include/asm-s390/pgtable.h

   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 <linux/mm_types.h>
  35 #include <asm/bug.h>
  36 #include <asm/processor.h>
  37
  38 struct vm_area_struct; /* forward declaration (include/linux/mm.h) */
  39 struct mm_struct;
  40
  41 extern pgd_t swapper_pg_dir[] __attribute__ ((aligned (4096)));
  42 extern void paging_init(void);
  43 extern void vmem_map_init(void);
  44
  45 /*
  46  * The S390 doesn't have any external MMU info: the kernel page
  47  * tables contain all the necessary information.
  48  */
  49 #define update_mmu_cache(vma, address, pte)     do { } while (0)
  50
  51 /*
  52  * ZERO_PAGE is a global shared page that is always zero: used
  53  * for zero-mapped memory areas etc..
  54  */
  55 extern char empty_zero_page[PAGE_SIZE];
  56 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
  57 #endif /* !__ASSEMBLY__ */
  58
  59 /*
  60  * PMD_SHIFT determines the size of the area a second-level page
  61  * table can map
  62  * PGDIR_SHIFT determines what a third-level page table entry can map
  63  */
  64 #ifndef __s390x__
  65 # define PMD_SHIFT      22
  66 # define PGDIR_SHIFT    22
  67 #else /* __s390x__ */
  68 # define PMD_SHIFT      21
  69 # define PGDIR_SHIFT    31
  70 #endif /* __s390x__ */
  71
  72 #define PMD_SIZE        (1UL << PMD_SHIFT)
  73 #define PMD_MASK        (~(PMD_SIZE-1))
  74 #define PGDIR_SIZE      (1UL << PGDIR_SHIFT)
  75 #define PGDIR_MASK      (~(PGDIR_SIZE-1))
  76
  77 /*
  78  * entries per page directory level: the S390 is two-level, so
  79  * we don't really have any PMD directory physically.
  80  * for S390 segment-table entries are combined to one PGD
  81  * that leads to 1024 pte per pgd
  82  */
  83 #ifndef __s390x__
  84 # define PTRS_PER_PTE    1024
  85 # define PTRS_PER_PMD    1
  86 # define PTRS_PER_PGD    512
  87 #else /* __s390x__ */
  88 # define PTRS_PER_PTE    512
  89 # define PTRS_PER_PMD    1024
  90 # define PTRS_PER_PGD    2048
  91 #endif /* __s390x__ */
  92
  93 #define FIRST_USER_ADDRESS  0
  94
  95 #define pte_ERROR(e) \
  96         printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e))
  97 #define pmd_ERROR(e) \
  98         printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e))
  99 #define pgd_ERROR(e) \
 100         printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e))
 101
 102 #ifndef __ASSEMBLY__
 103 /*
 104  * Just any arbitrary offset to the start of the vmalloc VM area: the
 105  * current 8MB value just means that there will be a 8MB "hole" after the
 106  * physical memory until the kernel virtual memory starts.  That means that
 107  * any out-of-bounds memory accesses will hopefully be caught.
 108  * The vmalloc() routines leaves a hole of 4kB between each vmalloced
 109  * area for the same reason. ;)
 110  */
 111 extern unsigned long vmalloc_end;
 112 #define VMALLOC_OFFSET  (8*1024*1024)
 113 #define VMALLOC_START   (((unsigned long) high_memory + VMALLOC_OFFSET) \
 114                          & ~(VMALLOC_OFFSET-1))
 115 #define VMALLOC_END     vmalloc_end
 116
 117 /*
 118  * We need some free virtual space to be able to do vmalloc.
 119  * VMALLOC_MIN_SIZE defines the minimum size of the vmalloc
 120  * area. On a machine with 2GB memory we make sure that we
 121  * have at least 128MB free space for vmalloc. On a machine
 122  * with 4TB we make sure we have at least 128GB.
