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 #include <asm/e820.h>
   6
   7 #include <asm/pgtable_types.h>
   8
   9 /*
  10  * Macro to mark a page protection value as UC-
  11  */
  12 #define pgprot_noncached(prot)                                          \
  13         ((boot_cpu_data.x86 > 3)                                        \
  14          ? (__pgprot(pgprot_val(prot) |                                 \
  15                      cachemode2protval(_PAGE_CACHE_MODE_UC_MINUS)))     \
  16          : (prot))
  17
  18 #ifndef __ASSEMBLY__
  19 #include <asm/x86_init.h>
  20
  21 void ptdump_walk_pgd_level(struct seq_file *m, pgd_t *pgd);
  22
  23 /*
  24  * ZERO_PAGE is a global shared page that is always zero: used
  25  * for zero-mapped memory areas etc..
  26  */
  27 extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)]
  28         __visible;
  29 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
  30
  31 extern spinlock_t pgd_lock;
  32 extern struct list_head pgd_list;
  33
  34 extern struct mm_struct *pgd_page_get_mm(struct page *page);
  35
  36 #ifdef CONFIG_PARAVIRT
  37 #include <asm/paravirt.h>
  38 #else  /* !CONFIG_PARAVIRT */
  39 #define set_pte(ptep, pte)              native_set_pte(ptep, pte)
  40 #define set_pte_at(mm, addr, ptep, pte) native_set_pte_at(mm, addr, ptep, pte)
  41 #define set_pmd_at(mm, addr, pmdp, pmd) native_set_pmd_at(mm, addr, pmdp, pmd)
  42
  43 #define set_pte_atomic(ptep, pte)                                       \
  44         native_set_pte_atomic(ptep, pte)
  45
  46 #define set_pmd(pmdp, pmd)              native_set_pmd(pmdp, pmd)
  47
  48 #ifndef __PAGETABLE_PUD_FOLDED
  49 #define set_pgd(pgdp, pgd)              native_set_pgd(pgdp, pgd)
  50 #define pgd_clear(pgd)                  native_pgd_clear(pgd)
  51 #endif
  52
  53 #ifndef set_pud
  54 # define set_pud(pudp, pud)             native_set_pud(pudp, pud)
  55 #endif
  56
  57 #ifndef __PAGETABLE_PMD_FOLDED
  58 #define pud_clear(pud)                  native_pud_clear(pud)
  59 #endif
  60
  61 #define pte_clear(mm, addr, ptep)       native_pte_clear(mm, addr, ptep)
  62 #define pmd_clear(pmd)                  native_pmd_clear(pmd)
  63
  64 #define pte_update(mm, addr, ptep)              do { } while (0)
  65 #define pte_update_defer(mm, addr, ptep)        do { } while (0)
  66 #define pmd_update(mm, addr, ptep)              do { } while (0)
  67 #define pmd_update_defer(mm, addr, ptep)        do { } while (0)
  68
  69 #define pgd_val(x)      native_pgd_val(x)
  70 #define __pgd(x)        native_make_pgd(x)
  71
  72 #ifndef __PAGETABLE_PUD_FOLDED
  73 #define pud_val(x)      native_pud_val(x)
  74 #define __pud(x)        native_make_pud(x)
  75 #endif
  76
  77 #ifndef __PAGETABLE_PMD_FOLDED
  78 #define pmd_val(x)      native_pmd_val(x)
  79 #define __pmd(x)        native_make_pmd(x)
  80 #endif
  81
  82 #define pte_val(x)      native_pte_val(x)
  83 #define __pte(x)        native_make_pte(x)
  84
  85 #define arch_end_context_switch(prev)   do {} while(0)
  86
  87 #endif  /* CONFIG_PARAVIRT */
  88
  89 /*
  90  * The following only work if pte_present() is true.
  91  * Undefined behaviour if not..
