Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * include/asm-s390/pgtable.h | |
3 | * | |
4 | * S390 version | |
5 | * Copyright (C) 1999,2000 IBM Deutschland Entwicklung GmbH, IBM Corporation | |
6 | * Author(s): Hartmut Penner (hp@de.ibm.com) | |
7 | * Ulrich Weigand (weigand@de.ibm.com) | |
8 | * Martin Schwidefsky (schwidefsky@de.ibm.com) | |
9 | * | |
10 | * Derived from "include/asm-i386/pgtable.h" | |
11 | */ | |
12 | ||
13 | #ifndef _ASM_S390_PGTABLE_H | |
14 | #define _ASM_S390_PGTABLE_H | |
15 | ||
1da177e4 LT |
16 | /* |
17 | * The Linux memory management assumes a three-level page table setup. For | |
18 | * s390 31 bit we "fold" the mid level into the top-level page table, so | |
19 | * that we physically have the same two-level page table as the s390 mmu | |
20 | * expects in 31 bit mode. For s390 64 bit we use three of the five levels | |
21 | * the hardware provides (region first and region second tables are not | |
22 | * used). | |
23 | * | |
24 | * The "pgd_xxx()" functions are trivial for a folded two-level | |
25 | * setup: the pgd is never bad, and a pmd always exists (as it's folded | |
26 | * into the pgd entry) | |
27 | * | |
28 | * This file contains the functions and defines necessary to modify and use | |
29 | * the S390 page table tree. | |
30 | */ | |
31 | #ifndef __ASSEMBLY__ | |
9789db08 | 32 | #include <linux/sched.h> |
2dcea57a | 33 | #include <linux/mm_types.h> |
1da177e4 | 34 | #include <asm/bug.h> |
b2fa47e6 | 35 | #include <asm/page.h> |
1da177e4 | 36 | |
1da177e4 LT |
37 | extern pgd_t swapper_pg_dir[] __attribute__ ((aligned (4096))); |
38 | extern void paging_init(void); | |
2b67fc46 | 39 | extern void vmem_map_init(void); |
92f842ea | 40 | extern void fault_init(void); |
1da177e4 LT |
41 | |
42 | /* | |
43 | * The S390 doesn't have any external MMU info: the kernel page | |
44 | * tables contain all the necessary information. | |
45 | */ | |
4b3073e1 | 46 | #define update_mmu_cache(vma, address, ptep) do { } while (0) |
1da177e4 LT |
47 | |
48 | /* | |
238ec4ef | 49 | * ZERO_PAGE is a global shared page that is always zero; used |
1da177e4 LT |
50 | * for zero-mapped memory areas etc.. |
51 | */ | |
238ec4ef MS |
52 | |
53 | extern unsigned long empty_zero_page; | |
54 | extern unsigned long zero_page_mask; | |
55 | ||
56 | #define ZERO_PAGE(vaddr) \ | |
57 | (virt_to_page((void *)(empty_zero_page + \ | |
58 | (((unsigned long)(vaddr)) &zero_page_mask)))) | |
59 | ||
60 | #define is_zero_pfn is_zero_pfn | |
61 | static inline int is_zero_pfn(unsigned long pfn) | |
62 | { | |
63 | extern unsigned long zero_pfn; | |
64 | unsigned long offset_from_zero_pfn = pfn - zero_pfn; | |
65 | return offset_from_zero_pfn <= (zero_page_mask >> PAGE_SHIFT); | |
66 | } | |
67 | ||
68 | #define my_zero_pfn(addr) page_to_pfn(ZERO_PAGE(addr)) | |
69 | ||
1da177e4 LT |
70 | #endif /* !__ASSEMBLY__ */ |
71 | ||
72 | /* | |
73 | * PMD_SHIFT determines the size of the area a second-level page | |
74 | * table can map | |
75 | * PGDIR_SHIFT determines what a third-level page table entry can map | |
76 | */ | |
77 | #ifndef __s390x__ | |
146e4b3c MS |
78 | # define PMD_SHIFT 20 |
79 | # define PUD_SHIFT 20 | |
80 | # define PGDIR_SHIFT 20 | |
1da177e4 | 81 | #else /* __s390x__ */ |
146e4b3c | 82 | # define PMD_SHIFT 20 |
190a1d72 | 83 | # define PUD_SHIFT 31 |
5a216a20 | 84 | # define PGDIR_SHIFT 42 |
1da177e4 LT |
85 | #endif /* __s390x__ */ |
86 | ||
87 | #define PMD_SIZE (1UL << PMD_SHIFT) | |
88 | #define PMD_MASK (~(PMD_SIZE-1)) | |
190a1d72 MS |
89 | #define PUD_SIZE (1UL << PUD_SHIFT) |
90 | #define PUD_MASK (~(PUD_SIZE-1)) | |
5a216a20 MS |
91 | #define PGDIR_SIZE (1UL << PGDIR_SHIFT) |
92 | #define PGDIR_MASK (~(PGDIR_SIZE-1)) | |
1da177e4 LT |
93 | |
94 | /* | |
95 | * entries per page directory level: the S390 is two-level, so | |
96 | * we don't really have any PMD directory physically. | |
97 | * for S390 segment-table entries are combined to one PGD | |
98 | * that leads to 1024 pte per pgd | |
99 | */ | |
146e4b3c | 100 | #define PTRS_PER_PTE 256 |
1da177e4 | 101 | #ifndef __s390x__ |
146e4b3c | 102 | #define PTRS_PER_PMD 1 |
5a216a20 | 103 | #define PTRS_PER_PUD 1 |
1da177e4 | 104 | #else /* __s390x__ */ |
146e4b3c | 105 | #define PTRS_PER_PMD 2048 |
5a216a20 | 106 | #define PTRS_PER_PUD 2048 |
1da177e4 | 107 | #endif /* __s390x__ */ |
146e4b3c | 108 | #define PTRS_PER_PGD 2048 |
1da177e4 | 109 | |
d455a369 HD |
110 | #define FIRST_USER_ADDRESS 0 |
111 | ||
1da177e4 LT |
112 | #define pte_ERROR(e) \ |
113 | printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e)) | |
114 | #define pmd_ERROR(e) \ | |
115 | printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e)) | |
190a1d72 MS |
116 | #define pud_ERROR(e) \ |
117 | printk("%s:%d: bad pud %p.\n", __FILE__, __LINE__, (void *) pud_val(e)) | |
1da177e4 LT |
118 | #define pgd_ERROR(e) \ |
119 | printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e)) | |
120 | ||
121 | #ifndef __ASSEMBLY__ | |
122 | /* | |
5fd9c6e2 | 123 | * The vmalloc area will always be on the topmost area of the kernel |
7d3f661e | 124 | * mapping. We reserve 96MB (31bit) / 128GB (64bit) for vmalloc, |
5fd9c6e2 CB |
125 | * which should be enough for any sane case. |
126 | * By putting vmalloc at the top, we maximise the gap between physical | |
127 | * memory and vmalloc to catch misplaced memory accesses. As a side | |
128 | * effect, this also makes sure that 64 bit module code cannot be used | |
129 | * as system call address. | |
8b62bc96 | 130 | */ |
239a6425 HC |
131 | |
132 | extern unsigned long VMALLOC_START; | |
133 | ||
1da177e4 | 134 | #ifndef __s390x__ |
239a6425 | 135 | #define VMALLOC_SIZE (96UL << 20) |
5fd9c6e2 | 136 | #define VMALLOC_END 0x7e000000UL |
0189103c | 137 | #define VMEM_MAP_END 0x80000000UL |
1da177e4 | 138 | #else /* __s390x__ */ |
7d3f661e MS |
139 | #define VMALLOC_SIZE (128UL << 30) |
140 | #define VMALLOC_END 0x3e000000000UL | |
0189103c | 141 | #define VMEM_MAP_END 0x40000000000UL |
1da177e4 LT |
142 | #endif /* __s390x__ */ |
143 | ||
0189103c HC |
144 | /* |
145 | * VMEM_MAX_PHYS is the highest physical address that can be added to the 1:1 | |
146 | * mapping. This needs to be calculated at compile time since the size of the | |
147 | * VMEM_MAP is static but the size of struct page can change. | |
148 | */ | |
522d8dc0 MS |
149 | #define VMEM_MAX_PAGES ((VMEM_MAP_END - VMALLOC_END) / sizeof(struct page)) |
150 | #define VMEM_MAX_PFN min(VMALLOC_START >> PAGE_SHIFT, VMEM_MAX_PAGES) | |
151 | #define VMEM_MAX_PHYS ((VMEM_MAX_PFN << PAGE_SHIFT) & ~((16 << 20) - 1)) | |
17f34580 | 152 | #define vmemmap ((struct page *) VMALLOC_END) |
