]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | #ifndef _ASM_GENERIC_PGTABLE_H |
2 | #define _ASM_GENERIC_PGTABLE_H | |
3 | ||
f25748e3 DW |
4 | #include <linux/pfn.h> |
5 | ||
673eae82 | 6 | #ifndef __ASSEMBLY__ |
9535239f | 7 | #ifdef CONFIG_MMU |
673eae82 | 8 | |
fbd71844 | 9 | #include <linux/mm_types.h> |
187f1882 | 10 | #include <linux/bug.h> |
e61ce6ad | 11 | #include <linux/errno.h> |
fbd71844 | 12 | |
c2febafc KS |
13 | #if 5 - defined(__PAGETABLE_P4D_FOLDED) - defined(__PAGETABLE_PUD_FOLDED) - \ |
14 | defined(__PAGETABLE_PMD_FOLDED) != CONFIG_PGTABLE_LEVELS | |
15 | #error CONFIG_PGTABLE_LEVELS is not consistent with __PAGETABLE_{P4D,PUD,PMD}_FOLDED | |
235a8f02 KS |
16 | #endif |
17 | ||
6ee8630e HD |
18 | /* |
19 | * On almost all architectures and configurations, 0 can be used as the | |
20 | * upper ceiling to free_pgtables(): on many architectures it has the same | |
21 | * effect as using TASK_SIZE. However, there is one configuration which | |
22 | * must impose a more careful limit, to avoid freeing kernel pgtables. | |
23 | */ | |
24 | #ifndef USER_PGTABLES_CEILING | |
25 | #define USER_PGTABLES_CEILING 0UL | |
26 | #endif | |
27 | ||
1da177e4 | 28 | #ifndef __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS |
e2cda322 AA |
29 | extern int ptep_set_access_flags(struct vm_area_struct *vma, |
30 | unsigned long address, pte_t *ptep, | |
31 | pte_t entry, int dirty); | |
32 | #endif | |
33 | ||
34 | #ifndef __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS | |
bd5e88ad | 35 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
e2cda322 AA |
36 | extern int pmdp_set_access_flags(struct vm_area_struct *vma, |
37 | unsigned long address, pmd_t *pmdp, | |
38 | pmd_t entry, int dirty); | |
a00cc7d9 MW |
39 | extern int pudp_set_access_flags(struct vm_area_struct *vma, |
40 | unsigned long address, pud_t *pudp, | |
41 | pud_t entry, int dirty); | |
bd5e88ad VG |
42 | #else |
43 | static inline int pmdp_set_access_flags(struct vm_area_struct *vma, | |
44 | unsigned long address, pmd_t *pmdp, | |
45 | pmd_t entry, int dirty) | |
46 | { | |
47 | BUILD_BUG(); | |
48 | return 0; | |
49 | } | |
a00cc7d9 MW |
50 | static inline int pudp_set_access_flags(struct vm_area_struct *vma, |
51 | unsigned long address, pud_t *pudp, | |
52 | pud_t entry, int dirty) | |
53 | { | |
54 | BUILD_BUG(); | |
55 | return 0; | |
56 | } | |
bd5e88ad | 57 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE */ |
1da177e4 LT |
58 | #endif |
59 | ||
60 | #ifndef __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG | |
e2cda322 AA |
61 | static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, |
62 | unsigned long address, | |
63 | pte_t *ptep) | |
64 | { | |
65 | pte_t pte = *ptep; | |
66 | int r = 1; | |
67 | if (!pte_young(pte)) | |
68 | r = 0; | |
69 | else | |
70 | set_pte_at(vma->vm_mm, address, ptep, pte_mkold(pte)); | |
71 | return r; | |
72 | } | |
73 | #endif | |
74 | ||
75 | #ifndef __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG | |
76 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
77 | static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma, | |
78 | unsigned long address, | |
79 | pmd_t *pmdp) | |
80 | { | |
81 | pmd_t pmd = *pmdp; | |
82 | int r = 1; | |
83 | if (!pmd_young(pmd)) | |
84 | r = 0; | |
85 | else | |
86 | set_pmd_at(vma->vm_mm, address, pmdp, pmd_mkold(pmd)); | |
87 | return r; | |
88 | } | |
bd5e88ad | 89 | #else |
e2cda322 AA |
90 | static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma, |
91 | unsigned long address, | |
92 | pmd_t *pmdp) | |
93 | { | |
bd5e88ad | 94 | BUILD_BUG(); |
e2cda322 AA |
95 | return 0; |
96 | } | |
97 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE */ | |
1da177e4 LT |
98 | #endif |
99 | ||
100 | #ifndef __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH | |
e2cda322 AA |
101 | int ptep_clear_flush_young(struct vm_area_struct *vma, |
102 | unsigned long address, pte_t *ptep); | |
103 | #endif | |
104 | ||
105 | #ifndef __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH | |
bd5e88ad VG |
106 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
107 | extern int pmdp_clear_flush_young(struct vm_area_struct *vma, | |
108 | unsigned long address, pmd_t *pmdp); | |
109 | #else | |
110 | /* | |
111 | * Despite relevant to THP only, this API is called from generic rmap code | |
112 | * under PageTransHuge(), hence needs a dummy implementation for !THP | |
113 | */ | |
114 | static inline int pmdp_clear_flush_young(struct vm_area_struct *vma, | |
115 | unsigned long address, pmd_t *pmdp) | |
116 | { | |
117 | BUILD_BUG(); | |
118 | return 0; | |
119 | } | |
