include/asm-arm/pgtable.h

   1 /*
   2  *  linux/include/asm-arm/pgtable.h
   3  *
   4  *  Copyright (C) 1995-2002 Russell King
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License version 2 as
   8  * published by the Free Software Foundation.
   9  */
  10 #ifndef _ASMARM_PGTABLE_H
  11 #define _ASMARM_PGTABLE_H
  12
  13 #include <asm-generic/4level-fixup.h>
  14 #include <asm/proc-fns.h>
  15
  16 #ifndef CONFIG_MMU
  17
  18 #include "pgtable-nommu.h"
  19
  20 #else
  21
  22 #include <asm/memory.h>
  23 #include <asm/arch/vmalloc.h>
  24
  25 /*
  26  * Just any arbitrary offset to the start of the vmalloc VM area: the
  27  * current 8MB value just means that there will be a 8MB "hole" after the
  28  * physical memory until the kernel virtual memory starts.  That means that
  29  * any out-of-bounds memory accesses will hopefully be caught.
  30  * The vmalloc() routines leaves a hole of 4kB between each vmalloced
  31  * area for the same reason. ;)
  32  *
  33  * Note that platforms may override VMALLOC_START, but they must provide
  34  * VMALLOC_END.  VMALLOC_END defines the (exclusive) limit of this space,
  35  * which may not overlap IO space.
  36  */
  37 #ifndef VMALLOC_START
  38 #define VMALLOC_OFFSET          (8*1024*1024)
  39 #define VMALLOC_START           (((unsigned long)high_memory + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1))
  40 #endif
  41
  42 /*
  43  * Hardware-wise, we have a two level page table structure, where the first
  44  * level has 4096 entries, and the second level has 256 entries.  Each entry
  45  * is one 32-bit word.  Most of the bits in the second level entry are used
  46  * by hardware, and there aren't any "accessed" and "dirty" bits.
  47  *
  48  * Linux on the other hand has a three level page table structure, which can
  49  * be wrapped to fit a two level page table structure easily - using the PGD
  50  * and PTE only.  However, Linux also expects one "PTE" table per page, and
  51  * at least a "dirty" bit.
  52  *
  53  * Therefore, we tweak the implementation slightly - we tell Linux that we
  54  * have 2048 entries in the first level, each of which is 8 bytes (iow, two
  55  * hardware pointers to the second level.)  The second level contains two
  56  * hardware PTE tables arranged contiguously, followed by Linux versions
  57  * which contain the state information Linux needs.  We, therefore, end up
  58  * with 512 entries in the "PTE" level.
  59  *
  60  * This leads to the page tables having the following layout:
  61  *
  62  *    pgd             pte
  63  * |        |
  64  * +--------+ +0
  65  * |        |-----> +------------+ +0
  66  * +- - - - + +4    |  h/w pt 0  |
  67  * |        |-----> +------------+ +1024
  68  * +--------+ +8    |  h/w pt 1  |
  69  * |        |       +------------+ +2048
  70  * +- - - - +       | Linux pt 0 |
  71  * |        |       +------------+ +3072
  72  * +--------+       | Linux pt 1 |
  73  * |        |       +------------+ +4096
  74  *
  75  * See L_PTE_xxx below for definitions of bits in the "Linux pt", and
  76  * PTE_xxx for definitions of bits appearing in the "h/w pt".
  77  *
  78  * PMD_xxx definitions refer to bits in the first level page table.
  79  *
  80  * The "dirty" bit is emulated by only granting hardware write permission
  81  * iff the page is marked "writable" and "dirty" in the Linux PTE.  This
  82  * means that a write to a clean page will cause a permission fault, and
  83  * the Linux MM layer will mark the page dirty via handle_pte_fault().
  84  * For the hardware to notice the permission change, the TLB entry must
  85  * be flushed, and ptep_establish() does that for us.
