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 #define _PAGE_TYPE_EX_RO        0x202
 228 #define _PAGE_TYPE_EX_RW        0x002
 229
 230 /*
 231  * PTE type bits are rather complicated. handle_pte_fault uses pte_present,
 232  * pte_none and pte_file to find out the pte type WITHOUT holding the page
 233  * table lock. ptep_clear_flush on the other hand uses ptep_clear_flush to
 234  * invalidate a given pte. ipte sets the hw invalid bit and clears all tlbs
 235  * for the page. The page table entry is set to _PAGE_TYPE_EMPTY afterwards.
 236  * This change is done while holding the lock, but the intermediate step
 237  * of a previously valid pte with the hw invalid bit set can be observed by
 238  * handle_pte_fault. That makes it necessary that all valid pte types with
 239  * the hw invalid bit set must be distinguishable from the four pte types
 240  * empty, none, swap and file.
 241  *
 242  *                      irxt  ipte  irxt
 243  * _PAGE_TYPE_EMPTY     1000   ->   1000
 244  * _PAGE_TYPE_NONE      1001   ->   1001
 245  * _PAGE_TYPE_SWAP      1011   ->   1011
 246  * _PAGE_TYPE_FILE      11?1   ->   11?1
 247  * _PAGE_TYPE_RO        0100   ->   1100
 248  * _PAGE_TYPE_RW        0000   ->   1000
 249  * _PAGE_TYPE_EX_RO     0110   ->   1110
 250  * _PAGE_TYPE_EX_RW     0010   ->   1010
 251  *
 252  * pte_none is true for bits combinations 1000, 1010, 1100, 1110
 253  * pte_present is true for bits combinations 0000, 0010, 0100, 0110, 1001
 254  * pte_file is true for bits combinations 1101, 1111
 255  * swap pte is 1011 and 0001, 0011, 0101, 0111 are invalid.
 256  */
 257
 258 #ifndef __s390x__
 259
 260 /* Bits in the segment table entry */
 261 #define _PAGE_TABLE_LEN 0xf            /* only full page-tables            */
 262 #define _PAGE_TABLE_COM 0x10           /* common page-table                */
 263 #define _PAGE_TABLE_INV 0x20           /* invalid page-table               */
 264 #define _SEG_PRESENT    0x001          /* Software (overlap with PTL)      */
 265
 266 /* Bits int the storage key */
 267 #define _PAGE_CHANGED    0x02          /* HW changed bit                   */
 268 #define _PAGE_REFERENCED 0x04          /* HW referenced bit                */
 269
 270 #define _USER_SEG_TABLE_LEN    0x7f    /* user-segment-table up to 2 GB    */
 271 #define _KERNEL_SEG_TABLE_LEN  0x7f    /* kernel-segment-table up to 2 GB  */
 272
 273 /*
 274  * User and Kernel pagetables are identical
 275  */
 276 #define _PAGE_TABLE     _PAGE_TABLE_LEN
 277 #define _KERNPG_TABLE   _PAGE_TABLE_LEN
 278
 279 /*
 280  * The Kernel segment-tables includes the User segment-table
 281  */
 282
 283 #define _SEGMENT_TABLE  (_USER_SEG_TABLE_LEN|0x80000000|0x100)
 284 #define _KERNSEG_TABLE  _KERNEL_SEG_TABLE_LEN
 285
 286 #define USER_STD_MASK   0x00000080UL
 287
 288 #else /* __s390x__ */
 289
 290 /* Bits in the segment table entry */
 291 #define _PMD_ENTRY_INV   0x20          /* invalid segment table entry      */
 292 #define _PMD_ENTRY       0x00
 293
 294 /* Bits in the region third table entry */
 295 #define _PGD_ENTRY_INV   0x20          /* invalid region table entry       */
 296 #define _PGD_ENTRY       0x07
 297
 298 /*
 299  * User and kernel page directory
 300  */
 301 #define _REGION_THIRD       0x4
 302 #define _REGION_THIRD_LEN   0x3
 303 #define _REGION_TABLE       (_REGION_THIRD|_REGION_THIRD_LEN|0x40|0x100)
 304 #define _KERN_REGION_TABLE  (_REGION_THIRD|_REGION_THIRD_LEN)
 305
 306 #define USER_STD_MASK           0x0000000000000080UL
 307
 308 /* Bits in the storage key */
 309 #define _PAGE_CHANGED    0x02          /* HW changed bit                   */
 310 #define _PAGE_REFERENCED 0x04          /* HW referenced bit                */
 311
 312 #endif /* __s390x__ */
 313
 314 /*
 315  * Page protection definitions.