120 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE */ | |
1da177e4 LT |
121 | #endif |
122 | ||
1da177e4 | 123 | #ifndef __HAVE_ARCH_PTEP_GET_AND_CLEAR |
e2cda322 AA |
124 | static inline pte_t ptep_get_and_clear(struct mm_struct *mm, |
125 | unsigned long address, | |
126 | pte_t *ptep) | |
127 | { | |
128 | pte_t pte = *ptep; | |
129 | pte_clear(mm, address, ptep); | |
130 | return pte; | |
131 | } | |
132 | #endif | |
133 | ||
e2cda322 | 134 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
a00cc7d9 | 135 | #ifndef __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR |
8809aa2d AK |
136 | static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm, |
137 | unsigned long address, | |
138 | pmd_t *pmdp) | |
e2cda322 AA |
139 | { |
140 | pmd_t pmd = *pmdp; | |
2d28a227 | 141 | pmd_clear(pmdp); |
e2cda322 | 142 | return pmd; |
49b24d6b | 143 | } |
a00cc7d9 MW |
144 | #endif /* __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR */ |
145 | #ifndef __HAVE_ARCH_PUDP_HUGE_GET_AND_CLEAR | |
146 | static inline pud_t pudp_huge_get_and_clear(struct mm_struct *mm, | |
147 | unsigned long address, | |
148 | pud_t *pudp) | |
149 | { | |
150 | pud_t pud = *pudp; | |
151 | ||
152 | pud_clear(pudp); | |
153 | return pud; | |
154 | } | |
155 | #endif /* __HAVE_ARCH_PUDP_HUGE_GET_AND_CLEAR */ | |
e2cda322 | 156 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE */ |
1da177e4 | 157 | |
fcbe08d6 | 158 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
a00cc7d9 | 159 | #ifndef __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR_FULL |
8809aa2d | 160 | static inline pmd_t pmdp_huge_get_and_clear_full(struct mm_struct *mm, |
fcbe08d6 MS |
161 | unsigned long address, pmd_t *pmdp, |
162 | int full) | |
163 | { | |
8809aa2d | 164 | return pmdp_huge_get_and_clear(mm, address, pmdp); |
fcbe08d6 | 165 | } |
fcbe08d6 MS |
166 | #endif |
167 | ||
a00cc7d9 MW |
168 | #ifndef __HAVE_ARCH_PUDP_HUGE_GET_AND_CLEAR_FULL |
169 | static inline pud_t pudp_huge_get_and_clear_full(struct mm_struct *mm, | |
170 | unsigned long address, pud_t *pudp, | |
171 | int full) | |
172 | { | |
173 | return pudp_huge_get_and_clear(mm, address, pudp); | |
174 | } | |
175 | #endif | |
176 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE */ | |
177 | ||
a600388d | 178 | #ifndef __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL |
e2cda322 AA |
179 | static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm, |
180 | unsigned long address, pte_t *ptep, | |
181 | int full) | |
182 | { | |
183 | pte_t pte; | |
184 | pte = ptep_get_and_clear(mm, address, ptep); | |
185 | return pte; | |
186 | } | |
a600388d ZA |
187 | #endif |
188 | ||
9888a1ca ZA |
189 | /* |
190 | * Some architectures may be able to avoid expensive synchronization | |
191 | * primitives when modifications are made to PTE's which are already | |
192 | * not present, or in the process of an address space destruction. | |
193 | */ | |
194 | #ifndef __HAVE_ARCH_PTE_CLEAR_NOT_PRESENT_FULL | |
e2cda322 AA |
195 | static inline void pte_clear_not_present_full(struct mm_struct *mm, |
196 | unsigned long address, | |
197 | pte_t *ptep, | |
198 | int full) | |
199 | { | |
200 | pte_clear(mm, address, ptep); | |
201 | } | |
a600388d ZA |
202 | #endif |
203 | ||
1da177e4 | 204 | #ifndef __HAVE_ARCH_PTEP_CLEAR_FLUSH |
e2cda322 AA |
205 | extern pte_t ptep_clear_flush(struct vm_area_struct *vma, |
206 | unsigned long address, | |
207 | pte_t *ptep); | |
208 | #endif | |
209 | ||
8809aa2d AK |
210 | #ifndef __HAVE_ARCH_PMDP_HUGE_CLEAR_FLUSH |
211 | extern pmd_t pmdp_huge_clear_flush(struct vm_area_struct *vma, | |
e2cda322 AA |
212 | unsigned long address, |
213 | pmd_t *pmdp); | |
a00cc7d9 MW |
214 | extern pud_t pudp_huge_clear_flush(struct vm_area_struct *vma, |
215 | unsigned long address, | |
216 | pud_t *pudp); | |
1da177e4 LT |
217 | #endif |
218 | ||
219 | #ifndef __HAVE_ARCH_PTEP_SET_WRPROTECT | |
8c65b4a6 | 220 | struct mm_struct; |
1da177e4 LT |
221 | static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long address, pte_t *ptep) |
222 | { | |
223 | pte_t old_pte = *ptep; | |
224 | set_pte_at(mm, address, ptep, pte_wrprotect(old_pte)); | |
225 | } | |
226 | #endif | |
227 | ||
288bc549 AK |
228 | #ifndef pte_savedwrite |
229 | #define pte_savedwrite pte_write | |
230 | #endif | |
231 | ||
232 | #ifndef pte_mk_savedwrite | |
233 | #define pte_mk_savedwrite pte_mkwrite | |
234 | #endif | |
235 | ||
595cd8f2 AK |
236 | #ifndef pte_clear_savedwrite |
237 | #define pte_clear_savedwrite pte_wrprotect | |
238 | #endif | |
239 | ||
288bc549 AK |
240 | #ifndef pmd_savedwrite |
241 | #define pmd_savedwrite pmd_write | |
