Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
1da177e4 | 2 | * S390 version |
a53c8fab | 3 | * Copyright IBM Corp. 1999, 2000 |
1da177e4 LT |
4 | * Author(s): Hartmut Penner (hp@de.ibm.com) |
5 | * Ulrich Weigand (weigand@de.ibm.com) | |
6 | * Martin Schwidefsky (schwidefsky@de.ibm.com) | |
7 | * | |
8 | * Derived from "include/asm-i386/pgtable.h" | |
9 | */ | |
10 | ||
11 | #ifndef _ASM_S390_PGTABLE_H | |
12 | #define _ASM_S390_PGTABLE_H | |
13 | ||
1da177e4 LT |
14 | /* |
15 | * The Linux memory management assumes a three-level page table setup. For | |
16 | * s390 31 bit we "fold" the mid level into the top-level page table, so | |
17 | * that we physically have the same two-level page table as the s390 mmu | |
18 | * expects in 31 bit mode. For s390 64 bit we use three of the five levels | |
19 | * the hardware provides (region first and region second tables are not | |
20 | * used). | |
21 | * | |
22 | * The "pgd_xxx()" functions are trivial for a folded two-level | |
23 | * setup: the pgd is never bad, and a pmd always exists (as it's folded | |
24 | * into the pgd entry) | |
25 | * | |
26 | * This file contains the functions and defines necessary to modify and use | |
27 | * the S390 page table tree. | |
28 | */ | |
29 | #ifndef __ASSEMBLY__ | |
9789db08 | 30 | #include <linux/sched.h> |
2dcea57a | 31 | #include <linux/mm_types.h> |
1da177e4 | 32 | #include <asm/bug.h> |
b2fa47e6 | 33 | #include <asm/page.h> |
1da177e4 | 34 | |
1da177e4 LT |
35 | extern pgd_t swapper_pg_dir[] __attribute__ ((aligned (4096))); |
36 | extern void paging_init(void); | |
2b67fc46 | 37 | extern void vmem_map_init(void); |
92f842ea | 38 | extern void fault_init(void); |
1da177e4 LT |
39 | |
40 | /* | |
41 | * The S390 doesn't have any external MMU info: the kernel page | |
42 | * tables contain all the necessary information. | |
43 | */ | |
4b3073e1 | 44 | #define update_mmu_cache(vma, address, ptep) do { } while (0) |
b113da65 | 45 | #define update_mmu_cache_pmd(vma, address, ptep) do { } while (0) |
1da177e4 LT |
46 | |
47 | /* | |
238ec4ef | 48 | * ZERO_PAGE is a global shared page that is always zero; used |
1da177e4 LT |
49 | * for zero-mapped memory areas etc.. |
50 | */ | |
238ec4ef MS |
51 | |
52 | extern unsigned long empty_zero_page; | |
53 | extern unsigned long zero_page_mask; | |
54 | ||
55 | #define ZERO_PAGE(vaddr) \ | |
56 | (virt_to_page((void *)(empty_zero_page + \ | |
57 | (((unsigned long)(vaddr)) &zero_page_mask)))) | |
58 | ||
59 | #define is_zero_pfn is_zero_pfn | |
60 | static inline int is_zero_pfn(unsigned long pfn) | |
61 | { | |
62 | extern unsigned long zero_pfn; | |
63 | unsigned long offset_from_zero_pfn = pfn - zero_pfn; | |
64 | return offset_from_zero_pfn <= (zero_page_mask >> PAGE_SHIFT); | |
65 | } | |
66 | ||
67 | #define my_zero_pfn(addr) page_to_pfn(ZERO_PAGE(addr)) | |
68 | ||
1da177e4 LT |
69 | #endif /* !__ASSEMBLY__ */ |
70 | ||
71 | /* | |
72 | * PMD_SHIFT determines the size of the area a second-level page | |
73 | * table can map | |
74 | * PGDIR_SHIFT determines what a third-level page table entry can map | |
75 | */ | |
f4815ac6 | 76 | #ifndef CONFIG_64BIT |
146e4b3c MS |
77 | # define PMD_SHIFT 20 |
78 | # define PUD_SHIFT 20 | |
79 | # define PGDIR_SHIFT 20 | |
f4815ac6 | 80 | #else /* CONFIG_64BIT */ |
146e4b3c | 81 | # define PMD_SHIFT 20 |
190a1d72 | 82 | # define PUD_SHIFT 31 |
5a216a20 | 83 | # define PGDIR_SHIFT 42 |
f4815ac6 | 84 | #endif /* CONFIG_64BIT */ |
1da177e4 LT |
85 | |
86 | #define PMD_SIZE (1UL << PMD_SHIFT) | |
87 | #define PMD_MASK (~(PMD_SIZE-1)) | |
190a1d72 MS |
88 | #define PUD_SIZE (1UL << PUD_SHIFT) |
89 | #define PUD_MASK (~(PUD_SIZE-1)) | |
5a216a20 MS |
90 | #define PGDIR_SIZE (1UL << PGDIR_SHIFT) |
91 | #define PGDIR_MASK (~(PGDIR_SIZE-1)) | |
1da177e4 LT |
92 | |
93 | /* | |
94 | * entries per page directory level: the S390 is two-level, so | |
95 | * we don't really have any PMD directory physically. | |
96 | * for S390 segment-table entries are combined to one PGD | |
97 | * that leads to 1024 pte per pgd | |
98 | */ | |
146e4b3c | 99 | #define PTRS_PER_PTE 256 |
f4815ac6 | 100 | #ifndef CONFIG_64BIT |
146e4b3c | 101 | #define PTRS_PER_PMD 1 |
5a216a20 | 102 | #define PTRS_PER_PUD 1 |
f4815ac6 | 103 | #else /* CONFIG_64BIT */ |
146e4b3c | 104 | #define PTRS_PER_PMD 2048 |
5a216a20 | 105 | #define PTRS_PER_PUD 2048 |
f4815ac6 | 106 | #endif /* CONFIG_64BIT */ |
146e4b3c | 107 | #define PTRS_PER_PGD 2048 |
1da177e4 | 108 | |
d455a369 HD |
109 | #define FIRST_USER_ADDRESS 0 |
110 | ||
1da177e4 LT |
111 | #define pte_ERROR(e) \ |
112 | printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e)) | |
113 | #define pmd_ERROR(e) \ | |
114 | printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e)) | |
190a1d72 MS |
115 | #define pud_ERROR(e) \ |
116 | printk("%s:%d: bad pud %p.\n", __FILE__, __LINE__, (void *) pud_val(e)) | |
1da177e4 LT |
117 | #define pgd_ERROR(e) \ |
118 | printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e)) | |
119 | ||
120 | #ifndef __ASSEMBLY__ | |
121 | /* | |
5fd9c6e2 | 122 | * The vmalloc area will always be on the topmost area of the kernel |
7d3f661e | 123 | * mapping. We reserve 96MB (31bit) / 128GB (64bit) for vmalloc, |
5fd9c6e2 CB |
124 | * which should be enough for any sane case. |
125 | * By putting vmalloc at the top, we maximise the gap between physical | |
126 | * memory and vmalloc to catch misplaced memory accesses. As a side | |
127 | * effect, this also makes sure that 64 bit module code cannot be used | |
128 | * as system call address. | |
8b62bc96 | 129 | */ |
239a6425 | 130 | extern unsigned long VMALLOC_START; |
14045ebf MS |
131 | extern unsigned long VMALLOC_END; |
132 | extern struct page *vmemmap; | |
239a6425 | 133 | |
14045ebf | 134 | #define VMEM_MAX_PHYS ((unsigned long) vmemmap) |
5fd9c6e2 | 135 | |
1da177e4 LT |
136 | /* |
137 | * A 31 bit pagetable entry of S390 has following format: | |
138 | * | PFRA | | OS | | |
139 | * 0 0IP0 | |
140 | * 00000000001111111111222222222233 | |
141 | * 01234567890123456789012345678901 | |
142 | * | |
143 | * I Page-Invalid Bit: Page is not available for address-translation | |
144 | * P Page-Protection Bit: Store access not possible for page | |
145 | * | |
146 | * A 31 bit segmenttable entry of S390 has following format: | |
147 | * | P-table origin | |PTL | |
148 | * 0 IC | |
149 | * 00000000001111111111222222222233 | |
150 | * 01234567890123456789012345678901 | |
151 | * | |
152 | * I Segment-Invalid Bit: Segment is not available for address-translation | |
153 | * C Common-Segment Bit: Segment is not private (PoP 3-30) | |
154 | * PTL Page-Table-Length: Page-table length (PTL+1*16 entries -> up to 256) | |
155 | * | |
156 | * The 31 bit segmenttable origin of S390 has following format: | |
157 | * | |
158 | * |S-table origin | | STL | | |
159 | * X **GPS | |
160 | * 00000000001111111111222222222233 | |
161 | * 01234567890123456789012345678901 | |
162 | * | |
163 | * X Space-Switch event: | |
164 | * G Segment-Invalid Bit: * | |
165 | * P Private-Space Bit: Segment is not private (PoP 3-30) | |
166 | * S Storage-Alteration: | |
167 | * STL Segment-Table-Length: Segment-table length (STL+1*16 entries -> up to 2048) | |
168 | * | |
169 | * A 64 bit pagetable entry of S390 has following format: | |
6a985c61 | 170 | * | PFRA |0IPC| OS | |
1da177e4 LT |
171 | * 0000000000111111111122222222223333333333444444444455555555556666 |
