xref: /linux/arch/arm64/include/asm/pgtable.h (revision b2d0f5d5dc53532e6f07bc546a476a55ebdfe0f3)
1 /*
2  * Copyright (C) 2012 ARM Ltd.
3  *
4  * This program is free software; you can redistribute it and/or modify
5  * it under the terms of the GNU General Public License version 2 as
6  * published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope that it will be useful,
9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
11  * GNU General Public License for more details.
12  *
13  * You should have received a copy of the GNU General Public License
14  * along with this program.  If not, see <http://www.gnu.org/licenses/>.
15  */
16 #ifndef __ASM_PGTABLE_H
17 #define __ASM_PGTABLE_H
18 
19 #include <asm/bug.h>
20 #include <asm/proc-fns.h>
21 
22 #include <asm/memory.h>
23 #include <asm/pgtable-hwdef.h>
24 #include <asm/pgtable-prot.h>
25 
26 /*
27  * VMALLOC range.
28  *
29  * VMALLOC_START: beginning of the kernel vmalloc space
30  * VMALLOC_END: extends to the available space below vmmemmap, PCI I/O space
31  *	and fixed mappings
32  */
33 #define VMALLOC_START		(MODULES_END)
34 #define VMALLOC_END		(PAGE_OFFSET - PUD_SIZE - VMEMMAP_SIZE - SZ_64K)
35 
36 #define vmemmap			((struct page *)VMEMMAP_START - (memstart_addr >> PAGE_SHIFT))
37 
38 #define FIRST_USER_ADDRESS	0UL
39 
40 #ifndef __ASSEMBLY__
41 
42 #include <asm/cmpxchg.h>
43 #include <asm/fixmap.h>
44 #include <linux/mmdebug.h>
45 
46 extern void __pte_error(const char *file, int line, unsigned long val);
47 extern void __pmd_error(const char *file, int line, unsigned long val);
48 extern void __pud_error(const char *file, int line, unsigned long val);
49 extern void __pgd_error(const char *file, int line, unsigned long val);
50 
51 /*
52  * ZERO_PAGE is a global shared page that is always zero: used
53  * for zero-mapped memory areas etc..
54  */
55 extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)];
56 #define ZERO_PAGE(vaddr)	phys_to_page(__pa_symbol(empty_zero_page))
57 
58 #define pte_ERROR(pte)		__pte_error(__FILE__, __LINE__, pte_val(pte))
59 
60 #define pte_pfn(pte)		((pte_val(pte) & PHYS_MASK) >> PAGE_SHIFT)
61 
62 #define pfn_pte(pfn,prot)	(__pte(((phys_addr_t)(pfn) << PAGE_SHIFT) | pgprot_val(prot)))
63 
64 #define pte_none(pte)		(!pte_val(pte))
65 #define pte_clear(mm,addr,ptep)	set_pte(ptep, __pte(0))
66 #define pte_page(pte)		(pfn_to_page(pte_pfn(pte)))
67 
68 /*
69  * The following only work if pte_present(). Undefined behaviour otherwise.
70  */
71 #define pte_present(pte)	(!!(pte_val(pte) & (PTE_VALID | PTE_PROT_NONE)))
72 #define pte_young(pte)		(!!(pte_val(pte) & PTE_AF))
73 #define pte_special(pte)	(!!(pte_val(pte) & PTE_SPECIAL))
74 #define pte_write(pte)		(!!(pte_val(pte) & PTE_WRITE))
75 #define pte_user_exec(pte)	(!(pte_val(pte) & PTE_UXN))
76 #define pte_cont(pte)		(!!(pte_val(pte) & PTE_CONT))
77 
78 #define pte_cont_addr_end(addr, end)						\
79 ({	unsigned long __boundary = ((addr) + CONT_PTE_SIZE) & CONT_PTE_MASK;	\
80 	(__boundary - 1 < (end) - 1) ? __boundary : (end);			\
81 })
82 
83 #define pmd_cont_addr_end(addr, end)						\
84 ({	unsigned long __boundary = ((addr) + CONT_PMD_SIZE) & CONT_PMD_MASK;	\
85 	(__boundary - 1 < (end) - 1) ? __boundary : (end);			\
86 })
87 
88 #define pte_hw_dirty(pte)	(pte_write(pte) && !(pte_val(pte) & PTE_RDONLY))
89 #define pte_sw_dirty(pte)	(!!(pte_val(pte) & PTE_DIRTY))
90 #define pte_dirty(pte)		(pte_sw_dirty(pte) || pte_hw_dirty(pte))
91 
92 #define pte_valid(pte)		(!!(pte_val(pte) & PTE_VALID))
93 /*
94  * Execute-only user mappings do not have the PTE_USER bit set. All valid
95  * kernel mappings have the PTE_UXN bit set.