 123  */
 124 #ifndef __s390x__
 125 #define VMALLOC_MIN_SIZE        0x8000000UL
 126 #define VMALLOC_END_INIT        0x80000000UL
 127 #else /* __s390x__ */
 128 #define VMALLOC_MIN_SIZE        0x2000000000UL
 129 #define VMALLOC_END_INIT        0x40000000000UL
 130 #endif /* __s390x__ */
 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 bit  */
 216 #define _PAGE_INVALID   0x400           /* HW invalid bit    */
 217 #define _PAGE_SWT       0x001           /* SW pte type bit t */
 218 #define _PAGE_SWX       0x002           /* SW pte type bit x */
 219
 220 /* Six different types of pages. */
 221 #define _PAGE_TYPE_EMPTY        0x400
 222 #define _PAGE_TYPE_NONE         0x401
 223 #define _PAGE_TYPE_SWAP         0x403
 224 #define _PAGE_TYPE_FILE         0x601   /* bit 0x002 is used for offset !! */
 225 #define _PAGE_TYPE_RO           0x200
 226 #define _PAGE_TYPE_RW           0x000
 227
 228 /*
 229  * PTE type bits are rather complicated. handle_pte_fault uses pte_present,
 230  * pte_none and pte_file to find out the pte type WITHOUT holding the page
 231  * table lock. ptep_clear_flush on the other hand uses ptep_clear_flush to
 232  * invalidate a given pte. ipte sets the hw invalid bit and clears all tlbs
 233  * for the page. The page table entry is set to _PAGE_TYPE_EMPTY afterwards.
 234  * This change is done while holding the lock, but the intermediate step
 235  * of a previously valid pte with the hw invalid bit set can be observed by
 236  * handle_pte_fault. That makes it necessary that all valid pte types with
 237  * the hw invalid bit set must be distinguishable from the four pte types
 238  * empty, none, swap and file.
 239  *
 240  *                      irxt  ipte  irxt
 241  * _PAGE_TYPE_EMPTY     1000   ->   1000
 242  * _PAGE_TYPE_NONE      1001   ->   1001
 243  * _PAGE_TYPE_SWAP      1011   ->   1011
 244  * _PAGE_TYPE_FILE      11?1   ->   11?1
 245  * _PAGE_TYPE_RO        0100   ->   1100
 246  * _PAGE_TYPE_RW        0000   ->   1000
 247  *
 248  * pte_none is true for bits combinations 1000, 1100
 249  * pte_present is true for bits combinations 0000, 0010, 0100, 0110, 1001
 250  * pte_file is true for bits combinations 1101, 1111
 251  * swap pte is 1011 and 0001, 0011, 0101, 0111, 1010 and 1110 are invalid.
 252  */
 253
 254 #ifndef __s390x__
 255
 256 /* Bits in the segment table entry */
 257 #define _PAGE_TABLE_LEN 0xf            /* only full page-tables            */
 258 #define _PAGE_TABLE_COM 0x10           /* common page-table                */
 259 #define _PAGE_TABLE_INV 0x20           /* invalid page-table               */
 260 #define _SEG_PRESENT    0x001          /* Software (overlap with PTL)      */
 261
 262 /* Bits int the storage key */
 263 #define _PAGE_CHANGED    0x02          /* HW changed bit                   */
 264 #define _PAGE_REFERENCED 0x04          /* HW referenced bit                */
 265
 266 #define _USER_SEG_TABLE_LEN    0x7f    /* user-segment-table up to 2 GB    */
 267 #define _KERNEL_SEG_TABLE_LEN  0x7f    /* kernel-segment-table up to 2 GB  */
 268
 269 /*
 270  * User and Kernel pagetables are identical
 271  */
 272 #define _PAGE_TABLE     _PAGE_TABLE_LEN
 273 #define _KERNPG_TABLE   _PAGE_TABLE_LEN
 274
 275 /*
 276  * The Kernel segment-tables includes the User segment-table
 277  */
 278
 279 #define _SEGMENT_TABLE  (_USER_SEG_TABLE_LEN|0x80000000|0x100)
 280 #define _KERNSEG_TABLE  _KERNEL_SEG_TABLE_LEN
 281
 282 #define USER_STD_MASK   0x00000080UL
 283
 284 #else /* __s390x__ */
 285
 286 /* Bits in the segment table entry */
 287 #define _PMD_ENTRY_INV   0x20          /* invalid segment table entry      */
 288 #define _PMD_ENTRY       0x00
 289
 290 /* Bits in the region third table entry */
 291 #define _PGD_ENTRY_INV   0x20          /* invalid region table entry       */
 292 #define _PGD_ENTRY       0x07
 293
 294 /*
 295  * User and kernel page directory
 296  */
 297 #define _REGION_THIRD       0x4
 298 #define _REGION_THIRD_LEN   0x3
 299 #define _REGION_TABLE       (_REGION_THIRD|_REGION_THIRD_LEN|0x40|0x100)
 300 #define _KERN_REGION_TABLE  (_REGION_THIRD|_REGION_THIRD_LEN)
 301
 302 #define USER_STD_MASK           0x0000000000000080UL
 303
 304 /* Bits in the storage key */
 305 #define _PAGE_CHANGED    0x02          /* HW changed bit                   */
 306 #define _PAGE_REFERENCED 0x04          /* HW referenced bit                */
 307
 308 #endif /* __s390x__ */
 309
 310 /*
 311  * Page protection definitions.
