Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 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 | |
| 16 | #include <asm-generic/4level-fixup.h> |
| 17 | |
| 18 | /* |
| 19 | * The Linux memory management assumes a three-level page table setup. For |
| 20 | * s390 31 bit we "fold" the mid level into the top-level page table, so |
| 21 | * that we physically have the same two-level page table as the s390 mmu |
| 22 | * expects in 31 bit mode. For s390 64 bit we use three of the five levels |
| 23 | * the hardware provides (region first and region second tables are not |
| 24 | * used). |
| 25 | * |
| 26 | * The "pgd_xxx()" functions are trivial for a folded two-level |
| 27 | * setup: the pgd is never bad, and a pmd always exists (as it's folded |
| 28 | * into the pgd entry) |
| 29 | * |
| 30 | * This file contains the functions and defines necessary to modify and use |
| 31 | * the S390 page table tree. |
| 32 | */ |
| 33 | #ifndef __ASSEMBLY__ |
| 34 | #include <asm/bug.h> |
| 35 | #include <asm/processor.h> |
| 36 | #include <linux/threads.h> |
| 37 | |
| 38 | struct vm_area_struct; /* forward declaration (include/linux/mm.h) */ |
| 39 | |
| 40 | extern pgd_t swapper_pg_dir[] __attribute__ ((aligned (4096))); |
| 41 | extern void paging_init(void); |
| 42 | |
| 43 | /* |
| 44 | * The S390 doesn't have any external MMU info: the kernel page |
| 45 | * tables contain all the necessary information. |
| 46 | */ |
| 47 | #define update_mmu_cache(vma, address, pte) do { } while (0) |
| 48 | |
| 49 | /* |
| 50 | * ZERO_PAGE is a global shared page that is always zero: used |
| 51 | * for zero-mapped memory areas etc.. |
| 52 | */ |
| 53 | extern char empty_zero_page[PAGE_SIZE]; |
| 54 | #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page)) |
| 55 | #endif /* !__ASSEMBLY__ */ |
| 56 | |
| 57 | /* |
| 58 | * PMD_SHIFT determines the size of the area a second-level page |
| 59 | * table can map |
| 60 | * PGDIR_SHIFT determines what a third-level page table entry can map |
| 61 | */ |
| 62 | #ifndef __s390x__ |
| 63 | # define PMD_SHIFT 22 |
| 64 | # define PGDIR_SHIFT 22 |
| 65 | #else /* __s390x__ */ |
| 66 | # define PMD_SHIFT 21 |
| 67 | # define PGDIR_SHIFT 31 |
| 68 | #endif /* __s390x__ */ |
| 69 | |
| 70 | #define PMD_SIZE (1UL << PMD_SHIFT) |
| 71 | #define PMD_MASK (~(PMD_SIZE-1)) |
| 72 | #define PGDIR_SIZE (1UL << PGDIR_SHIFT) |
| 73 | #define PGDIR_MASK (~(PGDIR_SIZE-1)) |
| 74 | |
| 75 | /* |
| 76 | * entries per page directory level: the S390 is two-level, so |
| 77 | * we don't really have any PMD directory physically. |
| 78 | * for S390 segment-table entries are combined to one PGD |
| 79 | * that leads to 1024 pte per pgd |
| 80 | */ |
| 81 | #ifndef __s390x__ |
| 82 | # define PTRS_PER_PTE 1024 |
| 83 | # define PTRS_PER_PMD 1 |
| 84 | # define PTRS_PER_PGD 512 |
| 85 | #else /* __s390x__ */ |
| 86 | # define PTRS_PER_PTE 512 |
| 87 | # define PTRS_PER_PMD 1024 |
| 88 | # define PTRS_PER_PGD 2048 |
| 89 | #endif /* __s390x__ */ |
| 90 | |
| 91 | /* |
| 92 | * pgd entries used up by user/kernel: |
| 93 | */ |
| 94 | #ifndef __s390x__ |
| 95 | # define USER_PTRS_PER_PGD 512 |
| 96 | # define USER_PGD_PTRS 512 |
| 97 | # define KERNEL_PGD_PTRS 512 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 98 | #else /* __s390x__ */ |
| 99 | # define USER_PTRS_PER_PGD 2048 |
| 100 | # define USER_PGD_PTRS 2048 |
| 101 | # define KERNEL_PGD_PTRS 2048 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 102 | #endif /* __s390x__ */ |
| 103 | |
Hugh Dickins | d455a36 | 2005-04-19 13:29:23 -0700 | [diff] [blame] | 104 | #define FIRST_USER_ADDRESS 0 |
| 105 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 106 | #define pte_ERROR(e) \ |
| 107 | printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e)) |
