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 void write_pkru(u32 pkru)
 111 {
 112         if (boot_cpu_has(X86_FEATURE_OSPKE))
 113                 __write_pkru(pkru);
 114 }
 115
 116 static inline int pte_young(pte_t pte)
 117 {
 118         return pte_flags(pte) & _PAGE_ACCESSED;
 119 }
 120
 121 static inline int pmd_dirty(pmd_t pmd)
 122 {
 123         return pmd_flags(pmd) & _PAGE_DIRTY;
 124 }
 125
 126 static inline int pmd_young(pmd_t pmd)
 127 {
 128         return pmd_flags(pmd) & _PAGE_ACCESSED;
 129 }
 130
 131 static inline int pte_write(pte_t pte)
 132 {
 133         return pte_flags(pte) & _PAGE_RW;
 134 }
 135
 136 static inline int pte_huge(pte_t pte)
 137 {
 138         return pte_flags(pte) & _PAGE_PSE;
 139 }
 140
 141 static inline int pte_global(pte_t pte)
 142 {
 143         return pte_flags(pte) & _PAGE_GLOBAL;
 144 }
 145
 146 static inline int pte_exec(pte_t pte)
 147 {
 148         return !(pte_flags(pte) & _PAGE_NX);
 149 }
 150
 151 static inline int pte_special(pte_t pte)
 152 {
 153         return pte_flags(pte) & _PAGE_SPECIAL;
 154 }
 155
 156 static inline unsigned long pte_pfn(pte_t pte)
 157 {
 158         return (pte_val(pte) & PTE_PFN_MASK) >> PAGE_SHIFT;
 159 }
 160
 161 static inline unsigned long pmd_pfn(pmd_t pmd)
 162 {
 163         return (pmd_val(pmd) & pmd_pfn_mask(pmd)) >> PAGE_SHIFT;
 164 }
 165
 166 static inline unsigned long pud_pfn(pud_t pud)
 167 {
 168         return (pud_val(pud) & pud_pfn_mask(pud)) >> PAGE_SHIFT;
 169 }
 170
 171 #define pte_page(pte)   pfn_to_page(pte_pfn(pte))
 172
 173 static inline int pmd_large(pmd_t pte)
 174 {
 175         return pmd_flags(pte) & _PAGE_PSE;
 176 }
 177
 178 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 179 static inline int pmd_trans_huge(pmd_t pmd)
 180 {
 181         return (pmd_val(pmd) & (_PAGE_PSE|_PAGE_DEVMAP)) == _PAGE_PSE;
 182 }
 183
 184 static inline int has_transparent_hugepage(void)
 185 {
 186         return cpu_has_pse;
 187 }
 188
 189 #ifdef __HAVE_ARCH_PTE_DEVMAP
 190 static inline int pmd_devmap(pmd_t pmd)
 191 {
 192         return !!(pmd_val(pmd) & _PAGE_DEVMAP);
 193 }
 194 #endif
 195 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
 196
 197 static inline pte_t pte_set_flags(pte_t pte, pteval_t set)
 198 {
 199         pteval_t v = native_pte_val(pte);
 200
 201         return native_make_pte(v | set);
 202 }
 203
 204 static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
 205 {
 206         pteval_t v = native_pte_val(pte);
 207
 208         return native_make_pte(v & ~clear);
 209 }
 210
 211 static inline pte_t pte_mkclean(pte_t pte)
 212 {
 213         return pte_clear_flags(pte, _PAGE_DIRTY);
 214 }
 215
 216 static inline pte_t pte_mkold(pte_t pte)
 217 {
 218         return pte_clear_flags(pte, _PAGE_ACCESSED);
 219 }
 220
 221 static inline pte_t pte_wrprotect(pte_t pte)
 222 {
 223         return pte_clear_flags(pte, _PAGE_RW);
 224 }
 225
 226 static inline pte_t pte_mkexec(pte_t pte)
 227 {
 228         return pte_clear_flags(pte, _PAGE_NX);
 229 }
 230
 231 static inline pte_t pte_mkdirty(pte_t pte)
 232 {
 233         return pte_set_flags(pte, _PAGE_DIRTY | _PAGE_SOFT_DIRTY);
 234 }
 235
 236 static inline pte_t pte_mkyoung(pte_t pte)
 237 {
 238         return pte_set_flags(pte, _PAGE_ACCESSED);
 239 }
 240
