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 void ptdump_walk_pgd_level_checkwx(void);
  23
  24 #ifdef CONFIG_DEBUG_WX
  25 #define debug_checkwx() ptdump_walk_pgd_level_checkwx()
  26 #else
  27 #define debug_checkwx() do { } while (0)
  28 #endif
  29
  30 /*
  31  * ZERO_PAGE is a global shared page that is always zero: used
  32  * for zero-mapped memory areas etc..
  33  */
  34 extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)]
  35         __visible;
  36 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
  37
  38 extern spinlock_t pgd_lock;
  39 extern struct list_head pgd_list;
  40
  41 extern struct mm_struct *pgd_page_get_mm(struct page *page);
  42
  43 #ifdef CONFIG_PARAVIRT
  44 #include <asm/paravirt.h>
  45 #else  /* !CONFIG_PARAVIRT */
  46 #define set_pte(ptep, pte)              native_set_pte(ptep, pte)
  47 #define set_pte_at(mm, addr, ptep, pte) native_set_pte_at(mm, addr, ptep, pte)
  48 #define set_pmd_at(mm, addr, pmdp, pmd) native_set_pmd_at(mm, addr, pmdp, pmd)
  49
  50 #define set_pte_atomic(ptep, pte)                                       \
  51         native_set_pte_atomic(ptep, pte)
  52
  53 #define set_pmd(pmdp, pmd)              native_set_pmd(pmdp, pmd)
  54
  55 #ifndef __PAGETABLE_PUD_FOLDED
  56 #define set_pgd(pgdp, pgd)              native_set_pgd(pgdp, pgd)
  57 #define pgd_clear(pgd)                  native_pgd_clear(pgd)
  58 #endif
  59
  60 #ifndef set_pud
  61 # define set_pud(pudp, pud)             native_set_pud(pudp, pud)
  62 #endif
  63
  64 #ifndef __PAGETABLE_PMD_FOLDED
  65 #define pud_clear(pud)                  native_pud_clear(pud)
  66 #endif
  67
  68 #define pte_clear(mm, addr, ptep)       native_pte_clear(mm, addr, ptep)
  69 #define pmd_clear(pmd)                  native_pmd_clear(pmd)
  70
  71 #define pte_update(mm, addr, ptep)              do { } while (0)
  72
  73 #define pgd_val(x)      native_pgd_val(x)
  74 #define __pgd(x)        native_make_pgd(x)
  75
  76 #ifndef __PAGETABLE_PUD_FOLDED
  77 #define pud_val(x)      native_pud_val(x)
  78 #define __pud(x)        native_make_pud(x)
  79 #endif
  80
  81 #ifndef __PAGETABLE_PMD_FOLDED
  82 #define pmd_val(x)      native_pmd_val(x)
  83 #define __pmd(x)        native_make_pmd(x)
  84 #endif
  85
  86 #define pte_val(x)      native_pte_val(x)
  87 #define __pte(x)        native_make_pte(x)
  88
  89 #define arch_end_context_switch(prev)   do {} while(0)
  90
  91 #endif  /* CONFIG_PARAVIRT */
  92
  93 /*
  94  * The following only work if pte_present() is true.
  95  * Undefined behaviour if not..
