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
   6 #include <asm/pgtable_types.h>
   7
   8 /*
   9  * Macro to mark a page protection value as UC-
  10  */
  11 #define pgprot_noncached(prot)                                  \
  12         ((boot_cpu_data.x86 > 3)                                \
  13          ? (__pgprot(pgprot_val(prot) | _PAGE_CACHE_UC_MINUS))  \
  14          : (prot))
  15
  16 #ifndef __ASSEMBLY__
  17
  18 /*
  19  * ZERO_PAGE is a global shared page that is always zero: used
  20  * for zero-mapped memory areas etc..
  21  */
  22 extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)];
  23 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
  24
  25 extern spinlock_t pgd_lock;
  26 extern struct list_head pgd_list;
  27
  28 #ifdef CONFIG_PARAVIRT
  29 #include <asm/paravirt.h>
  30 #else  /* !CONFIG_PARAVIRT */
  31 #define set_pte(ptep, pte)              native_set_pte(ptep, pte)
  32 #define set_pte_at(mm, addr, ptep, pte) native_set_pte_at(mm, addr, ptep, pte)
  33
  34 #define set_pte_atomic(ptep, pte)                                       \
  35         native_set_pte_atomic(ptep, pte)
  36
  37 #define set_pmd(pmdp, pmd)              native_set_pmd(pmdp, pmd)
  38
  39 #ifndef __PAGETABLE_PUD_FOLDED
  40 #define set_pgd(pgdp, pgd)              native_set_pgd(pgdp, pgd)
  41 #define pgd_clear(pgd)                  native_pgd_clear(pgd)
  42 #endif
  43
  44 #ifndef set_pud
  45 # define set_pud(pudp, pud)             native_set_pud(pudp, pud)
  46 #endif
  47
  48 #ifndef __PAGETABLE_PMD_FOLDED
  49 #define pud_clear(pud)                  native_pud_clear(pud)
  50 #endif
  51
  52 #define pte_clear(mm, addr, ptep)       native_pte_clear(mm, addr, ptep)
  53 #define pmd_clear(pmd)                  native_pmd_clear(pmd)
  54
  55 #define pte_update(mm, addr, ptep)              do { } while (0)
  56 #define pte_update_defer(mm, addr, ptep)        do { } while (0)
  57
  58 static inline void __init paravirt_pagetable_setup_start(pgd_t *base)
  59 {
  60         native_pagetable_setup_start(base);
  61 }
  62
  63 static inline void __init paravirt_pagetable_setup_done(pgd_t *base)
  64 {
  65         native_pagetable_setup_done(base);
  66 }
  67
  68 #define pgd_val(x)      native_pgd_val(x)
  69 #define __pgd(x)        native_make_pgd(x)
  70
  71 #ifndef __PAGETABLE_PUD_FOLDED
  72 #define pud_val(x)      native_pud_val(x)
  73 #define __pud(x)        native_make_pud(x)
  74 #endif
  75
  76 #ifndef __PAGETABLE_PMD_FOLDED
  77 #define pmd_val(x)      native_pmd_val(x)
  78 #define __pmd(x)        native_make_pmd(x)
  79 #endif
  80
  81 #define pte_val(x)      native_pte_val(x)
  82 #define __pte(x)        native_make_pte(x)
  83
  84 #define arch_end_context_switch(prev)   do {} while(0)
  85
  86 #endif  /* CONFIG_PARAVIRT */
  87
  88 /*
  89  * The following only work if pte_present() is true.
  90  * Undefined behaviour if not..
