include/asm-i386/pgtable.h

   1 #ifndef _I386_PGTABLE_H
   2 #define _I386_PGTABLE_H
   3
   4 #include <linux/config.h>
   5
   6 /*
   7  * The Linux memory management assumes a three-level page table setup. On
   8  * the i386, we use that, but "fold" the mid level into the top-level page
   9  * table, so that we physically have the same two-level page table as the
  10  * i386 mmu expects.
  11  *
  12  * This file contains the functions and defines necessary to modify and use
  13  * the i386 page table tree.
  14  */
  15 #ifndef __ASSEMBLY__
  16 #include <asm/processor.h>
  17 #include <asm/fixmap.h>
  18 #include <linux/threads.h>
  19
  20 #ifndef _I386_BITOPS_H
  21 #include <asm/bitops.h>
  22 #endif
  23
  24 #include <linux/slab.h>
  25 #include <linux/list.h>
  26 #include <linux/spinlock.h>
  27
  28 /*
  29  * ZERO_PAGE is a global shared page that is always zero: used
  30  * for zero-mapped memory areas etc..
  31  */
  32 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
  33 extern unsigned long empty_zero_page[1024];
  34 extern pgd_t swapper_pg_dir[1024];
  35 extern kmem_cache_t *pgd_cache;
  36 extern kmem_cache_t *pmd_cache;
  37 extern spinlock_t pgd_lock;
  38 extern struct page *pgd_list;
  39
  40 void pmd_ctor(void *, kmem_cache_t *, unsigned long);
  41 void pgd_ctor(void *, kmem_cache_t *, unsigned long);
  42 void pgd_dtor(void *, kmem_cache_t *, unsigned long);
  43 void pgtable_cache_init(void);
  44 void paging_init(void);
  45
  46 /*
  47  * The Linux x86 paging architecture is 'compile-time dual-mode', it
  48  * implements both the traditional 2-level x86 page tables and the
  49  * newer 3-level PAE-mode page tables.
  50  */
  51 #ifdef CONFIG_X86_PAE
  52 # include <asm/pgtable-3level-defs.h>
  53 # define PMD_SIZE       (1UL << PMD_SHIFT)
  54 # define PMD_MASK       (~(PMD_SIZE-1))
  55 #else
  56 # include <asm/pgtable-2level-defs.h>
  57 #endif
  58
  59 #define PGDIR_SIZE      (1UL << PGDIR_SHIFT)
  60 #define PGDIR_MASK      (~(PGDIR_SIZE-1))
  61
  62 #define USER_PTRS_PER_PGD       (TASK_SIZE/PGDIR_SIZE)
  63 #define FIRST_USER_ADDRESS      0
  64
  65 #define USER_PGD_PTRS (PAGE_OFFSET >> PGDIR_SHIFT)
  66 #define KERNEL_PGD_PTRS (PTRS_PER_PGD-USER_PGD_PTRS)
  67
  68 #define TWOLEVEL_PGDIR_SHIFT    22
  69 #define BOOT_USER_PGD_PTRS (__PAGE_OFFSET >> TWOLEVEL_PGDIR_SHIFT)
  70 #define BOOT_KERNEL_PGD_PTRS (1024-BOOT_USER_PGD_PTRS)
  71
  72 /* Just any arbitrary offset to the start of the vmalloc VM area: the
  73  * current 8MB value just means that there will be a 8MB "hole" after the
  74  * physical memory until the kernel virtual memory starts.  That means that
  75  * any out-of-bounds memory accesses will hopefully be caught.
  76  * The vmalloc() routines leaves a hole of 4kB between each vmalloced
  77  * area for the same reason. ;)
  78  */
  79 #define VMALLOC_OFFSET  (8*1024*1024)
  80 #define VMALLOC_START   (((unsigned long) high_memory + vmalloc_earlyreserve + \
  81                         2*VMALLOC_OFFSET-1) & ~(VMALLOC_OFFSET-1))
  82 #ifdef CONFIG_HIGHMEM
  83 # define VMALLOC_END    (PKMAP_BASE-2*PAGE_SIZE)
  84 #else
  85 # define VMALLOC_END    (FIXADDR_START-2*PAGE_SIZE)
  86 #endif
  87
  88 /*
  89  * _PAGE_PSE set in the page directory entry just means that
  90  * the page directory entry points directly to a 4MB-aligned block of
  91  * memory.