172 | * 0123456789012345678901234567890123456789012345678901234567890123 | |
173 | * | |
174 | * I Page-Invalid Bit: Page is not available for address-translation | |
175 | * P Page-Protection Bit: Store access not possible for page | |
6a985c61 | 176 | * C Change-bit override: HW is not required to set change bit |
1da177e4 LT |
177 | * |
178 | * A 64 bit segmenttable entry of S390 has following format: | |
179 | * | P-table origin | TT | |
180 | * 0000000000111111111122222222223333333333444444444455555555556666 | |
181 | * 0123456789012345678901234567890123456789012345678901234567890123 | |
182 | * | |
183 | * I Segment-Invalid Bit: Segment is not available for address-translation | |
184 | * C Common-Segment Bit: Segment is not private (PoP 3-30) | |
185 | * P Page-Protection Bit: Store access not possible for page | |
186 | * TT Type 00 | |
187 | * | |
188 | * A 64 bit region table entry of S390 has following format: | |
189 | * | S-table origin | TF TTTL | |
190 | * 0000000000111111111122222222223333333333444444444455555555556666 | |
191 | * 0123456789012345678901234567890123456789012345678901234567890123 | |
192 | * | |
193 | * I Segment-Invalid Bit: Segment is not available for address-translation | |
194 | * TT Type 01 | |
195 | * TF | |
190a1d72 | 196 | * TL Table length |
1da177e4 LT |
197 | * |
198 | * The 64 bit regiontable origin of S390 has following format: | |
199 | * | region table origon | DTTL | |
200 | * 0000000000111111111122222222223333333333444444444455555555556666 | |
201 | * 0123456789012345678901234567890123456789012345678901234567890123 | |
202 | * | |
203 | * X Space-Switch event: | |
204 | * G Segment-Invalid Bit: | |
205 | * P Private-Space Bit: | |
206 | * S Storage-Alteration: | |
207 | * R Real space | |
208 | * TL Table-Length: | |
209 | * | |
210 | * A storage key has the following format: | |
211 | * | ACC |F|R|C|0| | |
212 | * 0 3 4 5 6 7 | |
213 | * ACC: access key | |
214 | * F : fetch protection bit | |
215 | * R : referenced bit | |
216 | * C : changed bit | |
217 | */ | |
218 | ||
219 | /* Hardware bits in the page table entry */ | |
6a985c61 | 220 | #define _PAGE_CO 0x100 /* HW Change-bit override */ |
83377484 MS |
221 | #define _PAGE_RO 0x200 /* HW read-only bit */ |
222 | #define _PAGE_INVALID 0x400 /* HW invalid bit */ | |
3610cce8 MS |
223 | |
224 | /* Software bits in the page table entry */ | |
83377484 MS |
225 | #define _PAGE_SWT 0x001 /* SW pte type bit t */ |
226 | #define _PAGE_SWX 0x002 /* SW pte type bit x */ | |
b2fa47e6 MS |
227 | #define _PAGE_SWC 0x004 /* SW pte changed bit (for KVM) */ |
228 | #define _PAGE_SWR 0x008 /* SW pte referenced bit (for KVM) */ | |
229 | #define _PAGE_SPECIAL 0x010 /* SW associated with special page */ | |
a08cb629 | 230 | #define __HAVE_ARCH_PTE_SPECIAL |
1da177e4 | 231 | |
138c9021 | 232 | /* Set of bits not changed in pte_modify */ |
b2fa47e6 | 233 | #define _PAGE_CHG_MASK (PAGE_MASK | _PAGE_SPECIAL | _PAGE_SWC | _PAGE_SWR) |
138c9021 | 234 | |
83377484 | 235 | /* Six different types of pages. */ |
9282ed92 GS |
236 | #define _PAGE_TYPE_EMPTY 0x400 |
237 | #define _PAGE_TYPE_NONE 0x401 | |
83377484 MS |
238 | #define _PAGE_TYPE_SWAP 0x403 |
239 | #define _PAGE_TYPE_FILE 0x601 /* bit 0x002 is used for offset !! */ | |
9282ed92 GS |
240 | #define _PAGE_TYPE_RO 0x200 |
241 | #define _PAGE_TYPE_RW 0x000 | |
1da177e4 | 242 | |
53492b1d GS |
243 | /* |
244 | * Only four types for huge pages, using the invalid bit and protection bit | |
245 | * of a segment table entry. | |
246 | */ | |
247 | #define _HPAGE_TYPE_EMPTY 0x020 /* _SEGMENT_ENTRY_INV */ | |
248 | #define _HPAGE_TYPE_NONE 0x220 | |
249 | #define _HPAGE_TYPE_RO 0x200 /* _SEGMENT_ENTRY_RO */ | |
250 | #define _HPAGE_TYPE_RW 0x000 | |
251 | ||
83377484 MS |
252 | /* |
253 | * PTE type bits are rather complicated. handle_pte_fault uses pte_present, | |
254 | * pte_none and pte_file to find out the pte type WITHOUT holding the page | |
255 | * table lock. ptep_clear_flush on the other hand uses ptep_clear_flush to | |
256 | * invalidate a given pte. ipte sets the hw invalid bit and clears all tlbs | |
257 | * for the page. The page table entry is set to _PAGE_TYPE_EMPTY afterwards. | |
258 | * This change is done while holding the lock, but the intermediate step | |
259 | * of a previously valid pte with the hw invalid bit set can be observed by | |
260 | * handle_pte_fault. That makes it necessary that all valid pte types with | |
261 | * the hw invalid bit set must be distinguishable from the four pte types | |
262 | * empty, none, swap and file. | |
263 | * | |
264 | * irxt ipte irxt | |
265 | * _PAGE_TYPE_EMPTY 1000 -> 1000 | |
266 | * _PAGE_TYPE_NONE 1001 -> 1001 | |
267 | * _PAGE_TYPE_SWAP 1011 -> 1011 | |
268 | * _PAGE_TYPE_FILE 11?1 -> 11?1 | |
269 | * _PAGE_TYPE_RO 0100 -> 1100 | |
270 | * _PAGE_TYPE_RW 0000 -> 1000 | |
271 | * | |
c1821c2e | 272 | * pte_none is true for bits combinations 1000, 1010, 1100, 1110 |
83377484 MS |
273 | * pte_present is true for bits combinations 0000, 0010, 0100, 0110, 1001 |
274 | * pte_file is true for bits combinations 1101, 1111 | |
c1821c2e | 275 | * swap pte is 1011 and 0001, 0011, 0101, 0111 are invalid. |
83377484 MS |
276 | */ |
277 | ||
f4815ac6 | 278 | #ifndef CONFIG_64BIT |
1da177e4 | 279 | |
3610cce8 MS |
280 | /* Bits in the segment table address-space-control-element */ |
281 | #define _ASCE_SPACE_SWITCH 0x80000000UL /* space switch event */ | |
282 | #define _ASCE_ORIGIN_MASK 0x7ffff000UL /* segment table origin */ | |
283 | #define _ASCE_PRIVATE_SPACE 0x100 /* private space control */ | |
284 | #define _ASCE_ALT_EVENT 0x80 /* storage alteration event control */ | |
285 | #define _ASCE_TABLE_LENGTH 0x7f /* 128 x 64 entries = 8k */ | |
1da177e4 | 286 | |
3610cce8 MS |
287 | /* Bits in the segment table entry */ |
288 | #define _SEGMENT_ENTRY_ORIGIN 0x7fffffc0UL /* page table origin */ | |
80217147 | 289 | #define _SEGMENT_ENTRY_RO 0x200 /* page protection bit */ |
3610cce8 MS |
290 | #define _SEGMENT_ENTRY_INV 0x20 /* invalid segment table entry */ |
291 | #define _SEGMENT_ENTRY_COMMON 0x10 /* common segment bit */ | |
292 | #define _SEGMENT_ENTRY_PTL 0x0f /* page table length */ | |
1da177e4 | 293 | |
3610cce8 MS |
294 | #define _SEGMENT_ENTRY (_SEGMENT_ENTRY_PTL) |
295 | #define _SEGMENT_ENTRY_EMPTY (_SEGMENT_ENTRY_INV) | |
1da177e4 | 296 | |
6c61cfe9 MS |
297 | /* Page status table bits for virtualization */ |
298 | #define RCP_ACC_BITS 0xf0000000UL | |
299 | #define RCP_FP_BIT 0x08000000UL | |
300 | #define RCP_PCL_BIT 0x00800000UL | |
301 | #define RCP_HR_BIT 0x00400000UL | |
302 | #define RCP_HC_BIT 0x00200000UL | |
303 | #define RCP_GR_BIT 0x00040000UL | |
304 | #define RCP_GC_BIT 0x00020000UL | |
305 | ||
306 | /* User dirty / referenced bit for KVM's migration feature */ | |
307 | #define KVM_UR_BIT 0x00008000UL | |
308 | #define KVM_UC_BIT 0x00004000UL | |
309 | ||
f4815ac6 | 310 | #else /* CONFIG_64BIT */ |
1da177e4 | 311 | |
3610cce8 MS |
312 | /* Bits in the segment/region table address-space-control-element */ |
313 | #define _ASCE_ORIGIN ~0xfffUL/* segment table origin */ | |
314 | #define _ASCE_PRIVATE_SPACE 0x100 /* private space control */ | |
315 | #define _ASCE_ALT_EVENT 0x80 /* storage alteration event control */ | |
316 | #define _ASCE_SPACE_SWITCH 0x40 /* space switch event */ | |
317 | #define _ASCE_REAL_SPACE 0x20 /* real space control */ | |
318 | #define _ASCE_TYPE_MASK 0x0c /* asce table type mask */ | |
