Commit | Line | Data |
---|---|---|
b2441318 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
3dfcb315 AK |
2 | #ifndef _ASM_POWERPC_BOOK3S_64_PGTABLE_H_ |
3 | #define _ASM_POWERPC_BOOK3S_64_PGTABLE_H_ | |
2e873519 | 4 | |
9849a569 KS |
5 | #include <asm-generic/5level-fixup.h> |
6 | ||
c137a275 AK |
7 | #ifndef __ASSEMBLY__ |
8 | #include <linux/mmdebug.h> | |
ebd31197 | 9 | #include <linux/bug.h> |
c137a275 | 10 | #endif |
9849a569 | 11 | |
2e873519 AK |
12 | /* |
13 | * Common bits between hash and Radix page table | |
14 | */ | |
15 | #define _PAGE_BIT_SWAP_TYPE 0 | |
16 | ||
35175033 | 17 | #define _PAGE_NA 0 |
6b8cb66a | 18 | #define _PAGE_RO 0 |
812fadcb | 19 | #define _PAGE_USER 0 |
6b8cb66a | 20 | |
2e873519 AK |
21 | #define _PAGE_EXEC 0x00001 /* execute permission */ |
22 | #define _PAGE_WRITE 0x00002 /* write access allowed */ | |
23 | #define _PAGE_READ 0x00004 /* read access allowed */ | |
24 | #define _PAGE_RW (_PAGE_READ | _PAGE_WRITE) | |
25 | #define _PAGE_RWX (_PAGE_READ | _PAGE_WRITE | _PAGE_EXEC) | |
26 | #define _PAGE_PRIVILEGED 0x00008 /* kernel access only */ | |
27 | #define _PAGE_SAO 0x00010 /* Strong access order */ | |
28 | #define _PAGE_NON_IDEMPOTENT 0x00020 /* non idempotent memory */ | |
29 | #define _PAGE_TOLERANT 0x00030 /* tolerant memory, cache inhibited */ | |
30 | #define _PAGE_DIRTY 0x00080 /* C: page changed */ | |
31 | #define _PAGE_ACCESSED 0x00100 /* R: page referenced */ | |
3dfcb315 | 32 | /* |
2e873519 | 33 | * Software bits |
3dfcb315 | 34 | */ |
69dfbaeb AK |
35 | #define _RPAGE_SW0 0x2000000000000000UL |
36 | #define _RPAGE_SW1 0x00800 | |
37 | #define _RPAGE_SW2 0x00400 | |
38 | #define _RPAGE_SW3 0x00200 | |
049d567a AK |
39 | #define _RPAGE_RSV1 0x1000000000000000UL |
40 | #define _RPAGE_RSV2 0x0800000000000000UL | |
41 | #define _RPAGE_RSV3 0x0400000000000000UL | |
42 | #define _RPAGE_RSV4 0x0200000000000000UL | |
eb95d016 | 43 | #define _RPAGE_RSV5 0x00040UL |
6aa59f51 AK |
44 | |
45 | #define _PAGE_PTE 0x4000000000000000UL /* distinguishes PTEs from pointers */ | |
46 | #define _PAGE_PRESENT 0x8000000000000000UL /* pte contains a translation */ | |
bd0dbb73 AK |
47 | /* |
48 | * We need to mark a pmd pte invalid while splitting. We can do that by clearing | |
49 | * the _PAGE_PRESENT bit. But then that will be taken as a swap pte. In order to | |
50 | * differentiate between two use a SW field when invalidating. | |
51 | * | |
52 | * We do that temporary invalidate for regular pte entry in ptep_set_access_flags | |
53 | * | |
54 | * This is used only when _PAGE_PRESENT is cleared. | |
55 | */ | |
56 | #define _PAGE_INVALID _RPAGE_SW0 | |
6aa59f51 AK |
57 | |
58 | /* | |
59 | * Top and bottom bits of RPN which can be used by hash | |
60 | * translation mode, because we expect them to be zero | |
61 | * otherwise. | |
62 | */ | |
32789d38 AK |
63 | #define _RPAGE_RPN0 0x01000 |
64 | #define _RPAGE_RPN1 0x02000 | |
6aa59f51 AK |
65 | #define _RPAGE_RPN44 0x0100000000000000UL |
66 | #define _RPAGE_RPN43 0x0080000000000000UL | |
67 | #define _RPAGE_RPN42 0x0040000000000000UL | |
68 | #define _RPAGE_RPN41 0x0020000000000000UL | |
049d567a | 69 | |
2f18d533 AK |
70 | /* Max physical address bit as per radix table */ |
71 | #define _RPAGE_PA_MAX 57 | |
72 | ||
73 | /* | |
74 | * Max physical address bit we will use for now. | |
75 | * | |
76 | * This is mostly a hardware limitation and for now Power9 has | |
77 | * a 51 bit limit. | |
78 | * | |
79 | * This is different from the number of physical bit required to address | |
80 | * the last byte of memory. That is defined by MAX_PHYSMEM_BITS. | |
81 | * MAX_PHYSMEM_BITS is a linux limitation imposed by the maximum | |
82 | * number of sections we can support (SECTIONS_SHIFT). | |
83 | * | |
84 | * This is different from Radix page table limitation above and | |
85 | * should always be less than that. The limit is done such that | |
86 | * we can overload the bits between _RPAGE_PA_MAX and _PAGE_PA_MAX | |
87 | * for hash linux page table specific bits. | |
88 | * | |
89 | * In order to be compatible with future hardware generations we keep | |
90 | * some offsets and limit this for now to 53 | |
91 | */ | |
92 | #define _PAGE_PA_MAX 53 | |
93 | ||
69dfbaeb | 94 | #define _PAGE_SOFT_DIRTY _RPAGE_SW3 /* software: software dirty tracking */ |
69dfbaeb | 95 | #define _PAGE_SPECIAL _RPAGE_SW2 /* software: special page */ |
ebd31197 OH |
96 | #define _PAGE_DEVMAP _RPAGE_SW1 /* software: ZONE_DEVICE page */ |
97 | #define __HAVE_ARCH_PTE_DEVMAP | |
98 | ||
2e873519 AK |
99 | /* |
100 | * Drivers request for cache inhibited pte mapping using _PAGE_NO_CACHE | |
101 | * Instead of fixing all of them, add an alternate define which | |
102 | * maps CI pte mapping. | |
103 | */ | |
104 | #define _PAGE_NO_CACHE _PAGE_TOLERANT | |
105 | /* | |
2f18d533 AK |
106 | * We support _RPAGE_PA_MAX bit real address in pte. On the linux side |
107 | * we are limited by _PAGE_PA_MAX. Clear everything above _PAGE_PA_MAX | |
108 | * and every thing below PAGE_SHIFT; | |
2e873519 | 109 | */ |
2f18d533 | 110 | #define PTE_RPN_MASK (((1UL << _PAGE_PA_MAX) - 1) & (PAGE_MASK)) |
2e873519 AK |
111 | /* |
112 | * set of bits not changed in pmd_modify. Even though we have hash specific bits | |
113 | * in here, on radix we expect them to be zero. | |
114 | */ | |
115 | #define _HPAGE_CHG_MASK (PTE_RPN_MASK | _PAGE_HPTEFLAGS | _PAGE_DIRTY | \ | |
116 | _PAGE_ACCESSED | H_PAGE_THP_HUGE | _PAGE_PTE | \ | |
117 | _PAGE_SOFT_DIRTY) | |
118 | /* | |
119 | * user access blocked by key | |
120 | */ | |
121 | #define _PAGE_KERNEL_RW (_PAGE_PRIVILEGED | _PAGE_RW | _PAGE_DIRTY) | |
122 | #define _PAGE_KERNEL_RO (_PAGE_PRIVILEGED | _PAGE_READ) | |
123 | #define _PAGE_KERNEL_RWX (_PAGE_PRIVILEGED | _PAGE_DIRTY | \ | |
124 | _PAGE_RW | _PAGE_EXEC) | |
125 | /* | |
126 | * No page size encoding in the linux PTE | |
127 | */ | |
128 | #define _PAGE_PSIZE 0 | |
129 | /* | |
130 | * _PAGE_CHG_MASK masks of bits that are to be preserved across | |
131 | * pgprot changes | |
132 | */ | |
133 | #define _PAGE_CHG_MASK (PTE_RPN_MASK | _PAGE_HPTEFLAGS | _PAGE_DIRTY | \ | |
134 | _PAGE_ACCESSED | _PAGE_SPECIAL | _PAGE_PTE | \ | |
135 | _PAGE_SOFT_DIRTY) | |
eb95d016 RP |
136 | |
137 | #define H_PTE_PKEY (H_PTE_PKEY_BIT0 | H_PTE_PKEY_BIT1 | H_PTE_PKEY_BIT2 | \ | |
138 | H_PTE_PKEY_BIT3 | H_PTE_PKEY_BIT4) | |
2e873519 AK |
139 | /* |
140 | * Mask of bits returned by pte_pgprot() | |
141 | */ | |
142 | #define PAGE_PROT_BITS (_PAGE_SAO | _PAGE_NON_IDEMPOTENT | _PAGE_TOLERANT | \ | |
143 | H_PAGE_4K_PFN | _PAGE_PRIVILEGED | _PAGE_ACCESSED | \ | |
