LXR linux/include/asm-x86

   1#ifndef _X86_64_PGTABLE_H
   2#define _X86_64_PGTABLE_H
   3
   4/*
   5 * This file contains the functions and defines necessary to modify and use
   6 * the x86-64 page table tree.
   7 */
   8#include <asm/processor.h>
   9#include <asm/fixmap.h>
  10#include <asm/bitops.h>
  11#include <linux/threads.h>
  12#include <asm/pda.h>
  13
  14extern pud_t level3_kernel_pgt[512];
  15extern pud_t level3_physmem_pgt[512];
  16extern pud_t level3_ident_pgt[512];
  17extern pmd_t level2_kernel_pgt[512];
  18extern pgd_t init_level4_pgt[];
  19extern unsigned long __supported_pte_mask;
  20
  21#define swapper_pg_dir init_level4_pgt
  22
  23extern int nonx_setup(char *str);
  24extern void paging_init(void);
  25extern void clear_kernel_mapping(unsigned long addr, unsigned long size);
  26
  27extern unsigned long pgkern_mask;
  28
  29/*
  30 * ZERO_PAGE is a global shared page that is always zero: used
  31 * for zero-mapped memory areas etc..
  32 */
  33extern unsigned long empty_zero_page[PAGE_SIZE/sizeof(unsigned long)];
  34#define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
  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#define pte_ERROR(e) \
  61        printk("%s:%d: bad pte %p(%016lx).\n", __FILE__, __LINE__, &(e), pte_val(e))
  62#define pmd_ERROR(e) \
  63        printk("%s:%d: bad pmd %p(%016lx).\n", __FILE__, __LINE__, &(e), pmd_val(e))
  64#define pud_ERROR(e) \
  65        printk("%s:%d: bad pud %p(%016lx).\n", __FILE__, __LINE__, &(e), pud_val(e))
  66#define pgd_ERROR(e) \
  67        printk("%s:%d: bad pgd %p(%016lx).\n", __FILE__, __LINE__, &(e), pgd_val(e))
  68
  69#define pgd_none(x)     (!pgd_val(x))
  70#define pud_none(x)     (!pud_val(x))
  71
  72static inline void set_pte(pte_t *dst, pte_t val)
  73{
  74        pte_val(*dst) = pte_val(val);
  75} 
  76
  77static inline void set_pmd(pmd_t *dst, pmd_t val)
  78{
  79        pmd_val(*dst) = pmd_val(val); 
  80} 
  81
  82static inline void set_pud(pud_t *dst, pud_t val)
  83{
  84        pud_val(*dst) = pud_val(val);
  85}
  86
  87extern inline void pud_clear (pud_t *pud)
  88{
  89        set_pud(pud, __pud(0));
  90}
  91
  92static inline void set_pgd(pgd_t *dst, pgd_t val)
  93{
  94        pgd_val(*dst) = pgd_val(val); 
  95} 
  96
  97extern inline void pgd_clear (pgd_t * pgd)
  98{
  99        set_pgd(pgd, __pgd(0));
 100}
 101
 102#define pud_page(pud) \
 103((unsigned long) __va(pud_val(pud) & PHYSICAL_PAGE_MASK))
 104
 105#define ptep_get_and_clear(xp)  __pte(xchg(&(xp)->pte, 0))
 106#define pte_same(a, b)          ((a).pte == (b).pte)
 107
 108#define PMD_SIZE        (1UL << PMD_SHIFT)
 109#define PMD_MASK        (~(PMD_SIZE-1))
 110#define PUD_SIZE        (1UL << PUD_SHIFT)
 111#define PUD_MASK        (~(PUD_SIZE-1))
 112#define PGDIR_SIZE      (1UL << PGDIR_SHIFT)
 113#define PGDIR_MASK      (~(PGDIR_SIZE-1))
 114
 115#define USER_PTRS_PER_PGD       (TASK_SIZE/PGDIR_SIZE)
 116#define FIRST_USER_PGD_NR       0
 117
 118#ifndef __ASSEMBLY__
 119#define MAXMEM           0x3fffffffffffUL
 120#define VMALLOC_START    0xffffc20000000000UL
 121#define VMALLOC_END      0xffffe1ffffffffffUL
 122#define MODULES_VADDR    0xffffffff88000000UL
 123#define MODULES_END      0xfffffffffff00000UL
 124#define MODULES_LEN   (MODULES_END - MODULES_VADDR)
 125
 126#define _PAGE_BIT_PRESENT       0
 127#define _PAGE_BIT_RW            1
 128#define _PAGE_BIT_USER          2
 129#define _PAGE_BIT_PWT           3
 130#define _PAGE_BIT_PCD           4
 131#define _PAGE_BIT_ACCESSED      5
 132#define _PAGE_BIT_DIRTY         6
 133#define _PAGE_BIT_PSE           7       /* 4 MB (or 2MB) page */
 134#define _PAGE_BIT_GLOBAL        8       /* Global TLB entry PPro+ */