  96  */
  97 static inline int pte_dirty(pte_t pte)
  98 {
  99         return pte_flags(pte) & _PAGE_DIRTY;
 100 }
 101
 102 static inline int pte_young(pte_t pte)
 103 {
 104         return pte_flags(pte) & _PAGE_ACCESSED;
 105 }
 106
 107 static inline int pmd_dirty(pmd_t pmd)
 108 {
 109         return pmd_flags(pmd) & _PAGE_DIRTY;
 110 }
 111
 112 static inline int pmd_young(pmd_t pmd)
 113 {
 114         return pmd_flags(pmd) & _PAGE_ACCESSED;
 115 }
 116
 117 static inline int pte_write(pte_t pte)
 118 {
 119         return pte_flags(pte) & _PAGE_RW;
 120 }
 121
 122 static inline int pte_huge(pte_t pte)
 123 {
 124         return pte_flags(pte) & _PAGE_PSE;
 125 }
 126
 127 static inline int pte_global(pte_t pte)
 128 {
 129         return pte_flags(pte) & _PAGE_GLOBAL;
 130 }
 131
 132 static inline int pte_exec(pte_t pte)
 133 {
 134         return !(pte_flags(pte) & _PAGE_NX);
 135 }
 136
 137 static inline int pte_special(pte_t pte)
 138 {
 139         return pte_flags(pte) & _PAGE_SPECIAL;
 140 }
 141
 142 static inline unsigned long pte_pfn(pte_t pte)
 143 {
 144         return (pte_val(pte) & PTE_PFN_MASK) >> PAGE_SHIFT;
 145 }
 146
 147 static inline unsigned long pmd_pfn(pmd_t pmd)
 148 {
 149         return (pmd_val(pmd) & pmd_pfn_mask(pmd)) >> PAGE_SHIFT;
 150 }
 151
 152 static inline unsigned long pud_pfn(pud_t pud)
 153 {
 154         return (pud_val(pud) & pud_pfn_mask(pud)) >> PAGE_SHIFT;
 155 }
 156
 157 #define pte_page(pte)   pfn_to_page(pte_pfn(pte))
 158
 159 static inline int pmd_large(pmd_t pte)
 160 {
 161         return pmd_flags(pte) & _PAGE_PSE;
 162 }
 163
 164 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 165 static inline int pmd_trans_huge(pmd_t pmd)
 166 {
 167         return pmd_val(pmd) & _PAGE_PSE;
 168 }
 169
 170 static inline int has_transparent_hugepage(void)
 171 {
 172         return cpu_has_pse;
 173 }
 174 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
 175
 176 static inline pte_t pte_set_flags(pte_t pte, pteval_t set)
 177 {
 178         pteval_t v = native_pte_val(pte);
 179
 180         return native_make_pte(v | set);
 181 }
 182
 183 static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
 184 {
 185         pteval_t v = native_pte_val(pte);
 186
 187         return native_make_pte(v & ~clear);
 188 }
 189
 190 static inline pte_t pte_mkclean(pte_t pte)
 191 {
 192         return pte_clear_flags(pte, _PAGE_DIRTY);
 193 }
 194
 195 static inline pte_t pte_mkold(pte_t pte)
 196 {
 197         return pte_clear_flags(pte, _PAGE_ACCESSED);
 198 }
 199
 200 static inline pte_t pte_wrprotect(pte_t pte)
 201 {
 202         return pte_clear_flags(pte, _PAGE_RW);
 203 }
 204
 205 static inline pte_t pte_mkexec(pte_t pte)
 206 {
 207         return pte_clear_flags(pte, _PAGE_NX);
 208 }
 209
 210 static inline pte_t pte_mkdirty(pte_t pte)
 211 {
 212         return pte_set_flags(pte, _PAGE_DIRTY | _PAGE_SOFT_DIRTY);
 213 }
 214
 215 static inline pte_t pte_mkyoung(pte_t pte)
 216 {
 217         return pte_set_flags(pte, _PAGE_ACCESSED);
 218 }
 219
 220 static inline pte_t pte_mkwrite(pte_t pte)
 221 {
 222         return pte_set_flags(pte, _PAGE_RW);
 223 }
 224
 225 static inline pte_t pte_mkhuge(pte_t pte)
 226 {
 227         return pte_set_flags(pte, _PAGE_PSE);
 228 }
 229
 230 static inline pte_t pte_clrhuge(pte_t pte)
 231 {
 232         return pte_clear_flags(pte, _PAGE_PSE);
 233 }
 234
 235 static inline pte_t pte_mkglobal(pte_t pte)
 236 {
 237         return pte_set_flags(pte, _PAGE_GLOBAL);
 238 }
 239
 240 static inline pte_t pte_clrglobal(pte_t pte)
 241 {
 242         return pte_clear_flags(pte, _PAGE_GLOBAL);
 243 }
 244
 245 static inline pte_t pte_mkspecial(pte_t pte)
 246 {
 247         return pte_set_flags(pte, _PAGE_SPECIAL);
 248 }
 249
 250 static inline pmd_t pmd_set_flags(pmd_t pmd, pmdval_t set)
 251 {
 252         pmdval_t v = native_pmd_val(pmd);
 253
 254         return __pmd(v | set);
 255 }
 256
 257 static inline pmd_t pmd_clear_flags(pmd_t pmd, pmdval_t clear)
 258 {
 259         pmdval_t v = native_pmd_val(pmd);
 260
 261         return __pmd(v & ~clear);
 262 }
 263