5fd9c6e2 | 153 | |
1da177e4 LT |
154 | /* |
155 | * A 31 bit pagetable entry of S390 has following format: | |
156 | * | PFRA | | OS | | |
157 | * 0 0IP0 | |
158 | * 00000000001111111111222222222233 | |
159 | * 01234567890123456789012345678901 | |
160 | * | |
161 | * I Page-Invalid Bit: Page is not available for address-translation | |
162 | * P Page-Protection Bit: Store access not possible for page | |
163 | * | |
164 | * A 31 bit segmenttable entry of S390 has following format: | |
165 | * | P-table origin | |PTL | |
166 | * 0 IC | |
167 | * 00000000001111111111222222222233 | |
168 | * 01234567890123456789012345678901 | |
169 | * | |
170 | * I Segment-Invalid Bit: Segment is not available for address-translation | |
171 | * C Common-Segment Bit: Segment is not private (PoP 3-30) | |
172 | * PTL Page-Table-Length: Page-table length (PTL+1*16 entries -> up to 256) | |
173 | * | |
174 | * The 31 bit segmenttable origin of S390 has following format: | |
175 | * | |
176 | * |S-table origin | | STL | | |
177 | * X **GPS | |
178 | * 00000000001111111111222222222233 | |
179 | * 01234567890123456789012345678901 | |
180 | * | |
181 | * X Space-Switch event: | |
182 | * G Segment-Invalid Bit: * | |
183 | * P Private-Space Bit: Segment is not private (PoP 3-30) | |
184 | * S Storage-Alteration: | |
185 | * STL Segment-Table-Length: Segment-table length (STL+1*16 entries -> up to 2048) | |
186 | * | |
187 | * A 64 bit pagetable entry of S390 has following format: | |
6a985c61 | 188 | * | PFRA |0IPC| OS | |
1da177e4 LT |
189 | * 0000000000111111111122222222223333333333444444444455555555556666 |
190 | * 0123456789012345678901234567890123456789012345678901234567890123 | |
191 | * | |
192 | * I Page-Invalid Bit: Page is not available for address-translation | |
193 | * P Page-Protection Bit: Store access not possible for page | |
6a985c61 | 194 | * C Change-bit override: HW is not required to set change bit |
1da177e4 LT |
195 | * |
196 | * A 64 bit segmenttable entry of S390 has following format: | |
197 | * | P-table origin | TT | |
198 | * 0000000000111111111122222222223333333333444444444455555555556666 | |
199 | * 0123456789012345678901234567890123456789012345678901234567890123 | |
200 | * | |
201 | * I Segment-Invalid Bit: Segment is not available for address-translation | |
202 | * C Common-Segment Bit: Segment is not private (PoP 3-30) | |
203 | * P Page-Protection Bit: Store access not possible for page | |
204 | * TT Type 00 | |
205 | * | |
206 | * A 64 bit region table entry of S390 has following format: | |
207 | * | S-table origin | TF TTTL | |
208 | * 0000000000111111111122222222223333333333444444444455555555556666 | |
209 | * 0123456789012345678901234567890123456789012345678901234567890123 | |
210 | * | |
211 | * I Segment-Invalid Bit: Segment is not available for address-translation | |
212 | * TT Type 01 | |
213 | * TF | |
190a1d72 | 214 | * TL Table length |
1da177e4 LT |
215 | * |
216 | * The 64 bit regiontable origin of S390 has following format: | |
217 | * | region table origon | DTTL | |
218 | * 0000000000111111111122222222223333333333444444444455555555556666 | |
219 | * 0123456789012345678901234567890123456789012345678901234567890123 | |
220 | * | |
221 | * X Space-Switch event: | |
222 | * G Segment-Invalid Bit: | |
223 | * P Private-Space Bit: | |
224 | * S Storage-Alteration: | |
225 | * R Real space | |
226 | * TL Table-Length: | |
227 | * | |
228 | * A storage key has the following format: | |
229 | * | ACC |F|R|C|0| | |
230 | * 0 3 4 5 6 7 | |
231 | * ACC: access key | |
232 | * F : fetch protection bit | |
233 | * R : referenced bit | |
234 | * C : changed bit | |
235 | */ | |
236 | ||
237 | /* Hardware bits in the page table entry */ | |
6a985c61 | 238 | #define _PAGE_CO 0x100 /* HW Change-bit override */ |
83377484 MS |
239 | #define _PAGE_RO 0x200 /* HW read-only bit */ |
240 | #define _PAGE_INVALID 0x400 /* HW invalid bit */ | |
3610cce8 MS |
241 | |
242 | /* Software bits in the page table entry */ | |
83377484 MS |
243 | #define _PAGE_SWT 0x001 /* SW pte type bit t */ |
244 | #define _PAGE_SWX 0x002 /* SW pte type bit x */ | |
b2fa47e6 MS |
245 | #define _PAGE_SWC 0x004 /* SW pte changed bit (for KVM) */ |
246 | #define _PAGE_SWR 0x008 /* SW pte referenced bit (for KVM) */ | |
247 | #define _PAGE_SPECIAL 0x010 /* SW associated with special page */ | |
a08cb629 | 248 | #define __HAVE_ARCH_PTE_SPECIAL |
1da177e4 | 249 | |
138c9021 | 250 | /* Set of bits not changed in pte_modify */ |
b2fa47e6 | 251 | #define _PAGE_CHG_MASK (PAGE_MASK | _PAGE_SPECIAL | _PAGE_SWC | _PAGE_SWR) |
138c9021 | 252 | |
83377484 | 253 | /* Six different types of pages. */ |
9282ed92 GS |
254 | #define _PAGE_TYPE_EMPTY 0x400 |
255 | #define _PAGE_TYPE_NONE 0x401 | |
83377484 MS |
256 | #define _PAGE_TYPE_SWAP 0x403 |
257 | #define _PAGE_TYPE_FILE 0x601 /* bit 0x002 is used for offset !! */ | |
9282ed92 GS |
258 | #define _PAGE_TYPE_RO 0x200 |
259 | #define _PAGE_TYPE_RW 0x000 | |
1da177e4 | 260 | |
53492b1d GS |
261 | /* |
262 | * Only four types for huge pages, using the invalid bit and protection bit | |
263 | * of a segment table entry. | |
264 | */ | |
265 | #define _HPAGE_TYPE_EMPTY 0x020 /* _SEGMENT_ENTRY_INV */ | |
266 | #define _HPAGE_TYPE_NONE 0x220 | |
267 | #define _HPAGE_TYPE_RO 0x200 /* _SEGMENT_ENTRY_RO */ | |
268 | #define _HPAGE_TYPE_RW 0x000 | |
269 | ||
83377484 MS |
270 | /* |
271 | * PTE type bits are rather complicated. handle_pte_fault uses pte_present, | |
272 | * pte_none and pte_file to find out the pte type WITHOUT holding the page | |
273 | * table lock. ptep_clear_flush on the other hand uses ptep_clear_flush to | |
274 | * invalidate a given pte. ipte sets the hw invalid bit and clears all tlbs | |
275 | * for the page. The page table entry is set to _PAGE_TYPE_EMPTY afterwards. | |
276 | * This change is done while holding the lock, but the intermediate step | |
277 | * of a previously valid pte with the hw invalid bit set can be observed by | |
278 | * handle_pte_fault. That makes it necessary that all valid pte types with | |
279 | * the hw invalid bit set must be distinguishable from the four pte types | |
280 | * empty, none, swap and file. | |
281 | * | |
282 | * irxt ipte irxt | |
283 | * _PAGE_TYPE_EMPTY 1000 -> 1000 | |
284 | * _PAGE_TYPE_NONE 1001 -> 1001 | |
285 | * _PAGE_TYPE_SWAP 1011 -> 1011 | |
286 | * _PAGE_TYPE_FILE 11?1 -> 11?1 | |
287 | * _PAGE_TYPE_RO 0100 -> 1100 | |
288 | * _PAGE_TYPE_RW 0000 -> 1000 | |
289 | * | |
c1821c2e | 290 | * pte_none is true for bits combinations 1000, 1010, 1100, 1110 |
83377484 MS |
291 | * pte_present is true for bits combinations 0000, 0010, 0100, 0110, 1001 |
