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
 103 static inline u32 read_pkru(void)
 104 {
 105         if (boot_cpu_has(X86_FEATURE_OSPKE))
 106                 return __read_pkru();
 107         return 0;
 108 }
 109
 110 static inline int pte_young(pte_t pte)
 111 {
 112         return pte_flags(pte) & _PAGE_ACCESSED;
 113 }
 114
 115 static inline int pmd_dirty(pmd_t pmd)
 116 {
 117         return pmd_flags(pmd) & _PAGE_DIRTY;
 118 }
 119
 120 static inline int pmd_young(pmd_t pmd)
 121 {
 122         return pmd_flags(pmd) & _PAGE_ACCESSED;
 123 }
 124
 125 static inline int pte_write(pte_t pte)
 126 {
 127         return pte_flags(pte) & _PAGE_RW;
 128 }
 129
 130 static inline int pte_huge(pte_t pte)
 131 {
 132         return pte_flags(pte) & _PAGE_PSE;
 133 }
 134
 135 static inline int pte_global(pte_t pte)
 136 {
 137         return pte_flags(pte) & _PAGE_GLOBAL;
 138 }
 139
 140 static inline int pte_exec(pte_t pte)
 141 {
 142         return !(pte_flags(pte) & _PAGE_NX);
 143 }
 144
 145 static inline int pte_special(pte_t pte)
 146 {
 147         return pte_flags(pte) & _PAGE_SPECIAL;
 148 }
 149
 150 static inline unsigned long pte_pfn(pte_t pte)
 151 {
 152         return (pte_val(pte) & PTE_PFN_MASK) >> PAGE_SHIFT;
 153 }
 154
 155 static inline unsigned long pmd_pfn(pmd_t pmd)
 156 {
 157         return (pmd_val(pmd) & pmd_pfn_mask(pmd)) >> PAGE_SHIFT;
 158 }
 159
 160 static inline unsigned long pud_pfn(pud_t pud)
 161 {
 162         return (pud_val(pud) & pud_pfn_mask(pud)) >> PAGE_SHIFT;
 163 }
 164
 165 #define pte_page(pte)   pfn_to_page(pte_pfn(pte))
 166
 167 static inline int pmd_large(pmd_t pte)
 168 {
 169         return pmd_flags(pte) & _PAGE_PSE;
 170 }
 171
 172 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 173 static inline int pmd_trans_huge(pmd_t pmd)
 174 {
 175         return (pmd_val(pmd) & (_PAGE_PSE|_PAGE_DEVMAP)) == _PAGE_PSE;
 176 }
 177
 178 static inline int has_transparent_hugepage(void)
 179 {
 180         return cpu_has_pse;
 181 }
 182
 183 #ifdef __HAVE_ARCH_PTE_DEVMAP
 184 static inline int pmd_devmap(pmd_t pmd)
 185 {
 186         return !!(pmd_val(pmd) & _PAGE_DEVMAP);
 187 }
 188 #endif
 189 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
 190
 191 static inline pte_t pte_set_flags(pte_t pte, pteval_t set)
 192 {
 193         pteval_t v = native_pte_val(pte);
 194
 195         return native_make_pte(v | set);
 196 }
 197
 198 static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
 199 {
 200         pteval_t v = native_pte_val(pte);
 201
 202         return native_make_pte(v & ~clear);
 203 }
 204
 205 static inline pte_t pte_mkclean(pte_t pte)
 206 {
 207         return pte_clear_flags(pte, _PAGE_DIRTY);
 208 }
 209
 210 static inline pte_t pte_mkold(pte_t pte)
 211 {
 212         return pte_clear_flags(pte, _PAGE_ACCESSED);
 213 }
 214
 215 static inline pte_t pte_wrprotect(pte_t pte)
 216 {
 217         return pte_clear_flags(pte, _PAGE_RW);
 218 }
 219