319 | #define _ASCE_TYPE_REGION1 0x0c /* region first table type */ | |
320 | #define _ASCE_TYPE_REGION2 0x08 /* region second table type */ | |
321 | #define _ASCE_TYPE_REGION3 0x04 /* region third table type */ | |
322 | #define _ASCE_TYPE_SEGMENT 0x00 /* segment table type */ | |
323 | #define _ASCE_TABLE_LENGTH 0x03 /* region table length */ | |
324 | ||
325 | /* Bits in the region table entry */ | |
326 | #define _REGION_ENTRY_ORIGIN ~0xfffUL/* region/segment table origin */ | |
327 | #define _REGION_ENTRY_INV 0x20 /* invalid region table entry */ | |
328 | #define _REGION_ENTRY_TYPE_MASK 0x0c /* region/segment table type mask */ | |
329 | #define _REGION_ENTRY_TYPE_R1 0x0c /* region first table type */ | |
330 | #define _REGION_ENTRY_TYPE_R2 0x08 /* region second table type */ | |
331 | #define _REGION_ENTRY_TYPE_R3 0x04 /* region third table type */ | |
332 | #define _REGION_ENTRY_LENGTH 0x03 /* region third length */ | |
333 | ||
334 | #define _REGION1_ENTRY (_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH) | |
335 | #define _REGION1_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INV) | |
336 | #define _REGION2_ENTRY (_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH) | |
337 | #define _REGION2_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INV) | |
338 | #define _REGION3_ENTRY (_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH) | |
339 | #define _REGION3_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INV) | |
340 | ||
1da177e4 | 341 | /* Bits in the segment table entry */ |
3610cce8 MS |
342 | #define _SEGMENT_ENTRY_ORIGIN ~0x7ffUL/* segment table origin */ |
343 | #define _SEGMENT_ENTRY_RO 0x200 /* page protection bit */ | |
344 | #define _SEGMENT_ENTRY_INV 0x20 /* invalid segment table entry */ | |
1da177e4 | 345 | |
3610cce8 MS |
346 | #define _SEGMENT_ENTRY (0) |
347 | #define _SEGMENT_ENTRY_EMPTY (_SEGMENT_ENTRY_INV) | |
348 | ||
53492b1d GS |
349 | #define _SEGMENT_ENTRY_LARGE 0x400 /* STE-format control, large page */ |
350 | #define _SEGMENT_ENTRY_CO 0x100 /* change-recording override */ | |
75077afb GS |
351 | #define _SEGMENT_ENTRY_SPLIT_BIT 0 /* THP splitting bit number */ |
352 | #define _SEGMENT_ENTRY_SPLIT (1UL << _SEGMENT_ENTRY_SPLIT_BIT) | |
53492b1d | 353 | |
1ae1c1d0 GS |
354 | /* Set of bits not changed in pmd_modify */ |
355 | #define _SEGMENT_CHG_MASK (_SEGMENT_ENTRY_ORIGIN | _SEGMENT_ENTRY_LARGE \ | |
356 | | _SEGMENT_ENTRY_SPLIT | _SEGMENT_ENTRY_CO) | |
357 | ||
6c61cfe9 MS |
358 | /* Page status table bits for virtualization */ |
359 | #define RCP_ACC_BITS 0xf000000000000000UL | |
360 | #define RCP_FP_BIT 0x0800000000000000UL | |
361 | #define RCP_PCL_BIT 0x0080000000000000UL | |
362 | #define RCP_HR_BIT 0x0040000000000000UL | |
363 | #define RCP_HC_BIT 0x0020000000000000UL | |
364 | #define RCP_GR_BIT 0x0004000000000000UL | |
365 | #define RCP_GC_BIT 0x0002000000000000UL | |
366 | ||
367 | /* User dirty / referenced bit for KVM's migration feature */ | |
368 | #define KVM_UR_BIT 0x0000800000000000UL | |
369 | #define KVM_UC_BIT 0x0000400000000000UL | |
370 | ||
f4815ac6 | 371 | #endif /* CONFIG_64BIT */ |
1da177e4 LT |
372 | |
373 | /* | |
3610cce8 MS |
374 | * A user page table pointer has the space-switch-event bit, the |
375 | * private-space-control bit and the storage-alteration-event-control | |
376 | * bit set. A kernel page table pointer doesn't need them. | |
1da177e4 | 377 | */ |
3610cce8 MS |
378 | #define _ASCE_USER_BITS (_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \ |
379 | _ASCE_ALT_EVENT) | |
1da177e4 | 380 | |
1da177e4 | 381 | /* |
9282ed92 | 382 | * Page protection definitions. |
1da177e4 | 383 | */ |
9282ed92 GS |
384 | #define PAGE_NONE __pgprot(_PAGE_TYPE_NONE) |
385 | #define PAGE_RO __pgprot(_PAGE_TYPE_RO) | |
386 | #define PAGE_RW __pgprot(_PAGE_TYPE_RW) | |
387 | ||
388 | #define PAGE_KERNEL PAGE_RW | |
389 | #define PAGE_COPY PAGE_RO | |
1da177e4 LT |
390 | |
391 | /* | |
043d0708 MS |
392 | * On s390 the page table entry has an invalid bit and a read-only bit. |
393 | * Read permission implies execute permission and write permission | |
394 | * implies read permission. | |
1da177e4 LT |
395 | */ |
396 | /*xwr*/ | |
9282ed92 GS |
397 | #define __P000 PAGE_NONE |
398 | #define __P001 PAGE_RO | |
399 | #define __P010 PAGE_RO | |
400 | #define __P011 PAGE_RO | |
043d0708 MS |
401 | #define __P100 PAGE_RO |
402 | #define __P101 PAGE_RO | |
403 | #define __P110 PAGE_RO | |
404 | #define __P111 PAGE_RO | |
9282ed92 GS |
405 | |
406 | #define __S000 PAGE_NONE | |
407 | #define __S001 PAGE_RO | |
408 | #define __S010 PAGE_RW | |
409 | #define __S011 PAGE_RW | |
043d0708 MS |
410 | #define __S100 PAGE_RO |
411 | #define __S101 PAGE_RO | |
412 | #define __S110 PAGE_RW | |
413 | #define __S111 PAGE_RW | |
1da177e4 | 414 | |
b2fa47e6 | 415 | static inline int mm_exclusive(struct mm_struct *mm) |
1da177e4 | 416 | { |
b2fa47e6 MS |
417 | return likely(mm == current->active_mm && |
418 | atomic_read(&mm->context.attach_count) <= 1); | |
1da177e4 | 419 | } |
1da177e4 | 420 | |
b2fa47e6 MS |
421 | static inline int mm_has_pgste(struct mm_struct *mm) |
422 | { | |
423 | #ifdef CONFIG_PGSTE | |
424 | if (unlikely(mm->context.has_pgste)) | |
425 | return 1; | |
426 | #endif | |
427 | return 0; | |
428 | } | |
1da177e4 LT |
429 | /* |
430 | * pgd/pmd/pte query functions | |
431 | */ | |
f4815ac6 | 432 | #ifndef CONFIG_64BIT |
1da177e4 | 433 | |
4448aaf0 AB |
434 | static inline int pgd_present(pgd_t pgd) { return 1; } |
435 | static inline int pgd_none(pgd_t pgd) { return 0; } | |
436 | static inline int pgd_bad(pgd_t pgd) { return 0; } | |
1da177e4 | 437 | |
190a1d72 MS |
438 | static inline int pud_present(pud_t pud) { return 1; } |
439 | static inline int pud_none(pud_t pud) { return 0; } | |
440 | static inline int pud_bad(pud_t pud) { return 0; } | |
441 | ||
f4815ac6 | 442 | #else /* CONFIG_64BIT */ |
1da177e4 | 443 | |
5a216a20 MS |
444 | static inline int pgd_present(pgd_t pgd) |
445 | { | |
6252d702 MS |
446 | if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2) |
447 | return 1; | |
5a216a20 MS |
448 | return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL; |
449 | } | |
450 | ||
451 | static inline int pgd_none(pgd_t pgd) | |
452 | { | |
6252d702 MS |
453 | if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2) |
454 | return 0; | |
5a216a20 MS |
455 | return (pgd_val(pgd) & _REGION_ENTRY_INV) != 0UL; |
456 | } | |
457 | ||
458 | static inline int pgd_bad(pgd_t pgd) | |
459 | { | |
6252d702 MS |
460 | /* |
461 | * With dynamic page table levels the pgd can be a region table | |
462 | * entry or a segment table entry. Check for the bit that are | |
463 | * invalid for either table entry. | |
464 | */ | |
5a216a20 | 465 | unsigned long mask = |
6252d702 | 466 | ~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INV & |
5a216a20 MS |
467 | ~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH; |
468 | return (pgd_val(pgd) & mask) != 0; | |
469 | } | |
190a1d72 MS |
470 | |
471 | static inline int pud_present(pud_t pud) | |
1da177e4 | 472 | { |
6252d702 MS |
473 | if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3) |
474 | return 1; | |
0d017923 | 475 | return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL; |
1da177e4 LT |
476 | } |
477 | ||
190a1d72 | 478 | static inline int pud_none(pud_t pud) |
1da177e4 | 479 | { |
6252d702 MS |
480 | if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3) |
481 | return 0; | |