292 | * pte_file is true for bits combinations 1101, 1111 | |
c1821c2e | 293 | * swap pte is 1011 and 0001, 0011, 0101, 0111 are invalid. |
83377484 MS |
294 | */ |
295 | ||
1da177e4 LT |
296 | #ifndef __s390x__ |
297 | ||
3610cce8 MS |
298 | /* Bits in the segment table address-space-control-element */ |
299 | #define _ASCE_SPACE_SWITCH 0x80000000UL /* space switch event */ | |
300 | #define _ASCE_ORIGIN_MASK 0x7ffff000UL /* segment table origin */ | |
301 | #define _ASCE_PRIVATE_SPACE 0x100 /* private space control */ | |
302 | #define _ASCE_ALT_EVENT 0x80 /* storage alteration event control */ | |
303 | #define _ASCE_TABLE_LENGTH 0x7f /* 128 x 64 entries = 8k */ | |
1da177e4 | 304 | |
3610cce8 MS |
305 | /* Bits in the segment table entry */ |
306 | #define _SEGMENT_ENTRY_ORIGIN 0x7fffffc0UL /* page table origin */ | |
80217147 | 307 | #define _SEGMENT_ENTRY_RO 0x200 /* page protection bit */ |
3610cce8 MS |
308 | #define _SEGMENT_ENTRY_INV 0x20 /* invalid segment table entry */ |
309 | #define _SEGMENT_ENTRY_COMMON 0x10 /* common segment bit */ | |
310 | #define _SEGMENT_ENTRY_PTL 0x0f /* page table length */ | |
1da177e4 | 311 | |
3610cce8 MS |
312 | #define _SEGMENT_ENTRY (_SEGMENT_ENTRY_PTL) |
313 | #define _SEGMENT_ENTRY_EMPTY (_SEGMENT_ENTRY_INV) | |
1da177e4 | 314 | |
6c61cfe9 MS |
315 | /* Page status table bits for virtualization */ |
316 | #define RCP_ACC_BITS 0xf0000000UL | |
317 | #define RCP_FP_BIT 0x08000000UL | |
318 | #define RCP_PCL_BIT 0x00800000UL | |
319 | #define RCP_HR_BIT 0x00400000UL | |
320 | #define RCP_HC_BIT 0x00200000UL | |
321 | #define RCP_GR_BIT 0x00040000UL | |
322 | #define RCP_GC_BIT 0x00020000UL | |
323 | ||
324 | /* User dirty / referenced bit for KVM's migration feature */ | |
325 | #define KVM_UR_BIT 0x00008000UL | |
326 | #define KVM_UC_BIT 0x00004000UL | |
327 | ||
1da177e4 LT |
328 | #else /* __s390x__ */ |
329 | ||
3610cce8 MS |
330 | /* Bits in the segment/region table address-space-control-element */ |
331 | #define _ASCE_ORIGIN ~0xfffUL/* segment table origin */ | |
332 | #define _ASCE_PRIVATE_SPACE 0x100 /* private space control */ | |
333 | #define _ASCE_ALT_EVENT 0x80 /* storage alteration event control */ | |
334 | #define _ASCE_SPACE_SWITCH 0x40 /* space switch event */ | |
335 | #define _ASCE_REAL_SPACE 0x20 /* real space control */ | |
336 | #define _ASCE_TYPE_MASK 0x0c /* asce table type mask */ | |
337 | #define _ASCE_TYPE_REGION1 0x0c /* region first table type */ | |
338 | #define _ASCE_TYPE_REGION2 0x08 /* region second table type */ | |
339 | #define _ASCE_TYPE_REGION3 0x04 /* region third table type */ | |
340 | #define _ASCE_TYPE_SEGMENT 0x00 /* segment table type */ | |
341 | #define _ASCE_TABLE_LENGTH 0x03 /* region table length */ | |
342 | ||
343 | /* Bits in the region table entry */ | |
344 | #define _REGION_ENTRY_ORIGIN ~0xfffUL/* region/segment table origin */ | |
345 | #define _REGION_ENTRY_INV 0x20 /* invalid region table entry */ | |
346 | #define _REGION_ENTRY_TYPE_MASK 0x0c /* region/segment table type mask */ | |
347 | #define _REGION_ENTRY_TYPE_R1 0x0c /* region first table type */ | |
348 | #define _REGION_ENTRY_TYPE_R2 0x08 /* region second table type */ | |
349 | #define _REGION_ENTRY_TYPE_R3 0x04 /* region third table type */ | |
350 | #define _REGION_ENTRY_LENGTH 0x03 /* region third length */ | |
351 | ||
352 | #define _REGION1_ENTRY (_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH) | |
353 | #define _REGION1_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INV) | |
354 | #define _REGION2_ENTRY (_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH) | |
355 | #define _REGION2_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INV) | |
356 | #define _REGION3_ENTRY (_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH) | |
357 | #define _REGION3_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INV) | |
358 | ||
1da177e4 | 359 | /* Bits in the segment table entry */ |
3610cce8 MS |
360 | #define _SEGMENT_ENTRY_ORIGIN ~0x7ffUL/* segment table origin */ |
361 | #define _SEGMENT_ENTRY_RO 0x200 /* page protection bit */ | |
362 | #define _SEGMENT_ENTRY_INV 0x20 /* invalid segment table entry */ | |
1da177e4 | 363 | |
3610cce8 MS |
364 | #define _SEGMENT_ENTRY (0) |
365 | #define _SEGMENT_ENTRY_EMPTY (_SEGMENT_ENTRY_INV) | |
366 | ||
53492b1d GS |
367 | #define _SEGMENT_ENTRY_LARGE 0x400 /* STE-format control, large page */ |
368 | #define _SEGMENT_ENTRY_CO 0x100 /* change-recording override */ | |
369 | ||
6c61cfe9 MS |
370 | /* Page status table bits for virtualization */ |
371 | #define RCP_ACC_BITS 0xf000000000000000UL | |
372 | #define RCP_FP_BIT 0x0800000000000000UL | |
373 | #define RCP_PCL_BIT 0x0080000000000000UL | |
374 | #define RCP_HR_BIT 0x0040000000000000UL | |
375 | #define RCP_HC_BIT 0x0020000000000000UL | |
376 | #define RCP_GR_BIT 0x0004000000000000UL | |
377 | #define RCP_GC_BIT 0x0002000000000000UL | |
378 | ||
379 | /* User dirty / referenced bit for KVM's migration feature */ | |
380 | #define KVM_UR_BIT 0x0000800000000000UL | |
381 | #define KVM_UC_BIT 0x0000400000000000UL | |
382 | ||
3610cce8 | 383 | #endif /* __s390x__ */ |
1da177e4 LT |
384 | |
385 | /* | |
3610cce8 MS |
386 | * A user page table pointer has the space-switch-event bit, the |
387 | * private-space-control bit and the storage-alteration-event-control | |
388 | * bit set. A kernel page table pointer doesn't need them. | |
1da177e4 | 389 | */ |
3610cce8 MS |
390 | #define _ASCE_USER_BITS (_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \ |
391 | _ASCE_ALT_EVENT) | |
1da177e4 | 392 | |
1da177e4 | 393 | /* |
9282ed92 | 394 | * Page protection definitions. |
1da177e4 | 395 | */ |
9282ed92 GS |
396 | #define PAGE_NONE __pgprot(_PAGE_TYPE_NONE) |
397 | #define PAGE_RO __pgprot(_PAGE_TYPE_RO) | |
398 | #define PAGE_RW __pgprot(_PAGE_TYPE_RW) | |
399 | ||
400 | #define PAGE_KERNEL PAGE_RW | |
401 | #define PAGE_COPY PAGE_RO | |
1da177e4 LT |
402 | |
403 | /* | |
043d0708 MS |
404 | * On s390 the page table entry has an invalid bit and a read-only bit. |
405 | * Read permission implies execute permission and write permission | |
406 | * implies read permission. | |
1da177e4 LT |
407 | */ |
408 | /*xwr*/ | |
9282ed92 GS |
409 | #define __P000 PAGE_NONE |
410 | #define __P001 PAGE_RO | |
411 | #define __P010 PAGE_RO | |
412 | #define __P011 PAGE_RO | |
043d0708 MS |
413 | #define __P100 PAGE_RO |
414 | #define __P101 PAGE_RO | |
415 | #define __P110 PAGE_RO | |
416 | #define __P111 PAGE_RO | |
9282ed92 GS |
417 | |
418 | #define __S000 PAGE_NONE | |
419 | #define __S001 PAGE_RO | |
420 | #define __S010 PAGE_RW | |