| 108 | #define pmd_ERROR(e) \ |
| 109 | printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e)) |
| 110 | #define pgd_ERROR(e) \ |
| 111 | printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e)) |
| 112 | |
| 113 | #ifndef __ASSEMBLY__ |
| 114 | /* |
| 115 | * Just any arbitrary offset to the start of the vmalloc VM area: the |
| 116 | * current 8MB value just means that there will be a 8MB "hole" after the |
| 117 | * physical memory until the kernel virtual memory starts. That means that |
| 118 | * any out-of-bounds memory accesses will hopefully be caught. |
| 119 | * The vmalloc() routines leaves a hole of 4kB between each vmalloced |
| 120 | * area for the same reason. ;) |
| 121 | */ |
| 122 | #define VMALLOC_OFFSET (8*1024*1024) |
| 123 | #define VMALLOC_START (((unsigned long) high_memory + VMALLOC_OFFSET) \ |
| 124 | & ~(VMALLOC_OFFSET-1)) |
| 125 | #ifndef __s390x__ |
| 126 | # define VMALLOC_END (0x7fffffffL) |
| 127 | #else /* __s390x__ */ |
| 128 | # define VMALLOC_END (0x40000000000L) |
| 129 | #endif /* __s390x__ */ |
| 130 | |
| 131 | |
| 132 | /* |
| 133 | * A 31 bit pagetable entry of S390 has following format: |
| 134 | * | PFRA | | OS | |
| 135 | * 0 0IP0 |
| 136 | * 00000000001111111111222222222233 |
| 137 | * 01234567890123456789012345678901 |
| 138 | * |
| 139 | * I Page-Invalid Bit: Page is not available for address-translation |
| 140 | * P Page-Protection Bit: Store access not possible for page |
| 141 | * |
| 142 | * A 31 bit segmenttable entry of S390 has following format: |
| 143 | * | P-table origin | |PTL |
| 144 | * 0 IC |
| 145 | * 00000000001111111111222222222233 |
| 146 | * 01234567890123456789012345678901 |
| 147 | * |
| 148 | * I Segment-Invalid Bit: Segment is not available for address-translation |
| 149 | * C Common-Segment Bit: Segment is not private (PoP 3-30) |
| 150 | * PTL Page-Table-Length: Page-table length (PTL+1*16 entries -> up to 256) |
| 151 | * |
| 152 | * The 31 bit segmenttable origin of S390 has following format: |
| 153 | * |
| 154 | * |S-table origin | | STL | |
| 155 | * X **GPS |
| 156 | * 00000000001111111111222222222233 |
| 157 | * 01234567890123456789012345678901 |
| 158 | * |
| 159 | * X Space-Switch event: |
| 160 | * G Segment-Invalid Bit: * |
| 161 | * P Private-Space Bit: Segment is not private (PoP 3-30) |
| 162 | * S Storage-Alteration: |
| 163 | * STL Segment-Table-Length: Segment-table length (STL+1*16 entries -> up to 2048) |
| 164 | * |
| 165 | * A 64 bit pagetable entry of S390 has following format: |
| 166 | * | PFRA |0IP0| OS | |
| 167 | * 0000000000111111111122222222223333333333444444444455555555556666 |
| 168 | * 0123456789012345678901234567890123456789012345678901234567890123 |
| 169 | * |
| 170 | * I Page-Invalid Bit: Page is not available for address-translation |
| 171 | * P Page-Protection Bit: Store access not possible for page |
| 172 | * |
| 173 | * A 64 bit segmenttable entry of S390 has following format: |
| 174 | * | P-table origin | TT |
| 175 | * 0000000000111111111122222222223333333333444444444455555555556666 |
| 176 | * 0123456789012345678901234567890123456789012345678901234567890123 |
| 177 | * |
| 178 | * I Segment-Invalid Bit: Segment is not available for address-translation |
| 179 | * C Common-Segment Bit: Segment is not private (PoP 3-30) |
| 180 | * P Page-Protection Bit: Store access not possible for page |
| 181 | * TT Type 00 |
| 182 | * |
| 183 | * A 64 bit region table entry of S390 has following format: |
| 184 | * | S-table origin | TF TTTL |
| 185 | * 0000000000111111111122222222223333333333444444444455555555556666 |
| 186 | * 0123456789012345678901234567890123456789012345678901234567890123 |
| 187 | * |
| 188 | * I Segment-Invalid Bit: Segment is not available for address-translation |
| 189 | * TT Type 01 |
| 190 | * TF |