 220 static inline pte_t pte_mkexec(pte_t pte)
 221 {
 222         return pte_clear_flags(pte, _PAGE_NX);
 223 }
 224
 225 static inline pte_t pte_mkdirty(pte_t pte)
 226 {
 227         return pte_set_flags(pte, _PAGE_DIRTY | _PAGE_SOFT_DIRTY);
 228 }
 229
 230 static inline pte_t pte_mkyoung(pte_t pte)
 231 {
 232         return pte_set_flags(pte, _PAGE_ACCESSED);
 233 }
 234
 235 static inline pte_t pte_mkwrite(pte_t pte)
 236 {
 237         return pte_set_flags(pte, _PAGE_RW);
 238 }
 239
 240 static inline pte_t pte_mkhuge(pte_t pte)
 241 {
 242         return pte_set_flags(pte, _PAGE_PSE);
 243 }
 244
 245 static inline pte_t pte_clrhuge(pte_t pte)
 246 {
 247         return pte_clear_flags(pte, _PAGE_PSE);
 248 }
 249
 250 static inline pte_t pte_mkglobal(pte_t pte)
 251 {
 252         return pte_set_flags(pte, _PAGE_GLOBAL);
 253 }
 254
 255 static inline pte_t pte_clrglobal(pte_t pte)
 256 {
 257         return pte_clear_flags(pte, _PAGE_GLOBAL);
 258 }
 259
 260 static inline pte_t pte_mkspecial(pte_t pte)
 261 {
 262         return pte_set_flags(pte, _PAGE_SPECIAL);
 263 }
 264
 265 static inline pte_t pte_mkdevmap(pte_t pte)
 266 {
 267         return pte_set_flags(pte, _PAGE_SPECIAL|_PAGE_DEVMAP);
 268 }
 269
 270 static inline pmd_t pmd_set_flags(pmd_t pmd, pmdval_t set)
 271 {
 272         pmdval_t v = native_pmd_val(pmd);
 273
 274         return __pmd(v | set);
 275 }
 276
 277 static inline pmd_t pmd_clear_flags(pmd_t pmd, pmdval_t clear)
 278 {
 279         pmdval_t v = native_pmd_val(pmd);
 280
 281         return __pmd(v & ~clear);
 282 }
 283
 284 static inline pmd_t pmd_mkold(pmd_t pmd)
 285 {
 286         return pmd_clear_flags(pmd, _PAGE_ACCESSED);
 287 }
 288
 289 static inline pmd_t pmd_mkclean(pmd_t pmd)
 290 {
 291         return pmd_clear_flags(pmd, _PAGE_DIRTY);
 292 }
 293
 294 static inline pmd_t pmd_wrprotect(pmd_t pmd)
 295 {
 296         return pmd_clear_flags(pmd, _PAGE_RW);
 297 }
 298
 299 static inline pmd_t pmd_mkdirty(pmd_t pmd)
 300 {
 301         return pmd_set_flags(pmd, _PAGE_DIRTY | _PAGE_SOFT_DIRTY);
 302 }
 303
 304 static inline pmd_t pmd_mkdevmap(pmd_t pmd)
 305 {
 306         return pmd_set_flags(pmd, _PAGE_DEVMAP);
 307 }
 308
 309 static inline pmd_t pmd_mkhuge(pmd_t pmd)
 310 {
 311         return pmd_set_flags(pmd, _PAGE_PSE);
 312 }
 313
 314 static inline pmd_t pmd_mkyoung(pmd_t pmd)
 315 {
 316         return pmd_set_flags(pmd, _PAGE_ACCESSED);
 317 }
 318
 319 static inline pmd_t pmd_mkwrite(pmd_t pmd)
 320 {
 321         return pmd_set_flags(pmd, _PAGE_RW);
 322 }
 323
 324 static inline pmd_t pmd_mknotpresent(pmd_t pmd)
 325 {
 326         return pmd_clear_flags(pmd, _PAGE_PRESENT | _PAGE_PROTNONE);
 327 }
 328
 329 #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY
 330 static inline int pte_soft_dirty(pte_t pte)
 331 {
 332         return pte_flags(pte) & _PAGE_SOFT_DIRTY;
 333 }
 334
 335 static inline int pmd_soft_dirty(pmd_t pmd)
 336 {
 337         return pmd_flags(pmd) & _PAGE_SOFT_DIRTY;