  86  *
  87  * The "accessed" or "young" bit is emulated by a similar method; we only
  88  * allow accesses to the page if the "young" bit is set.  Accesses to the
  89  * page will cause a fault, and handle_pte_fault() will set the young bit
  90  * for us as long as the page is marked present in the corresponding Linux
  91  * PTE entry.  Again, ptep_establish() will ensure that the TLB is up to
  92  * date.
  93  *
  94  * However, when the "young" bit is cleared, we deny access to the page
  95  * by clearing the hardware PTE.  Currently Linux does not flush the TLB
  96  * for us in this case, which means the TLB will retain the transation
  97  * until either the TLB entry is evicted under pressure, or a context
  98  * switch which changes the user space mapping occurs.
  99  */
 100 #define PTRS_PER_PTE            512
 101 #define PTRS_PER_PMD            1
 102 #define PTRS_PER_PGD            2048
 103
 104 /*
 105  * PMD_SHIFT determines the size of the area a second-level page table can map
 106  * PGDIR_SHIFT determines what a third-level page table entry can map
 107  */
 108 #define PMD_SHIFT               21
 109 #define PGDIR_SHIFT             21
 110
 111 #define LIBRARY_TEXT_START      0x0c000000
 112
 113 #ifndef __ASSEMBLY__
 114 extern void __pte_error(const char *file, int line, unsigned long val);
 115 extern void __pmd_error(const char *file, int line, unsigned long val);
 116 extern void __pgd_error(const char *file, int line, unsigned long val);
 117
 118 #define pte_ERROR(pte)          __pte_error(__FILE__, __LINE__, pte_val(pte))
 119 #define pmd_ERROR(pmd)          __pmd_error(__FILE__, __LINE__, pmd_val(pmd))
 120 #define pgd_ERROR(pgd)          __pgd_error(__FILE__, __LINE__, pgd_val(pgd))
 121 #endif /* !__ASSEMBLY__ */
 122
 123 #define PMD_SIZE                (1UL << PMD_SHIFT)
 124 #define PMD_MASK                (~(PMD_SIZE-1))
 125 #define PGDIR_SIZE              (1UL << PGDIR_SHIFT)
 126 #define PGDIR_MASK              (~(PGDIR_SIZE-1))
 127
 128 /*
 129  * This is the lowest virtual address we can permit any user space
 130  * mapping to be mapped at.  This is particularly important for
 131  * non-high vector CPUs.
 132  */
 133 #define FIRST_USER_ADDRESS      PAGE_SIZE
 134
 135 #define FIRST_USER_PGD_NR       1
 136 #define USER_PTRS_PER_PGD       ((TASK_SIZE/PGDIR_SIZE) - FIRST_USER_PGD_NR)
 137
 138 /*
 139  * section address mask and size definitions.
 140  */
 141 #define SECTION_SHIFT           20
 142 #define SECTION_SIZE            (1UL << SECTION_SHIFT)
 143 #define SECTION_MASK            (~(SECTION_SIZE-1))
 144
 145 /*
 146  * ARMv6 supersection address mask and size definitions.
 147  */
 148 #define SUPERSECTION_SHIFT      24
 149 #define SUPERSECTION_SIZE       (1UL << SUPERSECTION_SHIFT)
 150 #define SUPERSECTION_MASK       (~(SUPERSECTION_SIZE-1))
 151
 152 /*
 153  * "Linux" PTE definitions.
 154  *
 155  * We keep two sets of PTEs - the hardware and the linux version.
 156  * This allows greater flexibility in the way we map the Linux bits
 157  * onto the hardware tables, and allows us to have YOUNG and DIRTY
 158  * bits.
 159  *
 160  * The PTE table pointer refers to the hardware entries; the "Linux"
 161  * entries are stored 1024 bytes below.