  92  */
  93 static inline int pte_dirty(pte_t pte)
  94 {
  95         return pte_flags(pte) & _PAGE_DIRTY;
  96 }
  97
  98 static inline int pte_young(pte_t pte)
  99 {
 100         return pte_flags(pte) & _PAGE_ACCESSED;
 101 }
 102
 103 static inline int pmd_dirty(pmd_t pmd)
 104 {
 105         return pmd_flags(pmd) & _PAGE_DIRTY;
 106 }
 107
 108 static inline int pmd_young(pmd_t pmd)
 109 {
 110         return pmd_flags(pmd) & _PAGE_ACCESSED;
 111 }
 112
 113 static inline int pte_write(pte_t pte)
 114 {
 115         return pte_flags(pte) & _PAGE_RW;
 116 }
 117
 118 static inline int pte_huge(pte_t pte)
 119 {
 120         return pte_flags(pte) & _PAGE_PSE;
 121 }
 122
 123 static inline int pte_global(pte_t pte)
 124 {
 125         return pte_flags(pte) & _PAGE_GLOBAL;
 126 }
 127
 128 static inline int pte_exec(pte_t pte)
 129 {
 130         return !(pte_flags(pte) & _PAGE_NX);
 131 }
 132
 133 static inline int pte_special(pte_t pte)
 134 {
 135         return pte_flags(pte) & _PAGE_SPECIAL;
 136 }
 137
 138 static inline unsigned long pte_pfn(pte_t pte)
 139 {
 140         return (pte_val(pte) & PTE_PFN_MASK) >> PAGE_SHIFT;
 141 }
 142
 143 static inline unsigned long pmd_pfn(pmd_t pmd)
 144 {
 145         return (pmd_val(pmd) & PTE_PFN_MASK) >> PAGE_SHIFT;
 146 }
 147
 148 static inline unsigned long pud_pfn(pud_t pud)
 149 {
 150         return (pud_val(pud) & PTE_PFN_MASK) >> PAGE_SHIFT;
 151 }
 152
 153 #define pte_page(pte)   pfn_to_page(pte_pfn(pte))
 154
 155 static inline int pmd_large(pmd_t pte)
 156 {
 157         return pmd_flags(pte) & _PAGE_PSE;
 158 }
 159
 160 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 161 static inline int pmd_trans_splitting(pmd_t pmd)
 162 {
 163         return pmd_val(pmd) & _PAGE_SPLITTING;
 164 }
 165
 166 static inline int pmd_trans_huge(pmd_t pmd)
 167 {
 168         return pmd_val(pmd) & _PAGE_PSE;
 169 }
 170
 171 static inline int has_transparent_hugepage(void)
 172 {
 173         return cpu_has_pse;
 174 }
 175 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
 176
 177 static inline pte_t pte_set_flags(pte_t pte, pteval_t set)
 178 {
 179         pteval_t v = native_pte_val(pte);
 180
 181         return native_make_pte(v | set);
 182 }
 183
 184 static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
 185 {
 186         pteval_t v = native_pte_val(pte);
 187
 188         return native_make_pte(v & ~clear);
 189 }
 190
 191 static inline pte_t pte_mkclean(pte_t pte)
 192 {
 193         return pte_clear_flags(pte, _PAGE_DIRTY);
 194 }
 195
 196 static inline pte_t pte_mkold(pte_t pte)
 197 {
 198         return pte_clear_flags(pte, _PAGE_ACCESSED);
 199 }
 200
 201 static inline pte_t pte_wrprotect(pte_t pte)
 202 {
 203         return pte_clear_flags(pte, _PAGE_RW);
 204 }
 205
 206 static inline pte_t pte_mkexec(pte_t pte)
 207 {
 208         return pte_clear_flags(pte, _PAGE_NX);
 209 }
 210
 211 static inline pte_t pte_mkdirty(pte_t pte)
 212 {
 213         return pte_set_flags(pte, _PAGE_DIRTY | _PAGE_SOFT_DIRTY);
 214 }
 215
 216 static inline pte_t pte_mkyoung(pte_t pte)
 217 {
 218         return pte_set_flags(pte, _PAGE_ACCESSED);
 219 }
 220
 221 static inline pte_t pte_mkwrite(pte_t pte)
 222 {
 223         return pte_set_flags(pte, _PAGE_RW);
 224 }
 225
 226 static inline pte_t pte_mkhuge(pte_t pte)
 227 {
 228         return pte_set_flags(pte, _PAGE_PSE);
 229 }
 230
 231 static inline pte_t pte_clrhuge(pte_t pte)
 232 {
 233         return pte_clear_flags(pte, _PAGE_PSE);
 234 }
 235
 236 static inline pte_t pte_mkglobal(pte_t pte)
 237 {
 238         return pte_set_flags(pte, _PAGE_GLOBAL);
 239 }
 240
 241 static inline pte_t pte_clrglobal(pte_t pte)
 242 {
 243         return pte_clear_flags(pte, _PAGE_GLOBAL);
 244 }
 245
 246 static inline pte_t pte_mkspecial(pte_t pte)
 247 {
 248         return pte_set_flags(pte, _PAGE_SPECIAL);
 249 }
 250
 251 static inline pmd_t pmd_set_flags(pmd_t pmd, pmdval_t set)
 252 {
 253         pmdval_t v = native_pmd_val(pmd);
 254
