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_atomic(ptep, pte)                                       \
  35         native_set_pte_atomic(ptep, pte)
  36
  37 #define set_pmd(pmdp, pmd)              native_set_pmd(pmdp, pmd)
  38
  39 #ifndef __PAGETABLE_PUD_FOLDED
  40 #define set_pgd(pgdp, pgd)              native_set_pgd(pgdp, pgd)
  41 #define pgd_clear(pgd)                  native_pgd_clear(pgd)
  42 #endif
  43
  44 #ifndef set_pud
  45 # define set_pud(pudp, pud)             native_set_pud(pudp, pud)
  46 #endif
  47
  48 #ifndef __PAGETABLE_PMD_FOLDED
  49 #define pud_clear(pud)                  native_pud_clear(pud)
  50 #endif
  51
  52 #define pte_clear(mm, addr, ptep)       native_pte_clear(mm, addr, ptep)
  53 #define pmd_clear(pmd)                  native_pmd_clear(pmd)
  54
  55 #define pte_update(mm, addr, ptep)              do { } while (0)
  56 #define pte_update_defer(mm, addr, ptep)        do { } while (0)
  57
  58 static inline void __init paravirt_pagetable_setup_start(pgd_t *base)
  59 {
  60         native_pagetable_setup_start(base);
  61 }
  62
  63 static inline void __init paravirt_pagetable_setup_done(pgd_t *base)
  64 {
  65         native_pagetable_setup_done(base);
  66 }
  67
  68 #define pgd_val(x)      native_pgd_val(x)
  69 #define __pgd(x)        native_make_pgd(x)
  70
  71 #ifndef __PAGETABLE_PUD_FOLDED
  72 #define pud_val(x)      native_pud_val(x)
  73 #define __pud(x)        native_make_pud(x)
  74 #endif
  75
  76 #ifndef __PAGETABLE_PMD_FOLDED
  77 #define pmd_val(x)      native_pmd_val(x)
  78 #define __pmd(x)        native_make_pmd(x)
  79 #endif
  80
  81 #define pte_val(x)      native_pte_val(x)
  82 #define __pte(x)        native_make_pte(x)
  83
  84 #endif  /* CONFIG_PARAVIRT */
  85
  86 /*
  87  * The following only work if pte_present() is true.
  88  * Undefined behaviour if not..
  89  */
  90 static inline int pte_dirty(pte_t pte)
  91 {
  92         return pte_flags(pte) & _PAGE_DIRTY;
  93 }
  94
  95 static inline int pte_young(pte_t pte)
  96 {
  97         return pte_flags(pte) & _PAGE_ACCESSED;
  98 }
  99
 100 static inline int pte_write(pte_t pte)
 101 {
 102         return pte_flags(pte) & _PAGE_RW;
 103 }
 104
 105 static inline int pte_file(pte_t pte)
 106 {
 107         return pte_flags(pte) & _PAGE_FILE;
 108 }
 109
 110 static inline int pte_huge(pte_t pte)
 111 {
 112         return pte_flags(pte) & _PAGE_PSE;
 113 }
 114
 115 static inline int pte_global(pte_t pte)
 116 {
 117         return pte_flags(pte) & _PAGE_GLOBAL;
 118 }
 119
 120 static inline int pte_exec(pte_t pte)
 121 {
 122         return !(pte_flags(pte) & _PAGE_NX);
 123 }
 124
 125 static inline int pte_special(pte_t pte)
 126 {
 127         return pte_flags(pte) & _PAGE_SPECIAL;
 128 }
 129
 130 static inline unsigned long pte_pfn(pte_t pte)
 131 {
 132         return (pte_val(pte) & PTE_PFN_MASK) >> PAGE_SHIFT;
 133 }
 134
 135 #define pte_page(pte)   pfn_to_page(pte_pfn(pte))
 136
 137 static inline int pmd_large(pmd_t pte)
 138 {
 139         return (pmd_flags(pte) & (_PAGE_PSE | _PAGE_PRESENT)) ==
 140                 (_PAGE_PSE | _PAGE_PRESENT);
 141 }
 142
 143 static inline pte_t pte_set_flags(pte_t pte, pteval_t set)
 144 {
 145         pteval_t v = native_pte_val(pte);
 146
 147         return native_make_pte(v | set);
 148 }
 149
 150 static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
 151 {
 152         pteval_t v = native_pte_val(pte);
 153
 154         return native_make_pte(v & ~clear);
 155 }
 156
 157 static inline pte_t pte_mkclean(pte_t pte)
 158 {
 159         return pte_clear_flags(pte, _PAGE_DIRTY);
 160 }
 161
 162 static inline pte_t pte_mkold(pte_t pte)
 163 {
 164         return pte_clear_flags(pte, _PAGE_ACCESSED);
 165 }
 166
 167 static inline pte_t pte_wrprotect(pte_t pte)
 168 {
 169         return pte_clear_flags(pte, _PAGE_RW);
 170 }
 171
 172 static inline pte_t pte_mkexec(pte_t pte)
 173 {
 174         return pte_clear_flags(pte, _PAGE_NX);
 175 }
 176
 177 static inline pte_t pte_mkdirty(pte_t pte)
 178 {
 179         return pte_set_flags(pte, _PAGE_DIRTY);
 180 }
 181
 182 static inline pte_t pte_mkyoung(pte_t pte)
 183 {
 184         return pte_set_flags(pte, _PAGE_ACCESSED);
 185 }
 186
 187 static inline pte_t pte_mkwrite(pte_t pte)
 188 {
 189         return pte_set_flags(pte, _PAGE_RW);
 190 }
 191
 192 static inline pte_t pte_mkhuge(pte_t pte)
 193 {
 194         return pte_set_flags(pte, _PAGE_PSE);
 195 }
 196
 197 static inline pte_t pte_clrhuge(pte_t pte)
 198 {
 199         return pte_clear_flags(pte, _PAGE_PSE);
 200 }
 201
 202 static inline pte_t pte_mkglobal(pte_t pte)
 203 {