96  */
97 #define pte_valid_not_user(pte) \
98 	((pte_val(pte) & (PTE_VALID | PTE_USER | PTE_UXN)) == (PTE_VALID | PTE_UXN))
99 #define pte_valid_young(pte) \
100 	((pte_val(pte) & (PTE_VALID | PTE_AF)) == (PTE_VALID | PTE_AF))
101 
102 /*
103  * Could the pte be present in the TLB? We must check mm_tlb_flush_pending
104  * so that we don't erroneously return false for pages that have been
105  * remapped as PROT_NONE but are yet to be flushed from the TLB.
106  */
107 #define pte_accessible(mm, pte)	\
108 	(mm_tlb_flush_pending(mm) ? pte_present(pte) : pte_valid_young(pte))
109 
110 static inline pte_t clear_pte_bit(pte_t pte, pgprot_t prot)
111 {
112 	pte_val(pte) &= ~pgprot_val(prot);
113 	return pte;
114 }
115 
116 static inline pte_t set_pte_bit(pte_t pte, pgprot_t prot)
117 {
118 	pte_val(pte) |= pgprot_val(prot);
119 	return pte;
120 }
121 
122 static inline pte_t pte_wrprotect(pte_t pte)
123 {
124 	pte = clear_pte_bit(pte, __pgprot(PTE_WRITE));
125 	pte = set_pte_bit(pte, __pgprot(PTE_RDONLY));
126 	return pte;
127 }
128 
129 static inline pte_t pte_mkwrite(pte_t pte)
130 {
131 	pte = set_pte_bit(pte, __pgprot(PTE_WRITE));
132 	pte = clear_pte_bit(pte, __pgprot(PTE_RDONLY));
133 	return pte;
134 }
135 
136 static inline pte_t pte_mkclean(pte_t pte)
137 {
138 	return clear_pte_bit(pte, __pgprot(PTE_DIRTY));
139 }
140 
141 static inline pte_t pte_mkdirty(pte_t pte)
142 {
143 	return set_pte_bit(pte, __pgprot(PTE_DIRTY));
144 }
145 
146 static inline pte_t pte_mkold(pte_t pte)
147 {
148 	return clear_pte_bit(pte, __pgprot(PTE_AF));
149 }
150 
151 static inline pte_t pte_mkyoung(pte_t pte)
152 {
153 	return set_pte_bit(pte, __pgprot(PTE_AF));
154 }
155 
156 static inline pte_t pte_mkspecial(pte_t pte)
157 {
158 	return set_pte_bit(pte, __pgprot(PTE_SPECIAL));
159 }
160 
161 static inline pte_t pte_mkcont(pte_t pte)
162 {
163 	pte = set_pte_bit(pte, __pgprot(PTE_CONT));
164 	return set_pte_bit(pte, __pgprot(PTE_TYPE_PAGE));
165 }
166 
167 static inline pte_t pte_mknoncont(pte_t pte)
168 {
169 	return clear_pte_bit(pte, __pgprot(PTE_CONT));
170 }
171 
172 static inline pte_t pte_mkpresent(pte_t pte)
173 {
174 	return set_pte_bit(pte, __pgprot(PTE_VALID));
175 }
176 
177 static inline pmd_t pmd_mkcont(pmd_t pmd)
178 {
179 	return __pmd(pmd_val(pmd) | PMD_SECT_CONT);
180 }
181 
182 static inline void set_pte(pte_t *ptep, pte_t pte)
183 {
184 	*ptep = pte;
185 
186 	/*
187 	 * Only if the new pte is valid and kernel, otherwise TLB maintenance
188 	 * or update_mmu_cache() have the necessary barriers.