  91  */
  92 static inline int pte_dirty(pte_t pte)
  93 {
  94         return pte_flags(pte) & _PAGE_DIRTY;
  95 }
  96
  97 static inline int pte_young(pte_t pte)
  98 {
  99         return pte_flags(pte) & _PAGE_ACCESSED;
 100 }
 101
 102 static inline int pte_write(pte_t pte)
 103 {
 104         return pte_flags(pte) & _PAGE_RW;
 105 }
 106
 107 static inline int pte_file(pte_t pte)
 108 {
 109         return pte_flags(pte) & _PAGE_FILE;
 110 }
 111
 112 static inline int pte_huge(pte_t pte)
 113 {
 114         return pte_flags(pte) & _PAGE_PSE;
 115 }
 116
 117 static inline int pte_global(pte_t pte)
 118 {
 119         return pte_flags(pte) & _PAGE_GLOBAL;
 120 }
 121
 122 static inline int pte_exec(pte_t pte)
 123 {
 124         return !(pte_flags(pte) & _PAGE_NX);
 125 }
 126
 127 static inline int pte_special(pte_t pte)
 128 {
 129         return pte_flags(pte) & _PAGE_SPECIAL;
 130 }
 131
 132 static inline unsigned long pte_pfn(pte_t pte)
 133 {
 134         return (pte_val(pte) & 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(unsigned long flags,
 273                                                 unsigned long new_flags)
 274 {
 275         /*
 276          * Certain new memtypes are not allowed with certain
 277          * requested memtype:
 278          * - request is uncached, return cannot be write-back
 279          * - request is write-combine, return cannot be write-back
 280          */
 281         if ((flags == _PAGE_CACHE_UC_MINUS &&
 282              new_flags == _PAGE_CACHE_WB) ||
 283             (flags == _PAGE_CACHE_WC &&
 284              new_flags == _PAGE_CACHE_WB)) {
 285                 return 0;
 286         }
 287
 288         return 1;
 289 }
 290
 291 pmd_t *populate_extra_pmd(unsigned long vaddr);
 292 pte_t *populate_extra_pte(unsigned long vaddr);
 293 #endif  /* __ASSEMBLY__ */
 294
 295 #ifdef CONFIG_X86_32
 296 # include "pgtable_32.h"
 297 #else
 298 # include "pgtable_64.h"
 299 #endif
 300
 301 #ifndef __ASSEMBLY__
 302 #include <linux/mm_types.h>
 303
 304 static inline int pte_none(pte_t pte)
 305 {
 306         return !pte.pte;
 307 }
 308
 309 #define __HAVE_ARCH_PTE_SAME
 310 static inline int pte_same(pte_t a, pte_t b)
 311 {
 312         return a.pte == b.pte;
 313 }
 314
 315 static inline int pte_present(pte_t a)
 316 {
 317         return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
 318 }
 319
 320 static inline int pmd_present(pmd_t pmd)
 321 {
 322         return pmd_flags(pmd) & _PAGE_PRESENT;
 323 }
 324
 325 static inline int pmd_none(pmd_t pmd)
 326 {
 327         /* Only check low word on 32-bit platforms, since it might be
 328            out of sync with upper half. */
 329         return (unsigned long)native_pmd_val(pmd) == 0;
 330 }
 331
 332 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
 333 {
 334         return (unsigned long)__va(pmd_val(pmd) & PTE_PFN_MASK);
 335 }
 336
 337 /*
 338  * Currently stuck as a macro due to indirect forward reference to
 339  * linux/mmzone.h's __section_mem_map_addr() definition:
 340  */
 341 #define pmd_page(pmd)   pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
 342
 343 /*
 344  * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
 345  *
 346  * this macro returns the index of the entry in the pmd page which would
 347  * control the given virtual address
 348  */
 349 static inline unsigned pmd_index(unsigned long address)
 350 {
 351         return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
 352 }
 353
 354 /*
 355  * Conversion functions: convert a page and protection to a page entry,
 356  * and a page entry and page directory to the page they refer to.
