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  * vmalloc area starts at 4GB to prevent syscall table entry exchanging
 111  * from modules.
 112  */
 113 extern unsigned long vmalloc_end;
 114
 115 #ifdef CONFIG_64BIT
 116 #define VMALLOC_ADDR    (max(0x100000000UL, (unsigned long) high_memory))
 117 #else
 118 #define VMALLOC_ADDR    ((unsigned long) high_memory)
 119 #endif
 120 #define VMALLOC_OFFSET  (8*1024*1024)
 121 #define VMALLOC_START   ((VMALLOC_ADDR + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1))
 122 #define VMALLOC_END     vmalloc_end
 123
 124 /*
 125  * We need some free virtual space to be able to do vmalloc.
 126  * VMALLOC_MIN_SIZE defines the minimum size of the vmalloc
 127  * area. On a machine with 2GB memory we make sure that we
 128  * have at least 128MB free space for vmalloc. On a machine
 129  * with 4TB we make sure we have at least 128GB.
 130  */
 131 #ifndef __s390x__
 132 #define VMALLOC_MIN_SIZE        0x8000000UL
 133 #define VMALLOC_END_INIT        0x80000000UL
 134 #else /* __s390x__ */
 135 #define VMALLOC_MIN_SIZE        0x2000000000UL
 136 #define VMALLOC_END_INIT        0x40000000000UL
 137 #endif /* __s390x__ */
 138
 139 /*
 140  * A 31 bit pagetable entry of S390 has following format:
 141  *  |   PFRA          |    |  OS  |
 142  * 0                   0IP0
 143  * 00000000001111111111222222222233
 144  * 01234567890123456789012345678901
 145  *
 146  * I Page-Invalid Bit:    Page is not available for address-translation
 147  * P Page-Protection Bit: Store access not possible for page
 148  *
 149  * A 31 bit segmenttable entry of S390 has following format:
 150  *  |   P-table origin      |  |PTL
 151  * 0                         IC
 152  * 00000000001111111111222222222233
 153  * 01234567890123456789012345678901
 154  *
 155  * I Segment-Invalid Bit:    Segment is not available for address-translation
 156  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
 157  * PTL Page-Table-Length:    Page-table length (PTL+1*16 entries -> up to 256)
 158  *
 159  * The 31 bit segmenttable origin of S390 has following format:
 160  *
 161  *  |S-table origin   |     | STL |
 162  * X                   **GPS
 163  * 00000000001111111111222222222233
 164  * 01234567890123456789012345678901
 165  *
 166  * X Space-Switch event:
 167  * G Segment-Invalid Bit:     *
 168  * P Private-Space Bit:       Segment is not private (PoP 3-30)
 169  * S Storage-Alteration:
 170  * STL Segment-Table-Length:  Segment-table length (STL+1*16 entries -> up to 2048)
 171  *
 172  * A 64 bit pagetable entry of S390 has following format:
 173  * |                     PFRA                         |0IP0|  OS  |
 174  * 0000000000111111111122222222223333333333444444444455555555556666
 175  * 0123456789012345678901234567890123456789012345678901234567890123
 176  *
 177  * I Page-Invalid Bit:    Page is not available for address-translation
 178  * P Page-Protection Bit: Store access not possible for page
 179  *
 180  * A 64 bit segmenttable entry of S390 has following format:
 181  * |        P-table origin                              |      TT
 182  * 0000000000111111111122222222223333333333444444444455555555556666
 183  * 0123456789012345678901234567890123456789012345678901234567890123
 184  *
 185  * I Segment-Invalid Bit:    Segment is not available for address-translation
 186  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