 204         return pte_set_flags(pte, _PAGE_GLOBAL);
 205 }
 206
 207 static inline pte_t pte_clrglobal(pte_t pte)
 208 {
 209         return pte_clear_flags(pte, _PAGE_GLOBAL);
 210 }
 211
 212 static inline pte_t pte_mkspecial(pte_t pte)
 213 {
 214         return pte_set_flags(pte, _PAGE_SPECIAL);
 215 }
 216
 217 /*
 218  * Mask out unsupported bits in a present pgprot.  Non-present pgprots
 219  * can use those bits for other purposes, so leave them be.
 220  */
 221 static inline pgprotval_t massage_pgprot(pgprot_t pgprot)
 222 {
 223         pgprotval_t protval = pgprot_val(pgprot);
 224
 225         if (protval & _PAGE_PRESENT)
 226                 protval &= __supported_pte_mask;
 227
 228         return protval;
 229 }
 230
 231 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
 232 {
 233         return __pte(((phys_addr_t)page_nr << PAGE_SHIFT) |
 234                      massage_pgprot(pgprot));
 235 }
 236
 237 static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
 238 {
 239         return __pmd(((phys_addr_t)page_nr << PAGE_SHIFT) |
 240                      massage_pgprot(pgprot));
 241 }
 242
 243 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 244 {
 245         pteval_t val = pte_val(pte);
 246
 247         /*
 248          * Chop off the NX bit (if present), and add the NX portion of
 249          * the newprot (if present):
 250          */
 251         val &= _PAGE_CHG_MASK;
 252         val |= massage_pgprot(newprot) & ~_PAGE_CHG_MASK;
 253
 254         return __pte(val);
 255 }
 256
 257 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
 258 #define pgprot_modify pgprot_modify
 259 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
 260 {
 261         pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
 262         pgprotval_t addbits = pgprot_val(newprot);
 263         return __pgprot(preservebits | addbits);
 264 }
 265
 266 #define pte_pgprot(x) __pgprot(pte_flags(x) & PTE_FLAGS_MASK)
 267
 268 #define canon_pgprot(p) __pgprot(massage_pgprot(p))
 269
 270 static inline int is_new_memtype_allowed(unsigned long flags,
 271                                                 unsigned long new_flags)
 272 {
 273         /*
 274          * Certain new memtypes are not allowed with certain
 275          * requested memtype:
 276          * - request is uncached, return cannot be write-back
 277          * - request is write-combine, return cannot be write-back
 278          */
 279         if ((flags == _PAGE_CACHE_UC_MINUS &&
 280              new_flags == _PAGE_CACHE_WB) ||
 281             (flags == _PAGE_CACHE_WC &&
 282              new_flags == _PAGE_CACHE_WB)) {
 283                 return 0;
 284         }
 285
 286         return 1;
 287 }
 288
 289 pmd_t *populate_extra_pmd(unsigned long vaddr);
 290 pte_t *populate_extra_pte(unsigned long vaddr);
 291 #endif  /* __ASSEMBLY__ */
 292
 293 #ifdef CONFIG_X86_32
 294 # include "pgtable_32.h"
 295 #else
 296 # include "pgtable_64.h"
 297 #endif
 298
 299 #ifndef __ASSEMBLY__
 300 #include <linux/mm_types.h>
 301
 302 static inline int pte_none(pte_t pte)
 303 {
 304         return !pte.pte;
 305 }
 306
 307 #define __HAVE_ARCH_PTE_SAME
 308 static inline int pte_same(pte_t a, pte_t b)
 309 {
 310         return a.pte == b.pte;
 311 }
 312
 313 static inline int pte_present(pte_t a)
 314 {
 315         return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
 316 }
 317
 318 static inline int pmd_present(pmd_t pmd)
 319 {
 320         return pmd_flags(pmd) & _PAGE_PRESENT;
 321 }
 322
 323 static inline int pmd_none(pmd_t pmd)
 324 {
 325         /* Only check low word on 32-bit platforms, since it might be
 326            out of sync with upper half. */
 327         return (unsigned long)native_pmd_val(pmd) == 0;
 328 }
 329
 330 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
 331 {
 332         return (unsigned long)__va(pmd_val(pmd) & PTE_PFN_MASK);
 333 }
 334
 335 /*
 336  * Currently stuck as a macro due to indirect forward reference to
 337  * linux/mmzone.h's __section_mem_map_addr() definition:
 338  */
 339 #define pmd_page(pmd)   pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
 340
 341 /*
 342  * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
 343  *
 344  * this macro returns the index of the entry in the pmd page which would
 345  * control the given virtual address
 346  */
 347 static inline unsigned pmd_index(unsigned long address)
 348 {
 349         return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
 350 }
 351
 352 /*
 353  * Conversion functions: convert a page and protection to a page entry,
 354  * and a page entry and page directory to the page they refer to.
