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 #include <linux/mm_types.h>
46 #include <linux/sched.h>
47 
48 extern void __pte_error(const char *file, int line, unsigned long val);
49 extern void __pmd_error(const char *file, int line, unsigned long val);
50 extern void __pud_error(const char *file, int line, unsigned long val);
51 extern void __pgd_error(const char *file, int line, unsigned long val);
52 
53 /*
54  * ZERO_PAGE is a global shared page that is always zero: used
55  * for zero-mapped memory areas etc..
56  */
57 extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)];
58 #define ZERO_PAGE(vaddr)	phys_to_page(__pa_symbol(empty_zero_page))
59 
60 #define pte_ERROR(pte)		__pte_error(__FILE__, __LINE__, pte_val(pte))
61 
62 /*
63  * Macros to convert between a physical address and its placement in a
64  * page table entry, taking care of 52-bit addresses.
65  */
66 #ifdef CONFIG_ARM64_PA_BITS_52
67 #define __pte_to_phys(pte)	\
68 	((pte_val(pte) & PTE_ADDR_LOW) | ((pte_val(pte) & PTE_ADDR_HIGH) << 36))
69 #define __phys_to_pte_val(phys)	(((phys) | ((phys) >> 36)) & PTE_ADDR_MASK)
70 #else
71 #define __pte_to_phys(pte)	(pte_val(pte) & PTE_ADDR_MASK)
72 #define __phys_to_pte_val(phys)	(phys)
73 #endif
74 
75 #define pte_pfn(pte)		(__pte_to_phys(pte) >> PAGE_SHIFT)
76 #define pfn_pte(pfn,prot)	\
77 	__pte(__phys_to_pte_val((phys_addr_t)(pfn) << PAGE_SHIFT) | pgprot_val(prot))
78 
79 #define pte_none(pte)		(!pte_val(pte))
80 #define pte_clear(mm,addr,ptep)	set_pte(ptep, __pte(0))
81 #define pte_page(pte)		(pfn_to_page(pte_pfn(pte)))
82 
83 /*
84  * The following only work if pte_present(). Undefined behaviour otherwise.
85  */
86 #define pte_present(pte)	(!!(pte_val(pte) & (PTE_VALID | PTE_PROT_NONE)))
87 #define pte_young(pte)		(!!(pte_val(pte) & PTE_AF))
88 #define pte_special(pte)	(!!(pte_val(pte) & PTE_SPECIAL))
89 #define pte_write(pte)		(!!(pte_val(pte) & PTE_WRITE))
90 #define pte_user_exec(pte)	(!(pte_val(pte) & PTE_UXN))
91 #define pte_cont(pte)		(!!(pte_val(pte) & PTE_CONT))
92 
93 #define pte_cont_addr_end(addr, end)						\
94 ({	unsigned long __boundary = ((addr) + CONT_PTE_SIZE) & CONT_PTE_MASK;	\
95 	(__boundary - 1 < (end) - 1) ? __boundary : (end);			\
96 })
97 
98 #define pmd_cont_addr_end(addr, end)						\
99 ({	unsigned long __boundary = ((addr) + CONT_PMD_SIZE) & CONT_PMD_MASK;	\
100 	(__boundary - 1 < (end) - 1) ? __boundary : (end);			\
101 })
102 
103 #define pte_hw_dirty(pte)	(pte_write(pte) && !(pte_val(pte) & PTE_RDONLY))
104 #define pte_sw_dirty(pte)	(!!(pte_val(pte) & PTE_DIRTY))
105 #define pte_dirty(pte)		(pte_sw_dirty(pte) || pte_hw_dirty(pte))
106 
107 #define pte_valid(pte)		(!!(pte_val(pte) & PTE_VALID))
108 /*
109  * Execute-only user mappings do not have the PTE_USER bit set. All valid
110  * kernel mappings have the PTE_UXN bit set.
111  */
112 #define pte_valid_not_user(pte) \
113 	((pte_val(pte) & (PTE_VALID | PTE_USER | PTE_UXN)) == (PTE_VALID | PTE_UXN))
114 #define pte_valid_young(pte) \
115 	((pte_val(pte) & (PTE_VALID | PTE_AF)) == (PTE_VALID | PTE_AF))
116 #define pte_valid_user(pte) \
117 	((pte_val(pte) & (PTE_VALID | PTE_USER)) == (PTE_VALID | PTE_USER))
118 
119 /*
120  * Could the pte be present in the TLB? We must check mm_tlb_flush_pending
121  * so that we don't erroneously return false for pages that have been
122  * remapped as PROT_NONE but are yet to be flushed from the TLB.