242 | #endif | |
243 | ||
244 | #ifndef pmd_mk_savedwrite | |
245 | #define pmd_mk_savedwrite pmd_mkwrite | |
246 | #endif | |
247 | ||
595cd8f2 AK |
248 | #ifndef pmd_clear_savedwrite |
249 | #define pmd_clear_savedwrite pmd_wrprotect | |
250 | #endif | |
251 | ||
e2cda322 AA |
252 | #ifndef __HAVE_ARCH_PMDP_SET_WRPROTECT |
253 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
254 | static inline void pmdp_set_wrprotect(struct mm_struct *mm, | |
255 | unsigned long address, pmd_t *pmdp) | |
256 | { | |
257 | pmd_t old_pmd = *pmdp; | |
258 | set_pmd_at(mm, address, pmdp, pmd_wrprotect(old_pmd)); | |
259 | } | |
bd5e88ad | 260 | #else |
e2cda322 AA |
261 | static inline void pmdp_set_wrprotect(struct mm_struct *mm, |
262 | unsigned long address, pmd_t *pmdp) | |
263 | { | |
bd5e88ad | 264 | BUILD_BUG(); |
e2cda322 AA |
265 | } |
266 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE */ | |
267 | #endif | |
a00cc7d9 MW |
268 | #ifndef __HAVE_ARCH_PUDP_SET_WRPROTECT |
269 | #ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD | |
270 | static inline void pudp_set_wrprotect(struct mm_struct *mm, | |
271 | unsigned long address, pud_t *pudp) | |
272 | { | |
273 | pud_t old_pud = *pudp; | |
274 | ||
275 | set_pud_at(mm, address, pudp, pud_wrprotect(old_pud)); | |
276 | } | |
277 | #else | |
278 | static inline void pudp_set_wrprotect(struct mm_struct *mm, | |
279 | unsigned long address, pud_t *pudp) | |
280 | { | |
281 | BUILD_BUG(); | |
282 | } | |
283 | #endif /* CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */ | |
284 | #endif | |
e2cda322 | 285 | |
15a25b2e AK |
286 | #ifndef pmdp_collapse_flush |
287 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
f28b6ff8 AK |
288 | extern pmd_t pmdp_collapse_flush(struct vm_area_struct *vma, |
289 | unsigned long address, pmd_t *pmdp); | |
15a25b2e AK |
290 | #else |
291 | static inline pmd_t pmdp_collapse_flush(struct vm_area_struct *vma, | |
292 | unsigned long address, | |
293 | pmd_t *pmdp) | |
294 | { | |
295 | BUILD_BUG(); | |
296 | return *pmdp; | |
297 | } | |
298 | #define pmdp_collapse_flush pmdp_collapse_flush | |
299 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE */ | |
300 | #endif | |
301 | ||
e3ebcf64 | 302 | #ifndef __HAVE_ARCH_PGTABLE_DEPOSIT |
6b0b50b0 AK |
303 | extern void pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp, |
304 | pgtable_t pgtable); | |
e3ebcf64 GS |
305 | #endif |
306 | ||
307 | #ifndef __HAVE_ARCH_PGTABLE_WITHDRAW | |
6b0b50b0 | 308 | extern pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp); |
e3ebcf64 GS |
309 | #endif |
310 | ||
46dcde73 GS |
311 | #ifndef __HAVE_ARCH_PMDP_INVALIDATE |
312 | extern void pmdp_invalidate(struct vm_area_struct *vma, unsigned long address, | |
313 | pmd_t *pmdp); | |
314 | #endif | |
315 | ||
c777e2a8 AK |
316 | #ifndef __HAVE_ARCH_PMDP_HUGE_SPLIT_PREPARE |
317 | static inline void pmdp_huge_split_prepare(struct vm_area_struct *vma, | |
318 | unsigned long address, pmd_t *pmdp) | |
319 | { | |
320 | ||
321 | } | |
322 | #endif | |
323 | ||
1da177e4 | 324 | #ifndef __HAVE_ARCH_PTE_SAME |
e2cda322 AA |
325 | static inline int pte_same(pte_t pte_a, pte_t pte_b) |
326 | { | |
327 | return pte_val(pte_a) == pte_val(pte_b); | |
328 | } | |
329 | #endif | |
330 | ||
45961722 KW |
331 | #ifndef __HAVE_ARCH_PTE_UNUSED |
332 | /* | |
333 | * Some architectures provide facilities to virtualization guests | |
334 | * so that they can flag allocated pages as unused. This allows the | |
335 | * host to transparently reclaim unused pages. This function returns | |
336 | * whether the pte's page is unused. | |
337 | */ | |
338 | static inline int pte_unused(pte_t pte) | |
339 | { | |
340 | return 0; | |
341 | } | |
342 | #endif | |
343 | ||
e2cda322 AA |
344 | #ifndef __HAVE_ARCH_PMD_SAME |
345 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
346 | static inline int pmd_same(pmd_t pmd_a, pmd_t pmd_b) | |
347 | { | |
348 | return pmd_val(pmd_a) == pmd_val(pmd_b); | |
349 | } | |
a00cc7d9 MW |
350 | |
351 | static inline int pud_same(pud_t pud_a, pud_t pud_b) | |
352 | { | |
353 | return pud_val(pud_a) == pud_val(pud_b); | |
354 | } | |
e2cda322 AA |
355 | #else /* CONFIG_TRANSPARENT_HUGEPAGE */ |
356 | static inline int pmd_same(pmd_t pmd_a, pmd_t pmd_b) | |
357 | { | |
bd5e88ad | 358 | BUILD_BUG(); |
e2cda322 AA |
359 | return 0; |
360 | } | |
a00cc7d9 MW |
361 | |
362 | static inline int pud_same(pud_t pud_a, pud_t pud_b) | |
363 | { | |
364 | BUILD_BUG(); | |
365 | return 0; | |
366 | } | |
e2cda322 | 367 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE */ |
1da177e4 LT |
368 | #endif |
369 | ||
1da177e4 LT |