| 191 | * TL Table lenght |
| 192 | * |
| 193 | * The 64 bit regiontable origin of S390 has following format: |
| 194 | * | region table origon | DTTL |
| 195 | * 0000000000111111111122222222223333333333444444444455555555556666 |
| 196 | * 0123456789012345678901234567890123456789012345678901234567890123 |
| 197 | * |
| 198 | * X Space-Switch event: |
| 199 | * G Segment-Invalid Bit: |
| 200 | * P Private-Space Bit: |
| 201 | * S Storage-Alteration: |
| 202 | * R Real space |
| 203 | * TL Table-Length: |
| 204 | * |
| 205 | * A storage key has the following format: |
| 206 | * | ACC |F|R|C|0| |
| 207 | * 0 3 4 5 6 7 |
| 208 | * ACC: access key |
| 209 | * F : fetch protection bit |
| 210 | * R : referenced bit |
| 211 | * C : changed bit |
| 212 | */ |
| 213 | |
| 214 | /* Hardware bits in the page table entry */ |
| 215 | #define _PAGE_RO 0x200 /* HW read-only */ |
| 216 | #define _PAGE_INVALID 0x400 /* HW invalid */ |
| 217 | |
| 218 | /* Mask and four different kinds of invalid pages. */ |
| 219 | #define _PAGE_INVALID_MASK 0x601 |
| 220 | #define _PAGE_INVALID_EMPTY 0x400 |
| 221 | #define _PAGE_INVALID_NONE 0x401 |
| 222 | #define _PAGE_INVALID_SWAP 0x600 |
| 223 | #define _PAGE_INVALID_FILE 0x601 |
| 224 | |
| 225 | #ifndef __s390x__ |
| 226 | |
| 227 | /* Bits in the segment table entry */ |
| 228 | #define _PAGE_TABLE_LEN 0xf /* only full page-tables */ |
| 229 | #define _PAGE_TABLE_COM 0x10 /* common page-table */ |
| 230 | #define _PAGE_TABLE_INV 0x20 /* invalid page-table */ |
| 231 | #define _SEG_PRESENT 0x001 /* Software (overlap with PTL) */ |
| 232 | |
| 233 | /* Bits int the storage key */ |
| 234 | #define _PAGE_CHANGED 0x02 /* HW changed bit */ |
| 235 | #define _PAGE_REFERENCED 0x04 /* HW referenced bit */ |
| 236 | |
| 237 | #define _USER_SEG_TABLE_LEN 0x7f /* user-segment-table up to 2 GB */ |
| 238 | #define _KERNEL_SEG_TABLE_LEN 0x7f /* kernel-segment-table up to 2 GB */ |
| 239 | |
| 240 | /* |
| 241 | * User and Kernel pagetables are identical |
| 242 | */ |
| 243 | #define _PAGE_TABLE _PAGE_TABLE_LEN |
| 244 | #define _KERNPG_TABLE _PAGE_TABLE_LEN |
| 245 | |
| 246 | /* |
| 247 | * The Kernel segment-tables includes the User segment-table |
| 248 | */ |
| 249 | |
| 250 | #define _SEGMENT_TABLE (_USER_SEG_TABLE_LEN|0x80000000|0x100) |
| 251 | #define _KERNSEG_TABLE _KERNEL_SEG_TABLE_LEN |
| 252 | |
| 253 | #define USER_STD_MASK 0x00000080UL |
| 254 | |
| 255 | #else /* __s390x__ */ |
| 256 | |
| 257 | /* Bits in the segment table entry */ |
| 258 | #define _PMD_ENTRY_INV 0x20 /* invalid segment table entry */ |
| 259 | #define _PMD_ENTRY 0x00 |
| 260 | |
| 261 | /* Bits in the region third table entry */ |
| 262 | #define _PGD_ENTRY_INV 0x20 /* invalid region table entry */ |
| 263 | #define _PGD_ENTRY 0x07 |
| 264 | |
| 265 | /* |
| 266 | * User and kernel page directory |
| 267 | */ |
| 268 | #define _REGION_THIRD 0x4 |
| 269 | #define _REGION_THIRD_LEN 0x3 |
| 270 | #define _REGION_TABLE (_REGION_THIRD|_REGION_THIRD_LEN|0x40|0x100) |
| 271 | #define _KERN_REGION_TABLE (_REGION_THIRD|_REGION_THIRD_LEN) |
| 272 | |
| 273 | #define USER_STD_MASK 0x0000000000000080UL |
| 274 | |
| 275 | /* Bits in the storage key */ |
| 276 | #define _PAGE_CHANGED 0x02 /* HW changed bit */ |
| 277 | #define _PAGE_REFERENCED 0x04 /* HW referenced bit */ |
| 278 | |
| 279 | #endif /* __s390x__ */ |
| 280 | |
| 281 | /* |
| 282 | * No mapping available |
| 283 | */ |
| 284 | #define PAGE_NONE_SHARED __pgprot(_PAGE_INVALID_NONE) |
| 285 | #define PAGE_NONE_PRIVATE __pgprot(_PAGE_INVALID_NONE) |
| 286 | #define PAGE_RO_SHARED __pgprot(_PAGE_RO) |
| 287 | #define PAGE_RO_PRIVATE __pgprot(_PAGE_RO) |
| 288 | #define PAGE_COPY __pgprot(_PAGE_RO) |
| 289 | #define PAGE_SHARED __pgprot(0) |
| 290 | #define PAGE_KERNEL __pgprot(0) |
| 291 | |
| 292 | /* |