 255         return __pmd(v | set);
 256 }
 257
 258 static inline pmd_t pmd_clear_flags(pmd_t pmd, pmdval_t clear)
 259 {
 260         pmdval_t v = native_pmd_val(pmd);
 261
 262         return __pmd(v & ~clear);
 263 }
 264
 265 static inline pmd_t pmd_mkold(pmd_t pmd)
 266 {
 267         return pmd_clear_flags(pmd, _PAGE_ACCESSED);
 268 }
 269
 270 static inline pmd_t pmd_wrprotect(pmd_t pmd)
 271 {
 272         return pmd_clear_flags(pmd, _PAGE_RW);
 273 }
 274
 275 static inline pmd_t pmd_mkdirty(pmd_t pmd)
 276 {
 277         return pmd_set_flags(pmd, _PAGE_DIRTY | _PAGE_SOFT_DIRTY);
 278 }
 279
 280 static inline pmd_t pmd_mkhuge(pmd_t pmd)
 281 {
 282         return pmd_set_flags(pmd, _PAGE_PSE);
 283 }
 284
 285 static inline pmd_t pmd_mkyoung(pmd_t pmd)
 286 {
 287         return pmd_set_flags(pmd, _PAGE_ACCESSED);
 288 }
 289
 290 static inline pmd_t pmd_mkwrite(pmd_t pmd)
 291 {
 292         return pmd_set_flags(pmd, _PAGE_RW);
 293 }
 294
 295 static inline pmd_t pmd_mknotpresent(pmd_t pmd)
 296 {
 297         return pmd_clear_flags(pmd, _PAGE_PRESENT | _PAGE_PROTNONE);
 298 }
 299
 300 #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY
 301 static inline int pte_soft_dirty(pte_t pte)
 302 {
 303         return pte_flags(pte) & _PAGE_SOFT_DIRTY;
 304 }
 305
 306 static inline int pmd_soft_dirty(pmd_t pmd)
 307 {
 308         return pmd_flags(pmd) & _PAGE_SOFT_DIRTY;
 309 }
 310
 311 static inline pte_t pte_mksoft_dirty(pte_t pte)
 312 {
 313         return pte_set_flags(pte, _PAGE_SOFT_DIRTY);
 314 }
 315
 316 static inline pmd_t pmd_mksoft_dirty(pmd_t pmd)
 317 {
 318         return pmd_set_flags(pmd, _PAGE_SOFT_DIRTY);
 319 }
 320
 321 #endif /* CONFIG_HAVE_ARCH_SOFT_DIRTY */
 322
 323 /*
 324  * Mask out unsupported bits in a present pgprot.  Non-present pgprots
 325  * can use those bits for other purposes, so leave them be.
 326  */
 327 static inline pgprotval_t massage_pgprot(pgprot_t pgprot)
 328 {
 329         pgprotval_t protval = pgprot_val(pgprot);
 330
 331         if (protval & _PAGE_PRESENT)
 332                 protval &= __supported_pte_mask;
 333
 334         return protval;
 335 }
 336
 337 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
 338 {
 339         return __pte(((phys_addr_t)page_nr << PAGE_SHIFT) |
 340                      massage_pgprot(pgprot));
 341 }
 342
 343 static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
 344 {
 345         return __pmd(((phys_addr_t)page_nr << PAGE_SHIFT) |
 346                      massage_pgprot(pgprot));
 347 }
 348
 349 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 350 {
 351         pteval_t val = pte_val(pte);
 352
 353         /*
 354          * Chop off the NX bit (if present), and add the NX portion of
 355          * the newprot (if present):
 356          */
 357         val &= _PAGE_CHG_MASK;
 358         val |= massage_pgprot(newprot) & ~_PAGE_CHG_MASK;
 359
 360         return __pte(val);
 361 }
 362
 363 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
 364 {
 365         pmdval_t val = pmd_val(pmd);
 366
 367         val &= _HPAGE_CHG_MASK;
 368         val |= massage_pgprot(newprot) & ~_HPAGE_CHG_MASK;
 369
 370         return __pmd(val);
 371 }
 372
 373 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
 374 #define pgprot_modify pgprot_modify
 375 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
 376 {
 377         pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
 378         pgprotval_t addbits = pgprot_val(newprot);
 379         return __pgprot(preservebits | addbits);
 380 }
 381
 382 #define pte_pgprot(x) __pgprot(pte_flags(x) & PTE_FLAGS_MASK)
 383
 384 #define canon_pgprot(p) __pgprot(massage_pgprot(p))
 385
 386 static inline int is_new_memtype_allowed(u64 paddr, unsigned long size,
 387                                          enum page_cache_mode pcm,
 388                                          enum page_cache_mode new_pcm)
 389 {
 390         /*
 391          * PAT type is always WB for untracked ranges, so no need to check.