0d017923 | 482 | return (pud_val(pud) & _REGION_ENTRY_INV) != 0UL; |
1da177e4 LT |
483 | } |
484 | ||
190a1d72 | 485 | static inline int pud_bad(pud_t pud) |
1da177e4 | 486 | { |
6252d702 MS |
487 | /* |
488 | * With dynamic page table levels the pud can be a region table | |
489 | * entry or a segment table entry. Check for the bit that are | |
490 | * invalid for either table entry. | |
491 | */ | |
5a216a20 | 492 | unsigned long mask = |
6252d702 | 493 | ~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INV & |
5a216a20 MS |
494 | ~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH; |
495 | return (pud_val(pud) & mask) != 0; | |
1da177e4 LT |
496 | } |
497 | ||
f4815ac6 | 498 | #endif /* CONFIG_64BIT */ |
3610cce8 | 499 | |
4448aaf0 | 500 | static inline int pmd_present(pmd_t pmd) |
1da177e4 | 501 | { |
0d017923 | 502 | return (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN) != 0UL; |
1da177e4 LT |
503 | } |
504 | ||
4448aaf0 | 505 | static inline int pmd_none(pmd_t pmd) |
1da177e4 | 506 | { |
0d017923 | 507 | return (pmd_val(pmd) & _SEGMENT_ENTRY_INV) != 0UL; |
1da177e4 LT |
508 | } |
509 | ||
4448aaf0 | 510 | static inline int pmd_bad(pmd_t pmd) |
1da177e4 | 511 | { |
3610cce8 MS |
512 | unsigned long mask = ~_SEGMENT_ENTRY_ORIGIN & ~_SEGMENT_ENTRY_INV; |
513 | return (pmd_val(pmd) & mask) != _SEGMENT_ENTRY; | |
1da177e4 LT |
514 | } |
515 | ||
75077afb GS |
516 | #define __HAVE_ARCH_PMDP_SPLITTING_FLUSH |
517 | extern void pmdp_splitting_flush(struct vm_area_struct *vma, | |
518 | unsigned long addr, pmd_t *pmdp); | |
519 | ||
1ae1c1d0 GS |
520 | #define __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS |
521 | extern int pmdp_set_access_flags(struct vm_area_struct *vma, | |
522 | unsigned long address, pmd_t *pmdp, | |
523 | pmd_t entry, int dirty); | |
524 | ||
525 | #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH | |
526 | extern int pmdp_clear_flush_young(struct vm_area_struct *vma, | |
527 | unsigned long address, pmd_t *pmdp); | |
528 | ||
529 | #define __HAVE_ARCH_PMD_WRITE | |
530 | static inline int pmd_write(pmd_t pmd) | |
531 | { | |
532 | return (pmd_val(pmd) & _SEGMENT_ENTRY_RO) == 0; | |
533 | } | |
534 | ||
535 | static inline int pmd_young(pmd_t pmd) | |
536 | { | |
537 | return 0; | |
538 | } | |
539 | ||
4448aaf0 | 540 | static inline int pte_none(pte_t pte) |
1da177e4 | 541 | { |
83377484 | 542 | return (pte_val(pte) & _PAGE_INVALID) && !(pte_val(pte) & _PAGE_SWT); |
1da177e4 LT |
543 | } |
544 | ||
4448aaf0 | 545 | static inline int pte_present(pte_t pte) |
1da177e4 | 546 | { |
83377484 MS |
547 | unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT | _PAGE_SWX; |
548 | return (pte_val(pte) & mask) == _PAGE_TYPE_NONE || | |
549 | (!(pte_val(pte) & _PAGE_INVALID) && | |
550 | !(pte_val(pte) & _PAGE_SWT)); | |
1da177e4 LT |
551 | } |
552 | ||
4448aaf0 | 553 | static inline int pte_file(pte_t pte) |
1da177e4 | 554 | { |
83377484 MS |
555 | unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT; |
556 | return (pte_val(pte) & mask) == _PAGE_TYPE_FILE; | |
1da177e4 LT |
557 | } |
558 | ||
7e675137 NP |
559 | static inline int pte_special(pte_t pte) |
560 | { | |
a08cb629 | 561 | return (pte_val(pte) & _PAGE_SPECIAL); |
7e675137 NP |
562 | } |
563 | ||
ba8a9229 | 564 | #define __HAVE_ARCH_PTE_SAME |
b2fa47e6 MS |
565 | static inline int pte_same(pte_t a, pte_t b) |
566 | { | |
567 | return pte_val(a) == pte_val(b); | |
568 | } | |
1da177e4 | 569 | |
b2fa47e6 | 570 | static inline pgste_t pgste_get_lock(pte_t *ptep) |
5b7baf05 | 571 | { |
b2fa47e6 | 572 | unsigned long new = 0; |
5b7baf05 | 573 | #ifdef CONFIG_PGSTE |
b2fa47e6 MS |
574 | unsigned long old; |
575 | ||
5b7baf05 | 576 | preempt_disable(); |
b2fa47e6 MS |
577 | asm( |
578 | " lg %0,%2\n" | |
579 | "0: lgr %1,%0\n" | |
580 | " nihh %0,0xff7f\n" /* clear RCP_PCL_BIT in old */ | |
581 | " oihh %1,0x0080\n" /* set RCP_PCL_BIT in new */ | |
582 | " csg %0,%1,%2\n" | |
583 | " jl 0b\n" | |
584 | : "=&d" (old), "=&d" (new), "=Q" (ptep[PTRS_PER_PTE]) | |
585 | : "Q" (ptep[PTRS_PER_PTE]) : "cc"); | |
5b7baf05 | 586 | #endif |
b2fa47e6 | 587 | return __pgste(new); |
5b7baf05 CB |
588 | } |
589 | ||
b2fa47e6 | 590 | static inline void pgste_set_unlock(pte_t *ptep, pgste_t pgste) |
5b7baf05 CB |
591 | { |
592 | #ifdef CONFIG_PGSTE | |
b2fa47e6 MS |
593 | asm( |
594 | " nihh %1,0xff7f\n" /* clear RCP_PCL_BIT */ | |
595 | " stg %1,%0\n" | |
596 | : "=Q" (ptep[PTRS_PER_PTE]) | |
597 | : "d" (pgste_val(pgste)), "Q" (ptep[PTRS_PER_PTE]) : "cc"); | |
5b7baf05 CB |
598 | preempt_enable(); |
599 | #endif | |
600 | } | |
601 | ||
b2fa47e6 | 602 | static inline pgste_t pgste_update_all(pte_t *ptep, pgste_t pgste) |
5b7baf05 CB |
603 | { |
604 | #ifdef CONFIG_PGSTE | |
a43a9d93 | 605 | unsigned long address, bits; |
b2fa47e6 MS |
606 | unsigned char skey; |
607 | ||
09b53883 MS |
608 | if (!pte_present(*ptep)) |
609 | return pgste; | |
a43a9d93 HC |
610 | address = pte_val(*ptep) & PAGE_MASK; |
611 | skey = page_get_storage_key(address); | |
b2fa47e6 MS |
612 | bits = skey & (_PAGE_CHANGED | _PAGE_REFERENCED); |
613 | /* Clear page changed & referenced bit in the storage key */ | |
7c81878b CO |
614 | if (bits & _PAGE_CHANGED) |
615 | page_set_storage_key(address, skey ^ bits, 1); | |
616 | else if (bits) | |
617 | page_reset_referenced(address); | |
b2fa47e6 MS |
618 | /* Transfer page changed & referenced bit to guest bits in pgste */ |
619 | pgste_val(pgste) |= bits << 48; /* RCP_GR_BIT & RCP_GC_BIT */ | |
620 | /* Get host changed & referenced bits from pgste */ | |
621 | bits |= (pgste_val(pgste) & (RCP_HR_BIT | RCP_HC_BIT)) >> 52; | |
622 | /* Clear host bits in pgste. */ | |
623 | pgste_val(pgste) &= ~(RCP_HR_BIT | RCP_HC_BIT); | |
624 | pgste_val(pgste) &= ~(RCP_ACC_BITS | RCP_FP_BIT); | |
625 | /* Copy page access key and fetch protection bit to pgste */ | |
626 | pgste_val(pgste) |= | |
627 | (unsigned long) (skey & (_PAGE_ACC_BITS | _PAGE_FP_BIT)) << 56; | |
628 | /* Transfer changed and referenced to kvm user bits */ | |
629 | pgste_val(pgste) |= bits << 45; /* KVM_UR_BIT & KVM_UC_BIT */ | |
630 | /* Transfer changed & referenced to pte sofware bits */ | |
631 | pte_val(*ptep) |= bits << 1; /* _PAGE_SWR & _PAGE_SWC */ | |
632 | #endif | |
633 | return pgste; | |
634 | ||
635 | } | |
636 | ||
637 | static inline pgste_t pgste_update_young(pte_t *ptep, pgste_t pgste) | |
638 | { | |
639 | #ifdef CONFIG_PGSTE | |
640 | int young; | |
641 | ||
09b53883 MS |
642 | if (!pte_present(*ptep)) |
643 | return pgste; | |
b2fa47e6 MS |
644 | young = page_reset_referenced(pte_val(*ptep) & PAGE_MASK); |
645 | /* Transfer page referenced bit to pte software bit (host view) */ | |
646 | if (young || (pgste_val(pgste) & RCP_HR_BIT)) | |
647 | pte_val(*ptep) |= _PAGE_SWR; | |
648 | /* Clear host referenced bit in pgste. */ | |
649 | pgste_val(pgste) &= ~RCP_HR_BIT; | |
650 | /* Transfer page referenced bit to guest bit in pgste */ | |
651 | pgste_val(pgste) |= (unsigned long) young << 50; /* set RCP_GR_BIT */ | |
652 | #endif | |
653 | return pgste; | |
654 | ||
655 | } | |
656 | ||
09b53883 | 657 | static inline void pgste_set_pte(pte_t *ptep, pgste_t pgste, pte_t entry) |
b2fa47e6 MS |
658 | { |
659 | #ifdef CONFIG_PGSTE | |
a43a9d93 | 660 | unsigned long address; |
b2fa47e6 MS |
661 | unsigned long okey, nkey; |
662 | ||
09b53883 MS |
663 | if (!pte_present(entry)) |
664 | return; | |
665 | address = pte_val(entry) & PAGE_MASK; | |
