arch/x86/include/asm/pgtable.h

   1 #ifndef _ASM_X86_PGTABLE_H
   2 #define _ASM_X86_PGTABLE_H
   3
   4 #include <asm/page.h>
   5
   6 #include <asm/pgtable_types.h>
   7
   8 /*
   9  * Macro to mark a page protection value as UC-
  10  */
  11 #define pgprot_noncached(prot)                                  \
  12         ((boot_cpu_data.x86 > 3)                                \
  13          ? (__pgprot(pgprot_val(prot) | _PAGE_CACHE_UC_MINUS))  \
  14          : (prot))
  15
  16 #ifndef __ASSEMBLY__
  17
  18 /*
  19  * ZERO_PAGE is a global shared page that is always zero: used
  20  * for zero-mapped memory areas etc..
  21  */
  22 extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)];
  23 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
  24
  25 extern spinlock_t pgd_lock;
  26 extern struct list_head pgd_list;
  27
  28 #ifdef CONFIG_PARAVIRT
  29 #include <asm/paravirt.h>
  30 #else  /* !CONFIG_PARAVIRT */
  31 #define set_pte(ptep, pte)              native_set_pte(ptep, pte)
  32 #define set_pte_at(mm, addr, ptep, pte) native_set_pte_at(mm, addr, ptep, pte)
  33
  34 #define set_pte_present(mm, addr, ptep, pte)                            \
  35         native_set_pte_present(mm, addr, ptep, pte)
  36 #define set_pte_atomic(ptep, pte)                                       \
  37         native_set_pte_atomic(ptep, pte)
  38
  39 #define set_pmd(pmdp, pmd)              native_set_pmd(pmdp, pmd)
  40
  41 #ifndef __PAGETABLE_PUD_FOLDED
  42 #define set_pgd(pgdp, pgd)              native_set_pgd(pgdp, pgd)
  43 #define pgd_clear(pgd)                  native_pgd_clear(pgd)
  44 #endif
  45
  46 #ifndef set_pud
  47 # define set_pud(pudp, pud)             native_set_pud(pudp, pud)
  48 #endif
  49
  50 #ifndef __PAGETABLE_PMD_FOLDED
  51 #define pud_clear(pud)                  native_pud_clear(pud)
  52 #endif
  53
  54 #define pte_clear(mm, addr, ptep)       native_pte_clear(mm, addr, ptep)
  55 #define pmd_clear(pmd)                  native_pmd_clear(pmd)
  56
  57 #define pte_update(mm, addr, ptep)              do { } while (0)
  58 #define pte_update_defer(mm, addr, ptep)        do { } while (0)
  59
  60 static inline void __init paravirt_pagetable_setup_start(pgd_t *base)
  61 {
  62         native_pagetable_setup_start(base);
  63 }
  64
  65 static inline void __init paravirt_pagetable_setup_done(pgd_t *base)
  66 {
  67         native_pagetable_setup_done(base);
  68 }
  69
  70 #define pgd_val(x)      native_pgd_val(x)
  71 #define __pgd(x)        native_make_pgd(x)
  72
  73 #ifndef __PAGETABLE_PUD_FOLDED
  74 #define pud_val(x)      native_pud_val(x)
  75 #define __pud(x)        native_make_pud(x)
  76 #endif
  77
  78 #ifndef __PAGETABLE_PMD_FOLDED
  79 #define pmd_val(x)      native_pmd_val(x)
  80 #define __pmd(x)        native_make_pmd(x)
  81 #endif
  82
  83 #define pte_val(x)      native_pte_val(x)
  84 #define __pte(x)        native_make_pte(x)
  85
  86 #endif  /* CONFIG_PARAVIRT */
  87
  88 /*
  89  * The following only work if pte_present() is true.
  90  * Undefined behaviour if not..