 316  */
 317 #define PAGE_NONE       __pgprot(_PAGE_TYPE_NONE)
 318 #define PAGE_RO         __pgprot(_PAGE_TYPE_RO)
 319 #define PAGE_RW         __pgprot(_PAGE_TYPE_RW)
 320 #define PAGE_EX_RO      __pgprot(_PAGE_TYPE_EX_RO)
 321 #define PAGE_EX_RW      __pgprot(_PAGE_TYPE_EX_RW)
 322
 323 #define PAGE_KERNEL     PAGE_RW
 324 #define PAGE_COPY       PAGE_RO
 325
 326 /*
 327  * Dependent on the EXEC_PROTECT option s390 can do execute protection.
 328  * Write permission always implies read permission. In theory with a
 329  * primary/secondary page table execute only can be implemented but
 330  * it would cost an additional bit in the pte to distinguish all the
 331  * different pte types. To avoid that execute permission currently
 332  * implies read permission as well.
 333  */
 334          /*xwr*/
 335 #define __P000  PAGE_NONE
 336 #define __P001  PAGE_RO
 337 #define __P010  PAGE_RO
 338 #define __P011  PAGE_RO
 339 #define __P100  PAGE_EX_RO
 340 #define __P101  PAGE_EX_RO
 341 #define __P110  PAGE_EX_RO
 342 #define __P111  PAGE_EX_RO
 343
 344 #define __S000  PAGE_NONE
 345 #define __S001  PAGE_RO
 346 #define __S010  PAGE_RW
 347 #define __S011  PAGE_RW
 348 #define __S100  PAGE_EX_RO
 349 #define __S101  PAGE_EX_RO
 350 #define __S110  PAGE_EX_RW
 351 #define __S111  PAGE_EX_RW
 352
 353 #ifndef __s390x__
 354 # define PMD_SHADOW_SHIFT       1
 355 # define PGD_SHADOW_SHIFT       1
 356 #else /* __s390x__ */
 357 # define PMD_SHADOW_SHIFT       2
 358 # define PGD_SHADOW_SHIFT       2
 359 #endif /* __s390x__ */
 360
 361 static inline struct page *get_shadow_page(struct page *page)
 362 {
 363         if (s390_noexec && !list_empty(&page->lru))
 364                 return virt_to_page(page->lru.next);
 365         return NULL;
 366 }
 367
 368 static inline pte_t *get_shadow_pte(pte_t *ptep)
 369 {
 370         unsigned long pteptr = (unsigned long) (ptep);
 371
 372         if (s390_noexec) {
 373                 unsigned long offset = pteptr & (PAGE_SIZE - 1);
 374                 void *addr = (void *) (pteptr ^ offset);
 375                 struct page *page = virt_to_page(addr);
 376                 if (!list_empty(&page->lru))
 377                         return (pte_t *) ((unsigned long) page->lru.next |
 378                                                                 offset);
 379         }
 380         return NULL;
 381 }
 382
 383 static inline pmd_t *get_shadow_pmd(pmd_t *pmdp)
 384 {
 385         unsigned long pmdptr = (unsigned long) (pmdp);
 386
 387         if (s390_noexec) {
 388                 unsigned long offset = pmdptr &
 389                                 ((PAGE_SIZE << PMD_SHADOW_SHIFT) - 1);
 390                 void *addr = (void *) (pmdptr ^ offset);
 391                 struct page *page = virt_to_page(addr);
 392                 if (!list_empty(&page->lru))
 393                         return (pmd_t *) ((unsigned long) page->lru.next |
 394                                                                 offset);
 395         }
 396         return NULL;
 397 }
 398
 399 static inline pgd_t *get_shadow_pgd(pgd_t *pgdp)
 400 {
 401         unsigned long pgdptr = (unsigned long) (pgdp);
 402
 403         if (s390_noexec) {
 404                 unsigned long offset = pgdptr &
 405                                 ((PAGE_SIZE << PGD_SHADOW_SHIFT) - 1);
 406                 void *addr = (void *) (pgdptr ^ offset);
 407                 struct page *page = virt_to_page(addr);
 408                 if (!list_empty(&page->lru))
 409                         return (pgd_t *) ((unsigned long) page->lru.next |
 410                                                                 offset);
 411         }
 412         return NULL;
 413 }
 414
 415 /*
 416  * Certain architectures need to do special things when PTEs
 417  * within a page table are directly modified.  Thus, the following
 418  * hook is made available.