144 | _PAGE_READ | _PAGE_WRITE | _PAGE_DIRTY | _PAGE_EXEC | \ | |
eb95d016 | 145 | _PAGE_SOFT_DIRTY | H_PTE_PKEY) |
3dfcb315 | 146 | /* |
2e873519 AK |
147 | * We define 2 sets of base prot bits, one for basic pages (ie, |
148 | * cacheable kernel and user pages) and one for non cacheable | |
149 | * pages. We always set _PAGE_COHERENT when SMP is enabled or | |
150 | * the processor might need it for DMA coherency. | |
3dfcb315 | 151 | */ |
2e873519 AK |
152 | #define _PAGE_BASE_NC (_PAGE_PRESENT | _PAGE_ACCESSED | _PAGE_PSIZE) |
153 | #define _PAGE_BASE (_PAGE_BASE_NC) | |
154 | ||
155 | /* Permission masks used to generate the __P and __S table, | |
156 | * | |
157 | * Note:__pgprot is defined in arch/powerpc/include/asm/page.h | |
158 | * | |
159 | * Write permissions imply read permissions for now (we could make write-only | |
160 | * pages on BookE but we don't bother for now). Execute permission control is | |
161 | * possible on platforms that define _PAGE_EXEC | |
162 | * | |
163 | * Note due to the way vm flags are laid out, the bits are XWR | |
164 | */ | |
165 | #define PAGE_NONE __pgprot(_PAGE_BASE | _PAGE_PRIVILEGED) | |
166 | #define PAGE_SHARED __pgprot(_PAGE_BASE | _PAGE_RW) | |
167 | #define PAGE_SHARED_X __pgprot(_PAGE_BASE | _PAGE_RW | _PAGE_EXEC) | |
168 | #define PAGE_COPY __pgprot(_PAGE_BASE | _PAGE_READ) | |
169 | #define PAGE_COPY_X __pgprot(_PAGE_BASE | _PAGE_READ | _PAGE_EXEC) | |
170 | #define PAGE_READONLY __pgprot(_PAGE_BASE | _PAGE_READ) | |
171 | #define PAGE_READONLY_X __pgprot(_PAGE_BASE | _PAGE_READ | _PAGE_EXEC) | |
172 | ||
173 | #define __P000 PAGE_NONE | |
174 | #define __P001 PAGE_READONLY | |
175 | #define __P010 PAGE_COPY | |
176 | #define __P011 PAGE_COPY | |
177 | #define __P100 PAGE_READONLY_X | |
178 | #define __P101 PAGE_READONLY_X | |
179 | #define __P110 PAGE_COPY_X | |
180 | #define __P111 PAGE_COPY_X | |
181 | ||
182 | #define __S000 PAGE_NONE | |
183 | #define __S001 PAGE_READONLY | |
184 | #define __S010 PAGE_SHARED | |
185 | #define __S011 PAGE_SHARED | |
186 | #define __S100 PAGE_READONLY_X | |
187 | #define __S101 PAGE_READONLY_X | |
188 | #define __S110 PAGE_SHARED_X | |
189 | #define __S111 PAGE_SHARED_X | |
190 | ||
191 | /* Permission masks used for kernel mappings */ | |
192 | #define PAGE_KERNEL __pgprot(_PAGE_BASE | _PAGE_KERNEL_RW) | |
193 | #define PAGE_KERNEL_NC __pgprot(_PAGE_BASE_NC | _PAGE_KERNEL_RW | \ | |
194 | _PAGE_TOLERANT) | |
195 | #define PAGE_KERNEL_NCG __pgprot(_PAGE_BASE_NC | _PAGE_KERNEL_RW | \ | |
196 | _PAGE_NON_IDEMPOTENT) | |
197 | #define PAGE_KERNEL_X __pgprot(_PAGE_BASE | _PAGE_KERNEL_RWX) | |
198 | #define PAGE_KERNEL_RO __pgprot(_PAGE_BASE | _PAGE_KERNEL_RO) | |
199 | #define PAGE_KERNEL_ROX __pgprot(_PAGE_BASE | _PAGE_KERNEL_ROX) | |
200 | ||
201 | /* | |
202 | * Protection used for kernel text. We want the debuggers to be able to | |
203 | * set breakpoints anywhere, so don't write protect the kernel text | |
204 | * on platforms where such control is possible. | |
205 | */ | |
206 | #if defined(CONFIG_KGDB) || defined(CONFIG_XMON) || defined(CONFIG_BDI_SWITCH) || \ | |
207 | defined(CONFIG_KPROBES) || defined(CONFIG_DYNAMIC_FTRACE) | |
208 | #define PAGE_KERNEL_TEXT PAGE_KERNEL_X | |
209 | #else | |
210 | #define PAGE_KERNEL_TEXT PAGE_KERNEL_ROX | |
211 | #endif | |
212 | ||
213 | /* Make modules code happy. We don't set RO yet */ | |
214 | #define PAGE_KERNEL_EXEC PAGE_KERNEL_X | |
215 | #define PAGE_AGP (PAGE_KERNEL_NC) | |
3dfcb315 | 216 | |
dd1842a2 AK |
217 | #ifndef __ASSEMBLY__ |
218 | /* | |
219 | * page table defines | |
220 | */ | |
221 | extern unsigned long __pte_index_size; | |
222 | extern unsigned long __pmd_index_size; | |
223 | extern unsigned long __pud_index_size; | |
224 | extern unsigned long __pgd_index_size; | |
fae22116 | 225 | extern unsigned long __pud_cache_index; |
dd1842a2 AK |
226 | #define PTE_INDEX_SIZE __pte_index_size |
227 | #define PMD_INDEX_SIZE __pmd_index_size | |
228 | #define PUD_INDEX_SIZE __pud_index_size | |
229 | #define PGD_INDEX_SIZE __pgd_index_size | |
738f9645 AK |
230 | /* pmd table use page table fragments */ |
231 | #define PMD_CACHE_INDEX 0 | |
fae22116 | 232 | #define PUD_CACHE_INDEX __pud_cache_index |
dd1842a2 AK |
233 | /* |
234 | * Because of use of pte fragments and THP, size of page table | |
235 | * are not always derived out of index size above. | |
236 | */ | |
237 | extern unsigned long __pte_table_size; | |
238 | extern unsigned long __pmd_table_size; | |
239 | extern unsigned long __pud_table_size; | |
240 | extern unsigned long __pgd_table_size; | |
241 | #define PTE_TABLE_SIZE __pte_table_size | |
242 | #define PMD_TABLE_SIZE __pmd_table_size | |
243 | #define PUD_TABLE_SIZE __pud_table_size | |
244 | #define PGD_TABLE_SIZE __pgd_table_size | |
a2f41eb9 AK |
245 | |
246 | extern unsigned long __pmd_val_bits; | |
247 | extern unsigned long __pud_val_bits; | |
248 | extern unsigned long __pgd_val_bits; | |
249 | #define PMD_VAL_BITS __pmd_val_bits | |
250 | #define PUD_VAL_BITS __pud_val_bits | |
251 | #define PGD_VAL_BITS __pgd_val_bits | |
5ed7ecd0 AK |
252 | |
253 | extern unsigned long __pte_frag_nr; | |
254 | #define PTE_FRAG_NR __pte_frag_nr | |
255 | extern unsigned long __pte_frag_size_shift; | |
256 | #define PTE_FRAG_SIZE_SHIFT __pte_frag_size_shift | |
257 | #define PTE_FRAG_SIZE (1UL << PTE_FRAG_SIZE_SHIFT) | |
dd1842a2 | 258 | |
8a6c697b AK |
259 | extern unsigned long __pmd_frag_nr; |
260 | #define PMD_FRAG_NR __pmd_frag_nr | |
261 | extern unsigned long __pmd_frag_size_shift; | |
262 | #define PMD_FRAG_SIZE_SHIFT __pmd_frag_size_shift | |
263 | #define PMD_FRAG_SIZE (1UL << PMD_FRAG_SIZE_SHIFT) | |
264 | ||
dd1842a2 AK |
265 | #define PTRS_PER_PTE (1 << PTE_INDEX_SIZE) |
266 | #define PTRS_PER_PMD (1 << PMD_INDEX_SIZE) | |
267 | #define PTRS_PER_PUD (1 << PUD_INDEX_SIZE) | |
268 | #define PTRS_PER_PGD (1 << PGD_INDEX_SIZE) | |
269 | ||
270 | /* PMD_SHIFT determines what a second-level page table entry can map */ | |
271 | #define PMD_SHIFT (PAGE_SHIFT + PTE_INDEX_SIZE) | |
272 | #define PMD_SIZE (1UL << PMD_SHIFT) | |
273 | #define PMD_MASK (~(PMD_SIZE-1)) | |
274 | ||
275 | /* PUD_SHIFT determines what a third-level page table entry can map */ | |
276 | #define PUD_SHIFT (PMD_SHIFT + PMD_INDEX_SIZE) | |
277 | #define PUD_SIZE (1UL << PUD_SHIFT) | |
278 | #define PUD_MASK (~(PUD_SIZE-1)) | |
279 | ||
280 | /* PGDIR_SHIFT determines what a fourth-level page table entry can map */ | |
281 | #define PGDIR_SHIFT (PUD_SHIFT + PUD_INDEX_SIZE) | |
282 | #define PGDIR_SIZE (1UL << PGDIR_SHIFT) | |