 187  * P Page-Protection Bit: Store access not possible for page
 188  * TT Type 00
 189  *
 190  * A 64 bit region table entry of S390 has following format:
 191  * |        S-table origin                             |   TF  TTTL
 192  * 0000000000111111111122222222223333333333444444444455555555556666
 193  * 0123456789012345678901234567890123456789012345678901234567890123
 194  *
 195  * I Segment-Invalid Bit:    Segment is not available for address-translation
 196  * TT Type 01
 197  * TF
 198  * TL Table lenght
 199  *
 200  * The 64 bit regiontable origin of S390 has following format:
 201  * |      region table origon                          |       DTTL
 202  * 0000000000111111111122222222223333333333444444444455555555556666
 203  * 0123456789012345678901234567890123456789012345678901234567890123
 204  *
 205  * X Space-Switch event:
 206  * G Segment-Invalid Bit:
 207  * P Private-Space Bit:
 208  * S Storage-Alteration:
 209  * R Real space
 210  * TL Table-Length:
 211  *
 212  * A storage key has the following format:
 213  * | ACC |F|R|C|0|
 214  *  0   3 4 5 6 7
 215  * ACC: access key
 216  * F  : fetch protection bit
 217  * R  : referenced bit
 218  * C  : changed bit
 219  */
 220
 221 /* Hardware bits in the page table entry */
 222 #define _PAGE_RO        0x200           /* HW read-only bit  */
 223 #define _PAGE_INVALID   0x400           /* HW invalid bit    */
 224 #define _PAGE_SWT       0x001           /* SW pte type bit t */
 225 #define _PAGE_SWX       0x002           /* SW pte type bit x */
 226
 227 /* Six different types of pages. */
 228 #define _PAGE_TYPE_EMPTY        0x400
 229 #define _PAGE_TYPE_NONE         0x401
 230 #define _PAGE_TYPE_SWAP         0x403
 231 #define _PAGE_TYPE_FILE         0x601   /* bit 0x002 is used for offset !! */
 232 #define _PAGE_TYPE_RO           0x200
 233 #define _PAGE_TYPE_RW           0x000
 234 #define _PAGE_TYPE_EX_RO        0x202
 235 #define _PAGE_TYPE_EX_RW        0x002
 236
 237 /*
 238  * PTE type bits are rather complicated. handle_pte_fault uses pte_present,
 239  * pte_none and pte_file to find out the pte type WITHOUT holding the page
 240  * table lock. ptep_clear_flush on the other hand uses ptep_clear_flush to
 241  * invalidate a given pte. ipte sets the hw invalid bit and clears all tlbs
 242  * for the page. The page table entry is set to _PAGE_TYPE_EMPTY afterwards.
 243  * This change is done while holding the lock, but the intermediate step
 244  * of a previously valid pte with the hw invalid bit set can be observed by
 245  * handle_pte_fault. That makes it necessary that all valid pte types with
 246  * the hw invalid bit set must be distinguishable from the four pte types
 247  * empty, none, swap and file.
 248  *
 249  *                      irxt  ipte  irxt
 250  * _PAGE_TYPE_EMPTY     1000   ->   1000
 251  * _PAGE_TYPE_NONE      1001   ->   1001
 252  * _PAGE_TYPE_SWAP      1011   ->   1011
 253  * _PAGE_TYPE_FILE      11?1   ->   11?1
 254  * _PAGE_TYPE_RO        0100   ->   1100
 255  * _PAGE_TYPE_RW        0000   ->   1000
 256  * _PAGE_TYPE_EX_RO     0110   ->   1110
 257  * _PAGE_TYPE_EX_RW     0010   ->   1010
 258  *
 259  * pte_none is true for bits combinations 1000, 1010, 1100, 1110
 260  * pte_present is true for bits combinations 0000, 0010, 0100, 0110, 1001
 261  * pte_file is true for bits combinations 1101, 1111
 262  * swap pte is 1011 and 0001, 0011, 0101, 0111 are invalid.
