include/asm-x86/pgtable.h

   1 #ifndef _ASM_X86_PGTABLE_H
   2 #define _ASM_X86_PGTABLE_H
   3
   4 #define FIRST_USER_ADDRESS      0
   5
   6 #define _PAGE_BIT_PRESENT       0       /* is present */
   7 #define _PAGE_BIT_RW            1       /* writeable */
   8 #define _PAGE_BIT_USER          2       /* userspace addressable */
   9 #define _PAGE_BIT_PWT           3       /* page write through */
  10 #define _PAGE_BIT_PCD           4       /* page cache disabled */
  11 #define _PAGE_BIT_ACCESSED      5       /* was accessed (raised by CPU) */
  12 #define _PAGE_BIT_DIRTY         6       /* was written to (raised by CPU) */
  13 #define _PAGE_BIT_FILE          6
  14 #define _PAGE_BIT_PSE           7       /* 4 MB (or 2MB) page */
  15 #define _PAGE_BIT_PAT           7       /* on 4KB pages */
  16 #define _PAGE_BIT_GLOBAL        8       /* Global TLB entry PPro+ */
  17 #define _PAGE_BIT_UNUSED1       9       /* available for programmer */
  18 #define _PAGE_BIT_UNUSED2       10
  19 #define _PAGE_BIT_UNUSED3       11
  20 #define _PAGE_BIT_PAT_LARGE     12      /* On 2MB or 1GB pages */
  21 #define _PAGE_BIT_NX           63       /* No execute: only valid after cpuid check */
  22
  23 /*
  24  * Note: we use _AC(1, L) instead of _AC(1, UL) so that we get a
  25  * sign-extended value on 32-bit with all 1's in the upper word,
  26  * which preserves the upper pte values on 64-bit ptes:
  27  */
  28 #define _PAGE_PRESENT   (_AC(1, L)<<_PAGE_BIT_PRESENT)
  29 #define _PAGE_RW        (_AC(1, L)<<_PAGE_BIT_RW)
  30 #define _PAGE_USER      (_AC(1, L)<<_PAGE_BIT_USER)
  31 #define _PAGE_PWT       (_AC(1, L)<<_PAGE_BIT_PWT)
  32 #define _PAGE_PCD       (_AC(1, L)<<_PAGE_BIT_PCD)
  33 #define _PAGE_ACCESSED  (_AC(1, L)<<_PAGE_BIT_ACCESSED)
  34 #define _PAGE_DIRTY     (_AC(1, L)<<_PAGE_BIT_DIRTY)
  35 #define _PAGE_PSE       (_AC(1, L)<<_PAGE_BIT_PSE)      /* 2MB page */
  36 #define _PAGE_GLOBAL    (_AC(1, L)<<_PAGE_BIT_GLOBAL)   /* Global TLB entry */
  37 #define _PAGE_UNUSED1   (_AC(1, L)<<_PAGE_BIT_UNUSED1)
  38 #define _PAGE_UNUSED2   (_AC(1, L)<<_PAGE_BIT_UNUSED2)
  39 #define _PAGE_UNUSED3   (_AC(1, L)<<_PAGE_BIT_UNUSED3)
  40 #define _PAGE_PAT       (_AC(1, L)<<_PAGE_BIT_PAT)
  41 #define _PAGE_PAT_LARGE (_AC(1, L)<<_PAGE_BIT_PAT_LARGE)
  42
  43 #if defined(CONFIG_X86_64) || defined(CONFIG_X86_PAE)
  44 #define _PAGE_NX        (_AC(1, ULL) << _PAGE_BIT_NX)
  45 #else
  46 #define _PAGE_NX        0
  47 #endif
  48
  49 /* If _PAGE_PRESENT is clear, we use these: */
  50 #define _PAGE_FILE      _PAGE_DIRTY     /* nonlinear file mapping,
  51                                          * saved PTE; unset:swap */
  52 #define _PAGE_PROTNONE  _PAGE_PSE       /* if the user mapped it with PROT_NONE;