 162  */
 163 #define L_PTE_PRESENT           (1 << 0)
 164 #define L_PTE_FILE              (1 << 1)        /* only when !PRESENT */
 165 #define L_PTE_YOUNG             (1 << 1)
 166 #define L_PTE_BUFFERABLE        (1 << 2)        /* matches PTE */
 167 #define L_PTE_CACHEABLE         (1 << 3)        /* matches PTE */
 168 #define L_PTE_USER              (1 << 4)
 169 #define L_PTE_WRITE             (1 << 5)
 170 #define L_PTE_EXEC              (1 << 6)
 171 #define L_PTE_DIRTY             (1 << 7)
 172 #define L_PTE_COHERENT          (1 << 9)        /* I/O coherent (xsc3) */
 173 #define L_PTE_SHARED            (1 << 10)       /* shared between CPUs (v6) */
 174 #define L_PTE_ASID              (1 << 11)       /* non-global (use ASID, v6) */
 175
 176 #ifndef __ASSEMBLY__
 177
 178 /*
 179  * The following macros handle the cache and bufferable bits...
 180  */
 181 #define _L_PTE_DEFAULT  L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_CACHEABLE | L_PTE_BUFFERABLE
 182 #define _L_PTE_READ     L_PTE_USER | L_PTE_EXEC
 183
 184 extern pgprot_t         pgprot_kernel;
 185
 186 #define PAGE_NONE       __pgprot(_L_PTE_DEFAULT)
 187 #define PAGE_COPY       __pgprot(_L_PTE_DEFAULT | _L_PTE_READ)
 188 #define PAGE_SHARED     __pgprot(_L_PTE_DEFAULT | _L_PTE_READ | L_PTE_WRITE)
 189 #define PAGE_READONLY   __pgprot(_L_PTE_DEFAULT | _L_PTE_READ)
 190 #define PAGE_KERNEL     pgprot_kernel
 191
 192 #endif /* __ASSEMBLY__ */
 193
 194 /*
 195  * The table below defines the page protection levels that we insert into our
 196  * Linux page table version.  These get translated into the best that the
 197  * architecture can perform.  Note that on most ARM hardware:
 198  *  1) We cannot do execute protection
 199  *  2) If we could do execute protection, then read is implied
 200  *  3) write implies read permissions
 201  */
 202 #define __P000  PAGE_NONE
 203 #define __P001  PAGE_READONLY
 204 #define __P010  PAGE_COPY
 205 #define __P011  PAGE_COPY
 206 #define __P100  PAGE_READONLY
 207 #define __P101  PAGE_READONLY
 208 #define __P110  PAGE_COPY
 209 #define __P111  PAGE_COPY
 210
 211 #define __S000  PAGE_NONE
 212 #define __S001  PAGE_READONLY
 213 #define __S010  PAGE_SHARED
 214 #define __S011  PAGE_SHARED
 215 #define __S100  PAGE_READONLY
 216 #define __S101  PAGE_READONLY
 217 #define __S110  PAGE_SHARED
 218 #define __S111  PAGE_SHARED
 219
 220 #ifndef __ASSEMBLY__
 221 /*
 222  * ZERO_PAGE is a global shared page that is always zero: used
 223  * for zero-mapped memory areas etc..
 224  */
 225 extern struct page *empty_zero_page;
 226 #define ZERO_PAGE(vaddr)        (empty_zero_page)
 227
 228 #define pte_pfn(pte)            (pte_val(pte) >> PAGE_SHIFT)
 229 #define pfn_pte(pfn,prot)       (__pte(((pfn) << PAGE_SHIFT) | pgprot_val(prot)))
 230
 231 #define pte_none(pte)           (!pte_val(pte))
 232 #define pte_clear(mm,addr,ptep) set_pte_at((mm),(addr),(ptep), __pte(0))
 233 #define pte_page(pte)           (pfn_to_page(pte_pfn(pte)))
 234 #define pte_offset_kernel(dir,addr)     (pmd_page_vaddr(*(dir)) + __pte_index(addr))
 235 #define pte_offset_map(dir,addr)        (pmd_page_vaddr(*(dir)) + __pte_index(addr))
 236 #define pte_offset_map_nested(dir,addr) (pmd_page_vaddr(*(dir)) + __pte_index(addr))
 237 #define pte_unmap(pte)          do { } while (0)
 238 #define pte_unmap_nested(pte)   do { } while (0)
 239
 240 #define set_pte(ptep, pte)      cpu_set_pte(ptep,pte)
 241 #define set_pte_at(mm,addr,ptep,pteval) set_pte(ptep,pteval)
 242
 243 /*
 244  * The following only work if pte_present() is true.