  91  */
  92 static inline int pte_dirty(pte_t pte)
  93 {
  94         return pte_flags(pte) & _PAGE_DIRTY;
  95 }
  96
  97 static inline int pte_young(pte_t pte)
  98 {
  99         return pte_flags(pte) & _PAGE_ACCESSED;
 100 }
 101
 102 static inline int pte_write(pte_t pte)
 103 {
 104         return pte_flags(pte) & _PAGE_RW;
 105 }
 106
 107 static inline int pte_file(pte_t pte)
 108 {
 109         return pte_flags(pte) & _PAGE_FILE;
 110 }
 111
 112 static inline int pte_huge(pte_t pte)
 113 {
 114         return pte_flags(pte) & _PAGE_PSE;
 115 }
 116
 117 static inline int pte_global(pte_t pte)
 118 {
 119         return pte_flags(pte) & _PAGE_GLOBAL;
 120 }
 121
 122 static inline int pte_exec(pte_t pte)
 123 {
 124         return !(pte_flags(pte) & _PAGE_NX);
 125 }
 126
 127 static inline int pte_special(pte_t pte)
 128 {
 129         return pte_flags(pte) & _PAGE_SPECIAL;
 130 }
 131
 132 static inline unsigned long pte_pfn(pte_t pte)
 133 {
 134         return (pte_val(pte) & PTE_PFN_MASK) >> PAGE_SHIFT;
 135 }
 136
 137 #define pte_page(pte)   pfn_to_page(pte_pfn(pte))
 138
 139 static inline int pmd_large(pmd_t pte)
 140 {
 141         return (pmd_flags(pte) & (_PAGE_PSE | _PAGE_PRESENT)) ==
 142                 (_PAGE_PSE | _PAGE_PRESENT);
 143 }
 144
 145 static inline pte_t pte_set_flags(pte_t pte, pteval_t set)
 146 {
 147         pteval_t v = native_pte_val(pte);
 148
 149         return native_make_pte(v | set);
 150 }
 151
 152 static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
 153 {
 154         pteval_t v = native_pte_val(pte);
 155
 156         return native_make_pte(v & ~clear);
 157 }
 158
 159 static inline pte_t pte_mkclean(pte_t pte)
 160 {
 161         return pte_clear_flags(pte, _PAGE_DIRTY);
 162 }
 163
 164 static inline pte_t pte_mkold(pte_t pte)
 165 {
 166         return pte_clear_flags(pte, _PAGE_ACCESSED);
 167 }
 168
 169 static inline pte_t pte_wrprotect(pte_t pte)
 170 {
 171         return pte_clear_flags(pte, _PAGE_RW);
 172 }
 173
 174 static inline pte_t pte_mkexec(pte_t pte)
 175 {
 176         return pte_clear_flags(pte, _PAGE_NX);
 177 }
 178
 179 static inline pte_t pte_mkdirty(pte_t pte)
 180 {
 181         return pte_set_flags(pte, _PAGE_DIRTY);
 182 }
 183
 184 static inline pte_t pte_mkyoung(pte_t pte)
 185 {
 186         return pte_set_flags(pte, _PAGE_ACCESSED);
 187 }
 188
 189 static inline pte_t pte_mkwrite(pte_t pte)
 190 {
 191         return pte_set_flags(pte, _PAGE_RW);
 192 }
 193
 194 static inline pte_t pte_mkhuge(pte_t pte)
 195 {
 196         return pte_set_flags(pte, _PAGE_PSE);
 197 }
 198
 199 static inline pte_t pte_clrhuge(pte_t pte)
 200 {
 201         return pte_clear_flags(pte, _PAGE_PSE);
 202 }
 203
 204 static inline pte_t pte_mkglobal(pte_t pte)
 205 {
 206         return pte_set_flags(pte, _PAGE_GLOBAL);
 207 }
 208
 209 static inline pte_t pte_clrglobal(pte_t pte)
 210 {
 211         return pte_clear_flags(pte, _PAGE_GLOBAL);
 212 }
 213
 214 static inline pte_t pte_mkspecial(pte_t pte)
 215 {
 216         return pte_set_flags(pte, _PAGE_SPECIAL);
 217 }
 218
 219 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
 220 {
 221         return __pte((((phys_addr_t)page_nr << PAGE_SHIFT) |
 222                       pgprot_val(pgprot)) & __supported_pte_mask);
 223 }
 224
 225 static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
 226 {
 227         return __pmd((((phys_addr_t)page_nr << PAGE_SHIFT) |
 228                       pgprot_val(pgprot)) & __supported_pte_mask);
