include/asm-x86/pgtable_64.h

   1 #ifndef _X86_64_PGTABLE_H
   2 #define _X86_64_PGTABLE_H
   3
   4 #include <linux/const.h>
   5 #ifndef __ASSEMBLY__
   6
   7 /*
   8  * This file contains the functions and defines necessary to modify and use
   9  * the x86-64 page table tree.
  10  */
  11 #include <asm/processor.h>
  12 #include <asm/bitops.h>
  13 #include <linux/threads.h>
  14 #include <asm/pda.h>
  15
  16 extern pud_t level3_kernel_pgt[512];
  17 extern pud_t level3_ident_pgt[512];
  18 extern pmd_t level2_kernel_pgt[512];
  19 extern pgd_t init_level4_pgt[];
  20 extern unsigned long __supported_pte_mask;
  21
  22 #define swapper_pg_dir init_level4_pgt
  23
  24 extern void paging_init(void);
  25 extern void clear_kernel_mapping(unsigned long addr, unsigned long size);
  26
  27 /*
  28  * ZERO_PAGE is a global shared page that is always zero: used
  29  * for zero-mapped memory areas etc..
  30  */
  31 extern unsigned long empty_zero_page[PAGE_SIZE/sizeof(unsigned long)];
  32 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
  33
  34 #endif /* !__ASSEMBLY__ */
  35
  36 /*
  37  * PGDIR_SHIFT determines what a top-level page table entry can map
  38  */
  39 #define PGDIR_SHIFT     39
  40 #define PTRS_PER_PGD    512
  41
  42 /*
  43  * 3rd level page
  44  */
  45 #define PUD_SHIFT       30
  46 #define PTRS_PER_PUD    512
  47
  48 /*
  49  * PMD_SHIFT determines the size of the area a middle-level
  50  * page table can map
  51  */
  52 #define PMD_SHIFT       21
  53 #define PTRS_PER_PMD    512
  54
  55 /*
  56  * entries per page directory level
  57  */
  58 #define PTRS_PER_PTE    512
  59
  60 #ifndef __ASSEMBLY__
  61
  62 #define pte_ERROR(e) \
  63         printk("%s:%d: bad pte %p(%016lx).\n", __FILE__, __LINE__, &(e), pte_val(e))
  64 #define pmd_ERROR(e) \
  65         printk("%s:%d: bad pmd %p(%016lx).\n", __FILE__, __LINE__, &(e), pmd_val(e))
  66 #define pud_ERROR(e) \
  67         printk("%s:%d: bad pud %p(%016lx).\n", __FILE__, __LINE__, &(e), pud_val(e))
  68 #define pgd_ERROR(e) \
  69         printk("%s:%d: bad pgd %p(%016lx).\n", __FILE__, __LINE__, &(e), pgd_val(e))
  70
  71 #define pgd_none(x)     (!pgd_val(x))
  72 #define pud_none(x)     (!pud_val(x))
  73
  74 static inline void set_pte(pte_t *dst, pte_t val)
  75 {
  76         pte_val(*dst) = pte_val(val);
  77 }
  78 #define set_pte_at(mm,addr,ptep,pteval) set_pte(ptep,pteval)
  79
  80 static inline void set_pmd(pmd_t *dst, pmd_t val)
  81 {
  82         pmd_val(*dst) = pmd_val(val);
  83 }
  84
  85 static inline void set_pud(pud_t *dst, pud_t val)
  86 {
  87         pud_val(*dst) = pud_val(val);
  88 }
  89
  90 static inline void pud_clear (pud_t *pud)
  91 {
  92         set_pud(pud, __pud(0));
  93 }
  94
  95 static inline void set_pgd(pgd_t *dst, pgd_t val)
  96 {
  97         pgd_val(*dst) = pgd_val(val);
  98 }
  99
 100 static inline void pgd_clear (pgd_t * pgd)
 101 {
 102         set_pgd(pgd, __pgd(0));
 103 }
 104
 105 #define ptep_get_and_clear(mm,addr,xp)  __pte(xchg(&(xp)->pte, 0))
 106
 107 struct mm_struct;
 108
 109 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm, unsigned long addr, pte_t *ptep, int full)
 110 {
 111         pte_t pte;
 112         if (full) {