  53                                            pte_present gives true */
  54
  55 #define _PAGE_TABLE     (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER |        \
  56                          _PAGE_ACCESSED | _PAGE_DIRTY)
  57 #define _KERNPG_TABLE   (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED |    \
  58                          _PAGE_DIRTY)
  59
  60 #define _PAGE_CHG_MASK  (PTE_MASK | _PAGE_PCD | _PAGE_PWT |             \
  61                          _PAGE_ACCESSED | _PAGE_DIRTY)
  62
  63 #define _PAGE_CACHE_MASK        (_PAGE_PCD | _PAGE_PWT)
  64 #define _PAGE_CACHE_WB          (0)
  65 #define _PAGE_CACHE_WC          (_PAGE_PWT)
  66 #define _PAGE_CACHE_UC_MINUS    (_PAGE_PCD)
  67 #define _PAGE_CACHE_UC          (_PAGE_PCD | _PAGE_PWT)
  68
  69 #define PAGE_NONE       __pgprot(_PAGE_PROTNONE | _PAGE_ACCESSED)
  70 #define PAGE_SHARED     __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | \
  71                                  _PAGE_ACCESSED | _PAGE_NX)
  72
  73 #define PAGE_SHARED_EXEC        __pgprot(_PAGE_PRESENT | _PAGE_RW |     \
  74                                          _PAGE_USER | _PAGE_ACCESSED)
  75 #define PAGE_COPY_NOEXEC        __pgprot(_PAGE_PRESENT | _PAGE_USER |   \
  76                                          _PAGE_ACCESSED | _PAGE_NX)
  77 #define PAGE_COPY_EXEC          __pgprot(_PAGE_PRESENT | _PAGE_USER |   \
  78                                          _PAGE_ACCESSED)
  79 #define PAGE_COPY               PAGE_COPY_NOEXEC
  80 #define PAGE_READONLY           __pgprot(_PAGE_PRESENT | _PAGE_USER |   \
  81                                          _PAGE_ACCESSED | _PAGE_NX)
  82 #define PAGE_READONLY_EXEC      __pgprot(_PAGE_PRESENT | _PAGE_USER |   \
  83                                          _PAGE_ACCESSED)
  84
  85 #ifdef CONFIG_X86_32
  86 #define _PAGE_KERNEL_EXEC \
  87         (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED)
  88 #define _PAGE_KERNEL (_PAGE_KERNEL_EXEC | _PAGE_NX)
  89
  90 #ifndef __ASSEMBLY__
  91 extern pteval_t __PAGE_KERNEL, __PAGE_KERNEL_EXEC;
  92 #endif  /* __ASSEMBLY__ */
  93 #else
  94 #define __PAGE_KERNEL_EXEC                                              \
  95         (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED)
  96 #define __PAGE_KERNEL           (__PAGE_KERNEL_EXEC | _PAGE_NX)
  97 #endif
  98
  99 #define __PAGE_KERNEL_RO                (__PAGE_KERNEL & ~_PAGE_RW)
 100 #define __PAGE_KERNEL_RX                (__PAGE_KERNEL_EXEC & ~_PAGE_RW)
 101 #define __PAGE_KERNEL_EXEC_NOCACHE      (__PAGE_KERNEL_EXEC | _PAGE_PCD | _PAGE_PWT)
 102 #define __PAGE_KERNEL_WC                (__PAGE_KERNEL | _PAGE_CACHE_WC)
