xref: /linux-6.15/include/linux/pgtable.h (revision ca5999fd)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_PGTABLE_H
3 #define _LINUX_PGTABLE_H
4 
5 #include <linux/pfn.h>
6 #include <asm/pgtable.h>
7 
8 #ifndef __ASSEMBLY__
9 #ifdef CONFIG_MMU
10 
11 #include <linux/mm_types.h>
12 #include <linux/bug.h>
13 #include <linux/errno.h>
14 #include <asm-generic/pgtable_uffd.h>
15 
16 #if 5 - defined(__PAGETABLE_P4D_FOLDED) - defined(__PAGETABLE_PUD_FOLDED) - \
17 	defined(__PAGETABLE_PMD_FOLDED) != CONFIG_PGTABLE_LEVELS
18 #error CONFIG_PGTABLE_LEVELS is not consistent with __PAGETABLE_{P4D,PUD,PMD}_FOLDED
19 #endif
20 
21 /*
22  * On almost all architectures and configurations, 0 can be used as the
23  * upper ceiling to free_pgtables(): on many architectures it has the same
24  * effect as using TASK_SIZE.  However, there is one configuration which
25  * must impose a more careful limit, to avoid freeing kernel pgtables.
26  */
27 #ifndef USER_PGTABLES_CEILING
28 #define USER_PGTABLES_CEILING	0UL
29 #endif
30 
31 #ifndef __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
32 extern int ptep_set_access_flags(struct vm_area_struct *vma,
33 				 unsigned long address, pte_t *ptep,
34 				 pte_t entry, int dirty);
35 #endif
36 
37 #ifndef __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
38 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
39 extern int pmdp_set_access_flags(struct vm_area_struct *vma,
40 				 unsigned long address, pmd_t *pmdp,
41 				 pmd_t entry, int dirty);
42 extern int pudp_set_access_flags(struct vm_area_struct *vma,
43 				 unsigned long address, pud_t *pudp,
44 				 pud_t entry, int dirty);
45 #else
46 static inline int pmdp_set_access_flags(struct vm_area_struct *vma,
47 					unsigned long address, pmd_t *pmdp,
48 					pmd_t entry, int dirty)
49 {
50 	BUILD_BUG();
51 	return 0;
52 }
53 static inline int pudp_set_access_flags(struct vm_area_struct *vma,
54 					unsigned long address, pud_t *pudp,
55 					pud_t entry, int dirty)
56 {
57 	BUILD_BUG();
58 	return 0;
59 }
60 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
61 #endif
62 
63 #ifndef __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
64 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
65 					    unsigned long address,
66 					    pte_t *ptep)
67 {
68 	pte_t pte = *ptep;
69 	int r = 1;
70 	if (!pte_young(pte))
71 		r = 0;
72 	else
73 		set_pte_at(vma->vm_mm, address, ptep, pte_mkold(pte));
74 	return r;
75 }
76 #endif
77 
78 #ifndef __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
79 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
80 static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
81 					    unsigned long address,
82 					    pmd_t *pmdp)
83 {
84 	pmd_t pmd = *pmdp;
85 	int r = 1;
86 	if (!pmd_young(pmd))
87 		r = 0;
88 	else
89 		set_pmd_at(vma->vm_mm, address, pmdp, pmd_mkold(pmd));
90 	return r;
91 }
92 #else
93 static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
94 					    unsigned long address,
95 					    pmd_t *pmdp)
96 {
97 	BUILD_BUG();
98 	return 0;
99 }
100 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
101 #endif
102 
103 #ifndef __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
104 int ptep_clear_flush_young(struct vm_area_struct *vma,
105 			   unsigned long address, pte_t *ptep);
106 #endif
107 
108 #ifndef __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
109 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
110 extern int pmdp_clear_flush_young(struct vm_area_struct *vma,
111 				  unsigned long address, pmd_t *pmdp);
112 #else
113 /*
114  * Despite relevant to THP only, this API is called from generic rmap code
115  * under PageTransHuge(), hence needs a dummy implementation for !THP
116  */
117 static inline int pmdp_clear_flush_young(struct vm_area_struct *vma,
118 					 unsigned long address, pmd_t *pmdp)
119 {
120 	BUILD_BUG();
121 	return 0;
122 }
123 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
124 #endif
125 
126 #ifndef __HAVE_ARCH_PTEP_GET_AND_CLEAR
127 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
128 				       unsigned long address,
129 				       pte_t *ptep)
130 {
131 	pte_t pte = *ptep;
132 	pte_clear(mm, address, ptep);
133 	return pte;
134 }
135 #endif
136 
137 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
138 #ifndef __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
139 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm,
140 					    unsigned long address,
141 					    pmd_t *pmdp)
142 {
143 	pmd_t pmd = *pmdp;
144 	pmd_clear(pmdp);
145 	return pmd;
146 }
147 #endif /* __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR */
148 #ifndef __HAVE_ARCH_PUDP_HUGE_GET_AND_CLEAR
149 static inline pud_t pudp_huge_get_and_clear(struct mm_struct *mm,
150 					    unsigned long address,
151 					    pud_t *pudp)
152 {
153 	pud_t pud = *pudp;
154 
155 	pud_clear(pudp);
156 	return pud;
157 }
158 #endif /* __HAVE_ARCH_PUDP_HUGE_GET_AND_CLEAR */
159 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
160 
161 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
162 #ifndef __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR_FULL
163 static inline pmd_t pmdp_huge_get_and_clear_full(struct vm_area_struct *vma,
164 					    unsigned long address, pmd_t *pmdp,
165 					    int full)
166 {
167 	return pmdp_huge_get_and_clear(vma->vm_mm, address, pmdp);
168 }
169 #endif
170 
171 #ifndef __HAVE_ARCH_PUDP_HUGE_GET_AND_CLEAR_FULL
172 static inline pud_t pudp_huge_get_and_clear_full(struct mm_struct *mm,
173 					    unsigned long address, pud_t *pudp,
174 					    int full)
175 {
176 	return pudp_huge_get_and_clear(mm, address, pudp);
177 }
178 #endif
179 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
180 
181 #ifndef __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
182 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
183 					    unsigned long address, pte_t *ptep,
184 					    int full)
185 {
186 	pte_t pte;
187 	pte = ptep_get_and_clear(mm, address, ptep);
188 	return pte;
189 }
190 #endif
191 
192 
193 /*
194  * If two threads concurrently fault at the same page, the thread that
195  * won the race updates the PTE and its local TLB/Cache. The other thread
196  * gives up, simply does nothing, and continues; on architectures where
197  * software can update TLB,  local TLB can be updated here to avoid next page
198  * fault. This function updates TLB only, do nothing with cache or others.