283 | #define PGDIR_MASK (~(PGDIR_SIZE-1)) | |
284 | ||
285 | /* Bits to mask out from a PMD to get to the PTE page */ | |
286 | #define PMD_MASKED_BITS 0xc0000000000000ffUL | |
287 | /* Bits to mask out from a PUD to get to the PMD page */ | |
288 | #define PUD_MASKED_BITS 0xc0000000000000ffUL | |
289 | /* Bits to mask out from a PGD to get to the PUD page */ | |
290 | #define PGD_MASKED_BITS 0xc0000000000000ffUL | |
d6a9996e | 291 | |
0c4d2680 AK |
292 | /* |
293 | * Used as an indicator for rcu callback functions | |
294 | */ | |
295 | enum pgtable_index { | |
296 | PTE_INDEX = 0, | |
297 | PMD_INDEX, | |
298 | PUD_INDEX, | |
299 | PGD_INDEX, | |
fadd03c6 AK |
300 | /* |
301 | * Below are used with 4k page size and hugetlb | |
302 | */ | |
303 | HTLB_16M_INDEX, | |
304 | HTLB_16G_INDEX, | |
0c4d2680 AK |
305 | }; |
306 | ||
d6a9996e AK |
307 | extern unsigned long __vmalloc_start; |
308 | extern unsigned long __vmalloc_end; | |
309 | #define VMALLOC_START __vmalloc_start | |
310 | #define VMALLOC_END __vmalloc_end | |
311 | ||
312 | extern unsigned long __kernel_virt_start; | |
313 | extern unsigned long __kernel_virt_size; | |
63ee9b2f | 314 | extern unsigned long __kernel_io_start; |
d6a9996e AK |
315 | #define KERN_VIRT_START __kernel_virt_start |
316 | #define KERN_VIRT_SIZE __kernel_virt_size | |
63ee9b2f | 317 | #define KERN_IO_START __kernel_io_start |
d6a9996e AK |
318 | extern struct page *vmemmap; |
319 | extern unsigned long ioremap_bot; | |
bfa37087 | 320 | extern unsigned long pci_io_base; |
dd1842a2 | 321 | #endif /* __ASSEMBLY__ */ |
3dfcb315 | 322 | |
ab537dca | 323 | #include <asm/book3s/64/hash.h> |
b0b5e9b1 | 324 | #include <asm/book3s/64/radix.h> |
3dfcb315 | 325 | |
a9252aae AK |
326 | #ifdef CONFIG_PPC_64K_PAGES |
327 | #include <asm/book3s/64/pgtable-64k.h> | |
328 | #else | |
329 | #include <asm/book3s/64/pgtable-4k.h> | |
330 | #endif | |
331 | ||
3dfcb315 | 332 | #include <asm/barrier.h> |
3dfcb315 AK |
333 | /* |
334 | * The second half of the kernel virtual space is used for IO mappings, | |
335 | * it's itself carved into the PIO region (ISA and PHB IO space) and | |
336 | * the ioremap space | |
337 | * | |
338 | * ISA_IO_BASE = KERN_IO_START, 64K reserved area | |
339 | * PHB_IO_BASE = ISA_IO_BASE + 64K to ISA_IO_BASE + 2G, PHB IO spaces | |
340 | * IOREMAP_BASE = ISA_IO_BASE + 2G to VMALLOC_START + PGTABLE_RANGE | |
341 | */ | |
3dfcb315 AK |
342 | #define FULL_IO_SIZE 0x80000000ul |
343 | #define ISA_IO_BASE (KERN_IO_START) | |
344 | #define ISA_IO_END (KERN_IO_START + 0x10000ul) | |
345 | #define PHB_IO_BASE (ISA_IO_END) | |
346 | #define PHB_IO_END (KERN_IO_START + FULL_IO_SIZE) | |
347 | #define IOREMAP_BASE (PHB_IO_END) | |
348 | #define IOREMAP_END (KERN_VIRT_START + KERN_VIRT_SIZE) | |
349 | ||
b0412ea9 | 350 | /* Advertise special mapping type for AGP */ |
b0412ea9 AK |
351 | #define HAVE_PAGE_AGP |
352 | ||
3dfcb315 AK |
353 | #ifndef __ASSEMBLY__ |
354 | ||
355 | /* | |
356 | * This is the default implementation of various PTE accessors, it's | |
357 | * used in all cases except Book3S with 64K pages where we have a | |
358 | * concept of sub-pages | |
359 | */ | |
360 | #ifndef __real_pte | |
361 | ||
ff31e105 | 362 | #define __real_pte(e, p, o) ((real_pte_t){(e)}) |
3dfcb315 | 363 | #define __rpte_to_pte(r) ((r).pte) |
945537df | 364 | #define __rpte_to_hidx(r,index) (pte_val(__rpte_to_pte(r)) >> H_PAGE_F_GIX_SHIFT) |
3dfcb315 AK |
365 | |
366 | #define pte_iterate_hashed_subpages(rpte, psize, va, index, shift) \ | |
367 | do { \ | |
368 | index = 0; \ | |
369 | shift = mmu_psize_defs[psize].shift; \ | |
370 | ||
371 | #define pte_iterate_hashed_end() } while(0) | |
372 | ||
373 | /* | |
374 | * We expect this to be called only for user addresses or kernel virtual | |
375 | * addresses other than the linear mapping. | |
376 | */ | |
377 | #define pte_pagesize_index(mm, addr, pte) MMU_PAGE_4K | |
378 | ||
379 | #endif /* __real_pte */ | |
380 | ||
ac94ac79 AK |
381 | static inline unsigned long pte_update(struct mm_struct *mm, unsigned long addr, |
382 | pte_t *ptep, unsigned long clr, | |
383 | unsigned long set, int huge) | |
384 | { | |
385 | if (radix_enabled()) | |
386 | return radix__pte_update(mm, addr, ptep, clr, set, huge); | |
387 | return hash__pte_update(mm, addr, ptep, clr, set, huge); | |
388 | } | |
13f829a5 AK |
389 | /* |
390 | * For hash even if we have _PAGE_ACCESSED = 0, we do a pte_update. | |
391 | * We currently remove entries from the hashtable regardless of whether | |
392 | * the entry was young or dirty. | |
393 | * | |
394 | * We should be more intelligent about this but for the moment we override | |
395 | * these functions and force a tlb flush unconditionally | |
396 | * For radix: H_PAGE_HASHPTE should be zero. Hence we can use the same | |
397 | * function for both hash and radix. | |
398 | */ | |
399 | static inline int __ptep_test_and_clear_young(struct mm_struct *mm, | |
400 | unsigned long addr, pte_t *ptep) | |
401 | { | |
402 | unsigned long old; | |
403 | ||
66c570f5 | 404 | if ((pte_raw(*ptep) & cpu_to_be64(_PAGE_ACCESSED | H_PAGE_HASHPTE)) == 0) |
13f829a5 AK |
405 | return 0; |
406 | old = pte_update(mm, addr, ptep, _PAGE_ACCESSED, 0, 0); | |
407 | return (old & _PAGE_ACCESSED) != 0; | |
408 | } | |
409 | ||
410 | #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG | |
411 | #define ptep_test_and_clear_young(__vma, __addr, __ptep) \ | |
412 | ({ \ | |
413 | int __r; \ | |
414 | __r = __ptep_test_and_clear_young((__vma)->vm_mm, __addr, __ptep); \ | |
415 | __r; \ | |
416 | }) | |
417 | ||
d19469e8 | 418 | static inline int __pte_write(pte_t pte) |
52c50ca7 AK |
419 | { |
420 | return !!(pte_raw(pte) & cpu_to_be64(_PAGE_WRITE)); | |
421 | } | |
422 | ||
423 | #ifdef CONFIG_NUMA_BALANCING | |
424 | #define pte_savedwrite pte_savedwrite | |
425 | static inline bool pte_savedwrite(pte_t pte) | |
426 | { | |
427 | /* | |
428 | * Saved write ptes are prot none ptes that doesn't have | |
429 | * privileged bit sit. We mark prot none as one which has | |
430 | * present and pviliged bit set and RWX cleared. To mark | |
431 | * protnone which used to have _PAGE_WRITE set we clear | |
432 | * the privileged bit. | |
433 | */ | |
434 | return !(pte_raw(pte) & cpu_to_be64(_PAGE_RWX | _PAGE_PRIVILEGED)); | |
435 | } | |
436 | #else | |
437 | #define pte_savedwrite pte_savedwrite | |
438 | static inline bool pte_savedwrite(pte_t pte) | |
439 | { | |
440 | return false; | |
441 | } | |
442 | #endif | |
443 | ||
d19469e8 AK |
444 | static inline int pte_write(pte_t pte) |
445 | { | |
