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) | _PAGE_CACHE_UC_MINUS))  \
  15          : (prot))
  16
  17 #ifndef __ASSEMBLY__
  18
  19 #include <asm/x86_init.h>
  20
  21 /*
  22  * ZERO_PAGE is a global shared page that is always zero: used
  23  * for zero-mapped memory areas etc..
  24  */
  25 extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)];
  26 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
  27
  28 extern spinlock_t pgd_lock;
  29 extern struct list_head pgd_list;
  30
  31 extern struct mm_struct *pgd_page_get_mm(struct page *page);
  32
  33 #ifdef CONFIG_PARAVIRT
  34 #include <asm/paravirt.h>
  35 #else  /* !CONFIG_PARAVIRT */
  36 #define set_pte(ptep, pte)              native_set_pte(ptep, pte)
  37 #define set_pte_at(mm, addr, ptep, pte) native_set_pte_at(mm, addr, ptep, pte)
  38
  39 #define set_pte_atomic(ptep, pte)                                       \
  40         native_set_pte_atomic(ptep, pte)
  41
  42 #define set_pmd(pmdp, pmd)              native_set_pmd(pmdp, pmd)
  43
  44 #ifndef __PAGETABLE_PUD_FOLDED
  45 #define set_pgd(pgdp, pgd)              native_set_pgd(pgdp, pgd)
  46 #define pgd_clear(pgd)                  native_pgd_clear(pgd)
  47 #endif
  48
  49 #ifndef set_pud
  50 # define set_pud(pudp, pud)             native_set_pud(pudp, pud)
  51 #endif
  52
  53 #ifndef __PAGETABLE_PMD_FOLDED
  54 #define pud_clear(pud)                  native_pud_clear(pud)
  55 #endif
  56
  57 #define pte_clear(mm, addr, ptep)       native_pte_clear(mm, addr, ptep)
  58 #define pmd_clear(pmd)                  native_pmd_clear(pmd)
  59
  60 #define pte_update(mm, addr, ptep)              do { } while (0)
  61 #define pte_update_defer(mm, addr, ptep)        do { } while (0)
  62
  63 #define pgd_val(x)      native_pgd_val(x)
  64 #define __pgd(x)        native_make_pgd(x)
  65
  66 #ifndef __PAGETABLE_PUD_FOLDED
  67 #define pud_val(x)      native_pud_val(x)
  68 #define __pud(x)        native_make_pud(x)
  69 #endif
  70
  71 #ifndef __PAGETABLE_PMD_FOLDED
  72 #define pmd_val(x)      native_pmd_val(x)
  73 #define __pmd(x)        native_make_pmd(x)
  74 #endif
  75
  76 #define pte_val(x)      native_pte_val(x)
  77 #define __pte(x)        native_make_pte(x)
  78
  79 #define arch_end_context_switch(prev)   do {} while(0)
  80
  81 #endif  /* CONFIG_PARAVIRT */
  82
  83 /*
  84  * The following only work if pte_present() is true.
  85  * Undefined behaviour if not..