 103 #define __PAGE_KERNEL_NOCACHE           (__PAGE_KERNEL | _PAGE_PCD | _PAGE_PWT)
 104 #define __PAGE_KERNEL_UC_MINUS          (__PAGE_KERNEL | _PAGE_PCD)
 105 #define __PAGE_KERNEL_VSYSCALL          (__PAGE_KERNEL_RX | _PAGE_USER)
 106 #define __PAGE_KERNEL_VSYSCALL_NOCACHE  (__PAGE_KERNEL_VSYSCALL | _PAGE_PCD | _PAGE_PWT)
 107 #define __PAGE_KERNEL_LARGE             (__PAGE_KERNEL | _PAGE_PSE)
 108 #define __PAGE_KERNEL_LARGE_EXEC        (__PAGE_KERNEL_EXEC | _PAGE_PSE)
 109
 110 #ifdef CONFIG_X86_32
 111 # define MAKE_GLOBAL(x)                 __pgprot((x))
 112 #else
 113 # define MAKE_GLOBAL(x)                 __pgprot((x) | _PAGE_GLOBAL)
 114 #endif
 115
 116 #define PAGE_KERNEL                     MAKE_GLOBAL(__PAGE_KERNEL)
 117 #define PAGE_KERNEL_RO                  MAKE_GLOBAL(__PAGE_KERNEL_RO)
 118 #define PAGE_KERNEL_EXEC                MAKE_GLOBAL(__PAGE_KERNEL_EXEC)
 119 #define PAGE_KERNEL_RX                  MAKE_GLOBAL(__PAGE_KERNEL_RX)
 120 #define PAGE_KERNEL_WC                  MAKE_GLOBAL(__PAGE_KERNEL_WC)
 121 #define PAGE_KERNEL_NOCACHE             MAKE_GLOBAL(__PAGE_KERNEL_NOCACHE)
 122 #define PAGE_KERNEL_UC_MINUS            MAKE_GLOBAL(__PAGE_KERNEL_UC_MINUS)
 123 #define PAGE_KERNEL_EXEC_NOCACHE        MAKE_GLOBAL(__PAGE_KERNEL_EXEC_NOCACHE)
 124 #define PAGE_KERNEL_LARGE               MAKE_GLOBAL(__PAGE_KERNEL_LARGE)
 125 #define PAGE_KERNEL_LARGE_EXEC          MAKE_GLOBAL(__PAGE_KERNEL_LARGE_EXEC)
 126 #define PAGE_KERNEL_VSYSCALL            MAKE_GLOBAL(__PAGE_KERNEL_VSYSCALL)
 127 #define PAGE_KERNEL_VSYSCALL_NOCACHE    MAKE_GLOBAL(__PAGE_KERNEL_VSYSCALL_NOCACHE)
 128
 129 /*         xwr */
 130 #define __P000  PAGE_NONE
 131 #define __P001  PAGE_READONLY
 132 #define __P010  PAGE_COPY
 133 #define __P011  PAGE_COPY
 134 #define __P100  PAGE_READONLY_EXEC
 135 #define __P101  PAGE_READONLY_EXEC
 136 #define __P110  PAGE_COPY_EXEC
 137 #define __P111  PAGE_COPY_EXEC
 138
 139 #define __S000  PAGE_NONE
 140 #define __S001  PAGE_READONLY
 141 #define __S010  PAGE_SHARED
 142 #define __S011  PAGE_SHARED
 143 #define __S100  PAGE_READONLY_EXEC
 144 #define __S101  PAGE_READONLY_EXEC
 145 #define __S110  PAGE_SHARED_EXEC
 146 #define __S111  PAGE_SHARED_EXEC
 147
 148 #ifndef __ASSEMBLY__
 149
 150 /*
 151  * ZERO_PAGE is a global shared page that is always zero: used
 152  * for zero-mapped memory areas etc..
 153  */
 154 extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)];
 155 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
 156
 157 extern spinlock_t pgd_lock;
 158 extern struct list_head pgd_list;
 159
 160 /*
 161  * The following only work if pte_present() is true.
 162  * Undefined behaviour if not..