446 | return __pte_write(pte) || pte_savedwrite(pte); | |
447 | } | |
448 | ||
ca8afd40 CL |
449 | static inline int pte_read(pte_t pte) |
450 | { | |
451 | return !!(pte_raw(pte) & cpu_to_be64(_PAGE_READ)); | |
452 | } | |
453 | ||
13f829a5 AK |
454 | #define __HAVE_ARCH_PTEP_SET_WRPROTECT |
455 | static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, | |
456 | pte_t *ptep) | |
457 | { | |
d19469e8 | 458 | if (__pte_write(*ptep)) |
52c50ca7 AK |
459 | pte_update(mm, addr, ptep, _PAGE_WRITE, 0, 0); |
460 | else if (unlikely(pte_savedwrite(*ptep))) | |
461 | pte_update(mm, addr, ptep, 0, _PAGE_PRIVILEGED, 0); | |
13f829a5 AK |
462 | } |
463 | ||
464 | static inline void huge_ptep_set_wrprotect(struct mm_struct *mm, | |
465 | unsigned long addr, pte_t *ptep) | |
466 | { | |
52c50ca7 AK |
467 | /* |
468 | * We should not find protnone for hugetlb, but this complete the | |
469 | * interface. | |
470 | */ | |
d19469e8 | 471 | if (__pte_write(*ptep)) |
52c50ca7 AK |
472 | pte_update(mm, addr, ptep, _PAGE_WRITE, 0, 1); |
473 | else if (unlikely(pte_savedwrite(*ptep))) | |
474 | pte_update(mm, addr, ptep, 0, _PAGE_PRIVILEGED, 1); | |
13f829a5 AK |
475 | } |
476 | ||
477 | #define __HAVE_ARCH_PTEP_GET_AND_CLEAR | |
478 | static inline pte_t ptep_get_and_clear(struct mm_struct *mm, | |
479 | unsigned long addr, pte_t *ptep) | |
480 | { | |
481 | unsigned long old = pte_update(mm, addr, ptep, ~0UL, 0, 0); | |
482 | return __pte(old); | |
483 | } | |
484 | ||
f4894b80 AK |
485 | #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL |
486 | static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm, | |
487 | unsigned long addr, | |
488 | pte_t *ptep, int full) | |
489 | { | |
490 | if (full && radix_enabled()) { | |
491 | /* | |
2bf1071a NP |
492 | * We know that this is a full mm pte clear and |
493 | * hence can be sure there is no parallel set_pte. | |
f4894b80 AK |
494 | */ |
495 | return radix__ptep_get_and_clear_full(mm, addr, ptep, full); | |
496 | } | |
497 | return ptep_get_and_clear(mm, addr, ptep); | |
498 | } | |
499 | ||
500 | ||
13f829a5 AK |
501 | static inline void pte_clear(struct mm_struct *mm, unsigned long addr, |
502 | pte_t * ptep) | |
503 | { | |
504 | pte_update(mm, addr, ptep, ~0UL, 0, 0); | |
505 | } | |
66c570f5 | 506 | |
66c570f5 AK |
507 | static inline int pte_dirty(pte_t pte) |
508 | { | |
509 | return !!(pte_raw(pte) & cpu_to_be64(_PAGE_DIRTY)); | |
510 | } | |
511 | ||
512 | static inline int pte_young(pte_t pte) | |
513 | { | |
514 | return !!(pte_raw(pte) & cpu_to_be64(_PAGE_ACCESSED)); | |
515 | } | |
516 | ||
517 | static inline int pte_special(pte_t pte) | |
518 | { | |
519 | return !!(pte_raw(pte) & cpu_to_be64(_PAGE_SPECIAL)); | |
520 | } | |
521 | ||
13f829a5 AK |
522 | static inline pgprot_t pte_pgprot(pte_t pte) { return __pgprot(pte_val(pte) & PAGE_PROT_BITS); } |
523 | ||
524 | #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY | |
525 | static inline bool pte_soft_dirty(pte_t pte) | |
526 | { | |
66c570f5 | 527 | return !!(pte_raw(pte) & cpu_to_be64(_PAGE_SOFT_DIRTY)); |
13f829a5 | 528 | } |
66c570f5 | 529 | |
13f829a5 AK |
530 | static inline pte_t pte_mksoft_dirty(pte_t pte) |
531 | { | |
532 | return __pte(pte_val(pte) | _PAGE_SOFT_DIRTY); | |
533 | } | |
534 | ||
535 | static inline pte_t pte_clear_soft_dirty(pte_t pte) | |
536 | { | |
537 | return __pte(pte_val(pte) & ~_PAGE_SOFT_DIRTY); | |
538 | } | |
539 | #endif /* CONFIG_HAVE_ARCH_SOFT_DIRTY */ | |
540 | ||
541 | #ifdef CONFIG_NUMA_BALANCING | |
13f829a5 AK |
542 | static inline int pte_protnone(pte_t pte) |
543 | { | |
c137a275 AK |
544 | return (pte_raw(pte) & cpu_to_be64(_PAGE_PRESENT | _PAGE_PTE | _PAGE_RWX)) == |
545 | cpu_to_be64(_PAGE_PRESENT | _PAGE_PTE); | |
546 | } | |
547 | ||
548 | #define pte_mk_savedwrite pte_mk_savedwrite | |
549 | static inline pte_t pte_mk_savedwrite(pte_t pte) | |
550 | { | |
551 | /* | |
552 | * Used by Autonuma subsystem to preserve the write bit | |
553 | * while marking the pte PROT_NONE. Only allow this | |
554 | * on PROT_NONE pte | |
555 | */ | |
556 | VM_BUG_ON((pte_raw(pte) & cpu_to_be64(_PAGE_PRESENT | _PAGE_RWX | _PAGE_PRIVILEGED)) != | |
557 | cpu_to_be64(_PAGE_PRESENT | _PAGE_PRIVILEGED)); | |
558 | return __pte(pte_val(pte) & ~_PAGE_PRIVILEGED); | |
559 | } | |
560 | ||
561 | #define pte_clear_savedwrite pte_clear_savedwrite | |
562 | static inline pte_t pte_clear_savedwrite(pte_t pte) | |
563 | { | |
564 | /* | |
565 | * Used by KSM subsystem to make a protnone pte readonly. | |
566 | */ | |
567 | VM_BUG_ON(!pte_protnone(pte)); | |
568 | return __pte(pte_val(pte) | _PAGE_PRIVILEGED); | |
569 | } | |
d19469e8 AK |
570 | #else |
571 | #define pte_clear_savedwrite pte_clear_savedwrite | |
572 | static inline pte_t pte_clear_savedwrite(pte_t pte) | |
573 | { | |
574 | VM_WARN_ON(1); | |
575 | return __pte(pte_val(pte) & ~_PAGE_WRITE); | |
576 | } | |
13f829a5 AK |
577 | #endif /* CONFIG_NUMA_BALANCING */ |
578 | ||
579 | static inline int pte_present(pte_t pte) | |
580 | { | |
bd0dbb73 AK |
581 | /* |
582 | * A pte is considerent present if _PAGE_PRESENT is set. | |
583 | * We also need to consider the pte present which is marked | |
584 | * invalid during ptep_set_access_flags. Hence we look for _PAGE_INVALID | |
585 | * if we find _PAGE_PRESENT cleared. | |
586 | */ | |
587 | return !!(pte_raw(pte) & cpu_to_be64(_PAGE_PRESENT | _PAGE_INVALID)); | |
13f829a5 | 588 | } |
f72a85e3 | 589 | |
bca7aacf | 590 | #ifdef CONFIG_PPC_MEM_KEYS |
f2407ef3 | 591 | extern bool arch_pte_access_permitted(u64 pte, bool write, bool execute); |
bca7aacf RP |
592 | #else |
593 | static inline bool arch_pte_access_permitted(u64 pte, bool write, bool execute) | |
594 | { | |
595 | return true; | |
596 | } | |
597 | #endif /* CONFIG_PPC_MEM_KEYS */ | |
f2407ef3 | 598 | |
f72a85e3 AK |
599 | #define pte_access_permitted pte_access_permitted |
600 | static inline bool pte_access_permitted(pte_t pte, bool write) | |
601 | { | |
602 | unsigned long pteval = pte_val(pte); | |
603 | /* Also check for pte_user */ | |
604 | unsigned long clear_pte_bits = _PAGE_PRIVILEGED; | |
605 | /* | |
606 | * _PAGE_READ is needed for any access and will be | |
607 | * cleared for PROT_NONE | |
608 | */ | |
609 | unsigned long need_pte_bits = _PAGE_PRESENT | _PAGE_READ; | |
610 | ||
611 | if (write) | |
612 | need_pte_bits |= _PAGE_WRITE; | |
613 | ||
614 | if ((pteval & need_pte_bits) != need_pte_bits) | |
615 | return false; | |
616 | ||
617 | if ((pteval & clear_pte_bits) == clear_pte_bits) | |
618 | return false; | |
bca7aacf RP |
619 | |