  86  */
  87 static inline int pte_dirty(pte_t pte)
  88 {
  89         return pte_flags(pte) & _PAGE_DIRTY;
  90 }
  91
  92 static inline int pte_young(pte_t pte)
  93 {
  94         return pte_flags(pte) & _PAGE_ACCESSED;
  95 }
  96
  97 static inline int pte_write(pte_t pte)
  98 {
  99         return pte_flags(pte) & _PAGE_RW;
 100 }
 101
 102 static inline int pte_file(pte_t pte)
 103 {
 104         return pte_flags(pte) & _PAGE_FILE;
 105 }
 106
 107 static inline int pte_huge(pte_t pte)
 108 {
 109         return pte_flags(pte) & _PAGE_PSE;
 110 }
 111
 112 static inline int pte_global(pte_t pte)
 113 {
 114         return pte_flags(pte) & _PAGE_GLOBAL;
 115 }
 116
 117 static inline int pte_exec(pte_t pte)
 118 {
 119         return !(pte_flags(pte) & _PAGE_NX);
 120 }
 121
 122 static inline int pte_special(pte_t pte)
 123 {
 124         return pte_flags(pte) & _PAGE_SPECIAL;
 125 }
 126
 127 static inline unsigned long pte_pfn(pte_t pte)
 128 {
 129         return (pte_val(pte) & PTE_PFN_MASK) >> PAGE_SHIFT;
 130 }
 131
 132 static inline unsigned long pmd_pfn(pmd_t pmd)
 133 {
 134         return (pmd_val(pmd) & PTE_PFN_MASK) >> PAGE_SHIFT;
 135 }
 136
 137 #define pte_page(pte)   pfn_to_page(pte_pfn(pte))
 138
 139 static inline int pmd_large(pmd_t pte)
 140 {
 141         return (pmd_flags(pte) & (_PAGE_PSE | _PAGE_PRESENT)) ==
 142                 (_PAGE_PSE | _PAGE_PRESENT);
 143 }
 144
 145 static inline pte_t pte_set_flags(pte_t pte, pteval_t set)
 146 {
 147         pteval_t v = native_pte_val(pte);
 148
 149         return native_make_pte(v | set);
 150 }
 151
 152 static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
 153 {
 154         pteval_t v = native_pte_val(pte);
 155
 156         return native_make_pte(v & ~clear);
 157 }
 158
 159 static inline pte_t pte_mkclean(pte_t pte)
 160 {
 161         return pte_clear_flags(pte, _PAGE_DIRTY);
 162 }
 163
 164 static inline pte_t pte_mkold(pte_t pte)
 165 {
 166         return pte_clear_flags(pte, _PAGE_ACCESSED);
 167 }
 168
 169 static inline pte_t pte_wrprotect(pte_t pte)
 170 {
 171         return pte_clear_flags(pte, _PAGE_RW);
 172 }
 173
 174 static inline pte_t pte_mkexec(pte_t pte)
 175 {
 176         return pte_clear_flags(pte, _PAGE_NX);
 177 }
 178
 179 static inline pte_t pte_mkdirty(pte_t pte)
 180 {
 181         return pte_set_flags(pte, _PAGE_DIRTY);
 182 }
 183
 184 static inline pte_t pte_mkyoung(pte_t pte)
 185 {
 186         return pte_set_flags(pte, _PAGE_ACCESSED);
 187 }
 188
 189 static inline pte_t pte_mkwrite(pte_t pte)
 190 {
 191         return pte_set_flags(pte, _PAGE_RW);
 192 }
 193
 194 static inline pte_t pte_mkhuge(pte_t pte)
 195 {
 196         return pte_set_flags(pte, _PAGE_PSE);
 197 }
 198
 199 static inline pte_t pte_clrhuge(pte_t pte)
 200 {
 201         return pte_clear_flags(pte, _PAGE_PSE);
 202 }
 203
 204 static inline pte_t pte_mkglobal(pte_t pte)
 205 {
 206         return pte_set_flags(pte, _PAGE_GLOBAL);
 207 }
 208
 209 static inline pte_t pte_clrglobal(pte_t pte)
 210 {
 211         return pte_clear_flags(pte, _PAGE_GLOBAL);
 212 }
 213
 214 static inline pte_t pte_mkspecial(pte_t pte)
 215 {
 216         return pte_set_flags(pte, _PAGE_SPECIAL);
 217 }
 218
 219 /*
 220  * Mask out unsupported bits in a present pgprot.  Non-present pgprots
 221  * can use those bits for other purposes, so leave them be.