 241 static inline pte_t pte_mkwrite(pte_t pte)
 242 {
 243         return pte_set_flags(pte, _PAGE_RW);
 244 }
 245
 246 static inline pte_t pte_mkhuge(pte_t pte)
 247 {
 248         return pte_set_flags(pte, _PAGE_PSE);
 249 }
 250
 251 static inline pte_t pte_clrhuge(pte_t pte)
 252 {
 253         return pte_clear_flags(pte, _PAGE_PSE);
 254 }
 255
 256 static inline pte_t pte_mkglobal(pte_t pte)
 257 {
 258         return pte_set_flags(pte, _PAGE_GLOBAL);
 259 }
 260
 261 static inline pte_t pte_clrglobal(pte_t pte)
 262 {
 263         return pte_clear_flags(pte, _PAGE_GLOBAL);
 264 }
 265
 266 static inline pte_t pte_mkspecial(pte_t pte)
 267 {
 268         return pte_set_flags(pte, _PAGE_SPECIAL);
 269 }
 270
 271 static inline pte_t pte_mkdevmap(pte_t pte)
 272 {
 273         return pte_set_flags(pte, _PAGE_SPECIAL|_PAGE_DEVMAP);
 274 }
 275
 276 static inline pmd_t pmd_set_flags(pmd_t pmd, pmdval_t set)
 277 {
 278         pmdval_t v = native_pmd_val(pmd);
 279
 280         return __pmd(v | set);
 281 }
 282
 283 static inline pmd_t pmd_clear_flags(pmd_t pmd, pmdval_t clear)
 284 {
 285         pmdval_t v = native_pmd_val(pmd);
 286
 287         return __pmd(v & ~clear);
 288 }
 289
 290 static inline pmd_t pmd_mkold(pmd_t pmd)
 291 {
 292         return pmd_clear_flags(pmd, _PAGE_ACCESSED);
 293 }
 294
 295 static inline pmd_t pmd_mkclean(pmd_t pmd)
 296 {
 297         return pmd_clear_flags(pmd, _PAGE_DIRTY);
 298 }
 299
 300 static inline pmd_t pmd_wrprotect(pmd_t pmd)
 301 {
 302         return pmd_clear_flags(pmd, _PAGE_RW);
 303 }
 304
 305 static inline pmd_t pmd_mkdirty(pmd_t pmd)
 306 {
 307         return pmd_set_flags(pmd, _PAGE_DIRTY | _PAGE_SOFT_DIRTY);
 308 }
 309
 310 static inline pmd_t pmd_mkdevmap(pmd_t pmd)
 311 {
 312         return pmd_set_flags(pmd, _PAGE_DEVMAP);
 313 }
 314
 315 static inline pmd_t pmd_mkhuge(pmd_t pmd)
 316 {
 317         return pmd_set_flags(pmd, _PAGE_PSE);
 318 }
 319
 320 static inline pmd_t pmd_mkyoung(pmd_t pmd)
 321 {
 322         return pmd_set_flags(pmd, _PAGE_ACCESSED);
 323 }
 324
 325 static inline pmd_t pmd_mkwrite(pmd_t pmd)
 326 {
 327         return pmd_set_flags(pmd, _PAGE_RW);
 328 }
 329
 330 static inline pmd_t pmd_mknotpresent(pmd_t pmd)
 331 {
 332         return pmd_clear_flags(pmd, _PAGE_PRESENT | _PAGE_PROTNONE);
 333 }
 334
 335 #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY
 336 static inline int pte_soft_dirty(pte_t pte)
 337 {
 338         return pte_flags(pte) & _PAGE_SOFT_DIRTY;
 339 }
 340
 341 static inline int pmd_soft_dirty(pmd_t pmd)
 342 {
 343         return pmd_flags(pmd) & _PAGE_SOFT_DIRTY;
 344 }
 345
 346 static inline pte_t pte_mksoft_dirty(pte_t pte)
 347 {
 348         return pte_set_flags(pte, _PAGE_SOFT_DIRTY);
 349 }
 350
 351 static inline pmd_t pmd_mksoft_dirty(pmd_t pmd)
 352 {
 353         return pmd_set_flags(pmd, _PAGE_SOFT_DIRTY);
 354 }
 355
 356 static inline pte_t pte_clear_soft_dirty(pte_t pte)
 357 {
 358         return pte_clear_flags(pte, _PAGE_SOFT_DIRTY);
 359 }
 360
 361 static inline pmd_t pmd_clear_soft_dirty(pmd_t pmd)
 362 {
 363         return pmd_clear_flags(pmd, _PAGE_SOFT_DIRTY);
 364 }
 365
 366 #endif /* CONFIG_HAVE_ARCH_SOFT_DIRTY */
 367
 368 /*
 369  * Mask out unsupported bits in a present pgprot.  Non-present pgprots
 370  * can use those bits for other purposes, so leave them be.
