include/asm-x86/pgtable.h

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