421 | #define __S011 PAGE_RW | |
043d0708 MS |
422 | #define __S100 PAGE_RO |
423 | #define __S101 PAGE_RO | |
424 | #define __S110 PAGE_RW | |
425 | #define __S111 PAGE_RW | |
1da177e4 | 426 | |
b2fa47e6 | 427 | static inline int mm_exclusive(struct mm_struct *mm) |
1da177e4 | 428 | { |
b2fa47e6 MS |
429 | return likely(mm == current->active_mm && |
430 | atomic_read(&mm->context.attach_count) <= 1); | |
1da177e4 | 431 | } |
1da177e4 | 432 | |
b2fa47e6 MS |
433 | static inline int mm_has_pgste(struct mm_struct *mm) |
434 | { | |
435 | #ifdef CONFIG_PGSTE | |
436 | if (unlikely(mm->context.has_pgste)) | |
437 | return 1; | |
438 | #endif | |
439 | return 0; | |
440 | } | |
1da177e4 LT |
441 | /* |
442 | * pgd/pmd/pte query functions | |
443 | */ | |
444 | #ifndef __s390x__ | |
445 | ||
4448aaf0 AB |
446 | static inline int pgd_present(pgd_t pgd) { return 1; } |
447 | static inline int pgd_none(pgd_t pgd) { return 0; } | |
448 | static inline int pgd_bad(pgd_t pgd) { return 0; } | |
1da177e4 | 449 | |
190a1d72 MS |
450 | static inline int pud_present(pud_t pud) { return 1; } |
451 | static inline int pud_none(pud_t pud) { return 0; } | |
452 | static inline int pud_bad(pud_t pud) { return 0; } | |
453 | ||
1da177e4 LT |
454 | #else /* __s390x__ */ |
455 | ||
5a216a20 MS |
456 | static inline int pgd_present(pgd_t pgd) |
457 | { | |
6252d702 MS |
458 | if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2) |
459 | return 1; | |
5a216a20 MS |
460 | return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL; |
461 | } | |
462 | ||
463 | static inline int pgd_none(pgd_t pgd) | |
464 | { | |
6252d702 MS |
465 | if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2) |
466 | return 0; | |
5a216a20 MS |
467 | return (pgd_val(pgd) & _REGION_ENTRY_INV) != 0UL; |
468 | } | |
469 | ||
470 | static inline int pgd_bad(pgd_t pgd) | |
471 | { | |
6252d702 MS |
472 | /* |
473 | * With dynamic page table levels the pgd can be a region table | |
474 | * entry or a segment table entry. Check for the bit that are | |
475 | * invalid for either table entry. | |
476 | */ | |
5a216a20 | 477 | unsigned long mask = |
6252d702 | 478 | ~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INV & |
5a216a20 MS |
479 | ~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH; |
480 | return (pgd_val(pgd) & mask) != 0; | |
481 | } | |
190a1d72 MS |
482 | |
483 | static inline int pud_present(pud_t pud) | |
1da177e4 | 484 | { |
6252d702 MS |
485 | if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3) |
486 | return 1; | |
0d017923 | 487 | return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL; |
1da177e4 LT |
488 | } |
489 | ||
190a1d72 | 490 | static inline int pud_none(pud_t pud) |
1da177e4 | 491 | { |
6252d702 MS |
492 | if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3) |
493 | return 0; | |
0d017923 | 494 | return (pud_val(pud) & _REGION_ENTRY_INV) != 0UL; |
1da177e4 LT |
495 | } |
496 | ||
190a1d72 | 497 | static inline int pud_bad(pud_t pud) |
1da177e4 | 498 | { |
6252d702 MS |
499 | /* |
500 | * With dynamic page table levels the pud can be a region table | |
501 | * entry or a segment table entry. Check for the bit that are | |
502 | * invalid for either table entry. | |
503 | */ | |
5a216a20 | 504 | unsigned long mask = |
6252d702 | 505 | ~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INV & |
5a216a20 MS |
506 | ~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH; |
507 | return (pud_val(pud) & mask) != 0; | |
1da177e4 LT |
508 | } |
509 | ||
3610cce8 MS |
510 | #endif /* __s390x__ */ |
511 | ||
4448aaf0 | 512 | static inline int pmd_present(pmd_t pmd) |
1da177e4 | 513 | { |
0d017923 | 514 | return (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN) != 0UL; |
1da177e4 LT |
515 | } |
516 | ||
4448aaf0 | 517 | static inline int pmd_none(pmd_t pmd) |
1da177e4 | 518 | { |
0d017923 | 519 | return (pmd_val(pmd) & _SEGMENT_ENTRY_INV) != 0UL; |
1da177e4 LT |
520 | } |
521 | ||
4448aaf0 | 522 | static inline int pmd_bad(pmd_t pmd) |
1da177e4 | 523 | { |
3610cce8 MS |
524 | unsigned long mask = ~_SEGMENT_ENTRY_ORIGIN & ~_SEGMENT_ENTRY_INV; |
525 | return (pmd_val(pmd) & mask) != _SEGMENT_ENTRY; | |
1da177e4 LT |
526 | } |
527 | ||
4448aaf0 | 528 | static inline int pte_none(pte_t pte) |
1da177e4 | 529 | { |
83377484 | 530 | return (pte_val(pte) & _PAGE_INVALID) && !(pte_val(pte) & _PAGE_SWT); |
1da177e4 LT |
531 | } |
532 | ||
4448aaf0 | 533 | static inline int pte_present(pte_t pte) |
1da177e4 | 534 | { |
83377484 MS |
535 | unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT | _PAGE_SWX; |
536 | return (pte_val(pte) & mask) == _PAGE_TYPE_NONE || | |
537 | (!(pte_val(pte) & _PAGE_INVALID) && | |
538 | !(pte_val(pte) & _PAGE_SWT)); | |
1da177e4 LT |
539 | } |
540 | ||
4448aaf0 | 541 | static inline int pte_file(pte_t pte) |
1da177e4 | 542 | { |
83377484 MS |
543 | unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT; |
544 | return (pte_val(pte) & mask) == _PAGE_TYPE_FILE; | |
1da177e4 LT |
545 | } |
546 | ||
7e675137 NP |
547 | static inline int pte_special(pte_t pte) |
548 | { | |
a08cb629 | 549 | return (pte_val(pte) & _PAGE_SPECIAL); |
7e675137 NP |
550 | } |
551 | ||
ba8a9229 | 552 | #define __HAVE_ARCH_PTE_SAME |
b2fa47e6 MS |
553 | static inline int pte_same(pte_t a, pte_t b) |
554 | { | |
555 | return pte_val(a) == pte_val(b); | |
556 | } | |
1da177e4 | 557 | |
b2fa47e6 | 558 | static inline pgste_t pgste_get_lock(pte_t *ptep) |
5b7baf05 | 559 | { |
b2fa47e6 | 560 | unsigned long new = 0; |
5b7baf05 | 561 | #ifdef CONFIG_PGSTE |
b2fa47e6 MS |
562 | unsigned long old; |
563 | ||
5b7baf05 | 564 | preempt_disable(); |
b2fa47e6 MS |
565 | asm( |
566 | " lg %0,%2\n" | |
567 | "0: lgr %1,%0\n" | |
568 | " nihh %0,0xff7f\n" /* clear RCP_PCL_BIT in old */ | |
569 | " oihh %1,0x0080\n" /* set RCP_PCL_BIT in new */ | |
570 | " csg %0,%1,%2\n" | |
571 | " jl 0b\n" | |
572 | : "=&d" (old), "=&d" (new), "=Q" (ptep[PTRS_PER_PTE]) | |
573 | : "Q" (ptep[PTRS_PER_PTE]) : "cc"); | |
5b7baf05 | 574 | #endif |
b2fa47e6 | 575 | return __pgste(new); |
5b7baf05 CB |
576 | } |
577 | ||
b2fa47e6 | 578 | static inline void pgste_set_unlock(pte_t *ptep, pgste_t pgste) |
5b7baf05 CB |
579 | { |
580 | #ifdef CONFIG_PGSTE | |
b2fa47e6 MS |
581 | asm( |
582 | " nihh %1,0xff7f\n" /* clear RCP_PCL_BIT */ | |
583 | " stg %1,%0\n" | |
584 | : "=Q" (ptep[PTRS_PER_PTE]) | |
585 | : "d" (pgste_val(pgste)), "Q" (ptep[PTRS_PER_PTE]) : "cc"); | |
5b7baf05 CB |
586 | preempt_enable(); |
587 | #endif | |
588 | } | |
589 | ||
b2fa47e6 | 590 | static inline pgste_t pgste_update_all(pte_t *ptep, pgste_t pgste) |
5b7baf05 CB |