 371  */
 372 static inline pgprotval_t massage_pgprot(pgprot_t pgprot)
 373 {
 374         pgprotval_t protval = pgprot_val(pgprot);
 375
 376         if (protval & _PAGE_PRESENT)
 377                 protval &= __supported_pte_mask;
 378
 379         return protval;
 380 }
 381
 382 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
 383 {
 384         return __pte(((phys_addr_t)page_nr << PAGE_SHIFT) |
 385                      massage_pgprot(pgprot));
 386 }
 387
 388 static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
 389 {
 390         return __pmd(((phys_addr_t)page_nr << PAGE_SHIFT) |
 391                      massage_pgprot(pgprot));
 392 }
 393
 394 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 395 {
 396         pteval_t val = pte_val(pte);
 397
 398         /*
 399          * Chop off the NX bit (if present), and add the NX portion of
 400          * the newprot (if present):
 401          */
 402         val &= _PAGE_CHG_MASK;
 403         val |= massage_pgprot(newprot) & ~_PAGE_CHG_MASK;
 404
 405         return __pte(val);
 406 }
 407
 408 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
 409 {
 410         pmdval_t val = pmd_val(pmd);
 411
 412         val &= _HPAGE_CHG_MASK;
 413         val |= massage_pgprot(newprot) & ~_HPAGE_CHG_MASK;
 414
 415         return __pmd(val);
 416 }
 417
 418 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
 419 #define pgprot_modify pgprot_modify
 420 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
 421 {
 422         pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
 423         pgprotval_t addbits = pgprot_val(newprot);
 424         return __pgprot(preservebits | addbits);
 425 }
 426
 427 #define pte_pgprot(x) __pgprot(pte_flags(x))
 428 #define pmd_pgprot(x) __pgprot(pmd_flags(x))
 429 #define pud_pgprot(x) __pgprot(pud_flags(x))
 430
 431 #define canon_pgprot(p) __pgprot(massage_pgprot(p))
 432
 433 static inline int is_new_memtype_allowed(u64 paddr, unsigned long size,
 434                                          enum page_cache_mode pcm,
 435                                          enum page_cache_mode new_pcm)
 436 {
 437         /*
 438          * PAT type is always WB for untracked ranges, so no need to check.