a43a9d93 | 666 | okey = nkey = page_get_storage_key(address); |
b2fa47e6 MS |
667 | nkey &= ~(_PAGE_ACC_BITS | _PAGE_FP_BIT); |
668 | /* Set page access key and fetch protection bit from pgste */ | |
669 | nkey |= (pgste_val(pgste) & (RCP_ACC_BITS | RCP_FP_BIT)) >> 56; | |
670 | if (okey != nkey) | |
a43a9d93 | 671 | page_set_storage_key(address, nkey, 1); |
5b7baf05 CB |
672 | #endif |
673 | } | |
674 | ||
e5992f2e MS |
675 | /** |
676 | * struct gmap_struct - guest address space | |
677 | * @mm: pointer to the parent mm_struct | |
678 | * @table: pointer to the page directory | |
480e5926 | 679 | * @asce: address space control element for gmap page table |
e5992f2e MS |
680 | * @crst_list: list of all crst tables used in the guest address space |
681 | */ | |
682 | struct gmap { | |
683 | struct list_head list; | |
684 | struct mm_struct *mm; | |
685 | unsigned long *table; | |
480e5926 | 686 | unsigned long asce; |
e5992f2e MS |
687 | struct list_head crst_list; |
688 | }; | |
689 | ||
690 | /** | |
691 | * struct gmap_rmap - reverse mapping for segment table entries | |
692 | * @next: pointer to the next gmap_rmap structure in the list | |
693 | * @entry: pointer to a segment table entry | |
694 | */ | |
695 | struct gmap_rmap { | |
696 | struct list_head list; | |
697 | unsigned long *entry; | |
698 | }; | |
699 | ||
700 | /** | |
701 | * struct gmap_pgtable - gmap information attached to a page table | |
702 | * @vmaddr: address of the 1MB segment in the process virtual memory | |
703 | * @mapper: list of segment table entries maping a page table | |
704 | */ | |
705 | struct gmap_pgtable { | |
706 | unsigned long vmaddr; | |
707 | struct list_head mapper; | |
708 | }; | |
709 | ||
710 | struct gmap *gmap_alloc(struct mm_struct *mm); | |
711 | void gmap_free(struct gmap *gmap); | |
712 | void gmap_enable(struct gmap *gmap); | |
713 | void gmap_disable(struct gmap *gmap); | |
714 | int gmap_map_segment(struct gmap *gmap, unsigned long from, | |
715 | unsigned long to, unsigned long length); | |
716 | int gmap_unmap_segment(struct gmap *gmap, unsigned long to, unsigned long len); | |
499069e1 | 717 | unsigned long __gmap_fault(unsigned long address, struct gmap *); |
e5992f2e | 718 | unsigned long gmap_fault(unsigned long address, struct gmap *); |
388186bc | 719 | void gmap_discard(unsigned long from, unsigned long to, struct gmap *); |
e5992f2e | 720 | |
b2fa47e6 MS |
721 | /* |
722 | * Certain architectures need to do special things when PTEs | |
723 | * within a page table are directly modified. Thus, the following | |
724 | * hook is made available. | |
725 | */ | |
726 | static inline void set_pte_at(struct mm_struct *mm, unsigned long addr, | |
727 | pte_t *ptep, pte_t entry) | |
728 | { | |
729 | pgste_t pgste; | |
730 | ||
731 | if (mm_has_pgste(mm)) { | |
732 | pgste = pgste_get_lock(ptep); | |
09b53883 | 733 | pgste_set_pte(ptep, pgste, entry); |
b2fa47e6 MS |
734 | *ptep = entry; |
735 | pgste_set_unlock(ptep, pgste); | |
736 | } else | |
737 | *ptep = entry; | |
738 | } | |
739 | ||
1da177e4 LT |
740 | /* |
741 | * query functions pte_write/pte_dirty/pte_young only work if | |
742 | * pte_present() is true. Undefined behaviour if not.. | |
743 | */ | |
4448aaf0 | 744 | static inline int pte_write(pte_t pte) |
1da177e4 LT |
745 | { |
746 | return (pte_val(pte) & _PAGE_RO) == 0; | |
747 | } | |
748 | ||
4448aaf0 | 749 | static inline int pte_dirty(pte_t pte) |
1da177e4 | 750 | { |
b2fa47e6 MS |
751 | #ifdef CONFIG_PGSTE |
752 | if (pte_val(pte) & _PAGE_SWC) | |
753 | return 1; | |
754 | #endif | |
1da177e4 LT |
755 | return 0; |
756 | } | |
757 | ||
4448aaf0 | 758 | static inline int pte_young(pte_t pte) |
1da177e4 | 759 | { |
b2fa47e6 MS |
760 | #ifdef CONFIG_PGSTE |
761 | if (pte_val(pte) & _PAGE_SWR) | |
762 | return 1; | |
763 | #endif | |
1da177e4 LT |
764 | return 0; |
765 | } | |
766 | ||
1da177e4 LT |
767 | /* |
768 | * pgd/pmd/pte modification functions | |
769 | */ | |
770 | ||
b2fa47e6 | 771 | static inline void pgd_clear(pgd_t *pgd) |
5a216a20 | 772 | { |
f4815ac6 | 773 | #ifdef CONFIG_64BIT |
6252d702 MS |
774 | if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2) |
775 | pgd_val(*pgd) = _REGION2_ENTRY_EMPTY; | |
b2fa47e6 | 776 | #endif |
5a216a20 MS |
777 | } |
778 | ||
b2fa47e6 | 779 | static inline void pud_clear(pud_t *pud) |
1da177e4 | 780 | { |
f4815ac6 | 781 | #ifdef CONFIG_64BIT |
6252d702 MS |
782 | if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3) |
783 | pud_val(*pud) = _REGION3_ENTRY_EMPTY; | |
b2fa47e6 | 784 | #endif |
1da177e4 LT |
785 | } |
786 | ||
b2fa47e6 | 787 | static inline void pmd_clear(pmd_t *pmdp) |
1da177e4 | 788 | { |
3610cce8 | 789 | pmd_val(*pmdp) = _SEGMENT_ENTRY_EMPTY; |
1da177e4 LT |
790 | } |
791 | ||
4448aaf0 | 792 | static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep) |
1da177e4 | 793 | { |
9282ed92 | 794 | pte_val(*ptep) = _PAGE_TYPE_EMPTY; |
1da177e4 LT |
795 | } |
796 | ||
797 | /* | |
798 | * The following pte modification functions only work if | |
799 | * pte_present() is true. Undefined behaviour if not.. | |
800 | */ | |
4448aaf0 | 801 | static inline pte_t pte_modify(pte_t pte, pgprot_t newprot) |
1da177e4 | 802 | { |
138c9021 | 803 | pte_val(pte) &= _PAGE_CHG_MASK; |
1da177e4 LT |
804 | pte_val(pte) |= pgprot_val(newprot); |
805 | return pte; | |
806 | } | |
807 | ||
4448aaf0 | 808 | static inline pte_t pte_wrprotect(pte_t pte) |
1da177e4 | 809 | { |
9282ed92 | 810 | /* Do not clobber _PAGE_TYPE_NONE pages! */ |
1da177e4 LT |
811 | if (!(pte_val(pte) & _PAGE_INVALID)) |
812 | pte_val(pte) |= _PAGE_RO; | |
813 | return pte; | |
814 | } | |
815 | ||
4448aaf0 | 816 | static inline pte_t pte_mkwrite(pte_t pte) |
1da177e4 LT |
817 | { |
818 | pte_val(pte) &= ~_PAGE_RO; | |
819 | return pte; | |
820 | } | |
821 | ||
4448aaf0 | 822 | static inline pte_t pte_mkclean(pte_t pte) |
1da177e4 | 823 | { |
b2fa47e6 MS |
824 | #ifdef CONFIG_PGSTE |
825 | pte_val(pte) &= ~_PAGE_SWC; | |
826 | #endif | |
1da177e4 LT |
827 | return pte; |
828 | } | |
829 | ||
4448aaf0 | 830 | static inline pte_t pte_mkdirty(pte_t pte) |
1da177e4 | 831 | { |
1da177e4 LT |
832 | return pte; |
833 | } | |
834 | ||
4448aaf0 | 835 | static inline pte_t pte_mkold(pte_t pte) |
1da177e4 | 836 | { |
b2fa47e6 MS |
837 | #ifdef CONFIG_PGSTE |
838 | pte_val(pte) &= ~_PAGE_SWR; | |
839 | #endif | |
1da177e4 LT |
840 | return pte; |
841 | } | |
842 | ||
4448aaf0 | 843 | static inline pte_t pte_mkyoung(pte_t pte) |
1da177e4 | 844 | { |
1da177e4 LT |
845 | return pte; |
846 | } | |
847 | ||
7e675137 NP |
848 | static inline pte_t pte_mkspecial(pte_t pte) |
849 | { | |
a08cb629 | 850 | pte_val(pte) |= _PAGE_SPECIAL; |
7e675137 NP |
851 | return pte; |
852 | } | |
853 | ||
84afdcee HC |
854 | #ifdef CONFIG_HUGETLB_PAGE |
855 | static inline pte_t pte_mkhuge(pte_t pte) | |
856 | { | |
857 | /* | |
858 | * PROT_NONE needs to be remapped from the pte type to the ste type. | |
859 | * The HW invalid bit is also different for pte and ste. The pte | |
860 | * invalid bit happens to be the same as the ste _SEGMENT_ENTRY_LARGE | |
861 | * bit, so we don't have to clear it. | |
862 | */ | |
863 | if (pte_val(pte) & _PAGE_INVALID) { | |
864 | if (pte_val(pte) & _PAGE_SWT) | |
865 | pte_val(pte) |= _HPAGE_TYPE_NONE; | |
866 | pte_val(pte) |= _SEGMENT_ENTRY_INV; | |
867 | } | |
868 | /* | |
869 | * Clear SW pte bits SWT and SWX, there are no SW bits in a segment | |
870 | * table entry. | |
871 | */ | |