591 | { |
592 | #ifdef CONFIG_PGSTE | |
a43a9d93 | 593 | unsigned long address, bits; |
b2fa47e6 MS |
594 | unsigned char skey; |
595 | ||
09b53883 MS |
596 | if (!pte_present(*ptep)) |
597 | return pgste; | |
a43a9d93 HC |
598 | address = pte_val(*ptep) & PAGE_MASK; |
599 | skey = page_get_storage_key(address); | |
b2fa47e6 MS |
600 | bits = skey & (_PAGE_CHANGED | _PAGE_REFERENCED); |
601 | /* Clear page changed & referenced bit in the storage key */ | |
602 | if (bits) { | |
603 | skey ^= bits; | |
a43a9d93 | 604 | page_set_storage_key(address, skey, 1); |
15e86b0c | 605 | } |
b2fa47e6 MS |
606 | /* Transfer page changed & referenced bit to guest bits in pgste */ |
607 | pgste_val(pgste) |= bits << 48; /* RCP_GR_BIT & RCP_GC_BIT */ | |
608 | /* Get host changed & referenced bits from pgste */ | |
609 | bits |= (pgste_val(pgste) & (RCP_HR_BIT | RCP_HC_BIT)) >> 52; | |
610 | /* Clear host bits in pgste. */ | |
611 | pgste_val(pgste) &= ~(RCP_HR_BIT | RCP_HC_BIT); | |
612 | pgste_val(pgste) &= ~(RCP_ACC_BITS | RCP_FP_BIT); | |
613 | /* Copy page access key and fetch protection bit to pgste */ | |
614 | pgste_val(pgste) |= | |
615 | (unsigned long) (skey & (_PAGE_ACC_BITS | _PAGE_FP_BIT)) << 56; | |
616 | /* Transfer changed and referenced to kvm user bits */ | |
617 | pgste_val(pgste) |= bits << 45; /* KVM_UR_BIT & KVM_UC_BIT */ | |
618 | /* Transfer changed & referenced to pte sofware bits */ | |
619 | pte_val(*ptep) |= bits << 1; /* _PAGE_SWR & _PAGE_SWC */ | |
620 | #endif | |
621 | return pgste; | |
622 | ||
623 | } | |
624 | ||
625 | static inline pgste_t pgste_update_young(pte_t *ptep, pgste_t pgste) | |
626 | { | |
627 | #ifdef CONFIG_PGSTE | |
628 | int young; | |
629 | ||
09b53883 MS |
630 | if (!pte_present(*ptep)) |
631 | return pgste; | |
b2fa47e6 MS |
632 | young = page_reset_referenced(pte_val(*ptep) & PAGE_MASK); |
633 | /* Transfer page referenced bit to pte software bit (host view) */ | |
634 | if (young || (pgste_val(pgste) & RCP_HR_BIT)) | |
635 | pte_val(*ptep) |= _PAGE_SWR; | |
636 | /* Clear host referenced bit in pgste. */ | |
637 | pgste_val(pgste) &= ~RCP_HR_BIT; | |
638 | /* Transfer page referenced bit to guest bit in pgste */ | |
639 | pgste_val(pgste) |= (unsigned long) young << 50; /* set RCP_GR_BIT */ | |
640 | #endif | |
641 | return pgste; | |
642 | ||
643 | } | |
644 | ||
09b53883 | 645 | static inline void pgste_set_pte(pte_t *ptep, pgste_t pgste, pte_t entry) |
b2fa47e6 MS |
646 | { |
647 | #ifdef CONFIG_PGSTE | |
a43a9d93 | 648 | unsigned long address; |
b2fa47e6 MS |
649 | unsigned long okey, nkey; |
650 | ||
09b53883 MS |
651 | if (!pte_present(entry)) |
652 | return; | |
653 | address = pte_val(entry) & PAGE_MASK; | |
a43a9d93 | 654 | okey = nkey = page_get_storage_key(address); |
b2fa47e6 MS |
655 | nkey &= ~(_PAGE_ACC_BITS | _PAGE_FP_BIT); |
656 | /* Set page access key and fetch protection bit from pgste */ | |
657 | nkey |= (pgste_val(pgste) & (RCP_ACC_BITS | RCP_FP_BIT)) >> 56; | |
658 | if (okey != nkey) | |
a43a9d93 | 659 | page_set_storage_key(address, nkey, 1); |
5b7baf05 CB |
660 | #endif |
661 | } | |
662 | ||
e5992f2e MS |
663 | /** |
664 | * struct gmap_struct - guest address space | |
665 | * @mm: pointer to the parent mm_struct | |
666 | * @table: pointer to the page directory | |
480e5926 | 667 | * @asce: address space control element for gmap page table |
e5992f2e MS |
668 | * @crst_list: list of all crst tables used in the guest address space |
669 | */ | |
670 | struct gmap { | |
671 | struct list_head list; | |
672 | struct mm_struct *mm; | |
673 | unsigned long *table; | |
480e5926 | 674 | unsigned long asce; |
e5992f2e MS |
675 | struct list_head crst_list; |
676 | }; | |
677 | ||
678 | /** | |
679 | * struct gmap_rmap - reverse mapping for segment table entries | |
680 | * @next: pointer to the next gmap_rmap structure in the list | |
681 | * @entry: pointer to a segment table entry | |
682 | */ | |
683 | struct gmap_rmap { | |
684 | struct list_head list; | |
685 | unsigned long *entry; | |
686 | }; | |
687 | ||
688 | /** | |
689 | * struct gmap_pgtable - gmap information attached to a page table | |
690 | * @vmaddr: address of the 1MB segment in the process virtual memory | |
691 | * @mapper: list of segment table entries maping a page table | |
692 | */ | |
693 | struct gmap_pgtable { | |
694 | unsigned long vmaddr; | |
695 | struct list_head mapper; | |
696 | }; | |
697 | ||
698 | struct gmap *gmap_alloc(struct mm_struct *mm); | |
699 | void gmap_free(struct gmap *gmap); | |
700 | void gmap_enable(struct gmap *gmap); | |
701 | void gmap_disable(struct gmap *gmap); | |
702 | int gmap_map_segment(struct gmap *gmap, unsigned long from, | |
703 | unsigned long to, unsigned long length); | |
704 | int gmap_unmap_segment(struct gmap *gmap, unsigned long to, unsigned long len); | |
499069e1 | 705 | unsigned long __gmap_fault(unsigned long address, struct gmap *); |
e5992f2e | 706 | unsigned long gmap_fault(unsigned long address, struct gmap *); |
388186bc | 707 | void gmap_discard(unsigned long from, unsigned long to, struct gmap *); |
e5992f2e | 708 | |
b2fa47e6 MS |
709 | /* |
710 | * Certain architectures need to do special things when PTEs | |
711 | * within a page table are directly modified. Thus, the following | |
712 | * hook is made available. | |
713 | */ | |
714 | static inline void set_pte_at(struct mm_struct *mm, unsigned long addr, | |
715 | pte_t *ptep, pte_t entry) | |
716 | { | |
717 | pgste_t pgste; | |
718 | ||
719 | if (mm_has_pgste(mm)) { | |
720 | pgste = pgste_get_lock(ptep); | |
09b53883 | 721 | pgste_set_pte(ptep, pgste, entry); |
b2fa47e6 MS |
722 | *ptep = entry; |
723 | pgste_set_unlock(ptep, pgste); | |
724 | } else | |
725 | *ptep = entry; | |
726 | } | |
727 | ||
1da177e4 LT |
728 | /* |
729 | * query functions pte_write/pte_dirty/pte_young only work if | |
730 | * pte_present() is true. Undefined behaviour if not.. | |
731 | */ | |
4448aaf0 | 732 | static inline int pte_write(pte_t pte) |
1da177e4 LT |
733 | { |
734 | return (pte_val(pte) & _PAGE_RO) == 0; | |
735 | } | |
736 | ||
4448aaf0 | 737 | static inline int pte_dirty(pte_t pte) |
1da177e4 | 738 | { |
b2fa47e6 MS |
739 | #ifdef CONFIG_PGSTE |
740 | if (pte_val(pte) & _PAGE_SWC) | |
741 | return 1; | |
742 | #endif | |
1da177e4 LT |
743 | return 0; |
744 | } | |
745 | ||
4448aaf0 | 746 | static inline int pte_young(pte_t pte) |
1da177e4 | 747 | { |
b2fa47e6 MS |
748 | #ifdef CONFIG_PGSTE |
749 | if (pte_val(pte) & _PAGE_SWR) | |
750 | return 1; | |
751 | #endif | |
1da177e4 LT |
752 | return 0; |
753 | } | |
754 | ||
1da177e4 LT |
755 | /* |
756 | * pgd/pmd/pte modification functions | |