370 | #ifndef __HAVE_ARCH_PGD_OFFSET_GATE |
371 | #define pgd_offset_gate(mm, addr) pgd_offset(mm, addr) | |
372 | #endif | |
373 | ||
0b0968a3 | 374 | #ifndef __HAVE_ARCH_MOVE_PTE |
8b1f3124 | 375 | #define move_pte(pte, prot, old_addr, new_addr) (pte) |
8b1f3124 NP |
376 | #endif |
377 | ||
2c3cf556 | 378 | #ifndef pte_accessible |
20841405 | 379 | # define pte_accessible(mm, pte) ((void)(pte), 1) |
2c3cf556 RR |
380 | #endif |
381 | ||
61c77326 SL |
382 | #ifndef flush_tlb_fix_spurious_fault |
383 | #define flush_tlb_fix_spurious_fault(vma, address) flush_tlb_page(vma, address) | |
384 | #endif | |
385 | ||
0634a632 PM |
386 | #ifndef pgprot_noncached |
387 | #define pgprot_noncached(prot) (prot) | |
388 | #endif | |
389 | ||
2520bd31 | 390 | #ifndef pgprot_writecombine |
391 | #define pgprot_writecombine pgprot_noncached | |
392 | #endif | |
393 | ||
d1b4bfbf TK |
394 | #ifndef pgprot_writethrough |
395 | #define pgprot_writethrough pgprot_noncached | |
396 | #endif | |
397 | ||
8b921acf LD |
398 | #ifndef pgprot_device |
399 | #define pgprot_device pgprot_noncached | |
400 | #endif | |
401 | ||
64e45507 PF |
402 | #ifndef pgprot_modify |
403 | #define pgprot_modify pgprot_modify | |
404 | static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot) | |
405 | { | |
406 | if (pgprot_val(oldprot) == pgprot_val(pgprot_noncached(oldprot))) | |
407 | newprot = pgprot_noncached(newprot); | |
408 | if (pgprot_val(oldprot) == pgprot_val(pgprot_writecombine(oldprot))) | |
409 | newprot = pgprot_writecombine(newprot); | |
410 | if (pgprot_val(oldprot) == pgprot_val(pgprot_device(oldprot))) | |
411 | newprot = pgprot_device(newprot); | |
412 | return newprot; | |
413 | } | |
414 | #endif | |
415 | ||
1da177e4 | 416 | /* |
8f6c99c1 HD |
417 | * When walking page tables, get the address of the next boundary, |
418 | * or the end address of the range if that comes earlier. Although no | |
419 | * vma end wraps to 0, rounded up __boundary may wrap to 0 throughout. | |
1da177e4 LT |
420 | */ |
421 | ||
1da177e4 LT |
422 | #define pgd_addr_end(addr, end) \ |
423 | ({ unsigned long __boundary = ((addr) + PGDIR_SIZE) & PGDIR_MASK; \ | |
424 | (__boundary - 1 < (end) - 1)? __boundary: (end); \ | |
425 | }) | |
1da177e4 | 426 | |
c2febafc KS |
427 | #ifndef p4d_addr_end |
428 | #define p4d_addr_end(addr, end) \ | |
429 | ({ unsigned long __boundary = ((addr) + P4D_SIZE) & P4D_MASK; \ | |
430 | (__boundary - 1 < (end) - 1)? __boundary: (end); \ | |
431 | }) | |
432 | #endif | |
433 | ||
1da177e4 LT |
434 | #ifndef pud_addr_end |
435 | #define pud_addr_end(addr, end) \ | |
436 | ({ unsigned long __boundary = ((addr) + PUD_SIZE) & PUD_MASK; \ | |
437 | (__boundary - 1 < (end) - 1)? __boundary: (end); \ | |
438 | }) | |
439 | #endif | |
440 | ||
441 | #ifndef pmd_addr_end | |
442 | #define pmd_addr_end(addr, end) \ | |
443 | ({ unsigned long __boundary = ((addr) + PMD_SIZE) & PMD_MASK; \ | |
444 | (__boundary - 1 < (end) - 1)? __boundary: (end); \ | |
445 | }) | |
446 | #endif | |
447 | ||
1da177e4 LT |
448 | /* |
449 | * When walking page tables, we usually want to skip any p?d_none entries; | |
450 | * and any p?d_bad entries - reporting the error before resetting to none. | |
451 | * Do the tests inline, but report and clear the bad entry in mm/memory.c. | |
452 | */ | |
453 | void pgd_clear_bad(pgd_t *); | |
c2febafc | 454 | void p4d_clear_bad(p4d_t *); |
1da177e4 LT |
455 | void pud_clear_bad(pud_t *); |
456 | void pmd_clear_bad(pmd_t *); | |
457 | ||
458 | static inline int pgd_none_or_clear_bad(pgd_t *pgd) | |
459 | { | |
460 | if (pgd_none(*pgd)) | |
461 | return 1; | |
462 | if (unlikely(pgd_bad(*pgd))) { | |
463 | pgd_clear_bad(pgd); | |
464 | return 1; | |
465 | } | |
466 | return 0; | |
467 | } | |
468 | ||
c2febafc KS |
469 | static inline int p4d_none_or_clear_bad(p4d_t *p4d) |
470 | { | |
471 | if (p4d_none(*p4d)) | |
472 | return 1; | |
473 | if (unlikely(p4d_bad(*p4d))) { | |
474 | p4d_clear_bad(p4d); | |
475 | return 1; | |
476 | } | |
477 | return 0; | |
478 | } | |
479 | ||
1da177e4 LT |
480 | static inline int pud_none_or_clear_bad(pud_t *pud) |
481 | { | |
482 | if (pud_none(*pud)) | |
483 | return 1; | |
484 | if (unlikely(pud_bad(*pud))) { | |
485 | pud_clear_bad(pud); | |
486 | return 1; | |
487 | } | |
488 | return 0; | |
489 | } | |
490 | ||
491 | static inline int pmd_none_or_clear_bad(pmd_t *pmd) | |
492 | { | |
493 | if (pmd_none(*pmd)) | |
494 | return 1; | |
495 | if (unlikely(pmd_bad(*pmd))) { | |
496 | pmd_clear_bad(pmd); | |
497 | return 1; | |
498 | } | |
499 | return 0; | |
500 | } | |
9535239f | 501 | |
1ea0704e JF |