872 | pte_val(pte) &= ~(_PAGE_SWT | _PAGE_SWX); | |
873 | /* | |
874 | * Also set the change-override bit because we don't need dirty bit | |
875 | * tracking for hugetlbfs pages. | |
876 | */ | |
877 | pte_val(pte) |= (_SEGMENT_ENTRY_LARGE | _SEGMENT_ENTRY_CO); | |
878 | return pte; | |
879 | } | |
880 | #endif | |
881 | ||
15e86b0c | 882 | /* |
b2fa47e6 | 883 | * Get (and clear) the user dirty bit for a pte. |
15e86b0c | 884 | */ |
b2fa47e6 MS |
885 | static inline int ptep_test_and_clear_user_dirty(struct mm_struct *mm, |
886 | pte_t *ptep) | |
15e86b0c | 887 | { |
b2fa47e6 MS |
888 | pgste_t pgste; |
889 | int dirty = 0; | |
890 | ||
891 | if (mm_has_pgste(mm)) { | |
892 | pgste = pgste_get_lock(ptep); | |
893 | pgste = pgste_update_all(ptep, pgste); | |
894 | dirty = !!(pgste_val(pgste) & KVM_UC_BIT); | |
895 | pgste_val(pgste) &= ~KVM_UC_BIT; | |
896 | pgste_set_unlock(ptep, pgste); | |
897 | return dirty; | |
15e86b0c | 898 | } |
15e86b0c FF |
899 | return dirty; |
900 | } | |
b2fa47e6 MS |
901 | |
902 | /* | |
903 | * Get (and clear) the user referenced bit for a pte. | |
904 | */ | |
905 | static inline int ptep_test_and_clear_user_young(struct mm_struct *mm, | |
906 | pte_t *ptep) | |
907 | { | |
908 | pgste_t pgste; | |
909 | int young = 0; | |
910 | ||
911 | if (mm_has_pgste(mm)) { | |
912 | pgste = pgste_get_lock(ptep); | |
913 | pgste = pgste_update_young(ptep, pgste); | |
914 | young = !!(pgste_val(pgste) & KVM_UR_BIT); | |
915 | pgste_val(pgste) &= ~KVM_UR_BIT; | |
916 | pgste_set_unlock(ptep, pgste); | |
917 | } | |
918 | return young; | |
919 | } | |
15e86b0c | 920 | |
ba8a9229 MS |
921 | #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG |
922 | static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, | |
923 | unsigned long addr, pte_t *ptep) | |
1da177e4 | 924 | { |
b2fa47e6 MS |
925 | pgste_t pgste; |
926 | pte_t pte; | |
927 | ||
928 | if (mm_has_pgste(vma->vm_mm)) { | |
929 | pgste = pgste_get_lock(ptep); | |
930 | pgste = pgste_update_young(ptep, pgste); | |
931 | pte = *ptep; | |
932 | *ptep = pte_mkold(pte); | |
933 | pgste_set_unlock(ptep, pgste); | |
934 | return pte_young(pte); | |
935 | } | |
1da177e4 LT |
936 | return 0; |
937 | } | |
938 | ||
ba8a9229 MS |
939 | #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH |
940 | static inline int ptep_clear_flush_young(struct vm_area_struct *vma, | |
941 | unsigned long address, pte_t *ptep) | |
1da177e4 | 942 | { |
5b7baf05 CB |
943 | /* No need to flush TLB |
944 | * On s390 reference bits are in storage key and never in TLB | |
945 | * With virtualization we handle the reference bit, without we | |
946 | * we can simply return */ | |
5b7baf05 | 947 | return ptep_test_and_clear_young(vma, address, ptep); |
1da177e4 LT |
948 | } |
949 | ||
9282ed92 | 950 | static inline void __ptep_ipte(unsigned long address, pte_t *ptep) |
1da177e4 | 951 | { |
9282ed92 | 952 | if (!(pte_val(*ptep) & _PAGE_INVALID)) { |
f4815ac6 | 953 | #ifndef CONFIG_64BIT |
146e4b3c | 954 | /* pto must point to the start of the segment table */ |
1da177e4 | 955 | pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00); |
9282ed92 GS |
956 | #else |
957 | /* ipte in zarch mode can do the math */ | |
958 | pte_t *pto = ptep; | |
959 | #endif | |
94c12cc7 MS |
960 | asm volatile( |
961 | " ipte %2,%3" | |
962 | : "=m" (*ptep) : "m" (*ptep), | |
963 | "a" (pto), "a" (address)); | |
1da177e4 | 964 | } |
9282ed92 GS |
965 | } |
966 | ||
ba8a9229 MS |
967 | /* |
968 | * This is hard to understand. ptep_get_and_clear and ptep_clear_flush | |
969 | * both clear the TLB for the unmapped pte. The reason is that | |
970 | * ptep_get_and_clear is used in common code (e.g. change_pte_range) | |
971 | * to modify an active pte. The sequence is | |
972 | * 1) ptep_get_and_clear | |
973 | * 2) set_pte_at | |
974 | * 3) flush_tlb_range | |
975 | * On s390 the tlb needs to get flushed with the modification of the pte | |
976 | * if the pte is active. The only way how this can be implemented is to | |
977 | * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range | |
978 | * is a nop. | |
979 | */ | |
980 | #define __HAVE_ARCH_PTEP_GET_AND_CLEAR | |
b2fa47e6 MS |
981 | static inline pte_t ptep_get_and_clear(struct mm_struct *mm, |
982 | unsigned long address, pte_t *ptep) | |
983 | { | |
984 | pgste_t pgste; | |
985 | pte_t pte; | |
986 | ||
987 | mm->context.flush_mm = 1; | |
988 | if (mm_has_pgste(mm)) | |
989 | pgste = pgste_get_lock(ptep); | |
990 | ||
991 | pte = *ptep; | |
992 | if (!mm_exclusive(mm)) | |
993 | __ptep_ipte(address, ptep); | |
994 | pte_val(*ptep) = _PAGE_TYPE_EMPTY; | |
995 | ||
996 | if (mm_has_pgste(mm)) { | |
997 | pgste = pgste_update_all(&pte, pgste); | |
998 | pgste_set_unlock(ptep, pgste); | |
999 | } | |
1000 | return pte; | |
1001 | } | |
1002 | ||
1003 | #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION | |
1004 | static inline pte_t ptep_modify_prot_start(struct mm_struct *mm, | |
1005 | unsigned long address, | |
1006 | pte_t *ptep) | |
1007 | { | |
1008 | pte_t pte; | |
1009 | ||
1010 | mm->context.flush_mm = 1; | |
1011 | if (mm_has_pgste(mm)) | |
1012 | pgste_get_lock(ptep); | |
1013 | ||
1014 | pte = *ptep; | |
1015 | if (!mm_exclusive(mm)) | |
1016 | __ptep_ipte(address, ptep); | |
1017 | return pte; | |
1018 | } | |
1019 | ||
1020 | static inline void ptep_modify_prot_commit(struct mm_struct *mm, | |
1021 | unsigned long address, | |
1022 | pte_t *ptep, pte_t pte) | |
1023 | { | |
1024 | *ptep = pte; | |
1025 | if (mm_has_pgste(mm)) | |
1026 | pgste_set_unlock(ptep, *(pgste_t *)(ptep + PTRS_PER_PTE)); | |
1027 | } | |
ba8a9229 MS |
1028 | |
1029 | #define __HAVE_ARCH_PTEP_CLEAR_FLUSH | |
f0e47c22 MS |
1030 | static inline pte_t ptep_clear_flush(struct vm_area_struct *vma, |
1031 | unsigned long address, pte_t *ptep) | |
1032 | { | |
b2fa47e6 MS |
1033 | pgste_t pgste; |
1034 | pte_t pte; | |
1035 | ||
1036 | if (mm_has_pgste(vma->vm_mm)) | |
1037 | pgste = pgste_get_lock(ptep); | |
1038 | ||
1039 | pte = *ptep; | |
1040 | __ptep_ipte(address, ptep); | |
1041 | pte_val(*ptep) = _PAGE_TYPE_EMPTY; | |
1042 | ||
1043 | if (mm_has_pgste(vma->vm_mm)) { | |
1044 | pgste = pgste_update_all(&pte, pgste); | |
1045 | pgste_set_unlock(ptep, pgste); | |
1046 | } | |
1da177e4 LT |
1047 | return pte; |
1048 | } | |
1049 | ||
ba8a9229 MS |
1050 | /* |
1051 | * The batched pte unmap code uses ptep_get_and_clear_full to clear the | |
1052 | * ptes. Here an optimization is possible. tlb_gather_mmu flushes all | |
1053 | * tlbs of an mm if it can guarantee that the ptes of the mm_struct | |
1054 | * cannot be accessed while the batched unmap is running. In this case | |
1055 | * full==1 and a simple pte_clear is enough. See tlb.h. | |
1056 | */ | |
1057 | #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL | |
1058 | static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm, | |
b2fa47e6 | 1059 | unsigned long address, |
ba8a9229 | 1060 | pte_t *ptep, int full) |
1da177e4 | 1061 | { |
b2fa47e6 MS |
1062 | pgste_t pgste; |
1063 | pte_t pte; | |
1064 | ||
1065 | if (mm_has_pgste(mm)) | |
1066 | pgste = pgste_get_lock(ptep); | |
ba8a9229 | 1067 | |
b2fa47e6 MS |
1068 | pte = *ptep; |
1069 | if (!full) | |
1070 | __ptep_ipte(address, ptep); | |
1071 | pte_val(*ptep) = _PAGE_TYPE_EMPTY; | |
1072 | ||
1073 | if (mm_has_pgste(mm)) { | |
1074 | pgste = pgste_update_all(&pte, pgste); | |
1075 | pgste_set_unlock(ptep, pgste); | |
1076 | } | |
ba8a9229 | 1077 | return pte; |
1da177e4 LT |