 113                 pte = *ptep;
 114                 *ptep = __pte(0);
 115         } else {
 116                 pte = ptep_get_and_clear(mm, addr, ptep);
 117         }
 118         return pte;
 119 }
 120
 121 #define pte_same(a, b)          ((a).pte == (b).pte)
 122
 123 #define pte_pgprot(a)   (__pgprot((a).pte & ~PHYSICAL_PAGE_MASK))
 124
 125 #endif /* !__ASSEMBLY__ */
 126
 127 #define PMD_SIZE        (_AC(1,UL) << PMD_SHIFT)
 128 #define PMD_MASK        (~(PMD_SIZE-1))
 129 #define PUD_SIZE        (_AC(1,UL) << PUD_SHIFT)
 130 #define PUD_MASK        (~(PUD_SIZE-1))
 131 #define PGDIR_SIZE      (_AC(1,UL) << PGDIR_SHIFT)
 132 #define PGDIR_MASK      (~(PGDIR_SIZE-1))
 133
 134 #define USER_PTRS_PER_PGD       ((TASK_SIZE-1)/PGDIR_SIZE+1)
 135 #define FIRST_USER_ADDRESS      0
 136
 137 #define MAXMEM           _AC(0x3fffffffffff, UL)
 138 #define VMALLOC_START    _AC(0xffffc20000000000, UL)
 139 #define VMALLOC_END      _AC(0xffffe1ffffffffff, UL)
 140 #define MODULES_VADDR    _AC(0xffffffff88000000, UL)
 141 #define MODULES_END      _AC(0xfffffffffff00000, UL)
 142 #define MODULES_LEN   (MODULES_END - MODULES_VADDR)
 143
 144 #define _PAGE_BIT_PRESENT       0
 145 #define _PAGE_BIT_RW            1
 146 #define _PAGE_BIT_USER          2
 147 #define _PAGE_BIT_PWT           3
 148 #define _PAGE_BIT_PCD           4
 149 #define _PAGE_BIT_ACCESSED      5
 150 #define _PAGE_BIT_DIRTY         6
 151 #define _PAGE_BIT_PSE           7       /* 4 MB (or 2MB) page */
 152 #define _PAGE_BIT_GLOBAL        8       /* Global TLB entry PPro+ */
 153 #define _PAGE_BIT_NX           63       /* No execute: only valid after cpuid check */
 154
 155 #define _PAGE_PRESENT   0x001
 156 #define _PAGE_RW        0x002
 157 #define _PAGE_USER      0x004
 158 #define _PAGE_PWT       0x008
 159 #define _PAGE_PCD       0x010
 160 #define _PAGE_ACCESSED  0x020
 161 #define _PAGE_DIRTY     0x040
 162 #define _PAGE_PSE       0x080   /* 2MB page */
 163 #define _PAGE_FILE      0x040   /* nonlinear file mapping, saved PTE; unset:swap */
 164 #define _PAGE_GLOBAL    0x100   /* Global TLB entry */
 165
 166 #define _PAGE_PROTNONE  0x080   /* If not present */
 167 #define _PAGE_NX        (_AC(1,UL)<<_PAGE_BIT_NX)
 168
 169 #define _PAGE_TABLE     (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED | _PAGE_DIRTY)
 170 #define _KERNPG_TABLE   (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY)
 171
 172 #define _PAGE_CHG_MASK  (PTE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY)
 173
 174 #define PAGE_NONE       __pgprot(_PAGE_PROTNONE | _PAGE_ACCESSED)
 175 #define PAGE_SHARED     __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
 176 #define PAGE_SHARED_EXEC __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED)
 177 #define PAGE_COPY_NOEXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
 178 #define PAGE_COPY PAGE_COPY_NOEXEC
 179 #define PAGE_COPY_EXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
 180 #define PAGE_READONLY   __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
 181 #define PAGE_READONLY_EXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
 182 #define __PAGE_KERNEL \
 183         (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_NX)
 184 #define __PAGE_KERNEL_EXEC \