 264 static inline pmd_t pmd_mkold(pmd_t pmd)
 265 {
 266         return pmd_clear_flags(pmd, _PAGE_ACCESSED);
 267 }
 268
 269 static inline pmd_t pmd_wrprotect(pmd_t pmd)
 270 {
 271         return pmd_clear_flags(pmd, _PAGE_RW);
 272 }
 273
 274 static inline pmd_t pmd_mkdirty(pmd_t pmd)
 275 {
 276         return pmd_set_flags(pmd, _PAGE_DIRTY | _PAGE_SOFT_DIRTY);
 277 }
 278
 279 static inline pmd_t pmd_mkhuge(pmd_t pmd)
 280 {
 281         return pmd_set_flags(pmd, _PAGE_PSE);
 282 }
 283
 284 static inline pmd_t pmd_mkyoung(pmd_t pmd)
 285 {
 286         return pmd_set_flags(pmd, _PAGE_ACCESSED);
 287 }
 288
 289 static inline pmd_t pmd_mkwrite(pmd_t pmd)
 290 {
 291         return pmd_set_flags(pmd, _PAGE_RW);
 292 }
 293
 294 static inline pmd_t pmd_mknotpresent(pmd_t pmd)
 295 {
 296         return pmd_clear_flags(pmd, _PAGE_PRESENT | _PAGE_PROTNONE);
 297 }
 298
 299 #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY
 300 static inline int pte_soft_dirty(pte_t pte)
 301 {
 302         return pte_flags(pte) & _PAGE_SOFT_DIRTY;
 303 }
 304
 305 static inline int pmd_soft_dirty(pmd_t pmd)
 306 {
 307         return pmd_flags(pmd) & _PAGE_SOFT_DIRTY;
 308 }
 309
 310 static inline pte_t pte_mksoft_dirty(pte_t pte)
 311 {
 312         return pte_set_flags(pte, _PAGE_SOFT_DIRTY);
 313 }
 314
 315 static inline pmd_t pmd_mksoft_dirty(pmd_t pmd)
 316 {
 317         return pmd_set_flags(pmd, _PAGE_SOFT_DIRTY);
 318 }
 319
 320 static inline pte_t pte_clear_soft_dirty(pte_t pte)
 321 {
 322         return pte_clear_flags(pte, _PAGE_SOFT_DIRTY);
 323 }
 324
 325 static inline pmd_t pmd_clear_soft_dirty(pmd_t pmd)
 326 {
 327         return pmd_clear_flags(pmd, _PAGE_SOFT_DIRTY);
 328 }
 329
 330 #endif /* CONFIG_HAVE_ARCH_SOFT_DIRTY */
 331
 332 /*
 333  * Mask out unsupported bits in a present pgprot.  Non-present pgprots
 334  * can use those bits for other purposes, so leave them be.
 335  */
 336 static inline pgprotval_t massage_pgprot(pgprot_t pgprot)
 337 {
 338         pgprotval_t protval = pgprot_val(pgprot);
 339
 340         if (protval & _PAGE_PRESENT)
 341                 protval &= __supported_pte_mask;
 342
 343         return protval;
 344 }
 345
 346 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
 347 {
 348         return __pte(((phys_addr_t)page_nr << PAGE_SHIFT) |
 349                      massage_pgprot(pgprot));
 350 }
 351
 352 static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
 353 {
 354         return __pmd(((phys_addr_t)page_nr << PAGE_SHIFT) |
 355                      massage_pgprot(pgprot));
 356 }
 357
 358 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 359 {
 360         pteval_t val = pte_val(pte);
 361
 362         /*
 363          * Chop off the NX bit (if present), and add the NX portion of
 364          * the newprot (if present):
 365          */
 366         val &= _PAGE_CHG_MASK;
 367         val |= massage_pgprot(newprot) & ~_PAGE_CHG_MASK;
 368
 369         return __pte(val);
 370 }
 371
 372 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
 373 {
 374         pmdval_t val = pmd_val(pmd);
 375
 376         val &= _HPAGE_CHG_MASK;
 377         val |= massage_pgprot(newprot) & ~_HPAGE_CHG_MASK;
 378
 379         return __pmd(val);
 380 }
 381
 382 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
 383 #define pgprot_modify pgprot_modify
 384 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
 385 {
 386         pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
 387         pgprotval_t addbits = pgprot_val(newprot);
 388         return __pgprot(preservebits | addbits);
 389 }
 390
 391 #define pte_pgprot(x) __pgprot(pte_flags(x))
 392 #define pmd_pgprot(x) __pgprot(pmd_flags(x))
 393 #define pud_pgprot(x) __pgprot(pud_flags(x))
 394
 395 #define canon_pgprot(p) __pgprot(massage_pgprot(p))
 396
 397 static inline int is_new_memtype_allowed(u64 paddr, unsigned long size,
 398                                          enum page_cache_mode pcm,
 399                                          enum page_cache_mode new_pcm)
 400 {
 401         /*
 402          * PAT type is always WB for untracked ranges, so no need to check.