189 	 */
190 	if (pte_valid_not_user(pte)) {
191 		dsb(ishst);
192 		isb();
193 	}
194 }
195 
196 struct mm_struct;
197 struct vm_area_struct;
198 
199 extern void __sync_icache_dcache(pte_t pteval, unsigned long addr);
200 
201 /*
202  * PTE bits configuration in the presence of hardware Dirty Bit Management
203  * (PTE_WRITE == PTE_DBM):
204  *
205  * Dirty  Writable | PTE_RDONLY  PTE_WRITE  PTE_DIRTY (sw)
206  *   0      0      |   1           0          0
207  *   0      1      |   1           1          0
208  *   1      0      |   1           0          1
209  *   1      1      |   0           1          x
210  *
211  * When hardware DBM is not present, the sofware PTE_DIRTY bit is updated via
212  * the page fault mechanism. Checking the dirty status of a pte becomes:
213  *
214  *   PTE_DIRTY || (PTE_WRITE && !PTE_RDONLY)
215  */
216 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
217 			      pte_t *ptep, pte_t pte)
218 {
219 	if (pte_present(pte) && pte_user_exec(pte) && !pte_special(pte))
220 		__sync_icache_dcache(pte, addr);
221 
222 	/*
223 	 * If the existing pte is valid, check for potential race with
224 	 * hardware updates of the pte (ptep_set_access_flags safely changes
225 	 * valid ptes without going through an invalid entry).
226 	 */
227 	if (pte_valid(*ptep) && pte_valid(pte)) {
228 		VM_WARN_ONCE(!pte_young(pte),
229 			     "%s: racy access flag clearing: 0x%016llx -> 0x%016llx",
230 			     __func__, pte_val(*ptep), pte_val(pte));
231 		VM_WARN_ONCE(pte_write(*ptep) && !pte_dirty(pte),
232 			     "%s: racy dirty state clearing: 0x%016llx -> 0x%016llx",
233 			     __func__, pte_val(*ptep), pte_val(pte));
234 	}
235 
236 	set_pte(ptep, pte);
237 }
238 
239 #define __HAVE_ARCH_PTE_SAME
240 static inline int pte_same(pte_t pte_a, pte_t pte_b)
241 {
242 	pteval_t lhs, rhs;
243 
244 	lhs = pte_val(pte_a);
245 	rhs = pte_val(pte_b);
246 
247 	if (pte_present(pte_a))
248 		lhs &= ~PTE_RDONLY;
249 
250 	if (pte_present(pte_b))
251 		rhs &= ~PTE_RDONLY;
252 
253 	return (lhs == rhs);
254 }
255 
256 /*
257  * Huge pte definitions.
258  */
259 #define pte_huge(pte)		(!(pte_val(pte) & PTE_TABLE_BIT))
260 #define pte_mkhuge(pte)		(__pte(pte_val(pte) & ~PTE_TABLE_BIT))
261 
262 /*
263  * Hugetlb definitions.
264  */
265 #define HUGE_MAX_HSTATE		4
266 #define HPAGE_SHIFT		PMD_SHIFT
267 #define HPAGE_SIZE		(_AC(1, UL) << HPAGE_SHIFT)
268 #define HPAGE_MASK		(~(HPAGE_SIZE - 1))
269 #define HUGETLB_PAGE_ORDER	(HPAGE_SHIFT - PAGE_SHIFT)
270 
271 #define __HAVE_ARCH_PTE_SPECIAL
272 
273 static inline pte_t pud_pte(pud_t pud)
274 {
275 	return __pte(pud_val(pud));
276 }
277 
278 static inline pmd_t pud_pmd(pud_t pud)
279 {
280 	return __pmd(pud_val(pud));
281 }
282 
283 static inline pte_t pmd_pte(pmd_t pmd)
284 {
285 	return __pte(pmd_val(pmd));
286 }
287 
288 static inline pmd_t pte_pmd(pte_t pte)
289 {
290 	return __pmd(pte_val(pte));
291 }
292 
293 static inline pgprot_t mk_sect_prot(pgprot_t prot)
294 {
295 	return __pgprot(pgprot_val(prot) & ~PTE_TABLE_BIT);
296 }
297 
298 #ifdef CONFIG_NUMA_BALANCING
299 /*
300  * See the comment in include/asm-generic/pgtable.h
301  */
302 static inline int pte_protnone(pte_t pte)
303 {
304 	return (pte_val(pte) & (PTE_VALID | PTE_PROT_NONE)) == PTE_PROT_NONE;
305 }
306 
307 static inline int pmd_protnone(pmd_t pmd)
308 {
309 	return pte_protnone(pmd_pte(pmd));
310 }
311 #endif
312 
313 /*
314  * THP definitions.