 245  * Undefined behaviour if not..
 246  */
 247 #define pte_present(pte)        (pte_val(pte) & L_PTE_PRESENT)
 248 #define pte_read(pte)           (pte_val(pte) & L_PTE_USER)
 249 #define pte_write(pte)          (pte_val(pte) & L_PTE_WRITE)
 250 #define pte_exec(pte)           (pte_val(pte) & L_PTE_EXEC)
 251 #define pte_dirty(pte)          (pte_val(pte) & L_PTE_DIRTY)
 252 #define pte_young(pte)          (pte_val(pte) & L_PTE_YOUNG)
 253
 254 /*
 255  * The following only works if pte_present() is not true.
 256  */
 257 #define pte_file(pte)           (pte_val(pte) & L_PTE_FILE)
 258 #define pte_to_pgoff(x)         (pte_val(x) >> 2)
 259 #define pgoff_to_pte(x)         __pte(((x) << 2) | L_PTE_FILE)
 260
 261 #define PTE_FILE_MAX_BITS       30
 262
 263 #define PTE_BIT_FUNC(fn,op) \
 264 static inline pte_t pte_##fn(pte_t pte) { pte_val(pte) op; return pte; }
 265
 266 /*PTE_BIT_FUNC(rdprotect, &= ~L_PTE_USER);*/
 267 /*PTE_BIT_FUNC(mkread,    |= L_PTE_USER);*/
 268 PTE_BIT_FUNC(wrprotect, &= ~L_PTE_WRITE);
 269 PTE_BIT_FUNC(mkwrite,   |= L_PTE_WRITE);
 270 PTE_BIT_FUNC(exprotect, &= ~L_PTE_EXEC);
 271 PTE_BIT_FUNC(mkexec,    |= L_PTE_EXEC);
 272 PTE_BIT_FUNC(mkclean,   &= ~L_PTE_DIRTY);
 273 PTE_BIT_FUNC(mkdirty,   |= L_PTE_DIRTY);
 274 PTE_BIT_FUNC(mkold,     &= ~L_PTE_YOUNG);
 275 PTE_BIT_FUNC(mkyoung,   |= L_PTE_YOUNG);
 276
 277 /*
 278  * Mark the prot value as uncacheable and unbufferable.
 279  */
 280 #define pgprot_noncached(prot)  __pgprot(pgprot_val(prot) & ~(L_PTE_CACHEABLE | L_PTE_BUFFERABLE))
 281 #define pgprot_writecombine(prot) __pgprot(pgprot_val(prot) & ~L_PTE_CACHEABLE)
 282
 283 #define pmd_none(pmd)           (!pmd_val(pmd))
 284 #define pmd_present(pmd)        (pmd_val(pmd))
 285 #define pmd_bad(pmd)            (pmd_val(pmd) & 2)
 286
 287 #define copy_pmd(pmdpd,pmdps)           \
 288         do {                            \
 289                 pmdpd[0] = pmdps[0];    \
 290                 pmdpd[1] = pmdps[1];    \
 291                 flush_pmd_entry(pmdpd); \
 292         } while (0)
 293
 294 #define pmd_clear(pmdp)                 \
 295         do {                            \
 296                 pmdp[0] = __pmd(0);     \
 297                 pmdp[1] = __pmd(0);     \
 298                 clean_pmd_entry(pmdp);  \
 299         } while (0)
 300
 301 static inline pte_t *pmd_page_vaddr(pmd_t pmd)
 302 {
 303         unsigned long ptr;
 304
 305         ptr = pmd_val(pmd) & ~(PTRS_PER_PTE * sizeof(void *) - 1);
 306         ptr += PTRS_PER_PTE * sizeof(void *);
 307
 308         return __va(ptr);
 309 }
 310
 311 #define pmd_page(pmd) virt_to_page(__va(pmd_val(pmd)))
 312
 313 /*
 314  * Permanent address of a page. We never have highmem, so this is trivial.