 403          */
 404         if (x86_platform.is_untracked_pat_range(paddr, paddr + size))
 405                 return 1;
 406
 407         /*
 408          * Certain new memtypes are not allowed with certain
 409          * requested memtype:
 410          * - request is uncached, return cannot be write-back
 411          * - request is write-combine, return cannot be write-back
 412          * - request is write-through, return cannot be write-back
 413          * - request is write-through, return cannot be write-combine
 414          */
 415         if ((pcm == _PAGE_CACHE_MODE_UC_MINUS &&
 416              new_pcm == _PAGE_CACHE_MODE_WB) ||
 417             (pcm == _PAGE_CACHE_MODE_WC &&
 418              new_pcm == _PAGE_CACHE_MODE_WB) ||
 419             (pcm == _PAGE_CACHE_MODE_WT &&
 420              new_pcm == _PAGE_CACHE_MODE_WB) ||
 421             (pcm == _PAGE_CACHE_MODE_WT &&
 422              new_pcm == _PAGE_CACHE_MODE_WC)) {
 423                 return 0;
 424         }
 425
 426         return 1;
 427 }
 428
 429 pmd_t *populate_extra_pmd(unsigned long vaddr);
 430 pte_t *populate_extra_pte(unsigned long vaddr);
 431 #endif  /* __ASSEMBLY__ */
 432
 433 #ifdef CONFIG_X86_32
 434 # include <asm/pgtable_32.h>
 435 #else
 436 # include <asm/pgtable_64.h>
 437 #endif
 438
 439 #ifndef __ASSEMBLY__
 440 #include <linux/mm_types.h>
 441 #include <linux/mmdebug.h>
 442 #include <linux/log2.h>
 443
 444 static inline int pte_none(pte_t pte)
 445 {
 446         return !pte.pte;
 447 }
 448
 449 #define __HAVE_ARCH_PTE_SAME
 450 static inline int pte_same(pte_t a, pte_t b)
 451 {
 452         return a.pte == b.pte;
 453 }
 454
 455 static inline int pte_present(pte_t a)
 456 {
 457         return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
 458 }
 459
 460 #define pte_accessible pte_accessible
 461 static inline bool pte_accessible(struct mm_struct *mm, pte_t a)
 462 {
 463         if (pte_flags(a) & _PAGE_PRESENT)
 464                 return true;
 465
 466         if ((pte_flags(a) & _PAGE_PROTNONE) &&
 467                         mm_tlb_flush_pending(mm))
 468                 return true;
 469
 470         return false;
 471 }
 472
 473 static inline int pte_hidden(pte_t pte)
 474 {
 475         return pte_flags(pte) & _PAGE_HIDDEN;
 476 }
 477
 478 static inline int pmd_present(pmd_t pmd)
 479 {
 480         /*
 481          * Checking for _PAGE_PSE is needed too because
 482          * split_huge_page will temporarily clear the present bit (but
 483          * the _PAGE_PSE flag will remain set at all times while the
 484          * _PAGE_PRESENT bit is clear).