 229 }
 230
 231 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 232 {
 233         pteval_t val = pte_val(pte);
 234
 235         /*
 236          * Chop off the NX bit (if present), and add the NX portion of
 237          * the newprot (if present):
 238          */
 239         val &= _PAGE_CHG_MASK;
 240         val |= pgprot_val(newprot) & (~_PAGE_CHG_MASK) & __supported_pte_mask;
 241
 242         return __pte(val);
 243 }
 244
 245 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
 246 #define pgprot_modify pgprot_modify
 247 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
 248 {
 249         pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
 250         pgprotval_t addbits = pgprot_val(newprot);
 251         return __pgprot(preservebits | addbits);
 252 }
 253
 254 #define pte_pgprot(x) __pgprot(pte_flags(x) & PTE_FLAGS_MASK)
 255
 256 #define canon_pgprot(p) __pgprot(pgprot_val(p) & __supported_pte_mask)
 257
 258 static inline int is_new_memtype_allowed(unsigned long flags,
 259                                                 unsigned long new_flags)
 260 {
 261         /*
 262          * Certain new memtypes are not allowed with certain
 263          * requested memtype:
 264          * - request is uncached, return cannot be write-back
 265          * - request is write-combine, return cannot be write-back
 266          */
 267         if ((flags == _PAGE_CACHE_UC_MINUS &&
 268              new_flags == _PAGE_CACHE_WB) ||
 269             (flags == _PAGE_CACHE_WC &&
 270              new_flags == _PAGE_CACHE_WB)) {
 271                 return 0;
 272         }
 273
 274         return 1;
 275 }
 276
 277 #endif  /* __ASSEMBLY__ */
 278
 279 #ifdef CONFIG_X86_32
 280 # include "pgtable_32.h"
 281 #else
 282 # include "pgtable_64.h"
 283 #endif
 284
 285 #ifndef __ASSEMBLY__
 286 #include <linux/mm_types.h>
 287
 288 static inline int pte_none(pte_t pte)
 289 {
 290         return !pte.pte;
 291 }
 292
 293 #define __HAVE_ARCH_PTE_SAME
 294 static inline int pte_same(pte_t a, pte_t b)
 295 {
 296         return a.pte == b.pte;
 297 }
 298
 299 static inline int pte_present(pte_t a)
 300 {
 301         return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
 302 }
 303
 304 static inline int pmd_present(pmd_t pmd)
 305 {
 306         return pmd_flags(pmd) & _PAGE_PRESENT;
 307 }
 308
 309 static inline int pmd_none(pmd_t pmd)
 310 {
 311         /* Only check low word on 32-bit platforms, since it might be
 312            out of sync with upper half. */
 313         return (unsigned long)native_pmd_val(pmd) == 0;
 314 }
 315
 316 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
 317 {
 318         return (unsigned long)__va(pmd_val(pmd) & PTE_PFN_MASK);
 319 }
 320
 321 /*
 322  * Currently stuck as a macro due to indirect forward reference to
 323  * linux/mmzone.h's __section_mem_map_addr() definition:
 324  */
 325 #define pmd_page(pmd)   pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
 326
 327 /*
 328  * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
 329  *
 330  * this macro returns the index of the entry in the pmd page which would
 331  * control the given virtual address
 332  */
 333 static inline unsigned pmd_index(unsigned long address)
 334 {
 335         return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
 336 }
 337
 338 /*
 339  * Conversion functions: convert a page and protection to a page entry,
 340  * and a page entry and page directory to the page they refer to.
