTY */
353 
354 /*
355  * Mask out unsupported bits in a present pgprot.  Non-present pgprots
356  * can use those bits for other purposes, so leave them be.
357  */
358 static inline pgprotval_t massage_pgprot(pgprot_t pgprot)
359 {
360 	pgprotval_t protval = pgprot_val(pgprot);
361 
362 	if (protval & _PAGE_PRESENT)
363 		protval &= __supported_pte_mask;
364 
365 	return protval;
366 }
367 
368 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
369 {
370 	return __pte(((phys_addr_t)page_nr << PAGE_SHIFT) |
371 		     massage_pgprot(pgprot));
372 }
373 
374 static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
375 {
376 	return __pmd(((phys_addr_t)page_nr << PAGE_SHIFT) |
377 		     massage_pgprot(pgprot));
378 }
379 
380 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
381 {
382 	pteval_t val = pte_val(pte);
383 
384 	/*
385 	 * Chop off the NX bit (if present), and add the NX portion of
386 	 * the newprot (if present):
387 	 */
388 	val &= _PAGE_CHG_MASK;
389 	val |= massage_pgprot(newprot) & ~_PAGE_CHG_MASK;
390 
391 	return __pte(val);
392 }
393 
394 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
395 {
396 	pmdval_t val = pmd_val(pmd);
397 
398 	val &= _HPAGE_CHG_MASK;
399 	val |= massage_pgprot(newprot) & ~_HPAGE_CHG_MASK;
400 
401 	return __pmd(val);
402 }
403 
404 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
405 #define pgprot_modify pgprot_modify
406 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
407 {
408 	pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
409 	pgprotval_t addbits = pgprot_val(newprot);
410 	return __pgprot(preservebits | addbits);
411 }
412 
413 #define pte_pgprot(x) __pgprot(pte_flags(x))
414 #define pmd_pgprot(x) __pgprot(pmd_flags(x))
415 #define pud_pgprot(x) __pgprot(pud_flags(x))
416 
417 #define canon_pgprot(p) __pgprot(massage_pgprot(p))
418 
419 static inline int is_new_memtype_allowed(u64 paddr, unsigned long size,
420 					 enum page_cache_mode pcm,
421 					 enum page_cache_mode new_pcm)
422 {
423 	/*
424 	 * PAT type is always WB for untracked ranges, so no need to check.
425 	 */
426 	if (x86_platform.is_untracked_pat_range(paddr, paddr + size))
427 		return 1;
428 
429 	/*
430 	 * Certain new memtypes are not allowed with certain
431 	 * requested memtype:
432 	 * - request is uncached, return cannot be write-back
433 	 * - request is write-combine, return cannot be write-back
434 	 * - request is write-through, return cannot be write-back
435 	 * - request is write-through, return cannot be write-combine
436 	 */
437 	if ((pcm == _PAGE_CACHE_MODE_UC_MINUS &&
438 	     new_pcm == _PAGE_CACHE_MODE_WB) ||
439 	    (pcm == _PAGE_CACHE_MODE_WC &&
440 	     new_pcm == _PAGE_CACHE_MODE_WB) ||
441 	    (pcm == _PAGE_CACHE_MODE_WT &&
442 	     new_pcm == _PAGE_CACHE_MODE_WB) ||
443 	    (pcm == _PAGE_CACHE_MODE_WT &&
444 	     new_pcm == _PAGE_CACHE_MODE_WC)) {
445 		return 0;
446 	}
447 
448 	return 1;
449 }
450 
451 pmd_t *populate_extra_pmd(unsigned long vaddr);
452 pte_t *populate_extra_pte(unsigned long vaddr);
453 #endif	/* __ASSEMBLY__ */
454 
455 #ifdef CONFIG_X86_32
456 # include <asm/pgtable_32.h>
457 #else
458 # include <asm/pgtable_64.h>
459 #endif
460 
461 #ifndef __ASSEMBLY__
462 #include <linux/mm_types.h>
463 #include <linux/mmdebug.h>
464 #include <linux/log2.h>
465 
466 static inline int pte_none(pte_t pte)
467 {
468 	return !pte.pte;
469 }
470 
471 #define __HAVE_ARCH_PTE_SAME
472 static inline int pte_same(pte_t a, pte_t b)
473 {
474 	return a.pte == b.pte;
475 }
476 
477 static inline int pte_present(pte_t a)
478 {
479 	return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
480 }
481 
482 #ifdef __HAVE_ARCH_PTE_DEVMAP
483 static inline int pte_devmap(pte_t a)
484 {
485 	return (pte_flags(a) & _PAGE_DEVMAP) == _PAGE_DEVMAP;
486 }
487 #endif
488 
489 #define pte_accessible pte_accessible
490 static inline bool pte_accessible(struct mm_struct *mm, pte_t a)
491 {
492 	if (pte_flags(a) & _PAGE_PRESENT)
493 		return true;
494 
495 	if ((pte_flags(a) & _PAGE_PROTNONE) &&
496 			mm_tlb_flush_pending(mm))
497 		return true;
498 
499 	return false;
500 }
501 
502 static inline int pte_hidden(pte_t pte)
503 {
504 	return pte_flags(pte) & _PAGE_HIDDEN;
505 }
506 
507 static inline int pmd_present(pmd_t pmd)
508 {
509 	/*
510 	 * Checking for _PAGE_PSE is needed too because
511 	 * split_huge_page will temporarily clear the present bit (but
512 	 * the _PAGE_PSE flag will remain set at all times while the
513 	 * _PAGE_PRESENT bit is clear).