  92  */
  93 #define _PAGE_BIT_PRESENT       0
  94 #define _PAGE_BIT_RW            1
  95 #define _PAGE_BIT_USER          2
  96 #define _PAGE_BIT_PWT           3
  97 #define _PAGE_BIT_PCD           4
  98 #define _PAGE_BIT_ACCESSED      5
  99 #define _PAGE_BIT_DIRTY         6
 100 #define _PAGE_BIT_PSE           7       /* 4 MB (or 2MB) page, Pentium+, if present.. */
 101 #define _PAGE_BIT_GLOBAL        8       /* Global TLB entry PPro+ */
 102 #define _PAGE_BIT_UNUSED1       9       /* available for programmer */
 103 #define _PAGE_BIT_UNUSED2       10
 104 #define _PAGE_BIT_UNUSED3       11
 105 #define _PAGE_BIT_NX            63
 106
 107 #define _PAGE_PRESENT   0x001
 108 #define _PAGE_RW        0x002
 109 #define _PAGE_USER      0x004
 110 #define _PAGE_PWT       0x008
 111 #define _PAGE_PCD       0x010
 112 #define _PAGE_ACCESSED  0x020
 113 #define _PAGE_DIRTY     0x040
 114 #define _PAGE_PSE       0x080   /* 4 MB (or 2MB) page, Pentium+, if present.. */
 115 #define _PAGE_GLOBAL    0x100   /* Global TLB entry PPro+ */
 116 #define _PAGE_UNUSED1   0x200   /* available for programmer */
 117 #define _PAGE_UNUSED2   0x400
 118 #define _PAGE_UNUSED3   0x800
 119
 120 /* If _PAGE_PRESENT is clear, we use these: */
 121 #define _PAGE_FILE      0x040   /* nonlinear file mapping, saved PTE; unset:swap */
 122 #define _PAGE_PROTNONE  0x080   /* if the user mapped it with PROT_NONE;
 123                                    pte_present gives true */
 124 #ifdef CONFIG_X86_PAE
 125 #define _PAGE_NX        (1ULL<<_PAGE_BIT_NX)
 126 #else
 127 #define _PAGE_NX        0
 128 #endif
 129
 130 #define _PAGE_TABLE     (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED | _PAGE_DIRTY)
 131 #define _KERNPG_TABLE   (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY)
 132 #define _PAGE_CHG_MASK  (PTE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY)
 133
 134 #define PAGE_NONE \
 135         __pgprot(_PAGE_PROTNONE | _PAGE_ACCESSED)
 136 #define PAGE_SHARED \
 137         __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED)
 138
 139 #define PAGE_SHARED_EXEC \
 140         __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED)
 141 #define PAGE_COPY_NOEXEC \
 142         __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
 143 #define PAGE_COPY_EXEC \
 144         __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
 145 #define PAGE_COPY \
 146         PAGE_COPY_NOEXEC
 147 #define PAGE_READONLY \
 148         __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
 149 #define PAGE_READONLY_EXEC \
 150         __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
 151
 152 #define _PAGE_KERNEL \
 153         (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_NX)
 154 #define _PAGE_KERNEL_EXEC \
 155         (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED)
 156
 157 extern unsigned long long __PAGE_KERNEL, __PAGE_KERNEL_EXEC;
 158 #define __PAGE_KERNEL_RO                (__PAGE_KERNEL & ~_PAGE_RW)
 159 #define __PAGE_KERNEL_NOCACHE           (__PAGE_KERNEL | _PAGE_PCD)
 160 #define __PAGE_KERNEL_LARGE             (__PAGE_KERNEL | _PAGE_PSE)
 161 #define __PAGE_KERNEL_LARGE_EXEC        (__PAGE_KERNEL_EXEC | _PAGE_PSE)
 162
 163 #define PAGE_KERNEL             __pgprot(__PAGE_KERNEL)
 164 #define PAGE_KERNEL_RO          __pgprot(__PAGE_KERNEL_RO)
 165 #define PAGE_KERNEL_EXEC        __pgprot(__PAGE_KERNEL_EXEC)
 166 #define PAGE_KERNEL_NOCACHE     __pgprot(__PAGE_KERNEL_NOCACHE)
 167 #define PAGE_KERNEL_LARGE       __pgprot(__PAGE_KERNEL_LARGE)
 168 #define PAGE_KERNEL_LARGE_EXEC  __pgprot(__PAGE_KERNEL_LARGE_EXEC)
 169
 170 /*
 171  * The i386 can't do page protection for execute, and considers that
 172  * the same are read. Also, write permissions imply read permissions.