 355  *
 356  * (Currently stuck as a macro because of indirect forward reference
 357  * to linux/mm.h:page_to_nid())
 358  */
 359 #define mk_pte(page, pgprot)   pfn_pte(page_to_pfn(page), (pgprot))
 360
 361 /*
 362  * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
 363  *
 364  * this function returns the index of the entry in the pte page which would
 365  * control the given virtual address
 366  */
 367 static inline unsigned pte_index(unsigned long address)
 368 {
 369         return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
 370 }
 371
 372 static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
 373 {
 374         return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
 375 }
 376
 377 static inline int pmd_bad(pmd_t pmd)
 378 {
 379         return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
 380 }
 381
 382 static inline unsigned long pages_to_mb(unsigned long npg)
 383 {
 384         return npg >> (20 - PAGE_SHIFT);
 385 }
 386
 387 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
 388         remap_pfn_range(vma, vaddr, pfn, size, prot)
 389
 390 #if PAGETABLE_LEVELS > 2
 391 static inline int pud_none(pud_t pud)
 392 {
 393         return native_pud_val(pud) == 0;
 394 }
 395
 396 static inline int pud_present(pud_t pud)
 397 {
 398         return pud_flags(pud) & _PAGE_PRESENT;
 399 }
 400
 401 static inline unsigned long pud_page_vaddr(pud_t pud)
 402 {
 403         return (unsigned long)__va((unsigned long)pud_val(pud) & PTE_PFN_MASK);
 404 }
 405
 406 /*
 407  * Currently stuck as a macro due to indirect forward reference to
 408  * linux/mmzone.h's __section_mem_map_addr() definition:
 409  */
 410 #define pud_page(pud)           pfn_to_page(pud_val(pud) >> PAGE_SHIFT)
 411
 412 /* Find an entry in the second-level page table.. */
 413 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
 414 {
 415         return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
 416 }
 417
 418 static inline unsigned long pmd_pfn(pmd_t pmd)
 419 {
 420         return (pmd_val(pmd) & PTE_PFN_MASK) >> PAGE_SHIFT;
 421 }
 422
 423 static inline int pud_large(pud_t pud)
 424 {
 425         return (pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
 426                 (_PAGE_PSE | _PAGE_PRESENT);
 427 }
 428
 429 static inline int pud_bad(pud_t pud)
 430 {
 431         return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
 432 }
 433 #else
 434 static inline int pud_large(pud_t pud)
 435 {
 436         return 0;
 437 }
 438 #endif  /* PAGETABLE_LEVELS > 2 */
 439
 440 #if PAGETABLE_LEVELS > 3
 441 static inline int pgd_present(pgd_t pgd)
 442 {
 443         return pgd_flags(pgd) & _PAGE_PRESENT;
 444 }
 445
 446 static inline unsigned long pgd_page_vaddr(pgd_t pgd)
 447 {
 448         return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
 449 }
 450
 451 /*
 452  * Currently stuck as a macro due to indirect forward reference to
 453  * linux/mmzone.h's __section_mem_map_addr() definition:
 454  */
 455 #define pgd_page(pgd)           pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
 456
 457 /* to find an entry in a page-table-directory. */
 458 static inline unsigned pud_index(unsigned long address)
 459 {
 460         return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
 461 }
 462
 463 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
 464 {
 465         return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
 466 }
 467
 468 static inline int pgd_bad(pgd_t pgd)
 469 {
 470         return (pgd_flags(pgd) & ~_PAGE_USER) != _KERNPG_TABLE;
 471 }
 472
 473 static inline int pgd_none(pgd_t pgd)
 474 {
 475         return !native_pgd_val(pgd);
 476 }
 477 #endif  /* PAGETABLE_LEVELS > 3 */
 478
 479 #endif  /* __ASSEMBLY__ */
 480
 481 /*
 482  * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
 483  *
 484  * this macro returns the index of the entry in the pgd page which would
 485  * control the given virtual address
 486  */
 487 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
 488
 489 /*
 490  * pgd_offset() returns a (pgd_t *)
 491  * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
 492  */
 493 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
 494 /*
 495  * a shortcut which implies the use of the kernel's pgd, instead
 496  * of a process's
 497  */
 498 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
 499
 500
 501 #define KERNEL_PGD_BOUNDARY     pgd_index(PAGE_OFFSET)
 502 #define KERNEL_PGD_PTRS         (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