 419  */
 420 static inline void set_pte(pte_t *pteptr, pte_t pteval)
 421 {
 422         pte_t *shadow_pte = get_shadow_pte(pteptr);
 423
 424         *pteptr = pteval;
 425         if (shadow_pte) {
 426                 if (!(pte_val(pteval) & _PAGE_INVALID) &&
 427                     (pte_val(pteval) & _PAGE_SWX))
 428                         pte_val(*shadow_pte) = pte_val(pteval) | _PAGE_RO;
 429                 else
 430                         pte_val(*shadow_pte) = _PAGE_TYPE_EMPTY;
 431         }
 432 }
 433 #define set_pte_at(mm,addr,ptep,pteval) set_pte(ptep,pteval)
 434
 435 /*
 436  * pgd/pmd/pte query functions
 437  */
 438 #ifndef __s390x__
 439
 440 static inline int pgd_present(pgd_t pgd) { return 1; }
 441 static inline int pgd_none(pgd_t pgd)    { return 0; }
 442 static inline int pgd_bad(pgd_t pgd)     { return 0; }
 443
 444 static inline int pmd_present(pmd_t pmd) { return pmd_val(pmd) & _SEG_PRESENT; }
 445 static inline int pmd_none(pmd_t pmd)    { return pmd_val(pmd) & _PAGE_TABLE_INV; }
 446 static inline int pmd_bad(pmd_t pmd)
 447 {
 448         return (pmd_val(pmd) & (~PAGE_MASK & ~_PAGE_TABLE_INV)) != _PAGE_TABLE;
 449 }
 450
 451 #else /* __s390x__ */
 452
 453 static inline int pgd_present(pgd_t pgd)
 454 {
 455         return (pgd_val(pgd) & ~PAGE_MASK) == _PGD_ENTRY;
 456 }
 457
 458 static inline int pgd_none(pgd_t pgd)
 459 {
 460         return pgd_val(pgd) & _PGD_ENTRY_INV;
 461 }
 462
 463 static inline int pgd_bad(pgd_t pgd)
 464 {
 465         return (pgd_val(pgd) & (~PAGE_MASK & ~_PGD_ENTRY_INV)) != _PGD_ENTRY;
 466 }
 467
 468 static inline int pmd_present(pmd_t pmd)
 469 {
 470         return (pmd_val(pmd) & ~PAGE_MASK) == _PMD_ENTRY;
 471 }
 472
 473 static inline int pmd_none(pmd_t pmd)
 474 {
 475         return pmd_val(pmd) & _PMD_ENTRY_INV;
 476 }
 477
 478 static inline int pmd_bad(pmd_t pmd)
 479 {
 480         return (pmd_val(pmd) & (~PAGE_MASK & ~_PMD_ENTRY_INV)) != _PMD_ENTRY;
 481 }
 482
 483 #endif /* __s390x__ */
 484
 485 static inline int pte_none(pte_t pte)
 486 {
 487         return (pte_val(pte) & _PAGE_INVALID) && !(pte_val(pte) & _PAGE_SWT);
 488 }
 489
 490 static inline int pte_present(pte_t pte)
 491 {
 492         unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT | _PAGE_SWX;
 493         return (pte_val(pte) & mask) == _PAGE_TYPE_NONE ||
 494                 (!(pte_val(pte) & _PAGE_INVALID) &&
 495                  !(pte_val(pte) & _PAGE_SWT));
 496 }
 497
 498 static inline int pte_file(pte_t pte)
 499 {
 500         unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT;
 501         return (pte_val(pte) & mask) == _PAGE_TYPE_FILE;
 502 }
 503
 504 #define pte_same(a,b)   (pte_val(a) == pte_val(b))
 505
 506 /*
 507  * query functions pte_write/pte_dirty/pte_young only work if
 508  * pte_present() is true. Undefined behaviour if not..