 338 }
 339
 340 static inline pte_t pte_mksoft_dirty(pte_t pte)
 341 {
 342         return pte_set_flags(pte, _PAGE_SOFT_DIRTY);
 343 }
 344
 345 static inline pmd_t pmd_mksoft_dirty(pmd_t pmd)
 346 {
 347         return pmd_set_flags(pmd, _PAGE_SOFT_DIRTY);
 348 }
 349
 350 static inline pte_t pte_clear_soft_dirty(pte_t pte)
 351 {
 352         return pte_clear_flags(pte, _PAGE_SOFT_DIRTY);
 353 }
 354
 355 static inline pmd_t pmd_clear_soft_dirty(pmd_t pmd)
 356 {
 357         return pmd_clear_flags(pmd, _PAGE_SOFT_DIRTY);
 358 }
 359
 360 #endif /* CONFIG_HAVE_ARCH_SOFT_DIRTY */
 361
 362 /*
 363  * Mask out unsupported bits in a present pgprot.  Non-present pgprots
 364  * can use those bits for other purposes, so leave them be.
 365  */
 366 static inline pgprotval_t massage_pgprot(pgprot_t pgprot)
 367 {
 368         pgprotval_t protval = pgprot_val(pgprot);
 369
 370         if (protval & _PAGE_PRESENT)
 371                 protval &= __supported_pte_mask;
 372
 373         return protval;
 374 }
 375
 376 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
 377 {
 378         return __pte(((phys_addr_t)page_nr << PAGE_SHIFT) |
 379                      massage_pgprot(pgprot));
 380 }
 381
 382 static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
 383 {
 384         return __pmd(((phys_addr_t)page_nr << PAGE_SHIFT) |
 385                      massage_pgprot(pgprot));
 386 }
 387
 388 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 389 {
 390         pteval_t val = pte_val(pte);
 391
 392         /*
 393          * Chop off the NX bit (if present), and add the NX portion of
 394          * the newprot (if present):
 395          */
 396         val &= _PAGE_CHG_MASK;
 397         val |= massage_pgprot(newprot) & ~_PAGE_CHG_MASK;
 398
 399         return __pte(val);
 400 }
 401
 402 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
 403 {
 404         pmdval_t val = pmd_val(pmd);
 405
 406         val &= _HPAGE_CHG_MASK;
 407         val |= massage_pgprot(newprot) & ~_HPAGE_CHG_MASK;
 408
 409         return __pmd(val);
 410 }
 411
 412 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
 413 #define pgprot_modify pgprot_modify
 414 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
 415 {
 416         pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
 417         pgprotval_t addbits = pgprot_val(newprot);
 418         return __pgprot(preservebits | addbits);
 419 }
 420
 421 #define pte_pgprot(x) __pgprot(pte_flags(x))
 422 #define pmd_pgprot(x) __pgprot(pmd_flags(x))
 423 #define pud_pgprot(x) __pgprot(pud_flags(x))
 424
 425 #define canon_pgprot(p) __pgprot(massage_pgprot(p))
 426
 427 static inline int is_new_memtype_allowed(u64 paddr, unsigned long size,
 428                                          enum page_cache_mode pcm,
 429                                          enum page_cache_mode new_pcm)
 430 {
 431         /*
 432          * PAT type is always WB for untracked ranges, so no need to check.