 263  */
 264
 265 #ifndef __s390x__
 266
 267 /* Bits in the segment table entry */
 268 #define _PAGE_TABLE_LEN 0xf            /* only full page-tables            */
 269 #define _PAGE_TABLE_COM 0x10           /* common page-table                */
 270 #define _PAGE_TABLE_INV 0x20           /* invalid page-table               */
 271 #define _SEG_PRESENT    0x001          /* Software (overlap with PTL)      */
 272
 273 /* Bits int the storage key */
 274 #define _PAGE_CHANGED    0x02          /* HW changed bit                   */
 275 #define _PAGE_REFERENCED 0x04          /* HW referenced bit                */
 276
 277 #define _USER_SEG_TABLE_LEN    0x7f    /* user-segment-table up to 2 GB    */
 278 #define _KERNEL_SEG_TABLE_LEN  0x7f    /* kernel-segment-table up to 2 GB  */
 279
 280 /*
 281  * User and Kernel pagetables are identical
 282  */
 283 #define _PAGE_TABLE     _PAGE_TABLE_LEN
 284 #define _KERNPG_TABLE   _PAGE_TABLE_LEN
 285
 286 /*
 287  * The Kernel segment-tables includes the User segment-table
 288  */
 289
 290 #define _SEGMENT_TABLE  (_USER_SEG_TABLE_LEN|0x80000000|0x100)
 291 #define _KERNSEG_TABLE  _KERNEL_SEG_TABLE_LEN
 292
 293 #define USER_STD_MASK   0x00000080UL
 294
 295 #else /* __s390x__ */
 296
 297 /* Bits in the segment table entry */
 298 #define _PMD_ENTRY_INV   0x20          /* invalid segment table entry      */
 299 #define _PMD_ENTRY       0x00
 300
 301 /* Bits in the region third table entry */
 302 #define _PGD_ENTRY_INV   0x20          /* invalid region table entry       */
 303 #define _PGD_ENTRY       0x07
 304
 305 /*
 306  * User and kernel page directory
 307  */
 308 #define _REGION_THIRD       0x4
 309 #define _REGION_THIRD_LEN   0x3
 310 #define _REGION_TABLE       (_REGION_THIRD|_REGION_THIRD_LEN|0x40|0x100)
 311 #define _KERN_REGION_TABLE  (_REGION_THIRD|_REGION_THIRD_LEN)
 312
 313 #define USER_STD_MASK           0x0000000000000080UL
 314
 315 /* Bits in the storage key */
 316 #define _PAGE_CHANGED    0x02          /* HW changed bit                   */
 317 #define _PAGE_REFERENCED 0x04          /* HW referenced bit                */
 318
 319 #endif /* __s390x__ */
 320
 321 /*
 322  * Page protection definitions.
 323  */
 324 #define PAGE_NONE       __pgprot(_PAGE_TYPE_NONE)
 325 #define PAGE_RO         __pgprot(_PAGE_TYPE_RO)
 326 #define PAGE_RW         __pgprot(_PAGE_TYPE_RW)
 327 #define PAGE_EX_RO      __pgprot(_PAGE_TYPE_EX_RO)
 328 #define PAGE_EX_RW      __pgprot(_PAGE_TYPE_EX_RW)
 329
 330 #define PAGE_KERNEL     PAGE_RW
 331 #define PAGE_COPY       PAGE_RO
 332
 333 /*
 334  * Dependent on the EXEC_PROTECT option s390 can do execute protection.
 335  * Write permission always implies read permission. In theory with a
 336  * primary/secondary page table execute only can be implemented but
 337  * it would cost an additional bit in the pte to distinguish all the
 338  * different pte types. To avoid that execute permission currently
 339  * implies read permission as well.
 340  */
 341          /*xwr*/
 342 #define __P000  PAGE_NONE
 343 #define __P001  PAGE_RO
 344 #define __P010  PAGE_RO
 345 #define __P011  PAGE_RO
 346 #define __P100  PAGE_EX_RO
 347 #define __P101  PAGE_EX_RO
 348 #define __P110  PAGE_EX_RO
 349 #define __P111  PAGE_EX_RO
 350
 351 #define __S000  PAGE_NONE
 352 #define __S001  PAGE_RO
 353 #define __S010  PAGE_RW
 354 #define __S011  PAGE_RW
 355 #define __S100  PAGE_EX_RO
 356 #define __S101  PAGE_EX_RO
 357 #define __S110  PAGE_EX_RW
 358 #define __S111  PAGE_EX_RW
 359
 360 #ifndef __s390x__
 361 # define PMD_SHADOW_SHIFT       1
 362 # define PGD_SHADOW_SHIFT       1
 363 #else /* __s390x__ */
 364 # define PMD_SHADOW_SHIFT       2
 365 # define PGD_SHADOW_SHIFT       2
 366 #endif /* __s390x__ */
 367
 368 static inline struct page *get_shadow_page(struct page *page)
 369 {