 392          */
 393         if (x86_platform.is_untracked_pat_range(paddr, paddr + size))
 394                 return 1;
 395
 396         /*
 397          * Certain new memtypes are not allowed with certain
 398          * requested memtype:
 399          * - request is uncached, return cannot be write-back
 400          * - request is write-combine, return cannot be write-back
 401          */
 402         if ((pcm == _PAGE_CACHE_MODE_UC_MINUS &&
 403              new_pcm == _PAGE_CACHE_MODE_WB) ||
 404             (pcm == _PAGE_CACHE_MODE_WC &&
 405              new_pcm == _PAGE_CACHE_MODE_WB)) {
 406                 return 0;
 407         }
 408
 409         return 1;
 410 }
 411
 412 pmd_t *populate_extra_pmd(unsigned long vaddr);
 413 pte_t *populate_extra_pte(unsigned long vaddr);
 414 #endif  /* __ASSEMBLY__ */
 415
 416 #ifdef CONFIG_X86_32
 417 # include <asm/pgtable_32.h>
 418 #else
 419 # include <asm/pgtable_64.h>
 420 #endif
 421
 422 #ifndef __ASSEMBLY__
 423 #include <linux/mm_types.h>
 424 #include <linux/mmdebug.h>
 425 #include <linux/log2.h>
 426
 427 static inline int pte_none(pte_t pte)
 428 {
 429         return !pte.pte;
 430 }
 431
 432 #define __HAVE_ARCH_PTE_SAME
 433 static inline int pte_same(pte_t a, pte_t b)
 434 {
 435         return a.pte == b.pte;
 436 }
 437
 438 static inline int pte_present(pte_t a)
 439 {
 440         return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
 441 }
 442
 443 #define pte_accessible pte_accessible
 444 static inline bool pte_accessible(struct mm_struct *mm, pte_t a)
 445 {
 446         if (pte_flags(a) & _PAGE_PRESENT)
 447                 return true;
 448
 449         if ((pte_flags(a) & _PAGE_PROTNONE) &&
 450                         mm_tlb_flush_pending(mm))
 451                 return true;
 452
 453         return false;
 454 }
 455
 456 static inline int pte_hidden(pte_t pte)
 457 {
 458         return pte_flags(pte) & _PAGE_HIDDEN;
 459 }
 460
 461 static inline int pmd_present(pmd_t pmd)
 462 {
 463         /*
 464          * Checking for _PAGE_PSE is needed too because
 465          * split_huge_page will temporarily clear the present bit (but
 466          * the _PAGE_PSE flag will remain set at all times while the
 467          * _PAGE_PRESENT bit is clear).