 433          */
 434         if (x86_platform.is_untracked_pat_range(paddr, paddr + size))
 435                 return 1;
 436
 437         /*
 438          * Certain new memtypes are not allowed with certain
 439          * requested memtype:
 440          * - request is uncached, return cannot be write-back
 441          * - request is write-combine, return cannot be write-back
 442          * - request is write-through, return cannot be write-back
 443          * - request is write-through, return cannot be write-combine
 444          */
 445         if ((pcm == _PAGE_CACHE_MODE_UC_MINUS &&
 446              new_pcm == _PAGE_CACHE_MODE_WB) ||
 447             (pcm == _PAGE_CACHE_MODE_WC &&
 448              new_pcm == _PAGE_CACHE_MODE_WB) ||
 449             (pcm == _PAGE_CACHE_MODE_WT &&
 450              new_pcm == _PAGE_CACHE_MODE_WB) ||
 451             (pcm == _PAGE_CACHE_MODE_WT &&
 452              new_pcm == _PAGE_CACHE_MODE_WC)) {
 453                 return 0;
 454         }
 455
 456         return 1;
 457 }
 458
 459 pmd_t *populate_extra_pmd(unsigned long vaddr);
 460 pte_t *populate_extra_pte(unsigned long vaddr);
 461 #endif  /* __ASSEMBLY__ */
 462
 463 #ifdef CONFIG_X86_32
 464 # include <asm/pgtable_32.h>
 465 #else
 466 # include <asm/pgtable_64.h>
 467 #endif
 468
 469 #ifndef __ASSEMBLY__
 470 #include <linux/mm_types.h>
 471 #include <linux/mmdebug.h>
 472 #include <linux/log2.h>
 473
 474 static inline int pte_none(pte_t pte)
 475 {
 476         return !pte.pte;
 477 }
 478
 479 #define __HAVE_ARCH_PTE_SAME
 480 static inline int pte_same(pte_t a, pte_t b)
 481 {
 482         return a.pte == b.pte;
 483 }
 484
 485 static inline int pte_present(pte_t a)
 486 {
 487         return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
 488 }
 489
 490 #ifdef __HAVE_ARCH_PTE_DEVMAP
 491 static inline int pte_devmap(pte_t a)
 492 {
 493         return (pte_flags(a) & _PAGE_DEVMAP) == _PAGE_DEVMAP;
 494 }
 495 #endif
 496
 497 #define pte_accessible pte_accessible
 498 static inline bool pte_accessible(struct mm_struct *mm, pte_t a)
 499 {
 500         if (pte_flags(a) & _PAGE_PRESENT)
 501                 return true;
 502
 503         if ((pte_flags(a) & _PAGE_PROTNONE) &&
 504                         mm_tlb_flush_pending(mm))
 505                 return true;
 506
 507         return false;
 508 }
 509
 510 static inline int pte_hidden(pte_t pte)
 511 {
 512         return pte_flags(pte) & _PAGE_HIDDEN;
 513 }
 514
 515 static inline int pmd_present(pmd_t pmd)
 516 {
 517         /*
 518          * Checking for _PAGE_PSE is needed too because
 519          * split_huge_page will temporarily clear the present bit (but
 520          * the _PAGE_PSE flag will remain set at all times while the
 521          * _PAGE_PRESENT bit is clear).
 522          */
 523         return pmd_flags(pmd) & (_PAGE_PRESENT | _PAGE_PROTNONE | _PAGE_PSE);
 524 }
 525
 526 #ifdef CONFIG_NUMA_BALANCING
 527 /*
 528  * These work without NUMA balancing but the kernel does not care. See the
 529  * comment in include/asm-generic/pgtable.h
 530  */
 531 static inline int pte_protnone(pte_t pte)
 532 {
 533         return (pte_flags(pte) & (_PAGE_PROTNONE | _PAGE_PRESENT))
 534                 == _PAGE_PROTNONE;
 535 }
 536
 537 static inline int pmd_protnone(pmd_t pmd)
 538 {
 539         return (pmd_flags(pmd) & (_PAGE_PROTNONE | _PAGE_PRESENT))
 540                 == _PAGE_PROTNONE;
 541 }
 542 #endif /* CONFIG_NUMA_BALANCING */
 543
 544 static inline int pmd_none(pmd_t pmd)
 545 {
 546         /* Only check low word on 32-bit platforms, since it might be
 547            out of sync with upper half. */
 548         return (unsigned long)native_pmd_val(pmd) == 0;
 549 }
 550
 551 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
 552 {
 553         return (unsigned long)__va(pmd_val(pmd) & pmd_pfn_mask(pmd));
 554 }
 555
 556 /*
 557  * Currently stuck as a macro due to indirect forward reference to
 558  * linux/mmzone.h's __section_mem_map_addr() definition:
 559  */
 560 #define pmd_page(pmd)           \
 561         pfn_to_page((pmd_val(pmd) & pmd_pfn_mask(pmd)) >> PAGE_SHIFT)
 562
 563 /*
 564  * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
 565  *
 566  * this macro returns the index of the entry in the pmd page which would
 567  * control the given virtual address
 568  */
 569 static inline unsigned long pmd_index(unsigned long address)
 570 {
 571         return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
 572 }
 573
 574 /*
 575  * Conversion functions: convert a page and protection to a page entry,
 576  * and a page entry and page directory to the page they refer to.