123  */
124 #define pte_accessible(mm, pte)	\
125 	(mm_tlb_flush_pending(mm) ? pte_present(pte) : pte_valid_young(pte))
126 
127 /*
128  * p??_access_permitted() is true for valid user mappings (subject to the
129  * write permission check) other than user execute-only which do not have the
130  * PTE_USER bit set. PROT_NONE mappings do not have the PTE_VALID bit set.
131  */
132 #define pte_access_permitted(pte, write) \
133 	(pte_valid_user(pte) && (!(write) || pte_write(pte)))
134 #define pmd_access_permitted(pmd, write) \
135 	(pte_access_permitted(pmd_pte(pmd), (write)))
136 #define pud_access_permitted(pud, write) \
137 	(pte_access_permitted(pud_pte(pud), (write)))
138 
clear_pte_bit(pte_t pte,pgprot_t prot)139 static inline pte_t clear_pte_bit(pte_t pte, pgprot_t prot)
140 {
141 	pte_val(pte) &= ~pgprot_val(prot);
142 	return pte;
143 }
144 
set_pte_bit(pte_t pte,pgprot_t prot)145 static inline pte_t set_pte_bit(pte_t pte, pgprot_t prot)
146 {
147 	pte_val(pte) |= pgprot_val(prot);
148 	return pte;
149 }
150 
pte_wrprotect(pte_t pte)151 static inline pte_t pte_wrprotect(pte_t pte)
152 {
153 	pte = clear_pte_bit(pte, __pgprot(PTE_WRITE));
154 	pte = set_pte_bit(pte, __pgprot(PTE_RDONLY));
155 	return pte;
156 }
157 
pte_mkwrite(pte_t pte)158 static inline pte_t pte_mkwrite(pte_t pte)
159 {
160 	pte = set_pte_bit(pte, __pgprot(PTE_WRITE));
161 	pte = clear_pte_bit(pte, __pgprot(PTE_RDONLY));
162 	return pte;
163 }
164 
pte_mkclean(pte_t pte)165 static inline pte_t pte_mkclean(pte_t pte)
166 {
167 	pte = clear_pte_bit(pte, __pgprot(PTE_DIRTY));
168 	pte = set_pte_bit(pte, __pgprot(PTE_RDONLY));
169 
170 	return pte;
171 }
172 
pte_mkdirty(pte_t pte)173 static inline pte_t pte_mkdirty(pte_t pte)
174 {
175 	pte = set_pte_bit(pte, __pgprot(PTE_DIRTY));
176 
177 	if (pte_write(pte))
178 		pte = clear_pte_bit(pte, __pgprot(PTE_RDONLY));
179 
180 	return pte;
181 }
182 
pte_mkold(pte_t pte)183 static inline pte_t pte_mkold(pte_t pte)
184 {
185 	return clear_pte_bit(pte, __pgprot(PTE_AF));
186 }
187 
pte_mkyoung(pte_t pte)188 static inline pte_t pte_mkyoung(pte_t pte)
189 {
190 	return set_pte_bit(pte, __pgprot(PTE_AF));
191 }
192 
pte_mkspecial(pte_t pte)193 static inline pte_t pte_mkspecial(pte_t pte)
194 {
195 	return set_pte_bit(pte, __pgprot(PTE_SPECIAL));
196 }
197 
pte_mkcont(pte_t pte)198 static inline pte_t pte_mkcont(pte_t pte)
199 {
200 	pte = set_pte_bit(pte, __pgprot(PTE_CONT));
201 	return set_pte_bit(pte, __pgprot(PTE_TYPE_PAGE));
202 }
203 
pte_mknoncont(pte_t pte)204 static inline pte_t pte_mknoncont(pte_t pte)
205 {
206 	return clear_pte_bit(pte, __pgprot(PTE_CONT));
207 }
208 
pte_mkpresent(pte_t pte)209 static inline pte_t pte_mkpresent(pte_t pte)
210 {
211 	return set_pte_bit(pte, __pgprot(PTE_VALID));
212 }
213 
pmd_mkcont(pmd_t pmd)214 static inline pmd_t pmd_mkcont(pmd_t pmd)
215 {
216 	return __pmd(pmd_val(pmd) | PMD_SECT_CONT);
217 }
218 
set_pte(pte_t * ptep,pte_t pte)219 static inline void set_pte(pte_t *ptep, pte_t pte)
220 {
221 	WRITE_ONCE(*ptep, pte);
222 
223 	/*
224 	 * Only if the new pte is valid and kernel, otherwise TLB maintenance
225 	 * or update_mmu_cache() have the necessary barriers.