 173  * This is the closest we can get..
 174  */
 175 #define __P000  PAGE_NONE
 176 #define __P001  PAGE_READONLY
 177 #define __P010  PAGE_COPY
 178 #define __P011  PAGE_COPY
 179 #define __P100  PAGE_READONLY_EXEC
 180 #define __P101  PAGE_READONLY_EXEC
 181 #define __P110  PAGE_COPY_EXEC
 182 #define __P111  PAGE_COPY_EXEC
 183
 184 #define __S000  PAGE_NONE
 185 #define __S001  PAGE_READONLY
 186 #define __S010  PAGE_SHARED
 187 #define __S011  PAGE_SHARED
 188 #define __S100  PAGE_READONLY_EXEC
 189 #define __S101  PAGE_READONLY_EXEC
 190 #define __S110  PAGE_SHARED_EXEC
 191 #define __S111  PAGE_SHARED_EXEC
 192
 193 /*
 194  * Define this if things work differently on an i386 and an i486:
 195  * it will (on an i486) warn about kernel memory accesses that are
 196  * done without a 'access_ok(VERIFY_WRITE,..)'
 197  */
 198 #undef TEST_ACCESS_OK
 199
 200 /* The boot page tables (all created as a single array) */
 201 extern unsigned long pg0[];
 202
 203 #define pte_present(x)  ((x).pte_low & (_PAGE_PRESENT | _PAGE_PROTNONE))
 204 #define pte_clear(mm,addr,xp)   do { set_pte_at(mm, addr, xp, __pte(0)); } while (0)
 205
 206 /* To avoid harmful races, pmd_none(x) should check only the lower when PAE */
 207 #define pmd_none(x)     (!(unsigned long)pmd_val(x))
 208 #define pmd_present(x)  (pmd_val(x) & _PAGE_PRESENT)
 209 #define pmd_clear(xp)   do { set_pmd(xp, __pmd(0)); } while (0)
 210 #define pmd_bad(x)      ((pmd_val(x) & (~PAGE_MASK & ~_PAGE_USER)) != _KERNPG_TABLE)
 211
 212
 213 #define pages_to_mb(x) ((x) >> (20-PAGE_SHIFT))
 214
 215 /*
 216  * The following only work if pte_present() is true.
 217  * Undefined behaviour if not..
 218  */
 219 #define __LARGE_PTE (_PAGE_PSE | _PAGE_PRESENT)
 220 static inline int pte_user(pte_t pte)           { return (pte).pte_low & _PAGE_USER; }
 221 static inline int pte_read(pte_t pte)           { return (pte).pte_low & _PAGE_USER; }
 222 static inline int pte_dirty(pte_t pte)          { return (pte).pte_low & _PAGE_DIRTY; }
 223 static inline int pte_young(pte_t pte)          { return (pte).pte_low & _PAGE_ACCESSED; }
 224 static inline int pte_write(pte_t pte)          { return (pte).pte_low & _PAGE_RW; }
 225 static inline int pte_huge(pte_t pte)           { return ((pte).pte_low & __LARGE_PTE) == __LARGE_PTE; }
 226
 227 /*
 228  * The following only works if pte_present() is not true.