| 293 | * The S390 can't do page protection for execute, and considers that the |
| 294 | * same are read. Also, write permissions imply read permissions. This is |
| 295 | * the closest we can get.. |
| 296 | */ |
| 297 | /*xwr*/ |
| 298 | #define __P000 PAGE_NONE_PRIVATE |
| 299 | #define __P001 PAGE_RO_PRIVATE |
| 300 | #define __P010 PAGE_COPY |
| 301 | #define __P011 PAGE_COPY |
| 302 | #define __P100 PAGE_RO_PRIVATE |
| 303 | #define __P101 PAGE_RO_PRIVATE |
| 304 | #define __P110 PAGE_COPY |
| 305 | #define __P111 PAGE_COPY |
| 306 | |
| 307 | #define __S000 PAGE_NONE_SHARED |
| 308 | #define __S001 PAGE_RO_SHARED |
| 309 | #define __S010 PAGE_SHARED |
| 310 | #define __S011 PAGE_SHARED |
| 311 | #define __S100 PAGE_RO_SHARED |
| 312 | #define __S101 PAGE_RO_SHARED |
| 313 | #define __S110 PAGE_SHARED |
| 314 | #define __S111 PAGE_SHARED |
| 315 | |
| 316 | /* |
| 317 | * Certain architectures need to do special things when PTEs |
| 318 | * within a page table are directly modified. Thus, the following |
| 319 | * hook is made available. |
| 320 | */ |
| 321 | extern inline void set_pte(pte_t *pteptr, pte_t pteval) |
| 322 | { |
| 323 | *pteptr = pteval; |
| 324 | } |
| 325 | #define set_pte_at(mm,addr,ptep,pteval) set_pte(ptep,pteval) |
| 326 | |
| 327 | /* |
| 328 | * pgd/pmd/pte query functions |
| 329 | */ |
| 330 | #ifndef __s390x__ |
| 331 | |
| 332 | extern inline int pgd_present(pgd_t pgd) { return 1; } |
| 333 | extern inline int pgd_none(pgd_t pgd) { return 0; } |
| 334 | extern inline int pgd_bad(pgd_t pgd) { return 0; } |
| 335 | |
| 336 | extern inline int pmd_present(pmd_t pmd) { return pmd_val(pmd) & _SEG_PRESENT; } |
| 337 | extern inline int pmd_none(pmd_t pmd) { return pmd_val(pmd) & _PAGE_TABLE_INV; } |
| 338 | extern inline int pmd_bad(pmd_t pmd) |
| 339 | { |
| 340 | return (pmd_val(pmd) & (~PAGE_MASK & ~_PAGE_TABLE_INV)) != _PAGE_TABLE; |
| 341 | } |
| 342 | |
| 343 | #else /* __s390x__ */ |
| 344 | |
| 345 | extern inline int pgd_present(pgd_t pgd) |
| 346 | { |
| 347 | return (pgd_val(pgd) & ~PAGE_MASK) == _PGD_ENTRY; |
| 348 | } |
| 349 | |
| 350 | extern inline int pgd_none(pgd_t pgd) |
| 351 | { |
| 352 | return pgd_val(pgd) & _PGD_ENTRY_INV; |
| 353 | } |
| 354 | |
| 355 | extern inline int pgd_bad(pgd_t pgd) |
| 356 | { |
| 357 | return (pgd_val(pgd) & (~PAGE_MASK & ~_PGD_ENTRY_INV)) != _PGD_ENTRY; |
| 358 | } |
| 359 | |
| 360 | extern inline int pmd_present(pmd_t pmd) |
| 361 | { |
| 362 | return (pmd_val(pmd) & ~PAGE_MASK) == _PMD_ENTRY; |
| 363 | } |
| 364 | |
| 365 | extern inline int pmd_none(pmd_t pmd) |
| 366 | { |
| 367 | return pmd_val(pmd) & _PMD_ENTRY_INV; |
| 368 | } |
| 369 | |
| 370 | extern inline int pmd_bad(pmd_t pmd) |
| 371 | { |
| 372 | return (pmd_val(pmd) & (~PAGE_MASK & ~_PMD_ENTRY_INV)) != _PMD_ENTRY; |
| 373 | } |
| 374 | |
| 375 | #endif /* __s390x__ */ |
| 376 | |
| 377 | extern inline int pte_none(pte_t pte) |
| 378 | { |
| 379 | return (pte_val(pte) & _PAGE_INVALID_MASK) == _PAGE_INVALID_EMPTY; |
| 380 | } |
| 381 | |
| 382 | extern inline int pte_present(pte_t pte) |
| 383 | { |
| 384 | return !(pte_val(pte) & _PAGE_INVALID) || |
| 385 | (pte_val(pte) & _PAGE_INVALID_MASK) == _PAGE_INVALID_NONE; |
| 386 | } |
| 387 | |
| 388 | extern inline int pte_file(pte_t pte) |
| 389 | { |
| 390 | return (pte_val(pte) & _PAGE_INVALID_MASK) == _PAGE_INVALID_FILE; |
| 391 | } |
| 392 | |
| 393 | #define pte_same(a,b) (pte_val(a) == pte_val(b)) |
| 394 | |
| 395 | /* |
| 396 | * query functions pte_write/pte_dirty/pte_young only work if |
| 397 | * pte_present() is true. Undefined behaviour if not.. |
| 398 | */ |
| 399 | extern inline int pte_write(pte_t pte) |
| 400 | { |
| 401 | return (pte_val(pte) & _PAGE_RO) == 0; |
| 402 | } |
| 403 | |
| 404 | extern inline int pte_dirty(pte_t pte) |