 312  */
 313 #define PAGE_NONE       __pgprot(_PAGE_TYPE_NONE)
 314 #define PAGE_RO         __pgprot(_PAGE_TYPE_RO)
 315 #define PAGE_RW         __pgprot(_PAGE_TYPE_RW)
 316
 317 #define PAGE_KERNEL     PAGE_RW
 318 #define PAGE_COPY       PAGE_RO
 319
 320 /*
 321  * The S390 can't do page protection for execute, and considers that the
 322  * same are read. Also, write permissions imply read permissions. This is
 323  * the closest we can get..
 324  */
 325          /*xwr*/
 326 #define __P000  PAGE_NONE
 327 #define __P001  PAGE_RO
 328 #define __P010  PAGE_RO
 329 #define __P011  PAGE_RO
 330 #define __P100  PAGE_RO
 331 #define __P101  PAGE_RO
 332 #define __P110  PAGE_RO
 333 #define __P111  PAGE_RO
 334
 335 #define __S000  PAGE_NONE
 336 #define __S001  PAGE_RO
 337 #define __S010  PAGE_RW
 338 #define __S011  PAGE_RW
 339 #define __S100  PAGE_RO
 340 #define __S101  PAGE_RO
 341 #define __S110  PAGE_RW
 342 #define __S111  PAGE_RW
 343
 344 /*
 345  * Certain architectures need to do special things when PTEs
 346  * within a page table are directly modified.  Thus, the following
 347  * hook is made available.
 348  */
 349 static inline void set_pte(pte_t *pteptr, pte_t pteval)
 350 {
 351         *pteptr = pteval;
 352 }
 353 #define set_pte_at(mm,addr,ptep,pteval) set_pte(ptep,pteval)
 354
 355 /*
 356  * pgd/pmd/pte query functions
 357  */
 358 #ifndef __s390x__
 359
 360 static inline int pgd_present(pgd_t pgd) { return 1; }
 361 static inline int pgd_none(pgd_t pgd)    { return 0; }
 362 static inline int pgd_bad(pgd_t pgd)     { return 0; }
 363
 364 static inline int pmd_present(pmd_t pmd) { return pmd_val(pmd) & _SEG_PRESENT; }
 365 static inline int pmd_none(pmd_t pmd)    { return pmd_val(pmd) & _PAGE_TABLE_INV; }
 366 static inline int pmd_bad(pmd_t pmd)
 367 {
 368         return (pmd_val(pmd) & (~PAGE_MASK & ~_PAGE_TABLE_INV)) != _PAGE_TABLE;
 369 }
 370
 371 #else /* __s390x__ */
 372
 373 static inline int pgd_present(pgd_t pgd)
 374 {
 375         return (pgd_val(pgd) & ~PAGE_MASK) == _PGD_ENTRY;
 376 }
 377
 378 static inline int pgd_none(pgd_t pgd)
 379 {
 380         return pgd_val(pgd) & _PGD_ENTRY_INV;
 381 }
 382
 383 static inline int pgd_bad(pgd_t pgd)
 384 {
 385         return (pgd_val(pgd) & (~PAGE_MASK & ~_PGD_ENTRY_INV)) != _PGD_ENTRY;
 386 }
 387
 388 static inline int pmd_present(pmd_t pmd)
 389 {
 390         return (pmd_val(pmd) & ~PAGE_MASK) == _PMD_ENTRY;
 391 }
 392
 393 static inline int pmd_none(pmd_t pmd)
 394 {
 395         return pmd_val(pmd) & _PMD_ENTRY_INV;
 396 }
 397
 398 static inline int pmd_bad(pmd_t pmd)
 399 {
 400         return (pmd_val(pmd) & (~PAGE_MASK & ~_PMD_ENTRY_INV)) != _PMD_ENTRY;
 401 }
 402
 403 #endif /* __s390x__ */
 404
 405 static inline int pte_none(pte_t pte)