199  * It is the difference with function update_mmu_cache.
200  */
201 #ifndef __HAVE_ARCH_UPDATE_MMU_TLB
202 static inline void update_mmu_tlb(struct vm_area_struct *vma,
203 				unsigned long address, pte_t *ptep)
204 {
205 }
206 #define __HAVE_ARCH_UPDATE_MMU_TLB
207 #endif
208 
209 /*
210  * Some architectures may be able to avoid expensive synchronization
211  * primitives when modifications are made to PTE's which are already
212  * not present, or in the process of an address space destruction.
213  */
214 #ifndef __HAVE_ARCH_PTE_CLEAR_NOT_PRESENT_FULL
215 static inline void pte_clear_not_present_full(struct mm_struct *mm,
216 					      unsigned long address,
217 					      pte_t *ptep,
218 					      int full)
219 {
220 	pte_clear(mm, address, ptep);
221 }
222 #endif
223 
224 #ifndef __HAVE_ARCH_PTEP_CLEAR_FLUSH
225 extern pte_t ptep_clear_flush(struct vm_area_struct *vma,
226 			      unsigned long address,
227 			      pte_t *ptep);
228 #endif
229 
230 #ifndef __HAVE_ARCH_PMDP_HUGE_CLEAR_FLUSH
231 extern pmd_t pmdp_huge_clear_flush(struct vm_area_struct *vma,
232 			      unsigned long address,
233 			      pmd_t *pmdp);
234 extern pud_t pudp_huge_clear_flush(struct vm_area_struct *vma,
235 			      unsigned long address,
236 			      pud_t *pudp);
237 #endif
238 
239 #ifndef __HAVE_ARCH_PTEP_SET_WRPROTECT
240 struct mm_struct;
241 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long address, pte_t *ptep)
242 {
243 	pte_t old_pte = *ptep;
244 	set_pte_at(mm, address, ptep, pte_wrprotect(old_pte));
245 }
246 #endif
247 
248 /*
249  * On some architectures hardware does not set page access bit when accessing
250  * memory page, it is responsibilty of software setting this bit. It brings
251  * out extra page fault penalty to track page access bit. For optimization page
252  * access bit can be set during all page fault flow on these arches.
253  * To be differentiate with macro pte_mkyoung, this macro is used on platforms
254  * where software maintains page access bit.
255  */
256 #ifndef pte_sw_mkyoung
257 static inline pte_t pte_sw_mkyoung(pte_t pte)
258 {
259 	return pte;
260 }
261 #define pte_sw_mkyoung	pte_sw_mkyoung
262 #endif
263 
264 #ifndef pte_savedwrite
265 #define pte_savedwrite pte_write
266 #endif
267 
268 #ifndef pte_mk_savedwrite
269 #define pte_mk_savedwrite pte_mkwrite
270 #endif
271 
272 #ifndef pte_clear_savedwrite
273 #define pte_clear_savedwrite pte_wrprotect
274 #endif
275 
276 #ifndef pmd_savedwrite
277 #define pmd_savedwrite pmd_write
278 #endif
279 
280 #ifndef pmd_mk_savedwrite
281 #define pmd_mk_savedwrite pmd_mkwrite
282 #endif
283 
284 #ifndef pmd_clear_savedwrite
285 #define pmd_clear_savedwrite pmd_wrprotect
286 #endif
287 
288 #ifndef __HAVE_ARCH_PMDP_SET_WRPROTECT
289 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
290 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
291 				      unsigned long address, pmd_t *pmdp)
292 {
293 	pmd_t old_pmd = *pmdp;
294 	set_pmd_at(mm, address, pmdp, pmd_wrprotect(old_pmd));
295 }
296 #else
297 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
298 				      unsigned long address, pmd_t *pmdp)
299 {
300 	BUILD_BUG();
301 }
302 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
303 #endif
304 #ifndef __HAVE_ARCH_PUDP_SET_WRPROTECT
305 #ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
306 static inline void pudp_set_wrprotect(struct mm_struct *mm,
307 				      unsigned long address, pud_t *pudp)
308 {
309 	pud_t old_pud = *pudp;
310 
311 	set_pud_at(mm, address, pudp, pud_wrprotect(old_pud));
312 }
313 #else
314 static inline void pudp_set_wrprotect(struct mm_struct *mm,
315 				      unsigned long address, pud_t *pudp)
316 {
317 	BUILD_BUG();
318 }
319 #endif /* CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */
320 #endif
321 
322 #ifndef pmdp_collapse_flush
323 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
324 extern pmd_t pmdp_collapse_flush(struct vm_area_struct *vma,
325 				 unsigned long address, pmd_t *pmdp);
326 #else
327 static inline pmd_t pmdp_collapse_flush(struct vm_area_struct *vma,
328 					unsigned long address,
329 					pmd_t *pmdp)
330 {
331 	BUILD_BUG();
332 	return *pmdp;
333 }
334 #define pmdp_collapse_flush pmdp_collapse_flush
335 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
336 #endif
337 
338 #ifndef __HAVE_ARCH_PGTABLE_DEPOSIT
339 extern void pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp,
340 				       pgtable_t pgtable);
341 #endif
342 
343 #ifndef __HAVE_ARCH_PGTABLE_WITHDRAW
344 extern pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp);
345 #endif
346 
347 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
348 /*
349  * This is an implementation of pmdp_establish() that is only suitable for an
350  * architecture that doesn't have hardware dirty/accessed bits. In this case we
351  * can't race with CPU which sets these bits and non-atomic aproach is fine.