 163  */
 164 static inline int pte_dirty(pte_t pte)
 165 {
 166         return pte_val(pte) & _PAGE_DIRTY;
 167 }
 168
 169 static inline int pte_young(pte_t pte)
 170 {
 171         return pte_val(pte) & _PAGE_ACCESSED;
 172 }
 173
 174 static inline int pte_write(pte_t pte)
 175 {
 176         return pte_val(pte) & _PAGE_RW;
 177 }
 178
 179 static inline int pte_file(pte_t pte)
 180 {
 181         return pte_val(pte) & _PAGE_FILE;
 182 }
 183
 184 static inline int pte_huge(pte_t pte)
 185 {
 186         return pte_val(pte) & _PAGE_PSE;
 187 }
 188
 189 static inline int pte_global(pte_t pte)
 190 {
 191         return pte_val(pte) & _PAGE_GLOBAL;
 192 }
 193
 194 static inline int pte_exec(pte_t pte)
 195 {
 196         return !(pte_val(pte) & _PAGE_NX);
 197 }
 198
 199 static inline int pte_special(pte_t pte)
 200 {
 201         return 0;
 202 }
 203
 204 static inline int pmd_large(pmd_t pte)
 205 {
 206         return (pmd_val(pte) & (_PAGE_PSE | _PAGE_PRESENT)) ==
 207                 (_PAGE_PSE | _PAGE_PRESENT);
 208 }
 209
 210 static inline pte_t pte_mkclean(pte_t pte)
 211 {
 212         return __pte(pte_val(pte) & ~(pteval_t)_PAGE_DIRTY);
 213 }
 214
 215 static inline pte_t pte_mkold(pte_t pte)
 216 {
 217         return __pte(pte_val(pte) & ~(pteval_t)_PAGE_ACCESSED);
 218 }
 219
 220 static inline pte_t pte_wrprotect(pte_t pte)
 221 {
 222         return __pte(pte_val(pte) & ~(pteval_t)_PAGE_RW);
 223 }
 224
 225 static inline pte_t pte_mkexec(pte_t pte)
 226 {
 227         return __pte(pte_val(pte) & ~(pteval_t)_PAGE_NX);
 228 }
 229
 230 static inline pte_t pte_mkdirty(pte_t pte)
 231 {
 232         return __pte(pte_val(pte) | _PAGE_DIRTY);
 233 }
 234
 235 static inline pte_t pte_mkyoung(pte_t pte)
 236 {
 237         return __pte(pte_val(pte) | _PAGE_ACCESSED);
 238 }
 239
 240 static inline pte_t pte_mkwrite(pte_t pte)
 241 {
 242         return __pte(pte_val(pte) | _PAGE_RW);
 243 }
 244
 245 static inline pte_t pte_mkhuge(pte_t pte)
 246 {
 247         return __pte(pte_val(pte) | _PAGE_PSE);
 248 }
 249
 250 static inline pte_t pte_clrhuge(pte_t pte)
 251 {
 252         return __pte(pte_val(pte) & ~(pteval_t)_PAGE_PSE);
 253 }
 254
 255 static inline pte_t pte_mkglobal(pte_t pte)
 256 {
 257         return __pte(pte_val(pte) | _PAGE_GLOBAL);
 258 }
 259
 260 static inline pte_t pte_clrglobal(pte_t pte)
 261 {
 262         return __pte(pte_val(pte) & ~(pteval_t)_PAGE_GLOBAL);
 263 }
 264
 265 static inline pte_t pte_mkspecial(pte_t pte)
 266 {
 267         return pte;
 268 }
 269
 270 extern pteval_t __supported_pte_mask;
 271
 272 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
 273 {
 274         return __pte((((phys_addr_t)page_nr << PAGE_SHIFT) |
 275                       pgprot_val(pgprot)) & __supported_pte_mask);
 276 }
 277
 278 static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
 279 {
 280         return __pmd((((phys_addr_t)page_nr << PAGE_SHIFT) |
 281                       pgprot_val(pgprot)) & __supported_pte_mask);
 282 }
 283
 284 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 285 {
 286         pteval_t val = pte_val(pte);
 287
 288         /*