 406 {
 407         return (pte_val(pte) & _PAGE_INVALID) && !(pte_val(pte) & _PAGE_SWT);
 408 }
 409
 410 static inline int pte_present(pte_t pte)
 411 {
 412         unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT | _PAGE_SWX;
 413         return (pte_val(pte) & mask) == _PAGE_TYPE_NONE ||
 414                 (!(pte_val(pte) & _PAGE_INVALID) &&
 415                  !(pte_val(pte) & _PAGE_SWT));
 416 }
 417
 418 static inline int pte_file(pte_t pte)
 419 {
 420         unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT;
 421         return (pte_val(pte) & mask) == _PAGE_TYPE_FILE;
 422 }
 423
 424 #define pte_same(a,b)   (pte_val(a) == pte_val(b))
 425
 426 /*
 427  * query functions pte_write/pte_dirty/pte_young only work if
 428  * pte_present() is true. Undefined behaviour if not..
 429  */
 430 static inline int pte_write(pte_t pte)
 431 {
 432         return (pte_val(pte) & _PAGE_RO) == 0;
 433 }
 434
 435 static inline int pte_dirty(pte_t pte)
 436 {
 437         /* A pte is neither clean nor dirty on s/390. The dirty bit
 438          * is in the storage key. See page_test_and_clear_dirty for
 439          * details.
 440          */
 441         return 0;
 442 }
 443
 444 static inline int pte_young(pte_t pte)
 445 {
 446         /* A pte is neither young nor old on s/390. The young bit
 447          * is in the storage key. See page_test_and_clear_young for
 448          * details.
 449          */
 450         return 0;
 451 }
 452
 453 static inline int pte_read(pte_t pte)
 454 {
 455         /* All pages are readable since we don't use the fetch
 456          * protection bit in the storage key.
 457          */
 458         return 1;
 459 }
 460
 461 /*
 462  * pgd/pmd/pte modification functions
 463  */
 464
 465 #ifndef __s390x__
 466
 467 static inline void pgd_clear(pgd_t * pgdp)      { }
 468
 469 static inline void pmd_clear(pmd_t * pmdp)
 470 {
 471         pmd_val(pmdp[0]) = _PAGE_TABLE_INV;
 472         pmd_val(pmdp[1]) = _PAGE_TABLE_INV;
 473         pmd_val(pmdp[2]) = _PAGE_TABLE_INV;
 474         pmd_val(pmdp[3]) = _PAGE_TABLE_INV;
 475 }
 476
 477 #else /* __s390x__ */
 478
 479 static inline void pgd_clear(pgd_t * pgdp)
 480 {
 481         pgd_val(*pgdp) = _PGD_ENTRY_INV | _PGD_ENTRY;
 482 }
 483
 484 static inline void pmd_clear(pmd_t * pmdp)
 485 {
 486         pmd_val(*pmdp) = _PMD_ENTRY_INV | _PMD_ENTRY;
 487         pmd_val1(*pmdp) = _PMD_ENTRY_INV | _PMD_ENTRY;
 488 }
 489
 490 #endif /* __s390x__ */
 491
 492 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
 493 {
 494         pte_val(*ptep) = _PAGE_TYPE_EMPTY;
 495 }
 496
 497 /*
 498  * The following pte modification functions only work if
 499  * pte_present() is true. Undefined behaviour if not..