757 | */ | |
758 | ||
b2fa47e6 | 759 | static inline void pgd_clear(pgd_t *pgd) |
5a216a20 | 760 | { |
b2fa47e6 | 761 | #ifdef __s390x__ |
6252d702 MS |
762 | if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2) |
763 | pgd_val(*pgd) = _REGION2_ENTRY_EMPTY; | |
b2fa47e6 | 764 | #endif |
5a216a20 MS |
765 | } |
766 | ||
b2fa47e6 | 767 | static inline void pud_clear(pud_t *pud) |
1da177e4 | 768 | { |
b2fa47e6 | 769 | #ifdef __s390x__ |
6252d702 MS |
770 | if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3) |
771 | pud_val(*pud) = _REGION3_ENTRY_EMPTY; | |
b2fa47e6 | 772 | #endif |
1da177e4 LT |
773 | } |
774 | ||
b2fa47e6 | 775 | static inline void pmd_clear(pmd_t *pmdp) |
1da177e4 | 776 | { |
3610cce8 | 777 | pmd_val(*pmdp) = _SEGMENT_ENTRY_EMPTY; |
1da177e4 LT |
778 | } |
779 | ||
4448aaf0 | 780 | static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep) |
1da177e4 | 781 | { |
9282ed92 | 782 | pte_val(*ptep) = _PAGE_TYPE_EMPTY; |
1da177e4 LT |
783 | } |
784 | ||
785 | /* | |
786 | * The following pte modification functions only work if | |
787 | * pte_present() is true. Undefined behaviour if not.. | |
788 | */ | |
4448aaf0 | 789 | static inline pte_t pte_modify(pte_t pte, pgprot_t newprot) |
1da177e4 | 790 | { |
138c9021 | 791 | pte_val(pte) &= _PAGE_CHG_MASK; |
1da177e4 LT |
792 | pte_val(pte) |= pgprot_val(newprot); |
793 | return pte; | |
794 | } | |
795 | ||
4448aaf0 | 796 | static inline pte_t pte_wrprotect(pte_t pte) |
1da177e4 | 797 | { |
9282ed92 | 798 | /* Do not clobber _PAGE_TYPE_NONE pages! */ |
1da177e4 LT |
799 | if (!(pte_val(pte) & _PAGE_INVALID)) |
800 | pte_val(pte) |= _PAGE_RO; | |
801 | return pte; | |
802 | } | |
803 | ||
4448aaf0 | 804 | static inline pte_t pte_mkwrite(pte_t pte) |
1da177e4 LT |
805 | { |
806 | pte_val(pte) &= ~_PAGE_RO; | |
807 | return pte; | |
808 | } | |
809 | ||
4448aaf0 | 810 | static inline pte_t pte_mkclean(pte_t pte) |
1da177e4 | 811 | { |
b2fa47e6 MS |
812 | #ifdef CONFIG_PGSTE |
813 | pte_val(pte) &= ~_PAGE_SWC; | |
814 | #endif | |
1da177e4 LT |
815 | return pte; |
816 | } | |
817 | ||
4448aaf0 | 818 | static inline pte_t pte_mkdirty(pte_t pte) |
1da177e4 | 819 | { |
1da177e4 LT |
820 | return pte; |
821 | } | |
822 | ||
4448aaf0 | 823 | static inline pte_t pte_mkold(pte_t pte) |
1da177e4 | 824 | { |
b2fa47e6 MS |
825 | #ifdef CONFIG_PGSTE |
826 | pte_val(pte) &= ~_PAGE_SWR; | |
827 | #endif | |
1da177e4 LT |
828 | return pte; |
829 | } | |
830 | ||
4448aaf0 | 831 | static inline pte_t pte_mkyoung(pte_t pte) |
1da177e4 | 832 | { |
1da177e4 LT |
833 | return pte; |
834 | } | |
835 | ||
7e675137 NP |
836 | static inline pte_t pte_mkspecial(pte_t pte) |
837 | { | |
a08cb629 | 838 | pte_val(pte) |= _PAGE_SPECIAL; |
7e675137 NP |
839 | return pte; |
840 | } | |
841 | ||
84afdcee HC |
842 | #ifdef CONFIG_HUGETLB_PAGE |
843 | static inline pte_t pte_mkhuge(pte_t pte) | |
844 | { | |
845 | /* | |
846 | * PROT_NONE needs to be remapped from the pte type to the ste type. | |
847 | * The HW invalid bit is also different for pte and ste. The pte | |
848 | * invalid bit happens to be the same as the ste _SEGMENT_ENTRY_LARGE | |
849 | * bit, so we don't have to clear it. | |
850 | */ | |
851 | if (pte_val(pte) & _PAGE_INVALID) { | |
852 | if (pte_val(pte) & _PAGE_SWT) | |
853 | pte_val(pte) |= _HPAGE_TYPE_NONE; | |
854 | pte_val(pte) |= _SEGMENT_ENTRY_INV; | |
855 | } | |
856 | /* | |
857 | * Clear SW pte bits SWT and SWX, there are no SW bits in a segment | |
858 | * table entry. | |
859 | */ | |
860 | pte_val(pte) &= ~(_PAGE_SWT | _PAGE_SWX); | |
861 | /* | |
862 | * Also set the change-override bit because we don't need dirty bit | |
863 | * tracking for hugetlbfs pages. | |
864 | */ | |
865 | pte_val(pte) |= (_SEGMENT_ENTRY_LARGE | _SEGMENT_ENTRY_CO); | |
866 | return pte; | |
867 | } | |
868 | #endif | |
869 | ||
15e86b0c | 870 | /* |
b2fa47e6 | 871 | * Get (and clear) the user dirty bit for a pte. |
15e86b0c | 872 | */ |
b2fa47e6 MS |
873 | static inline int ptep_test_and_clear_user_dirty(struct mm_struct *mm, |
874 | pte_t *ptep) | |
15e86b0c | 875 | { |
b2fa47e6 MS |
876 | pgste_t pgste; |
877 | int dirty = 0; | |
878 | ||
879 | if (mm_has_pgste(mm)) { | |
880 | pgste = pgste_get_lock(ptep); | |
881 | pgste = pgste_update_all(ptep, pgste); | |
882 | dirty = !!(pgste_val(pgste) & KVM_UC_BIT); | |
883 | pgste_val(pgste) &= ~KVM_UC_BIT; | |
884 | pgste_set_unlock(ptep, pgste); | |
885 | return dirty; | |
15e86b0c | 886 | } |
15e86b0c FF |
887 | return dirty; |
888 | } | |
b2fa47e6 MS |
889 | |
890 | /* | |
891 | * Get (and clear) the user referenced bit for a pte. | |
892 | */ | |
893 | static inline int ptep_test_and_clear_user_young(struct mm_struct *mm, | |
894 | pte_t *ptep) | |
895 | { | |
896 | pgste_t pgste; | |
897 | int young = 0; | |
898 | ||
899 | if (mm_has_pgste(mm)) { | |
900 | pgste = pgste_get_lock(ptep); | |
901 | pgste = pgste_update_young(ptep, pgste); | |
902 | young = !!(pgste_val(pgste) & KVM_UR_BIT); | |
903 | pgste_val(pgste) &= ~KVM_UR_BIT; | |
904 | pgste_set_unlock(ptep, pgste); | |
905 | } | |
906 | return young; | |
907 | } | |
15e86b0c | 908 | |
ba8a9229 MS |
909 | #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG |
910 | static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, | |
911 | unsigned long addr, pte_t *ptep) | |
1da177e4 | 912 | { |
b2fa47e6 MS |
913 | pgste_t pgste; |
914 | pte_t pte; | |
915 | ||
916 | if (mm_has_pgste(vma->vm_mm)) { | |
917 | pgste = pgste_get_lock(ptep); | |
918 | pgste = pgste_update_young(ptep, pgste); | |
919 | pte = *ptep; | |
920 | *ptep = pte_mkold(pte); | |
921 | pgste_set_unlock(ptep, pgste); | |
922 | return pte_young(pte); | |
923 | } | |
1da177e4 LT |
924 | return 0; |
925 | } | |
926 | ||
ba8a9229 MS |
927 | #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH |
928 | static inline int ptep_clear_flush_young(struct vm_area_struct *vma, | |
929 | unsigned long address, pte_t *ptep) | |
1da177e4 | 930 | { |
5b7baf05 CB |
931 | /* No need to flush TLB |
932 | * On s390 reference bits are in storage key and never in TLB | |
933 | * With virtualization we handle the reference bit, without we | |
934 | * we can simply return */ | |
5b7baf05 | 935 | return ptep_test_and_clear_young(vma, address, ptep); |
1da177e4 LT |
936 | } |
937 | ||
9282ed92 | 938 | static inline void __ptep_ipte(unsigned long address, pte_t *ptep) |
1da177e4 | 939 | { |
9282ed92 | 940 | if (!(pte_val(*ptep) & _PAGE_INVALID)) { |
1da177e4 | 941 | #ifndef __s390x__ |
146e4b3c | 942 | /* pto must point to the start of the segment table */ |
1da177e4 | 943 | pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00); |