| 405 | { |
| 406 | /* A pte is neither clean nor dirty on s/390. The dirty bit |
| 407 | * is in the storage key. See page_test_and_clear_dirty for |
| 408 | * details. |
| 409 | */ |
| 410 | return 0; |
| 411 | } |
| 412 | |
| 413 | extern inline int pte_young(pte_t pte) |
| 414 | { |
| 415 | /* A pte is neither young nor old on s/390. The young bit |
| 416 | * is in the storage key. See page_test_and_clear_young for |
| 417 | * details. |
| 418 | */ |
| 419 | return 0; |
| 420 | } |
| 421 | |
| 422 | extern inline int pte_read(pte_t pte) |
| 423 | { |
| 424 | /* All pages are readable since we don't use the fetch |
| 425 | * protection bit in the storage key. |
| 426 | */ |
| 427 | return 1; |
| 428 | } |
| 429 | |
| 430 | /* |
| 431 | * pgd/pmd/pte modification functions |
| 432 | */ |
| 433 | |
| 434 | #ifndef __s390x__ |
| 435 | |
| 436 | extern inline void pgd_clear(pgd_t * pgdp) { } |
| 437 | |
| 438 | extern inline void pmd_clear(pmd_t * pmdp) |
| 439 | { |
| 440 | pmd_val(pmdp[0]) = _PAGE_TABLE_INV; |
| 441 | pmd_val(pmdp[1]) = _PAGE_TABLE_INV; |
| 442 | pmd_val(pmdp[2]) = _PAGE_TABLE_INV; |
| 443 | pmd_val(pmdp[3]) = _PAGE_TABLE_INV; |
| 444 | } |
| 445 | |
| 446 | #else /* __s390x__ */ |
| 447 | |
| 448 | extern inline void pgd_clear(pgd_t * pgdp) |
| 449 | { |
| 450 | pgd_val(*pgdp) = _PGD_ENTRY_INV | _PGD_ENTRY; |
| 451 | } |
| 452 | |
| 453 | extern inline void pmd_clear(pmd_t * pmdp) |
| 454 | { |
| 455 | pmd_val(*pmdp) = _PMD_ENTRY_INV | _PMD_ENTRY; |
| 456 | pmd_val1(*pmdp) = _PMD_ENTRY_INV | _PMD_ENTRY; |
| 457 | } |
| 458 | |
| 459 | #endif /* __s390x__ */ |
| 460 | |
| 461 | extern inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep) |
| 462 | { |
| 463 | pte_val(*ptep) = _PAGE_INVALID_EMPTY; |
| 464 | } |
| 465 | |
| 466 | /* |
| 467 | * The following pte modification functions only work if |
| 468 | * pte_present() is true. Undefined behaviour if not.. |
| 469 | */ |
| 470 | extern inline pte_t pte_modify(pte_t pte, pgprot_t newprot) |
| 471 | { |
| 472 | pte_val(pte) &= PAGE_MASK; |
| 473 | pte_val(pte) |= pgprot_val(newprot); |
| 474 | return pte; |
| 475 | } |
| 476 | |
| 477 | extern inline pte_t pte_wrprotect(pte_t pte) |
| 478 | { |
| 479 | /* Do not clobber _PAGE_INVALID_NONE pages! */ |
| 480 | if (!(pte_val(pte) & _PAGE_INVALID)) |
| 481 | pte_val(pte) |= _PAGE_RO; |
| 482 | return pte; |
| 483 | } |
| 484 | |
| 485 | extern inline pte_t pte_mkwrite(pte_t pte) |
| 486 | { |
| 487 | pte_val(pte) &= ~_PAGE_RO; |
| 488 | return pte; |
| 489 | } |
| 490 | |
| 491 | extern inline pte_t pte_mkclean(pte_t pte) |
| 492 | { |
| 493 | /* The only user of pte_mkclean is the fork() code. |
| 494 | We must *not* clear the *physical* page dirty bit |
| 495 | just because fork() wants to clear the dirty bit in |
| 496 | *one* of the page's mappings. So we just do nothing. */ |
| 497 | return pte; |
| 498 | } |
| 499 | |
| 500 | extern inline pte_t pte_mkdirty(pte_t pte) |
| 501 | { |
| 502 | /* We do not explicitly set the dirty bit because the |
| 503 | * sske instruction is slow. It is faster to let the |
| 504 | * next instruction set the dirty bit. |
| 505 | */ |
| 506 | return pte; |
| 507 | } |
| 508 | |
| 509 | extern inline pte_t pte_mkold(pte_t pte) |
| 510 | { |
| 511 | /* S/390 doesn't keep its dirty/referenced bit in the pte. |
| 512 | * There is no point in clearing the real referenced bit. |
| 513 | */ |
| 514 | return pte; |
| 515 | } |
| 516 | |
| 517 | extern inline pte_t pte_mkyoung(pte_t pte) |
| 518 | { |
| 519 | /* S/390 doesn't keep its dirty/referenced bit in the pte. |
| 520 | * There is no point in setting the real referenced bit. |
| 521 | */ |
| 522 | return pte; |
| 523 | } |
| 524 | |
| 525 | static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep) |
| 526 | { |
| 527 | return 0; |
| 528 | } |