 468          */
 469         return pmd_flags(pmd) & (_PAGE_PRESENT | _PAGE_PROTNONE | _PAGE_PSE);
 470 }
 471
 472 #ifdef CONFIG_NUMA_BALANCING
 473 /*
 474  * These work without NUMA balancing but the kernel does not care. See the
 475  * comment in include/asm-generic/pgtable.h
 476  */
 477 static inline int pte_protnone(pte_t pte)
 478 {
 479         return (pte_flags(pte) & (_PAGE_PROTNONE | _PAGE_PRESENT))
 480                 == _PAGE_PROTNONE;
 481 }
 482
 483 static inline int pmd_protnone(pmd_t pmd)
 484 {
 485         return (pmd_flags(pmd) & (_PAGE_PROTNONE | _PAGE_PRESENT))
 486                 == _PAGE_PROTNONE;
 487 }
 488 #endif /* CONFIG_NUMA_BALANCING */
 489
 490 static inline int pmd_none(pmd_t pmd)
 491 {
 492         /* Only check low word on 32-bit platforms, since it might be
 493            out of sync with upper half. */
 494         return (unsigned long)native_pmd_val(pmd) == 0;
 495 }
 496
 497 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
 498 {
 499         return (unsigned long)__va(pmd_val(pmd) & PTE_PFN_MASK);
 500 }
 501
 502 /*
 503  * Currently stuck as a macro due to indirect forward reference to
 504  * linux/mmzone.h's __section_mem_map_addr() definition:
 505  */
 506 #define pmd_page(pmd)   pfn_to_page((pmd_val(pmd) & PTE_PFN_MASK) >> PAGE_SHIFT)
 507
 508 /*
 509  * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
 510  *
 511  * this macro returns the index of the entry in the pmd page which would
 512  * control the given virtual address
 513  */
 514 static inline unsigned long pmd_index(unsigned long address)
 515 {
 516         return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
 517 }
 518
 519 /*
 520  * Conversion functions: convert a page and protection to a page entry,
 521  * and a page entry and page directory to the page they refer to.
 522  *
 523  * (Currently stuck as a macro because of indirect forward reference
 524  * to linux/mm.h:page_to_nid())
 525  */
 526 #define mk_pte(page, pgprot)   pfn_pte(page_to_pfn(page), (pgprot))
 527
 528 /*
 529  * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
 530  *
 531  * this function returns the index of the entry in the pte page which would
 532  * control the given virtual address
 533  */
 534 static inline unsigned long pte_index(unsigned long address)
 535 {
 536         return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
 537 }
 538
 539 static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
 540 {
 541         return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
 542 }
 543
 544 static inline int pmd_bad(pmd_t pmd)
 545 {
 546         return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
 547 }
 548
 549 static inline unsigned long pages_to_mb(unsigned long npg)
 550 {
 551         return npg >> (20 - PAGE_SHIFT);
 552 }
 553
 554 #if CONFIG_PGTABLE_LEVELS > 2
 555 static inline int pud_none(pud_t pud)
 556 {
 557         return native_pud_val(pud) == 0;
 558 }
 559
 560 static inline int pud_present(pud_t pud)
 561 {
 562         return pud_flags(pud) & _PAGE_PRESENT;
 563 }
 564
 565 static inline unsigned long pud_page_vaddr(pud_t pud)
 566 {
 567         return (unsigned long)__va((unsigned long)pud_val(pud) & PTE_PFN_MASK);
 568 }
 569
 570 /*
 571  * Currently stuck as a macro due to indirect forward reference to
 572  * linux/mmzone.h's __section_mem_map_addr() definition:
 573  */
 574 #define pud_page(pud)           pfn_to_page(pud_val(pud) >> PAGE_SHIFT)
 575
 576 /* Find an entry in the second-level page table.. */
 577 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
 578 {
 579         return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
 580 }
 581
 582 static inline int pud_large(pud_t pud)
 583 {
 584         return (pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
 585                 (_PAGE_PSE | _PAGE_PRESENT);
 586 }
 587
 588 static inline int pud_bad(pud_t pud)
 589 {
 590         return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
 591 }
 592 #else
 593 static inline int pud_large(pud_t pud)
 594 {
 595         return 0;
 596 }
 597 #endif  /* CONFIG_PGTABLE_LEVELS > 2 */
 598
 599 #if CONFIG_PGTABLE_LEVELS > 3
 600 static inline int pgd_present(pgd_t pgd)
 601 {
 602         return pgd_flags(pgd) & _PAGE_PRESENT;