315  */
316 
317 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
318 #define pmd_trans_huge(pmd)	(pmd_val(pmd) && !(pmd_val(pmd) & PMD_TABLE_BIT))
319 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
320 
321 #define pmd_present(pmd)	pte_present(pmd_pte(pmd))
322 #define pmd_dirty(pmd)		pte_dirty(pmd_pte(pmd))
323 #define pmd_young(pmd)		pte_young(pmd_pte(pmd))
324 #define pmd_wrprotect(pmd)	pte_pmd(pte_wrprotect(pmd_pte(pmd)))
325 #define pmd_mkold(pmd)		pte_pmd(pte_mkold(pmd_pte(pmd)))
326 #define pmd_mkwrite(pmd)	pte_pmd(pte_mkwrite(pmd_pte(pmd)))
327 #define pmd_mkclean(pmd)	pte_pmd(pte_mkclean(pmd_pte(pmd)))
328 #define pmd_mkdirty(pmd)	pte_pmd(pte_mkdirty(pmd_pte(pmd)))
329 #define pmd_mkyoung(pmd)	pte_pmd(pte_mkyoung(pmd_pte(pmd)))
330 #define pmd_mknotpresent(pmd)	(__pmd(pmd_val(pmd) & ~PMD_SECT_VALID))
331 
332 #define pmd_thp_or_huge(pmd)	(pmd_huge(pmd) || pmd_trans_huge(pmd))
333 
334 #define __HAVE_ARCH_PMD_WRITE
335 #define pmd_write(pmd)		pte_write(pmd_pte(pmd))
336 
337 #define pmd_mkhuge(pmd)		(__pmd(pmd_val(pmd) & ~PMD_TABLE_BIT))
338 
339 #define pmd_pfn(pmd)		(((pmd_val(pmd) & PMD_MASK) & PHYS_MASK) >> PAGE_SHIFT)
340 #define pfn_pmd(pfn,prot)	(__pmd(((phys_addr_t)(pfn) << PAGE_SHIFT) | pgprot_val(prot)))
341 #define mk_pmd(page,prot)	pfn_pmd(page_to_pfn(page),prot)
342 
343 #define pud_write(pud)		pte_write(pud_pte(pud))
344 #define pud_pfn(pud)		(((pud_val(pud) & PUD_MASK) & PHYS_MASK) >> PAGE_SHIFT)
345 
346 #define set_pmd_at(mm, addr, pmdp, pmd)	set_pte_at(mm, addr, (pte_t *)pmdp, pmd_pte(pmd))
347 
348 #define __pgprot_modify(prot,mask,bits) \
349 	__pgprot((pgprot_val(prot) & ~(mask)) | (bits))
350 
351 /*
352  * Mark the prot value as uncacheable and unbufferable.