 577  *
 578  * (Currently stuck as a macro because of indirect forward reference
 579  * to linux/mm.h:page_to_nid())
 580  */
 581 #define mk_pte(page, pgprot)   pfn_pte(page_to_pfn(page), (pgprot))
 582
 583 /*
 584  * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
 585  *
 586  * this function returns the index of the entry in the pte page which would
 587  * control the given virtual address
 588  */
 589 static inline unsigned long pte_index(unsigned long address)
 590 {
 591         return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
 592 }
 593
 594 static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
 595 {
 596         return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
 597 }
 598
 599 static inline int pmd_bad(pmd_t pmd)
 600 {
 601         return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
 602 }
 603
 604 static inline unsigned long pages_to_mb(unsigned long npg)
 605 {
 606         return npg >> (20 - PAGE_SHIFT);
 607 }
 608
 609 #if CONFIG_PGTABLE_LEVELS > 2
 610 static inline int pud_none(pud_t pud)
 611 {
 612         return native_pud_val(pud) == 0;
 613 }
 614
 615 static inline int pud_present(pud_t pud)
 616 {
 617         return pud_flags(pud) & _PAGE_PRESENT;
 618 }
 619
 620 static inline unsigned long pud_page_vaddr(pud_t pud)
 621 {
 622         return (unsigned long)__va(pud_val(pud) & pud_pfn_mask(pud));
 623 }
 624
 625 /*
 626  * Currently stuck as a macro due to indirect forward reference to
 627  * linux/mmzone.h's __section_mem_map_addr() definition:
 628  */
 629 #define pud_page(pud)           \
 630         pfn_to_page((pud_val(pud) & pud_pfn_mask(pud)) >> PAGE_SHIFT)
 631
 632 /* Find an entry in the second-level page table.. */
 633 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
 634 {
 635         return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
 636 }
 637
 638 static inline int pud_large(pud_t pud)
 639 {
 640         return (pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
 641                 (_PAGE_PSE | _PAGE_PRESENT);
 642 }
 643
 644 static inline int pud_bad(pud_t pud)
 645 {
 646         return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
 647 }
 648 #else
 649 static inline int pud_large(pud_t pud)
 650 {
 651         return 0;
 652 }
 653 #endif  /* CONFIG_PGTABLE_LEVELS > 2 */
 654
 655 #if CONFIG_PGTABLE_LEVELS > 3
 656 static inline int pgd_present(pgd_t pgd)
 657 {
 658         return pgd_flags(pgd) & _PAGE_PRESENT;
 659 }
 660
 661 static inline unsigned long pgd_page_vaddr(pgd_t pgd)
 662 {
 663         return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
 664 }
 665
 666 /*
 667  * Currently stuck as a macro due to indirect forward reference to
 668  * linux/mmzone.h's __section_mem_map_addr() definition:
 669  */
 670 #define pgd_page(pgd)           pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
 671
 672 /* to find an entry in a page-table-directory. */
 673 static inline unsigned long pud_index(unsigned long address)
 674 {
 675         return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
 676 }
 677
 678 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
 679 {