352  */
353 static inline pmd_t generic_pmdp_establish(struct vm_area_struct *vma,
354 		unsigned long address, pmd_t *pmdp, pmd_t pmd)
355 {
356 	pmd_t old_pmd = *pmdp;
357 	set_pmd_at(vma->vm_mm, address, pmdp, pmd);
358 	return old_pmd;
359 }
360 #endif
361 
362 #ifndef __HAVE_ARCH_PMDP_INVALIDATE
363 extern pmd_t pmdp_invalidate(struct vm_area_struct *vma, unsigned long address,
364 			    pmd_t *pmdp);
365 #endif
366 
367 #ifndef __HAVE_ARCH_PTE_SAME
368 static inline int pte_same(pte_t pte_a, pte_t pte_b)
369 {
370 	return pte_val(pte_a) == pte_val(pte_b);
371 }
372 #endif
373 
374 #ifndef __HAVE_ARCH_PTE_UNUSED
375 /*
376  * Some architectures provide facilities to virtualization guests
377  * so that they can flag allocated pages as unused. This allows the
378  * host to transparently reclaim unused pages. This function returns
379  * whether the pte's page is unused.
380  */
381 static inline int pte_unused(pte_t pte)
382 {
383 	return 0;
384 }
385 #endif
386 
387 #ifndef pte_access_permitted
388 #define pte_access_permitted(pte, write) \
389 	(pte_present(pte) && (!(write) || pte_write(pte)))
390 #endif
391 
392 #ifndef pmd_access_permitted
393 #define pmd_access_permitted(pmd, write) \
394 	(pmd_present(pmd) && (!(write) || pmd_write(pmd)))
395 #endif
396 
397 #ifndef pud_access_permitted
398 #define pud_access_permitted(pud, write) \
399 	(pud_present(pud) && (!(write) || pud_write(pud)))
400 #endif
401 
402 #ifndef p4d_access_permitted
403 #define p4d_access_permitted(p4d, write) \
404 	(p4d_present(p4d) && (!(write) || p4d_write(p4d)))
405 #endif
406 
407 #ifndef pgd_access_permitted
408 #define pgd_access_permitted(pgd, write) \
409 	(pgd_present(pgd) && (!(write) || pgd_write(pgd)))
410 #endif
411 
412 #ifndef __HAVE_ARCH_PMD_SAME
413 static inline int pmd_same(pmd_t pmd_a, pmd_t pmd_b)
414 {
415 	return pmd_val(pmd_a) == pmd_val(pmd_b);
416 }
417 
418 static inline int pud_same(pud_t pud_a, pud_t pud_b)
419 {
420 	return pud_val(pud_a) == pud_val(pud_b);
421 }
422 #endif
423 
424 #ifndef __HAVE_ARCH_P4D_SAME
425 static inline int p4d_same(p4d_t p4d_a, p4d_t p4d_b)
426 {
427 	return p4d_val(p4d_a) == p4d_val(p4d_b);
428 }
429 #endif
430 
431 #ifndef __HAVE_ARCH_PGD_SAME
432 static inline int pgd_same(pgd_t pgd_a, pgd_t pgd_b)
433 {
434 	return pgd_val(pgd_a) == pgd_val(pgd_b);
435 }
436 #endif
437 
438 /*
439  * Use set_p*_safe(), and elide TLB flushing, when confident that *no*
440  * TLB flush will be required as a result of the "set". For example, use
441  * in scenarios where it is known ahead of time that the routine is
442  * setting non-present entries, or re-setting an existing entry to the
443  * same value. Otherwise, use the typical "set" helpers and flush the
444  * TLB.
445  */
446 #define set_pte_safe(ptep, pte) \
447 ({ \
448 	WARN_ON_ONCE(pte_present(*ptep) && !pte_same(*ptep, pte)); \
449 	set_pte(ptep, pte); \
450 })
451 
452 #define set_pmd_safe(pmdp, pmd) \
453 ({ \
454 	WARN_ON_ONCE(pmd_present(*pmdp) && !pmd_same(*pmdp, pmd)); \
455 	set_pmd(pmdp, pmd); \
456 })
457 
458 #define set_pud_safe(pudp, pud) \
459 ({ \
460 	WARN_ON_ONCE(pud_present(*pudp) && !pud_same(*pudp, pud)); \
461 	set_pud(pudp, pud); \
462 })
463 
464 #define set_p4d_safe(p4dp, p4d) \
465 ({ \
466 	WARN_ON_ONCE(p4d_present(*p4dp) && !p4d_same(*p4dp, p4d)); \
467 	set_p4d(p4dp, p4d); \
468 })
469 
470 #define set_pgd_safe(pgdp, pgd) \
471 ({ \
472 	WARN_ON_ONCE(pgd_present(*pgdp) && !pgd_same(*pgdp, pgd)); \
473 	set_pgd(pgdp, pgd); \
474 })
475 
476 #ifndef __HAVE_ARCH_DO_SWAP_PAGE
477 /*
478  * Some architectures support metadata associated with a page. When a
479  * page is being swapped out, this metadata must be saved so it can be
480  * restored when the page is swapped back in. SPARC M7 and newer
481  * processors support an ADI (Application Data Integrity) tag for the
482  * page as metadata for the page. arch_do_swap_page() can restore this
483  * metadata when a page is swapped back in.
484  */
485 static inline void arch_do_swap_page(struct mm_struct *mm,
486 				     struct vm_area_struct *vma,
487 				     unsigned long addr,
488 				     pte_t pte, pte_t oldpte)
489 {
490 
491 }
492 #endif
493 
494 #ifndef __HAVE_ARCH_UNMAP_ONE
495 /*
496  * Some architectures support metadata associated with a page. When a
497  * page is being swapped out, this metadata must be saved so it can be
498  * restored when the page is swapped back in. SPARC M7 and newer
499  * processors support an ADI (Application Data Integrity) tag for the
500  * page as metadata for the page. arch_unmap_one() can save this
501  * metadata on a swap-out of a page.