9282ed92 GS |
944 | #else |
945 | /* ipte in zarch mode can do the math */ | |
946 | pte_t *pto = ptep; | |
947 | #endif | |
94c12cc7 MS |
948 | asm volatile( |
949 | " ipte %2,%3" | |
950 | : "=m" (*ptep) : "m" (*ptep), | |
951 | "a" (pto), "a" (address)); | |
1da177e4 | 952 | } |
9282ed92 GS |
953 | } |
954 | ||
ba8a9229 MS |
955 | /* |
956 | * This is hard to understand. ptep_get_and_clear and ptep_clear_flush | |
957 | * both clear the TLB for the unmapped pte. The reason is that | |
958 | * ptep_get_and_clear is used in common code (e.g. change_pte_range) | |
959 | * to modify an active pte. The sequence is | |
960 | * 1) ptep_get_and_clear | |
961 | * 2) set_pte_at | |
962 | * 3) flush_tlb_range | |
963 | * On s390 the tlb needs to get flushed with the modification of the pte | |
964 | * if the pte is active. The only way how this can be implemented is to | |
965 | * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range | |
966 | * is a nop. | |
967 | */ | |
968 | #define __HAVE_ARCH_PTEP_GET_AND_CLEAR | |
b2fa47e6 MS |
969 | static inline pte_t ptep_get_and_clear(struct mm_struct *mm, |
970 | unsigned long address, pte_t *ptep) | |
971 | { | |
972 | pgste_t pgste; | |
973 | pte_t pte; | |
974 | ||
975 | mm->context.flush_mm = 1; | |
976 | if (mm_has_pgste(mm)) | |
977 | pgste = pgste_get_lock(ptep); | |
978 | ||
979 | pte = *ptep; | |
980 | if (!mm_exclusive(mm)) | |
981 | __ptep_ipte(address, ptep); | |
982 | pte_val(*ptep) = _PAGE_TYPE_EMPTY; | |
983 | ||
984 | if (mm_has_pgste(mm)) { | |
985 | pgste = pgste_update_all(&pte, pgste); | |
986 | pgste_set_unlock(ptep, pgste); | |
987 | } | |
988 | return pte; | |
989 | } | |
990 | ||
991 | #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION | |
992 | static inline pte_t ptep_modify_prot_start(struct mm_struct *mm, | |
993 | unsigned long address, | |
994 | pte_t *ptep) | |
995 | { | |
996 | pte_t pte; | |
997 | ||
998 | mm->context.flush_mm = 1; | |
999 | if (mm_has_pgste(mm)) | |
1000 | pgste_get_lock(ptep); | |
1001 | ||
1002 | pte = *ptep; | |
1003 | if (!mm_exclusive(mm)) | |
1004 | __ptep_ipte(address, ptep); | |
1005 | return pte; | |
1006 | } | |
1007 | ||
1008 | static inline void ptep_modify_prot_commit(struct mm_struct *mm, | |
1009 | unsigned long address, | |
1010 | pte_t *ptep, pte_t pte) | |
1011 | { | |
1012 | *ptep = pte; | |
1013 | if (mm_has_pgste(mm)) | |
1014 | pgste_set_unlock(ptep, *(pgste_t *)(ptep + PTRS_PER_PTE)); | |
1015 | } | |
ba8a9229 MS |
1016 | |
1017 | #define __HAVE_ARCH_PTEP_CLEAR_FLUSH | |
f0e47c22 MS |
1018 | static inline pte_t ptep_clear_flush(struct vm_area_struct *vma, |
1019 | unsigned long address, pte_t *ptep) | |
1020 | { | |
b2fa47e6 MS |
1021 | pgste_t pgste; |
1022 | pte_t pte; | |
1023 | ||
1024 | if (mm_has_pgste(vma->vm_mm)) | |
1025 | pgste = pgste_get_lock(ptep); | |
1026 | ||
1027 | pte = *ptep; | |
1028 | __ptep_ipte(address, ptep); | |
1029 | pte_val(*ptep) = _PAGE_TYPE_EMPTY; | |
1030 | ||
1031 | if (mm_has_pgste(vma->vm_mm)) { | |
1032 | pgste = pgste_update_all(&pte, pgste); | |
1033 | pgste_set_unlock(ptep, pgste); | |
1034 | } | |
1da177e4 LT |
1035 | return pte; |
1036 | } | |
1037 | ||
ba8a9229 MS |
1038 | /* |
1039 | * The batched pte unmap code uses ptep_get_and_clear_full to clear the | |
1040 | * ptes. Here an optimization is possible. tlb_gather_mmu flushes all | |
1041 | * tlbs of an mm if it can guarantee that the ptes of the mm_struct | |
1042 | * cannot be accessed while the batched unmap is running. In this case | |
1043 | * full==1 and a simple pte_clear is enough. See tlb.h. | |
1044 | */ | |
1045 | #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL | |
1046 | static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm, | |
b2fa47e6 | 1047 | unsigned long address, |
ba8a9229 | 1048 | pte_t *ptep, int full) |
1da177e4 | 1049 | { |
b2fa47e6 MS |
1050 | pgste_t pgste; |
1051 | pte_t pte; | |
1052 | ||
1053 | if (mm_has_pgste(mm)) | |
1054 | pgste = pgste_get_lock(ptep); | |
ba8a9229 | 1055 | |
b2fa47e6 MS |
1056 | pte = *ptep; |
1057 | if (!full) | |
1058 | __ptep_ipte(address, ptep); | |
1059 | pte_val(*ptep) = _PAGE_TYPE_EMPTY; | |
1060 | ||
1061 | if (mm_has_pgste(mm)) { | |
1062 | pgste = pgste_update_all(&pte, pgste); | |
1063 | pgste_set_unlock(ptep, pgste); | |
1064 | } | |
ba8a9229 | 1065 | return pte; |
1da177e4 LT |
1066 | } |
1067 | ||
ba8a9229 | 1068 | #define __HAVE_ARCH_PTEP_SET_WRPROTECT |
b2fa47e6 MS |
1069 | static inline pte_t ptep_set_wrprotect(struct mm_struct *mm, |
1070 | unsigned long address, pte_t *ptep) | |
1071 | { | |
1072 | pgste_t pgste; | |
1073 | pte_t pte = *ptep; | |
1074 | ||
1075 | if (pte_write(pte)) { | |
1076 | mm->context.flush_mm = 1; | |
1077 | if (mm_has_pgste(mm)) | |
1078 | pgste = pgste_get_lock(ptep); | |
1079 | ||
1080 | if (!mm_exclusive(mm)) | |
1081 | __ptep_ipte(address, ptep); | |
1082 | *ptep = pte_wrprotect(pte); | |
1083 | ||
1084 | if (mm_has_pgste(mm)) | |
1085 | pgste_set_unlock(ptep, pgste); | |
1086 | } | |
1087 | return pte; | |
1088 | } | |
ba8a9229 MS |
1089 | |
1090 | #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS | |
b2fa47e6 MS |
1091 | static inline int ptep_set_access_flags(struct vm_area_struct *vma, |
1092 | unsigned long address, pte_t *ptep, | |
1093 | pte_t entry, int dirty) | |
1094 | { | |
1095 | pgste_t pgste; | |
1096 | ||
1097 | if (pte_same(*ptep, entry)) | |
1098 | return 0; | |
1099 | if (mm_has_pgste(vma->vm_mm)) | |
1100 | pgste = pgste_get_lock(ptep); | |
1101 | ||
1102 | __ptep_ipte(address, ptep); | |
1103 | *ptep = entry; | |
1104 | ||
1105 | if (mm_has_pgste(vma->vm_mm)) | |
1106 | pgste_set_unlock(ptep, pgste); | |
1107 | return 1; | |
1108 | } | |
1da177e4 | 1109 | |
1da177e4 LT |
1110 | /* |
1111 | * Conversion functions: convert a page and protection to a page entry, | |
1112 | * and a page entry and page directory to the page they refer to. | |
1113 | */ | |
1114 | static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot) | |
1115 | { | |
1116 | pte_t __pte; | |
1117 | pte_val(__pte) = physpage + pgprot_val(pgprot); | |
1118 | return __pte; | |
1119 | } | |
1120 | ||
2dcea57a HC |
1121 | static inline pte_t mk_pte(struct page *page, pgprot_t pgprot) |
1122 | { | |
0b2b6e1d | 1123 | unsigned long physpage = page_to_phys(page); |
1da177e4 | 1124 | |
2dcea57a HC |
1125 | return mk_pte_phys(physpage, pgprot); |
1126 | } | |
1127 | ||
190a1d72 MS |
1128 | #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1)) |
1129 | #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1)) | |
1130 | #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1)) | |