 503
 504 #ifndef __ASSEMBLY__
 505
 506 /* local pte updates need not use xchg for locking */
 507 static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
 508 {
 509         pte_t res = *ptep;
 510
 511         /* Pure native function needs no input for mm, addr */
 512         native_pte_clear(NULL, 0, ptep);
 513         return res;
 514 }
 515
 516 static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
 517                                      pte_t *ptep , pte_t pte)
 518 {
 519         native_set_pte(ptep, pte);
 520 }
 521
 522 #ifndef CONFIG_PARAVIRT
 523 /*
 524  * Rules for using pte_update - it must be called after any PTE update which
 525  * has not been done using the set_pte / clear_pte interfaces.  It is used by
 526  * shadow mode hypervisors to resynchronize the shadow page tables.  Kernel PTE
 527  * updates should either be sets, clears, or set_pte_atomic for P->P
 528  * transitions, which means this hook should only be called for user PTEs.
 529  * This hook implies a P->P protection or access change has taken place, which
 530  * requires a subsequent TLB flush.  The notification can optionally be delayed
 531  * until the TLB flush event by using the pte_update_defer form of the
 532  * interface, but care must be taken to assure that the flush happens while
 533  * still holding the same page table lock so that the shadow and primary pages
 534  * do not become out of sync on SMP.
 535  */
 536 #define pte_update(mm, addr, ptep)              do { } while (0)
 537 #define pte_update_defer(mm, addr, ptep)        do { } while (0)
 538 #endif
 539
 540 /*
 541  * We only update the dirty/accessed state if we set
 542  * the dirty bit by hand in the kernel, since the hardware
 543  * will do the accessed bit for us, and we don't want to
 544  * race with other CPU's that might be updating the dirty
 545  * bit at the same time.
 546  */
 547 struct vm_area_struct;
 548
 549 #define  __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
 550 extern int ptep_set_access_flags(struct vm_area_struct *vma,
 551                                  unsigned long address, pte_t *ptep,
 552                                  pte_t entry, int dirty);
 553
 554 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
 555 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
 556                                      unsigned long addr, pte_t *ptep);
 557
 558 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
 559 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
 560                                   unsigned long address, pte_t *ptep);
 561
 562 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
 563 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
 564                                        pte_t *ptep)
 565 {
 566         pte_t pte = native_ptep_get_and_clear(ptep);
 567         pte_update(mm, addr, ptep);
 568         return pte;
 569 }
 570
 571 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
 572 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
 573                                             unsigned long addr, pte_t *ptep,
 574                                             int full)
 575 {
 576         pte_t pte;
 577         if (full) {
 578                 /*
 579                  * Full address destruction in progress; paravirt does not
 580                  * care about updates and native needs no locking
 581                  */
 582                 pte = native_local_ptep_get_and_clear(ptep);
 583         } else {
 584                 pte = ptep_get_and_clear(mm, addr, ptep);
 585         }
 586         return pte;
 587 }
 588
 589 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
 590 static inline void ptep_set_wrprotect(struct mm_struct *mm,
 591                                       unsigned long addr, pte_t *ptep)
 592 {
 593         clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
 594         pte_update(mm, addr, ptep);
 595 }
 596
 597 /*
 598  * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
 599  *
 600  *  dst - pointer to pgd range anwhere on a pgd page
 601  *  src - ""
 602  *  count - the number of pgds to copy.
 603  *
 604  * dst and src can be on the same page, but the range must not overlap,
 605  * and must not cross a page boundary.
 606  */
 607 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
 608 {
 609        memcpy(dst, src, count * sizeof(pgd_t));
 610 }
 611
 612
 613 #include <asm-generic/pgtable.h>
 614 #endif  /* __ASSEMBLY__ */
 615
 616 #endif /* _ASM_X86_PGTABLE_H */