 370         if (s390_noexec && !list_empty(&page->lru))
 371                 return virt_to_page(page->lru.next);
 372         return NULL;
 373 }
 374
 375 static inline pte_t *get_shadow_pte(pte_t *ptep)
 376 {
 377         unsigned long pteptr = (unsigned long) (ptep);
 378
 379         if (s390_noexec) {
 380                 unsigned long offset = pteptr & (PAGE_SIZE - 1);
 381                 void *addr = (void *) (pteptr ^ offset);
 382                 struct page *page = virt_to_page(addr);
 383                 if (!list_empty(&page->lru))
 384                         return (pte_t *) ((unsigned long) page->lru.next |
 385                                                                 offset);
 386         }
 387         return NULL;
 388 }
 389
 390 static inline pmd_t *get_shadow_pmd(pmd_t *pmdp)
 391 {
 392         unsigned long pmdptr = (unsigned long) (pmdp);
 393
 394         if (s390_noexec) {
 395                 unsigned long offset = pmdptr &
 396                                 ((PAGE_SIZE << PMD_SHADOW_SHIFT) - 1);
 397                 void *addr = (void *) (pmdptr ^ offset);
 398                 struct page *page = virt_to_page(addr);
 399                 if (!list_empty(&page->lru))
 400                         return (pmd_t *) ((unsigned long) page->lru.next |
 401                                                                 offset);
 402         }
 403         return NULL;
 404 }
 405
 406 static inline pgd_t *get_shadow_pgd(pgd_t *pgdp)
 407 {
 408         unsigned long pgdptr = (unsigned long) (pgdp);
 409
 410         if (s390_noexec) {
 411                 unsigned long offset = pgdptr &
 412                                 ((PAGE_SIZE << PGD_SHADOW_SHIFT) - 1);
 413                 void *addr = (void *) (pgdptr ^ offset);
 414                 struct page *page = virt_to_page(addr);
 415                 if (!list_empty(&page->lru))
 416                         return (pgd_t *) ((unsigned long) page->lru.next |
 417                                                                 offset);
 418         }
 419         return NULL;
 420 }
 421
 422 /*
 423  * Certain architectures need to do special things when PTEs
 424  * within a page table are directly modified.  Thus, the following
 425  * hook is made available.
 426  */
 427 static inline void set_pte(pte_t *pteptr, pte_t pteval)
 428 {
 429         pte_t *shadow_pte = get_shadow_pte(pteptr);
 430
 431         *pteptr = pteval;
 432         if (shadow_pte) {
 433                 if (!(pte_val(pteval) & _PAGE_INVALID) &&
 434                     (pte_val(pteval) & _PAGE_SWX))
 435                         pte_val(*shadow_pte) = pte_val(pteval) | _PAGE_RO;
 436                 else
 437                         pte_val(*shadow_pte) = _PAGE_TYPE_EMPTY;
 438         }
 439 }
 440 #define set_pte_at(mm,addr,ptep,pteval) set_pte(ptep,pteval)
 441
 442 /*
 443  * pgd/pmd/pte query functions
 444  */
 445 #ifndef __s390x__
 446
 447 static inline int pgd_present(pgd_t pgd) { return 1; }
 448 static inline int pgd_none(pgd_t pgd)    { return 0; }
 449 static inline int pgd_bad(pgd_t pgd)     { return 0; }
 450
 451 static inline int pmd_present(pmd_t pmd) { return pmd_val(pmd) & _SEG_PRESENT; }
 452 static inline int pmd_none(pmd_t pmd)    { return pmd_val(pmd) & _PAGE_TABLE_INV; }
 453 static inline int pmd_bad(pmd_t pmd)
 454 {
 455         return (pmd_val(pmd) & (~PAGE_MASK & ~_PAGE_TABLE_INV)) != _PAGE_TABLE;
 456 }
 457
 458 #else /* __s390x__ */
 459
 460 static inline int pgd_present(pgd_t pgd)
 461 {
 462         return (pgd_val(pgd) & ~PAGE_MASK) == _PGD_ENTRY;
 463 }
 464
 465 static inline int pgd_none(pgd_t pgd)
 466 {
 467         return pgd_val(pgd) & _PGD_ENTRY_INV;
 468 }
 469
 470 static inline int pgd_bad(pgd_t pgd)
 471 {
 472         return (pgd_val(pgd) & (~PAGE_MASK & ~_PGD_ENTRY_INV)) != _PGD_ENTRY;
 473 }
 474
 475 static inline int pmd_present(pmd_t pmd)