226 	 */
227 	if (pte_valid_not_user(pte))
228 		dsb(ishst);
229 }
230 
231 extern void __sync_icache_dcache(pte_t pteval);
232 
233 /*
234  * PTE bits configuration in the presence of hardware Dirty Bit Management
235  * (PTE_WRITE == PTE_DBM):
236  *
237  * Dirty  Writable | PTE_RDONLY  PTE_WRITE  PTE_DIRTY (sw)
238  *   0      0      |   1           0          0
239  *   0      1      |   1           1          0
240  *   1      0      |   1           0          1
241  *   1      1      |   0           1          x
242  *
243  * When hardware DBM is not present, the sofware PTE_DIRTY bit is updated via
244  * the page fault mechanism. Checking the dirty status of a pte becomes:
245  *
246  *   PTE_DIRTY || (PTE_WRITE && !PTE_RDONLY)
247  */
set_pte_at(struct mm_struct * mm,unsigned long addr,pte_t * ptep,pte_t pte)248 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
249 			      pte_t *ptep, pte_t pte)
250 {
251 	pte_t old_pte;
252 
253 	if (pte_present(pte) && pte_user_exec(pte) && !pte_special(pte))
254 		__sync_icache_dcache(pte);
255 
256 	/*
257 	 * If the existing pte is valid, check for potential race with
258 	 * hardware updates of the pte (ptep_set_access_flags safely changes
259 	 * valid ptes without going through an invalid entry).
260 	 */
261 	old_pte = READ_ONCE(*ptep);
262 	if (IS_ENABLED(CONFIG_DEBUG_VM) && pte_valid(old_pte) && pte_valid(pte) &&
263 	   (mm == current->active_mm || atomic_read(&mm->mm_users) > 1)) {
264 		VM_WARN_ONCE(!pte_young(pte),
265 			     "%s: racy access flag clearing: 0x%016llx -> 0x%016llx",
266 			     __func__, pte_val(old_pte), pte_val(pte));
267 		VM_WARN_ONCE(pte_write(old_pte) && !pte_dirty(pte),
268 			     "%s: racy dirty state clearing: 0x%016llx -> 0x%016llx",
269 			     __func__, pte_val(old_pte), pte_val(pte));
270 	}
271 
272 	set_pte(ptep, pte);
273 }
274 
275 #define __HAVE_ARCH_PTE_SAME
pte_same(pte_t pte_a,pte_t pte_b)276 static inline int pte_same(pte_t pte_a, pte_t pte_b)
277 {
278 	pteval_t lhs, rhs;
279 
280 	lhs = pte_val(pte_a);
281 	rhs = pte_val(pte_b);
282 
283 	if (pte_present(pte_a))
284 		lhs &= ~PTE_RDONLY;
285 
286 	if (pte_present(pte_b))
287 		rhs &= ~PTE_RDONLY;
288 
289 	return (lhs == rhs);
290 }
291 
292 /*
293  * Huge pte definitions.
294  */
295 #define pte_huge(pte)		(!(pte_val(pte) & PTE_TABLE_BIT))
296 #define pte_mkhuge(pte)		(__pte(pte_val(pte) & ~PTE_TABLE_BIT))
297 
298 /*
299  * Hugetlb definitions.
300  */
301 #define HUGE_MAX_HSTATE		4
302 #define HPAGE_SHIFT		PMD_SHIFT
303 #define HPAGE_SIZE		(_AC(1, UL) << HPAGE_SHIFT)
304 #define HPAGE_MASK		(~(HPAGE_SIZE - 1))
305 #define HUGETLB_PAGE_ORDER	(HPAGE_SHIFT - PAGE_SHIFT)
306 
pgd_pte(pgd_t pgd)307 static inline pte_t pgd_pte(pgd_t pgd)
308 {
309 	return __pte(pgd_val(pgd));
310 }
311 
pud_pte(pud_t pud)312 static inline pte_t pud_pte(pud_t pud)
313 {
314 	return __pte(pud_val(pud));
315 }
316 
pud_pmd(pud_t pud)317 static inline pmd_t pud_pmd(pud_t pud)
318 {
319 	return __pmd(pud_val(pud));
320 }
321 
pmd_pte(pmd_t pmd)322 static inline pte_t pmd_pte(pmd_t pmd)
323 {
324 	return __pte(pmd_val(pmd));
325 }
326 
pte_pmd(pte_t pte)327 static inline pmd_t pte_pmd(pte_t pte)
328 {
329 	return __pmd(pte_val(pte));
330 }
331 
mk_sect_prot(pgprot_t prot)332 static inline pgprot_t mk_sect_prot(pgprot_t prot)
333 {
334 	return __pgprot(pgprot_val(prot) & ~PTE_TABLE_BIT);
335 }
336 
337 #ifdef CONFIG_NUMA_BALANCING
338 /*
339  * See the comment in include/asm-generic/pgtable.h
340  */
pte_protnone(pte_t pte)341 static inline int pte_protnone(pte_t pte)
342 {
343 	return (pte_val(pte) & (PTE_VALID | PTE_PROT_NONE)) == PTE_PROT_NONE;
344 }
345 
pmd_protnone(pmd_t pmd)346 static inline int pmd_protnone(pmd_t pmd)
347 {
348 	return pte_protnone(pmd_pte(pmd));
349 }
350 #endif
351 
352 /*
353  * THP definitions.