 341  *
 342  * (Currently stuck as a macro because of indirect forward reference
 343  * to linux/mm.h:page_to_nid())
 344  */
 345 #define mk_pte(page, pgprot)   pfn_pte(page_to_pfn(page), (pgprot))
 346
 347 /*
 348  * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
 349  *
 350  * this function returns the index of the entry in the pte page which would
 351  * control the given virtual address
 352  */
 353 static inline unsigned pte_index(unsigned long address)
 354 {
 355         return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
 356 }
 357
 358 static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
 359 {
 360         return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
 361 }
 362
 363 static inline int pmd_bad(pmd_t pmd)
 364 {
 365         return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
 366 }
 367
 368 static inline unsigned long pages_to_mb(unsigned long npg)
 369 {
 370         return npg >> (20 - PAGE_SHIFT);
 371 }
 372
 373 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
 374         remap_pfn_range(vma, vaddr, pfn, size, prot)
 375
 376 #if PAGETABLE_LEVELS == 2
 377 static inline int pud_large(pud_t pud)
 378 {
 379         return 0;
 380 }
 381 #endif
 382
 383 #if PAGETABLE_LEVELS > 2
 384 static inline int pud_none(pud_t pud)
 385 {
 386         return native_pud_val(pud) == 0;
 387 }
 388
 389 static inline int pud_present(pud_t pud)
 390 {
 391         return pud_flags(pud) & _PAGE_PRESENT;
 392 }
 393
 394 static inline unsigned long pud_page_vaddr(pud_t pud)
 395 {
 396         return (unsigned long)__va((unsigned long)pud_val(pud) & PTE_PFN_MASK);
 397 }
 398
 399 /*
 400  * Currently stuck as a macro due to indirect forward reference to
 401  * linux/mmzone.h's __section_mem_map_addr() definition:
 402  */
 403 #define pud_page(pud)           pfn_to_page(pud_val(pud) >> PAGE_SHIFT)
 404
 405 /* Find an entry in the second-level page table.. */
 406 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
 407 {
 408         return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
 409 }
 410
 411 static inline unsigned long pmd_pfn(pmd_t pmd)
 412 {
 413         return (pmd_val(pmd) & PTE_PFN_MASK) >> PAGE_SHIFT;
 414 }
 415
 416 static inline int pud_large(pud_t pud)
 417 {
 418         return (pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
 419                 (_PAGE_PSE | _PAGE_PRESENT);
 420 }
 421
 422 static inline int pud_bad(pud_t pud)
 423 {
 424         return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
 425 }
 426 #else
 427 static inline int pud_large(pud_t pud)
 428 {
 429         return 0;
 430 }
 431 #endif  /* PAGETABLE_LEVELS > 2 */
 432
 433 #if PAGETABLE_LEVELS > 3
 434 static inline int pgd_present(pgd_t pgd)
 435 {
 436         return pgd_flags(pgd) & _PAGE_PRESENT;
 437 }
 438
 439 static inline unsigned long pgd_page_vaddr(pgd_t pgd)
 440 {
 441         return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
 442 }
 443
 444 /*
 445  * Currently stuck as a macro due to indirect forward reference to
 446  * linux/mmzone.h's __section_mem_map_addr() definition:
 447  */
 448 #define pgd_page(pgd)           pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
 449
 450 /* to find an entry in a page-table-directory. */
 451 static inline unsigned pud_index(unsigned long address)
 452 {
 453         return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
 454 }
 455
 456 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
 457 {
 458         return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
 459 }
 460
 461 static inline int pgd_bad(pgd_t pgd)
 462 {
 463         return (pgd_flags(pgd) & ~_PAGE_USER) != _KERNPG_TABLE;
 464 }
 465
 466 static inline int pgd_none(pgd_t pgd)
 467 {
 468         return !native_pgd_val(pgd);
 469 }
 470 #endif  /* PAGETABLE_LEVELS > 3 */
 471
 472 #endif  /* __ASSEMBLY__ */
 473
 474 /*
 475  * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
 476  *
 477  * this macro returns the index of the entry in the pgd page which would
 478  * control the given virtual address
 479  */
 480 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
 481
 482 /*
 483  * pgd_offset() returns a (pgd_t *)
 484  * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
 485  */
 486 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
 487 /*
 488  * a shortcut which implies the use of the kernel's pgd, instead
 489  * of a process's
 490  */
 491 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
 492
 493
 494 #define KERNEL_PGD_BOUNDARY     pgd_index(PAGE_OFFSET)
 495 #define KERNEL_PGD_PTRS         (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
 496
 497 #ifndef __ASSEMBLY__
 498
 499 /* local pte updates need not use xchg for locking */
 500 static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
 501 {
 502         pte_t res = *ptep;
 503
 504         /* Pure native function needs no input for mm, addr */
 505         native_pte_clear(NULL, 0, ptep);
 506         return res;
 507 }
 508
 509 static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
 510                                      pte_t *ptep , pte_t pte)
 511 {
 512         native_set_pte(ptep, pte);
 513 }
 514
 515 #ifndef CONFIG_PARAVIRT
 516 /*
 517  * Rules for using pte_update - it must be called after any PTE update which
 518  * has not been done using the set_pte / clear_pte interfaces.  It is used by
 519  * shadow mode hypervisors to resynchronize the shadow page tables.  Kernel PTE
 520  * updates should either be sets, clears, or set_pte_atomic for P->P
 521  * transitions, which means this hook should only be called for user PTEs.