1131 | #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1)) | |
1da177e4 | 1132 | |
190a1d72 MS |
1133 | #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address)) |
1134 | #define pgd_offset_k(address) pgd_offset(&init_mm, address) | |
1da177e4 | 1135 | |
190a1d72 | 1136 | #ifndef __s390x__ |
1da177e4 | 1137 | |
190a1d72 MS |
1138 | #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN) |
1139 | #define pud_deref(pmd) ({ BUG(); 0UL; }) | |
1140 | #define pgd_deref(pmd) ({ BUG(); 0UL; }) | |
46a82b2d | 1141 | |
190a1d72 MS |
1142 | #define pud_offset(pgd, address) ((pud_t *) pgd) |
1143 | #define pmd_offset(pud, address) ((pmd_t *) pud + pmd_index(address)) | |
1da177e4 | 1144 | |
190a1d72 | 1145 | #else /* __s390x__ */ |
1da177e4 | 1146 | |
190a1d72 MS |
1147 | #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN) |
1148 | #define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN) | |
5a216a20 | 1149 | #define pgd_deref(pgd) (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) |
1da177e4 | 1150 | |
5a216a20 MS |
1151 | static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address) |
1152 | { | |
6252d702 MS |
1153 | pud_t *pud = (pud_t *) pgd; |
1154 | if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2) | |
1155 | pud = (pud_t *) pgd_deref(*pgd); | |
5a216a20 MS |
1156 | return pud + pud_index(address); |
1157 | } | |
1da177e4 | 1158 | |
190a1d72 | 1159 | static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address) |
1da177e4 | 1160 | { |
6252d702 MS |
1161 | pmd_t *pmd = (pmd_t *) pud; |
1162 | if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3) | |
1163 | pmd = (pmd_t *) pud_deref(*pud); | |
190a1d72 | 1164 | return pmd + pmd_index(address); |
1da177e4 LT |
1165 | } |
1166 | ||
190a1d72 | 1167 | #endif /* __s390x__ */ |
1da177e4 | 1168 | |
190a1d72 MS |
1169 | #define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot)) |
1170 | #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT) | |
1171 | #define pte_page(x) pfn_to_page(pte_pfn(x)) | |
1da177e4 | 1172 | |
190a1d72 | 1173 | #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT) |
1da177e4 | 1174 | |
190a1d72 MS |
1175 | /* Find an entry in the lowest level page table.. */ |
1176 | #define pte_offset(pmd, addr) ((pte_t *) pmd_deref(*(pmd)) + pte_index(addr)) | |
1177 | #define pte_offset_kernel(pmd, address) pte_offset(pmd,address) | |
1da177e4 | 1178 | #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address) |
1da177e4 | 1179 | #define pte_unmap(pte) do { } while (0) |
1da177e4 LT |
1180 | |
1181 | /* | |
1182 | * 31 bit swap entry format: | |
1183 | * A page-table entry has some bits we have to treat in a special way. | |
1184 | * Bits 0, 20 and bit 23 have to be zero, otherwise an specification | |
1185 | * exception will occur instead of a page translation exception. The | |
1186 | * specifiation exception has the bad habit not to store necessary | |
1187 | * information in the lowcore. | |
1188 | * Bit 21 and bit 22 are the page invalid bit and the page protection | |
1189 | * bit. We set both to indicate a swapped page. | |
1190 | * Bit 30 and 31 are used to distinguish the different page types. For | |
1191 | * a swapped page these bits need to be zero. | |
1192 | * This leaves the bits 1-19 and bits 24-29 to store type and offset. | |
1193 | * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19 | |
1194 | * plus 24 for the offset. | |
1195 | * 0| offset |0110|o|type |00| | |
1196 | * 0 0000000001111111111 2222 2 22222 33 | |
1197 | * 0 1234567890123456789 0123 4 56789 01 | |
1198 | * | |
1199 | * 64 bit swap entry format: | |
1200 | * A page-table entry has some bits we have to treat in a special way. | |
1201 | * Bits 52 and bit 55 have to be zero, otherwise an specification | |
1202 | * exception will occur instead of a page translation exception. The | |
1203 | * specifiation exception has the bad habit not to store necessary | |
1204 | * information in the lowcore. | |
1205 | * Bit 53 and bit 54 are the page invalid bit and the page protection | |
1206 | * bit. We set both to indicate a swapped page. | |
1207 | * Bit 62 and 63 are used to distinguish the different page types. For | |
1208 | * a swapped page these bits need to be zero. | |
1209 | * This leaves the bits 0-51 and bits 56-61 to store type and offset. | |
1210 | * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51 | |
1211 | * plus 56 for the offset. | |
1212 | * | offset |0110|o|type |00| | |
1213 | * 0000000000111111111122222222223333333333444444444455 5555 5 55566 66 | |
1214 | * 0123456789012345678901234567890123456789012345678901 2345 6 78901 23 | |
1215 | */ | |
1216 | #ifndef __s390x__ | |
1217 | #define __SWP_OFFSET_MASK (~0UL >> 12) | |
1218 | #else | |
1219 | #define __SWP_OFFSET_MASK (~0UL >> 11) | |
1220 | #endif | |
4448aaf0 | 1221 | static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset) |
1da177e4 LT |
1222 | { |
1223 | pte_t pte; | |
1224 | offset &= __SWP_OFFSET_MASK; | |
9282ed92 | 1225 | pte_val(pte) = _PAGE_TYPE_SWAP | ((type & 0x1f) << 2) | |
1da177e4 LT |
1226 | ((offset & 1UL) << 7) | ((offset & ~1UL) << 11); |
1227 | return pte; | |
1228 | } | |
1229 | ||
1230 | #define __swp_type(entry) (((entry).val >> 2) & 0x1f) | |
1231 | #define __swp_offset(entry) (((entry).val >> 11) | (((entry).val >> 7) & 1)) | |
1232 | #define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) }) | |
1233 | ||
1234 | #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) }) | |
1235 | #define __swp_entry_to_pte(x) ((pte_t) { (x).val }) | |
1236 | ||
1237 | #ifndef __s390x__ | |
1238 | # define PTE_FILE_MAX_BITS 26 | |
1239 | #else /* __s390x__ */ | |
1240 | # define PTE_FILE_MAX_BITS 59 | |
1241 | #endif /* __s390x__ */ | |
1242 | ||
1243 | #define pte_to_pgoff(__pte) \ | |
1244 | ((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f)) | |
1245 | ||
1246 | #define pgoff_to_pte(__off) \ | |
1247 | ((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \ | |
9282ed92 | 1248 | | _PAGE_TYPE_FILE }) |
1da177e4 LT |
1249 | |
1250 | #endif /* !__ASSEMBLY__ */ | |
1251 | ||
1252 | #define kern_addr_valid(addr) (1) | |
1253 | ||
17f34580 HC |
1254 | extern int vmem_add_mapping(unsigned long start, unsigned long size); |
1255 | extern int vmem_remove_mapping(unsigned long start, unsigned long size); | |
402b0862 | 1256 | extern int s390_enable_sie(void); |
f4eb07c1 | 1257 | |
1da177e4 LT |
1258 | /* |
1259 | * No page table caches to initialise | |
1260 | */ | |
1261 | #define pgtable_cache_init() do { } while (0) | |
1262 | ||
1da177e4 LT |
1263 | #include <asm-generic/pgtable.h> |
1264 | ||
1265 | #endif /* _S390_PAGE_H */ |