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