502 | static inline pte_t __ptep_modify_prot_start(struct mm_struct *mm, |
503 | unsigned long addr, | |
504 | pte_t *ptep) | |
505 | { | |
506 | /* | |
507 | * Get the current pte state, but zero it out to make it | |
508 | * non-present, preventing the hardware from asynchronously | |
509 | * updating it. | |
510 | */ | |
511 | return ptep_get_and_clear(mm, addr, ptep); | |
512 | } | |
513 | ||
514 | static inline void __ptep_modify_prot_commit(struct mm_struct *mm, | |
515 | unsigned long addr, | |
516 | pte_t *ptep, pte_t pte) | |
517 | { | |
518 | /* | |
519 | * The pte is non-present, so there's no hardware state to | |
520 | * preserve. | |
521 | */ | |
522 | set_pte_at(mm, addr, ptep, pte); | |
523 | } | |
524 | ||
525 | #ifndef __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION | |
526 | /* | |
527 | * Start a pte protection read-modify-write transaction, which | |
528 | * protects against asynchronous hardware modifications to the pte. | |
529 | * The intention is not to prevent the hardware from making pte | |
530 | * updates, but to prevent any updates it may make from being lost. | |
531 | * | |
532 | * This does not protect against other software modifications of the | |
533 | * pte; the appropriate pte lock must be held over the transation. | |
534 | * | |
535 | * Note that this interface is intended to be batchable, meaning that | |
536 | * ptep_modify_prot_commit may not actually update the pte, but merely | |
537 | * queue the update to be done at some later time. The update must be | |
538 | * actually committed before the pte lock is released, however. | |
539 | */ | |
540 | static inline pte_t ptep_modify_prot_start(struct mm_struct *mm, | |
541 | unsigned long addr, | |
542 | pte_t *ptep) | |
543 | { | |
544 | return __ptep_modify_prot_start(mm, addr, ptep); | |
545 | } | |
546 | ||
547 | /* | |
548 | * Commit an update to a pte, leaving any hardware-controlled bits in | |
549 | * the PTE unmodified. | |
550 | */ | |
551 | static inline void ptep_modify_prot_commit(struct mm_struct *mm, | |
552 | unsigned long addr, | |
553 | pte_t *ptep, pte_t pte) | |
554 | { | |
555 | __ptep_modify_prot_commit(mm, addr, ptep, pte); | |
556 | } | |
557 | #endif /* __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION */ | |
fe1a6875 | 558 | #endif /* CONFIG_MMU */ |
1ea0704e | 559 | |
9535239f GU |
560 | /* |
561 | * A facility to provide lazy MMU batching. This allows PTE updates and | |
562 | * page invalidations to be delayed until a call to leave lazy MMU mode | |
563 | * is issued. Some architectures may benefit from doing this, and it is | |
564 | * beneficial for both shadow and direct mode hypervisors, which may batch | |
565 | * the PTE updates which happen during this window. Note that using this | |
566 | * interface requires that read hazards be removed from the code. A read | |
567 | * hazard could result in the direct mode hypervisor case, since the actual | |
568 | * write to the page tables may not yet have taken place, so reads though | |
569 | * a raw PTE pointer after it has been modified are not guaranteed to be | |
570 | * up to date. This mode can only be entered and left under the protection of | |
571 | * the page table locks for all page tables which may be modified. In the UP | |
572 | * case, this is required so that preemption is disabled, and in the SMP case, | |
573 | * it must synchronize the delayed page table writes properly on other CPUs. | |
574 | */ | |
575 | #ifndef __HAVE_ARCH_ENTER_LAZY_MMU_MODE | |
576 | #define arch_enter_lazy_mmu_mode() do {} while (0) | |
577 | #define arch_leave_lazy_mmu_mode() do {} while (0) | |
578 | #define arch_flush_lazy_mmu_mode() do {} while (0) | |
579 | #endif | |
580 | ||
581 | /* | |
7fd7d83d JF |
582 | * A facility to provide batching of the reload of page tables and |
583 | * other process state with the actual context switch code for | |
584 | * paravirtualized guests. By convention, only one of the batched | |
585 | * update (lazy) modes (CPU, MMU) should be active at any given time, | |
586 | * entry should never be nested, and entry and exits should always be | |
587 | * paired. This is for sanity of maintaining and reasoning about the | |
588 | * kernel code. In this case, the exit (end of the context switch) is | |
589 | * in architecture-specific code, and so doesn't need a generic | |
590 | * definition. | |
9535239f | 591 | */ |
7fd7d83d | 592 | #ifndef __HAVE_ARCH_START_CONTEXT_SWITCH |
224101ed | 593 | #define arch_start_context_switch(prev) do {} while (0) |
9535239f GU |
594 | #endif |
595 | ||
0f8975ec PE |
596 | #ifndef CONFIG_HAVE_ARCH_SOFT_DIRTY |
597 | static inline int pte_soft_dirty(pte_t pte) | |