 476 {
 477         return (pmd_val(pmd) & ~PAGE_MASK) == _PMD_ENTRY;
 478 }
 479
 480 static inline int pmd_none(pmd_t pmd)
 481 {
 482         return pmd_val(pmd) & _PMD_ENTRY_INV;
 483 }
 484
 485 static inline int pmd_bad(pmd_t pmd)
 486 {
 487         return (pmd_val(pmd) & (~PAGE_MASK & ~_PMD_ENTRY_INV)) != _PMD_ENTRY;
 488 }
 489
 490 #endif /* __s390x__ */
 491
 492 static inline int pte_none(pte_t pte)
 493 {
 494         return (pte_val(pte) & _PAGE_INVALID) && !(pte_val(pte) & _PAGE_SWT);
 495 }
 496
 497 static inline int pte_present(pte_t pte)
 498 {
 499         unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT | _PAGE_SWX;
 500         return (pte_val(pte) & mask) == _PAGE_TYPE_NONE ||
 501                 (!(pte_val(pte) & _PAGE_INVALID) &&
 502                  !(pte_val(pte) & _PAGE_SWT));
 503 }
 504
 505 static inline int pte_file(pte_t pte)
 506 {
 507         unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT;
 508         return (pte_val(pte) & mask) == _PAGE_TYPE_FILE;
 509 }
 510
 511 #define pte_same(a,b)   (pte_val(a) == pte_val(b))
 512
 513 /*
 514  * query functions pte_write/pte_dirty/pte_young only work if
 515  * pte_present() is true. Undefined behaviour if not..
 516  */
 517 static inline int pte_write(pte_t pte)
 518 {
 519         return (pte_val(pte) & _PAGE_RO) == 0;
 520 }
 521
 522 static inline int pte_dirty(pte_t pte)
 523 {
 524         /* A pte is neither clean nor dirty on s/390. The dirty bit
 525          * is in the storage key. See page_test_and_clear_dirty for
 526          * details.
 527          */
 528         return 0;
 529 }
 530
 531 static inline int pte_young(pte_t pte)
 532 {
 533         /* A pte is neither young nor old on s/390. The young bit
 534          * is in the storage key. See page_test_and_clear_young for
 535          * details.
 536          */
 537         return 0;
 538 }
 539
 540 /*
 541  * pgd/pmd/pte modification functions
 542  */
 543
 544 #ifndef __s390x__
 545
 546 static inline void pgd_clear(pgd_t * pgdp)      { }
 547
 548 static inline void pmd_clear_kernel(pmd_t * pmdp)
 549 {
 550         pmd_val(pmdp[0]) = _PAGE_TABLE_INV;
 551         pmd_val(pmdp[1]) = _PAGE_TABLE_INV;
 552         pmd_val(pmdp[2]) = _PAGE_TABLE_INV;
 553         pmd_val(pmdp[3]) = _PAGE_TABLE_INV;
 554 }
 555
 556 static inline void pmd_clear(pmd_t * pmdp)
 557 {
 558         pmd_t *shadow_pmd = get_shadow_pmd(pmdp);
 559
 560         pmd_clear_kernel(pmdp);
 561         if (shadow_pmd)
 562                 pmd_clear_kernel(shadow_pmd);
 563 }
 564
 565 #else /* __s390x__ */
 566
 567 static inline void pgd_clear_kernel(pgd_t * pgdp)
 568 {
 569         pgd_val(*pgdp) = _PGD_ENTRY_INV | _PGD_ENTRY;
 570 }
 571
 572 static inline void pgd_clear(pgd_t * pgdp)
 573 {
 574         pgd_t *shadow_pgd = get_shadow_pgd(pgdp);
 575
 576         pgd_clear_kernel(pgdp);
 577         if (shadow_pgd)
 578                 pgd_clear_kernel(shadow_pgd);
 579 }
 580
 581 static inline void pmd_clear_kernel(pmd_t * pmdp)
 582 {
 583         pmd_val(*pmdp) = _PMD_ENTRY_INV | _PMD_ENTRY;
 584         pmd_val1(*pmdp) = _PMD_ENTRY_INV | _PMD_ENTRY;
 585 }
 586
 587 static inline void pmd_clear(pmd_t * pmdp)
 588 {
 589         pmd_t *shadow_pmd = get_shadow_pmd(pmdp);
 590
 591         pmd_clear_kernel(pmdp);
 592         if (shadow_pmd)
 593                 pmd_clear_kernel(shadow_pmd);
 594 }
 595
 596 #endif /* __s390x__ */
 597
 598 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
 599 {
 600         pte_t *shadow_pte = get_shadow_pte(ptep);
 601
 602         pte_val(*ptep) = _PAGE_TYPE_EMPTY;
 603         if (shadow_pte)
 604                 pte_val(*shadow_pte) = _PAGE_TYPE_EMPTY;
 605 }
 606
 607 /*
 608  * The following pte modification functions only work if
 609  * pte_present() is true. Undefined behaviour if not..