 289          * Chop off the NX bit (if present), and add the NX portion of
 290          * the newprot (if present):
 291          */
 292         val &= _PAGE_CHG_MASK;
 293         val |= pgprot_val(newprot) & (~_PAGE_CHG_MASK) & __supported_pte_mask;
 294
 295         return __pte(val);
 296 }
 297
 298 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
 299 #define pgprot_modify pgprot_modify
 300 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
 301 {
 302         pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
 303         pgprotval_t addbits = pgprot_val(newprot);
 304         return __pgprot(preservebits | addbits);
 305 }
 306
 307 #define pte_pgprot(x) __pgprot(pte_val(x) & (0xfff | _PAGE_NX))
 308
 309 #define canon_pgprot(p) __pgprot(pgprot_val(p) & __supported_pte_mask)
 310
 311 #ifndef __ASSEMBLY__
 312 #define __HAVE_PHYS_MEM_ACCESS_PROT
 313 struct file;
 314 pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
 315                               unsigned long size, pgprot_t vma_prot);
 316 int phys_mem_access_prot_allowed(struct file *file, unsigned long pfn,
 317                               unsigned long size, pgprot_t *vma_prot);
 318 #endif
 319
 320 #ifdef CONFIG_PARAVIRT
 321 #include <asm/paravirt.h>
 322 #else  /* !CONFIG_PARAVIRT */
 323 #define set_pte(ptep, pte)              native_set_pte(ptep, pte)
 324 #define set_pte_at(mm, addr, ptep, pte) native_set_pte_at(mm, addr, ptep, pte)
 325
 326 #define set_pte_present(mm, addr, ptep, pte)                            \
 327         native_set_pte_present(mm, addr, ptep, pte)
 328 #define set_pte_atomic(ptep, pte)                                       \
 329         native_set_pte_atomic(ptep, pte)
 330
 331 #define set_pmd(pmdp, pmd)              native_set_pmd(pmdp, pmd)
 332
 333 #ifndef __PAGETABLE_PUD_FOLDED
 334 #define set_pgd(pgdp, pgd)              native_set_pgd(pgdp, pgd)
 335 #define pgd_clear(pgd)                  native_pgd_clear(pgd)
 336 #endif
 337
 338 #ifndef set_pud
 339 # define set_pud(pudp, pud)             native_set_pud(pudp, pud)
 340 #endif
 341
 342 #ifndef __PAGETABLE_PMD_FOLDED
 343 #define pud_clear(pud)                  native_pud_clear(pud)
 344 #endif
 345
 346 #define pte_clear(mm, addr, ptep)       native_pte_clear(mm, addr, ptep)
 347 #define pmd_clear(pmd)                  native_pmd_clear(pmd)
 348
 349 #define pte_update(mm, addr, ptep)              do { } while (0)
 350 #define pte_update_defer(mm, addr, ptep)        do { } while (0)
 351 #endif  /* CONFIG_PARAVIRT */
 352
 353 #endif  /* __ASSEMBLY__ */
 354
 355 #ifdef CONFIG_X86_32
 356 # include "pgtable_32.h"
 357 #else
 358 # include "pgtable_64.h"
 359 #endif
 360
 361 #define KERNEL_PGD_BOUNDARY     pgd_index(PAGE_OFFSET)
 362 #define KERNEL_PGD_PTRS         (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
 363
 364 #ifndef __ASSEMBLY__
 365
 366 enum {
 367         PG_LEVEL_NONE,
 368         PG_LEVEL_4K,
 369         PG_LEVEL_2M,
 370         PG_LEVEL_1G,
 371 };
 372
 373 /*
 374  * Helper function that returns the kernel pagetable entry controlling
 375  * the virtual address 'address'. NULL means no pagetable entry present.