 222  */
 223 static inline pgprotval_t massage_pgprot(pgprot_t pgprot)
 224 {
 225         pgprotval_t protval = pgprot_val(pgprot);
 226
 227         if (protval & _PAGE_PRESENT)
 228                 protval &= __supported_pte_mask;
 229
 230         return protval;
 231 }
 232
 233 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
 234 {
 235         return __pte(((phys_addr_t)page_nr << PAGE_SHIFT) |
 236                      massage_pgprot(pgprot));
 237 }
 238
 239 static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
 240 {
 241         return __pmd(((phys_addr_t)page_nr << PAGE_SHIFT) |
 242                      massage_pgprot(pgprot));
 243 }
 244
 245 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 246 {
 247         pteval_t val = pte_val(pte);
 248
 249         /*
 250          * Chop off the NX bit (if present), and add the NX portion of
 251          * the newprot (if present):
 252          */
 253         val &= _PAGE_CHG_MASK;
 254         val |= massage_pgprot(newprot) & ~_PAGE_CHG_MASK;
 255
 256         return __pte(val);
 257 }
 258
 259 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
 260 #define pgprot_modify pgprot_modify
 261 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
 262 {
 263         pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
 264         pgprotval_t addbits = pgprot_val(newprot);
 265         return __pgprot(preservebits | addbits);
 266 }
 267
 268 #define pte_pgprot(x) __pgprot(pte_flags(x) & PTE_FLAGS_MASK)
 269
 270 #define canon_pgprot(p) __pgprot(massage_pgprot(p))
 271
 272 static inline int is_new_memtype_allowed(u64 paddr, unsigned long size,
 273                                          unsigned long flags,
 274                                          unsigned long new_flags)
 275 {
 276         /*
 277          * PAT type is always WB for untracked ranges, so no need to check.
 278          */
 279         if (x86_platform.is_untracked_pat_range(paddr, paddr + size))
 280                 return 1;
 281
 282         /*
 283          * Certain new memtypes are not allowed with certain
 284          * requested memtype:
 285          * - request is uncached, return cannot be write-back
 286          * - request is write-combine, return cannot be write-back
 287          */
 288         if ((flags == _PAGE_CACHE_UC_MINUS &&
 289              new_flags == _PAGE_CACHE_WB) ||
 290             (flags == _PAGE_CACHE_WC &&
 291              new_flags == _PAGE_CACHE_WB)) {
 292                 return 0;
 293         }
 294
 295         return 1;
 296 }
 297
 298 pmd_t *populate_extra_pmd(unsigned long vaddr);
 299 pte_t *populate_extra_pte(unsigned long vaddr);
 300 #endif  /* __ASSEMBLY__ */
 301
 302 #ifdef CONFIG_X86_32
 303 # include "pgtable_32.h"
 304 #else
 305 # include "pgtable_64.h"
 306 #endif
 307
 308 #ifndef __ASSEMBLY__
 309 #include <linux/mm_types.h>
 310
 311 static inline int pte_none(pte_t pte)
 312 {
 313         return !pte.pte;
 314 }
 315
 316 #define __HAVE_ARCH_PTE_SAME
 317 static inline int pte_same(pte_t a, pte_t b)
 318 {
 319         return a.pte == b.pte;
 320 }
 321
 322 static inline int pte_present(pte_t a)
 323 {
 324         return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
 325 }
 326
 327 static inline int pte_hidden(pte_t pte)
 328 {
 329         return pte_flags(pte) & _PAGE_HIDDEN;
 330 }
 331
 332 static inline int pmd_present(pmd_t pmd)
 333 {
 334         return pmd_flags(pmd) & _PAGE_PRESENT;
 335 }
 336
 337 static inline int pmd_none(pmd_t pmd)
 338 {
 339         /* Only check low word on 32-bit platforms, since it might be
 340            out of sync with upper half. */
 341         return (unsigned long)native_pmd_val(pmd) == 0;
 342 }
 343
 344 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
 345 {
 346         return (unsigned long)__va(pmd_val(pmd) & PTE_PFN_MASK);
 347 }
 348
 349 /*
 350  * Currently stuck as a macro due to indirect forward reference to
 351  * linux/mmzone.h's __section_mem_map_addr() definition:
 352  */
 353 #define pmd_page(pmd)   pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
 354
 355 /*
 356  * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
 357  *
 358  * this macro returns the index of the entry in the pmd page which would
 359  * control the given virtual address
 360  */
 361 static inline unsigned long pmd_index(unsigned long address)
 362 {
 363         return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
 364 }
 365
 366 /*
 367  * Conversion functions: convert a page and protection to a page entry,
 368  * and a page entry and page directory to the page they refer to.