 610  */
 611 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 612 {
 613         pte_val(pte) &= PAGE_MASK;
 614         pte_val(pte) |= pgprot_val(newprot);
 615         return pte;
 616 }
 617
 618 static inline pte_t pte_wrprotect(pte_t pte)
 619 {
 620         /* Do not clobber _PAGE_TYPE_NONE pages!  */
 621         if (!(pte_val(pte) & _PAGE_INVALID))
 622                 pte_val(pte) |= _PAGE_RO;
 623         return pte;
 624 }
 625
 626 static inline pte_t pte_mkwrite(pte_t pte)
 627 {
 628         pte_val(pte) &= ~_PAGE_RO;
 629         return pte;
 630 }
 631
 632 static inline pte_t pte_mkclean(pte_t pte)
 633 {
 634         /* The only user of pte_mkclean is the fork() code.
 635            We must *not* clear the *physical* page dirty bit
 636            just because fork() wants to clear the dirty bit in
 637            *one* of the page's mappings.  So we just do nothing. */
 638         return pte;
 639 }
 640
 641 static inline pte_t pte_mkdirty(pte_t pte)
 642 {
 643         /* We do not explicitly set the dirty bit because the
 644          * sske instruction is slow. It is faster to let the
 645          * next instruction set the dirty bit.
 646          */
 647         return pte;
 648 }
 649
 650 static inline pte_t pte_mkold(pte_t pte)
 651 {
 652         /* S/390 doesn't keep its dirty/referenced bit in the pte.
 653          * There is no point in clearing the real referenced bit.
 654          */
 655         return pte;
 656 }
 657
 658 static inline pte_t pte_mkyoung(pte_t pte)
 659 {
 660         /* S/390 doesn't keep its dirty/referenced bit in the pte.
 661          * There is no point in setting the real referenced bit.
 662          */
 663         return pte;
 664 }
 665
 666 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
 667 {
 668         return 0;
 669 }
 670
 671 static inline int
 672 ptep_clear_flush_young(struct vm_area_struct *vma,
 673                         unsigned long address, pte_t *ptep)
 674 {
 675         /* No need to flush TLB; bits are in storage key */
 676         return ptep_test_and_clear_young(vma, address, ptep);
 677 }
 678
 679 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
 680 {
 681         pte_t pte = *ptep;
 682         pte_clear(mm, addr, ptep);
 683         return pte;
 684 }
 685
 686 static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
 687 {
 688         if (!(pte_val(*ptep) & _PAGE_INVALID)) {
 689 #ifndef __s390x__
 690                 /* S390 has 1mb segments, we are emulating 4MB segments */
 691                 pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00);
 692 #else
 693                 /* ipte in zarch mode can do the math */
 694                 pte_t *pto = ptep;
 695 #endif
 696                 asm volatile(
 697                         "       ipte    %2,%3"
 698                         : "=m" (*ptep) : "m" (*ptep),
 699                           "a" (pto), "a" (address));
 700         }
 701         pte_val(*ptep) = _PAGE_TYPE_EMPTY;
 702 }
 703
 704 static inline void ptep_invalidate(unsigned long address, pte_t *ptep)
 705 {
 706         __ptep_ipte(address, ptep);
 707         ptep = get_shadow_pte(ptep);
 708         if (ptep)
 709                 __ptep_ipte(address, ptep);
 710 }
 711
 712 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
 713                                      unsigned long address, pte_t *ptep)
 714 {
 715         pte_t pte = *ptep;
 716         ptep_invalidate(address, ptep);
 717         return pte;
 718 }
 719
 720 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
 721 {
 722         pte_t old_pte = *ptep;
 723         set_pte_at(mm, addr, ptep, pte_wrprotect(old_pte));
 724 }
 725
 726 #define ptep_set_access_flags(__vma, __addr, __ptep, __entry, __dirty)  \
 727 ({                                                                      \
 728         int __changed = !pte_same(*(__ptep), __entry);                  \
 729         if (__changed) {                                                \
 730                 ptep_invalidate(__addr, __ptep);                        \
 731                 set_pte_at((__vma)->vm_mm, __addr, __ptep, __entry);    \
 732         }                                                               \
 733         __changed;                                                      \
 734 })
 735
 736 /*
 737  * Test and clear dirty bit in storage key.