 376  * NOTE: the return type is pte_t but if the pmd is PSE then we return it
 377  * as a pte too.
 378  */
 379 extern pte_t *lookup_address(unsigned long address, unsigned int *level);
 380
 381 /* local pte updates need not use xchg for locking */
 382 static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
 383 {
 384         pte_t res = *ptep;
 385
 386         /* Pure native function needs no input for mm, addr */
 387         native_pte_clear(NULL, 0, ptep);
 388         return res;
 389 }
 390
 391 static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
 392                                      pte_t *ptep , pte_t pte)
 393 {
 394         native_set_pte(ptep, pte);
 395 }
 396
 397 #ifndef CONFIG_PARAVIRT
 398 /*
 399  * Rules for using pte_update - it must be called after any PTE update which
 400  * has not been done using the set_pte / clear_pte interfaces.  It is used by
 401  * shadow mode hypervisors to resynchronize the shadow page tables.  Kernel PTE
 402  * updates should either be sets, clears, or set_pte_atomic for P->P
 403  * transitions, which means this hook should only be called for user PTEs.
 404  * This hook implies a P->P protection or access change has taken place, which
 405  * requires a subsequent TLB flush.  The notification can optionally be delayed
 406  * until the TLB flush event by using the pte_update_defer form of the
 407  * interface, but care must be taken to assure that the flush happens while
 408  * still holding the same page table lock so that the shadow and primary pages
 409  * do not become out of sync on SMP.
 410  */
 411 #define pte_update(mm, addr, ptep)              do { } while (0)
 412 #define pte_update_defer(mm, addr, ptep)        do { } while (0)
 413 #endif
 414
 415 /*
 416  * We only update the dirty/accessed state if we set
 417  * the dirty bit by hand in the kernel, since the hardware
 418  * will do the accessed bit for us, and we don't want to
 419  * race with other CPU's that might be updating the dirty
 420  * bit at the same time.
 421  */
 422 #define  __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
 423 extern int ptep_set_access_flags(struct vm_area_struct *vma,
 424                                  unsigned long address, pte_t *ptep,
 425                                  pte_t entry, int dirty);
 426
 427 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
 428 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
 429                                      unsigned long addr, pte_t *ptep);
 430
 431 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
 432 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
 433                                   unsigned long address, pte_t *ptep);
 434
 435 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
 436 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
 437                                        pte_t *ptep)
 438 {
 439         pte_t pte = native_ptep_get_and_clear(ptep);
 440         pte_update(mm, addr, ptep);
 441         return pte;
 442 }
 443
 444 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
 445 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
 446                                             unsigned long addr, pte_t *ptep,
 447                                             int full)
 448 {
 449         pte_t pte;
 450         if (full) {
 451                 /*
 452                  * Full address destruction in progress; paravirt does not
 453                  * care about updates and native needs no locking
 454                  */
 455                 pte = native_local_ptep_get_and_clear(ptep);
 456         } else {
 457                 pte = ptep_get_and_clear(mm, addr, ptep);
 458         }
 459         return pte;
 460 }
 461
 462 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
 463 static inline void ptep_set_wrprotect(struct mm_struct *mm,
 464                                       unsigned long addr, pte_t *ptep)
 465 {
 466         clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
 467         pte_update(mm, addr, ptep);
 468 }
 469
 470 /*
 471  * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
 472  *
 473  *  dst - pointer to pgd range anwhere on a pgd page
 474  *  src - ""
 475  *  count - the number of pgds to copy.
 476  *
 477  * dst and src can be on the same page, but the range must not overlap,
 478  * and must not cross a page boundary.
 479  */
 480 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
 481 {
 482        memcpy(dst, src, count * sizeof(pgd_t));
 483 }
 484
 485
 486 #include <asm-generic/pgtable.h>
 487 #endif  /* __ASSEMBLY__ */
 488
 489 #endif  /* _ASM_X86_PGTABLE_H */