 509  */
 510 static inline int pte_write(pte_t pte)
 511 {
 512         return (pte_val(pte) & _PAGE_RO) == 0;
 513 }
 514
 515 static inline int pte_dirty(pte_t pte)
 516 {
 517         /* A pte is neither clean nor dirty on s/390. The dirty bit
 518          * is in the storage key. See page_test_and_clear_dirty for
 519          * details.
 520          */
 521         return 0;
 522 }
 523
 524 static inline int pte_young(pte_t pte)
 525 {
 526         /* A pte is neither young nor old on s/390. The young bit
 527          * is in the storage key. See page_test_and_clear_young for
 528          * details.
 529          */
 530         return 0;
 531 }
 532
 533 static inline int pte_read(pte_t pte)
 534 {
 535         /* All pages are readable since we don't use the fetch
 536          * protection bit in the storage key.
 537          */
 538         return 1;
 539 }
 540
 541 /*
 542  * pgd/pmd/pte modification functions
 543  */
 544
 545 #ifndef __s390x__
 546
 547 static inline void pgd_clear(pgd_t * pgdp)      { }
 548
 549 static inline void pmd_clear_kernel(pmd_t * pmdp)
 550 {
 551         pmd_val(pmdp[0]) = _PAGE_TABLE_INV;
 552         pmd_val(pmdp[1]) = _PAGE_TABLE_INV;
 553         pmd_val(pmdp[2]) = _PAGE_TABLE_INV;
 554         pmd_val(pmdp[3]) = _PAGE_TABLE_INV;
 555 }
 556
 557 static inline void pmd_clear(pmd_t * pmdp)
 558 {
 559         pmd_t *shadow_pmd = get_shadow_pmd(pmdp);
 560
 561         pmd_clear_kernel(pmdp);
 562         if (shadow_pmd)
 563                 pmd_clear_kernel(shadow_pmd);
 564 }
 565
 566 #else /* __s390x__ */
 567
 568 static inline void pgd_clear_kernel(pgd_t * pgdp)
 569 {
 570         pgd_val(*pgdp) = _PGD_ENTRY_INV | _PGD_ENTRY;
 571 }
 572
 573 static inline void pgd_clear(pgd_t * pgdp)
 574 {
 575         pgd_t *shadow_pgd = get_shadow_pgd(pgdp);
 576
 577         pgd_clear_kernel(pgdp);
 578         if (shadow_pgd)
 579                 pgd_clear_kernel(shadow_pgd);
 580 }
 581
 582 static inline void pmd_clear_kernel(pmd_t * pmdp)
 583 {
 584         pmd_val(*pmdp) = _PMD_ENTRY_INV | _PMD_ENTRY;
 585         pmd_val1(*pmdp) = _PMD_ENTRY_INV | _PMD_ENTRY;
 586 }
 587
 588 static inline void pmd_clear(pmd_t * pmdp)
 589 {
 590         pmd_t *shadow_pmd = get_shadow_pmd(pmdp);
 591
 592         pmd_clear_kernel(pmdp);
 593         if (shadow_pmd)
 594                 pmd_clear_kernel(shadow_pmd);
 595 }
 596
 597 #endif /* __s390x__ */
 598
 599 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
 600 {
 601         pte_t *shadow_pte = get_shadow_pte(ptep);
 602
 603         pte_val(*ptep) = _PAGE_TYPE_EMPTY;
 604         if (shadow_pte)
 605                 pte_val(*shadow_pte) = _PAGE_TYPE_EMPTY;
 606 }
 607
 608 /*
 609  * The following pte modification functions only work if
 610  * pte_present() is true. Undefined behaviour if not..