354  */
355 
356 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
357 #define pmd_trans_huge(pmd)	(pmd_val(pmd) && !(pmd_val(pmd) & PMD_TABLE_BIT))
358 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
359 
360 #define pmd_present(pmd)	pte_present(pmd_pte(pmd))
361 #define pmd_dirty(pmd)		pte_dirty(pmd_pte(pmd))
362 #define pmd_young(pmd)		pte_young(pmd_pte(pmd))
363 #define pmd_wrprotect(pmd)	pte_pmd(pte_wrprotect(pmd_pte(pmd)))
364 #define pmd_mkold(pmd)		pte_pmd(pte_mkold(pmd_pte(pmd)))
365 #define pmd_mkwrite(pmd)	pte_pmd(pte_mkwrite(pmd_pte(pmd)))
366 #define pmd_mkclean(pmd)	pte_pmd(pte_mkclean(pmd_pte(pmd)))
367 #define pmd_mkdirty(pmd)	pte_pmd(pte_mkdirty(pmd_pte(pmd)))
368 #define pmd_mkyoung(pmd)	pte_pmd(pte_mkyoung(pmd_pte(pmd)))
369 #define pmd_mknotpresent(pmd)	(__pmd(pmd_val(pmd) & ~PMD_SECT_VALID))
370 
371 #define pmd_thp_or_huge(pmd)	(pmd_huge(pmd) || pmd_trans_huge(pmd))
372 
373 #define pmd_write(pmd)		pte_write(pmd_pte(pmd))
374 
375 #define pmd_mkhuge(pmd)		(__pmd(pmd_val(pmd) & ~PMD_TABLE_BIT))
376 
377 #define __pmd_to_phys(pmd)	__pte_to_phys(pmd_pte(pmd))
378 #define __phys_to_pmd_val(phys)	__phys_to_pte_val(phys)
379 #define pmd_pfn(pmd)		((__pmd_to_phys(pmd) & PMD_MASK) >> PAGE_SHIFT)
380 #define pfn_pmd(pfn,prot)	__pmd(__phys_to_pmd_val((phys_addr_t)(pfn) << PAGE_SHIFT) | pgprot_val(prot))
381 #define mk_pmd(page,prot)	pfn_pmd(page_to_pfn(page),prot)
382 
383 #define pud_write(pud)		pte_write(pud_pte(pud))
384 
385 #define __pud_to_phys(pud)	__pte_to_phys(pud_pte(pud))
386 #define __phys_to_pud_val(phys)	__phys_to_pte_val(phys)
387 #define pud_pfn(pud)		((__pud_to_phys(pud) & PUD_MASK) >> PAGE_SHIFT)
388 #define pfn_pud(pfn,prot)	__pud(__phys_to_pud_val((phys_addr_t)(pfn) << PAGE_SHIFT) | pgprot_val(prot))
389 
390 #define set_pmd_at(mm, addr, pmdp, pmd)	set_pte_at(mm, addr, (pte_t *)pmdp, pmd_pte(pmd))
391 
392 #define __pgd_to_phys(pgd)	__pte_to_phys(pgd_pte(pgd))
393 #define __phys_to_pgd_val(phys)	__phys_to_pte_val(phys)
394 
395 #define __pgprot_modify(prot,mask,bits) \
396 	__pgprot((pgprot_val(prot) & ~(mask)) | (bits))
397 
398 /*
399  * Mark the prot value as uncacheable and unbufferable.