 500  */
 501 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 502 {
 503         pte_val(pte) &= PAGE_MASK;
 504         pte_val(pte) |= pgprot_val(newprot);
 505         return pte;
 506 }
 507
 508 static inline pte_t pte_wrprotect(pte_t pte)
 509 {
 510         /* Do not clobber _PAGE_TYPE_NONE pages!  */
 511         if (!(pte_val(pte) & _PAGE_INVALID))
 512                 pte_val(pte) |= _PAGE_RO;
 513         return pte;
 514 }
 515
 516 static inline pte_t pte_mkwrite(pte_t pte)
 517 {
 518         pte_val(pte) &= ~_PAGE_RO;
 519         return pte;
 520 }
 521
 522 static inline pte_t pte_mkclean(pte_t pte)
 523 {
 524         /* The only user of pte_mkclean is the fork() code.
 525            We must *not* clear the *physical* page dirty bit
 526            just because fork() wants to clear the dirty bit in
 527            *one* of the page's mappings.  So we just do nothing. */
 528         return pte;
 529 }
 530
 531 static inline pte_t pte_mkdirty(pte_t pte)
 532 {
 533         /* We do not explicitly set the dirty bit because the
 534          * sske instruction is slow. It is faster to let the
 535          * next instruction set the dirty bit.
 536          */
 537         return pte;
 538 }
 539
 540 static inline pte_t pte_mkold(pte_t pte)
 541 {
 542         /* S/390 doesn't keep its dirty/referenced bit in the pte.
 543          * There is no point in clearing the real referenced bit.
 544          */
 545         return pte;
 546 }
 547
 548 static inline pte_t pte_mkyoung(pte_t pte)
 549 {
 550         /* S/390 doesn't keep its dirty/referenced bit in the pte.
 551          * There is no point in setting the real referenced bit.
 552          */
 553         return pte;
 554 }
 555
 556 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
 557 {
 558         return 0;
 559 }
 560
 561 static inline int
 562 ptep_clear_flush_young(struct vm_area_struct *vma,
 563                         unsigned long address, pte_t *ptep)
 564 {
 565         /* No need to flush TLB; bits are in storage key */
 566         return ptep_test_and_clear_young(vma, address, ptep);
 567 }
 568
 569 static inline int ptep_test_and_clear_dirty(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
 570 {
 571         return 0;
 572 }
 573
 574 static inline int
 575 ptep_clear_flush_dirty(struct vm_area_struct *vma,
 576                         unsigned long address, pte_t *ptep)
 577 {
 578         /* No need to flush TLB; bits are in storage key */
 579         return ptep_test_and_clear_dirty(vma, address, ptep);
 580 }
 581
 582 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
 583 {
 584         pte_t pte = *ptep;
 585         pte_clear(mm, addr, ptep);
 586         return pte;
 587 }
 588
 589 static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
 590 {
 591         if (!(pte_val(*ptep) & _PAGE_INVALID)) {
 592 #ifndef __s390x__
 593                 /* S390 has 1mb segments, we are emulating 4MB segments */
 594                 pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00);
 595 #else
 596                 /* ipte in zarch mode can do the math */
 597                 pte_t *pto = ptep;
 598 #endif
 599                 asm volatile(
 600                         "       ipte    %2,%3"
 601                         : "=m" (*ptep) : "m" (*ptep),
 602                           "a" (pto), "a" (address));
 603         }
 604         pte_val(*ptep) = _PAGE_TYPE_EMPTY;
 605 }
 606
 607 static inline pte_t
 608 ptep_clear_flush(struct vm_area_struct *vma,
 609                  unsigned long address, pte_t *ptep)
 610 {
 611         pte_t pte = *ptep;
 612
 613         __ptep_ipte(address, ptep);
 614         return pte;
 615 }
 616
 617 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
 618 {
 619         pte_t old_pte = *ptep;
 620         set_pte_at(mm, addr, ptep, pte_wrprotect(old_pte));
 621 }
 622
 623 static inline void
 624 ptep_establish(struct vm_area_struct *vma,
 625                unsigned long address, pte_t *ptep,
 626                pte_t entry)
 627 {
 628         ptep_clear_flush(vma, address, ptep);
 629         set_pte(ptep, entry);
 630 }
 631
 632 #define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
 633         ptep_establish(__vma, __address, __ptep, __entry)
 634
 635 /*
 636  * Test and clear dirty bit in storage key.