 357  *
 358  * (Currently stuck as a macro because of indirect forward reference
 359  * to linux/mm.h:page_to_nid())
 360  */
 361 #define mk_pte(page, pgprot)   pfn_pte(page_to_pfn(page), (pgprot))
 362
 363 /*
 364  * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
 365  *
 366  * this function returns the index of the entry in the pte page which would
 367  * control the given virtual address
 368  */
 369 static inline unsigned pte_index(unsigned long address)
 370 {
 371         return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
 372 }
 373
 374 static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
 375 {
 376         return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
 377 }
 378
 379 static inline int pmd_bad(pmd_t pmd)
 380 {
 381         return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
 382 }
 383
 384 static inline unsigned long pages_to_mb(unsigned long npg)
 385 {
 386         return npg >> (20 - PAGE_SHIFT);
 387 }
 388
 389 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
 390         remap_pfn_range(vma, vaddr, pfn, size, prot)
 391
 392 #if PAGETABLE_LEVELS > 2
 393 static inline int pud_none(pud_t pud)
 394 {
 395         return native_pud_val(pud) == 0;
 396 }
 397
 398 static inline int pud_present(pud_t pud)
 399 {
 400         return pud_flags(pud) & _PAGE_PRESENT;
 401 }
 402
 403 static inline unsigned long pud_page_vaddr(pud_t pud)
 404 {
 405         return (unsigned long)__va((unsigned long)pud_val(pud) & PTE_PFN_MASK);
 406 }
 407
 408 /*
 409  * Currently stuck as a macro due to indirect forward reference to
 410  * linux/mmzone.h's __section_mem_map_addr() definition:
 411  */
 412 #define pud_page(pud)           pfn_to_page(pud_val(pud) >> PAGE_SHIFT)
 413
 414 /* Find an entry in the second-level page table.. */
 415 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
 416 {
 417         return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
 418 }
 419
 420 static inline unsigned long pmd_pfn(pmd_t pmd)
 421 {
 422         return (pmd_val(pmd) & PTE_PFN_MASK) >> PAGE_SHIFT;
 423 }
 424
 425 static inline int pud_large(pud_t pud)
 426 {
 427         return (pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
 428                 (_PAGE_PSE | _PAGE_PRESENT);
 429 }
 430
 431 static inline int pud_bad(pud_t pud)
 432 {
 433         return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
 434 }
 435 #else
 436 static inline int pud_large(pud_t pud)
 437 {
 438         return 0;
 439 }
 440 #endif  /* PAGETABLE_LEVELS > 2 */
 441
 442 #if PAGETABLE_LEVELS > 3
 443 static inline int pgd_present(pgd_t pgd)
 444 {
 445         return pgd_flags(pgd) & _PAGE_PRESENT;
 446 }
 447
 448 static inline unsigned long pgd_page_vaddr(pgd_t pgd)
 449 {
 450         return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
 451 }
 452
 453 /*
 454  * Currently stuck as a macro due to indirect forward reference to
 455  * linux/mmzone.h's __section_mem_map_addr() definition:
 456  */
 457 #define pgd_page(pgd)           pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
 458
 459 /* to find an entry in a page-table-directory. */
 460 static inline unsigned pud_index(unsigned long address)
 461 {
 462         return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
 463 }
 464
 465 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
 466 {
 467         return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
 468 }
 469
 470 static inline int pgd_bad(pgd_t pgd)
 471 {
 472         return (pgd_flags(pgd) & ~_PAGE_USER) != _KERNPG_TABLE;
 473 }
 474
 475 static inline int pgd_none(pgd_t pgd)
 476 {
 477         return !native_pgd_val(pgd);
 478 }
 479 #endif  /* PAGETABLE_LEVELS > 3 */
 480
 481 #endif  /* __ASSEMBLY__ */
 482
 483 /*
 484  * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
 485  *
 486  * this macro returns the index of the entry in the pgd page which would
 487  * control the given virtual address
 488  */
 489 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
 490
 491 /*
 492  * pgd_offset() returns a (pgd_t *)
 493  * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
 494  */
 495 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
 496 /*
 497  * a shortcut which implies the use of the kernel's pgd, instead
 498  * of a process's
 499  */
 500 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
 501
 502
 503 #define KERNEL_PGD_BOUNDARY     pgd_index(PAGE_OFFSET)
 504 #define KERNEL_PGD_PTRS         (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
 505
 506 #ifndef __ASSEMBLY__
 507
 508 extern int direct_gbpages;
 509
 510 /* local pte updates need not use xchg for locking */
 511 static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
 512 {
 513         pte_t res = *ptep;
 514
 515         /* Pure native function needs no input for mm, addr */
 516         native_pte_clear(NULL, 0, ptep);
 517         return res;
 518 }
 519
 520 static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
 521                                      pte_t *ptep , pte_t pte)
 522 {
 523         native_set_pte(ptep, pte);
 524 }
 525
 526 #ifndef CONFIG_PARAVIRT
 527 /*
 528  * Rules for using pte_update - it must be called after any PTE update which
 529  * has not been done using the set_pte / clear_pte interfaces.  It is used by
 530  * shadow mode hypervisors to resynchronize the shadow page tables.  Kernel PTE
 531  * updates should either be sets, clears, or set_pte_atomic for P->P
 532  * transitions, which means this hook should only be called for user PTEs.