400  */
401 #define pgprot_noncached(prot) \
402 	__pgprot_modify(prot, PTE_ATTRINDX_MASK, PTE_ATTRINDX(MT_DEVICE_nGnRnE) | PTE_PXN | PTE_UXN)
403 #define pgprot_writecombine(prot) \
404 	__pgprot_modify(prot, PTE_ATTRINDX_MASK, PTE_ATTRINDX(MT_NORMAL_NC) | PTE_PXN | PTE_UXN)
405 #define pgprot_device(prot) \
406 	__pgprot_modify(prot, PTE_ATTRINDX_MASK, PTE_ATTRINDX(MT_DEVICE_nGnRE) | PTE_PXN | PTE_UXN)
407 #define __HAVE_PHYS_MEM_ACCESS_PROT
408 struct file;
409 extern pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
410 				     unsigned long size, pgprot_t vma_prot);
411 
412 #define pmd_none(pmd)		(!pmd_val(pmd))
413 
414 #define pmd_bad(pmd)		(!(pmd_val(pmd) & PMD_TABLE_BIT))
415 
416 #define pmd_table(pmd)		((pmd_val(pmd) & PMD_TYPE_MASK) == \
417 				 PMD_TYPE_TABLE)
418 #define pmd_sect(pmd)		((pmd_val(pmd) & PMD_TYPE_MASK) == \
419 				 PMD_TYPE_SECT)
420 
421 #if defined(CONFIG_ARM64_64K_PAGES) || CONFIG_PGTABLE_LEVELS < 3
422 #define pud_sect(pud)		(0)
423 #define pud_table(pud)		(1)
424 #else
425 #define pud_sect(pud)		((pud_val(pud) & PUD_TYPE_MASK) == \
426 				 PUD_TYPE_SECT)
427 #define pud_table(pud)		((pud_val(pud) & PUD_TYPE_MASK) == \
428 				 PUD_TYPE_TABLE)
429 #endif
430 
set_pmd(pmd_t * pmdp,pmd_t pmd)431 static inline void set_pmd(pmd_t *pmdp, pmd_t pmd)
432 {
433 	WRITE_ONCE(*pmdp, pmd);
434 	dsb(ishst);
435 }
436 
pmd_clear(pmd_t * pmdp)437 static inline void pmd_clear(pmd_t *pmdp)
438 {
439 	set_pmd(pmdp, __pmd(0));
440 }
441 
pmd_page_paddr(pmd_t pmd)442 static inline phys_addr_t pmd_page_paddr(pmd_t pmd)
443 {
444 	return __pmd_to_phys(pmd);
445 }
446 
447 /* Find an entry in the third-level page table. */
448 #define pte_index(addr)		(((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
449 
450 #define pte_offset_phys(dir,addr)	(pmd_page_paddr(READ_ONCE(*(dir))) + pte_index(addr) * sizeof(pte_t))
451 #define pte_offset_kernel(dir,addr)	((pte_t *)__va(pte_offset_phys((dir), (addr))))
452 
453 #define pte_offset_map(dir,addr)	pte_offset_kernel((dir), (addr))
454 #define pte_offset_map_nested(dir,addr)	pte_offset_kernel((dir), (addr))
455 #define pte_unmap(pte)			do { } while (0)
456 #define pte_unmap_nested(pte)		do { } while (0)
457 
458 #define pte_set_fixmap(addr)		((pte_t *)set_fixmap_offset(FIX_PTE, addr))
459 #define pte_set_fixmap_offset(pmd, addr)	pte_set_fixmap(pte_offset_phys(pmd, addr))
460 #define pte_clear_fixmap()		clear_fixmap(FIX_PTE)
461 
462 #define pmd_page(pmd)		pfn_to_page(__phys_to_pfn(__pmd_to_phys(pmd)))
463 
464 /* use ONLY for statically allocated translation tables */
465 #define pte_offset_kimg(dir,addr)	((pte_t *)__phys_to_kimg(pte_offset_phys((dir), (addr))))
466 
467 /*
468  * Conversion functions: convert a page and protection to a page entry,
469  * and a page entry and page directory to the page they refer to.
