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 /*
  20  * ZERO_PAGE is a global shared page that is always zero: used
  21  * for zero-mapped memory areas etc..
  22  */
  23 extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)];
  24 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
  25
  26 extern spinlock_t pgd_lock;
  27 extern struct list_head pgd_list;
  28
  29 #ifdef CONFIG_PARAVIRT
  30 #include <asm/paravirt.h>
  31 #else  /* !CONFIG_PARAVIRT */
  32 #define set_pte(ptep, pte)              native_set_pte(ptep, pte)
  33 #define set_pte_at(mm, addr, ptep, pte) native_set_pte_at(mm, addr, ptep, pte)
  34
  35 #define set_pte_atomic(ptep, pte)                                       \
  36         native_set_pte_atomic(ptep, pte)
  37
  38 #define set_pmd(pmdp, pmd)              native_set_pmd(pmdp, pmd)
  39
  40 #ifndef __PAGETABLE_PUD_FOLDED
  41 #define set_pgd(pgdp, pgd)              native_set_pgd(pgdp, pgd)
  42 #define pgd_clear(pgd)                  native_pgd_clear(pgd)
  43 #endif
  44
  45 #ifndef set_pud
  46 # define set_pud(pudp, pud)             native_set_pud(pudp, pud)
  47 #endif
  48
  49 #ifndef __PAGETABLE_PMD_FOLDED
  50 #define pud_clear(pud)                  native_pud_clear(pud)
  51 #endif
  52
  53 #define pte_clear(mm, addr, ptep)       native_pte_clear(mm, addr, ptep)
  54 #define pmd_clear(pmd)                  native_pmd_clear(pmd)
  55
  56 #define pte_update(mm, addr, ptep)              do { } while (0)
  57 #define pte_update_defer(mm, addr, ptep)        do { } while (0)
  58
  59 static inline void __init paravirt_pagetable_setup_start(pgd_t *base)
  60 {
  61         native_pagetable_setup_start(base);
  62 }
  63
  64 static inline void __init paravirt_pagetable_setup_done(pgd_t *base)
  65 {
  66         native_pagetable_setup_done(base);
  67 }
  68
  69 #define pgd_val(x)      native_pgd_val(x)
  70 #define __pgd(x)        native_make_pgd(x)
  71
  72 #ifndef __PAGETABLE_PUD_FOLDED
  73 #define pud_val(x)      native_pud_val(x)
  74 #define __pud(x)        native_make_pud(x)
  75 #endif
  76
  77 #ifndef __PAGETABLE_PMD_FOLDED
  78 #define pmd_val(x)      native_pmd_val(x)
  79 #define __pmd(x)        native_make_pmd(x)
  80 #endif
  81
  82 #define pte_val(x)      native_pte_val(x)
  83 #define __pte(x)        native_make_pte(x)
  84
  85 #define arch_end_context_switch(prev)   do {} while(0)
  86
  87 #endif  /* CONFIG_PARAVIRT */
  88
  89 /*
  90  * The following only work if pte_present() is true.
  91  * Undefined behaviour if not..