353  */
354 #define pgprot_noncached(prot) \
355 	__pgprot_modify(prot, PTE_ATTRINDX_MASK, PTE_ATTRINDX(MT_DEVICE_nGnRnE) | PTE_PXN | PTE_UXN)
356 #define pgprot_writecombine(prot) \
357 	__pgprot_modify(prot, PTE_ATTRINDX_MASK, PTE_ATTRINDX(MT_NORMAL_NC) | PTE_PXN | PTE_UXN)
358 #define pgprot_device(prot) \
359 	__pgprot_modify(prot, PTE_ATTRINDX_MASK, PTE_ATTRINDX(MT_DEVICE_nGnRE) | PTE_PXN | PTE_UXN)
360 #define __HAVE_PHYS_MEM_ACCESS_PROT
361 struct file;
362 extern pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
363 				     unsigned long size, pgprot_t vma_prot);
364 
365 #define pmd_none(pmd)		(!pmd_val(pmd))
366 
367 #define pmd_bad(pmd)		(!(pmd_val(pmd) & PMD_TABLE_BIT))
368 
369 #define pmd_table(pmd)		((pmd_val(pmd) & PMD_TYPE_MASK) == \
370 				 PMD_TYPE_TABLE)
371 #define pmd_sect(pmd)		((pmd_val(pmd) & PMD_TYPE_MASK) == \
372 				 PMD_TYPE_SECT)
373 
374 #if defined(CONFIG_ARM64_64K_PAGES) || CONFIG_PGTABLE_LEVELS < 3
375 #define pud_sect(pud)		(0)
376 #define pud_table(pud)		(1)
377 #else
378 #define pud_sect(pud)		((pud_val(pud) & PUD_TYPE_MASK) == \
379 				 PUD_TYPE_SECT)
380 #define pud_table(pud)		((pud_val(pud) & PUD_TYPE_MASK) == \
381 				 PUD_TYPE_TABLE)
382 #endif
383 
384 static inline void set_pmd(pmd_t *pmdp, pmd_t pmd)
385 {
386 	*pmdp = pmd;
387 	dsb(ishst);
388 	isb();
389 }
390 
391 static inline void pmd_clear(pmd_t *pmdp)
392 {
393 	set_pmd(pmdp, __pmd(0));
394 }
395 
396 static inline phys_addr_t pmd_page_paddr(pmd_t pmd)
397 {
398 	return pmd_val(pmd) & PHYS_MASK & (s32)PAGE_MASK;
399 }
400 
401 /* Find an entry in the third-level page table. */
402 #define pte_index(addr)		(((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
403 
404 #define pte_offset_phys(dir,addr)	(pmd_page_paddr(READ_ONCE(*(dir))) + pte_index(addr) * sizeof(pte_t))
405 #define pte_offset_kernel(dir,addr)	((pte_t *)__va(pte_offset_phys((dir), (addr))))
406 
407 #define pte_offset_map(dir,addr)	pte_offset_kernel((dir), (addr))
408 #define pte_offset_map_nested(dir,addr)	pte_offset_kernel((dir), (addr))
409 #define pte_unmap(pte)			do { } while (0)
410 #define pte_unmap_nested(pte)		do { } while (0)
411 
412 #define pte_set_fixmap(addr)		((pte_t *)set_fixmap_offset(FIX_PTE, addr))
413 #define pte_set_fixmap_offset(pmd, addr)	pte_set_fixmap(pte_offset_phys(pmd, addr))
414 #define pte_clear_fixmap()		clear_fixmap(FIX_PTE)
415 
416 #define pmd_page(pmd)		pfn_to_page(__phys_to_pfn(pmd_val(pmd) & PHYS_MASK))
417 
418 /* use ONLY for statically allocated translation tables */
419 #define pte_offset_kimg(dir,addr)	((pte_t *)__phys_to_kimg(pte_offset_phys((dir), (addr))))
420 
421 /*
422  * Conversion functions: convert a page and protection to a page entry,
423  * and a page entry and page directory to the page they refer to.