 185         (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED)
 186 #define __PAGE_KERNEL_NOCACHE \
 187         (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_PCD | _PAGE_ACCESSED | _PAGE_NX)
 188 #define __PAGE_KERNEL_RO \
 189         (_PAGE_PRESENT | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_NX)
 190 #define __PAGE_KERNEL_VSYSCALL \
 191         (_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
 192 #define __PAGE_KERNEL_VSYSCALL_NOCACHE \
 193         (_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_PCD)
 194 #define __PAGE_KERNEL_LARGE \
 195         (__PAGE_KERNEL | _PAGE_PSE)
 196 #define __PAGE_KERNEL_LARGE_EXEC \
 197         (__PAGE_KERNEL_EXEC | _PAGE_PSE)
 198
 199 #define MAKE_GLOBAL(x) __pgprot((x) | _PAGE_GLOBAL)
 200
 201 #define PAGE_KERNEL MAKE_GLOBAL(__PAGE_KERNEL)
 202 #define PAGE_KERNEL_EXEC MAKE_GLOBAL(__PAGE_KERNEL_EXEC)
 203 #define PAGE_KERNEL_RO MAKE_GLOBAL(__PAGE_KERNEL_RO)
 204 #define PAGE_KERNEL_NOCACHE MAKE_GLOBAL(__PAGE_KERNEL_NOCACHE)
 205 #define PAGE_KERNEL_VSYSCALL32 __pgprot(__PAGE_KERNEL_VSYSCALL)
 206 #define PAGE_KERNEL_VSYSCALL MAKE_GLOBAL(__PAGE_KERNEL_VSYSCALL)
 207 #define PAGE_KERNEL_LARGE MAKE_GLOBAL(__PAGE_KERNEL_LARGE)
 208 #define PAGE_KERNEL_VSYSCALL_NOCACHE MAKE_GLOBAL(__PAGE_KERNEL_VSYSCALL_NOCACHE)
 209
 210 /*         xwr */
 211 #define __P000  PAGE_NONE
 212 #define __P001  PAGE_READONLY
 213 #define __P010  PAGE_COPY
 214 #define __P011  PAGE_COPY
 215 #define __P100  PAGE_READONLY_EXEC
 216 #define __P101  PAGE_READONLY_EXEC
 217 #define __P110  PAGE_COPY_EXEC
 218 #define __P111  PAGE_COPY_EXEC
 219
 220 #define __S000  PAGE_NONE
 221 #define __S001  PAGE_READONLY
 222 #define __S010  PAGE_SHARED
 223 #define __S011  PAGE_SHARED
 224 #define __S100  PAGE_READONLY_EXEC
 225 #define __S101  PAGE_READONLY_EXEC
 226 #define __S110  PAGE_SHARED_EXEC
 227 #define __S111  PAGE_SHARED_EXEC
 228
 229 #ifndef __ASSEMBLY__
 230
 231 static inline unsigned long pgd_bad(pgd_t pgd)
 232 {
 233         return pgd_val(pgd) & ~(PTE_MASK | _KERNPG_TABLE | _PAGE_USER);
 234 }
 235
 236 static inline unsigned long pud_bad(pud_t pud)
 237 {
 238         return pud_val(pud) & ~(PTE_MASK | _KERNPG_TABLE | _PAGE_USER);
 239 }
 240
 241 static inline unsigned long pmd_bad(pmd_t pmd)
 242 {
 243         return pmd_val(pmd) & ~(PTE_MASK | _KERNPG_TABLE | _PAGE_USER);
 244 }
 245
 246 #define pte_none(x)     (!pte_val(x))
 247 #define pte_present(x)  (pte_val(x) & (_PAGE_PRESENT | _PAGE_PROTNONE))
 248 #define pte_clear(mm,addr,xp)   do { set_pte_at(mm, addr, xp, __pte(0)); } while (0)
 249
 250 #define pages_to_mb(x) ((x) >> (20-PAGE_SHIFT)) /* FIXME: is this
 251                                                    right? */
 252 #define pte_page(x)     pfn_to_page(pte_pfn(x))
 253 #define pte_pfn(x)  ((pte_val(x) & __PHYSICAL_MASK) >> PAGE_SHIFT)
 254
 255 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
 256 {
 257         pte_t pte;
 258         pte_val(pte) = (page_nr << PAGE_SHIFT);
 259         pte_val(pte) |= pgprot_val(pgprot);
 260         pte_val(pte) &= __supported_pte_mask;
 261         return pte;
 262 }
 263
 264 /*
 265  * The following only work if pte_present() is true.
 266  * Undefined behaviour if not..