| 529 | |
| 530 | static inline int |
| 531 | ptep_clear_flush_young(struct vm_area_struct *vma, |
| 532 | unsigned long address, pte_t *ptep) |
| 533 | { |
| 534 | /* No need to flush TLB; bits are in storage key */ |
| 535 | return ptep_test_and_clear_young(vma, address, ptep); |
| 536 | } |
| 537 | |
| 538 | static inline int ptep_test_and_clear_dirty(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep) |
| 539 | { |
| 540 | return 0; |
| 541 | } |
| 542 | |
| 543 | static inline int |
| 544 | ptep_clear_flush_dirty(struct vm_area_struct *vma, |
| 545 | unsigned long address, pte_t *ptep) |
| 546 | { |
| 547 | /* No need to flush TLB; bits are in storage key */ |
| 548 | return ptep_test_and_clear_dirty(vma, address, ptep); |
| 549 | } |
| 550 | |
| 551 | static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep) |
| 552 | { |
| 553 | pte_t pte = *ptep; |
| 554 | pte_clear(mm, addr, ptep); |
| 555 | return pte; |
| 556 | } |
| 557 | |
| 558 | static inline pte_t |
| 559 | ptep_clear_flush(struct vm_area_struct *vma, |
| 560 | unsigned long address, pte_t *ptep) |
| 561 | { |
| 562 | pte_t pte = *ptep; |
| 563 | #ifndef __s390x__ |
| 564 | if (!(pte_val(pte) & _PAGE_INVALID)) { |
| 565 | /* S390 has 1mb segments, we are emulating 4MB segments */ |
| 566 | pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00); |
| 567 | __asm__ __volatile__ ("ipte %2,%3" |
| 568 | : "=m" (*ptep) : "m" (*ptep), |
| 569 | "a" (pto), "a" (address) ); |
| 570 | } |
| 571 | #else /* __s390x__ */ |
| 572 | if (!(pte_val(pte) & _PAGE_INVALID)) |
| 573 | __asm__ __volatile__ ("ipte %2,%3" |
| 574 | : "=m" (*ptep) : "m" (*ptep), |
| 575 | "a" (ptep), "a" (address) ); |
| 576 | #endif /* __s390x__ */ |
| 577 | pte_val(*ptep) = _PAGE_INVALID_EMPTY; |
| 578 | return pte; |
| 579 | } |
| 580 | |
| 581 | static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep) |
| 582 | { |
| 583 | pte_t old_pte = *ptep; |
| 584 | set_pte_at(mm, addr, ptep, pte_wrprotect(old_pte)); |
| 585 | } |
| 586 | |
| 587 | static inline void |
| 588 | ptep_establish(struct vm_area_struct *vma, |
| 589 | unsigned long address, pte_t *ptep, |
| 590 | pte_t entry) |
| 591 | { |
| 592 | ptep_clear_flush(vma, address, ptep); |
| 593 | set_pte(ptep, entry); |
| 594 | } |
| 595 | |
| 596 | #define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \ |
| 597 | ptep_establish(__vma, __address, __ptep, __entry) |
| 598 | |
| 599 | /* |
| 600 | * Test and clear dirty bit in storage key. |
| 601 | * We can't clear the changed bit atomically. This is a potential |
| 602 | * race against modification of the referenced bit. This function |
| 603 | * should therefore only be called if it is not mapped in any |
| 604 | * address space. |
| 605 | */ |
| 606 | #define page_test_and_clear_dirty(_page) \ |
| 607 | ({ \ |
| 608 | struct page *__page = (_page); \ |
| 609 | unsigned long __physpage = __pa((__page-mem_map) << PAGE_SHIFT); \ |
| 610 | int __skey = page_get_storage_key(__physpage); \ |
| 611 | if (__skey & _PAGE_CHANGED) \ |
| 612 | page_set_storage_key(__physpage, __skey & ~_PAGE_CHANGED);\ |
| 613 | (__skey & _PAGE_CHANGED); \ |
| 614 | }) |
| 615 | |
| 616 | /* |
| 617 | * Test and clear referenced bit in storage key. |
| 618 | */ |
| 619 | #define page_test_and_clear_young(page) \ |
| 620 | ({ \ |
| 621 | struct page *__page = (page); \ |
| 622 | unsigned long __physpage = __pa((__page-mem_map) << PAGE_SHIFT); \ |
| 623 | int __ccode; \ |
| 624 | asm volatile ("rrbe 0,%1\n\t" \ |
| 625 | "ipm %0\n\t" \ |
| 626 | "srl %0,28\n\t" \ |
| 627 | : "=d" (__ccode) : "a" (__physpage) : "cc" ); \ |
| 628 | (__ccode & 2); \ |
| 629 | }) |
| 630 | |
| 631 | /* |
| 632 | * Conversion functions: convert a page and protection to a page entry, |
| 633 | * and a page entry and page directory to the page they refer to. |