 738  * We can't clear the changed bit atomically. This is a potential
 739  * race against modification of the referenced bit. This function
 740  * should therefore only be called if it is not mapped in any
 741  * address space.
 742  */
 743 static inline int page_test_dirty(struct page *page)
 744 {
 745         return (page_get_storage_key(page_to_phys(page)) & _PAGE_CHANGED) != 0;
 746 }
 747
 748 static inline void page_clear_dirty(struct page *page)
 749 {
 750         page_set_storage_key(page_to_phys(page), PAGE_DEFAULT_KEY);
 751 }
 752
 753 /*
 754  * Test and clear referenced bit in storage key.
 755  */
 756 static inline int page_test_and_clear_young(struct page *page)
 757 {
 758         unsigned long physpage = page_to_phys(page);
 759         int ccode;
 760
 761         asm volatile(
 762                 "       rrbe    0,%1\n"
 763                 "       ipm     %0\n"
 764                 "       srl     %0,28\n"
 765                 : "=d" (ccode) : "a" (physpage) : "cc" );
 766         return ccode & 2;
 767 }
 768
 769 /*
 770  * Conversion functions: convert a page and protection to a page entry,
 771  * and a page entry and page directory to the page they refer to.
 772  */
 773 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
 774 {
 775         pte_t __pte;
 776         pte_val(__pte) = physpage + pgprot_val(pgprot);
 777         return __pte;
 778 }
 779
 780 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
 781 {
 782         unsigned long physpage = page_to_phys(page);
 783
 784         return mk_pte_phys(physpage, pgprot);
 785 }
 786
 787 static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot)
 788 {
 789         unsigned long physpage = __pa((pfn) << PAGE_SHIFT);
 790
 791         return mk_pte_phys(physpage, pgprot);
 792 }
 793
 794 #ifdef __s390x__
 795
 796 static inline pmd_t pfn_pmd(unsigned long pfn, pgprot_t pgprot)
 797 {
 798         unsigned long physpage = __pa((pfn) << PAGE_SHIFT);
 799
 800         return __pmd(physpage + pgprot_val(pgprot));
 801 }
 802
 803 #endif /* __s390x__ */
 804
 805 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
 806 #define pte_page(x) pfn_to_page(pte_pfn(x))
 807
 808 #define pmd_page_vaddr(pmd) (pmd_val(pmd) & PAGE_MASK)
 809
 810 #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
 811
 812 #define pgd_page_vaddr(pgd) (pgd_val(pgd) & PAGE_MASK)
 813
 814 #define pgd_page(pgd) pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
 815
 816 /* to find an entry in a page-table-directory */
 817 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
 818 #define pgd_offset(mm, address) ((mm)->pgd+pgd_index(address))
 819
 820 /* to find an entry in a kernel page-table-directory */
 821 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
 822
 823 #ifndef __s390x__
 824
 825 /* Find an entry in the second-level page table.. */
 826 static inline pmd_t * pmd_offset(pgd_t * dir, unsigned long address)
 827 {
 828         return (pmd_t *) dir;
 829 }
 830
 831 #else /* __s390x__ */
 832
 833 /* Find an entry in the second-level page table.. */
 834 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
 835 #define pmd_offset(dir,addr) \
 836         ((pmd_t *) pgd_page_vaddr(*(dir)) + pmd_index(addr))
 837
 838 #endif /* __s390x__ */
 839
 840 /* Find an entry in the third-level page table.. */
 841 #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
 842 #define pte_offset_kernel(pmd, address) \
 843         ((pte_t *) pmd_page_vaddr(*(pmd)) + pte_index(address))
 844 #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
 845 #define pte_offset_map_nested(pmd, address) pte_offset_kernel(pmd, address)
 846 #define pte_unmap(pte) do { } while (0)
 847 #define pte_unmap_nested(pte) do { } while (0)
 848
 849 /*
 850  * 31 bit swap entry format:
 851  * A page-table entry has some bits we have to treat in a special way.