424  */
425 #define mk_pte(page,prot)	pfn_pte(page_to_pfn(page),prot)
426 
427 #if CONFIG_PGTABLE_LEVELS > 2
428 
429 #define pmd_ERROR(pmd)		__pmd_error(__FILE__, __LINE__, pmd_val(pmd))
430 
431 #define pud_none(pud)		(!pud_val(pud))
432 #define pud_bad(pud)		(!(pud_val(pud) & PUD_TABLE_BIT))
433 #define pud_present(pud)	pte_present(pud_pte(pud))
434 
435 static inline void set_pud(pud_t *pudp, pud_t pud)
436 {
437 	*pudp = pud;
438 	dsb(ishst);
439 	isb();
440 }
441 
442 static inline void pud_clear(pud_t *pudp)
443 {
444 	set_pud(pudp, __pud(0));
445 }
446 
447 static inline phys_addr_t pud_page_paddr(pud_t pud)
448 {
449 	return pud_val(pud) & PHYS_MASK & (s32)PAGE_MASK;
450 }
451 
452 /* Find an entry in the second-level page table. */
453 #define pmd_index(addr)		(((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1))
454 
455 #define pmd_offset_phys(dir, addr)	(pud_page_paddr(*(dir)) + pmd_index(addr) * sizeof(pmd_t))
456 #define pmd_offset(dir, addr)		((pmd_t *)__va(pmd_offset_phys((dir), (addr))))
457 
458 #define pmd_set_fixmap(addr)		((pmd_t *)set_fixmap_offset(FIX_PMD, addr))
459 #define pmd_set_fixmap_offset(pud, addr)	pmd_set_fixmap(pmd_offset_phys(pud, addr))
460 #define pmd_clear_fixmap()		clear_fixmap(FIX_PMD)
461 
462 #define pud_page(pud)		pfn_to_page(__phys_to_pfn(pud_val(pud) & PHYS_MASK))
463 
464 /* use ONLY for statically allocated translation tables */
465 #define pmd_offset_kimg(dir,addr)	((pmd_t *)__phys_to_kimg(pmd_offset_phys((dir), (addr))))
466 
467 #else
468 
469 #define pud_page_paddr(pud)	({ BUILD_BUG(); 0; })
470 
471 /* Match pmd_offset folding in <asm/generic/pgtable-nopmd.h> */
472 #define pmd_set_fixmap(addr)		NULL
473 #define pmd_set_fixmap_offset(pudp, addr)	((pmd_t *)pudp)
474 #define pmd_clear_fixmap()
475 
476 #define pmd_offset_kimg(dir,addr)	((pmd_t *)dir)
477 
478 #endif	/* CONFIG_PGTABLE_LEVELS > 2 */
479 
480 #if CONFIG_PGTABLE_LEVELS > 3
481 
482 #define pud_ERROR(pud)		__pud_error(__FILE__, __LINE__, pud_val(pud))
483 
484 #define pgd_none(pgd)		(!pgd_val(pgd))
485 #define pgd_bad(pgd)		(!(pgd_val(pgd) & 2))
486 #define pgd_present(pgd)	(pgd_val(pgd))
487 
488 static inline void set_pgd(pgd_t *pgdp, pgd_t pgd)
489 {
490 	*pgdp = pgd;
491 	dsb(ishst);
492 }
493 
494 static inline void pgd_clear(pgd_t *pgdp)
495 {
496 	set_pgd(pgdp, __pgd(0));
497 }
498 
499 static inline phys_addr_t pgd_page_paddr(pgd_t pgd)
500 {
501 	return pgd_val(pgd) & PHYS_MASK & (s32)PAGE_MASK;
502 }
503 
504 /* Find an entry in the frst-level page table. */
505 #define pud_index(addr)		(((addr) >> PUD_SHIFT) & (PTRS_PER_PUD - 1))
506 
507 #define pud_offset_phys(dir, addr)	(pgd_page_paddr(*(dir)) + pud_index(addr) * sizeof(pud_t))
508 #define pud_offset(dir, addr)		((pud_t *)__va(pud_offset_phys((dir), (addr))))
509 
510 #define pud_set_fixmap(addr)		((pud_t *)set_fixmap_offset(FIX_PUD, addr))
511 #define pud_set_fixmap_offset(pgd, addr)	pud_set_fixmap(pud_offset_phys(pgd, addr))
512 #define pud_clear_fixmap()		clear_fixmap(FIX_PUD)
513 
514 #define pgd_page(pgd)		pfn_to_page(__phys_to_pfn(pgd_val(pgd) & PHYS_MASK))
515 
516 /* use ONLY for statically allocated translation tables */
517 #define pud_offset_kimg(dir,addr)	((pud_t *)__phys_to_kimg(pud_offset_phys((dir), (addr))))
518 
519 #else
520 
521 #define pgd_page_paddr(pgd)	({ BUILD_BUG(); 0;})
522 
523 /* Match pud_offset folding in <asm/generic/pgtable-nopud.h> */
524 #define pud_set_fixmap(addr)		NULL
525 #define pud_set_fixmap_offset(pgdp, addr)	((pud_t *)pgdp)
526 #define pud_clear_fixmap()
527 
528 #define pud_offset_kimg(dir,addr)	((pud_t *)dir)