 522  * This hook implies a P->P protection or access change has taken place, which
 523  * requires a subsequent TLB flush.  The notification can optionally be delayed
 524  * until the TLB flush event by using the pte_update_defer form of the
 525  * interface, but care must be taken to assure that the flush happens while
 526  * still holding the same page table lock so that the shadow and primary pages
 527  * do not become out of sync on SMP.
 528  */
 529 #define pte_update(mm, addr, ptep)              do { } while (0)
 530 #define pte_update_defer(mm, addr, ptep)        do { } while (0)
 531 #endif
 532
 533 /*
 534  * We only update the dirty/accessed state if we set
 535  * the dirty bit by hand in the kernel, since the hardware
 536  * will do the accessed bit for us, and we don't want to
 537  * race with other CPU's that might be updating the dirty
 538  * bit at the same time.
 539  */
 540 struct vm_area_struct;
 541
 542 #define  __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
 543 extern int ptep_set_access_flags(struct vm_area_struct *vma,
 544                                  unsigned long address, pte_t *ptep,
 545                                  pte_t entry, int dirty);
 546
 547 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
 548 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
 549                                      unsigned long addr, pte_t *ptep);
 550
 551 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
 552 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
 553                                   unsigned long address, pte_t *ptep);
 554
 555 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
 556 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
 557                                        pte_t *ptep)
 558 {
 559         pte_t pte = native_ptep_get_and_clear(ptep);
 560         pte_update(mm, addr, ptep);
 561         return pte;
 562 }
 563
 564 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
 565 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
 566                                             unsigned long addr, pte_t *ptep,
 567                                             int full)
 568 {
 569         pte_t pte;
 570         if (full) {
 571                 /*
 572                  * Full address destruction in progress; paravirt does not
 573                  * care about updates and native needs no locking
 574                  */
 575                 pte = native_local_ptep_get_and_clear(ptep);
 576         } else {
 577                 pte = ptep_get_and_clear(mm, addr, ptep);
 578         }
 579         return pte;
 580 }
 581
 582 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
 583 static inline void ptep_set_wrprotect(struct mm_struct *mm,
 584                                       unsigned long addr, pte_t *ptep)
 585 {
 586         clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
 587         pte_update(mm, addr, ptep);
 588 }
 589
 590 /*
 591  * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
 592  *
 593  *  dst - pointer to pgd range anwhere on a pgd page
 594  *  src - ""
 595  *  count - the number of pgds to copy.
 596  *
 597  * dst and src can be on the same page, but the range must not overlap,
 598  * and must not cross a page boundary.
 599  */
 600 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
 601 {
 602        memcpy(dst, src, count * sizeof(pgd_t));
 603 }
 604
 605
 606 #include <asm-generic/pgtable.h>
 607 #endif  /* __ASSEMBLY__ */
 608
 609 #endif /* _ASM_X86_PGTABLE_H */