1078 | } |
1079 | ||
ba8a9229 | 1080 | #define __HAVE_ARCH_PTEP_SET_WRPROTECT |
b2fa47e6 MS |
1081 | static inline pte_t ptep_set_wrprotect(struct mm_struct *mm, |
1082 | unsigned long address, pte_t *ptep) | |
1083 | { | |
1084 | pgste_t pgste; | |
1085 | pte_t pte = *ptep; | |
1086 | ||
1087 | if (pte_write(pte)) { | |
1088 | mm->context.flush_mm = 1; | |
1089 | if (mm_has_pgste(mm)) | |
1090 | pgste = pgste_get_lock(ptep); | |
1091 | ||
1092 | if (!mm_exclusive(mm)) | |
1093 | __ptep_ipte(address, ptep); | |
1094 | *ptep = pte_wrprotect(pte); | |
1095 | ||
1096 | if (mm_has_pgste(mm)) | |
1097 | pgste_set_unlock(ptep, pgste); | |
1098 | } | |
1099 | return pte; | |
1100 | } | |
ba8a9229 MS |
1101 | |
1102 | #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS | |
b2fa47e6 MS |
1103 | static inline int ptep_set_access_flags(struct vm_area_struct *vma, |
1104 | unsigned long address, pte_t *ptep, | |
1105 | pte_t entry, int dirty) | |
1106 | { | |
1107 | pgste_t pgste; | |
1108 | ||
1109 | if (pte_same(*ptep, entry)) | |
1110 | return 0; | |
1111 | if (mm_has_pgste(vma->vm_mm)) | |
1112 | pgste = pgste_get_lock(ptep); | |
1113 | ||
1114 | __ptep_ipte(address, ptep); | |
1115 | *ptep = entry; | |
1116 | ||
1117 | if (mm_has_pgste(vma->vm_mm)) | |
1118 | pgste_set_unlock(ptep, pgste); | |
1119 | return 1; | |
1120 | } | |
1da177e4 | 1121 | |
1da177e4 LT |
1122 | /* |
1123 | * Conversion functions: convert a page and protection to a page entry, | |
1124 | * and a page entry and page directory to the page they refer to. | |
1125 | */ | |
1126 | static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot) | |
1127 | { | |
1128 | pte_t __pte; | |
1129 | pte_val(__pte) = physpage + pgprot_val(pgprot); | |
1130 | return __pte; | |
1131 | } | |
1132 | ||
2dcea57a HC |
1133 | static inline pte_t mk_pte(struct page *page, pgprot_t pgprot) |
1134 | { | |
0b2b6e1d | 1135 | unsigned long physpage = page_to_phys(page); |
1da177e4 | 1136 | |
2dcea57a HC |
1137 | return mk_pte_phys(physpage, pgprot); |
1138 | } | |
1139 | ||
190a1d72 MS |
1140 | #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1)) |
1141 | #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1)) | |
1142 | #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1)) | |
1143 | #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1)) | |
1da177e4 | 1144 | |
190a1d72 MS |
1145 | #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address)) |
1146 | #define pgd_offset_k(address) pgd_offset(&init_mm, address) | |
1da177e4 | 1147 | |
f4815ac6 | 1148 | #ifndef CONFIG_64BIT |
1da177e4 | 1149 | |
190a1d72 MS |
1150 | #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN) |
1151 | #define pud_deref(pmd) ({ BUG(); 0UL; }) | |
1152 | #define pgd_deref(pmd) ({ BUG(); 0UL; }) | |
46a82b2d | 1153 | |
190a1d72 MS |
1154 | #define pud_offset(pgd, address) ((pud_t *) pgd) |
1155 | #define pmd_offset(pud, address) ((pmd_t *) pud + pmd_index(address)) | |
1da177e4 | 1156 | |
f4815ac6 | 1157 | #else /* CONFIG_64BIT */ |
1da177e4 | 1158 | |
190a1d72 MS |
1159 | #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN) |
1160 | #define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN) | |
5a216a20 | 1161 | #define pgd_deref(pgd) (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) |
1da177e4 | 1162 | |
5a216a20 MS |
1163 | static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address) |
1164 | { | |
6252d702 MS |
1165 | pud_t *pud = (pud_t *) pgd; |
1166 | if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2) | |
1167 | pud = (pud_t *) pgd_deref(*pgd); | |
5a216a20 MS |
1168 | return pud + pud_index(address); |
1169 | } | |
1da177e4 | 1170 | |
190a1d72 | 1171 | static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address) |
1da177e4 | 1172 | { |
6252d702 MS |
1173 | pmd_t *pmd = (pmd_t *) pud; |
1174 | if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3) | |
1175 | pmd = (pmd_t *) pud_deref(*pud); | |
190a1d72 | 1176 | return pmd + pmd_index(address); |
1da177e4 LT |
1177 | } |
1178 | ||
f4815ac6 | 1179 | #endif /* CONFIG_64BIT */ |
1da177e4 | 1180 | |
190a1d72 MS |
1181 | #define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot)) |
1182 | #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT) | |
1183 | #define pte_page(x) pfn_to_page(pte_pfn(x)) | |
1da177e4 | 1184 | |
190a1d72 | 1185 | #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT) |
1da177e4 | 1186 | |
190a1d72 MS |
1187 | /* Find an entry in the lowest level page table.. */ |
1188 | #define pte_offset(pmd, addr) ((pte_t *) pmd_deref(*(pmd)) + pte_index(addr)) | |
1189 | #define pte_offset_kernel(pmd, address) pte_offset(pmd,address) | |
1da177e4 | 1190 | #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address) |
1da177e4 | 1191 | #define pte_unmap(pte) do { } while (0) |
1da177e4 | 1192 | |
1ae1c1d0 GS |
1193 | static inline void __pmd_idte(unsigned long address, pmd_t *pmdp) |
1194 | { | |
1195 | unsigned long sto = (unsigned long) pmdp - | |
1196 | pmd_index(address) * sizeof(pmd_t); | |
1197 | ||
1198 | if (!(pmd_val(*pmdp) & _SEGMENT_ENTRY_INV)) { | |
1199 | asm volatile( | |
1200 | " .insn rrf,0xb98e0000,%2,%3,0,0" | |
1201 | : "=m" (*pmdp) | |
1202 | : "m" (*pmdp), "a" (sto), | |
1203 | "a" ((address & HPAGE_MASK)) | |
1204 | : "cc" | |
1205 | ); | |
1206 | } | |
1207 | } | |
1208 | ||
75077afb | 1209 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
9501d09f GS |
1210 | #define __HAVE_ARCH_PGTABLE_DEPOSIT |
1211 | extern void pgtable_trans_huge_deposit(struct mm_struct *mm, pgtable_t pgtable); | |
1212 | ||
1213 | #define __HAVE_ARCH_PGTABLE_WITHDRAW | |
1214 | extern pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm); | |
1215 | ||
75077afb GS |
1216 | static inline int pmd_trans_splitting(pmd_t pmd) |
1217 | { | |
1218 | return pmd_val(pmd) & _SEGMENT_ENTRY_SPLIT; | |
1219 | } | |
1ae1c1d0 GS |
1220 | |
1221 | static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr, | |
1222 | pmd_t *pmdp, pmd_t entry) | |
1223 | { | |
1224 | *pmdp = entry; | |
1225 | } | |
1226 | ||
1227 | static inline unsigned long massage_pgprot_pmd(pgprot_t pgprot) | |
1228 | { | |
1229 | unsigned long pgprot_pmd = 0; | |
1230 | ||
1231 | if (pgprot_val(pgprot) & _PAGE_INVALID) { | |
1232 | if (pgprot_val(pgprot) & _PAGE_SWT) | |
1233 | pgprot_pmd |= _HPAGE_TYPE_NONE; | |
1234 | pgprot_pmd |= _SEGMENT_ENTRY_INV; | |
1235 | } | |
1236 | if (pgprot_val(pgprot) & _PAGE_RO) | |
1237 | pgprot_pmd |= _SEGMENT_ENTRY_RO; | |
1238 | return pgprot_pmd; | |
1239 | } | |
1240 | ||
1241 | static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot) | |
1242 | { | |
1243 | pmd_val(pmd) &= _SEGMENT_CHG_MASK; | |
1244 | pmd_val(pmd) |= massage_pgprot_pmd(newprot); | |
1245 | return pmd; | |
1246 | } | |
1247 | ||
1248 | static inline pmd_t pmd_mkhuge(pmd_t pmd) | |
1249 | { | |
1250 | pmd_val(pmd) |= _SEGMENT_ENTRY_LARGE; | |
1251 | return pmd; | |
1252 | } | |
1253 | ||
1254 | static inline pmd_t pmd_mkwrite(pmd_t pmd) | |
1255 | { | |
1256 | pmd_val(pmd) &= ~_SEGMENT_ENTRY_RO; | |
1257 | return pmd; | |
1258 | } | |
1259 | ||
1260 | static inline pmd_t pmd_wrprotect(pmd_t pmd) | |
1261 | { | |
1262 | pmd_val(pmd) |= _SEGMENT_ENTRY_RO; | |
1263 | return pmd; | |
1264 | } | |
1265 | ||
1266 | static inline pmd_t pmd_mkdirty(pmd_t pmd) | |
1267 | { | |
1268 | /* No dirty bit in the segment table entry. */ | |
1269 | return pmd; | |
1270 | } | |
1271 | ||
1272 | static inline pmd_t pmd_mkold(pmd_t pmd) | |