514 	 */
515 	return pmd_flags(pmd) & (_PAGE_PRESENT | _PAGE_PROTNONE | _PAGE_PSE);
516 }
517 
518 #ifdef CONFIG_NUMA_BALANCING
519 /*
520  * These work without NUMA balancing but the kernel does not care. See the
521  * comment in include/asm-generic/pgtable.h
522  */
523 static inline int pte_protnone(pte_t pte)
524 {
525 	return (pte_flags(pte) & (_PAGE_PROTNONE | _PAGE_PRESENT))
526 		== _PAGE_PROTNONE;
527 }
528 
529 static inline int pmd_protnone(pmd_t pmd)
530 {
531 	return (pmd_flags(pmd) & (_PAGE_PROTNONE | _PAGE_PRESENT))
532 		== _PAGE_PROTNONE;
533 }
534 #endif /* CONFIG_NUMA_BALANCING */
535 
536 static inline int pmd_none(pmd_t pmd)
537 {
538 	/* Only check low word on 32-bit platforms, since it might be
539 	   out of sync with upper half. */
540 	return (unsigned long)native_pmd_val(pmd) == 0;
541 }
542 
543 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
544 {
545 	return (unsigned long)__va(pmd_val(pmd) & pmd_pfn_mask(pmd));
546 }
547 
548 /*
549  * Currently stuck as a macro due to indirect forward reference to
550  * linux/mmzone.h's __section_mem_map_addr() definition:
551  */
552 #define pmd_page(pmd)		\
553 	pfn_to_page((pmd_val(pmd) & pmd_pfn_mask(pmd)) >> PAGE_SHIFT)
554 
555 /*
556  * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
557  *
558  * this macro returns the index of the entry in the pmd page which would
559  * control the given virtual address
560  */
561 static inline unsigned long pmd_index(unsigned long address)
562 {
563 	return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
564 }
565 
566 /*
567  * Conversion functions: convert a page and protection to a page entry,
568  * and a page entry and page directory to the page they refer to.
569  *
570  * (Currently stuck as a macro because of indirect forward reference
571  * to linux/mm.h:page_to_nid())
572  */
573 #define mk_pte(page, pgprot)   pfn_pte(page_to_pfn(page), (pgprot))
574 
575 /*
576  * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
577  *
578  * this function returns the index of the entry in the pte page which would
579  * control the given virtual address
580  */
581 static inline unsigned long pte_index(unsigned long address)
582 {
583 	return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
584 }
585 
586 static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
587 {
588 	return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
589 }
590 
591 static inline int pmd_bad(pmd_t pmd)
592 {
593 	return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
594 }
595 
596 static inline unsigned long pages_to_mb(unsigned long npg)
597 {
598 	return npg >> (20 - PAGE_SHIFT);
599 }
600 
601 #if CONFIG_PGTABLE_LEVELS > 2
602 static inline int pud_none(pud_t pud)
603 {
604 	return native_pud_val(pud) == 0;
605 }
606 
607 static inline int pud_present(pud_t pud)
608 {
609 	return pud_flags(pud) & _PAGE_PRESENT;
610 }
611 
612 static inline unsigned long pud_page_vaddr(pud_t pud)
613 {
614 	return (unsigned long)__va(pud_val(pud) & pud_pfn_mask(pud));
615 }
616 
617 /*
618  * Currently stuck as a macro due to indirect forward reference to
619  * linux/mmzone.h's __section_mem_map_addr() definition:
620  */
621 #define pud_page(pud)		\
622 	pfn_to_page((pud_val(pud) & pud_pfn_mask(pud)) >> PAGE_SHIFT)
623 
624 /* Find an entry in the second-level page table.. */
625 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
626 {
627 	return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
628 }
629 
630 static inline int pud_large(pud_t pud)
631 {
632 	return (pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
633 		(_PAGE_PSE | _PAGE_PRESENT);
634 }
635 
636 static inline int pud_bad(pud_t pud)
637 {
638 	return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
639 }
640 #else
641 static inline int pud_large(pud_t pud)