 485          */
 486         return pmd_flags(pmd) & (_PAGE_PRESENT | _PAGE_PROTNONE | _PAGE_PSE);
 487 }
 488
 489 #ifdef CONFIG_NUMA_BALANCING
 490 /*
 491  * These work without NUMA balancing but the kernel does not care. See the
 492  * comment in include/asm-generic/pgtable.h
 493  */
 494 static inline int pte_protnone(pte_t pte)
 495 {
 496         return (pte_flags(pte) & (_PAGE_PROTNONE | _PAGE_PRESENT))
 497                 == _PAGE_PROTNONE;
 498 }
 499
 500 static inline int pmd_protnone(pmd_t pmd)
 501 {
 502         return (pmd_flags(pmd) & (_PAGE_PROTNONE | _PAGE_PRESENT))
 503                 == _PAGE_PROTNONE;
 504 }
 505 #endif /* CONFIG_NUMA_BALANCING */
 506
 507 static inline int pmd_none(pmd_t pmd)
 508 {
 509         /* Only check low word on 32-bit platforms, since it might be
 510            out of sync with upper half. */
 511         return (unsigned long)native_pmd_val(pmd) == 0;
 512 }
 513
 514 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
 515 {
 516         return (unsigned long)__va(pmd_val(pmd) & pmd_pfn_mask(pmd));
 517 }
 518
 519 /*
 520  * Currently stuck as a macro due to indirect forward reference to
 521  * linux/mmzone.h's __section_mem_map_addr() definition:
 522  */
 523 #define pmd_page(pmd)           \
 524         pfn_to_page((pmd_val(pmd) & pmd_pfn_mask(pmd)) >> PAGE_SHIFT)
 525
 526 /*
 527  * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
 528  *
 529  * this macro returns the index of the entry in the pmd page which would
 530  * control the given virtual address
 531  */
 532 static inline unsigned long pmd_index(unsigned long address)
 533 {
 534         return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
 535 }
 536
 537 /*
 538  * Conversion functions: convert a page and protection to a page entry,
 539  * and a page entry and page directory to the page they refer to.
 540  *
 541  * (Currently stuck as a macro because of indirect forward reference
 542  * to linux/mm.h:page_to_nid())
 543  */
 544 #define mk_pte(page, pgprot)   pfn_pte(page_to_pfn(page), (pgprot))
 545
 546 /*
 547  * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
 548  *
 549  * this function returns the index of the entry in the pte page which would
 550  * control the given virtual address
 551  */
 552 static inline unsigned long pte_index(unsigned long address)
 553 {
 554         return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
 555 }
 556
 557 static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
 558 {
 559         return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
 560 }
 561
 562 static inline int pmd_bad(pmd_t pmd)
 563 {
 564         return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
 565 }
 566
 567 static inline unsigned long pages_to_mb(unsigned long npg)
 568 {
 569         return npg >> (20 - PAGE_SHIFT);
 570 }
 571
 572 #if CONFIG_PGTABLE_LEVELS > 2
 573 static inline int pud_none(pud_t pud)
 574 {
 575         return native_pud_val(pud) == 0;
 576 }
 577
 578 static inline int pud_present(pud_t pud)
 579 {
 580         return pud_flags(pud) & _PAGE_PRESENT;
 581 }
 582
 583 static inline unsigned long pud_page_vaddr(pud_t pud)
 584 {
 585         return (unsigned long)__va(pud_val(pud) & pud_pfn_mask(pud));
 586 }
 587
 588 /*
 589  * Currently stuck as a macro due to indirect forward reference to
 590  * linux/mmzone.h's __section_mem_map_addr() definition:
 591  */
 592 #define pud_page(pud)           \
 593         pfn_to_page((pud_val(pud) & pud_pfn_mask(pud)) >> PAGE_SHIFT)
 594
 595 /* Find an entry in the second-level page table.. */
 596 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
 597 {
 598         return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
 599 }
 600
 601 static inline int pud_large(pud_t pud)
 602 {
 603         return (pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
 604                 (_PAGE_PSE | _PAGE_PRESENT);
 605 }
 606
 607 static inline int pud_bad(pud_t pud)
 608 {
 609         return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
 610 }
 611 #else
 612 static inline int pud_large(pud_t pud)