 603 }
 604
 605 static inline unsigned long pgd_page_vaddr(pgd_t pgd)
 606 {
 607         return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
 608 }
 609
 610 /*
 611  * Currently stuck as a macro due to indirect forward reference to
 612  * linux/mmzone.h's __section_mem_map_addr() definition:
 613  */
 614 #define pgd_page(pgd)           pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
 615
 616 /* to find an entry in a page-table-directory. */
 617 static inline unsigned long pud_index(unsigned long address)
 618 {
 619         return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
 620 }
 621
 622 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
 623 {
 624         return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
 625 }
 626
 627 static inline int pgd_bad(pgd_t pgd)
 628 {
 629         return (pgd_flags(pgd) & ~_PAGE_USER) != _KERNPG_TABLE;
 630 }
 631
 632 static inline int pgd_none(pgd_t pgd)
 633 {
 634         return !native_pgd_val(pgd);
 635 }
 636 #endif  /* CONFIG_PGTABLE_LEVELS > 3 */
 637
 638 #endif  /* __ASSEMBLY__ */
 639
 640 /*
 641  * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
 642  *
 643  * this macro returns the index of the entry in the pgd page which would
 644  * control the given virtual address
 645  */
 646 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
 647
 648 /*
 649  * pgd_offset() returns a (pgd_t *)
 650  * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
 651  */
 652 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
 653 /*
 654  * a shortcut which implies the use of the kernel's pgd, instead
 655  * of a process's
 656  */
 657 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
 658
 659
 660 #define KERNEL_PGD_BOUNDARY     pgd_index(PAGE_OFFSET)
 661 #define KERNEL_PGD_PTRS         (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
 662
 663 #ifndef __ASSEMBLY__
 664
 665 extern int direct_gbpages;
 666 void init_mem_mapping(void);
 667 void early_alloc_pgt_buf(void);
 668
 669 /* local pte updates need not use xchg for locking */
 670 static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
 671 {
 672         pte_t res = *ptep;
 673
 674         /* Pure native function needs no input for mm, addr */
 675         native_pte_clear(NULL, 0, ptep);
 676         return res;
 677 }
 678
 679 static inline pmd_t native_local_pmdp_get_and_clear(pmd_t *pmdp)
 680 {
 681         pmd_t res = *pmdp;
 682
 683         native_pmd_clear(pmdp);
 684         return res;
 685 }
 686
 687 static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
 688                                      pte_t *ptep , pte_t pte)
 689 {
 690         native_set_pte(ptep, pte);
 691 }
 692
 693 static inline void native_set_pmd_at(struct mm_struct *mm, unsigned long addr,
 694                                      pmd_t *pmdp , pmd_t pmd)
 695 {
 696         native_set_pmd(pmdp, pmd);
 697 }
 698
 699 #ifndef CONFIG_PARAVIRT
 700 /*
 701  * Rules for using pte_update - it must be called after any PTE update which
 702  * has not been done using the set_pte / clear_pte interfaces.  It is used by
 703  * shadow mode hypervisors to resynchronize the shadow page tables.  Kernel PTE
 704  * updates should either be sets, clears, or set_pte_atomic for P->P
 705  * transitions, which means this hook should only be called for user PTEs.
 706  * This hook implies a P->P protection or access change has taken place, which
 707  * requires a subsequent TLB flush.  The notification can optionally be delayed
 708  * until the TLB flush event by using the pte_update_defer form of the
 709  * interface, but care must be taken to assure that the flush happens while
 710  * still holding the same page table lock so that the shadow and primary pages
 711  * do not become out of sync on SMP.
 712  */
 713 #define pte_update(mm, addr, ptep)              do { } while (0)
 714 #define pte_update_defer(mm, addr, ptep)        do { } while (0)
 715 #endif
 716
 717 /*
 718  * We only update the dirty/accessed state if we set
 719  * the dirty bit by hand in the kernel, since the hardware
 720  * will do the accessed bit for us, and we don't want to
 721  * race with other CPU's that might be updating the dirty
 722  * bit at the same time.