| 634 | */ |
| 635 | static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot) |
| 636 | { |
| 637 | pte_t __pte; |
| 638 | pte_val(__pte) = physpage + pgprot_val(pgprot); |
| 639 | return __pte; |
| 640 | } |
| 641 | |
| 642 | #define mk_pte(pg, pgprot) \ |
| 643 | ({ \ |
| 644 | struct page *__page = (pg); \ |
| 645 | pgprot_t __pgprot = (pgprot); \ |
| 646 | unsigned long __physpage = __pa((__page-mem_map) << PAGE_SHIFT); \ |
| 647 | pte_t __pte = mk_pte_phys(__physpage, __pgprot); \ |
| 648 | __pte; \ |
| 649 | }) |
| 650 | |
| 651 | #define pfn_pte(pfn, pgprot) \ |
| 652 | ({ \ |
| 653 | pgprot_t __pgprot = (pgprot); \ |
| 654 | unsigned long __physpage = __pa((pfn) << PAGE_SHIFT); \ |
| 655 | pte_t __pte = mk_pte_phys(__physpage, __pgprot); \ |
| 656 | __pte; \ |
| 657 | }) |
| 658 | |
| 659 | #define SetPageUptodate(_page) \ |
| 660 | do { \ |
| 661 | struct page *__page = (_page); \ |
| 662 | if (!test_and_set_bit(PG_uptodate, &__page->flags)) \ |
| 663 | page_test_and_clear_dirty(_page); \ |
| 664 | } while (0) |
| 665 | |
| 666 | #ifdef __s390x__ |
| 667 | |
| 668 | #define pfn_pmd(pfn, pgprot) \ |
| 669 | ({ \ |
| 670 | pgprot_t __pgprot = (pgprot); \ |
| 671 | unsigned long __physpage = __pa((pfn) << PAGE_SHIFT); \ |
| 672 | pmd_t __pmd = __pmd(__physpage + pgprot_val(__pgprot)); \ |
| 673 | __pmd; \ |
| 674 | }) |
| 675 | |
| 676 | #endif /* __s390x__ */ |
| 677 | |
| 678 | #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT) |
| 679 | #define pte_page(x) pfn_to_page(pte_pfn(x)) |
| 680 | |
| 681 | #define pmd_page_kernel(pmd) (pmd_val(pmd) & PAGE_MASK) |
| 682 | |
| 683 | #define pmd_page(pmd) (mem_map+(pmd_val(pmd) >> PAGE_SHIFT)) |
| 684 | |
| 685 | #define pgd_page_kernel(pgd) (pgd_val(pgd) & PAGE_MASK) |
| 686 | |
| 687 | /* to find an entry in a page-table-directory */ |
| 688 | #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1)) |
| 689 | #define pgd_offset(mm, address) ((mm)->pgd+pgd_index(address)) |
| 690 | |
| 691 | /* to find an entry in a kernel page-table-directory */ |
| 692 | #define pgd_offset_k(address) pgd_offset(&init_mm, address) |
| 693 | |
| 694 | #ifndef __s390x__ |
| 695 | |
| 696 | /* Find an entry in the second-level page table.. */ |
| 697 | extern inline pmd_t * pmd_offset(pgd_t * dir, unsigned long address) |
| 698 | { |
| 699 | return (pmd_t *) dir; |
| 700 | } |
| 701 | |
| 702 | #else /* __s390x__ */ |
| 703 | |
| 704 | /* Find an entry in the second-level page table.. */ |
| 705 | #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1)) |
| 706 | #define pmd_offset(dir,addr) \ |
| 707 | ((pmd_t *) pgd_page_kernel(*(dir)) + pmd_index(addr)) |
| 708 | |
| 709 | #endif /* __s390x__ */ |
| 710 | |
| 711 | /* Find an entry in the third-level page table.. */ |
| 712 | #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1)) |
| 713 | #define pte_offset_kernel(pmd, address) \ |
| 714 | ((pte_t *) pmd_page_kernel(*(pmd)) + pte_index(address)) |
| 715 | #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address) |
| 716 | #define pte_offset_map_nested(pmd, address) pte_offset_kernel(pmd, address) |
| 717 | #define pte_unmap(pte) do { } while (0) |
| 718 | #define pte_unmap_nested(pte) do { } while (0) |
| 719 | |
| 720 | /* |
| 721 | * 31 bit swap entry format: |
| 722 | * A page-table entry has some bits we have to treat in a special way. |
| 723 | * Bits 0, 20 and bit 23 have to be zero, otherwise an specification |
| 724 | * exception will occur instead of a page translation exception. The |
| 725 | * specifiation exception has the bad habit not to store necessary |
| 726 | * information in the lowcore. |
| 727 | * Bit 21 and bit 22 are the page invalid bit and the page protection |
| 728 | * bit. We set both to indicate a swapped page. |