529 
530 #endif  /* CONFIG_PGTABLE_LEVELS > 3 */
531 
532 #define pgd_ERROR(pgd)		__pgd_error(__FILE__, __LINE__, pgd_val(pgd))
533 
534 /* to find an entry in a page-table-directory */
535 #define pgd_index(addr)		(((addr) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
536 
537 #define pgd_offset_raw(pgd, addr)	((pgd) + pgd_index(addr))
538 
539 #define pgd_offset(mm, addr)	(pgd_offset_raw((mm)->pgd, (addr)))
540 
541 /* to find an entry in a kernel page-table-directory */
542 #define pgd_offset_k(addr)	pgd_offset(&init_mm, addr)
543 
544 #define pgd_set_fixmap(addr)	((pgd_t *)set_fixmap_offset(FIX_PGD, addr))
545 #define pgd_clear_fixmap()	clear_fixmap(FIX_PGD)
546 
547 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
548 {
549 	const pteval_t mask = PTE_USER | PTE_PXN | PTE_UXN | PTE_RDONLY |
550 			      PTE_PROT_NONE | PTE_VALID | PTE_WRITE;
551 	/* preserve the hardware dirty information */
552 	if (pte_hw_dirty(pte))
553 		pte = pte_mkdirty(pte);
554 	pte_val(pte) = (pte_val(pte) & ~mask) | (pgprot_val(newprot) & mask);
555 	return pte;
556 }
557 
558 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
559 {
560 	return pte_pmd(pte_modify(pmd_pte(pmd), newprot));
561 }
562 
563 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
564 extern int ptep_set_access_flags(struct vm_area_struct *vma,
565 				 unsigned long address, pte_t *ptep,
566 				 pte_t entry, int dirty);
567 
568 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
569 #define __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
570 static inline int pmdp_set_access_flags(struct vm_area_struct *vma,
571 					unsigned long address, pmd_t *pmdp,
572 					pmd_t entry, int dirty)
573 {
574 	return ptep_set_access_flags(vma, address, (pte_t *)pmdp, pmd_pte(entry), dirty);
575 }
576 #endif
577 
578 /*
579  * Atomic pte/pmd modifications.
580  */
581 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
582 static inline int __ptep_test_and_clear_young(pte_t *ptep)
583 {
584 	pte_t old_pte, pte;
585 
586 	pte = READ_ONCE(*ptep);
587 	do {
588 		old_pte = pte;
589 		pte = pte_mkold(pte);
590 		pte_val(pte) = cmpxchg_relaxed(&pte_val(*ptep),
591 					       pte_val(old_pte), pte_val(pte));
592 	} while (pte_val(pte) != pte_val(old_pte));
593 
594 	return pte_young(pte);
595 }
596 
597 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
598 					    unsigned long address,
599 					    pte_t *ptep)
600 {
601 	return __ptep_test_and_clear_young(ptep);
602 }
603 
604 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
605 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
606 static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
607 					    unsigned long address,
608 					    pmd_t *pmdp)
609 {
610 	return ptep_test_and_clear_young(vma, address, (pte_t *)pmdp);
611 }
612 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
613 
614 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
615 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
616 				       unsigned long address, pte_t *ptep)
617 {
618 	return __pte(xchg_relaxed(&pte_val(*ptep), 0));
619 }
620 
621 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
622 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
623 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm,
624 					    unsigned long address, pmd_t *pmdp)
625 {
626 	return pte_pmd(ptep_get_and_clear(mm, address, (pte_t *)pmdp));
627 }
628 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
629 
630 /*
631  * ptep_set_wrprotect - mark read-only while preserving the hardware update of
632  * the Access Flag.