 613 {
 614         return 0;
 615 }
 616 #endif  /* CONFIG_PGTABLE_LEVELS > 2 */
 617
 618 #if CONFIG_PGTABLE_LEVELS > 3
 619 static inline int pgd_present(pgd_t pgd)
 620 {
 621         return pgd_flags(pgd) & _PAGE_PRESENT;
 622 }
 623
 624 static inline unsigned long pgd_page_vaddr(pgd_t pgd)
 625 {
 626         return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
 627 }
 628
 629 /*
 630  * Currently stuck as a macro due to indirect forward reference to
 631  * linux/mmzone.h's __section_mem_map_addr() definition:
 632  */
 633 #define pgd_page(pgd)           pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
 634
 635 /* to find an entry in a page-table-directory. */
 636 static inline unsigned long pud_index(unsigned long address)
 637 {
 638         return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
 639 }
 640
 641 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
 642 {
 643         return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
 644 }
 645
 646 static inline int pgd_bad(pgd_t pgd)
 647 {
 648         return (pgd_flags(pgd) & ~_PAGE_USER) != _KERNPG_TABLE;
 649 }
 650
 651 static inline int pgd_none(pgd_t pgd)
 652 {
 653         return !native_pgd_val(pgd);
 654 }
 655 #endif  /* CONFIG_PGTABLE_LEVELS > 3 */
 656
 657 #endif  /* __ASSEMBLY__ */
 658
 659 /*
 660  * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
 661  *
 662  * this macro returns the index of the entry in the pgd page which would
 663  * control the given virtual address
 664  */
 665 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
 666
 667 /*
 668  * pgd_offset() returns a (pgd_t *)
 669  * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
 670  */
 671 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
 672 /*
 673  * a shortcut which implies the use of the kernel's pgd, instead
 674  * of a process's
 675  */
 676 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
 677
 678
 679 #define KERNEL_PGD_BOUNDARY     pgd_index(PAGE_OFFSET)
 680 #define KERNEL_PGD_PTRS         (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
 681
 682 #ifndef __ASSEMBLY__
 683
 684 extern int direct_gbpages;
 685 void init_mem_mapping(void);
 686 void early_alloc_pgt_buf(void);
 687
 688 /* local pte updates need not use xchg for locking */
 689 static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
 690 {
 691         pte_t res = *ptep;
 692
 693         /* Pure native function needs no input for mm, addr */
 694         native_pte_clear(NULL, 0, ptep);
 695         return res;
 696 }
 697
 698 static inline pmd_t native_local_pmdp_get_and_clear(pmd_t *pmdp)
 699 {
 700         pmd_t res = *pmdp;
 701
 702         native_pmd_clear(pmdp);
 703         return res;
 704 }
 705
 706 static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
 707                                      pte_t *ptep , pte_t pte)
 708 {
 709         native_set_pte(ptep, pte);
 710 }
 711
 712 static inline void native_set_pmd_at(struct mm_struct *mm, unsigned long addr,
 713                                      pmd_t *pmdp , pmd_t pmd)
 714 {
 715         native_set_pmd(pmdp, pmd);
 716 }
 717
 718 #ifndef CONFIG_PARAVIRT
 719 /*
 720  * Rules for using pte_update - it must be called after any PTE update which
 721  * has not been done using the set_pte / clear_pte interfaces.  It is used by
 722  * shadow mode hypervisors to resynchronize the shadow page tables.  Kernel PTE
 723  * updates should either be sets, clears, or set_pte_atomic for P->P
 724  * transitions, which means this hook should only be called for user PTEs.
 725  * This hook implies a P->P protection or access change has taken place, which
 726  * requires a subsequent TLB flush.
 727  */
 728 #define pte_update(mm, addr, ptep)              do { } while (0)
 729 #endif
 730
 731 /*
 732  * We only update the dirty/accessed state if we set
 733  * the dirty bit by hand in the kernel, since the hardware
 734  * will do the accessed bit for us, and we don't want to
 735  * race with other CPU's that might be updating the dirty
 736  * bit at the same time.