620 | return arch_pte_access_permitted(pte_val(pte), write, 0); | |
f72a85e3 AK |
621 | } |
622 | ||
13f829a5 AK |
623 | /* |
624 | * Conversion functions: convert a page and protection to a page entry, | |
625 | * and a page entry and page directory to the page they refer to. | |
626 | * | |
627 | * Even if PTEs can be unsigned long long, a PFN is always an unsigned | |
628 | * long for now. | |
629 | */ | |
630 | static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot) | |
631 | { | |
632 | return __pte((((pte_basic_t)(pfn) << PAGE_SHIFT) & PTE_RPN_MASK) | | |
633 | pgprot_val(pgprot)); | |
634 | } | |
635 | ||
636 | static inline unsigned long pte_pfn(pte_t pte) | |
637 | { | |
638 | return (pte_val(pte) & PTE_RPN_MASK) >> PAGE_SHIFT; | |
639 | } | |
640 | ||
641 | /* Generic modifiers for PTE bits */ | |
642 | static inline pte_t pte_wrprotect(pte_t pte) | |
643 | { | |
d19469e8 AK |
644 | if (unlikely(pte_savedwrite(pte))) |
645 | return pte_clear_savedwrite(pte); | |
13f829a5 AK |
646 | return __pte(pte_val(pte) & ~_PAGE_WRITE); |
647 | } | |
648 | ||
649 | static inline pte_t pte_mkclean(pte_t pte) | |
650 | { | |
651 | return __pte(pte_val(pte) & ~_PAGE_DIRTY); | |
652 | } | |
653 | ||
654 | static inline pte_t pte_mkold(pte_t pte) | |
655 | { | |
656 | return __pte(pte_val(pte) & ~_PAGE_ACCESSED); | |
657 | } | |
658 | ||
659 | static inline pte_t pte_mkwrite(pte_t pte) | |
660 | { | |
661 | /* | |
662 | * write implies read, hence set both | |
663 | */ | |
664 | return __pte(pte_val(pte) | _PAGE_RW); | |
665 | } | |
666 | ||
667 | static inline pte_t pte_mkdirty(pte_t pte) | |
668 | { | |
669 | return __pte(pte_val(pte) | _PAGE_DIRTY | _PAGE_SOFT_DIRTY); | |
670 | } | |
671 | ||
672 | static inline pte_t pte_mkyoung(pte_t pte) | |
673 | { | |
674 | return __pte(pte_val(pte) | _PAGE_ACCESSED); | |
675 | } | |
676 | ||
677 | static inline pte_t pte_mkspecial(pte_t pte) | |
678 | { | |
679 | return __pte(pte_val(pte) | _PAGE_SPECIAL); | |
680 | } | |
681 | ||
682 | static inline pte_t pte_mkhuge(pte_t pte) | |
683 | { | |
684 | return pte; | |
685 | } | |
686 | ||
ebd31197 OH |
687 | static inline pte_t pte_mkdevmap(pte_t pte) |
688 | { | |
689 | return __pte(pte_val(pte) | _PAGE_SPECIAL|_PAGE_DEVMAP); | |
690 | } | |
691 | ||
c9c98bc5 OH |
692 | /* |
693 | * This is potentially called with a pmd as the argument, in which case it's not | |
694 | * safe to check _PAGE_DEVMAP unless we also confirm that _PAGE_PTE is set. | |
695 | * That's because the bit we use for _PAGE_DEVMAP is not reserved for software | |
696 | * use in page directory entries (ie. non-ptes). | |
697 | */ | |
ebd31197 OH |
698 | static inline int pte_devmap(pte_t pte) |
699 | { | |
c9c98bc5 OH |
700 | u64 mask = cpu_to_be64(_PAGE_DEVMAP | _PAGE_PTE); |
701 | ||
702 | return (pte_raw(pte) & mask) == mask; | |
ebd31197 OH |
703 | } |
704 | ||
13f829a5 AK |
705 | static inline pte_t pte_modify(pte_t pte, pgprot_t newprot) |
706 | { | |
707 | /* FIXME!! check whether this need to be a conditional */ | |
708 | return __pte((pte_val(pte) & _PAGE_CHG_MASK) | pgprot_val(newprot)); | |
709 | } | |
710 | ||
34fbadd8 AK |
711 | static inline bool pte_user(pte_t pte) |
712 | { | |
66c570f5 | 713 | return !(pte_raw(pte) & cpu_to_be64(_PAGE_PRIVILEGED)); |
34fbadd8 AK |
714 | } |
715 | ||
716 | /* Encode and de-code a swap entry */ | |
717 | #define MAX_SWAPFILES_CHECK() do { \ | |
718 | BUILD_BUG_ON(MAX_SWAPFILES_SHIFT > SWP_TYPE_BITS); \ | |
719 | /* \ | |
720 | * Don't have overlapping bits with _PAGE_HPTEFLAGS \ | |
721 | * We filter HPTEFLAGS on set_pte. \ | |
722 | */ \ | |
723 | BUILD_BUG_ON(_PAGE_HPTEFLAGS & (0x1f << _PAGE_BIT_SWAP_TYPE)); \ | |
724 | BUILD_BUG_ON(_PAGE_HPTEFLAGS & _PAGE_SWP_SOFT_DIRTY); \ | |
725 | } while (0) | |
726 | /* | |
727 | * on pte we don't need handle RADIX_TREE_EXCEPTIONAL_SHIFT; | |
728 | */ | |
729 | #define SWP_TYPE_BITS 5 | |
730 | #define __swp_type(x) (((x).val >> _PAGE_BIT_SWAP_TYPE) \ | |
731 | & ((1UL << SWP_TYPE_BITS) - 1)) | |
732 | #define __swp_offset(x) (((x).val & PTE_RPN_MASK) >> PAGE_SHIFT) | |
733 | #define __swp_entry(type, offset) ((swp_entry_t) { \ | |
734 | ((type) << _PAGE_BIT_SWAP_TYPE) \ | |
735 | | (((offset) << PAGE_SHIFT) & PTE_RPN_MASK)}) | |
736 | /* | |
737 | * swp_entry_t must be independent of pte bits. We build a swp_entry_t from | |
738 | * swap type and offset we get from swap and convert that to pte to find a | |
739 | * matching pte in linux page table. | |
740 | * Clear bits not found in swap entries here. | |
741 | */ | |
742 | #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val((pte)) & ~_PAGE_PTE }) | |
743 | #define __swp_entry_to_pte(x) __pte((x).val | _PAGE_PTE) | |
744 | ||
745 | #ifdef CONFIG_MEM_SOFT_DIRTY | |
746 | #define _PAGE_SWP_SOFT_DIRTY (1UL << (SWP_TYPE_BITS + _PAGE_BIT_SWAP_TYPE)) | |
747 | #else | |
748 | #define _PAGE_SWP_SOFT_DIRTY 0UL | |
749 | #endif /* CONFIG_MEM_SOFT_DIRTY */ | |
750 | ||
751 | #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY | |
752 | static inline pte_t pte_swp_mksoft_dirty(pte_t pte) | |
753 | { | |
754 | return __pte(pte_val(pte) | _PAGE_SWP_SOFT_DIRTY); | |
755 | } | |
66c570f5 | 756 | |
34fbadd8 AK |
757 | static inline bool pte_swp_soft_dirty(pte_t pte) |
758 | { | |
66c570f5 | 759 | return !!(pte_raw(pte) & cpu_to_be64(_PAGE_SWP_SOFT_DIRTY)); |
34fbadd8 | 760 | } |
66c570f5 | 761 | |
34fbadd8 AK |
762 | static inline pte_t pte_swp_clear_soft_dirty(pte_t pte) |
763 | { | |
764 | return __pte(pte_val(pte) & ~_PAGE_SWP_SOFT_DIRTY); | |
765 | } | |
766 | #endif /* CONFIG_HAVE_ARCH_SOFT_DIRTY */ | |
767 | ||
768 | static inline bool check_pte_access(unsigned long access, unsigned long ptev) | |
769 | { | |
770 | /* | |
771 | * This check for _PAGE_RWX and _PAGE_PRESENT bits | |
772 | */ | |
773 | if (access & ~ptev) | |
774 | return false; | |
775 | /* | |
776 | * This check for access to privilege space | |
777 | */ | |
778 | if ((access & _PAGE_PRIVILEGED) != (ptev & _PAGE_PRIVILEGED)) | |
779 | return false; | |
780 | ||
781 | return true; | |
782 | } | |
ac94ac79 AK |
783 | /* |
784 | * Generic functions with hash/radix callbacks | |
785 | */ | |
786 | ||
e4c1112c | 787 | static inline void __ptep_set_access_flags(struct vm_area_struct *vma, |
b3603e17 | 788 | pte_t *ptep, pte_t entry, |
e4c1112c AK |
789 | unsigned long address, |
790 | int psize) | |
ac94ac79 AK |
791 | { |
792 | if (radix_enabled()) | |
e4c1112c AK |
793 | return radix__ptep_set_access_flags(vma, ptep, entry, |
794 | address, psize); | |
ac94ac79 AK |
795 | return hash__ptep_set_access_flags(ptep, entry); |
796 | } | |
797 | ||