 680         return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
 681 }
 682
 683 static inline int pgd_bad(pgd_t pgd)
 684 {
 685         return (pgd_flags(pgd) & ~_PAGE_USER) != _KERNPG_TABLE;
 686 }
 687
 688 static inline int pgd_none(pgd_t pgd)
 689 {
 690         return !native_pgd_val(pgd);
 691 }
 692 #endif  /* CONFIG_PGTABLE_LEVELS > 3 */
 693
 694 #endif  /* __ASSEMBLY__ */
 695
 696 /*
 697  * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
 698  *
 699  * this macro returns the index of the entry in the pgd page which would
 700  * control the given virtual address
 701  */
 702 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
 703
 704 /*
 705  * pgd_offset() returns a (pgd_t *)
 706  * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
 707  */
 708 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
 709 /*
 710  * a shortcut which implies the use of the kernel's pgd, instead
 711  * of a process's
 712  */
 713 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
 714
 715
 716 #define KERNEL_PGD_BOUNDARY     pgd_index(PAGE_OFFSET)
 717 #define KERNEL_PGD_PTRS         (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
 718
 719 #ifndef __ASSEMBLY__
 720
 721 extern int direct_gbpages;
 722 void init_mem_mapping(void);
 723 void early_alloc_pgt_buf(void);
 724
 725 /* local pte updates need not use xchg for locking */
 726 static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
 727 {
 728         pte_t res = *ptep;
 729
 730         /* Pure native function needs no input for mm, addr */
 731         native_pte_clear(NULL, 0, ptep);
 732         return res;
 733 }
 734
 735 static inline pmd_t native_local_pmdp_get_and_clear(pmd_t *pmdp)
 736 {
 737         pmd_t res = *pmdp;
 738
 739         native_pmd_clear(pmdp);
 740         return res;
 741 }
 742
 743 static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
 744                                      pte_t *ptep , pte_t pte)
 745 {
 746         native_set_pte(ptep, pte);
 747 }
 748
 749 static inline void native_set_pmd_at(struct mm_struct *mm, unsigned long addr,
 750                                      pmd_t *pmdp , pmd_t pmd)
 751 {
 752         native_set_pmd(pmdp, pmd);
 753 }
 754
 755 #ifndef CONFIG_PARAVIRT
 756 /*
 757  * Rules for using pte_update - it must be called after any PTE update which
 758  * has not been done using the set_pte / clear_pte interfaces.  It is used by
 759  * shadow mode hypervisors to resynchronize the shadow page tables.  Kernel PTE
 760  * updates should either be sets, clears, or set_pte_atomic for P->P
 761  * transitions, which means this hook should only be called for user PTEs.
 762  * This hook implies a P->P protection or access change has taken place, which
 763  * requires a subsequent TLB flush.
 764  */
 765 #define pte_update(mm, addr, ptep)              do { } while (0)
 766 #endif
 767
 768 /*
 769  * We only update the dirty/accessed state if we set
 770  * the dirty bit by hand in the kernel, since the hardware
 771  * will do the accessed bit for us, and we don't want to
 772  * race with other CPU's that might be updating the dirty
 773  * bit at the same time.