 611  */
 612 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 613 {
 614         pte_val(pte) &= PAGE_MASK;
 615         pte_val(pte) |= pgprot_val(newprot);
 616         return pte;
 617 }
 618
 619 static inline pte_t pte_wrprotect(pte_t pte)
 620 {
 621         /* Do not clobber _PAGE_TYPE_NONE pages!  */
 622         if (!(pte_val(pte) & _PAGE_INVALID))
 623                 pte_val(pte) |= _PAGE_RO;
 624         return pte;
 625 }
 626
 627 static inline pte_t pte_mkwrite(pte_t pte)
 628 {
 629         pte_val(pte) &= ~_PAGE_RO;
 630         return pte;
 631 }
 632
 633 static inline pte_t pte_mkclean(pte_t pte)
 634 {
 635         /* The only user of pte_mkclean is the fork() code.
 636            We must *not* clear the *physical* page dirty bit
 637            just because fork() wants to clear the dirty bit in
 638            *one* of the page's mappings.  So we just do nothing. */
 639         return pte;
 640 }
 641
 642 static inline pte_t pte_mkdirty(pte_t pte)
 643 {
 644         /* We do not explicitly set the dirty bit because the
 645          * sske instruction is slow. It is faster to let the
 646          * next instruction set the dirty bit.
 647          */
 648         return pte;
 649 }
 650
 651 static inline pte_t pte_mkold(pte_t pte)
 652 {
 653         /* S/390 doesn't keep its dirty/referenced bit in the pte.
 654          * There is no point in clearing the real referenced bit.
 655          */
 656         return pte;
 657 }
 658
 659 static inline pte_t pte_mkyoung(pte_t pte)
 660 {
 661         /* S/390 doesn't keep its dirty/referenced bit in the pte.
 662          * There is no point in setting the real referenced bit.
 663          */
 664         return pte;
 665 }
 666
 667 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
 668 {
 669         return 0;
 670 }
 671
 672 static inline int
 673 ptep_clear_flush_young(struct vm_area_struct *vma,
 674                         unsigned long address, pte_t *ptep)
 675 {
 676         /* No need to flush TLB; bits are in storage key */
 677         return ptep_test_and_clear_young(vma, address, ptep);
 678 }
 679
 680 static inline int ptep_test_and_clear_dirty(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
 681 {
 682         return 0;
 683 }
 684
 685 static inline int
 686 ptep_clear_flush_dirty(struct vm_area_struct *vma,
 687                         unsigned long address, pte_t *ptep)
 688 {
 689         /* No need to flush TLB; bits are in storage key */
 690         return ptep_test_and_clear_dirty(vma, address, ptep);
 691 }
 692
 693 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
 694 {
 695         pte_t pte = *ptep;
 696         pte_clear(mm, addr, ptep);
 697         return pte;
 698 }
 699
 700 static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
 701 {
 702         if (!(pte_val(*ptep) & _PAGE_INVALID)) {
 703 #ifndef __s390x__
 704                 /* S390 has 1mb segments, we are emulating 4MB segments */
 705                 pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00);
 706 #else
 707                 /* ipte in zarch mode can do the math */
 708                 pte_t *pto = ptep;
 709 #endif
 710                 asm volatile(
 711                         "       ipte    %2,%3"
 712                         : "=m" (*ptep) : "m" (*ptep),
 713                           "a" (pto), "a" (address));
 714         }
 715         pte_val(*ptep) = _PAGE_TYPE_EMPTY;
 716 }
 717
 718 static inline pte_t
 719 ptep_clear_flush(struct vm_area_struct *vma,
 720                  unsigned long address, pte_t *ptep)
 721 {
 722         pte_t pte = *ptep;
 723         pte_t *shadow_pte = get_shadow_pte(ptep);
 724
 725         __ptep_ipte(address, ptep);
 726         if (shadow_pte)
 727                 __ptep_ipte(address, shadow_pte);
 728         return pte;
 729 }
 730
 731 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
 732 {
 733         pte_t old_pte = *ptep;
 734         set_pte_at(mm, addr, ptep, pte_wrprotect(old_pte));
 735 }
 736
 737 static inline void
 738 ptep_establish(struct vm_area_struct *vma,
 739                unsigned long address, pte_t *ptep,
 740                pte_t entry)
 741 {
 742         ptep_clear_flush(vma, address, ptep);
 743         set_pte(ptep, entry);
 744 }
 745
 746 #define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
 747 ({                                                                        \
 748         int __changed = !pte_same(*(__ptep), __entry);                    \
 749         if (__changed)                                                    \
 750                 ptep_establish(__vma, __address, __ptep, __entry);        \
 751         __changed;                                                        \
 752 })
 753
 754 /*
 755  * Test and clear dirty bit in storage key.