  92  */
  93 static inline int pte_dirty(pte_t pte)
  94 {
  95         return pte_flags(pte) & _PAGE_DIRTY;
  96 }
  97
  98 static inline int pte_young(pte_t pte)
  99 {
 100         return pte_flags(pte) & _PAGE_ACCESSED;
 101 }
 102
 103 static inline int pte_write(pte_t pte)
 104 {
 105         return pte_flags(pte) & _PAGE_RW;
 106 }
 107
 108 static inline int pte_file(pte_t pte)
 109 {
 110         return pte_flags(pte) & _PAGE_FILE;
 111 }
 112
 113 static inline int pte_huge(pte_t pte)
 114 {
 115         return pte_flags(pte) & _PAGE_PSE;
 116 }
 117
 118 static inline int pte_global(pte_t pte)
 119 {
 120         return pte_flags(pte) & _PAGE_GLOBAL;
 121 }
 122
 123 static inline int pte_exec(pte_t pte)
 124 {
 125         return !(pte_flags(pte) & _PAGE_NX);
 126 }
 127
 128 static inline int pte_special(pte_t pte)
 129 {
 130         return pte_flags(pte) & _PAGE_SPECIAL;
 131 }
 132
 133 static inline unsigned long pte_pfn(pte_t pte)
 134 {
 135         return (pte_val(pte) & PTE_PFN_MASK) >> PAGE_SHIFT;
 136 }
 137
 138 static inline unsigned long pmd_pfn(pmd_t pmd)
 139 {
 140         return (pmd_val(pmd) & PTE_PFN_MASK) >> PAGE_SHIFT;
 141 }
 142
 143 #define pte_page(pte)   pfn_to_page(pte_pfn(pte))
 144
 145 static inline int pmd_large(pmd_t pte)
 146 {
 147         return (pmd_flags(pte) & (_PAGE_PSE | _PAGE_PRESENT)) ==
 148                 (_PAGE_PSE | _PAGE_PRESENT);
 149 }
 150
 151 static inline pte_t pte_set_flags(pte_t pte, pteval_t set)
 152 {
 153         pteval_t v = native_pte_val(pte);
 154
 155         return native_make_pte(v | set);
 156 }
 157
 158 static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
 159 {
 160         pteval_t v = native_pte_val(pte);
 161
 162         return native_make_pte(v & ~clear);
 163 }
 164
 165 static inline pte_t pte_mkclean(pte_t pte)
 166 {
 167         return pte_clear_flags(pte, _PAGE_DIRTY);
 168 }
 169
 170 static inline pte_t pte_mkold(pte_t pte)
 171 {
 172         return pte_clear_flags(pte, _PAGE_ACCESSED);
 173 }
 174
 175 static inline pte_t pte_wrprotect(pte_t pte)
 176 {
 177         return pte_clear_flags(pte, _PAGE_RW);
 178 }
 179
 180 static inline pte_t pte_mkexec(pte_t pte)
 181 {
 182         return pte_clear_flags(pte, _PAGE_NX);
 183 }
 184
 185 static inline pte_t pte_mkdirty(pte_t pte)
 186 {
 187         return pte_set_flags(pte, _PAGE_DIRTY);
 188 }
 189
 190 static inline pte_t pte_mkyoung(pte_t pte)
 191 {
 192         return pte_set_flags(pte, _PAGE_ACCESSED);
 193 }
 194
 195 static inline pte_t pte_mkwrite(pte_t pte)
 196 {
 197         return pte_set_flags(pte, _PAGE_RW);
 198 }
 199
 200 static inline pte_t pte_mkhuge(pte_t pte)
 201 {
 202         return pte_set_flags(pte, _PAGE_PSE);
 203 }
 204
 205 static inline pte_t pte_clrhuge(pte_t pte)
 206 {
 207         return pte_clear_flags(pte, _PAGE_PSE);
 208 }
 209
 210 static inline pte_t pte_mkglobal(pte_t pte)
 211 {
 212         return pte_set_flags(pte, _PAGE_GLOBAL);
 213 }
 214
 215 static inline pte_t pte_clrglobal(pte_t pte)
 216 {
 217         return pte_clear_flags(pte, _PAGE_GLOBAL);
 218 }
 219
 220 static inline pte_t pte_mkspecial(pte_t pte)
 221 {
 222         return pte_set_flags(pte, _PAGE_SPECIAL);
 223 }
 224
 225 /*
 226  * Mask out unsupported bits in a present pgprot.  Non-present pgprots
 227  * can use those bits for other purposes, so leave them be.