 723  */
 724 struct vm_area_struct;
 725
 726 #define  __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
 727 extern int ptep_set_access_flags(struct vm_area_struct *vma,
 728                                  unsigned long address, pte_t *ptep,
 729                                  pte_t entry, int dirty);
 730
 731 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
 732 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
 733                                      unsigned long addr, pte_t *ptep);
 734
 735 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
 736 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
 737                                   unsigned long address, pte_t *ptep);
 738
 739 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
 740 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
 741                                        pte_t *ptep)
 742 {
 743         pte_t pte = native_ptep_get_and_clear(ptep);
 744         pte_update(mm, addr, ptep);
 745         return pte;
 746 }
 747
 748 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
 749 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
 750                                             unsigned long addr, pte_t *ptep,
 751                                             int full)
 752 {
 753         pte_t pte;
 754         if (full) {
 755                 /*
 756                  * Full address destruction in progress; paravirt does not
 757                  * care about updates and native needs no locking
 758                  */
 759                 pte = native_local_ptep_get_and_clear(ptep);
 760         } else {
 761                 pte = ptep_get_and_clear(mm, addr, ptep);
 762         }
 763         return pte;
 764 }
 765
 766 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
 767 static inline void ptep_set_wrprotect(struct mm_struct *mm,
 768                                       unsigned long addr, pte_t *ptep)
 769 {
 770         clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
 771         pte_update(mm, addr, ptep);
 772 }
 773
 774 #define flush_tlb_fix_spurious_fault(vma, address) do { } while (0)
 775
 776 #define mk_pmd(page, pgprot)   pfn_pmd(page_to_pfn(page), (pgprot))
 777
 778 #define  __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
 779 extern int pmdp_set_access_flags(struct vm_area_struct *vma,
 780                                  unsigned long address, pmd_t *pmdp,
 781                                  pmd_t entry, int dirty);
 782
 783 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
 784 extern int pmdp_test_and_clear_young(struct vm_area_struct *vma,
 785                                      unsigned long addr, pmd_t *pmdp);
 786
 787 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
 788 extern int pmdp_clear_flush_young(struct vm_area_struct *vma,
 789                                   unsigned long address, pmd_t *pmdp);
 790
 791
 792 #define __HAVE_ARCH_PMDP_SPLITTING_FLUSH
 793 extern void pmdp_splitting_flush(struct vm_area_struct *vma,
 794                                  unsigned long addr, pmd_t *pmdp);
 795
 796 #define __HAVE_ARCH_PMD_WRITE
 797 static inline int pmd_write(pmd_t pmd)
 798 {
 799         return pmd_flags(pmd) & _PAGE_RW;
 800 }
 801
 802 #define __HAVE_ARCH_PMDP_GET_AND_CLEAR
 803 static inline pmd_t pmdp_get_and_clear(struct mm_struct *mm, unsigned long addr,
 804                                        pmd_t *pmdp)
 805 {
 806         pmd_t pmd = native_pmdp_get_and_clear(pmdp);
 807         pmd_update(mm, addr, pmdp);
 808         return pmd;
 809 }
 810
 811 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
 812 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
 813                                       unsigned long addr, pmd_t *pmdp)
 814 {
 815         clear_bit(_PAGE_BIT_RW, (unsigned long *)pmdp);
 816         pmd_update(mm, addr, pmdp);
 817 }
 818
 819 /*
 820  * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
 821  *
 822  *  dst - pointer to pgd range anwhere on a pgd page
 823  *  src - ""
 824  *  count - the number of pgds to copy.
 825  *
 826  * dst and src can be on the same page, but the range must not overlap,
 827  * and must not cross a page boundary.
 828  */
 829 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
 830 {
 831        memcpy(dst, src, count * sizeof(pgd_t));
 832 }
 833
 834 #define PTE_SHIFT ilog2(PTRS_PER_PTE)
 835 static inline int page_level_shift(enum pg_level level)
 836 {
 837         return (PAGE_SHIFT - PTE_SHIFT) + level * PTE_SHIFT;
 838 }
 839 static inline unsigned long page_level_size(enum pg_level level)
 840 {
 841         return 1UL << page_level_shift(level);
 842 }
 843 static inline unsigned long page_level_mask(enum pg_level level)
 844 {
 845         return ~(page_level_size(level) - 1);
 846 }
 847
 848 /*
 849  * The x86 doesn't have any external MMU info: the kernel page
 850  * tables contain all the necessary information.
 851  */
 852 static inline void update_mmu_cache(struct vm_area_struct *vma,
 853                 unsigned long addr, pte_t *ptep)
 854 {
 855 }
 856 static inline void update_mmu_cache_pmd(struct vm_area_struct *vma,
 857                 unsigned long addr, pmd_t *pmd)
 858 {
 859 }
 860
 861 #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY
 862 static inline pte_t pte_swp_mksoft_dirty(pte_t pte)
 863 {
 864         return pte_set_flags(pte, _PAGE_SWP_SOFT_DIRTY);
 865 }
 866
 867 static inline int pte_swp_soft_dirty(pte_t pte)
 868 {
 869         return pte_flags(pte) & _PAGE_SWP_SOFT_DIRTY;
 870 }
 871
 872 static inline pte_t pte_swp_clear_soft_dirty(pte_t pte)
 873 {
 874         return pte_clear_flags(pte, _PAGE_SWP_SOFT_DIRTY);
 875 }
 876 #endif
 877
 878 #include <asm-generic/pgtable.h>
 879 #endif  /* __ASSEMBLY__ */
 880
 881 #endif /* _ASM_X86_PGTABLE_H */