| 729 | * Bit 30 and 31 are used to distinguish the different page types. For |
| 730 | * a swapped page these bits need to be zero. |
| 731 | * This leaves the bits 1-19 and bits 24-29 to store type and offset. |
| 732 | * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19 |
| 733 | * plus 24 for the offset. |
| 734 | * 0| offset |0110|o|type |00| |
| 735 | * 0 0000000001111111111 2222 2 22222 33 |
| 736 | * 0 1234567890123456789 0123 4 56789 01 |
| 737 | * |
| 738 | * 64 bit swap entry format: |
| 739 | * A page-table entry has some bits we have to treat in a special way. |
| 740 | * Bits 52 and bit 55 have to be zero, otherwise an specification |
| 741 | * exception will occur instead of a page translation exception. The |
| 742 | * specifiation exception has the bad habit not to store necessary |
| 743 | * information in the lowcore. |
| 744 | * Bit 53 and bit 54 are the page invalid bit and the page protection |
| 745 | * bit. We set both to indicate a swapped page. |
| 746 | * Bit 62 and 63 are used to distinguish the different page types. For |
| 747 | * a swapped page these bits need to be zero. |
| 748 | * This leaves the bits 0-51 and bits 56-61 to store type and offset. |
| 749 | * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51 |
| 750 | * plus 56 for the offset. |
| 751 | * | offset |0110|o|type |00| |
| 752 | * 0000000000111111111122222222223333333333444444444455 5555 5 55566 66 |
| 753 | * 0123456789012345678901234567890123456789012345678901 2345 6 78901 23 |
| 754 | */ |
| 755 | #ifndef __s390x__ |
| 756 | #define __SWP_OFFSET_MASK (~0UL >> 12) |
| 757 | #else |
| 758 | #define __SWP_OFFSET_MASK (~0UL >> 11) |
| 759 | #endif |
| 760 | extern inline pte_t mk_swap_pte(unsigned long type, unsigned long offset) |
| 761 | { |
| 762 | pte_t pte; |
| 763 | offset &= __SWP_OFFSET_MASK; |
| 764 | pte_val(pte) = _PAGE_INVALID_SWAP | ((type & 0x1f) << 2) | |
| 765 | ((offset & 1UL) << 7) | ((offset & ~1UL) << 11); |
| 766 | return pte; |
| 767 | } |
| 768 | |
| 769 | #define __swp_type(entry) (((entry).val >> 2) & 0x1f) |
| 770 | #define __swp_offset(entry) (((entry).val >> 11) | (((entry).val >> 7) & 1)) |
| 771 | #define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) }) |
| 772 | |
| 773 | #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) }) |
| 774 | #define __swp_entry_to_pte(x) ((pte_t) { (x).val }) |
| 775 | |
| 776 | #ifndef __s390x__ |
| 777 | # define PTE_FILE_MAX_BITS 26 |
| 778 | #else /* __s390x__ */ |
| 779 | # define PTE_FILE_MAX_BITS 59 |
| 780 | #endif /* __s390x__ */ |
| 781 | |
| 782 | #define pte_to_pgoff(__pte) \ |
| 783 | ((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f)) |
| 784 | |
| 785 | #define pgoff_to_pte(__off) \ |
| 786 | ((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \ |
| 787 | | _PAGE_INVALID_FILE }) |
| 788 | |
| 789 | #endif /* !__ASSEMBLY__ */ |
| 790 | |
| 791 | #define kern_addr_valid(addr) (1) |
| 792 | |
| 793 | /* |
| 794 | * No page table caches to initialise |
| 795 | */ |
| 796 | #define pgtable_cache_init() do { } while (0) |
| 797 | |
| 798 | #define __HAVE_ARCH_PTEP_ESTABLISH |
| 799 | #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS |
| 800 | #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG |
| 801 | #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH |
| 802 | #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY |
| 803 | #define __HAVE_ARCH_PTEP_CLEAR_DIRTY_FLUSH |
| 804 | #define __HAVE_ARCH_PTEP_GET_AND_CLEAR |
| 805 | #define __HAVE_ARCH_PTEP_CLEAR_FLUSH |
| 806 | #define __HAVE_ARCH_PTEP_SET_WRPROTECT |
| 807 | #define __HAVE_ARCH_PTE_SAME |
| 808 | #define __HAVE_ARCH_PAGE_TEST_AND_CLEAR_DIRTY |
| 809 | #define __HAVE_ARCH_PAGE_TEST_AND_CLEAR_YOUNG |
| 810 | #include <asm-generic/pgtable.h> |
| 811 | |
| 812 | #endif /* _S390_PAGE_H */ |
| 813 | |