 637  * We can't clear the changed bit atomically. This is a potential
 638  * race against modification of the referenced bit. This function
 639  * should therefore only be called if it is not mapped in any
 640  * address space.
 641  */
 642 static inline int page_test_and_clear_dirty(struct page *page)
 643 {
 644         unsigned long physpage = page_to_phys(page);
 645         int skey = page_get_storage_key(physpage);
 646
 647         if (skey & _PAGE_CHANGED)
 648                 page_set_storage_key(physpage, skey & ~_PAGE_CHANGED);
 649         return skey & _PAGE_CHANGED;
 650 }
 651
 652 /*
 653  * Test and clear referenced bit in storage key.
 654  */
 655 static inline int page_test_and_clear_young(struct page *page)
 656 {
 657         unsigned long physpage = page_to_phys(page);
 658         int ccode;
 659
 660         asm volatile(
 661                 "       rrbe    0,%1\n"
 662                 "       ipm     %0\n"
 663                 "       srl     %0,28\n"
 664                 : "=d" (ccode) : "a" (physpage) : "cc" );
 665         return ccode & 2;
 666 }
 667
 668 /*
 669  * Conversion functions: convert a page and protection to a page entry,
 670  * and a page entry and page directory to the page they refer to.
 671  */
 672 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
 673 {
 674         pte_t __pte;
 675         pte_val(__pte) = physpage + pgprot_val(pgprot);
 676         return __pte;
 677 }
 678
 679 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
 680 {
 681         unsigned long physpage = page_to_phys(page);
 682
 683         return mk_pte_phys(physpage, pgprot);
 684 }
 685
 686 static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot)
 687 {
 688         unsigned long physpage = __pa((pfn) << PAGE_SHIFT);
 689
 690         return mk_pte_phys(physpage, pgprot);
 691 }
 692
 693 #ifdef __s390x__
 694
 695 static inline pmd_t pfn_pmd(unsigned long pfn, pgprot_t pgprot)
 696 {
 697         unsigned long physpage = __pa((pfn) << PAGE_SHIFT);
 698
 699         return __pmd(physpage + pgprot_val(pgprot));
 700 }
 701
 702 #endif /* __s390x__ */
 703
 704 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
 705 #define pte_page(x) pfn_to_page(pte_pfn(x))
 706
 707 #define pmd_page_vaddr(pmd) (pmd_val(pmd) & PAGE_MASK)
 708
 709 #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
 710
 711 #define pgd_page_vaddr(pgd) (pgd_val(pgd) & PAGE_MASK)
 712
 713 #define pgd_page(pgd) pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
 714
 715 /* to find an entry in a page-table-directory */
 716 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
 717 #define pgd_offset(mm, address) ((mm)->pgd+pgd_index(address))
 718
 719 /* to find an entry in a kernel page-table-directory */
 720 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
 721
 722 #ifndef __s390x__
 723
 724 /* Find an entry in the second-level page table.. */
 725 static inline pmd_t * pmd_offset(pgd_t * dir, unsigned long address)
 726 {
 727         return (pmd_t *) dir;
 728 }
 729
 730 #else /* __s390x__ */
 731
 732 /* Find an entry in the second-level page table.. */
 733 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
 734 #define pmd_offset(dir,addr) \
 735         ((pmd_t *) pgd_page_vaddr(*(dir)) + pmd_index(addr))
 736
 737 #endif /* __s390x__ */
 738
 739 /* Find an entry in the third-level page table.. */
 740 #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
 741 #define pte_offset_kernel(pmd, address) \
 742         ((pte_t *) pmd_page_vaddr(*(pmd)) + pte_index(address))
 743 #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
 744 #define pte_offset_map_nested(pmd, address) pte_offset_kernel(pmd, address)
 745 #define pte_unmap(pte) do { } while (0)
 746 #define pte_unmap_nested(pte) do { } while (0)
 747
 748 /*
 749  * 31 bit swap entry format:
 750  * A page-table entry has some bits we have to treat in a special way.
 751  * Bits 0, 20 and bit 23 have to be zero, otherwise an specification
 752  * exception will occur instead of a page translation exception. The
 753  * specifiation exception has the bad habit not to store necessary
 754  * information in the lowcore.