 737  */
 738 struct vm_area_struct;
 739
 740 #define  __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
 741 extern int ptep_set_access_flags(struct vm_area_struct *vma,
 742                                  unsigned long address, pte_t *ptep,
 743                                  pte_t entry, int dirty);
 744
 745 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
 746 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
 747                                      unsigned long addr, pte_t *ptep);
 748
 749 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
 750 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
 751                                   unsigned long address, pte_t *ptep);
 752
 753 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
 754 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
 755                                        pte_t *ptep)
 756 {
 757         pte_t pte = native_ptep_get_and_clear(ptep);
 758         pte_update(mm, addr, ptep);
 759         return pte;
 760 }
 761
 762 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
 763 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
 764                                             unsigned long addr, pte_t *ptep,
 765                                             int full)
 766 {
 767         pte_t pte;
 768         if (full) {
 769                 /*
 770                  * Full address destruction in progress; paravirt does not
 771                  * care about updates and native needs no locking
 772                  */
 773                 pte = native_local_ptep_get_and_clear(ptep);
 774         } else {
 775                 pte = ptep_get_and_clear(mm, addr, ptep);
 776         }
 777         return pte;
 778 }
 779
 780 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
 781 static inline void ptep_set_wrprotect(struct mm_struct *mm,
 782                                       unsigned long addr, pte_t *ptep)
 783 {
 784         clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
 785         pte_update(mm, addr, ptep);
 786 }
 787
 788 #define flush_tlb_fix_spurious_fault(vma, address) do { } while (0)
 789
 790 #define mk_pmd(page, pgprot)   pfn_pmd(page_to_pfn(page), (pgprot))
 791
 792 #define  __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
 793 extern int pmdp_set_access_flags(struct vm_area_struct *vma,
 794                                  unsigned long address, pmd_t *pmdp,
 795                                  pmd_t entry, int dirty);
 796
 797 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
 798 extern int pmdp_test_and_clear_young(struct vm_area_struct *vma,
 799                                      unsigned long addr, pmd_t *pmdp);
 800
 801 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
 802 extern int pmdp_clear_flush_young(struct vm_area_struct *vma,
 803                                   unsigned long address, pmd_t *pmdp);
 804
 805
 806 #define __HAVE_ARCH_PMD_WRITE
 807 static inline int pmd_write(pmd_t pmd)
 808 {
 809         return pmd_flags(pmd) & _PAGE_RW;
 810 }
 811
 812 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
 813 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm, unsigned long addr,
 814                                        pmd_t *pmdp)
 815 {
 816         return native_pmdp_get_and_clear(pmdp);
 817 }
 818
 819 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
 820 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
 821                                       unsigned long addr, pmd_t *pmdp)
 822 {
 823         clear_bit(_PAGE_BIT_RW, (unsigned long *)pmdp);
 824 }
 825
 826 /*
 827  * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
 828  *
 829  *  dst - pointer to pgd range anwhere on a pgd page
 830  *  src - ""
 831  *  count - the number of pgds to copy.
 832  *
 833  * dst and src can be on the same page, but the range must not overlap,
 834  * and must not cross a page boundary.
 835  */
 836 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
 837 {
 838        memcpy(dst, src, count * sizeof(pgd_t));
 839 }
 840
 841 #define PTE_SHIFT ilog2(PTRS_PER_PTE)
 842 static inline int page_level_shift(enum pg_level level)
 843 {
 844         return (PAGE_SHIFT - PTE_SHIFT) + level * PTE_SHIFT;
 845 }
 846 static inline unsigned long page_level_size(enum pg_level level)
 847 {
 848         return 1UL << page_level_shift(level);
 849 }
 850 static inline unsigned long page_level_mask(enum pg_level level)
 851 {
 852         return ~(page_level_size(level) - 1);
 853 }
 854
 855 /*
 856  * The x86 doesn't have any external MMU info: the kernel page
 857  * tables contain all the necessary information.
 858  */
 859 static inline void update_mmu_cache(struct vm_area_struct *vma,
 860                 unsigned long addr, pte_t *ptep)
 861 {
 862 }
 863 static inline void update_mmu_cache_pmd(struct vm_area_struct *vma,
 864                 unsigned long addr, pmd_t *pmd)
 865 {
 866 }
 867
 868 #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY
 869 static inline pte_t pte_swp_mksoft_dirty(pte_t pte)
 870 {
 871         return pte_set_flags(pte, _PAGE_SWP_SOFT_DIRTY);
 872 }
 873
 874 static inline int pte_swp_soft_dirty(pte_t pte)
 875 {
 876         return pte_flags(pte) & _PAGE_SWP_SOFT_DIRTY;
 877 }
 878
 879 static inline pte_t pte_swp_clear_soft_dirty(pte_t pte)
 880 {
 881         return pte_clear_flags(pte, _PAGE_SWP_SOFT_DIRTY);
 882 }
 883 #endif
 884
 885 #include <asm-generic/pgtable.h>
 886 #endif  /* __ASSEMBLY__ */
 887
 888 #endif /* _ASM_X86_PGTABLE_H */