 439          */
 440         if (x86_platform.is_untracked_pat_range(paddr, paddr + size))
 441                 return 1;
 442
 443         /*
 444          * Certain new memtypes are not allowed with certain
 445          * requested memtype:
 446          * - request is uncached, return cannot be write-back
 447          * - request is write-combine, return cannot be write-back
 448          * - request is write-through, return cannot be write-back
 449          * - request is write-through, return cannot be write-combine
 450          */
 451         if ((pcm == _PAGE_CACHE_MODE_UC_MINUS &&
 452              new_pcm == _PAGE_CACHE_MODE_WB) ||
 453             (pcm == _PAGE_CACHE_MODE_WC &&
 454              new_pcm == _PAGE_CACHE_MODE_WB) ||
 455             (pcm == _PAGE_CACHE_MODE_WT &&
 456              new_pcm == _PAGE_CACHE_MODE_WB) ||
 457             (pcm == _PAGE_CACHE_MODE_WT &&
 458              new_pcm == _PAGE_CACHE_MODE_WC)) {
 459                 return 0;
 460         }
 461
 462         return 1;
 463 }
 464
 465 pmd_t *populate_extra_pmd(unsigned long vaddr);
 466 pte_t *populate_extra_pte(unsigned long vaddr);
 467 #endif  /* __ASSEMBLY__ */
 468
 469 #ifdef CONFIG_X86_32
 470 # include <asm/pgtable_32.h>
 471 #else
 472 # include <asm/pgtable_64.h>
 473 #endif
 474
 475 #ifndef __ASSEMBLY__
 476 #include <linux/mm_types.h>
 477 #include <linux/mmdebug.h>
 478 #include <linux/log2.h>
 479
 480 static inline int pte_none(pte_t pte)
 481 {
 482         return !pte.pte;
 483 }
 484
 485 #define __HAVE_ARCH_PTE_SAME
 486 static inline int pte_same(pte_t a, pte_t b)
 487 {
 488         return a.pte == b.pte;
 489 }
 490
 491 static inline int pte_present(pte_t a)
 492 {
 493         return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
 494 }
 495
 496 #ifdef __HAVE_ARCH_PTE_DEVMAP
 497 static inline int pte_devmap(pte_t a)
 498 {
 499         return (pte_flags(a) & _PAGE_DEVMAP) == _PAGE_DEVMAP;
 500 }
 501 #endif
 502
 503 #define pte_accessible pte_accessible
 504 static inline bool pte_accessible(struct mm_struct *mm, pte_t a)
 505 {
 506         if (pte_flags(a) & _PAGE_PRESENT)
 507                 return true;
 508
 509         if ((pte_flags(a) & _PAGE_PROTNONE) &&
 510                         mm_tlb_flush_pending(mm))
 511                 return true;
 512
 513         return false;
 514 }
 515
 516 static inline int pte_hidden(pte_t pte)
 517 {
 518         return pte_flags(pte) & _PAGE_HIDDEN;
 519 }
 520
 521 static inline int pmd_present(pmd_t pmd)
 522 {
 523         /*
 524          * Checking for _PAGE_PSE is needed too because
 525          * split_huge_page will temporarily clear the present bit (but
 526          * the _PAGE_PSE flag will remain set at all times while the
 527          * _PAGE_PRESENT bit is clear).
 528          */
 529         return pmd_flags(pmd) & (_PAGE_PRESENT | _PAGE_PROTNONE | _PAGE_PSE);
 530 }
 531
 532 #ifdef CONFIG_NUMA_BALANCING
 533 /*
 534  * These work without NUMA balancing but the kernel does not care. See the
 535  * comment in include/asm-generic/pgtable.h
 536  */
 537 static inline int pte_protnone(pte_t pte)
 538 {
 539         return (pte_flags(pte) & (_PAGE_PROTNONE | _PAGE_PRESENT))
 540                 == _PAGE_PROTNONE;
 541 }
 542
 543 static inline int pmd_protnone(pmd_t pmd)
 544 {
 545         return (pmd_flags(pmd) & (_PAGE_PROTNONE | _PAGE_PRESENT))
 546                 == _PAGE_PROTNONE;
 547 }
 548 #endif /* CONFIG_NUMA_BALANCING */
 549
 550 static inline int pmd_none(pmd_t pmd)
 551 {
 552         /* Only check low word on 32-bit platforms, since it might be
 553            out of sync with upper half. */
 554         return (unsigned long)native_pmd_val(pmd) == 0;
 555 }
 556
 557 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
 558 {
 559         return (unsigned long)__va(pmd_val(pmd) & pmd_pfn_mask(pmd));
 560 }
 561
 562 /*
 563  * Currently stuck as a macro due to indirect forward reference to
 564  * linux/mmzone.h's __section_mem_map_addr() definition:
 565  */
 566 #define pmd_page(pmd)           \
 567         pfn_to_page((pmd_val(pmd) & pmd_pfn_mask(pmd)) >> PAGE_SHIFT)
 568
 569 /*
 570  * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
 571  *
 572  * this macro returns the index of the entry in the pmd page which would
 573  * control the given virtual address
 574  */
 575 static inline unsigned long pmd_index(unsigned long address)
 576 {
 577         return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
 578 }
 579
 580 /*
 581  * Conversion functions: convert a page and protection to a page entry,
 582  * and a page entry and page directory to the page they refer to.