642 {
643 	return 0;
644 }
645 #endif	/* CONFIG_PGTABLE_LEVELS > 2 */
646 
647 #if CONFIG_PGTABLE_LEVELS > 3
648 static inline int pgd_present(pgd_t pgd)
649 {
650 	return pgd_flags(pgd) & _PAGE_PRESENT;
651 }
652 
653 static inline unsigned long pgd_page_vaddr(pgd_t pgd)
654 {
655 	return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
656 }
657 
658 /*
659  * Currently stuck as a macro due to indirect forward reference to
660  * linux/mmzone.h's __section_mem_map_addr() definition:
661  */
662 #define pgd_page(pgd)		pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
663 
664 /* to find an entry in a page-table-directory. */
665 static inline unsigned long pud_index(unsigned long address)
666 {
667 	return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
668 }
669 
670 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
671 {
672 	return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
673 }
674 
675 static inline int pgd_bad(pgd_t pgd)
676 {
677 	return (pgd_flags(pgd) & ~_PAGE_USER) != _KERNPG_TABLE;
678 }
679 
680 static inline int pgd_none(pgd_t pgd)
681 {
682 	return !native_pgd_val(pgd);
683 }
684 #endif	/* CONFIG_PGTABLE_LEVELS > 3 */
685 
686 #endif	/* __ASSEMBLY__ */
687 
688 /*
689  * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
690  *
691  * this macro returns the index of the entry in the pgd page which would
692  * control the given virtual address
693  */
694 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
695 
696 /*
697  * pgd_offset() returns a (pgd_t *)
698  * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
699  */
700 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
701 /*
702  * a shortcut which implies the use of the kernel's pgd, instead
703  * of a process's
704  */
705 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
706 
707 
708 #define KERNEL_PGD_BOUNDARY	pgd_index(PAGE_OFFSET)
709 #define KERNEL_PGD_PTRS		(PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
710 
711 #ifndef __ASSEMBLY__
712 
713 extern int direct_gbpages;
714 void init_mem_mapping(void);
715 void early_alloc_pgt_buf(void);
716 
717 /* local pte updates need not use xchg for locking */
718 static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
719 {
720 	pte_t res = *ptep;
721 
722 	/* Pure native function needs no input for mm, addr */
723 	native_pte_clear(NULL, 0, ptep);
724 	return res;
725 }
726 
727 static inline pmd_t native_local_pmdp_get_and_clear(pmd_t *pmdp)
728 {
729 	pmd_t res = *pmdp;
730 
731 	native_pmd_clear(pmdp);
732 	return res;
733 }
734 
735 static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
736 				     pte_t *ptep , pte_t pte)
737 {
738 	native_set_pte(ptep, pte);
739 }
740 
741 static inline void native_set_pmd_at(struct mm_struct *mm, unsigned long addr,
742 				     pmd_t *pmdp , pmd_t pmd)
743 {
744 	native_set_pmd(pmdp, pmd);
745 }
746 
747 #ifndef CONFIG_PARAVIRT
748 /*
749  * Rules for using pte_update - it must be called after any PTE update which
750  * has not been done using the set_pte / clear_pte interfaces.  It is used by
751  * shadow mode hypervisors to resynchronize the shadow page tables.  Kernel PTE
752  * updates should either be sets, clears, or set_pte_atomic for P->P
753  * transitions, which means this hook should only be called for user PTEs.
754  * This hook implies a P->P protection or access change has taken place, which
755  * requires a subsequent TLB flush.
756  */
757 #define pte_update(mm, addr, ptep)		do { } while (0)
758 #endif
759 
760 /*
761  * We only update the dirty/accessed state if we set
762  * the dirty bit by hand in the kernel, since the hardware
763  * will do the accessed bit for us, and we don't want to
764  * race with other CPU's that might be updating the dirty
765  * bit at the same time.