633  */
634 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
635 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long address, pte_t *ptep)
636 {
637 	pte_t old_pte, pte;
638 
639 	/*
640 	 * ptep_set_wrprotect() is only called on CoW mappings which are
641 	 * private (!VM_SHARED) with the pte either read-only (!PTE_WRITE &&
642 	 * PTE_RDONLY) or writable and software-dirty (PTE_WRITE &&
643 	 * !PTE_RDONLY && PTE_DIRTY); see is_cow_mapping() and
644 	 * protection_map[]. There is no race with the hardware update of the
645 	 * dirty state: clearing of PTE_RDONLY when PTE_WRITE (a.k.a. PTE_DBM)
646 	 * is set.
647 	 */
648 	VM_WARN_ONCE(pte_write(*ptep) && !pte_dirty(*ptep),
649 		     "%s: potential race with hardware DBM", __func__);
650 	pte = READ_ONCE(*ptep);
651 	do {
652 		old_pte = pte;
653 		pte = pte_wrprotect(pte);
654 		pte_val(pte) = cmpxchg_relaxed(&pte_val(*ptep),
655 					       pte_val(old_pte), pte_val(pte));
656 	} while (pte_val(pte) != pte_val(old_pte));
657 }
658 
659 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
660 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
661 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
662 				      unsigned long address, pmd_t *pmdp)
663 {
664 	ptep_set_wrprotect(mm, address, (pte_t *)pmdp);
665 }
666 #endif
667 
668 extern pgd_t swapper_pg_dir[PTRS_PER_PGD];
669 extern pgd_t idmap_pg_dir[PTRS_PER_PGD];
670 
671 /*
672  * Encode and decode a swap entry:
673  *	bits 0-1:	present (must be zero)
674  *	bits 2-7:	swap type
675  *	bits 8-57:	swap offset
676  *	bit  58:	PTE_PROT_NONE (must be zero)
677  */
678 #define __SWP_TYPE_SHIFT	2
679 #define __SWP_TYPE_BITS		6
680 #define __SWP_OFFSET_BITS	50
681 #define __SWP_TYPE_MASK		((1 << __SWP_TYPE_BITS) - 1)
682 #define __SWP_OFFSET_SHIFT	(__SWP_TYPE_BITS + __SWP_TYPE_SHIFT)
683 #define __SWP_OFFSET_MASK	((1UL << __SWP_OFFSET_BITS) - 1)
684 
685 #define __swp_type(x)		(((x).val >> __SWP_TYPE_SHIFT) & __SWP_TYPE_MASK)
686 #define __swp_offset(x)		(((x).val >> __SWP_OFFSET_SHIFT) & __SWP_OFFSET_MASK)
687 #define __swp_entry(type,offset) ((swp_entry_t) { ((type) << __SWP_TYPE_SHIFT) | ((offset) << __SWP_OFFSET_SHIFT) })
688 
689 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
690 #define __swp_entry_to_pte(swp)	((pte_t) { (swp).val })
691 
692 /*
693  * Ensure that there are not more swap files than can be encoded in the kernel
694  * PTEs.
695  */
696 #define MAX_SWAPFILES_CHECK() BUILD_BUG_ON(MAX_SWAPFILES_SHIFT > __SWP_TYPE_BITS)
697 
698 extern int kern_addr_valid(unsigned long addr);
699 
700 #include <asm-generic/pgtable.h>
701 
702 void pgd_cache_init(void);
703 #define pgtable_cache_init	pgd_cache_init
704 
705 /*
706  * On AArch64, the cache coherency is handled via the set_pte_at() function.
707  */
708 static inline void update_mmu_cache(struct vm_area_struct *vma,
709 				    unsigned long addr, pte_t *ptep)
710 {
711 	/*
712 	 * We don't do anything here, so there's a very small chance of
713 	 * us retaking a user fault which we just fixed up. The alternative
714 	 * is doing a dsb(ishst), but that penalises the fastpath.
715 	 */
716 }
717 
718 #define update_mmu_cache_pmd(vma, address, pmd) do { } while (0)
719 
720 #define kc_vaddr_to_offset(v)	((v) & ~VA_START)
721 #define kc_offset_to_vaddr(o)	((o) | VA_START)
722 
723 #endif /* !__ASSEMBLY__ */
724 
725 #endif /* __ASM_PGTABLE_H */
726