 228  */
 229 static inline pgprotval_t massage_pgprot(pgprot_t pgprot)
 230 {
 231         pgprotval_t protval = pgprot_val(pgprot);
 232
 233         if (protval & _PAGE_PRESENT)
 234                 protval &= __supported_pte_mask;
 235
 236         return protval;
 237 }
 238
 239 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
 240 {
 241         return __pte(((phys_addr_t)page_nr << PAGE_SHIFT) |
 242                      massage_pgprot(pgprot));
 243 }
 244
 245 static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
 246 {
 247         return __pmd(((phys_addr_t)page_nr << PAGE_SHIFT) |
 248                      massage_pgprot(pgprot));
 249 }
 250
 251 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 252 {
 253         pteval_t val = pte_val(pte);
 254
 255         /*
 256          * Chop off the NX bit (if present), and add the NX portion of
 257          * the newprot (if present):
 258          */
 259         val &= _PAGE_CHG_MASK;
 260         val |= massage_pgprot(newprot) & ~_PAGE_CHG_MASK;
 261
 262         return __pte(val);
 263 }
 264
 265 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
 266 #define pgprot_modify pgprot_modify
 267 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
 268 {
 269         pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
 270         pgprotval_t addbits = pgprot_val(newprot);
 271         return __pgprot(preservebits | addbits);
 272 }
 273
 274 #define pte_pgprot(x) __pgprot(pte_flags(x) & PTE_FLAGS_MASK)
 275
 276 #define canon_pgprot(p) __pgprot(massage_pgprot(p))
 277
 278 static inline int is_new_memtype_allowed(u64 paddr, unsigned long size,
 279                                          unsigned long flags,
 280                                          unsigned long new_flags)
 281 {
 282         /*
 283          * PAT type is always WB for ISA. So no need to check.
 284          */
 285         if (is_ISA_range(paddr, paddr + size - 1))
 286                 return 1;
 287
 288         /*
 289          * Certain new memtypes are not allowed with certain
 290          * requested memtype:
 291          * - request is uncached, return cannot be write-back
 292          * - request is write-combine, return cannot be write-back
 293          */
 294         if ((flags == _PAGE_CACHE_UC_MINUS &&
 295              new_flags == _PAGE_CACHE_WB) ||
 296             (flags == _PAGE_CACHE_WC &&
 297              new_flags == _PAGE_CACHE_WB)) {
 298                 return 0;
 299         }
 300
 301         return 1;
 302 }
 303
 304 pmd_t *populate_extra_pmd(unsigned long vaddr);
 305 pte_t *populate_extra_pte(unsigned long vaddr);
 306 #endif  /* __ASSEMBLY__ */
 307
 308 #ifdef CONFIG_X86_32
 309 # include "pgtable_32.h"
 310 #else
 311 # include "pgtable_64.h"
 312 #endif
 313
 314 #ifndef __ASSEMBLY__
 315 #include <linux/mm_types.h>
 316
 317 static inline int pte_none(pte_t pte)
 318 {
 319         return !pte.pte;
 320 }
 321
 322 #define __HAVE_ARCH_PTE_SAME
 323 static inline int pte_same(pte_t a, pte_t b)
 324 {
 325         return a.pte == b.pte;
 326 }
 327
 328 static inline int pte_present(pte_t a)
 329 {
 330         return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
 331 }
 332
 333 static inline int pte_hidden(pte_t pte)
 334 {
 335         return pte_flags(pte) & _PAGE_HIDDEN;
 336 }
 337
 338 static inline int pmd_present(pmd_t pmd)
 339 {
 340         return pmd_flags(pmd) & _PAGE_PRESENT;
 341 }
 342
 343 static inline int pmd_none(pmd_t pmd)
 344 {
 345         /* Only check low word on 32-bit platforms, since it might be
 346            out of sync with upper half. */
 347         return (unsigned long)native_pmd_val(pmd) == 0;
 348 }
 349
 350 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
 351 {
 352         return (unsigned long)__va(pmd_val(pmd) & PTE_PFN_MASK);
 353 }
 354
 355 /*
 356  * Currently stuck as a macro due to indirect forward reference to
 357  * linux/mmzone.h's __section_mem_map_addr() definition:
 358  */
 359 #define pmd_page(pmd)   pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
 360
 361 /*
 362  * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
 363  *
 364  * this macro returns the index of the entry in the pmd page which would
 365  * control the given virtual address
 366  */
 367 static inline unsigned long pmd_index(unsigned long address)
 368 {
 369         return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
 370 }
 371
 372 /*
 373  * Conversion functions: convert a page and protection to a page entry,
 374  * and a page entry and page directory to the page they refer to.