798 | #define __HAVE_ARCH_PTE_SAME | |
799 | static inline int pte_same(pte_t pte_a, pte_t pte_b) | |
800 | { | |
801 | if (radix_enabled()) | |
802 | return radix__pte_same(pte_a, pte_b); | |
803 | return hash__pte_same(pte_a, pte_b); | |
804 | } | |
805 | ||
806 | static inline int pte_none(pte_t pte) | |
807 | { | |
808 | if (radix_enabled()) | |
809 | return radix__pte_none(pte); | |
810 | return hash__pte_none(pte); | |
811 | } | |
812 | ||
813 | static inline void __set_pte_at(struct mm_struct *mm, unsigned long addr, | |
814 | pte_t *ptep, pte_t pte, int percpu) | |
815 | { | |
816 | if (radix_enabled()) | |
817 | return radix__set_pte_at(mm, addr, ptep, pte, percpu); | |
818 | return hash__set_pte_at(mm, addr, ptep, pte, percpu); | |
819 | } | |
34fbadd8 | 820 | |
13f829a5 AK |
821 | #define _PAGE_CACHE_CTL (_PAGE_NON_IDEMPOTENT | _PAGE_TOLERANT) |
822 | ||
823 | #define pgprot_noncached pgprot_noncached | |
824 | static inline pgprot_t pgprot_noncached(pgprot_t prot) | |
825 | { | |
826 | return __pgprot((pgprot_val(prot) & ~_PAGE_CACHE_CTL) | | |
827 | _PAGE_NON_IDEMPOTENT); | |
828 | } | |
829 | ||
830 | #define pgprot_noncached_wc pgprot_noncached_wc | |
831 | static inline pgprot_t pgprot_noncached_wc(pgprot_t prot) | |
832 | { | |
833 | return __pgprot((pgprot_val(prot) & ~_PAGE_CACHE_CTL) | | |
834 | _PAGE_TOLERANT); | |
835 | } | |
836 | ||
837 | #define pgprot_cached pgprot_cached | |
838 | static inline pgprot_t pgprot_cached(pgprot_t prot) | |
839 | { | |
840 | return __pgprot((pgprot_val(prot) & ~_PAGE_CACHE_CTL)); | |
841 | } | |
842 | ||
843 | #define pgprot_writecombine pgprot_writecombine | |
844 | static inline pgprot_t pgprot_writecombine(pgprot_t prot) | |
845 | { | |
846 | return pgprot_noncached_wc(prot); | |
847 | } | |
848 | /* | |
849 | * check a pte mapping have cache inhibited property | |
850 | */ | |
851 | static inline bool pte_ci(pte_t pte) | |
852 | { | |
853 | unsigned long pte_v = pte_val(pte); | |
854 | ||
855 | if (((pte_v & _PAGE_CACHE_CTL) == _PAGE_TOLERANT) || | |
856 | ((pte_v & _PAGE_CACHE_CTL) == _PAGE_NON_IDEMPOTENT)) | |
857 | return true; | |
858 | return false; | |
859 | } | |
860 | ||
f281b5d5 AK |
861 | static inline void pmd_set(pmd_t *pmdp, unsigned long val) |
862 | { | |
863 | *pmdp = __pmd(val); | |
864 | } | |
865 | ||
866 | static inline void pmd_clear(pmd_t *pmdp) | |
867 | { | |
868 | *pmdp = __pmd(0); | |
869 | } | |
870 | ||
66c570f5 AK |
871 | static inline int pmd_none(pmd_t pmd) |
872 | { | |
873 | return !pmd_raw(pmd); | |
874 | } | |
875 | ||
876 | static inline int pmd_present(pmd_t pmd) | |
877 | { | |
878 | ||
879 | return !pmd_none(pmd); | |
880 | } | |
3dfcb315 | 881 | |
ac94ac79 AK |
882 | static inline int pmd_bad(pmd_t pmd) |
883 | { | |
884 | if (radix_enabled()) | |
885 | return radix__pmd_bad(pmd); | |
886 | return hash__pmd_bad(pmd); | |
887 | } | |
888 | ||
f281b5d5 AK |
889 | static inline void pud_set(pud_t *pudp, unsigned long val) |
890 | { | |
891 | *pudp = __pud(val); | |
892 | } | |
893 | ||
894 | static inline void pud_clear(pud_t *pudp) | |
895 | { | |
896 | *pudp = __pud(0); | |
897 | } | |
898 | ||
66c570f5 AK |
899 | static inline int pud_none(pud_t pud) |
900 | { | |
901 | return !pud_raw(pud); | |
902 | } | |
903 | ||
904 | static inline int pud_present(pud_t pud) | |
905 | { | |
906 | return !pud_none(pud); | |
907 | } | |
3dfcb315 AK |
908 | |
909 | extern struct page *pud_page(pud_t pud); | |
371352ca | 910 | extern struct page *pmd_page(pmd_t pmd); |
3dfcb315 AK |
911 | static inline pte_t pud_pte(pud_t pud) |
912 | { | |
66c570f5 | 913 | return __pte_raw(pud_raw(pud)); |
3dfcb315 AK |
914 | } |
915 | ||
916 | static inline pud_t pte_pud(pte_t pte) | |
917 | { | |
66c570f5 | 918 | return __pud_raw(pte_raw(pte)); |
3dfcb315 AK |
919 | } |
920 | #define pud_write(pud) pte_write(pud_pte(pud)) | |
ac94ac79 AK |
921 | |
922 | static inline int pud_bad(pud_t pud) | |
923 | { | |
924 | if (radix_enabled()) | |
925 | return radix__pud_bad(pud); | |
926 | return hash__pud_bad(pud); | |
927 | } | |
928 | ||
f72a85e3 AK |
929 | #define pud_access_permitted pud_access_permitted |
930 | static inline bool pud_access_permitted(pud_t pud, bool write) | |
931 | { | |
932 | return pte_access_permitted(pud_pte(pud), write); | |
933 | } | |
ac94ac79 | 934 | |
3dfcb315 | 935 | #define pgd_write(pgd) pte_write(pgd_pte(pgd)) |
f281b5d5 AK |
936 | static inline void pgd_set(pgd_t *pgdp, unsigned long val) |
937 | { | |
938 | *pgdp = __pgd(val); | |
939 | } | |
3dfcb315 | 940 | |
368ced78 AK |
941 | static inline void pgd_clear(pgd_t *pgdp) |
942 | { | |
943 | *pgdp = __pgd(0); | |
944 | } | |
945 | ||
66c570f5 AK |
946 | static inline int pgd_none(pgd_t pgd) |
947 | { | |
948 | return !pgd_raw(pgd); | |
949 | } | |
950 | ||
951 | static inline int pgd_present(pgd_t pgd) | |
952 | { | |
953 | return !pgd_none(pgd); | |
954 | } | |
368ced78 AK |
955 | |
956 | static inline pte_t pgd_pte(pgd_t pgd) | |
957 | { | |
66c570f5 | 958 | return __pte_raw(pgd_raw(pgd)); |
368ced78 AK |
959 | } |
960 | ||
961 | static inline pgd_t pte_pgd(pte_t pte) | |
962 | { | |
66c570f5 | 963 | return __pgd_raw(pte_raw(pte)); |
368ced78 AK |
964 | } |
965 | ||
ac94ac79 AK |
966 | static inline int pgd_bad(pgd_t pgd) |
967 | { | |
968 | if (radix_enabled()) | |
969 | return radix__pgd_bad(pgd); | |
970 | return hash__pgd_bad(pgd); | |
971 | } | |
972 | ||
f72a85e3 AK |
973 | #define pgd_access_permitted pgd_access_permitted |
974 | static inline bool pgd_access_permitted(pgd_t pgd, bool write) | |
975 | { | |
976 | return pte_access_permitted(pgd_pte(pgd), write); | |
977 | } | |
978 | ||
368ced78 AK |
979 | extern struct page *pgd_page(pgd_t pgd); |
980 | ||
aba480e1 AK |
981 | /* Pointers in the page table tree are physical addresses */ |
982 | #define __pgtable_ptr_val(ptr) __pa(ptr) | |
983 | ||
984 | #define pmd_page_vaddr(pmd) __va(pmd_val(pmd) & ~PMD_MASKED_BITS) | |
985 | #define pud_page_vaddr(pud) __va(pud_val(pud) & ~PUD_MASKED_BITS) | |
986 | #define pgd_page_vaddr(pgd) __va(pgd_val(pgd) & ~PGD_MASKED_BITS) | |
987 | ||
988 | #define pgd_index(address) (((address) >> (PGDIR_SHIFT)) & (PTRS_PER_PGD - 1)) | |
989 | #define pud_index(address) (((address) >> (PUD_SHIFT)) & (PTRS_PER_PUD - 1)) | |
990 | #define pmd_index(address) (((address) >> (PMD_SHIFT)) & (PTRS_PER_PMD - 1)) | |
991 | #define pte_index(address) (((address) >> (PAGE_SHIFT)) & (PTRS_PER_PTE - 1)) | |
992 | ||
3dfcb315 AK |
993 | /* |
994 | * Find an entry in a page-table-directory. We combine the address region | |