 583  *
 584  * (Currently stuck as a macro because of indirect forward reference
 585  * to linux/mm.h:page_to_nid())
 586  */
 587 #define mk_pte(page, pgprot)   pfn_pte(page_to_pfn(page), (pgprot))
 588
 589 /*
 590  * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
 591  *
 592  * this function returns the index of the entry in the pte page which would
 593  * control the given virtual address
 594  */
 595 static inline unsigned long pte_index(unsigned long address)
 596 {
 597         return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
 598 }
 599
 600 static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
 601 {
 602         return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
 603 }
 604
 605 static inline int pmd_bad(pmd_t pmd)
 606 {
 607         return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
 608 }
 609
 610 static inline unsigned long pages_to_mb(unsigned long npg)
 611 {
 612         return npg >> (20 - PAGE_SHIFT);
 613 }
 614
 615 #if CONFIG_PGTABLE_LEVELS > 2
 616 static inline int pud_none(pud_t pud)
 617 {
 618         return native_pud_val(pud) == 0;
 619 }
 620
 621 static inline int pud_present(pud_t pud)
 622 {
 623         return pud_flags(pud) & _PAGE_PRESENT;
 624 }
 625
 626 static inline unsigned long pud_page_vaddr(pud_t pud)
 627 {
 628         return (unsigned long)__va(pud_val(pud) & pud_pfn_mask(pud));
 629 }
 630
 631 /*
 632  * Currently stuck as a macro due to indirect forward reference to
 633  * linux/mmzone.h's __section_mem_map_addr() definition:
 634  */
 635 #define pud_page(pud)           \
 636         pfn_to_page((pud_val(pud) & pud_pfn_mask(pud)) >> PAGE_SHIFT)
 637
 638 /* Find an entry in the second-level page table.. */
 639 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
 640 {
 641         return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
 642 }
 643
 644 static inline int pud_large(pud_t pud)
 645 {
 646         return (pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
 647                 (_PAGE_PSE | _PAGE_PRESENT);
 648 }
 649
 650 static inline int pud_bad(pud_t pud)
 651 {
 652         return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
 653 }
 654 #else
 655 static inline int pud_large(pud_t pud)
 656 {
 657         return 0;
 658 }
 659 #endif  /* CONFIG_PGTABLE_LEVELS > 2 */
 660
 661 #if CONFIG_PGTABLE_LEVELS > 3
 662 static inline int pgd_present(pgd_t pgd)
 663 {
 664         return pgd_flags(pgd) & _PAGE_PRESENT;
 665 }
 666
 667 static inline unsigned long pgd_page_vaddr(pgd_t pgd)
 668 {
 669         return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
 670 }
 671
 672 /*
 673  * Currently stuck as a macro due to indirect forward reference to
 674  * linux/mmzone.h's __section_mem_map_addr() definition:
 675  */
 676 #define pgd_page(pgd)           pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
 677
 678 /* to find an entry in a page-table-directory. */
 679 static inline unsigned long pud_index(unsigned long address)
 680 {
 681         return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
 682 }
 683
 684 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
 685 {
 686         return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
 687 }
 688
 689 static inline int pgd_bad(pgd_t pgd)
 690 {
 691         return (pgd_flags(pgd) & ~_PAGE_USER) != _KERNPG_TABLE;
 692 }
 693
 694 static inline int pgd_none(pgd_t pgd)
 695 {
 696         return !native_pgd_val(pgd);
 697 }
 698 #endif  /* CONFIG_PGTABLE_LEVELS > 3 */
 699
 700 #endif  /* __ASSEMBLY__ */
 701
 702 /*
 703  * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
 704  *
 705  * this macro returns the index of the entry in the pgd page which would
 706  * control the given virtual address
 707  */
 708 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
 709
 710 /*
 711  * pgd_offset() returns a (pgd_t *)
 712  * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
 713  */
 714 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
 715 /*
 716  * a shortcut which implies the use of the kernel's pgd, instead
 717  * of a process's
 718  */
 719 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
 720
 721
 722 #define KERNEL_PGD_BOUNDARY     pgd_index(PAGE_OFFSET)
 723 #define KERNEL_PGD_PTRS         (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
 724
 725 #ifndef __ASSEMBLY__
 726
 727 extern int direct_gbpages;
 728 void init_mem_mapping(void);
 729 void early_alloc_pgt_buf(void);
 730
 731 /* local pte updates need not use xchg for locking */
 732 static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
 733 {
 734         pte_t res = *ptep;
 735
 736         /* Pure native function needs no input for mm, addr */
 737         native_pte_clear(NULL, 0, ptep);
 738         return res;
 739 }
 740
 741 static inline pmd_t native_local_pmdp_get_and_clear(pmd_t *pmdp)
 742 {
 743         pmd_t res = *pmdp;
 744
 745         native_pmd_clear(pmdp);
 746         return res;
 747 }
 748
 749 static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