 375  *
 376  * (Currently stuck as a macro because of indirect forward reference
 377  * to linux/mm.h:page_to_nid())
 378  */
 379 #define mk_pte(page, pgprot)   pfn_pte(page_to_pfn(page), (pgprot))
 380
 381 /*
 382  * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
 383  *
 384  * this function returns the index of the entry in the pte page which would
 385  * control the given virtual address
 386  */
 387 static inline unsigned long pte_index(unsigned long address)
 388 {
 389         return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
 390 }
 391
 392 static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
 393 {
 394         return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
 395 }
 396
 397 static inline int pmd_bad(pmd_t pmd)
 398 {
 399         return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
 400 }
 401
 402 static inline unsigned long pages_to_mb(unsigned long npg)
 403 {
 404         return npg >> (20 - PAGE_SHIFT);
 405 }
 406
 407 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
 408         remap_pfn_range(vma, vaddr, pfn, size, prot)
 409
 410 #if PAGETABLE_LEVELS > 2
 411 static inline int pud_none(pud_t pud)
 412 {
 413         return native_pud_val(pud) == 0;
 414 }
 415
 416 static inline int pud_present(pud_t pud)
 417 {
 418         return pud_flags(pud) & _PAGE_PRESENT;
 419 }
 420
 421 static inline unsigned long pud_page_vaddr(pud_t pud)
 422 {
 423         return (unsigned long)__va((unsigned long)pud_val(pud) & PTE_PFN_MASK);
 424 }
 425
 426 /*
 427  * Currently stuck as a macro due to indirect forward reference to
 428  * linux/mmzone.h's __section_mem_map_addr() definition:
 429  */
 430 #define pud_page(pud)           pfn_to_page(pud_val(pud) >> PAGE_SHIFT)
 431
 432 /* Find an entry in the second-level page table.. */
 433 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
 434 {
 435         return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
 436 }
 437
 438 static inline int pud_large(pud_t pud)
 439 {
 440         return (pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
 441                 (_PAGE_PSE | _PAGE_PRESENT);
 442 }
 443
 444 static inline int pud_bad(pud_t pud)
 445 {
 446         return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
 447 }
 448 #else
 449 static inline int pud_large(pud_t pud)
 450 {
 451         return 0;
 452 }
 453 #endif  /* PAGETABLE_LEVELS > 2 */
 454
 455 #if PAGETABLE_LEVELS > 3
 456 static inline int pgd_present(pgd_t pgd)
 457 {
 458         return pgd_flags(pgd) & _PAGE_PRESENT;
 459 }
 460
 461 static inline unsigned long pgd_page_vaddr(pgd_t pgd)
 462 {
 463         return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
 464 }
 465
 466 /*
 467  * Currently stuck as a macro due to indirect forward reference to
 468  * linux/mmzone.h's __section_mem_map_addr() definition:
 469  */
 470 #define pgd_page(pgd)           pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
 471
 472 /* to find an entry in a page-table-directory. */
 473 static inline unsigned long pud_index(unsigned long address)
 474 {
 475         return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
 476 }
 477
 478 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
 479 {
 480         return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
 481 }
 482
 483 static inline int pgd_bad(pgd_t pgd)
 484 {
 485         return (pgd_flags(pgd) & ~_PAGE_USER) != _KERNPG_TABLE;
 486 }
 487
 488 static inline int pgd_none(pgd_t pgd)
 489 {
 490         return !native_pgd_val(pgd);
 491 }
 492 #endif  /* PAGETABLE_LEVELS > 3 */
 493
 494 #endif  /* __ASSEMBLY__ */
 495
 496 /*
 497  * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
 498  *
 499  * this macro returns the index of the entry in the pgd page which would
 500  * control the given virtual address
 501  */
 502 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
 503
 504 /*
 505  * pgd_offset() returns a (pgd_t *)
 506  * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
 507  */
 508 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
 509 /*
 510  * a shortcut which implies the use of the kernel's pgd, instead
 511  * of a process's
 512  */
 513 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
 514
 515
 516 #define KERNEL_PGD_BOUNDARY     pgd_index(PAGE_OFFSET)
 517 #define KERNEL_PGD_PTRS         (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
 518
 519 #ifndef __ASSEMBLY__
 520
 521 extern int direct_gbpages;
 522
 523 /* local pte updates need not use xchg for locking */
 524 static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
 525 {
 526         pte_t res = *ptep;
 527
 528         /* Pure native function needs no input for mm, addr */
 529         native_pte_clear(NULL, 0, ptep);
 530         return res;
 531 }
 532
 533 static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
 534                                      pte_t *ptep , pte_t pte)
 535 {
 536         native_set_pte(ptep, pte);
 537 }
 538
 539 #ifndef CONFIG_PARAVIRT
 540 /*
 541  * Rules for using pte_update - it must be called after any PTE update which
 542  * has not been done using the set_pte / clear_pte interfaces.  It is used by
 543  * shadow mode hypervisors to resynchronize the shadow page tables.  Kernel PTE
 544  * updates should either be sets, clears, or set_pte_atomic for P->P
 545  * transitions, which means this hook should only be called for user PTEs.