 755  * Bit 21 and bit 22 are the page invalid bit and the page protection
 756  * bit. We set both to indicate a swapped page.
 757  * Bit 30 and 31 are used to distinguish the different page types. For
 758  * a swapped page these bits need to be zero.
 759  * This leaves the bits 1-19 and bits 24-29 to store type and offset.
 760  * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19
 761  * plus 24 for the offset.
 762  * 0|     offset        |0110|o|type |00|
 763  * 0 0000000001111111111 2222 2 22222 33
 764  * 0 1234567890123456789 0123 4 56789 01
 765  *
 766  * 64 bit swap entry format:
 767  * A page-table entry has some bits we have to treat in a special way.
 768  * Bits 52 and bit 55 have to be zero, otherwise an specification
 769  * exception will occur instead of a page translation exception. The
 770  * specifiation exception has the bad habit not to store necessary
 771  * information in the lowcore.
 772  * Bit 53 and bit 54 are the page invalid bit and the page protection
 773  * bit. We set both to indicate a swapped page.
 774  * Bit 62 and 63 are used to distinguish the different page types. For
 775  * a swapped page these bits need to be zero.
 776  * This leaves the bits 0-51 and bits 56-61 to store type and offset.
 777  * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51
 778  * plus 56 for the offset.
 779  * |                      offset                        |0110|o|type |00|
 780  *  0000000000111111111122222222223333333333444444444455 5555 5 55566 66
 781  *  0123456789012345678901234567890123456789012345678901 2345 6 78901 23
 782  */
 783 #ifndef __s390x__
 784 #define __SWP_OFFSET_MASK (~0UL >> 12)
 785 #else
 786 #define __SWP_OFFSET_MASK (~0UL >> 11)
 787 #endif
 788 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
 789 {
 790         pte_t pte;
 791         offset &= __SWP_OFFSET_MASK;
 792         pte_val(pte) = _PAGE_TYPE_SWAP | ((type & 0x1f) << 2) |
 793                 ((offset & 1UL) << 7) | ((offset & ~1UL) << 11);
 794         return pte;
 795 }
 796
 797 #define __swp_type(entry)       (((entry).val >> 2) & 0x1f)
 798 #define __swp_offset(entry)     (((entry).val >> 11) | (((entry).val >> 7) & 1))
 799 #define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) })
 800
 801 #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
 802 #define __swp_entry_to_pte(x)   ((pte_t) { (x).val })
 803
 804 #ifndef __s390x__
 805 # define PTE_FILE_MAX_BITS      26
 806 #else /* __s390x__ */
 807 # define PTE_FILE_MAX_BITS      59
 808 #endif /* __s390x__ */
 809
 810 #define pte_to_pgoff(__pte) \
 811         ((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f))
 812
 813 #define pgoff_to_pte(__off) \
 814         ((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \
 815                    | _PAGE_TYPE_FILE })
 816
 817 #endif /* !__ASSEMBLY__ */
 818
 819 #define kern_addr_valid(addr)   (1)
 820
 821 extern int add_shared_memory(unsigned long start, unsigned long size);
 822 extern int remove_shared_memory(unsigned long start, unsigned long size);
 823
 824 /*
 825  * No page table caches to initialise
 826  */
 827 #define pgtable_cache_init()    do { } while (0)
 828
 829 #define __HAVE_ARCH_MEMMAP_INIT
 830 extern void memmap_init(unsigned long, int, unsigned long, unsigned long);
 831
 832 #define __HAVE_ARCH_PTEP_ESTABLISH
 833 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
 834 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
 835 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
 836 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
 837 #define __HAVE_ARCH_PTEP_CLEAR_DIRTY_FLUSH
 838 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
 839 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
 840 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
 841 #define __HAVE_ARCH_PTE_SAME
 842 #define __HAVE_ARCH_PAGE_TEST_AND_CLEAR_DIRTY
 843 #define __HAVE_ARCH_PAGE_TEST_AND_CLEAR_YOUNG
 844 #include <asm-generic/pgtable.h>
 845
 846 #endif /* _S390_PAGE_H */
 847