 750                                      pte_t *ptep , pte_t pte)
 751 {
 752         native_set_pte(ptep, pte);
 753 }
 754
 755 static inline void native_set_pmd_at(struct mm_struct *mm, unsigned long addr,
 756                                      pmd_t *pmdp , pmd_t pmd)
 757 {
 758         native_set_pmd(pmdp, pmd);
 759 }
 760
 761 #ifndef CONFIG_PARAVIRT
 762 /*
 763  * Rules for using pte_update - it must be called after any PTE update which
 764  * has not been done using the set_pte / clear_pte interfaces.  It is used by
 765  * shadow mode hypervisors to resynchronize the shadow page tables.  Kernel PTE
 766  * updates should either be sets, clears, or set_pte_atomic for P->P
 767  * transitions, which means this hook should only be called for user PTEs.
 768  * This hook implies a P->P protection or access change has taken place, which
 769  * requires a subsequent TLB flush.
 770  */
 771 #define pte_update(mm, addr, ptep)              do { } while (0)
 772 #endif
 773
 774 /*
 775  * We only update the dirty/accessed state if we set
 776  * the dirty bit by hand in the kernel, since the hardware
 777  * will do the accessed bit for us, and we don't want to
 778  * race with other CPU's that might be updating the dirty
 779  * bit at the same time.
 780  */
 781 struct vm_area_struct;
 782
 783 #define  __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
 784 extern int ptep_set_access_flags(struct vm_area_struct *vma,
 785                                  unsigned long address, pte_t *ptep,
 786                                  pte_t entry, int dirty);
 787
 788 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
 789 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
 790                                      unsigned long addr, pte_t *ptep);
 791
 792 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
 793 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
 794                                   unsigned long address, pte_t *ptep);
 795
 796 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
 797 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
 798                                        pte_t *ptep)
 799 {
 800         pte_t pte = native_ptep_get_and_clear(ptep);
 801         pte_update(mm, addr, ptep);
 802         return pte;
 803 }
 804
 805 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
 806 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
 807                                             unsigned long addr, pte_t *ptep,
 808                                             int full)
 809 {
 810         pte_t pte;
 811         if (full) {
 812                 /*
 813                  * Full address destruction in progress; paravirt does not
 814                  * care about updates and native needs no locking
 815                  */
 816                 pte = native_local_ptep_get_and_clear(ptep);
 817         } else {
 818                 pte = ptep_get_and_clear(mm, addr, ptep);
 819         }
 820         return pte;
 821 }
 822
 823 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
 824 static inline void ptep_set_wrprotect(struct mm_struct *mm,
 825                                       unsigned long addr, pte_t *ptep)
 826 {
 827         clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
 828         pte_update(mm, addr, ptep);
 829 }
 830
 831 #define flush_tlb_fix_spurious_fault(vma, address) do { } while (0)
 832
 833 #define mk_pmd(page, pgprot)   pfn_pmd(page_to_pfn(page), (pgprot))
 834
 835 #define  __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
 836 extern int pmdp_set_access_flags(struct vm_area_struct *vma,
 837                                  unsigned long address, pmd_t *pmdp,
 838                                  pmd_t entry, int dirty);
 839
 840 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
 841 extern int pmdp_test_and_clear_young(struct vm_area_struct *vma,
 842                                      unsigned long addr, pmd_t *pmdp);
 843
 844 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
 845 extern int pmdp_clear_flush_young(struct vm_area_struct *vma,
 846                                   unsigned long address, pmd_t *pmdp);
 847
 848
 849 #define __HAVE_ARCH_PMD_WRITE
 850 static inline int pmd_write(pmd_t pmd)
 851 {
 852         return pmd_flags(pmd) & _PAGE_RW;
 853 }
 854
 855 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
 856 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm, unsigned long addr,
 857                                        pmd_t *pmdp)
 858 {
 859         return native_pmdp_get_and_clear(pmdp);
 860 }
 861
 862 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
 863 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
 864                                       unsigned long addr, pmd_t *pmdp)
 865 {
 866         clear_bit(_PAGE_BIT_RW, (unsigned long *)pmdp);
 867 }
 868
 869 /*
 870  * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
 871  *
 872  *  dst - pointer to pgd range anwhere on a pgd page
 873  *  src - ""
 874  *  count - the number of pgds to copy.