 756  * We can't clear the changed bit atomically. This is a potential
 757  * race against modification of the referenced bit. This function
 758  * should therefore only be called if it is not mapped in any
 759  * address space.
 760  */
 761 static inline int page_test_dirty(struct page *page)
 762 {
 763         return (page_get_storage_key(page_to_phys(page)) & _PAGE_CHANGED) != 0;
 764 }
 765
 766 static inline void page_clear_dirty(struct page *page)
 767 {
 768         page_set_storage_key(page_to_phys(page), PAGE_DEFAULT_KEY);
 769 }
 770
 771 /*
 772  * Test and clear referenced bit in storage key.
 773  */
 774 static inline int page_test_and_clear_young(struct page *page)
 775 {
 776         unsigned long physpage = page_to_phys(page);
 777         int ccode;
 778
 779         asm volatile(
 780                 "       rrbe    0,%1\n"
 781                 "       ipm     %0\n"
 782                 "       srl     %0,28\n"
 783                 : "=d" (ccode) : "a" (physpage) : "cc" );
 784         return ccode & 2;
 785 }
 786
 787 /*
 788  * Conversion functions: convert a page and protection to a page entry,
 789  * and a page entry and page directory to the page they refer to.
 790  */
 791 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
 792 {
 793         pte_t __pte;
 794         pte_val(__pte) = physpage + pgprot_val(pgprot);
 795         return __pte;
 796 }
 797
 798 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
 799 {
 800         unsigned long physpage = page_to_phys(page);
 801
 802         return mk_pte_phys(physpage, pgprot);
 803 }
 804
 805 static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot)
 806 {
 807         unsigned long physpage = __pa((pfn) << PAGE_SHIFT);
 808
 809         return mk_pte_phys(physpage, pgprot);
 810 }
 811
 812 #ifdef __s390x__
 813
 814 static inline pmd_t pfn_pmd(unsigned long pfn, pgprot_t pgprot)
 815 {
 816         unsigned long physpage = __pa((pfn) << PAGE_SHIFT);
 817
 818         return __pmd(physpage + pgprot_val(pgprot));
 819 }
 820
 821 #endif /* __s390x__ */
 822
 823 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
 824 #define pte_page(x) pfn_to_page(pte_pfn(x))
 825
 826 #define pmd_page_vaddr(pmd) (pmd_val(pmd) & PAGE_MASK)
 827
 828 #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
 829
 830 #define pgd_page_vaddr(pgd) (pgd_val(pgd) & PAGE_MASK)
 831
 832 #define pgd_page(pgd) pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
 833
 834 /* to find an entry in a page-table-directory */
 835 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
 836 #define pgd_offset(mm, address) ((mm)->pgd+pgd_index(address))
 837
 838 /* to find an entry in a kernel page-table-directory */
 839 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
 840
 841 #ifndef __s390x__
 842
 843 /* Find an entry in the second-level page table.. */
 844 static inline pmd_t * pmd_offset(pgd_t * dir, unsigned long address)
 845 {
 846         return (pmd_t *) dir;
 847 }
 848
 849 #else /* __s390x__ */
 850
 851 /* Find an entry in the second-level page table.. */
 852 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
 853 #define pmd_offset(dir,addr) \
 854         ((pmd_t *) pgd_page_vaddr(*(dir)) + pmd_index(addr))
 855
 856 #endif /* __s390x__ */
 857
 858 /* Find an entry in the third-level page table.. */
 859 #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
 860 #define pte_offset_kernel(pmd, address) \
 861         ((pte_t *) pmd_page_vaddr(*(pmd)) + pte_index(address))
 862 #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
 863 #define pte_offset_map_nested(pmd, address) pte_offset_kernel(pmd, address)
 864 #define pte_unmap(pte) do { } while (0)
 865 #define pte_unmap_nested(pte) do { } while (0)
 866
 867 /*
 868  * 31 bit swap entry format:
 869  * A page-table entry has some bits we have to treat in a special way.
 870  * Bits 0, 20 and bit 23 have to be zero, otherwise an specification
 871  * exception will occur instead of a page translation exception. The
 872  * specifiation exception has the bad habit not to store necessary
 873  * information in the lowcore.