 229  */
 230 static inline int pte_file(pte_t pte)           { return (pte).pte_low & _PAGE_FILE; }
 231
 232 static inline pte_t pte_rdprotect(pte_t pte)    { (pte).pte_low &= ~_PAGE_USER; return pte; }
 233 static inline pte_t pte_exprotect(pte_t pte)    { (pte).pte_low &= ~_PAGE_USER; return pte; }
 234 static inline pte_t pte_mkclean(pte_t pte)      { (pte).pte_low &= ~_PAGE_DIRTY; return pte; }
 235 static inline pte_t pte_mkold(pte_t pte)        { (pte).pte_low &= ~_PAGE_ACCESSED; return pte; }
 236 static inline pte_t pte_wrprotect(pte_t pte)    { (pte).pte_low &= ~_PAGE_RW; return pte; }
 237 static inline pte_t pte_mkread(pte_t pte)       { (pte).pte_low |= _PAGE_USER; return pte; }
 238 static inline pte_t pte_mkexec(pte_t pte)       { (pte).pte_low |= _PAGE_USER; return pte; }
 239 static inline pte_t pte_mkdirty(pte_t pte)      { (pte).pte_low |= _PAGE_DIRTY; return pte; }
 240 static inline pte_t pte_mkyoung(pte_t pte)      { (pte).pte_low |= _PAGE_ACCESSED; return pte; }
 241 static inline pte_t pte_mkwrite(pte_t pte)      { (pte).pte_low |= _PAGE_RW; return pte; }
 242 static inline pte_t pte_mkhuge(pte_t pte)       { (pte).pte_low |= __LARGE_PTE; return pte; }
 243
 244 #ifdef CONFIG_X86_PAE
 245 # include <asm/pgtable-3level.h>
 246 #else
 247 # include <asm/pgtable-2level.h>
 248 #endif
 249
 250 static inline int ptep_test_and_clear_dirty(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
 251 {
 252         if (!pte_dirty(*ptep))
 253                 return 0;
 254         return test_and_clear_bit(_PAGE_BIT_DIRTY, &ptep->pte_low);
 255 }
 256
 257 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
 258 {
 259         if (!pte_young(*ptep))
 260                 return 0;
 261         return test_and_clear_bit(_PAGE_BIT_ACCESSED, &ptep->pte_low);
 262 }
 263
 264 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm, unsigned long addr, pte_t *ptep, int full)
 265 {
 266         pte_t pte;
 267         if (full) {
 268                 pte = *ptep;
 269                 *ptep = __pte(0);
 270         } else {
 271                 pte = ptep_get_and_clear(mm, addr, ptep);
 272         }
 273         return pte;
 274 }
 275
 276 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
 277 {
 278         clear_bit(_PAGE_BIT_RW, &ptep->pte_low);
 279 }
 280
 281 /*
 282  * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
 283  *
 284  *  dst - pointer to pgd range anwhere on a pgd page
 285  *  src - ""
 286  *  count - the number of pgds to copy.
 287  *
 288  * dst and src can be on the same page, but the range must not overlap,
 289  * and must not cross a page boundary.
 290  */
 291 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
 292 {
 293        memcpy(dst, src, count * sizeof(pgd_t));
 294 }
 295
 296 /*
 297  * Macro to mark a page protection value as "uncacheable".  On processors which do not support
 298  * it, this is a no-op.
 299  */
 300 #define pgprot_noncached(prot)  ((boot_cpu_data.x86 > 3)                                          \
 301                                  ? (__pgprot(pgprot_val(prot) | _PAGE_PCD | _PAGE_PWT)) : (prot))
 302
 303 /*
 304  * Conversion functions: convert a page and protection to a page entry,
 305  * and a page entry and page directory to the page they refer to.
 306  */
 307
 308 #define mk_pte(page, pgprot)    pfn_pte(page_to_pfn(page), (pgprot))
 309
 310 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 311 {
 312         pte.pte_low &= _PAGE_CHG_MASK;
 313         pte.pte_low |= pgprot_val(newprot);
 314 #ifdef CONFIG_X86_PAE
 315         /*
 316          * Chop off the NX bit (if present), and add the NX portion of
 317          * the newprot (if present):
 318          */
 319         pte.pte_high &= ~(1 << (_PAGE_BIT_NX - 32));
 320         pte.pte_high |= (pgprot_val(newprot) >> 32) & \
 321                                         (__supported_pte_mask >> 32);
 322 #endif
 323         return pte;
 324 }
 325
 326 #define page_pte(page) page_pte_prot(page, __pgprot(0))
 327
 328 #define pmd_large(pmd) \
 329 ((pmd_val(pmd) & (_PAGE_PSE|_PAGE_PRESENT)) == (_PAGE_PSE|_PAGE_PRESENT))
 330
 331 /*
 332  * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
 333  *
 334  * this macro returns the index of the entry in the pgd page which would
 335  * control the given virtual address
 336  */
 337 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
 338 #define pgd_index_k(addr) pgd_index(addr)
 339
 340 /*
 341  * pgd_offset() returns a (pgd_t *)
 342  * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
 343  */
 344 #define pgd_offset(mm, address) ((mm)->pgd+pgd_index(address))
 345
 346 /*
 347  * a shortcut which implies the use of the kernel's pgd, instead
 348  * of a process's
 349  */
 350 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
 351
 352 /*
 353  * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
 354  *
 355  * this macro returns the index of the entry in the pmd page which would
 356  * control the given virtual address
 357  */
 358 #define pmd_index(address) \
 359                 (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
 360
 361 /*
 362  * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
 363  *
 364  * this macro returns the index of the entry in the pte page which would
 365  * control the given virtual address
 366  */
 367 #define pte_index(address) \
 368                 (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
 369 #define pte_offset_kernel(dir, address) \
 370         ((pte_t *) pmd_page_kernel(*(dir)) +  pte_index(address))
 371
 372 #define pmd_page(pmd) (pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT))
 373
 374 #define pmd_page_kernel(pmd) \
 375                 ((unsigned long) __va(pmd_val(pmd) & PAGE_MASK))
 376
 377 /*
 378  * Helper function that returns the kernel pagetable entry controlling
 379  * the virtual address 'address'. NULL means no pagetable entry present.