 546  * This hook implies a P->P protection or access change has taken place, which
 547  * requires a subsequent TLB flush.  The notification can optionally be delayed
 548  * until the TLB flush event by using the pte_update_defer form of the
 549  * interface, but care must be taken to assure that the flush happens while
 550  * still holding the same page table lock so that the shadow and primary pages
 551  * do not become out of sync on SMP.
 552  */
 553 #define pte_update(mm, addr, ptep)              do { } while (0)
 554 #define pte_update_defer(mm, addr, ptep)        do { } while (0)
 555 #endif
 556
 557 /*
 558  * We only update the dirty/accessed state if we set
 559  * the dirty bit by hand in the kernel, since the hardware
 560  * will do the accessed bit for us, and we don't want to
 561  * race with other CPU's that might be updating the dirty
 562  * bit at the same time.
 563  */
 564 struct vm_area_struct;
 565
 566 #define  __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
 567 extern int ptep_set_access_flags(struct vm_area_struct *vma,
 568                                  unsigned long address, pte_t *ptep,
 569                                  pte_t entry, int dirty);
 570
 571 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
 572 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
 573                                      unsigned long addr, pte_t *ptep);
 574
 575 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
 576 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
 577                                   unsigned long address, pte_t *ptep);
 578
 579 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
 580 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
 581                                        pte_t *ptep)
 582 {
 583         pte_t pte = native_ptep_get_and_clear(ptep);
 584         pte_update(mm, addr, ptep);
 585         return pte;
 586 }
 587
 588 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
 589 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
 590                                             unsigned long addr, pte_t *ptep,
 591                                             int full)
 592 {
 593         pte_t pte;
 594         if (full) {
 595                 /*
 596                  * Full address destruction in progress; paravirt does not
 597                  * care about updates and native needs no locking
 598                  */
 599                 pte = native_local_ptep_get_and_clear(ptep);
 600         } else {
 601                 pte = ptep_get_and_clear(mm, addr, ptep);
 602         }
 603         return pte;
 604 }
 605
 606 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
 607 static inline void ptep_set_wrprotect(struct mm_struct *mm,
 608                                       unsigned long addr, pte_t *ptep)
 609 {
 610         clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
 611         pte_update(mm, addr, ptep);
 612 }
 613
 614 /*
 615  * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
 616  *
 617  *  dst - pointer to pgd range anwhere on a pgd page
 618  *  src - ""
 619  *  count - the number of pgds to copy.
 620  *
 621  * dst and src can be on the same page, but the range must not overlap,
 622  * and must not cross a page boundary.
 623  */
 624 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
 625 {
 626        memcpy(dst, src, count * sizeof(pgd_t));
 627 }
 628
 629
 630 #include <asm-generic/pgtable.h>
 631 #endif  /* __ASSEMBLY__ */
 632
 633 #endif /* _ASM_X86_PGTABLE_H */