 369  *
 370  * (Currently stuck as a macro because of indirect forward reference
 371  * to linux/mm.h:page_to_nid())
 372  */
 373 #define mk_pte(page, pgprot)   pfn_pte(page_to_pfn(page), (pgprot))
 374
 375 /*
 376  * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
 377  *
 378  * this function returns the index of the entry in the pte page which would
 379  * control the given virtual address
 380  */
 381 static inline unsigned long pte_index(unsigned long address)
 382 {
 383         return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
 384 }
 385
 386 static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
 387 {
 388         return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
 389 }
 390
 391 static inline int pmd_bad(pmd_t pmd)
 392 {
 393         return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
 394 }
 395
 396 static inline unsigned long pages_to_mb(unsigned long npg)
 397 {
 398         return npg >> (20 - PAGE_SHIFT);
 399 }
 400
 401 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
 402         remap_pfn_range(vma, vaddr, pfn, size, prot)
 403
 404 #if PAGETABLE_LEVELS > 2
 405 static inline int pud_none(pud_t pud)
 406 {
 407         return native_pud_val(pud) == 0;
 408 }
 409
 410 static inline int pud_present(pud_t pud)
 411 {
 412         return pud_flags(pud) & _PAGE_PRESENT;
 413 }
 414
 415 static inline unsigned long pud_page_vaddr(pud_t pud)
 416 {
 417         return (unsigned long)__va((unsigned long)pud_val(pud) & PTE_PFN_MASK);
 418 }
 419
 420 /*
 421  * Currently stuck as a macro due to indirect forward reference to
 422  * linux/mmzone.h's __section_mem_map_addr() definition:
 423  */
 424 #define pud_page(pud)           pfn_to_page(pud_val(pud) >> PAGE_SHIFT)
 425
 426 /* Find an entry in the second-level page table.. */
 427 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
 428 {
 429         return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
 430 }
 431
 432 static inline int pud_large(pud_t pud)
 433 {
 434         return (pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
 435                 (_PAGE_PSE | _PAGE_PRESENT);
 436 }
 437
 438 static inline int pud_bad(pud_t pud)
 439 {
 440         return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
 441 }
 442 #else
 443 static inline int pud_large(pud_t pud)
 444 {
 445         return 0;
 446 }
 447 #endif  /* PAGETABLE_LEVELS > 2 */
 448
 449 #if PAGETABLE_LEVELS > 3
 450 static inline int pgd_present(pgd_t pgd)
 451 {
 452         return pgd_flags(pgd) & _PAGE_PRESENT;
 453 }
 454
 455 static inline unsigned long pgd_page_vaddr(pgd_t pgd)
 456 {
 457         return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
 458 }
 459
 460 /*
 461  * Currently stuck as a macro due to indirect forward reference to
 462  * linux/mmzone.h's __section_mem_map_addr() definition:
 463  */
 464 #define pgd_page(pgd)           pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
 465
 466 /* to find an entry in a page-table-directory. */
 467 static inline unsigned long pud_index(unsigned long address)
 468 {
 469         return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
 470 }
 471
 472 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
 473 {
 474         return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
 475 }
 476
 477 static inline int pgd_bad(pgd_t pgd)
 478 {
 479         return (pgd_flags(pgd) & ~_PAGE_USER) != _KERNPG_TABLE;
 480 }
 481
 482 static inline int pgd_none(pgd_t pgd)
 483 {
 484         return !native_pgd_val(pgd);
 485 }
 486 #endif  /* PAGETABLE_LEVELS > 3 */
 487
 488 #endif  /* __ASSEMBLY__ */
 489
 490 /*
 491  * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
 492  *
 493  * this macro returns the index of the entry in the pgd page which would
 494  * control the given virtual address
 495  */
 496 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
 497
 498 /*
 499  * pgd_offset() returns a (pgd_t *)
 500  * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
 501  */
 502 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
 503 /*
 504  * a shortcut which implies the use of the kernel's pgd, instead
 505  * of a process's
 506  */
 507 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
 508
 509
 510 #define KERNEL_PGD_BOUNDARY     pgd_index(PAGE_OFFSET)
 511 #define KERNEL_PGD_PTRS         (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
 512
 513 #ifndef __ASSEMBLY__
 514
 515 extern int direct_gbpages;
 516
 517 /* local pte updates need not use xchg for locking */
 518 static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
 519 {
 520         pte_t res = *ptep;
 521
 522         /* Pure native function needs no input for mm, addr */
 523         native_pte_clear(NULL, 0, ptep);
 524         return res;
 525 }
 526
 527 static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
 528                                      pte_t *ptep , pte_t pte)
 529 {
 530         native_set_pte(ptep, pte);
 531 }
 532
 533 #ifndef CONFIG_PARAVIRT
 534 /*
 535  * Rules for using pte_update - it must be called after any PTE update which
 536  * has not been done using the set_pte / clear_pte interfaces.  It is used by
 537  * shadow mode hypervisors to resynchronize the shadow page tables.  Kernel PTE
 538  * updates should either be sets, clears, or set_pte_atomic for P->P
 539  * transitions, which means this hook should only be called for user PTEs.