766  */
767 struct vm_area_struct;
768 
769 #define  __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
770 extern int ptep_set_access_flags(struct vm_area_struct *vma,
771 				 unsigned long address, pte_t *ptep,
772 				 pte_t entry, int dirty);
773 
774 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
775 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
776 				     unsigned long addr, pte_t *ptep);
777 
778 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
779 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
780 				  unsigned long address, pte_t *ptep);
781 
782 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
783 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
784 				       pte_t *ptep)
785 {
786 	pte_t pte = native_ptep_get_and_clear(ptep);
787 	pte_update(mm, addr, ptep);
788 	return pte;
789 }
790 
791 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
792 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
793 					    unsigned long addr, pte_t *ptep,
794 					    int full)
795 {
796 	pte_t pte;
797 	if (full) {
798 		/*
799 		 * Full address destruction in progress; paravirt does not
800 		 * care about updates and native needs no locking
801 		 */
802 		pte = native_local_ptep_get_and_clear(ptep);
803 	} else {
804 		pte = ptep_get_and_clear(mm, addr, ptep);
805 	}
806 	return pte;
807 }
808 
809 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
810 static inline void ptep_set_wrprotect(struct mm_struct *mm,
811 				      unsigned long addr, pte_t *ptep)
812 {
813 	clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
814 	pte_update(mm, addr, ptep);
815 }
816 
817 #define flush_tlb_fix_spurious_fault(vma, address) do { } while (0)
818 
819 #define mk_pmd(page, pgprot)   pfn_pmd(page_to_pfn(page), (pgprot))
820 
821 #define  __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
822 extern int pmdp_set_access_flags(struct vm_area_struct *vma,
823 				 unsigned long address, pmd_t *pmdp,
824 				 pmd_t entry, int dirty);
825 
826 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
827 extern int pmdp_test_and_clear_young(struct vm_area_struct *vma,
828 				     unsigned long addr, pmd_t *pmdp);
829 
830 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
831 extern int pmdp_clear_flush_young(struct vm_area_struct *vma,
832 				  unsigned long address, pmd_t *pmdp);
833 
834 
835 #define __HAVE_ARCH_PMD_WRITE
836 static inline int pmd_write(pmd_t pmd)
837 {
838 	return pmd_flags(pmd) & _PAGE_RW;
839 }
840 
841 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
842 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm, unsigned long addr,
843 				       pmd_t *pmdp)
844 {
845 	return native_pmdp_get_and_clear(pmdp);
846 }
847 
848 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
849 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
850 				      unsigned long addr, pmd_t *pmdp)
851 {
852 	clear_bit(_PAGE_BIT_RW, (unsigned long *)pmdp);
853 }
854 
855 /*
856  * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
857  *
858  *  dst - pointer to pgd range anwhere on a pgd page
859  *  src - ""
860  *  count - the number of pgds to copy.
861  *
862  * dst and src can be on the same page, but the range must not overlap,
863  * and must not cross a page boundary.
864  */
865 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
866 {
867        memcpy(dst, src, count * sizeof(pgd_t));
868 }
869 
870 #define PTE_SHIFT ilog2(PTRS_PER_PTE)
871 static inline int page_level_shift(enum pg_level level)
872 {
873 	return (PAGE_SHIFT - PTE_SHIFT) + level * PTE_SHIFT;
874 }
875 static inline unsigned long page_level_size(enum pg_level level)
876 {
877 	return 1UL << page_level_shift(level);
878 }
879 static inline unsigned long page_level_mask(enum pg_level level)
880 {
881 	return ~(page_level_size(level) - 1);
882 }
883 
884 /*
885  * The x86 doesn't have any external MMU info: the kernel page
886  * tables contain all the necessary information.
887  */
888 static inline void update_mmu_cache(struct vm_area_struct *vma,
889 		unsigned long addr, pte_t *ptep)
890 {
891 }
892 static inline void update_mmu_cache_pmd(struct vm_area_struct *vma,
893 		unsigned long addr, pmd_t *pmd)
894 {
895 }
896 
897 #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY
898 static inline pte_t pte_swp_mksoft_dirty(pte_t pte)
899 {
900 	return pte_set_flags(pte, _PAGE_SWP_SOFT_DIRTY);
901 }
902 
903 static inline int pte_swp_soft_dirty(pte_t pte)
904 {
905 	return pte_flags(pte) & _PAGE_SWP_SOFT_DIRTY;
906 }
907 
908 static inline pte_t pte_swp_clear_soft_dirty(pte_t pte)
909 {
910 	return pte_clear_flags(pte, _PAGE_SWP_SOFT_DIRTY);
911 }
912 #endif
913 
914 #include <asm-generic/pgtable.h>
915 #endif	/* __ASSEMBLY__ */
916 
917 #endif /* _ASM_X86_PGTABLE_H */
918