995 | * (the high order N bits) and the pgd portion of the address. | |
996 | */ | |
3dfcb315 AK |
997 | |
998 | #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address)) | |
999 | ||
368ced78 AK |
1000 | #define pud_offset(pgdp, addr) \ |
1001 | (((pud_t *) pgd_page_vaddr(*(pgdp))) + pud_index(addr)) | |
3dfcb315 | 1002 | #define pmd_offset(pudp,addr) \ |
371352ca | 1003 | (((pmd_t *) pud_page_vaddr(*(pudp))) + pmd_index(addr)) |
3dfcb315 | 1004 | #define pte_offset_kernel(dir,addr) \ |
371352ca | 1005 | (((pte_t *) pmd_page_vaddr(*(dir))) + pte_index(addr)) |
3dfcb315 AK |
1006 | |
1007 | #define pte_offset_map(dir,addr) pte_offset_kernel((dir), (addr)) | |
1008 | #define pte_unmap(pte) do { } while(0) | |
1009 | ||
1010 | /* to find an entry in a kernel page-table-directory */ | |
1011 | /* This now only contains the vmalloc pages */ | |
1012 | #define pgd_offset_k(address) pgd_offset(&init_mm, address) | |
3dfcb315 AK |
1013 | |
1014 | #define pte_ERROR(e) \ | |
1015 | pr_err("%s:%d: bad pte %08lx.\n", __FILE__, __LINE__, pte_val(e)) | |
1016 | #define pmd_ERROR(e) \ | |
1017 | pr_err("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pmd_val(e)) | |
368ced78 AK |
1018 | #define pud_ERROR(e) \ |
1019 | pr_err("%s:%d: bad pud %08lx.\n", __FILE__, __LINE__, pud_val(e)) | |
3dfcb315 AK |
1020 | #define pgd_ERROR(e) \ |
1021 | pr_err("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e)) | |
1022 | ||
31a14fae AK |
1023 | static inline int map_kernel_page(unsigned long ea, unsigned long pa, |
1024 | unsigned long flags) | |
7207f436 | 1025 | { |
d9225ad9 AK |
1026 | if (radix_enabled()) { |
1027 | #if defined(CONFIG_PPC_RADIX_MMU) && defined(DEBUG_VM) | |
1028 | unsigned long page_size = 1 << mmu_psize_defs[mmu_io_psize].shift; | |
1029 | WARN((page_size != PAGE_SIZE), "I/O page size != PAGE_SIZE"); | |
1030 | #endif | |
1031 | return radix__map_kernel_page(ea, pa, __pgprot(flags), PAGE_SIZE); | |
1032 | } | |
31a14fae | 1033 | return hash__map_kernel_page(ea, pa, flags); |
7207f436 | 1034 | } |
31a14fae AK |
1035 | |
1036 | static inline int __meminit vmemmap_create_mapping(unsigned long start, | |
1037 | unsigned long page_size, | |
1038 | unsigned long phys) | |
7207f436 | 1039 | { |
d9225ad9 AK |
1040 | if (radix_enabled()) |
1041 | return radix__vmemmap_create_mapping(start, page_size, phys); | |
31a14fae | 1042 | return hash__vmemmap_create_mapping(start, page_size, phys); |
7207f436 | 1043 | } |
31a14fae AK |
1044 | |
1045 | #ifdef CONFIG_MEMORY_HOTPLUG | |
1046 | static inline void vmemmap_remove_mapping(unsigned long start, | |
1047 | unsigned long page_size) | |
7207f436 | 1048 | { |
d9225ad9 AK |
1049 | if (radix_enabled()) |
1050 | return radix__vmemmap_remove_mapping(start, page_size); | |
31a14fae | 1051 | return hash__vmemmap_remove_mapping(start, page_size); |
7207f436 | 1052 | } |
31a14fae | 1053 | #endif |
3dfcb315 AK |
1054 | struct page *realmode_pfn_to_page(unsigned long pfn); |
1055 | ||
3dfcb315 AK |
1056 | static inline pte_t pmd_pte(pmd_t pmd) |
1057 | { | |
66c570f5 | 1058 | return __pte_raw(pmd_raw(pmd)); |
3dfcb315 AK |
1059 | } |
1060 | ||
1061 | static inline pmd_t pte_pmd(pte_t pte) | |
1062 | { | |
66c570f5 | 1063 | return __pmd_raw(pte_raw(pte)); |
3dfcb315 AK |
1064 | } |
1065 | ||
1066 | static inline pte_t *pmdp_ptep(pmd_t *pmd) | |
1067 | { | |
1068 | return (pte_t *)pmd; | |
1069 | } | |
3dfcb315 AK |
1070 | #define pmd_pfn(pmd) pte_pfn(pmd_pte(pmd)) |
1071 | #define pmd_dirty(pmd) pte_dirty(pmd_pte(pmd)) | |
1072 | #define pmd_young(pmd) pte_young(pmd_pte(pmd)) | |
1073 | #define pmd_mkold(pmd) pte_pmd(pte_mkold(pmd_pte(pmd))) | |
1074 | #define pmd_wrprotect(pmd) pte_pmd(pte_wrprotect(pmd_pte(pmd))) | |
1075 | #define pmd_mkdirty(pmd) pte_pmd(pte_mkdirty(pmd_pte(pmd))) | |
d5d6a443 | 1076 | #define pmd_mkclean(pmd) pte_pmd(pte_mkclean(pmd_pte(pmd))) |
3dfcb315 AK |
1077 | #define pmd_mkyoung(pmd) pte_pmd(pte_mkyoung(pmd_pte(pmd))) |
1078 | #define pmd_mkwrite(pmd) pte_pmd(pte_mkwrite(pmd_pte(pmd))) | |
c137a275 AK |
1079 | #define pmd_mk_savedwrite(pmd) pte_pmd(pte_mk_savedwrite(pmd_pte(pmd))) |
1080 | #define pmd_clear_savedwrite(pmd) pte_pmd(pte_clear_savedwrite(pmd_pte(pmd))) | |
7207f436 LD |
1081 | |
1082 | #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY | |
1083 | #define pmd_soft_dirty(pmd) pte_soft_dirty(pmd_pte(pmd)) | |
1084 | #define pmd_mksoft_dirty(pmd) pte_pmd(pte_mksoft_dirty(pmd_pte(pmd))) | |
1085 | #define pmd_clear_soft_dirty(pmd) pte_pmd(pte_clear_soft_dirty(pmd_pte(pmd))) | |
1086 | #endif /* CONFIG_HAVE_ARCH_SOFT_DIRTY */ | |
1087 | ||
1ca72129 AK |
1088 | #ifdef CONFIG_NUMA_BALANCING |
1089 | static inline int pmd_protnone(pmd_t pmd) | |
1090 | { | |
1091 | return pte_protnone(pmd_pte(pmd)); | |
1092 | } | |
1093 | #endif /* CONFIG_NUMA_BALANCING */ | |
3dfcb315 | 1094 | |
3dfcb315 | 1095 | #define pmd_write(pmd) pte_write(pmd_pte(pmd)) |
d19469e8 | 1096 | #define __pmd_write(pmd) __pte_write(pmd_pte(pmd)) |
c137a275 | 1097 | #define pmd_savedwrite(pmd) pte_savedwrite(pmd_pte(pmd)) |
3dfcb315 | 1098 | |
f72a85e3 AK |
1099 | #define pmd_access_permitted pmd_access_permitted |
1100 | static inline bool pmd_access_permitted(pmd_t pmd, bool write) | |
1101 | { | |
1102 | return pte_access_permitted(pmd_pte(pmd), write); | |
1103 | } | |
1104 | ||
6a1ea362 AK |
1105 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
1106 | extern pmd_t pfn_pmd(unsigned long pfn, pgprot_t pgprot); | |
1107 | extern pmd_t mk_pmd(struct page *page, pgprot_t pgprot); | |
1108 | extern pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot); | |
1109 | extern void set_pmd_at(struct mm_struct *mm, unsigned long addr, | |
1110 | pmd_t *pmdp, pmd_t pmd); | |
1111 | extern void update_mmu_cache_pmd(struct vm_area_struct *vma, unsigned long addr, | |
1112 | pmd_t *pmd); | |
3df33f12 AK |
1113 | extern int hash__has_transparent_hugepage(void); |
1114 | static inline int has_transparent_hugepage(void) | |
1115 | { | |
bde3eb62 AK |
1116 | if (radix_enabled()) |
1117 | return radix__has_transparent_hugepage(); | |
3df33f12 AK |
1118 | return hash__has_transparent_hugepage(); |
1119 | } | |
c04a5880 | 1120 | #define has_transparent_hugepage has_transparent_hugepage |
6a1ea362 | 1121 | |
3df33f12 AK |
1122 | static inline unsigned long |
1123 | pmd_hugepage_update(struct mm_struct *mm, unsigned long addr, pmd_t *pmdp, | |
1124 | unsigned long clr, unsigned long set) | |
3dfcb315 | 1125 | { |
bde3eb62 AK |
1126 | if (radix_enabled()) |
1127 | return radix__pmd_hugepage_update(mm, addr, pmdp, clr, set); | |
3df33f12 AK |