 135#define _PAGE_BIT_NX           63       /* No execute: only valid after cpuid check */
 136
 137#define _PAGE_PRESENT   0x001
 138#define _PAGE_RW        0x002
 139#define _PAGE_USER      0x004
 140#define _PAGE_PWT       0x008
 141#define _PAGE_PCD       0x010
 142#define _PAGE_ACCESSED  0x020
 143#define _PAGE_DIRTY     0x040
 144#define _PAGE_PSE       0x080   /* 2MB page */
 145#define _PAGE_FILE      0x040   /* set:pagecache, unset:swap */
 146#define _PAGE_GLOBAL    0x100   /* Global TLB entry */
 147
 148#define _PAGE_PROTNONE  0x080   /* If not present */
 149#define _PAGE_NX        (1UL<<_PAGE_BIT_NX)
 150
 151#define _PAGE_TABLE     (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED | _PAGE_DIRTY)
 152#define _KERNPG_TABLE   (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY)
 153
 154#define _PAGE_CHG_MASK  (PTE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY)
 155
 156#define PAGE_NONE       __pgprot(_PAGE_PROTNONE | _PAGE_ACCESSED)
 157#define PAGE_SHARED     __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
 158#define PAGE_SHARED_EXEC __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED)
 159#define PAGE_COPY_NOEXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
 160#define PAGE_COPY PAGE_COPY_NOEXEC
 161#define PAGE_COPY_EXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
 162#define PAGE_READONLY   __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
 163#define PAGE_READONLY_EXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
 164#define __PAGE_KERNEL \
 165        (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_NX)
 166#define __PAGE_KERNEL_EXEC \
 167        (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED)
 168#define __PAGE_KERNEL_NOCACHE \
 169        (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_PCD | _PAGE_ACCESSED | _PAGE_NX)
 170#define __PAGE_KERNEL_RO \
 171        (_PAGE_PRESENT | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_NX)
 172#define __PAGE_KERNEL_VSYSCALL \
 173        (_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
 174#define __PAGE_KERNEL_VSYSCALL_NOCACHE \
 175        (_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_PCD)
 176#define __PAGE_KERNEL_LARGE \
 177        (__PAGE_KERNEL | _PAGE_PSE)
 178
 179#define MAKE_GLOBAL(x) __pgprot((x) | _PAGE_GLOBAL)
 180
 181#define PAGE_KERNEL MAKE_GLOBAL(__PAGE_KERNEL)
 182#define PAGE_KERNEL_EXEC MAKE_GLOBAL(__PAGE_KERNEL_EXEC)
 183#define PAGE_KERNEL_RO MAKE_GLOBAL(__PAGE_KERNEL_RO)
 184#define PAGE_KERNEL_NOCACHE MAKE_GLOBAL(__PAGE_KERNEL_NOCACHE)
 185#define PAGE_KERNEL_VSYSCALL32 __pgprot(__PAGE_KERNEL_VSYSCALL)
 186#define PAGE_KERNEL_VSYSCALL MAKE_GLOBAL(__PAGE_KERNEL_VSYSCALL)
 187#define PAGE_KERNEL_LARGE MAKE_GLOBAL(__PAGE_KERNEL_LARGE)
 188#define PAGE_KERNEL_VSYSCALL_NOCACHE MAKE_GLOBAL(__PAGE_KERNEL_VSYSCALL_NOCACHE)
 189
 190/*         xwr */
 191#define __P000  PAGE_NONE
 192#define __P001  PAGE_READONLY
 193#define __P010  PAGE_COPY
 194#define __P011  PAGE_COPY
 195#define __P100  PAGE_READONLY_EXEC
 196#define __P101  PAGE_READONLY_EXEC
 197#define __P110  PAGE_COPY_EXEC
 198#define __P111  PAGE_COPY_EXEC
 199
 200#define __S000  PAGE_NONE
 201#define __S001  PAGE_READONLY
 202#define __S010  PAGE_SHARED
 203#define __S011  PAGE_SHARED
 204#define __S100  PAGE_READONLY_EXEC
 205#define __S101  PAGE_READONLY_EXEC
 206#define __S110  PAGE_SHARED_EXEC
 207#define __S111  PAGE_SHARED_EXEC
 208
 209static inline unsigned long pgd_bad(pgd_t pgd) 
 210{ 
 211       unsigned long val = pgd_val(pgd);
 212       val &= ~PTE_MASK; 
 213       val &= ~(_PAGE_USER | _PAGE_DIRTY); 
 214       return val & ~(_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED);      