 533  * This hook implies a P->P protection or access change has taken place, which
 534  * requires a subsequent TLB flush.  The notification can optionally be delayed
 535  * until the TLB flush event by using the pte_update_defer form of the
 536  * interface, but care must be taken to assure that the flush happens while
 537  * still holding the same page table lock so that the shadow and primary pages
 538  * do not become out of sync on SMP.
 539  */
 540 #define pte_update(mm, addr, ptep)              do { } while (0)
 541 #define pte_update_defer(mm, addr, ptep)        do { } while (0)
 542 #endif
 543
 544 /*
 545  * We only update the dirty/accessed state if we set
 546  * the dirty bit by hand in the kernel, since the hardware
 547  * will do the accessed bit for us, and we don't want to
 548  * race with other CPU's that might be updating the dirty
 549  * bit at the same time.
 550  */
 551 struct vm_area_struct;
 552
 553 #define  __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
 554 extern int ptep_set_access_flags(struct vm_area_struct *vma,
 555                                  unsigned long address, pte_t *ptep,
 556                                  pte_t entry, int dirty);
 557
 558 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
 559 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
 560                                      unsigned long addr, pte_t *ptep);
 561
 562 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
 563 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
 564                                   unsigned long address, pte_t *ptep);
 565
 566 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
 567 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
 568                                        pte_t *ptep)
 569 {
 570         pte_t pte = native_ptep_get_and_clear(ptep);
 571         pte_update(mm, addr, ptep);
 572         return pte;
 573 }
 574
 575 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
 576 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
 577                                             unsigned long addr, pte_t *ptep,
 578                                             int full)
 579 {
 580         pte_t pte;
 581         if (full) {
 582                 /*
 583                  * Full address destruction in progress; paravirt does not
 584                  * care about updates and native needs no locking
 585                  */
 586                 pte = native_local_ptep_get_and_clear(ptep);
 587         } else {
 588                 pte = ptep_get_and_clear(mm, addr, ptep);
 589         }
 590         return pte;
 591 }
 592
 593 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
 594 static inline void ptep_set_wrprotect(struct mm_struct *mm,
 595                                       unsigned long addr, pte_t *ptep)
 596 {
 597         clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
 598         pte_update(mm, addr, ptep);
 599 }
 600
 601 /*
 602  * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
 603  *
 604  *  dst - pointer to pgd range anwhere on a pgd page
 605  *  src - ""
 606  *  count - the number of pgds to copy.
 607  *
 608  * dst and src can be on the same page, but the range must not overlap,
 609  * and must not cross a page boundary.
 610  */
 611 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
 612 {
 613        memcpy(dst, src, count * sizeof(pgd_t));
 614 }
 615
 616
 617 #include <asm-generic/pgtable.h>
 618 #endif  /* __ASSEMBLY__ */
 619
 620 #endif /* _ASM_X86_PGTABLE_H */