 315  */
 316 #define pages_to_mb(x)          ((x) >> (20 - PAGE_SHIFT))
 317
 318 /*
 319  * Conversion functions: convert a page and protection to a page entry,
 320  * and a page entry and page directory to the page they refer to.
 321  */
 322 #define mk_pte(page,prot)       pfn_pte(page_to_pfn(page),prot)
 323
 324 /*
 325  * The "pgd_xxx()" functions here are trivial for a folded two-level
 326  * setup: the pgd is never bad, and a pmd always exists (as it's folded
 327  * into the pgd entry)
 328  */
 329 #define pgd_none(pgd)           (0)
 330 #define pgd_bad(pgd)            (0)
 331 #define pgd_present(pgd)        (1)
 332 #define pgd_clear(pgdp)         do { } while (0)
 333 #define set_pgd(pgd,pgdp)       do { } while (0)
 334
 335 /* to find an entry in a page-table-directory */
 336 #define pgd_index(addr)         ((addr) >> PGDIR_SHIFT)
 337
 338 #define pgd_offset(mm, addr)    ((mm)->pgd+pgd_index(addr))
 339
 340 /* to find an entry in a kernel page-table-directory */
 341 #define pgd_offset_k(addr)      pgd_offset(&init_mm, addr)
 342
 343 /* Find an entry in the second-level page table.. */
 344 #define pmd_offset(dir, addr)   ((pmd_t *)(dir))
 345
 346 /* Find an entry in the third-level page table.. */
 347 #define __pte_index(addr)       (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
 348
 349 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 350 {
 351         const unsigned long mask = L_PTE_EXEC | L_PTE_WRITE | L_PTE_USER;
 352         pte_val(pte) = (pte_val(pte) & ~mask) | (pgprot_val(newprot) & mask);
 353         return pte;
 354 }
 355
 356 extern pgd_t swapper_pg_dir[PTRS_PER_PGD];
 357
 358 /* Encode and decode a swap entry.
 359  *
 360  * We support up to 32GB of swap on 4k machines
 361  */
 362 #define __swp_type(x)           (((x).val >> 2) & 0x7f)
 363 #define __swp_offset(x)         ((x).val >> 9)
 364 #define __swp_entry(type,offset) ((swp_entry_t) { ((type) << 2) | ((offset) << 9) })
 365 #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
 366 #define __swp_entry_to_pte(swp) ((pte_t) { (swp).val })
 367
 368 /* Needs to be defined here and not in linux/mm.h, as it is arch dependent */
 369 /* FIXME: this is not correct */
 370 #define kern_addr_valid(addr)   (1)
 371
 372 #include <asm-generic/pgtable.h>
 373
 374 /*
 375  * We provide our own arch_get_unmapped_area to cope with VIPT caches.
 376  */
 377 #define HAVE_ARCH_UNMAPPED_AREA
 378
 379 /*
 380  * remap a physical page `pfn' of size `size' with page protection `prot'
 381  * into virtual address `from'
 382  */
 383 #define io_remap_pfn_range(vma,from,pfn,size,prot) \
 384                 remap_pfn_range(vma, from, pfn, size, prot)
 385
 386 #define MK_IOSPACE_PFN(space, pfn)      (pfn)
 387 #define GET_IOSPACE(pfn)                0
 388 #define GET_PFN(pfn)                    (pfn)
 389
 390 #define pgtable_cache_init() do { } while (0)
 391
 392 #endif /* !__ASSEMBLY__ */
 393
 394 #endif /* CONFIG_MMU */
 395
 396 #endif /* _ASMARM_PGTABLE_H */