502  */
503 static inline int arch_unmap_one(struct mm_struct *mm,
504 				  struct vm_area_struct *vma,
505 				  unsigned long addr,
506 				  pte_t orig_pte)
507 {
508 	return 0;
509 }
510 #endif
511 
512 #ifndef __HAVE_ARCH_PGD_OFFSET_GATE
513 #define pgd_offset_gate(mm, addr)	pgd_offset(mm, addr)
514 #endif
515 
516 #ifndef __HAVE_ARCH_MOVE_PTE
517 #define move_pte(pte, prot, old_addr, new_addr)	(pte)
518 #endif
519 
520 #ifndef pte_accessible
521 # define pte_accessible(mm, pte)	((void)(pte), 1)
522 #endif
523 
524 #ifndef flush_tlb_fix_spurious_fault
525 #define flush_tlb_fix_spurious_fault(vma, address) flush_tlb_page(vma, address)
526 #endif
527 
528 #ifndef pgprot_nx
529 #define pgprot_nx(prot)	(prot)
530 #endif
531 
532 #ifndef pgprot_noncached
533 #define pgprot_noncached(prot)	(prot)
534 #endif
535 
536 #ifndef pgprot_writecombine
537 #define pgprot_writecombine pgprot_noncached
538 #endif
539 
540 #ifndef pgprot_writethrough
541 #define pgprot_writethrough pgprot_noncached
542 #endif
543 
544 #ifndef pgprot_device
545 #define pgprot_device pgprot_noncached
546 #endif
547 
548 #ifndef pgprot_modify
549 #define pgprot_modify pgprot_modify
550 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
551 {
552 	if (pgprot_val(oldprot) == pgprot_val(pgprot_noncached(oldprot)))
553 		newprot = pgprot_noncached(newprot);
554 	if (pgprot_val(oldprot) == pgprot_val(pgprot_writecombine(oldprot)))
555 		newprot = pgprot_writecombine(newprot);
556 	if (pgprot_val(oldprot) == pgprot_val(pgprot_device(oldprot)))
557 		newprot = pgprot_device(newprot);
558 	return newprot;
559 }
560 #endif
561 
562 /*
563  * When walking page tables, get the address of the next boundary,
564  * or the end address of the range if that comes earlier.  Although no
565  * vma end wraps to 0, rounded up __boundary may wrap to 0 throughout.
566  */
567 
568 #define pgd_addr_end(addr, end)						\
569 ({	unsigned long __boundary = ((addr) + PGDIR_SIZE) & PGDIR_MASK;	\
570 	(__boundary - 1 < (end) - 1)? __boundary: (end);		\
571 })
572 
573 #ifndef p4d_addr_end
574 #define p4d_addr_end(addr, end)						\
575 ({	unsigned long __boundary = ((addr) + P4D_SIZE) & P4D_MASK;	\
576 	(__boundary - 1 < (end) - 1)? __boundary: (end);		\
577 })
578 #endif
579 
580 #ifndef pud_addr_end
581 #define pud_addr_end(addr, end)						\
582 ({	unsigned long __boundary = ((addr) + PUD_SIZE) & PUD_MASK;	\
583 	(__boundary - 1 < (end) - 1)? __boundary: (end);		\
584 })
585 #endif
586 
587 #ifndef pmd_addr_end
588 #define pmd_addr_end(addr, end)						\
589 ({	unsigned long __boundary = ((addr) + PMD_SIZE) & PMD_MASK;	\
590 	(__boundary - 1 < (end) - 1)? __boundary: (end);		\
591 })
592 #endif
593 
594 /*
595  * When walking page tables, we usually want to skip any p?d_none entries;
596  * and any p?d_bad entries - reporting the error before resetting to none.
597  * Do the tests inline, but report and clear the bad entry in mm/memory.c.
598  */
599 void pgd_clear_bad(pgd_t *);
600 
601 #ifndef __PAGETABLE_P4D_FOLDED
602 void p4d_clear_bad(p4d_t *);
603 #else
604 #define p4d_clear_bad(p4d)        do { } while (0)
605 #endif
606 
607 #ifndef __PAGETABLE_PUD_FOLDED
608 void pud_clear_bad(pud_t *);
609 #else
610 #define pud_clear_bad(p4d)        do { } while (0)
611 #endif
612 
613 void pmd_clear_bad(pmd_t *);
614 
615 static inline int pgd_none_or_clear_bad(pgd_t *pgd)
616 {
617 	if (pgd_none(*pgd))
618 		return 1;
619 	if (unlikely(pgd_bad(*pgd))) {
620 		pgd_clear_bad(pgd);
621 		return 1;
622 	}
623 	return 0;
624 }
625 
626 static inline int p4d_none_or_clear_bad(p4d_t *p4d)
627 {
628 	if (p4d_none(*p4d))
629 		return 1;
630 	if (unlikely(p4d_bad(*p4d))) {
631 		p4d_clear_bad(p4d);
632 		return 1;
633 	}
634 	return 0;
635 }
636 
637 static inline int pud_none_or_clear_bad(pud_t *pud)
638 {
639 	if (pud_none(*pud))
640 		return 1;
641 	if (unlikely(pud_bad(*pud))) {
642 		pud_clear_bad(pud);
643 		return 1;
644 	}
645 	return 0;
646 }
647 
648 static inline int pmd_none_or_clear_bad(pmd_t *pmd)
649 {
650 	if (pmd_none(*pmd))
651 		return 1;
652 	if (unlikely(pmd_bad(*pmd))) {
653 		pmd_clear_bad(pmd);
654 		return 1;
655 	}
656 	return 0;
657 }
658 
659 static inline pte_t __ptep_modify_prot_start(struct vm_area_struct *vma,
660 					     unsigned long addr,
661 					     pte_t *ptep)
662 {
663 	/*
664 	 * Get the current pte state, but zero it out to make it
665 	 * non-present, preventing the hardware from asynchronously
666 	 * updating it.
667 	 */
668 	return ptep_get_and_clear(vma->vm_mm, addr, ptep);
669 }
670 
671 static inline void __ptep_modify_prot_commit(struct vm_area_struct *vma,
672 					     unsigned long addr,
673 					     pte_t *ptep, pte_t pte)
674 {
675 	/*
676 	 * The pte is non-present, so there's no hardware state to
677 	 * preserve.
678 	 */
679 	set_pte_at(vma->vm_mm, addr, ptep, pte);
680 }
681 
682 #ifndef __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
683 /*
684  * Start a pte protection read-modify-write transaction, which
685  * protects against asynchronous hardware modifications to the pte.
686  * The intention is not to prevent the hardware from making pte
687  * updates, but to prevent any updates it may make from being lost.
688  *
689  * This does not protect against other software modifications of the
690  * pte; the appropriate pte lock must be held over the transation.
691  *
692  * Note that this interface is intended to be batchable, meaning that
693  * ptep_modify_prot_commit may not actually update the pte, but merely
694  * queue the update to be done at some later time.  The update must be
695  * actually committed before the pte lock is released, however.