 267  */
 268 #define __LARGE_PTE (_PAGE_PSE|_PAGE_PRESENT)
 269 static inline int pte_dirty(pte_t pte)          { return pte_val(pte) & _PAGE_DIRTY; }
 270 static inline int pte_young(pte_t pte)          { return pte_val(pte) & _PAGE_ACCESSED; }
 271 static inline int pte_write(pte_t pte)          { return pte_val(pte) & _PAGE_RW; }
 272 static inline int pte_file(pte_t pte)           { return pte_val(pte) & _PAGE_FILE; }
 273 static inline int pte_huge(pte_t pte)           { return pte_val(pte) & _PAGE_PSE; }
 274
 275 static inline pte_t pte_mkclean(pte_t pte)      { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_DIRTY)); return pte; }
 276 static inline pte_t pte_mkold(pte_t pte)        { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_ACCESSED)); return pte; }
 277 static inline pte_t pte_wrprotect(pte_t pte)    { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_RW)); return pte; }
 278 static inline pte_t pte_mkexec(pte_t pte)       { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_NX)); return pte; }
 279 static inline pte_t pte_mkdirty(pte_t pte)      { set_pte(&pte, __pte(pte_val(pte) | _PAGE_DIRTY)); return pte; }
 280 static inline pte_t pte_mkyoung(pte_t pte)      { set_pte(&pte, __pte(pte_val(pte) | _PAGE_ACCESSED)); return pte; }
 281 static inline pte_t pte_mkwrite(pte_t pte)      { set_pte(&pte, __pte(pte_val(pte) | _PAGE_RW)); return pte; }
 282 static inline pte_t pte_mkhuge(pte_t pte)       { set_pte(&pte, __pte(pte_val(pte) | _PAGE_PSE)); return pte; }
 283 static inline pte_t pte_clrhuge(pte_t pte)      { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_PSE)); return pte; }
 284
 285 struct vm_area_struct;
 286
 287 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
 288 {
 289         if (!pte_young(*ptep))
 290                 return 0;
 291         return test_and_clear_bit(_PAGE_BIT_ACCESSED, &ptep->pte);
 292 }
 293
 294 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
 295 {
 296         clear_bit(_PAGE_BIT_RW, &ptep->pte);
 297 }
 298
 299 /*
 300  * Macro to mark a page protection value as "uncacheable".
 301  */
 302 #define pgprot_noncached(prot)  (__pgprot(pgprot_val(prot) | _PAGE_PCD | _PAGE_PWT))
 303
 304 static inline int pmd_large(pmd_t pte) {
 305         return (pmd_val(pte) & __LARGE_PTE) == __LARGE_PTE;
 306 }
 307
 308
 309 /*
 310  * Conversion functions: convert a page and protection to a page entry,
 311  * and a page entry and page directory to the page they refer to.
 312  */
 313
 314 /*
 315  * Level 4 access.
 316  */
 317 #define pgd_page_vaddr(pgd) ((unsigned long) __va((unsigned long)pgd_val(pgd) & PTE_MASK))
 318 #define pgd_page(pgd)           (pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT))
 319 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
 320 #define pgd_offset(mm, addr) ((mm)->pgd + pgd_index(addr))
 321 #define pgd_offset_k(address) (init_level4_pgt + pgd_index(address))
 322 #define pgd_present(pgd) (pgd_val(pgd) & _PAGE_PRESENT)
 323 #define mk_kernel_pgd(address) ((pgd_t){ (address) | _KERNPG_TABLE })
 324
 325 /* PUD - Level3 access */
 326 /* to find an entry in a page-table-directory. */
 327 #define pud_page_vaddr(pud) ((unsigned long) __va(pud_val(pud) & PHYSICAL_PAGE_MASK))
 328 #define pud_page(pud)           (pfn_to_page(pud_val(pud) >> PAGE_SHIFT))
 329 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
 330 #define pud_offset(pgd, address) ((pud_t *) pgd_page_vaddr(*(pgd)) + pud_index(address))
 331 #define pud_present(pud) (pud_val(pud) & _PAGE_PRESENT)
 332
 333 /* PMD  - Level 2 access */
 334 #define pmd_page_vaddr(pmd) ((unsigned long) __va(pmd_val(pmd) & PTE_MASK))
 335 #define pmd_page(pmd)           (pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT))
 336
 337 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
 338 #define pmd_offset(dir, address) ((pmd_t *) pud_page_vaddr(*(dir)) + \
 339                         pmd_index(address))
 340 #define pmd_none(x)     (!pmd_val(x))
 341 #define pmd_present(x)  (pmd_val(x) & _PAGE_PRESENT)
 342 #define pmd_clear(xp)   do { set_pmd(xp, __pmd(0)); } while (0)