470  */
471 #define mk_pte(page,prot)	pfn_pte(page_to_pfn(page),prot)
472 
473 #if CONFIG_PGTABLE_LEVELS > 2
474 
475 #define pmd_ERROR(pmd)		__pmd_error(__FILE__, __LINE__, pmd_val(pmd))
476 
477 #define pud_none(pud)		(!pud_val(pud))
478 #define pud_bad(pud)		(!(pud_val(pud) & PUD_TABLE_BIT))
479 #define pud_present(pud)	pte_present(pud_pte(pud))
480 
set_pud(pud_t * pudp,pud_t pud)481 static inline void set_pud(pud_t *pudp, pud_t pud)
482 {
483 	WRITE_ONCE(*pudp, pud);
484 	dsb(ishst);
485 }
486 
pud_clear(pud_t * pudp)487 static inline void pud_clear(pud_t *pudp)
488 {
489 	set_pud(pudp, __pud(0));
490 }
491 
pud_page_paddr(pud_t pud)492 static inline phys_addr_t pud_page_paddr(pud_t pud)
493 {
494 	return __pud_to_phys(pud);
495 }
496 
497 /* Find an entry in the second-level page table. */
498 #define pmd_index(addr)		(((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1))
499 
500 #define pmd_offset_phys(dir, addr)	(pud_page_paddr(READ_ONCE(*(dir))) + pmd_index(addr) * sizeof(pmd_t))
501 #define pmd_offset(dir, addr)		((pmd_t *)__va(pmd_offset_phys((dir), (addr))))
502 
503 #define pmd_set_fixmap(addr)		((pmd_t *)set_fixmap_offset(FIX_PMD, addr))
504 #define pmd_set_fixmap_offset(pud, addr)	pmd_set_fixmap(pmd_offset_phys(pud, addr))
505 #define pmd_clear_fixmap()		clear_fixmap(FIX_PMD)
506 
507 #define pud_page(pud)		pfn_to_page(__phys_to_pfn(__pud_to_phys(pud)))
508 
509 /* use ONLY for statically allocated translation tables */
510 #define pmd_offset_kimg(dir,addr)	((pmd_t *)__phys_to_kimg(pmd_offset_phys((dir), (addr))))
511 
512 #else
513 
514 #define pud_page_paddr(pud)	({ BUILD_BUG(); 0; })
515 
516 /* Match pmd_offset folding in <asm/generic/pgtable-nopmd.h> */
517 #define pmd_set_fixmap(addr)		NULL
518 #define pmd_set_fixmap_offset(pudp, addr)	((pmd_t *)pudp)
519 #define pmd_clear_fixmap()
520 
521 #define pmd_offset_kimg(dir,addr)	((pmd_t *)dir)
522 
523 #endif	/* CONFIG_PGTABLE_LEVELS > 2 */
524 
525 #if CONFIG_PGTABLE_LEVELS > 3
526 
527 #define pud_ERROR(pud)		__pud_error(__FILE__, __LINE__, pud_val(pud))
528 
529 #define pgd_none(pgd)		(!pgd_val(pgd))
530 #define pgd_bad(pgd)		(!(pgd_val(pgd) & 2))
531 #define pgd_present(pgd)	(pgd_val(pgd))
532 
set_pgd(pgd_t * pgdp,pgd_t pgd)533 static inline void set_pgd(pgd_t *pgdp, pgd_t pgd)
534 {
535 	WRITE_ONCE(*pgdp, pgd);
536 	dsb(ishst);
537 }
538 
pgd_clear(pgd_t * pgdp)539 static inline void pgd_clear(pgd_t *pgdp)
540 {
541 	set_pgd(pgdp, __pgd(0));
542 }
543 
pgd_page_paddr(pgd_t pgd)544 static inline phys_addr_t pgd_page_paddr(pgd_t pgd)
545 {
546 	return __pgd_to_phys(pgd);
547 }
548 
549 /* Find an entry in the frst-level page table. */
550 #define pud_index(addr)		(((addr) >> PUD_SHIFT) & (PTRS_PER_PUD - 1))
551 
552 #define pud_offset_phys(dir, addr)	(pgd_page_paddr(READ_ONCE(*(dir))) + pud_index(addr) * sizeof(pud_t))
553 #define pud_offset(dir, addr)		((pud_t *)__va(pud_offset_phys((dir), (addr))))
554 
555 #define pud_set_fixmap(addr)		((pud_t *)set_fixmap_offset(FIX_PUD, addr))
556 #define pud_set_fixmap_offset(pgd, addr)	pud_set_fixmap(pud_offset_phys(pgd, addr))
557 #define pud_clear_fixmap()		clear_fixmap(FIX_PUD)
558 
559 #define pgd_page(pgd)		pfn_to_page(__phys_to_pfn(__pgd_to_phys(pgd)))
560 
561 /* use ONLY for statically allocated translation tables */
562 #define pud_offset_kimg(dir,addr)	((pud_t *)__phys_to_kimg(pud_offset_phys((dir), (addr))))
563 
564 #else
565 
566 #define pgd_page_paddr(pgd)	({ BUILD_BUG(); 0;})
567 
568 /* Match pud_offset folding in <asm/generic/pgtable-nopud.h> */
569 #define pud_set_fixmap(addr)		NULL
570 #define pud_set_fixmap_offset(pgdp, addr)	((pud_t *)pgdp)
571 #define pud_clear_fixmap()
572 
573 #define pud_offset_kimg(dir,addr)	((pud_t *)dir)
574 
575 #endif  /* CONFIG_PGTABLE_LEVELS > 3 */
576 
577 #define pgd_ERROR(pgd)		__pgd_error(__FILE__, __LINE__, pgd_val(pgd))
578 
579 /* to find an entry in a page-table-directory */
580 #define pgd_index(addr)		(((addr) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
581 
582 #define pgd_offset_raw(pgd, addr)	((pgd) + pgd_index(addr))
583 
584 #define pgd_offset(mm, addr)	(pgd_offset_raw((mm)->pgd, (addr)))
585 