 215} 
 216
 217static inline unsigned long pud_bad(pud_t pud)
 218{
 219       unsigned long val = pud_val(pud);
 220       val &= ~PTE_MASK;
 221       val &= ~(_PAGE_USER | _PAGE_DIRTY);
 222       return val & ~(_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED);
 223}
 224
 225#define pte_none(x)     (!pte_val(x))
 226#define pte_present(x)  (pte_val(x) & (_PAGE_PRESENT | _PAGE_PROTNONE))
 227#define pte_clear(xp)   do { set_pte(xp, __pte(0)); } while (0)
 228
 229#define pages_to_mb(x) ((x) >> (20-PAGE_SHIFT)) /* FIXME: is this
 230                                                   right? */
 231#define pte_page(x)     pfn_to_page(pte_pfn(x))
 232#define pte_pfn(x)  ((pte_val(x) >> PAGE_SHIFT) & __PHYSICAL_MASK)
 233
 234static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
 235{
 236        pte_t pte;
 237        pte_val(pte) = (page_nr << PAGE_SHIFT);
 238        pte_val(pte) |= pgprot_val(pgprot);
 239        pte_val(pte) &= __supported_pte_mask;
 240        return pte;
 241}
 242
 243/*
 244 * The following only work if pte_present() is true.
 245 * Undefined behaviour if not..
 246 */
 247static inline int pte_user(pte_t pte)           { return pte_val(pte) & _PAGE_USER; }
 248extern inline int pte_read(pte_t pte)           { return pte_val(pte) & _PAGE_USER; }
 249extern inline int pte_exec(pte_t pte)           { return pte_val(pte) & _PAGE_USER; }
 250extern inline int pte_dirty(pte_t pte)          { return pte_val(pte) & _PAGE_DIRTY; }
 251extern inline int pte_young(pte_t pte)          { return pte_val(pte) & _PAGE_ACCESSED; }
 252extern inline int pte_write(pte_t pte)          { return pte_val(pte) & _PAGE_RW; }
 253static inline int pte_file(pte_t pte)           { return pte_val(pte) & _PAGE_FILE; }
 254
 255extern inline pte_t pte_rdprotect(pte_t pte)    { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_USER)); return pte; }
 256extern inline pte_t pte_exprotect(pte_t pte)    { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_USER)); return pte; }
 257extern inline pte_t pte_mkclean(pte_t pte)      { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_DIRTY)); return pte; }
 258extern inline pte_t pte_mkold(pte_t pte)        { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_ACCESSED)); return pte; }
 259extern inline pte_t pte_wrprotect(pte_t pte)    { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_RW)); return pte; }
 260extern inline pte_t pte_mkread(pte_t pte)       { set_pte(&pte, __pte(pte_val(pte) | _PAGE_USER)); return pte; }
 261extern inline pte_t pte_mkexec(pte_t pte)       { set_pte(&pte, __pte(pte_val(pte) | _PAGE_USER)); return pte; }
 262extern inline pte_t pte_mkdirty(pte_t pte)      { set_pte(&pte, __pte(pte_val(pte) | _PAGE_DIRTY)); return pte; }
 263extern inline pte_t pte_mkyoung(pte_t pte)      { set_pte(&pte, __pte(pte_val(pte) | _PAGE_ACCESSED)); return pte; }
 264extern inline pte_t pte_mkwrite(pte_t pte)      { set_pte(&pte, __pte(pte_val(pte) | _PAGE_RW)); return pte; }
 265
 266static inline int ptep_test_and_clear_dirty(pte_t *ptep)
 267{
 268        if (!pte_dirty(*ptep))
 269                return 0;
 270        return test_and_clear_bit(_PAGE_BIT_DIRTY, ptep);
 271}
 272
 273static inline int ptep_test_and_clear_young(pte_t *ptep)
 274{
 275        if (!pte_young(*ptep))
 276                return 0;
 277        return test_and_clear_bit(_PAGE_BIT_ACCESSED, ptep);
 278}
 279
 280static inline void ptep_set_wrprotect(pte_t *ptep)              { clear_bit(_PAGE_BIT_RW, ptep); }
 281static inline void ptep_mkdirty(pte_t *ptep)                    { set_bit(_PAGE_BIT_DIRTY, ptep); }
 282
 283/*
 284 * Macro to mark a page protection value as "uncacheable".