 380  * NOTE: the return type is pte_t but if the pmd is PSE then we return it
 381  * as a pte too.
 382  */
 383 extern pte_t *lookup_address(unsigned long address);
 384
 385 /*
 386  * Make a given kernel text page executable/non-executable.
 387  * Returns the previous executability setting of that page (which
 388  * is used to restore the previous state). Used by the SMP bootup code.
 389  * NOTE: this is an __init function for security reasons.
 390  */
 391 #ifdef CONFIG_X86_PAE
 392  extern int set_kernel_exec(unsigned long vaddr, int enable);
 393 #else
 394  static inline int set_kernel_exec(unsigned long vaddr, int enable) { return 0;}
 395 #endif
 396
 397 extern void noexec_setup(const char *str);
 398
 399 #if defined(CONFIG_HIGHPTE)
 400 #define pte_offset_map(dir, address) \
 401         ((pte_t *)kmap_atomic(pmd_page(*(dir)),KM_PTE0) + pte_index(address))
 402 #define pte_offset_map_nested(dir, address) \
 403         ((pte_t *)kmap_atomic(pmd_page(*(dir)),KM_PTE1) + pte_index(address))
 404 #define pte_unmap(pte) kunmap_atomic(pte, KM_PTE0)
 405 #define pte_unmap_nested(pte) kunmap_atomic(pte, KM_PTE1)
 406 #else
 407 #define pte_offset_map(dir, address) \
 408         ((pte_t *)page_address(pmd_page(*(dir))) + pte_index(address))
 409 #define pte_offset_map_nested(dir, address) pte_offset_map(dir, address)
 410 #define pte_unmap(pte) do { } while (0)
 411 #define pte_unmap_nested(pte) do { } while (0)
 412 #endif
 413
 414 /*
 415  * The i386 doesn't have any external MMU info: the kernel page
 416  * tables contain all the necessary information.
 417  *
 418  * Also, we only update the dirty/accessed state if we set
 419  * the dirty bit by hand in the kernel, since the hardware
 420  * will do the accessed bit for us, and we don't want to
 421  * race with other CPU's that might be updating the dirty
 422  * bit at the same time.
 423  */
 424 #define update_mmu_cache(vma,address,pte) do { } while (0)
 425 #define  __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
 426 #define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
 427         do {                                                              \
 428                 if (__dirty) {                                            \
 429                         (__ptep)->pte_low = (__entry).pte_low;            \
 430                         flush_tlb_page(__vma, __address);                 \
 431                 }                                                         \
 432         } while (0)
 433
 434 #endif /* !__ASSEMBLY__ */
 435
 436 #ifdef CONFIG_FLATMEM
 437 #define kern_addr_valid(addr)   (1)
 438 #endif /* CONFIG_FLATMEM */
 439
 440 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot)         \
 441                 remap_pfn_range(vma, vaddr, pfn, size, prot)
 442
 443 #define MK_IOSPACE_PFN(space, pfn)      (pfn)
 444 #define GET_IOSPACE(pfn)                0
 445 #define GET_PFN(pfn)                    (pfn)
 446
 447 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
 448 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
 449 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
 450 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
 451 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
 452 #define __HAVE_ARCH_PTE_SAME
 453 #include <asm-generic/pgtable.h>
 454
 455 #endif /* _I386_PGTABLE_H */