598 | { | |
599 | return 0; | |
600 | } | |
601 | ||
602 | static inline int pmd_soft_dirty(pmd_t pmd) | |
603 | { | |
604 | return 0; | |
605 | } | |
606 | ||
607 | static inline pte_t pte_mksoft_dirty(pte_t pte) | |
608 | { | |
609 | return pte; | |
610 | } | |
611 | ||
612 | static inline pmd_t pmd_mksoft_dirty(pmd_t pmd) | |
613 | { | |
614 | return pmd; | |
615 | } | |
179ef71c | 616 | |
a7b76174 MS |
617 | static inline pte_t pte_clear_soft_dirty(pte_t pte) |
618 | { | |
619 | return pte; | |
620 | } | |
621 | ||
622 | static inline pmd_t pmd_clear_soft_dirty(pmd_t pmd) | |
623 | { | |
624 | return pmd; | |
625 | } | |
626 | ||
179ef71c CG |
627 | static inline pte_t pte_swp_mksoft_dirty(pte_t pte) |
628 | { | |
629 | return pte; | |
630 | } | |
631 | ||
632 | static inline int pte_swp_soft_dirty(pte_t pte) | |
633 | { | |
634 | return 0; | |
635 | } | |
636 | ||
637 | static inline pte_t pte_swp_clear_soft_dirty(pte_t pte) | |
638 | { | |
639 | return pte; | |
640 | } | |
0f8975ec PE |
641 | #endif |
642 | ||
34801ba9 | 643 | #ifndef __HAVE_PFNMAP_TRACKING |
644 | /* | |
5180da41 SS |
645 | * Interfaces that can be used by architecture code to keep track of |
646 | * memory type of pfn mappings specified by the remap_pfn_range, | |
647 | * vm_insert_pfn. | |
648 | */ | |
649 | ||
650 | /* | |
651 | * track_pfn_remap is called when a _new_ pfn mapping is being established | |
652 | * by remap_pfn_range() for physical range indicated by pfn and size. | |
34801ba9 | 653 | */ |
5180da41 | 654 | static inline int track_pfn_remap(struct vm_area_struct *vma, pgprot_t *prot, |
b3b9c293 KK |
655 | unsigned long pfn, unsigned long addr, |
656 | unsigned long size) | |
34801ba9 | 657 | { |
658 | return 0; | |
659 | } | |
660 | ||
661 | /* | |
5180da41 SS |
662 | * track_pfn_insert is called when a _new_ single pfn is established |
663 | * by vm_insert_pfn(). | |
664 | */ | |
308a047c BP |
665 | static inline void track_pfn_insert(struct vm_area_struct *vma, pgprot_t *prot, |
666 | pfn_t pfn) | |
5180da41 | 667 | { |
5180da41 SS |
668 | } |
669 | ||
670 | /* | |
671 | * track_pfn_copy is called when vma that is covering the pfnmap gets | |
34801ba9 | 672 | * copied through copy_page_range(). |
673 | */ | |
5180da41 | 674 | static inline int track_pfn_copy(struct vm_area_struct *vma) |
34801ba9 | 675 | { |
676 | return 0; | |
677 | } | |
678 | ||
679 | /* | |
d9fe4fab | 680 | * untrack_pfn is called while unmapping a pfnmap for a region. |
34801ba9 | 681 | * untrack can be called for a specific region indicated by pfn and size or |
5180da41 | 682 | * can be for the entire vma (in which case pfn, size are zero). |
34801ba9 | 683 | */ |
5180da41 SS |
684 | static inline void untrack_pfn(struct vm_area_struct *vma, |
685 | unsigned long pfn, unsigned long size) | |
34801ba9 | 686 | { |
687 | } | |
d9fe4fab TK |
688 | |
689 | /* | |
690 | * untrack_pfn_moved is called while mremapping a pfnmap for a new region. | |
691 | */ | |
692 | static inline void untrack_pfn_moved(struct vm_area_struct *vma) | |
693 | { | |
694 | } | |
34801ba9 | 695 | #else |
5180da41 | 696 | extern int track_pfn_remap(struct vm_area_struct *vma, pgprot_t *prot, |
b3b9c293 KK |
697 | unsigned long pfn, unsigned long addr, |
698 | unsigned long size); | |
308a047c BP |
699 | extern void track_pfn_insert(struct vm_area_struct *vma, pgprot_t *prot, |
700 | pfn_t pfn); | |
5180da41 SS |
701 | extern int track_pfn_copy(struct vm_area_struct *vma); |
702 | extern void untrack_pfn(struct vm_area_struct *vma, unsigned long pfn, | |
703 | unsigned long size); | |
d9fe4fab | 704 | extern void untrack_pfn_moved(struct vm_area_struct *vma); |
34801ba9 | 705 | #endif |
706 | ||
816422ad KS |
707 | #ifdef __HAVE_COLOR_ZERO_PAGE |
708 | static inline int is_zero_pfn(unsigned long pfn) | |
709 | { | |
710 | extern unsigned long zero_pfn; | |
711 | unsigned long offset_from_zero_pfn = pfn - zero_pfn; | |
712 | return offset_from_zero_pfn <= (zero_page_mask >> PAGE_SHIFT); | |
713 | } | |
714 | ||
2f91ec8c KS |
715 | #define my_zero_pfn(addr) page_to_pfn(ZERO_PAGE(addr)) |
716 | ||
816422ad KS |
717 | #else |
718 | static inline int is_zero_pfn(unsigned long pfn) | |
719 | { | |
720 | extern unsigned long zero_pfn; | |
721 | return pfn == zero_pfn; | |
722 | } | |
723 | ||
724 | static inline unsigned long my_zero_pfn(unsigned long addr) | |
725 | { | |
726 | extern unsigned long zero_pfn; | |
727 | return zero_pfn; | |
728 | } | |
729 | #endif | |
730 | ||
1a5a9906 AA |
731 | #ifdef CONFIG_MMU |
732 | ||
5f6e8da7 AA |
733 | #ifndef CONFIG_TRANSPARENT_HUGEPAGE |
734 | static inline int pmd_trans_huge(pmd_t pmd) | |