1273 | { | |
1274 | /* No referenced bit in the segment table entry. */ | |
1275 | return pmd; | |
1276 | } | |
1277 | ||
1278 | static inline pmd_t pmd_mkyoung(pmd_t pmd) | |
1279 | { | |
1280 | /* No referenced bit in the segment table entry. */ | |
1281 | return pmd; | |
1282 | } | |
1283 | ||
1284 | #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG | |
1285 | static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma, | |
1286 | unsigned long address, pmd_t *pmdp) | |
1287 | { | |
1288 | unsigned long pmd_addr = pmd_val(*pmdp) & HPAGE_MASK; | |
1289 | long tmp, rc; | |
1290 | int counter; | |
1291 | ||
1292 | rc = 0; | |
1293 | if (MACHINE_HAS_RRBM) { | |
1294 | counter = PTRS_PER_PTE >> 6; | |
1295 | asm volatile( | |
1296 | "0: .insn rre,0xb9ae0000,%0,%3\n" /* rrbm */ | |
1297 | " ogr %1,%0\n" | |
1298 | " la %3,0(%4,%3)\n" | |
1299 | " brct %2,0b\n" | |
1300 | : "=&d" (tmp), "+&d" (rc), "+d" (counter), | |
1301 | "+a" (pmd_addr) | |
1302 | : "a" (64 * 4096UL) : "cc"); | |
1303 | rc = !!rc; | |
1304 | } else { | |
1305 | counter = PTRS_PER_PTE; | |
1306 | asm volatile( | |
1307 | "0: rrbe 0,%2\n" | |
1308 | " la %2,0(%3,%2)\n" | |
1309 | " brc 12,1f\n" | |
1310 | " lhi %0,1\n" | |
1311 | "1: brct %1,0b\n" | |
1312 | : "+d" (rc), "+d" (counter), "+a" (pmd_addr) | |
1313 | : "a" (4096UL) : "cc"); | |
1314 | } | |
1315 | return rc; | |
1316 | } | |
1317 | ||
1318 | #define __HAVE_ARCH_PMDP_GET_AND_CLEAR | |
1319 | static inline pmd_t pmdp_get_and_clear(struct mm_struct *mm, | |
1320 | unsigned long address, pmd_t *pmdp) | |
1321 | { | |
1322 | pmd_t pmd = *pmdp; | |
1323 | ||
1324 | __pmd_idte(address, pmdp); | |
1325 | pmd_clear(pmdp); | |
1326 | return pmd; | |
1327 | } | |
1328 | ||
1329 | #define __HAVE_ARCH_PMDP_CLEAR_FLUSH | |
1330 | static inline pmd_t pmdp_clear_flush(struct vm_area_struct *vma, | |
1331 | unsigned long address, pmd_t *pmdp) | |
1332 | { | |
1333 | return pmdp_get_and_clear(vma->vm_mm, address, pmdp); | |
1334 | } | |
1335 | ||
1336 | #define __HAVE_ARCH_PMDP_INVALIDATE | |
1337 | static inline void pmdp_invalidate(struct vm_area_struct *vma, | |
1338 | unsigned long address, pmd_t *pmdp) | |
1339 | { | |
1340 | __pmd_idte(address, pmdp); | |
1341 | } | |
1342 | ||
1343 | static inline pmd_t mk_pmd_phys(unsigned long physpage, pgprot_t pgprot) | |
1344 | { | |
1345 | pmd_t __pmd; | |
1346 | pmd_val(__pmd) = physpage + massage_pgprot_pmd(pgprot); | |
1347 | return __pmd; | |
1348 | } | |
1349 | ||
1350 | #define pfn_pmd(pfn, pgprot) mk_pmd_phys(__pa((pfn) << PAGE_SHIFT), (pgprot)) | |
1351 | #define mk_pmd(page, pgprot) pfn_pmd(page_to_pfn(page), (pgprot)) | |
1352 | ||
1353 | static inline int pmd_trans_huge(pmd_t pmd) | |
1354 | { | |
1355 | return pmd_val(pmd) & _SEGMENT_ENTRY_LARGE; | |
1356 | } | |
1357 | ||
1358 | static inline int has_transparent_hugepage(void) | |
1359 | { | |
1360 | return MACHINE_HAS_HPAGE ? 1 : 0; | |
1361 | } | |
1362 | ||
1363 | static inline unsigned long pmd_pfn(pmd_t pmd) | |
1364 | { | |
1365 | if (pmd_trans_huge(pmd)) | |
1366 | return pmd_val(pmd) >> HPAGE_SHIFT; | |
1367 | else | |
1368 | return pmd_val(pmd) >> PAGE_SHIFT; | |
1369 | } | |
75077afb GS |
1370 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE */ |
1371 | ||
1da177e4 LT |
1372 | /* |
1373 | * 31 bit swap entry format: | |
1374 | * A page-table entry has some bits we have to treat in a special way. | |
1375 | * Bits 0, 20 and bit 23 have to be zero, otherwise an specification | |
1376 | * exception will occur instead of a page translation exception. The | |
1377 | * specifiation exception has the bad habit not to store necessary | |
1378 | * information in the lowcore. | |
1379 | * Bit 21 and bit 22 are the page invalid bit and the page protection | |
1380 | * bit. We set both to indicate a swapped page. | |
1381 | * Bit 30 and 31 are used to distinguish the different page types. For | |
1382 | * a swapped page these bits need to be zero. | |
1383 | * This leaves the bits 1-19 and bits 24-29 to store type and offset. | |
1384 | * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19 | |
1385 | * plus 24 for the offset. | |
1386 | * 0| offset |0110|o|type |00| | |
1387 | * 0 0000000001111111111 2222 2 22222 33 | |
1388 | * 0 1234567890123456789 0123 4 56789 01 | |
1389 | * | |
1390 | * 64 bit swap entry format: | |
1391 | * A page-table entry has some bits we have to treat in a special way. | |
1392 | * Bits 52 and bit 55 have to be zero, otherwise an specification | |
1393 | * exception will occur instead of a page translation exception. The | |
1394 | * specifiation exception has the bad habit not to store necessary | |
1395 | * information in the lowcore. | |
1396 | * Bit 53 and bit 54 are the page invalid bit and the page protection | |
1397 | * bit. We set both to indicate a swapped page. | |
1398 | * Bit 62 and 63 are used to distinguish the different page types. For | |
1399 | * a swapped page these bits need to be zero. | |
1400 | * This leaves the bits 0-51 and bits 56-61 to store type and offset. | |
1401 | * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51 | |
1402 | * plus 56 for the offset. | |
1403 | * | offset |0110|o|type |00| | |
1404 | * 0000000000111111111122222222223333333333444444444455 5555 5 55566 66 | |
1405 | * 0123456789012345678901234567890123456789012345678901 2345 6 78901 23 | |
1406 | */ | |
f4815ac6 | 1407 | #ifndef CONFIG_64BIT |
1da177e4 LT |
1408 | #define __SWP_OFFSET_MASK (~0UL >> 12) |
1409 | #else | |
1410 | #define __SWP_OFFSET_MASK (~0UL >> 11) | |
1411 | #endif | |
4448aaf0 | 1412 | static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset) |
1da177e4 LT |
1413 | { |
1414 | pte_t pte; | |
1415 | offset &= __SWP_OFFSET_MASK; | |
9282ed92 | 1416 | pte_val(pte) = _PAGE_TYPE_SWAP | ((type & 0x1f) << 2) | |
1da177e4 LT |
1417 | ((offset & 1UL) << 7) | ((offset & ~1UL) << 11); |
1418 | return pte; | |
1419 | } | |
1420 | ||
1421 | #define __swp_type(entry) (((entry).val >> 2) & 0x1f) | |
1422 | #define __swp_offset(entry) (((entry).val >> 11) | (((entry).val >> 7) & 1)) | |
1423 | #define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) }) | |
1424 | ||
1425 | #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) }) | |
1426 | #define __swp_entry_to_pte(x) ((pte_t) { (x).val }) | |
1427 | ||
f4815ac6 | 1428 | #ifndef CONFIG_64BIT |
1da177e4 | 1429 | # define PTE_FILE_MAX_BITS 26 |
f4815ac6 | 1430 | #else /* CONFIG_64BIT */ |
1da177e4 | 1431 | # define PTE_FILE_MAX_BITS 59 |
f4815ac6 | 1432 | #endif /* CONFIG_64BIT */ |
1da177e4 LT |
1433 | |
1434 | #define pte_to_pgoff(__pte) \ | |
1435 | ((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f)) | |
1436 | ||
1437 | #define pgoff_to_pte(__off) \ | |
1438 | ((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \ | |
9282ed92 | 1439 | | _PAGE_TYPE_FILE }) |
1da177e4 LT |
1440 | |
1441 | #endif /* !__ASSEMBLY__ */ | |
1442 | ||
1443 | #define kern_addr_valid(addr) (1) | |
1444 | ||
17f34580 HC |
1445 | extern int vmem_add_mapping(unsigned long start, unsigned long size); |
1446 | extern int vmem_remove_mapping(unsigned long start, unsigned long size); | |
402b0862 | 1447 | extern int s390_enable_sie(void); |
f4eb07c1 | 1448 | |
1da177e4 LT |
1449 | /* |
1450 | * No page table caches to initialise | |
1451 | */ | |
1452 | #define pgtable_cache_init() do { } while (0) | |
1453 | ||
1da177e4 LT |
1454 | #include <asm-generic/pgtable.h> |
1455 | ||
1456 | #endif /* _S390_PAGE_H */ |