696  */
697 static inline pte_t ptep_modify_prot_start(struct vm_area_struct *vma,
698 					   unsigned long addr,
699 					   pte_t *ptep)
700 {
701 	return __ptep_modify_prot_start(vma, addr, ptep);
702 }
703 
704 /*
705  * Commit an update to a pte, leaving any hardware-controlled bits in
706  * the PTE unmodified.
707  */
708 static inline void ptep_modify_prot_commit(struct vm_area_struct *vma,
709 					   unsigned long addr,
710 					   pte_t *ptep, pte_t old_pte, pte_t pte)
711 {
712 	__ptep_modify_prot_commit(vma, addr, ptep, pte);
713 }
714 #endif /* __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION */
715 #endif /* CONFIG_MMU */
716 
717 /*
718  * No-op macros that just return the current protection value. Defined here
719  * because these macros can be used used even if CONFIG_MMU is not defined.
720  */
721 #ifndef pgprot_encrypted
722 #define pgprot_encrypted(prot)	(prot)
723 #endif
724 
725 #ifndef pgprot_decrypted
726 #define pgprot_decrypted(prot)	(prot)
727 #endif
728 
729 /*
730  * A facility to provide lazy MMU batching.  This allows PTE updates and
731  * page invalidations to be delayed until a call to leave lazy MMU mode
732  * is issued.  Some architectures may benefit from doing this, and it is
733  * beneficial for both shadow and direct mode hypervisors, which may batch
734  * the PTE updates which happen during this window.  Note that using this
735  * interface requires that read hazards be removed from the code.  A read
736  * hazard could result in the direct mode hypervisor case, since the actual
737  * write to the page tables may not yet have taken place, so reads though
738  * a raw PTE pointer after it has been modified are not guaranteed to be
739  * up to date.  This mode can only be entered and left under the protection of
740  * the page table locks for all page tables which may be modified.  In the UP
741  * case, this is required so that preemption is disabled, and in the SMP case,
742  * it must synchronize the delayed page table writes properly on other CPUs.
743  */
744 #ifndef __HAVE_ARCH_ENTER_LAZY_MMU_MODE
745 #define arch_enter_lazy_mmu_mode()	do {} while (0)
746 #define arch_leave_lazy_mmu_mode()	do {} while (0)
747 #define arch_flush_lazy_mmu_mode()	do {} while (0)
748 #endif
749 
750 /*
751  * A facility to provide batching of the reload of page tables and
752  * other process state with the actual context switch code for
753  * paravirtualized guests.  By convention, only one of the batched
754  * update (lazy) modes (CPU, MMU) should be active at any given time,
755  * entry should never be nested, and entry and exits should always be
756  * paired.  This is for sanity of maintaining and reasoning about the
757  * kernel code.  In this case, the exit (end of the context switch) is
758  * in architecture-specific code, and so doesn't need a generic
759  * definition.
760  */
761 #ifndef __HAVE_ARCH_START_CONTEXT_SWITCH
762 #define arch_start_context_switch(prev)	do {} while (0)
763 #endif
764 
765 #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY
766 #ifndef CONFIG_ARCH_ENABLE_THP_MIGRATION
767 static inline pmd_t pmd_swp_mksoft_dirty(pmd_t pmd)
768 {
769 	return pmd;
770 }
771 
772 static inline int pmd_swp_soft_dirty(pmd_t pmd)
773 {
774 	return 0;
775 }
776 
777 static inline pmd_t pmd_swp_clear_soft_dirty(pmd_t pmd)
778 {
779 	return pmd;
780 }
781 #endif
782 #else /* !CONFIG_HAVE_ARCH_SOFT_DIRTY */
783 static inline int pte_soft_dirty(pte_t pte)
784 {
785 	return 0;
786 }
787 
788 static inline int pmd_soft_dirty(pmd_t pmd)
789 {
790 	return 0;
791 }
792 
793 static inline pte_t pte_mksoft_dirty(pte_t pte)
794 {
795 	return pte;
796 }
797 
798 static inline pmd_t pmd_mksoft_dirty(pmd_t pmd)
799 {
800 	return pmd;
801 }
802 
803 static inline pte_t pte_clear_soft_dirty(pte_t pte)
804 {
805 	return pte;
806 }
807 
808 static inline pmd_t pmd_clear_soft_dirty(pmd_t pmd)
809 {
810 	return pmd;
811 }
812 
813 static inline pte_t pte_swp_mksoft_dirty(pte_t pte)
814 {
815 	return pte;
816 }
817 
818 static inline int pte_swp_soft_dirty(pte_t pte)
819 {
820 	return 0;
821 }
822 
823 static inline pte_t pte_swp_clear_soft_dirty(pte_t pte)
824 {
825 	return pte;
826 }
827 
828 static inline pmd_t pmd_swp_mksoft_dirty(pmd_t pmd)
829 {
830 	return pmd;
831 }
832 
833 static inline int pmd_swp_soft_dirty(pmd_t pmd)
834 {
835 	return 0;
836 }
837 
838 static inline pmd_t pmd_swp_clear_soft_dirty(pmd_t pmd)
839 {
840 	return pmd;
841 }
842 #endif
843 
844 #ifndef __HAVE_PFNMAP_TRACKING
845 /*
846  * Interfaces that can be used by architecture code to keep track of
847  * memory type of pfn mappings specified by the remap_pfn_range,
848  * vmf_insert_pfn.
849  */
850 
851 /*
852  * track_pfn_remap is called when a _new_ pfn mapping is being established
853  * by remap_pfn_range() for physical range indicated by pfn and size.
854  */
855 static inline int track_pfn_remap(struct vm_area_struct *vma, pgprot_t *prot,
856 				  unsigned long pfn, unsigned long addr,
857 				  unsigned long size)
858 {
859 	return 0;
860 }
861 
862 /*
863  * track_pfn_insert is called when a _new_ single pfn is established
864  * by vmf_insert_pfn().
865  */
866 static inline void track_pfn_insert(struct vm_area_struct *vma, pgprot_t *prot,
867 				    pfn_t pfn)
868 {
869 }
870 
871 /*
872  * track_pfn_copy is called when vma that is covering the pfnmap gets
873  * copied through copy_page_range().