 343 #define pfn_pmd(nr,prot) (__pmd(((nr) << PAGE_SHIFT) | pgprot_val(prot)))
 344 #define pmd_pfn(x)  ((pmd_val(x) & __PHYSICAL_MASK) >> PAGE_SHIFT)
 345
 346 #define pte_to_pgoff(pte) ((pte_val(pte) & PHYSICAL_PAGE_MASK) >> PAGE_SHIFT)
 347 #define pgoff_to_pte(off) ((pte_t) { ((off) << PAGE_SHIFT) | _PAGE_FILE })
 348 #define PTE_FILE_MAX_BITS __PHYSICAL_MASK_SHIFT
 349
 350 /* PTE - Level 1 access. */
 351
 352 /* page, protection -> pte */
 353 #define mk_pte(page, pgprot)    pfn_pte(page_to_pfn(page), (pgprot))
 354 #define mk_pte_huge(entry) (pte_val(entry) |= _PAGE_PRESENT | _PAGE_PSE)
 355
 356 /* Change flags of a PTE */
 357 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 358 {
 359         pte_val(pte) &= _PAGE_CHG_MASK;
 360         pte_val(pte) |= pgprot_val(newprot);
 361         pte_val(pte) &= __supported_pte_mask;
 362        return pte;
 363 }
 364
 365 #define pte_index(address) \
 366                 (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
 367 #define pte_offset_kernel(dir, address) ((pte_t *) pmd_page_vaddr(*(dir)) + \
 368                         pte_index(address))
 369
 370 /* x86-64 always has all page tables mapped. */
 371 #define pte_offset_map(dir,address) pte_offset_kernel(dir,address)
 372 #define pte_offset_map_nested(dir,address) pte_offset_kernel(dir,address)
 373 #define pte_unmap(pte) /* NOP */
 374 #define pte_unmap_nested(pte) /* NOP */
 375
 376 #define update_mmu_cache(vma,address,pte) do { } while (0)
 377
 378 /* We only update the dirty/accessed state if we set
 379  * the dirty bit by hand in the kernel, since the hardware
 380  * will do the accessed bit for us, and we don't want to
 381  * race with other CPU's that might be updating the dirty
 382  * bit at the same time. */
 383 #define  __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
 384 #define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
 385 ({                                                                        \
 386         int __changed = !pte_same(*(__ptep), __entry);                    \
 387         if (__changed && __dirty) {                                       \
 388                 set_pte(__ptep, __entry);                                 \
 389                 flush_tlb_page(__vma, __address);                         \
 390         }                                                                 \
 391         __changed;                                                        \
 392 })
 393
 394 /* Encode and de-code a swap entry */
 395 #define __swp_type(x)                   (((x).val >> 1) & 0x3f)
 396 #define __swp_offset(x)                 ((x).val >> 8)
 397 #define __swp_entry(type, offset)       ((swp_entry_t) { ((type) << 1) | ((offset) << 8) })
 398 #define __pte_to_swp_entry(pte)         ((swp_entry_t) { pte_val(pte) })
 399 #define __swp_entry_to_pte(x)           ((pte_t) { (x).val })
 400
 401 extern spinlock_t pgd_lock;
 402 extern struct list_head pgd_list;
 403
 404 extern int kern_addr_valid(unsigned long addr);
 405
 406 pte_t *lookup_address(unsigned long addr);
 407
 408 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot)         \
 409                 remap_pfn_range(vma, vaddr, pfn, size, prot)
 410
 411 #define HAVE_ARCH_UNMAPPED_AREA
 412
 413 #define pgtable_cache_init()   do { } while (0)
 414 #define check_pgt_cache()      do { } while (0)
 415
 416 #define PAGE_AGP    PAGE_KERNEL_NOCACHE
 417 #define HAVE_PAGE_AGP 1
 418
 419 /* fs/proc/kcore.c */
 420 #define kc_vaddr_to_offset(v) ((v) & __VIRTUAL_MASK)
 421 #define kc_offset_to_vaddr(o) \
 422    (((o) & (1UL << (__VIRTUAL_MASK_SHIFT-1))) ? ((o) | (~__VIRTUAL_MASK)) : (o))
 423
 424 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
 425 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
 426 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
 427 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
 428 #define __HAVE_ARCH_PTE_SAME
 429 #include <asm-generic/pgtable.h>
 430 #endif /* !__ASSEMBLY__ */
 431
 432 #endif /* _X86_64_PGTABLE_H */