586 /* to find an entry in a kernel page-table-directory */
587 #define pgd_offset_k(addr)	pgd_offset(&init_mm, addr)
588 
589 #define pgd_set_fixmap(addr)	((pgd_t *)set_fixmap_offset(FIX_PGD, addr))
590 #define pgd_clear_fixmap()	clear_fixmap(FIX_PGD)
591 
pte_modify(pte_t pte,pgprot_t newprot)592 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
593 {
594 	const pteval_t mask = PTE_USER | PTE_PXN | PTE_UXN | PTE_RDONLY |
595 			      PTE_PROT_NONE | PTE_VALID | PTE_WRITE;
596 	/* preserve the hardware dirty information */
597 	if (pte_hw_dirty(pte))
598 		pte = pte_mkdirty(pte);
599 	pte_val(pte) = (pte_val(pte) & ~mask) | (pgprot_val(newprot) & mask);
600 	return pte;
601 }
602 
pmd_modify(pmd_t pmd,pgprot_t newprot)603 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
604 {
605 	return pte_pmd(pte_modify(pmd_pte(pmd), newprot));
606 }
607 
608 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
609 extern int ptep_set_access_flags(struct vm_area_struct *vma,
610 				 unsigned long address, pte_t *ptep,
611 				 pte_t entry, int dirty);
612 
613 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
614 #define __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
pmdp_set_access_flags(struct vm_area_struct * vma,unsigned long address,pmd_t * pmdp,pmd_t entry,int dirty)615 static inline int pmdp_set_access_flags(struct vm_area_struct *vma,
616 					unsigned long address, pmd_t *pmdp,
617 					pmd_t entry, int dirty)
618 {
619 	return ptep_set_access_flags(vma, address, (pte_t *)pmdp, pmd_pte(entry), dirty);
620 }
621 #endif
622 
623 /*
624  * Atomic pte/pmd modifications.
625  */
626 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
__ptep_test_and_clear_young(pte_t * ptep)627 static inline int __ptep_test_and_clear_young(pte_t *ptep)
628 {
629 	pte_t old_pte, pte;
630 
631 	pte = READ_ONCE(*ptep);
632 	do {
633 		old_pte = pte;
634 		pte = pte_mkold(pte);
635 		pte_val(pte) = cmpxchg_relaxed(&pte_val(*ptep),
636 					       pte_val(old_pte), pte_val(pte));
637 	} while (pte_val(pte) != pte_val(old_pte));
638 
639 	return pte_young(pte);
640 }
641 
ptep_test_and_clear_young(struct vm_area_struct * vma,unsigned long address,pte_t * ptep)642 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
643 					    unsigned long address,
644 					    pte_t *ptep)
645 {
646 	return __ptep_test_and_clear_young(ptep);
647 }
648 
649 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
650 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
pmdp_test_and_clear_young(struct vm_area_struct * vma,unsigned long address,pmd_t * pmdp)651 static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
652 					    unsigned long address,
653 					    pmd_t *pmdp)
654 {
655 	return ptep_test_and_clear_young(vma, address, (pte_t *)pmdp);
656 }
657 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
658 
659 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
ptep_get_and_clear(struct mm_struct * mm,unsigned long address,pte_t * ptep)660 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
661 				       unsigned long address, pte_t *ptep)
662 {
663 	return __pte(xchg_relaxed(&pte_val(*ptep), 0));
664 }
665 
666 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
667 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
pmdp_huge_get_and_clear(struct mm_struct * mm,unsigned long address,pmd_t * pmdp)668 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm,
669 					    unsigned long address, pmd_t *pmdp)
670 {
671 	return pte_pmd(ptep_get_and_clear(mm, address, (pte_t *)pmdp));
672 }
673 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
674 
675 /*
676  * ptep_set_wrprotect - mark read-only while trasferring potential hardware
677  * dirty status (PTE_DBM && !PTE_RDONLY) to the software PTE_DIRTY bit.
678  */
679 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
ptep_set_wrprotect(struct mm_struct * mm,unsigned long address,pte_t * ptep)680 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long address, pte_t *ptep)
681 {
682 	pte_t old_pte, pte;
683 
684 	pte = READ_ONCE(*ptep);
685 	do {
686 		old_pte = pte;
687 		/*
688 		 * If hardware-dirty (PTE_WRITE/DBM bit set and PTE_RDONLY
689 		 * clear), set the PTE_DIRTY bit.