 875  *
 876  * dst and src can be on the same page, but the range must not overlap,
 877  * and must not cross a page boundary.
 878  */
 879 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
 880 {
 881        memcpy(dst, src, count * sizeof(pgd_t));
 882 }
 883
 884 #define PTE_SHIFT ilog2(PTRS_PER_PTE)
 885 static inline int page_level_shift(enum pg_level level)
 886 {
 887         return (PAGE_SHIFT - PTE_SHIFT) + level * PTE_SHIFT;
 888 }
 889 static inline unsigned long page_level_size(enum pg_level level)
 890 {
 891         return 1UL << page_level_shift(level);
 892 }
 893 static inline unsigned long page_level_mask(enum pg_level level)
 894 {
 895         return ~(page_level_size(level) - 1);
 896 }
 897
 898 /*
 899  * The x86 doesn't have any external MMU info: the kernel page
 900  * tables contain all the necessary information.
 901  */
 902 static inline void update_mmu_cache(struct vm_area_struct *vma,
 903                 unsigned long addr, pte_t *ptep)
 904 {
 905 }
 906 static inline void update_mmu_cache_pmd(struct vm_area_struct *vma,
 907                 unsigned long addr, pmd_t *pmd)
 908 {
 909 }
 910
 911 #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY
 912 static inline pte_t pte_swp_mksoft_dirty(pte_t pte)
 913 {
 914         return pte_set_flags(pte, _PAGE_SWP_SOFT_DIRTY);
 915 }
 916
 917 static inline int pte_swp_soft_dirty(pte_t pte)
 918 {
 919         return pte_flags(pte) & _PAGE_SWP_SOFT_DIRTY;
 920 }
 921
 922 static inline pte_t pte_swp_clear_soft_dirty(pte_t pte)
 923 {
 924         return pte_clear_flags(pte, _PAGE_SWP_SOFT_DIRTY);
 925 }
 926 #endif
 927
 928 #define PKRU_AD_BIT 0x1
 929 #define PKRU_WD_BIT 0x2
 930 #define PKRU_BITS_PER_PKEY 2
 931
 932 static inline bool __pkru_allows_read(u32 pkru, u16 pkey)
 933 {
 934         int pkru_pkey_bits = pkey * PKRU_BITS_PER_PKEY;
 935         return !(pkru & (PKRU_AD_BIT << pkru_pkey_bits));
 936 }
 937
 938 static inline bool __pkru_allows_write(u32 pkru, u16 pkey)
 939 {
 940         int pkru_pkey_bits = pkey * PKRU_BITS_PER_PKEY;
 941         /*
 942          * Access-disable disables writes too so we need to check
 943          * both bits here.
 944          */
 945         return !(pkru & ((PKRU_AD_BIT|PKRU_WD_BIT) << pkru_pkey_bits));
 946 }
 947
 948 static inline u16 pte_flags_pkey(unsigned long pte_flags)
 949 {
 950 #ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS
 951         /* ifdef to avoid doing 59-bit shift on 32-bit values */
 952         return (pte_flags & _PAGE_PKEY_MASK) >> _PAGE_BIT_PKEY_BIT0;
 953 #else
 954         return 0;
 955 #endif
 956 }
 957
 958 #include <asm-generic/pgtable.h>
 959 #endif  /* __ASSEMBLY__ */
 960
 961 #endif /* _ASM_X86_PGTABLE_H */