 285 */
 286#define pgprot_noncached(prot)  (__pgprot(pgprot_val(prot) | _PAGE_PCD | _PAGE_PWT))
 287
 288#define __LARGE_PTE (_PAGE_PSE|_PAGE_PRESENT) 
 289static inline int pmd_large(pmd_t pte) { 
 290        return (pmd_val(pte) & __LARGE_PTE) == __LARGE_PTE; 
 291}       
 292
 293
 294/*
 295 * Conversion functions: convert a page and protection to a page entry,
 296 * and a page entry and page directory to the page they refer to.
 297 */
 298
 299#define page_pte(page) page_pte_prot(page, __pgprot(0))
 300
 301/*
 302 * Level 4 access.
 303 */
 304#define pgd_page(pgd) ((unsigned long) __va((unsigned long)pgd_val(pgd) & PTE_MASK))
 305#define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
 306#define pgd_offset(mm, addr) ((mm)->pgd + pgd_index(addr))
 307#define pgd_offset_k(address) (init_level4_pgt + pgd_index(address))
 308#define pgd_present(pgd) (pgd_val(pgd) & _PAGE_PRESENT)
 309#define mk_kernel_pgd(address) ((pgd_t){ (address) | _KERNPG_TABLE })
 310
 311/* PUD - Level3 access */
 312/* to find an entry in a page-table-directory. */
 313#define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
 314#define pud_offset(pgd, address) ((pud_t *) pgd_page(*(pgd)) + pud_index(address))
 315#define pud_offset_k(pgd, addr) pud_offset(pgd, addr)
 316#define pud_present(pud) (pud_val(pud) & _PAGE_PRESENT)
 317
 318static inline pud_t *__pud_offset_k(pud_t *pud, unsigned long address)
 319{ 
 320        return pud + pud_index(address);
 321} 
 322
 323/* PMD  - Level 2 access */
 324#define pmd_page_kernel(pmd) ((unsigned long) __va(pmd_val(pmd) & PTE_MASK))
 325#define pmd_page(pmd)           (pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT))
 326
 327#define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
 328#define pmd_offset(dir, address) ((pmd_t *) pud_page(*(dir)) + \
 329                        pmd_index(address))
 330#define pmd_none(x)     (!pmd_val(x))
 331#define pmd_present(x)  (pmd_val(x) & _PAGE_PRESENT)
 332#define pmd_clear(xp)   do { set_pmd(xp, __pmd(0)); } while (0)
 333#define pmd_bad(x)      ((pmd_val(x) & (~PTE_MASK & ~_PAGE_USER)) != _KERNPG_TABLE )
 334#define pfn_pmd(nr,prot) (__pmd(((nr) << PAGE_SHIFT) | pgprot_val(prot)))
 335#define pmd_pfn(x)  ((pmd_val(x) >> PAGE_SHIFT) & __PHYSICAL_MASK)
 336
 337#define pte_to_pgoff(pte) ((pte_val(pte) & PHYSICAL_PAGE_MASK) >> PAGE_SHIFT)
 338#define pgoff_to_pte(off) ((pte_t) { ((off) << PAGE_SHIFT) | _PAGE_FILE })
 339#define PTE_FILE_MAX_BITS __PHYSICAL_MASK_SHIFT
 340
 341/* PTE - Level 1 access. */
 342
 343/* page, protection -> pte */
 344#define mk_pte(page, pgprot)    pfn_pte(page_to_pfn(page), (pgprot))
 345#define mk_pte_huge(entry) (pte_val(entry) |= _PAGE_PRESENT | _PAGE_PSE)
 346 
 347/* physical address -> PTE */
 348static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
 349{ 
 350        pte_t pte;
 351        pte_val(pte) = physpage | pgprot_val(pgprot); 
 352        return pte; 
 353}
 354 
 355/* Change flags of a PTE */
 356extern inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 357{ 