874  */
875 static inline int track_pfn_copy(struct vm_area_struct *vma)
876 {
877 	return 0;
878 }
879 
880 /*
881  * untrack_pfn is called while unmapping a pfnmap for a region.
882  * untrack can be called for a specific region indicated by pfn and size or
883  * can be for the entire vma (in which case pfn, size are zero).
884  */
885 static inline void untrack_pfn(struct vm_area_struct *vma,
886 			       unsigned long pfn, unsigned long size)
887 {
888 }
889 
890 /*
891  * untrack_pfn_moved is called while mremapping a pfnmap for a new region.
892  */
893 static inline void untrack_pfn_moved(struct vm_area_struct *vma)
894 {
895 }
896 #else
897 extern int track_pfn_remap(struct vm_area_struct *vma, pgprot_t *prot,
898 			   unsigned long pfn, unsigned long addr,
899 			   unsigned long size);
900 extern void track_pfn_insert(struct vm_area_struct *vma, pgprot_t *prot,
901 			     pfn_t pfn);
902 extern int track_pfn_copy(struct vm_area_struct *vma);
903 extern void untrack_pfn(struct vm_area_struct *vma, unsigned long pfn,
904 			unsigned long size);
905 extern void untrack_pfn_moved(struct vm_area_struct *vma);
906 #endif
907 
908 #ifdef __HAVE_COLOR_ZERO_PAGE
909 static inline int is_zero_pfn(unsigned long pfn)
910 {
911 	extern unsigned long zero_pfn;
912 	unsigned long offset_from_zero_pfn = pfn - zero_pfn;
913 	return offset_from_zero_pfn <= (zero_page_mask >> PAGE_SHIFT);
914 }
915 
916 #define my_zero_pfn(addr)	page_to_pfn(ZERO_PAGE(addr))
917 
918 #else
919 static inline int is_zero_pfn(unsigned long pfn)
920 {
921 	extern unsigned long zero_pfn;
922 	return pfn == zero_pfn;
923 }
924 
925 static inline unsigned long my_zero_pfn(unsigned long addr)
926 {
927 	extern unsigned long zero_pfn;
928 	return zero_pfn;
929 }
930 #endif
931 
932 #ifdef CONFIG_MMU
933 
934 #ifndef CONFIG_TRANSPARENT_HUGEPAGE
935 static inline int pmd_trans_huge(pmd_t pmd)
936 {
937 	return 0;
938 }
939 #ifndef pmd_write
940 static inline int pmd_write(pmd_t pmd)
941 {
942 	BUG();
943 	return 0;
944 }
945 #endif /* pmd_write */
946 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
947 
948 #ifndef pud_write
949 static inline int pud_write(pud_t pud)
950 {
951 	BUG();
952 	return 0;
953 }
954 #endif /* pud_write */
955 
956 #if !defined(CONFIG_ARCH_HAS_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
957 static inline int pmd_devmap(pmd_t pmd)
958 {
959 	return 0;
960 }
961 static inline int pud_devmap(pud_t pud)
962 {
963 	return 0;
964 }
965 static inline int pgd_devmap(pgd_t pgd)
966 {
967 	return 0;
968 }
969 #endif
970 
971 #if !defined(CONFIG_TRANSPARENT_HUGEPAGE) || \
972 	(defined(CONFIG_TRANSPARENT_HUGEPAGE) && \
973 	 !defined(CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD))
974 static inline int pud_trans_huge(pud_t pud)
975 {
976 	return 0;
977 }
978 #endif
979 
980 /* See pmd_none_or_trans_huge_or_clear_bad for discussion. */
981 static inline int pud_none_or_trans_huge_or_dev_or_clear_bad(pud_t *pud)
982 {
983 	pud_t pudval = READ_ONCE(*pud);
984 
985 	if (pud_none(pudval) || pud_trans_huge(pudval) || pud_devmap(pudval))
986 		return 1;
987 	if (unlikely(pud_bad(pudval))) {
988 		pud_clear_bad(pud);
989 		return 1;
990 	}
991 	return 0;
992 }
993 
994 /* See pmd_trans_unstable for discussion. */
995 static inline int pud_trans_unstable(pud_t *pud)
996 {
997 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) &&			\
998 	defined(CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD)
999 	return pud_none_or_trans_huge_or_dev_or_clear_bad(pud);
1000 #else
1001 	return 0;
1002 #endif
1003 }
1004 
1005 #ifndef pmd_read_atomic
1006 static inline pmd_t pmd_read_atomic(pmd_t *pmdp)
1007 {
1008 	/*
1009 	 * Depend on compiler for an atomic pmd read. NOTE: this is
1010 	 * only going to work, if the pmdval_t isn't larger than
1011 	 * an unsigned long.
1012 	 */
1013 	return *pmdp;
1014 }
1015 #endif
1016 
1017 #ifndef arch_needs_pgtable_deposit
1018 #define arch_needs_pgtable_deposit() (false)
1019 #endif
1020 /*
1021  * This function is meant to be used by sites walking pagetables with
1022  * the mmap_sem hold in read mode to protect against MADV_DONTNEED and
1023  * transhuge page faults. MADV_DONTNEED can convert a transhuge pmd
1024  * into a null pmd and the transhuge page fault can convert a null pmd
1025  * into an hugepmd or into a regular pmd (if the hugepage allocation
1026  * fails). While holding the mmap_sem in read mode the pmd becomes
1027  * stable and stops changing under us only if it's not null and not a
1028  * transhuge pmd. When those races occurs and this function makes a
1029  * difference vs the standard pmd_none_or_clear_bad, the result is
1030  * undefined so behaving like if the pmd was none is safe (because it
1031  * can return none anyway). The compiler level barrier() is critically
1032  * important to compute the two checks atomically on the same pmdval.
1033  *
1034  * For 32bit kernels with a 64bit large pmd_t this automatically takes
1035  * care of reading the pmd atomically to avoid SMP race conditions
1036  * against pmd_populate() when the mmap_sem is hold for reading by the
1037  * caller (a special atomic read not done by "gcc" as in the generic
1038  * version above, is also needed when THP is disabled because the page
1039  * fault can populate the pmd from under us).
1040  */
1041 static inline int pmd_none_or_trans_huge_or_clear_bad(pmd_t *pmd)
1042 {
1043 	pmd_t pmdval = pmd_read_atomic(pmd);
1044 	/*
1045 	 * The barrier will stabilize the pmdval in a register or on
1046 	 * the stack so that it will stop changing under the code.