 852  * Bits 0, 20 and bit 23 have to be zero, otherwise an specification
 853  * exception will occur instead of a page translation exception. The
 854  * specifiation exception has the bad habit not to store necessary
 855  * information in the lowcore.
 856  * Bit 21 and bit 22 are the page invalid bit and the page protection
 857  * bit. We set both to indicate a swapped page.
 858  * Bit 30 and 31 are used to distinguish the different page types. For
 859  * a swapped page these bits need to be zero.
 860  * This leaves the bits 1-19 and bits 24-29 to store type and offset.
 861  * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19
 862  * plus 24 for the offset.
 863  * 0|     offset        |0110|o|type |00|
 864  * 0 0000000001111111111 2222 2 22222 33
 865  * 0 1234567890123456789 0123 4 56789 01
 866  *
 867  * 64 bit swap entry format:
 868  * A page-table entry has some bits we have to treat in a special way.
 869  * Bits 52 and bit 55 have to be zero, otherwise an specification
 870  * exception will occur instead of a page translation exception. The
 871  * specifiation exception has the bad habit not to store necessary
 872  * information in the lowcore.
 873  * Bit 53 and bit 54 are the page invalid bit and the page protection
 874  * bit. We set both to indicate a swapped page.
 875  * Bit 62 and 63 are used to distinguish the different page types. For
 876  * a swapped page these bits need to be zero.
 877  * This leaves the bits 0-51 and bits 56-61 to store type and offset.
 878  * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51
 879  * plus 56 for the offset.
 880  * |                      offset                        |0110|o|type |00|
 881  *  0000000000111111111122222222223333333333444444444455 5555 5 55566 66
 882  *  0123456789012345678901234567890123456789012345678901 2345 6 78901 23
 883  */
 884 #ifndef __s390x__
 885 #define __SWP_OFFSET_MASK (~0UL >> 12)
 886 #else
 887 #define __SWP_OFFSET_MASK (~0UL >> 11)
 888 #endif
 889 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
 890 {
 891         pte_t pte;
 892         offset &= __SWP_OFFSET_MASK;
 893         pte_val(pte) = _PAGE_TYPE_SWAP | ((type & 0x1f) << 2) |
 894                 ((offset & 1UL) << 7) | ((offset & ~1UL) << 11);
 895         return pte;
 896 }
 897
 898 #define __swp_type(entry)       (((entry).val >> 2) & 0x1f)
 899 #define __swp_offset(entry)     (((entry).val >> 11) | (((entry).val >> 7) & 1))
 900 #define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) })
 901
 902 #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
 903 #define __swp_entry_to_pte(x)   ((pte_t) { (x).val })
 904
 905 #ifndef __s390x__
 906 # define PTE_FILE_MAX_BITS      26
 907 #else /* __s390x__ */
 908 # define PTE_FILE_MAX_BITS      59
 909 #endif /* __s390x__ */
 910
 911 #define pte_to_pgoff(__pte) \
 912         ((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f))
 913
 914 #define pgoff_to_pte(__off) \
 915         ((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \
 916                    | _PAGE_TYPE_FILE })
 917
 918 #endif /* !__ASSEMBLY__ */
 919
 920 #define kern_addr_valid(addr)   (1)
 921
 922 extern int add_shared_memory(unsigned long start, unsigned long size);
 923 extern int remove_shared_memory(unsigned long start, unsigned long size);
 924
 925 /*
 926  * No page table caches to initialise
 927  */
 928 #define pgtable_cache_init()    do { } while (0)
 929
 930 #define __HAVE_ARCH_MEMMAP_INIT
 931 extern void memmap_init(unsigned long, int, unsigned long, unsigned long);
 932
 933 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
 934 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
 935 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
 936 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
 937 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
 938 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
 939 #define __HAVE_ARCH_PTE_SAME
 940 #define __HAVE_ARCH_PAGE_TEST_DIRTY
 941 #define __HAVE_ARCH_PAGE_CLEAR_DIRTY
 942 #define __HAVE_ARCH_PAGE_TEST_AND_CLEAR_YOUNG
 943 #include <asm-generic/pgtable.h>
 944
 945 #endif /* _S390_PAGE_H */
 946