 540  * This hook implies a P->P protection or access change has taken place, which
 541  * requires a subsequent TLB flush.  The notification can optionally be delayed
 542  * until the TLB flush event by using the pte_update_defer form of the
 543  * interface, but care must be taken to assure that the flush happens while
 544  * still holding the same page table lock so that the shadow and primary pages
 545  * do not become out of sync on SMP.
 546  */
 547 #define pte_update(mm, addr, ptep)              do { } while (0)
 548 #define pte_update_defer(mm, addr, ptep)        do { } while (0)
 549 #endif
 550
 551 /*
 552  * We only update the dirty/accessed state if we set
 553  * the dirty bit by hand in the kernel, since the hardware
 554  * will do the accessed bit for us, and we don't want to
 555  * race with other CPU's that might be updating the dirty
 556  * bit at the same time.
 557  */
 558 struct vm_area_struct;
 559
 560 #define  __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
 561 extern int ptep_set_access_flags(struct vm_area_struct *vma,
 562                                  unsigned long address, pte_t *ptep,
 563                                  pte_t entry, int dirty);
 564
 565 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
 566 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
 567                                      unsigned long addr, pte_t *ptep);
 568
 569 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
 570 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
 571                                   unsigned long address, pte_t *ptep);
 572
 573 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
 574 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
 575                                        pte_t *ptep)
 576 {
 577         pte_t pte = native_ptep_get_and_clear(ptep);
 578         pte_update(mm, addr, ptep);
 579         return pte;
 580 }
 581
 582 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
 583 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
 584                                             unsigned long addr, pte_t *ptep,
 585                                             int full)
 586 {
 587         pte_t pte;
 588         if (full) {
 589                 /*
 590                  * Full address destruction in progress; paravirt does not
 591                  * care about updates and native needs no locking
 592                  */
 593                 pte = native_local_ptep_get_and_clear(ptep);
 594         } else {
 595                 pte = ptep_get_and_clear(mm, addr, ptep);
 596         }
 597         return pte;
 598 }
 599
 600 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
 601 static inline void ptep_set_wrprotect(struct mm_struct *mm,
 602                                       unsigned long addr, pte_t *ptep)
 603 {
 604         clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
 605         pte_update(mm, addr, ptep);
 606 }
 607
 608 #define flush_tlb_fix_spurious_fault(vma, address)
 609
 610 /*
 611  * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
 612  *
 613  *  dst - pointer to pgd range anwhere on a pgd page
 614  *  src - ""
 615  *  count - the number of pgds to copy.
 616  *
 617  * dst and src can be on the same page, but the range must not overlap,
 618  * and must not cross a page boundary.
 619  */
 620 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
 621 {
 622        memcpy(dst, src, count * sizeof(pgd_t));
 623 }
 624
 625
 626 #include <asm-generic/pgtable.h>
 627 #endif  /* __ASSEMBLY__ */
 628
 629 #endif /* _ASM_X86_PGTABLE_H */