 874  * Bit 21 and bit 22 are the page invalid bit and the page protection
 875  * bit. We set both to indicate a swapped page.
 876  * Bit 30 and 31 are used to distinguish the different page types. For
 877  * a swapped page these bits need to be zero.
 878  * This leaves the bits 1-19 and bits 24-29 to store type and offset.
 879  * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19
 880  * plus 24 for the offset.
 881  * 0|     offset        |0110|o|type |00|
 882  * 0 0000000001111111111 2222 2 22222 33
 883  * 0 1234567890123456789 0123 4 56789 01
 884  *
 885  * 64 bit swap entry format:
 886  * A page-table entry has some bits we have to treat in a special way.
 887  * Bits 52 and bit 55 have to be zero, otherwise an specification
 888  * exception will occur instead of a page translation exception. The
 889  * specifiation exception has the bad habit not to store necessary
 890  * information in the lowcore.
 891  * Bit 53 and bit 54 are the page invalid bit and the page protection
 892  * bit. We set both to indicate a swapped page.
 893  * Bit 62 and 63 are used to distinguish the different page types. For
 894  * a swapped page these bits need to be zero.
 895  * This leaves the bits 0-51 and bits 56-61 to store type and offset.
 896  * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51
 897  * plus 56 for the offset.
 898  * |                      offset                        |0110|o|type |00|
 899  *  0000000000111111111122222222223333333333444444444455 5555 5 55566 66
 900  *  0123456789012345678901234567890123456789012345678901 2345 6 78901 23
 901  */
 902 #ifndef __s390x__
 903 #define __SWP_OFFSET_MASK (~0UL >> 12)
 904 #else
 905 #define __SWP_OFFSET_MASK (~0UL >> 11)
 906 #endif
 907 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
 908 {
 909         pte_t pte;
 910         offset &= __SWP_OFFSET_MASK;
 911         pte_val(pte) = _PAGE_TYPE_SWAP | ((type & 0x1f) << 2) |
 912                 ((offset & 1UL) << 7) | ((offset & ~1UL) << 11);
 913         return pte;
 914 }
 915
 916 #define __swp_type(entry)       (((entry).val >> 2) & 0x1f)
 917 #define __swp_offset(entry)     (((entry).val >> 11) | (((entry).val >> 7) & 1))
 918 #define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) })
 919
 920 #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
 921 #define __swp_entry_to_pte(x)   ((pte_t) { (x).val })
 922
 923 #ifndef __s390x__
 924 # define PTE_FILE_MAX_BITS      26
 925 #else /* __s390x__ */
 926 # define PTE_FILE_MAX_BITS      59
 927 #endif /* __s390x__ */
 928
 929 #define pte_to_pgoff(__pte) \
 930         ((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f))
 931
 932 #define pgoff_to_pte(__off) \
 933         ((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \
 934                    | _PAGE_TYPE_FILE })
 935
 936 #endif /* !__ASSEMBLY__ */
 937
 938 #define kern_addr_valid(addr)   (1)
 939
 940 extern int add_shared_memory(unsigned long start, unsigned long size);
 941 extern int remove_shared_memory(unsigned long start, unsigned long size);
 942
 943 /*
 944  * No page table caches to initialise
 945  */
 946 #define pgtable_cache_init()    do { } while (0)
 947
 948 #define __HAVE_ARCH_MEMMAP_INIT
 949 extern void memmap_init(unsigned long, int, unsigned long, unsigned long);
 950
 951 #define __HAVE_ARCH_PTEP_ESTABLISH
 952 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
 953 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
 954 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
 955 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
 956 #define __HAVE_ARCH_PTEP_CLEAR_DIRTY_FLUSH
 957 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
 958 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
 959 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
 960 #define __HAVE_ARCH_PTE_SAME
 961 #define __HAVE_ARCH_PAGE_TEST_DIRTY
 962 #define __HAVE_ARCH_PAGE_CLEAR_DIRTY
 963 #define __HAVE_ARCH_PAGE_TEST_AND_CLEAR_YOUNG
 964 #include <asm-generic/pgtable.h>
 965
 966 #endif /* _S390_PAGE_H */
 967