735 | { | |
736 | return 0; | |
737 | } | |
e2cda322 AA |
738 | #ifndef __HAVE_ARCH_PMD_WRITE |
739 | static inline int pmd_write(pmd_t pmd) | |
740 | { | |
741 | BUG(); | |
742 | return 0; | |
743 | } | |
744 | #endif /* __HAVE_ARCH_PMD_WRITE */ | |
1a5a9906 AA |
745 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE */ |
746 | ||
a00cc7d9 MW |
747 | #if !defined(CONFIG_TRANSPARENT_HUGEPAGE) || \ |
748 | (defined(CONFIG_TRANSPARENT_HUGEPAGE) && \ | |
749 | !defined(CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD)) | |
750 | static inline int pud_trans_huge(pud_t pud) | |
751 | { | |
752 | return 0; | |
753 | } | |
754 | #endif | |
755 | ||
26c19178 AA |
756 | #ifndef pmd_read_atomic |
757 | static inline pmd_t pmd_read_atomic(pmd_t *pmdp) | |
758 | { | |
759 | /* | |
760 | * Depend on compiler for an atomic pmd read. NOTE: this is | |
761 | * only going to work, if the pmdval_t isn't larger than | |
762 | * an unsigned long. | |
763 | */ | |
764 | return *pmdp; | |
765 | } | |
766 | #endif | |
767 | ||
953c66c2 AK |
768 | #ifndef arch_needs_pgtable_deposit |
769 | #define arch_needs_pgtable_deposit() (false) | |
770 | #endif | |
1a5a9906 AA |
771 | /* |
772 | * This function is meant to be used by sites walking pagetables with | |
773 | * the mmap_sem hold in read mode to protect against MADV_DONTNEED and | |
774 | * transhuge page faults. MADV_DONTNEED can convert a transhuge pmd | |
775 | * into a null pmd and the transhuge page fault can convert a null pmd | |
776 | * into an hugepmd or into a regular pmd (if the hugepage allocation | |
777 | * fails). While holding the mmap_sem in read mode the pmd becomes | |
778 | * stable and stops changing under us only if it's not null and not a | |
779 | * transhuge pmd. When those races occurs and this function makes a | |
780 | * difference vs the standard pmd_none_or_clear_bad, the result is | |
781 | * undefined so behaving like if the pmd was none is safe (because it | |
782 | * can return none anyway). The compiler level barrier() is critically | |
783 | * important to compute the two checks atomically on the same pmdval. | |
26c19178 AA |
784 | * |
785 | * For 32bit kernels with a 64bit large pmd_t this automatically takes | |
786 | * care of reading the pmd atomically to avoid SMP race conditions | |
787 | * against pmd_populate() when the mmap_sem is hold for reading by the | |
788 | * caller (a special atomic read not done by "gcc" as in the generic | |
789 | * version above, is also needed when THP is disabled because the page | |
790 | * fault can populate the pmd from under us). | |
1a5a9906 AA |
791 | */ |
792 | static inline int pmd_none_or_trans_huge_or_clear_bad(pmd_t *pmd) | |
793 | { | |
26c19178 | 794 | pmd_t pmdval = pmd_read_atomic(pmd); |
1a5a9906 AA |
795 | /* |
796 | * The barrier will stabilize the pmdval in a register or on | |
797 | * the stack so that it will stop changing under the code. | |
e4eed03f AA |
798 | * |
799 | * When CONFIG_TRANSPARENT_HUGEPAGE=y on x86 32bit PAE, | |
800 | * pmd_read_atomic is allowed to return a not atomic pmdval | |
801 | * (for example pointing to an hugepage that has never been | |
802 | * mapped in the pmd). The below checks will only care about | |
803 | * the low part of the pmd with 32bit PAE x86 anyway, with the | |
804 | * exception of pmd_none(). So the important thing is that if | |
805 | * the low part of the pmd is found null, the high part will | |
806 | * be also null or the pmd_none() check below would be | |
807 | * confused. | |
1a5a9906 AA |
808 | */ |
809 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
810 | barrier(); | |
811 | #endif | |
ee53664b | 812 | if (pmd_none(pmdval) || pmd_trans_huge(pmdval)) |
1a5a9906 AA |
813 | return 1; |
814 | if (unlikely(pmd_bad(pmdval))) { | |
ee53664b | 815 | pmd_clear_bad(pmd); |
1a5a9906 AA |
816 | return 1; |
817 | } | |
818 | return 0; | |
819 | } | |
820 | ||
821 | /* | |
822 | * This is a noop if Transparent Hugepage Support is not built into | |
823 | * the kernel. Otherwise it is equivalent to | |
824 | * pmd_none_or_trans_huge_or_clear_bad(), and shall only be called in | |
825 | * places that already verified the pmd is not none and they want to | |
826 | * walk ptes while holding the mmap sem in read mode (write mode don't | |
827 | * need this). If THP is not enabled, the pmd can't go away under the | |
828 | * code even if MADV_DONTNEED runs, but if THP is enabled we need to | |
829 | * run a pmd_trans_unstable before walking the ptes after | |
830 | * split_huge_page_pmd returns (because it may have run when the pmd | |