1047 	 *
1048 	 * When CONFIG_TRANSPARENT_HUGEPAGE=y on x86 32bit PAE,
1049 	 * pmd_read_atomic is allowed to return a not atomic pmdval
1050 	 * (for example pointing to an hugepage that has never been
1051 	 * mapped in the pmd). The below checks will only care about
1052 	 * the low part of the pmd with 32bit PAE x86 anyway, with the
1053 	 * exception of pmd_none(). So the important thing is that if
1054 	 * the low part of the pmd is found null, the high part will
1055 	 * be also null or the pmd_none() check below would be
1056 	 * confused.
1057 	 */
1058 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1059 	barrier();
1060 #endif
1061 	/*
1062 	 * !pmd_present() checks for pmd migration entries
1063 	 *
1064 	 * The complete check uses is_pmd_migration_entry() in linux/swapops.h
1065 	 * But using that requires moving current function and pmd_trans_unstable()
1066 	 * to linux/swapops.h to resovle dependency, which is too much code move.
1067 	 *
1068 	 * !pmd_present() is equivalent to is_pmd_migration_entry() currently,
1069 	 * because !pmd_present() pages can only be under migration not swapped
1070 	 * out.
1071 	 *
1072 	 * pmd_none() is preseved for future condition checks on pmd migration
1073 	 * entries and not confusing with this function name, although it is
1074 	 * redundant with !pmd_present().
1075 	 */
1076 	if (pmd_none(pmdval) || pmd_trans_huge(pmdval) ||
1077 		(IS_ENABLED(CONFIG_ARCH_ENABLE_THP_MIGRATION) && !pmd_present(pmdval)))
1078 		return 1;
1079 	if (unlikely(pmd_bad(pmdval))) {
1080 		pmd_clear_bad(pmd);
1081 		return 1;
1082 	}
1083 	return 0;
1084 }
1085 
1086 /*
1087  * This is a noop if Transparent Hugepage Support is not built into
1088  * the kernel. Otherwise it is equivalent to
1089  * pmd_none_or_trans_huge_or_clear_bad(), and shall only be called in
1090  * places that already verified the pmd is not none and they want to
1091  * walk ptes while holding the mmap sem in read mode (write mode don't
1092  * need this). If THP is not enabled, the pmd can't go away under the
1093  * code even if MADV_DONTNEED runs, but if THP is enabled we need to
1094  * run a pmd_trans_unstable before walking the ptes after
1095  * split_huge_pmd returns (because it may have run when the pmd become
1096  * null, but then a page fault can map in a THP and not a regular page).
1097  */
1098 static inline int pmd_trans_unstable(pmd_t *pmd)
1099 {
1100 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1101 	return pmd_none_or_trans_huge_or_clear_bad(pmd);
1102 #else
1103 	return 0;
1104 #endif
1105 }
1106 
1107 #ifndef CONFIG_NUMA_BALANCING
1108 /*
1109  * Technically a PTE can be PROTNONE even when not doing NUMA balancing but
1110  * the only case the kernel cares is for NUMA balancing and is only ever set
1111  * when the VMA is accessible. For PROT_NONE VMAs, the PTEs are not marked
1112  * _PAGE_PROTNONE so by by default, implement the helper as "always no". It
1113  * is the responsibility of the caller to distinguish between PROT_NONE
1114  * protections and NUMA hinting fault protections.
1115  */
1116 static inline int pte_protnone(pte_t pte)
1117 {
1118 	return 0;
1119 }
1120 
1121 static inline int pmd_protnone(pmd_t pmd)
1122 {
1123 	return 0;
1124 }
1125 #endif /* CONFIG_NUMA_BALANCING */
1126 
1127 #endif /* CONFIG_MMU */
1128 
1129 #ifdef CONFIG_HAVE_ARCH_HUGE_VMAP
1130 
1131 #ifndef __PAGETABLE_P4D_FOLDED
1132 int p4d_set_huge(p4d_t *p4d, phys_addr_t addr, pgprot_t prot);
1133 int p4d_clear_huge(p4d_t *p4d);
1134 #else
1135 static inline int p4d_set_huge(p4d_t *p4d, phys_addr_t addr, pgprot_t prot)
1136 {
1137 	return 0;
1138 }
1139 static inline int p4d_clear_huge(p4d_t *p4d)
1140 {
1141 	return 0;
1142 }
1143 #endif /* !__PAGETABLE_P4D_FOLDED */
1144 
1145 int pud_set_huge(pud_t *pud, phys_addr_t addr, pgprot_t prot);
1146 int pmd_set_huge(pmd_t *pmd, phys_addr_t addr, pgprot_t prot);
1147 int pud_clear_huge(pud_t *pud);
1148 int pmd_clear_huge(pmd_t *pmd);
1149 int p4d_free_pud_page(p4d_t *p4d, unsigned long addr);
1150 int pud_free_pmd_page(pud_t *pud, unsigned long addr);
1151 int pmd_free_pte_page(pmd_t *pmd, unsigned long addr);
1152 #else	/* !CONFIG_HAVE_ARCH_HUGE_VMAP */
1153 static inline int p4d_set_huge(p4d_t *p4d, phys_addr_t addr, pgprot_t prot)
1154 {
1155 	return 0;
1156 }
1157 static inline int pud_set_huge(pud_t *pud, phys_addr_t addr, pgprot_t prot)
1158 {
1159 	return 0;
1160 }
1161 static inline int pmd_set_huge(pmd_t *pmd, phys_addr_t addr, pgprot_t prot)
1162 {
1163 	return 0;
1164 }
1165 static inline int p4d_clear_huge(p4d_t *p4d)
1166 {
1167 	return 0;
1168 }
1169 static inline int pud_clear_huge(pud_t *pud)
1170 {
1171 	return 0;
1172 }
1173 static inline int pmd_clear_huge(pmd_t *pmd)
1174 {
1175 	return 0;
1176 }
1177 static inline int p4d_free_pud_page(p4d_t *p4d, unsigned long addr)
1178 {
1179 	return 0;
1180 }
1181 static inline int pud_free_pmd_page(pud_t *pud, unsigned long addr)
1182 {
1183 	return 0;
1184 }
1185 static inline int pmd_free_pte_page(pmd_t *pmd, unsigned long addr)
1186 {
1187 	return 0;
1188 }
1189 #endif	/* CONFIG_HAVE_ARCH_HUGE_VMAP */
1190 
1191 #ifndef __HAVE_ARCH_FLUSH_PMD_TLB_RANGE
1192 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1193 /*
1194  * ARCHes with special requirements for evicting THP backing TLB entries can
1195  * implement this. Otherwise also, it can help optimize normal TLB flush in
1196  * THP regime. stock flush_tlb_range() typically has optimization to nuke the
1197  * entire TLB TLB if flush span is greater than a threshold, which will
1198  * likely be true for a single huge page. Thus a single thp flush will
1199  * invalidate the entire TLB which is not desitable.