 358        pte_val(pte) &= _PAGE_CHG_MASK;
 359        pte_val(pte) |= pgprot_val(newprot);
 360        pte_val(pte) &= __supported_pte_mask;
 361       return pte; 
 362}
 363
 364#define pte_index(address) \
 365                ((address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
 366#define pte_offset_kernel(dir, address) ((pte_t *) pmd_page_kernel(*(dir)) + \
 367                        pte_index(address))
 368
 369/* x86-64 always has all page tables mapped. */
 370#define pte_offset_map(dir,address) pte_offset_kernel(dir,address)
 371#define pte_offset_map_nested(dir,address) pte_offset_kernel(dir,address)
 372#define pte_unmap(pte) /* NOP */
 373#define pte_unmap_nested(pte) /* NOP */ 
 374
 375#define update_mmu_cache(vma,address,pte) do { } while (0)
 376
 377/* We only update the dirty/accessed state if we set
 378 * the dirty bit by hand in the kernel, since the hardware
 379 * will do the accessed bit for us, and we don't want to
 380 * race with other CPU's that might be updating the dirty
 381 * bit at the same time. */
 382#define  __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
 383#define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
 384        do {                                                              \
 385                if (__dirty) {                                            \
 386                        set_pte(__ptep, __entry);                         \
 387                        flush_tlb_page(__vma, __address);                 \
 388                }                                                         \
 389        } while (0)
 390
 391/* Encode and de-code a swap entry */
 392#define __swp_type(x)                   (((x).val >> 1) & 0x3f)
 393#define __swp_offset(x)                 ((x).val >> 8)
 394#define __swp_entry(type, offset)       ((swp_entry_t) { ((type) << 1) | ((offset) << 8) })
 395#define __pte_to_swp_entry(pte)         ((swp_entry_t) { pte_val(pte) })
 396#define __swp_entry_to_pte(x)           ((pte_t) { (x).val })
 397
 398#endif /* !__ASSEMBLY__ */
 399
 400extern int kern_addr_valid(unsigned long addr); 
 401
 402#define io_remap_page_range(vma, vaddr, paddr, size, prot)              \
 403                remap_pfn_range(vma, vaddr, (paddr) >> PAGE_SHIFT, size, prot)
 404
 405#define HAVE_ARCH_UNMAPPED_AREA
 406
 407#define pgtable_cache_init()   do { } while (0)
 408#define check_pgt_cache()      do { } while (0)
 409
 410#define PAGE_AGP    PAGE_KERNEL_NOCACHE
 411#define HAVE_PAGE_AGP 1
 412
 413/* fs/proc/kcore.c */
 414#define kc_vaddr_to_offset(v) ((v) & __VIRTUAL_MASK)
 415#define kc_offset_to_vaddr(o) \
 416   (((o) & (1UL << (__VIRTUAL_MASK_SHIFT-1))) ? ((o) | (~__VIRTUAL_MASK)) : (o))
 417
 418#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
 419#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
 420#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
 421#define __HAVE_ARCH_PTEP_SET_WRPROTECT
 422#define __HAVE_ARCH_PTEP_MKDIRTY
 423#define __HAVE_ARCH_PTE_SAME
 424#include <asm-generic/pgtable.h>
 425
 426#endif /* _X86_64_PGTABLE_H */
 427