831 | * become null, but then a page fault can map in a THP and not a | |
832 | * regular page). | |
833 | */ | |
834 | static inline int pmd_trans_unstable(pmd_t *pmd) | |
835 | { | |
836 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
837 | return pmd_none_or_trans_huge_or_clear_bad(pmd); | |
838 | #else | |
839 | return 0; | |
5f6e8da7 | 840 | #endif |
1a5a9906 AA |
841 | } |
842 | ||
e7bb4b6d MG |
843 | #ifndef CONFIG_NUMA_BALANCING |
844 | /* | |
845 | * Technically a PTE can be PROTNONE even when not doing NUMA balancing but | |
846 | * the only case the kernel cares is for NUMA balancing and is only ever set | |
847 | * when the VMA is accessible. For PROT_NONE VMAs, the PTEs are not marked | |
848 | * _PAGE_PROTNONE so by by default, implement the helper as "always no". It | |
849 | * is the responsibility of the caller to distinguish between PROT_NONE | |
850 | * protections and NUMA hinting fault protections. | |
851 | */ | |
852 | static inline int pte_protnone(pte_t pte) | |
853 | { | |
854 | return 0; | |
855 | } | |
856 | ||
857 | static inline int pmd_protnone(pmd_t pmd) | |
858 | { | |
859 | return 0; | |
860 | } | |
861 | #endif /* CONFIG_NUMA_BALANCING */ | |
862 | ||
1a5a9906 | 863 | #endif /* CONFIG_MMU */ |
5f6e8da7 | 864 | |
e61ce6ad | 865 | #ifdef CONFIG_HAVE_ARCH_HUGE_VMAP |
c2febafc KS |
866 | |
867 | #ifndef __PAGETABLE_P4D_FOLDED | |
868 | int p4d_set_huge(p4d_t *p4d, phys_addr_t addr, pgprot_t prot); | |
869 | int p4d_clear_huge(p4d_t *p4d); | |
870 | #else | |
871 | static inline int p4d_set_huge(p4d_t *p4d, phys_addr_t addr, pgprot_t prot) | |
872 | { | |
873 | return 0; | |
874 | } | |
875 | static inline int p4d_clear_huge(p4d_t *p4d) | |
876 | { | |
877 | return 0; | |
878 | } | |
879 | #endif /* !__PAGETABLE_P4D_FOLDED */ | |
880 | ||
e61ce6ad TK |
881 | int pud_set_huge(pud_t *pud, phys_addr_t addr, pgprot_t prot); |
882 | int pmd_set_huge(pmd_t *pmd, phys_addr_t addr, pgprot_t prot); | |
b9820d8f TK |
883 | int pud_clear_huge(pud_t *pud); |
884 | int pmd_clear_huge(pmd_t *pmd); | |
e61ce6ad | 885 | #else /* !CONFIG_HAVE_ARCH_HUGE_VMAP */ |
c2febafc KS |
886 | static inline int p4d_set_huge(p4d_t *p4d, phys_addr_t addr, pgprot_t prot) |
887 | { | |
888 | return 0; | |
889 | } | |
e61ce6ad TK |
890 | static inline int pud_set_huge(pud_t *pud, phys_addr_t addr, pgprot_t prot) |
891 | { | |
892 | return 0; | |
893 | } | |
894 | static inline int pmd_set_huge(pmd_t *pmd, phys_addr_t addr, pgprot_t prot) | |
895 | { | |
896 | return 0; | |
897 | } | |
c2febafc KS |
898 | static inline int p4d_clear_huge(p4d_t *p4d) |
899 | { | |
900 | return 0; | |
901 | } | |
b9820d8f TK |
902 | static inline int pud_clear_huge(pud_t *pud) |
903 | { | |
904 | return 0; | |
905 | } | |
906 | static inline int pmd_clear_huge(pmd_t *pmd) | |
907 | { | |
908 | return 0; | |
909 | } | |
e61ce6ad TK |
910 | #endif /* CONFIG_HAVE_ARCH_HUGE_VMAP */ |
911 | ||
458aa76d AK |
912 | #ifndef __HAVE_ARCH_FLUSH_PMD_TLB_RANGE |
913 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
914 | /* | |
915 | * ARCHes with special requirements for evicting THP backing TLB entries can | |
916 | * implement this. Otherwise also, it can help optimize normal TLB flush in | |
917 | * THP regime. stock flush_tlb_range() typically has optimization to nuke the | |
918 | * entire TLB TLB if flush span is greater than a threshold, which will | |
919 | * likely be true for a single huge page. Thus a single thp flush will | |
920 | * invalidate the entire TLB which is not desitable. | |
921 | * e.g. see arch/arc: flush_pmd_tlb_range | |
922 | */ | |
923 | #define flush_pmd_tlb_range(vma, addr, end) flush_tlb_range(vma, addr, end) | |
a00cc7d9 | 924 | #define flush_pud_tlb_range(vma, addr, end) flush_tlb_range(vma, addr, end) |
458aa76d AK |
925 | #else |
926 | #define flush_pmd_tlb_range(vma, addr, end) BUILD_BUG() | |
a00cc7d9 | 927 | #define flush_pud_tlb_range(vma, addr, end) BUILD_BUG() |
458aa76d AK |
928 | #endif |
929 | #endif | |
930 | ||
08ea8c07 BX |
931 | struct file; |
932 | int phys_mem_access_prot_allowed(struct file *file, unsigned long pfn, | |
933 | unsigned long size, pgprot_t *vma_prot); | |
1da177e4 LT |
934 | #endif /* !__ASSEMBLY__ */ |
935 | ||
40d158e6 AV |
936 | #ifndef io_remap_pfn_range |
937 | #define io_remap_pfn_range remap_pfn_range | |
938 | #endif | |
939 | ||
fd8cfd30 HD |
940 | #ifndef has_transparent_hugepage |
941 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
942 | #define has_transparent_hugepage() 1 | |
943 | #else | |
944 | #define has_transparent_hugepage() 0 | |
945 | #endif | |
946 | #endif | |
947 | ||
1da177e4 | 948 | #endif /* _ASM_GENERIC_PGTABLE_H */ |