 774  */
 775 struct vm_area_struct;
 776
 777 #define  __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
 778 extern int ptep_set_access_flags(struct vm_area_struct *vma,
 779                                  unsigned long address, pte_t *ptep,
 780                                  pte_t entry, int dirty);
 781
 782 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
 783 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
 784                                      unsigned long addr, pte_t *ptep);
 785
 786 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
 787 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
 788                                   unsigned long address, pte_t *ptep);
 789
 790 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
 791 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
 792                                        pte_t *ptep)
 793 {
 794         pte_t pte = native_ptep_get_and_clear(ptep);
 795         pte_update(mm, addr, ptep);
 796         return pte;
 797 }
 798
 799 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
 800 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
 801                                             unsigned long addr, pte_t *ptep,
 802                                             int full)
 803 {
 804         pte_t pte;
 805         if (full) {
 806                 /*
 807                  * Full address destruction in progress; paravirt does not
 808                  * care about updates and native needs no locking
 809                  */
 810                 pte = native_local_ptep_get_and_clear(ptep);
 811         } else {
 812                 pte = ptep_get_and_clear(mm, addr, ptep);
 813         }
 814         return pte;
 815 }
 816
 817 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
 818 static inline void ptep_set_wrprotect(struct mm_struct *mm,
 819                                       unsigned long addr, pte_t *ptep)
 820 {
 821         clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
 822         pte_update(mm, addr, ptep);
 823 }
 824
 825 #define flush_tlb_fix_spurious_fault(vma, address) do { } while (0)
 826
 827 #define mk_pmd(page, pgprot)   pfn_pmd(page_to_pfn(page), (pgprot))
 828
 829 #define  __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
 830 extern int pmdp_set_access_flags(struct vm_area_struct *vma,
 831                                  unsigned long address, pmd_t *pmdp,
 832                                  pmd_t entry, int dirty);
 833
 834 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
 835 extern int pmdp_test_and_clear_young(struct vm_area_struct *vma,
 836                                      unsigned long addr, pmd_t *pmdp);
 837
 838 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
 839 extern int pmdp_clear_flush_young(struct vm_area_struct *vma,
 840                                   unsigned long address, pmd_t *pmdp);
 841
 842
 843 #define __HAVE_ARCH_PMD_WRITE
 844 static inline int pmd_write(pmd_t pmd)
 845 {
 846         return pmd_flags(pmd) & _PAGE_RW;
 847 }
 848
 849 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
 850 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm, unsigned long addr,
 851                                        pmd_t *pmdp)
 852 {
 853         return native_pmdp_get_and_clear(pmdp);
 854 }
 855
 856 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
 857 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
 858                                       unsigned long addr, pmd_t *pmdp)
 859 {
 860         clear_bit(_PAGE_BIT_RW, (unsigned long *)pmdp);
 861 }
 862
 863 /*
 864  * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
 865  *
 866  *  dst - pointer to pgd range anwhere on a pgd page
 867  *  src - ""
 868  *  count - the number of pgds to copy.
 869  *
 870  * dst and src can be on the same page, but the range must not overlap,
 871  * and must not cross a page boundary.
 872  */
 873 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
 874 {
 875        memcpy(dst, src, count * sizeof(pgd_t));
 876 }
 877
 878 #define PTE_SHIFT ilog2(PTRS_PER_PTE)
 879 static inline int page_level_shift(enum pg_level level)
 880 {
 881         return (PAGE_SHIFT - PTE_SHIFT) + level * PTE_SHIFT;
 882 }
 883 static inline unsigned long page_level_size(enum pg_level level)
 884 {
 885         return 1UL << page_level_shift(level);
 886 }
 887 static inline unsigned long page_level_mask(enum pg_level level)
 888 {
 889         return ~(page_level_size(level) - 1);
 890 }
 891
 892 /*
 893  * The x86 doesn't have any external MMU info: the kernel page
 894  * tables contain all the necessary information.
 895  */
 896 static inline void update_mmu_cache(struct vm_area_struct *vma,
 897                 unsigned long addr, pte_t *ptep)
 898 {
 899 }
 900 static inline void update_mmu_cache_pmd(struct vm_area_struct *vma,
 901                 unsigned long addr, pmd_t *pmd)
 902 {
 903 }
 904
 905 #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY
 906 static inline pte_t pte_swp_mksoft_dirty(pte_t pte)
 907 {
 908         return pte_set_flags(pte, _PAGE_SWP_SOFT_DIRTY);
 909 }
 910
 911 static inline int pte_swp_soft_dirty(pte_t pte)
 912 {
 913         return pte_flags(pte) & _PAGE_SWP_SOFT_DIRTY;
 914 }
 915
 916 static inline pte_t pte_swp_clear_soft_dirty(pte_t pte)
 917 {
 918         return pte_clear_flags(pte, _PAGE_SWP_SOFT_DIRTY);
 919 }
 920 #endif
 921
 922 #include <asm-generic/pgtable.h>
 923 #endif  /* __ASSEMBLY__ */
 924
 925 #endif /* _ASM_X86_PGTABLE_H */