1128 | return hash__pmd_hugepage_update(mm, addr, pmdp, clr, set); |
1129 | } | |
1130 | ||
1131 | static inline int pmd_large(pmd_t pmd) | |
1132 | { | |
66c570f5 | 1133 | return !!(pmd_raw(pmd) & cpu_to_be64(_PAGE_PTE)); |
3df33f12 AK |
1134 | } |
1135 | ||
1136 | static inline pmd_t pmd_mknotpresent(pmd_t pmd) | |
1137 | { | |
1138 | return __pmd(pmd_val(pmd) & ~_PAGE_PRESENT); | |
1139 | } | |
1140 | /* | |
1141 | * For radix we should always find H_PAGE_HASHPTE zero. Hence | |
1142 | * the below will work for radix too | |
1143 | */ | |
1144 | static inline int __pmdp_test_and_clear_young(struct mm_struct *mm, | |
1145 | unsigned long addr, pmd_t *pmdp) | |
1146 | { | |
1147 | unsigned long old; | |
1148 | ||
66c570f5 | 1149 | if ((pmd_raw(*pmdp) & cpu_to_be64(_PAGE_ACCESSED | H_PAGE_HASHPTE)) == 0) |
3df33f12 AK |
1150 | return 0; |
1151 | old = pmd_hugepage_update(mm, addr, pmdp, _PAGE_ACCESSED, 0); | |
1152 | return ((old & _PAGE_ACCESSED) != 0); | |
1153 | } | |
1154 | ||
1155 | #define __HAVE_ARCH_PMDP_SET_WRPROTECT | |
1156 | static inline void pmdp_set_wrprotect(struct mm_struct *mm, unsigned long addr, | |
1157 | pmd_t *pmdp) | |
1158 | { | |
d19469e8 | 1159 | if (__pmd_write((*pmdp))) |
52c50ca7 AK |
1160 | pmd_hugepage_update(mm, addr, pmdp, _PAGE_WRITE, 0); |
1161 | else if (unlikely(pmd_savedwrite(*pmdp))) | |
1162 | pmd_hugepage_update(mm, addr, pmdp, 0, _PAGE_PRIVILEGED); | |
3dfcb315 AK |
1163 | } |
1164 | ||
ab624762 AK |
1165 | static inline int pmd_trans_huge(pmd_t pmd) |
1166 | { | |
1167 | if (radix_enabled()) | |
1168 | return radix__pmd_trans_huge(pmd); | |
1169 | return hash__pmd_trans_huge(pmd); | |
1170 | } | |
1171 | ||
1172 | #define __HAVE_ARCH_PMD_SAME | |
1173 | static inline int pmd_same(pmd_t pmd_a, pmd_t pmd_b) | |
1174 | { | |
1175 | if (radix_enabled()) | |
1176 | return radix__pmd_same(pmd_a, pmd_b); | |
1177 | return hash__pmd_same(pmd_a, pmd_b); | |
1178 | } | |
1179 | ||
3dfcb315 AK |
1180 | static inline pmd_t pmd_mkhuge(pmd_t pmd) |
1181 | { | |
ab624762 AK |
1182 | if (radix_enabled()) |
1183 | return radix__pmd_mkhuge(pmd); | |
1184 | return hash__pmd_mkhuge(pmd); | |
3dfcb315 AK |
1185 | } |
1186 | ||
3dfcb315 AK |
1187 | #define __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS |
1188 | extern int pmdp_set_access_flags(struct vm_area_struct *vma, | |
1189 | unsigned long address, pmd_t *pmdp, | |
1190 | pmd_t entry, int dirty); | |
1191 | ||
3dfcb315 AK |
1192 | #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG |
1193 | extern int pmdp_test_and_clear_young(struct vm_area_struct *vma, | |
1194 | unsigned long address, pmd_t *pmdp); | |
3dfcb315 AK |
1195 | |
1196 | #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR | |
3df33f12 AK |
1197 | static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm, |
1198 | unsigned long addr, pmd_t *pmdp) | |
1199 | { | |
bde3eb62 AK |
1200 | if (radix_enabled()) |
1201 | return radix__pmdp_huge_get_and_clear(mm, addr, pmdp); | |
3df33f12 AK |
1202 | return hash__pmdp_huge_get_and_clear(mm, addr, pmdp); |
1203 | } | |
3dfcb315 | 1204 | |
3df33f12 AK |
1205 | static inline pmd_t pmdp_collapse_flush(struct vm_area_struct *vma, |
1206 | unsigned long address, pmd_t *pmdp) | |
1207 | { | |
bde3eb62 AK |
1208 | if (radix_enabled()) |
1209 | return radix__pmdp_collapse_flush(vma, address, pmdp); | |
3df33f12 AK |
1210 | return hash__pmdp_collapse_flush(vma, address, pmdp); |
1211 | } | |
3dfcb315 AK |
1212 | #define pmdp_collapse_flush pmdp_collapse_flush |
1213 | ||
1214 | #define __HAVE_ARCH_PGTABLE_DEPOSIT | |
3df33f12 AK |
1215 | static inline void pgtable_trans_huge_deposit(struct mm_struct *mm, |
1216 | pmd_t *pmdp, pgtable_t pgtable) | |
1217 | { | |
bde3eb62 AK |
1218 | if (radix_enabled()) |
1219 | return radix__pgtable_trans_huge_deposit(mm, pmdp, pgtable); | |
3df33f12 AK |
1220 | return hash__pgtable_trans_huge_deposit(mm, pmdp, pgtable); |
1221 | } | |
1222 | ||
3dfcb315 | 1223 | #define __HAVE_ARCH_PGTABLE_WITHDRAW |
3df33f12 AK |
1224 | static inline pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, |
1225 | pmd_t *pmdp) | |
1226 | { | |
bde3eb62 AK |
1227 | if (radix_enabled()) |
1228 | return radix__pgtable_trans_huge_withdraw(mm, pmdp); | |
3df33f12 AK |
1229 | return hash__pgtable_trans_huge_withdraw(mm, pmdp); |
1230 | } | |
3dfcb315 AK |
1231 | |
1232 | #define __HAVE_ARCH_PMDP_INVALIDATE | |
8cc931e0 AK |
1233 | extern pmd_t pmdp_invalidate(struct vm_area_struct *vma, unsigned long address, |
1234 | pmd_t *pmdp); | |
3dfcb315 AK |
1235 | |
1236 | #define pmd_move_must_withdraw pmd_move_must_withdraw | |
1237 | struct spinlock; | |
1238 | static inline int pmd_move_must_withdraw(struct spinlock *new_pmd_ptl, | |
1dd38b6c AK |
1239 | struct spinlock *old_pmd_ptl, |
1240 | struct vm_area_struct *vma) | |
3dfcb315 | 1241 | { |
bde3eb62 AK |
1242 | if (radix_enabled()) |
1243 | return false; | |
3dfcb315 AK |
1244 | /* |
1245 | * Archs like ppc64 use pgtable to store per pmd | |
1246 | * specific information. So when we switch the pmd, | |
1247 | * we should also withdraw and deposit the pgtable | |
1248 | */ | |
1249 | return true; | |
1250 | } | |
953c66c2 AK |
1251 | |
1252 | ||
1253 | #define arch_needs_pgtable_deposit arch_needs_pgtable_deposit | |
1254 | static inline bool arch_needs_pgtable_deposit(void) | |
1255 | { | |
1256 | if (radix_enabled()) | |
1257 | return false; | |
1258 | return true; | |
1259 | } | |
fa4531f7 | 1260 | extern void serialize_against_pte_lookup(struct mm_struct *mm); |
953c66c2 | 1261 | |
ebd31197 OH |
1262 | |
1263 | static inline pmd_t pmd_mkdevmap(pmd_t pmd) | |
1264 | { | |
1265 | return __pmd(pmd_val(pmd) | (_PAGE_PTE | _PAGE_DEVMAP)); | |
1266 | } | |
1267 | ||
1268 | static inline int pmd_devmap(pmd_t pmd) | |
1269 | { | |
1270 | return pte_devmap(pmd_pte(pmd)); | |
1271 | } | |
1272 | ||
1273 | static inline int pud_devmap(pud_t pud) | |
1274 | { | |
1275 | return 0; | |
1276 | } | |
1277 | ||
1278 | static inline int pgd_devmap(pgd_t pgd) | |
1279 | { | |
1280 | return 0; | |
1281 | } | |
6a1ea362 | 1282 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE */ |
ebd31197 OH |
1283 | |
1284 | static inline const int pud_pfn(pud_t pud) | |
1285 | { | |
1286 | /* | |
1287 | * Currently all calls to pud_pfn() are gated around a pud_devmap() | |
1288 | * check so this should never be used. If it grows another user we | |
1289 | * want to know about it. | |
1290 | */ | |
1291 | BUILD_BUG(); | |
1292 | return 0; | |
1293 | } | |
029d9252 | 1294 | |
3dfcb315 AK |
1295 | #endif /* __ASSEMBLY__ */ |
1296 | #endif /* _ASM_POWERPC_BOOK3S_64_PGTABLE_H_ */ |