690 		 */
691 		if (pte_hw_dirty(pte))
692 			pte = pte_mkdirty(pte);
693 		pte = pte_wrprotect(pte);
694 		pte_val(pte) = cmpxchg_relaxed(&pte_val(*ptep),
695 					       pte_val(old_pte), pte_val(pte));
696 	} while (pte_val(pte) != pte_val(old_pte));
697 }
698 
699 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
700 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
pmdp_set_wrprotect(struct mm_struct * mm,unsigned long address,pmd_t * pmdp)701 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
702 				      unsigned long address, pmd_t *pmdp)
703 {
704 	ptep_set_wrprotect(mm, address, (pte_t *)pmdp);
705 }
706 
707 #define pmdp_establish pmdp_establish
pmdp_establish(struct vm_area_struct * vma,unsigned long address,pmd_t * pmdp,pmd_t pmd)708 static inline pmd_t pmdp_establish(struct vm_area_struct *vma,
709 		unsigned long address, pmd_t *pmdp, pmd_t pmd)
710 {
711 	return __pmd(xchg_relaxed(&pmd_val(*pmdp), pmd_val(pmd)));
712 }
713 #endif
714 
715 extern pgd_t swapper_pg_dir[PTRS_PER_PGD];
716 extern pgd_t swapper_pg_end[];
717 extern pgd_t idmap_pg_dir[PTRS_PER_PGD];
718 extern pgd_t tramp_pg_dir[PTRS_PER_PGD];
719 
720 /*
721  * Encode and decode a swap entry:
722  *	bits 0-1:	present (must be zero)
723  *	bits 2-7:	swap type
724  *	bits 8-57:	swap offset
725  *	bit  58:	PTE_PROT_NONE (must be zero)
726  */
727 #define __SWP_TYPE_SHIFT	2
728 #define __SWP_TYPE_BITS		6
729 #define __SWP_OFFSET_BITS	50
730 #define __SWP_TYPE_MASK		((1 << __SWP_TYPE_BITS) - 1)
731 #define __SWP_OFFSET_SHIFT	(__SWP_TYPE_BITS + __SWP_TYPE_SHIFT)
732 #define __SWP_OFFSET_MASK	((1UL << __SWP_OFFSET_BITS) - 1)
733 
734 #define __swp_type(x)		(((x).val >> __SWP_TYPE_SHIFT) & __SWP_TYPE_MASK)
735 #define __swp_offset(x)		(((x).val >> __SWP_OFFSET_SHIFT) & __SWP_OFFSET_MASK)
736 #define __swp_entry(type,offset) ((swp_entry_t) { ((type) << __SWP_TYPE_SHIFT) | ((offset) << __SWP_OFFSET_SHIFT) })
737 
738 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
739 #define __swp_entry_to_pte(swp)	((pte_t) { (swp).val })
740 
741 /*
742  * Ensure that there are not more swap files than can be encoded in the kernel
743  * PTEs.
744  */
745 #define MAX_SWAPFILES_CHECK() BUILD_BUG_ON(MAX_SWAPFILES_SHIFT > __SWP_TYPE_BITS)
746 
747 extern int kern_addr_valid(unsigned long addr);
748 
749 #include <asm-generic/pgtable.h>
750 
751 void pgd_cache_init(void);
752 #define pgtable_cache_init	pgd_cache_init
753 
754 /*
755  * On AArch64, the cache coherency is handled via the set_pte_at() function.
756  */
update_mmu_cache(struct vm_area_struct * vma,unsigned long addr,pte_t * ptep)757 static inline void update_mmu_cache(struct vm_area_struct *vma,
758 				    unsigned long addr, pte_t *ptep)
759 {
760 	/*
761 	 * We don't do anything here, so there's a very small chance of
762 	 * us retaking a user fault which we just fixed up. The alternative
763 	 * is doing a dsb(ishst), but that penalises the fastpath.
764 	 */
765 }
766 
767 #define update_mmu_cache_pmd(vma, address, pmd) do { } while (0)
768 
769 #define kc_vaddr_to_offset(v)	((v) & ~VA_START)
770 #define kc_offset_to_vaddr(o)	((o) | VA_START)
771 
772 #ifdef CONFIG_ARM64_PA_BITS_52
773 #define phys_to_ttbr(addr)	(((addr) | ((addr) >> 46)) & TTBR_BADDR_MASK_52)
774 #else
775 #define phys_to_ttbr(addr)	(addr)
776 #endif
777 
778 #endif /* !__ASSEMBLY__ */
779 
780 #endif /* __ASM_PGTABLE_H */
781