1200  * e.g. see arch/arc: flush_pmd_tlb_range
1201  */
1202 #define flush_pmd_tlb_range(vma, addr, end)	flush_tlb_range(vma, addr, end)
1203 #define flush_pud_tlb_range(vma, addr, end)	flush_tlb_range(vma, addr, end)
1204 #else
1205 #define flush_pmd_tlb_range(vma, addr, end)	BUILD_BUG()
1206 #define flush_pud_tlb_range(vma, addr, end)	BUILD_BUG()
1207 #endif
1208 #endif
1209 
1210 struct file;
1211 int phys_mem_access_prot_allowed(struct file *file, unsigned long pfn,
1212 			unsigned long size, pgprot_t *vma_prot);
1213 
1214 #ifndef CONFIG_X86_ESPFIX64
1215 static inline void init_espfix_bsp(void) { }
1216 #endif
1217 
1218 extern void __init pgtable_cache_init(void);
1219 
1220 #ifndef __HAVE_ARCH_PFN_MODIFY_ALLOWED
1221 static inline bool pfn_modify_allowed(unsigned long pfn, pgprot_t prot)
1222 {
1223 	return true;
1224 }
1225 
1226 static inline bool arch_has_pfn_modify_check(void)
1227 {
1228 	return false;
1229 }
1230 #endif /* !_HAVE_ARCH_PFN_MODIFY_ALLOWED */
1231 
1232 /*
1233  * Architecture PAGE_KERNEL_* fallbacks
1234  *
1235  * Some architectures don't define certain PAGE_KERNEL_* flags. This is either
1236  * because they really don't support them, or the port needs to be updated to
1237  * reflect the required functionality. Below are a set of relatively safe
1238  * fallbacks, as best effort, which we can count on in lieu of the architectures
1239  * not defining them on their own yet.
1240  */
1241 
1242 #ifndef PAGE_KERNEL_RO
1243 # define PAGE_KERNEL_RO PAGE_KERNEL
1244 #endif
1245 
1246 #ifndef PAGE_KERNEL_EXEC
1247 # define PAGE_KERNEL_EXEC PAGE_KERNEL
1248 #endif
1249 
1250 /*
1251  * Page Table Modification bits for pgtbl_mod_mask.
1252  *
1253  * These are used by the p?d_alloc_track*() set of functions an in the generic
1254  * vmalloc/ioremap code to track at which page-table levels entries have been
1255  * modified. Based on that the code can better decide when vmalloc and ioremap
1256  * mapping changes need to be synchronized to other page-tables in the system.
1257  */
1258 #define		__PGTBL_PGD_MODIFIED	0
1259 #define		__PGTBL_P4D_MODIFIED	1
1260 #define		__PGTBL_PUD_MODIFIED	2
1261 #define		__PGTBL_PMD_MODIFIED	3
1262 #define		__PGTBL_PTE_MODIFIED	4
1263 
1264 #define		PGTBL_PGD_MODIFIED	BIT(__PGTBL_PGD_MODIFIED)
1265 #define		PGTBL_P4D_MODIFIED	BIT(__PGTBL_P4D_MODIFIED)
1266 #define		PGTBL_PUD_MODIFIED	BIT(__PGTBL_PUD_MODIFIED)
1267 #define		PGTBL_PMD_MODIFIED	BIT(__PGTBL_PMD_MODIFIED)
1268 #define		PGTBL_PTE_MODIFIED	BIT(__PGTBL_PTE_MODIFIED)
1269 
1270 /* Page-Table Modification Mask */
1271 typedef unsigned int pgtbl_mod_mask;
1272 
1273 #endif /* !__ASSEMBLY__ */
1274 
1275 #ifndef io_remap_pfn_range
1276 #define io_remap_pfn_range remap_pfn_range
1277 #endif
1278 
1279 #ifndef has_transparent_hugepage
1280 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1281 #define has_transparent_hugepage() 1
1282 #else
1283 #define has_transparent_hugepage() 0
1284 #endif
1285 #endif
1286 
1287 /*
1288  * On some architectures it depends on the mm if the p4d/pud or pmd
1289  * layer of the page table hierarchy is folded or not.
1290  */
1291 #ifndef mm_p4d_folded
1292 #define mm_p4d_folded(mm)	__is_defined(__PAGETABLE_P4D_FOLDED)
1293 #endif
1294 
1295 #ifndef mm_pud_folded
1296 #define mm_pud_folded(mm)	__is_defined(__PAGETABLE_PUD_FOLDED)
1297 #endif
1298 
1299 #ifndef mm_pmd_folded
1300 #define mm_pmd_folded(mm)	__is_defined(__PAGETABLE_PMD_FOLDED)
1301 #endif
1302 
1303 /*
1304  * p?d_leaf() - true if this entry is a final mapping to a physical address.
1305  * This differs from p?d_huge() by the fact that they are always available (if
1306  * the architecture supports large pages at the appropriate level) even
1307  * if CONFIG_HUGETLB_PAGE is not defined.
1308  * Only meaningful when called on a valid entry.
1309  */
1310 #ifndef pgd_leaf
1311 #define pgd_leaf(x)	0
1312 #endif
1313 #ifndef p4d_leaf
1314 #define p4d_leaf(x)	0
1315 #endif
1316 #ifndef pud_leaf
1317 #define pud_leaf(x)	0
1318 #endif
1319 #ifndef pmd_leaf
1320 #define pmd_leaf(x)	0
1321 #endif
1322 
1323 #endif /* _LINUX_PGTABLE_H */
1324