Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | #ifndef _ASM_IA64_PGTABLE_H |
| 2 | #define _ASM_IA64_PGTABLE_H |
| 3 | |
| 4 | /* |
| 5 | * This file contains the functions and defines necessary to modify and use |
| 6 | * the IA-64 page table tree. |
| 7 | * |
| 8 | * This hopefully works with any (fixed) IA-64 page-size, as defined |
| 9 | * in <asm/page.h>. |
| 10 | * |
David Mosberger-Tang | ad597bd | 2005-06-08 10:45:00 -0700 | [diff] [blame] | 11 | * Copyright (C) 1998-2005 Hewlett-Packard Co |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 12 | * David Mosberger-Tang <davidm@hpl.hp.com> |
| 13 | */ |
| 14 | |
| 15 | #include <linux/config.h> |
| 16 | |
| 17 | #include <asm/mman.h> |
| 18 | #include <asm/page.h> |
| 19 | #include <asm/processor.h> |
| 20 | #include <asm/system.h> |
| 21 | #include <asm/types.h> |
| 22 | |
| 23 | #define IA64_MAX_PHYS_BITS 50 /* max. number of physical address bits (architected) */ |
| 24 | |
| 25 | /* |
| 26 | * First, define the various bits in a PTE. Note that the PTE format |
| 27 | * matches the VHPT short format, the firt doubleword of the VHPD long |
| 28 | * format, and the first doubleword of the TLB insertion format. |
| 29 | */ |
| 30 | #define _PAGE_P_BIT 0 |
| 31 | #define _PAGE_A_BIT 5 |
| 32 | #define _PAGE_D_BIT 6 |
| 33 | |
| 34 | #define _PAGE_P (1 << _PAGE_P_BIT) /* page present bit */ |
| 35 | #define _PAGE_MA_WB (0x0 << 2) /* write back memory attribute */ |
| 36 | #define _PAGE_MA_UC (0x4 << 2) /* uncacheable memory attribute */ |
| 37 | #define _PAGE_MA_UCE (0x5 << 2) /* UC exported attribute */ |
| 38 | #define _PAGE_MA_WC (0x6 << 2) /* write coalescing memory attribute */ |
| 39 | #define _PAGE_MA_NAT (0x7 << 2) /* not-a-thing attribute */ |
| 40 | #define _PAGE_MA_MASK (0x7 << 2) |
| 41 | #define _PAGE_PL_0 (0 << 7) /* privilege level 0 (kernel) */ |
| 42 | #define _PAGE_PL_1 (1 << 7) /* privilege level 1 (unused) */ |
| 43 | #define _PAGE_PL_2 (2 << 7) /* privilege level 2 (unused) */ |
| 44 | #define _PAGE_PL_3 (3 << 7) /* privilege level 3 (user) */ |
| 45 | #define _PAGE_PL_MASK (3 << 7) |
| 46 | #define _PAGE_AR_R (0 << 9) /* read only */ |
| 47 | #define _PAGE_AR_RX (1 << 9) /* read & execute */ |
| 48 | #define _PAGE_AR_RW (2 << 9) /* read & write */ |
| 49 | #define _PAGE_AR_RWX (3 << 9) /* read, write & execute */ |
| 50 | #define _PAGE_AR_R_RW (4 << 9) /* read / read & write */ |
| 51 | #define _PAGE_AR_RX_RWX (5 << 9) /* read & exec / read, write & exec */ |
| 52 | #define _PAGE_AR_RWX_RW (6 << 9) /* read, write & exec / read & write */ |
| 53 | #define _PAGE_AR_X_RX (7 << 9) /* exec & promote / read & exec */ |
| 54 | #define _PAGE_AR_MASK (7 << 9) |
| 55 | #define _PAGE_AR_SHIFT 9 |
| 56 | #define _PAGE_A (1 << _PAGE_A_BIT) /* page accessed bit */ |
| 57 | #define _PAGE_D (1 << _PAGE_D_BIT) /* page dirty bit */ |
| 58 | #define _PAGE_PPN_MASK (((__IA64_UL(1) << IA64_MAX_PHYS_BITS) - 1) & ~0xfffUL) |
| 59 | #define _PAGE_ED (__IA64_UL(1) << 52) /* exception deferral */ |
| 60 | #define _PAGE_PROTNONE (__IA64_UL(1) << 63) |
| 61 | |
| 62 | /* Valid only for a PTE with the present bit cleared: */ |
| 63 | #define _PAGE_FILE (1 << 1) /* see swap & file pte remarks below */ |
| 64 | |
| 65 | #define _PFN_MASK _PAGE_PPN_MASK |
| 66 | /* Mask of bits which may be changed by pte_modify(); the odd bits are there for _PAGE_PROTNONE */ |
| 67 | #define _PAGE_CHG_MASK (_PAGE_P | _PAGE_PROTNONE | _PAGE_PL_MASK | _PAGE_AR_MASK | _PAGE_ED) |
| 68 | |
| 69 | #define _PAGE_SIZE_4K 12 |
| 70 | #define _PAGE_SIZE_8K 13 |
| 71 | #define _PAGE_SIZE_16K 14 |
| 72 | #define _PAGE_SIZE_64K 16 |
| 73 | #define _PAGE_SIZE_256K 18 |
| 74 | #define _PAGE_SIZE_1M 20 |
| 75 | #define _PAGE_SIZE_4M 22 |
| 76 | #define _PAGE_SIZE_16M 24 |
| 77 | #define _PAGE_SIZE_64M 26 |
| 78 | #define _PAGE_SIZE_256M 28 |
| 79 | #define _PAGE_SIZE_1G 30 |
| 80 | #define _PAGE_SIZE_4G 32 |
| 81 | |
| 82 | #define __ACCESS_BITS _PAGE_ED | _PAGE_A | _PAGE_P | _PAGE_MA_WB |
| 83 | #define __DIRTY_BITS_NO_ED _PAGE_A | _PAGE_P | _PAGE_D | _PAGE_MA_WB |
| 84 | #define __DIRTY_BITS _PAGE_ED | __DIRTY_BITS_NO_ED |
| 85 | |
| 86 | /* |
| 87 | * Definitions for first level: |
| 88 | * |
| 89 | * PGDIR_SHIFT determines what a first-level page table entry can map. |
| 90 | */ |
| 91 | #define PGDIR_SHIFT (PAGE_SHIFT + 2*(PAGE_SHIFT-3)) |
| 92 | #define PGDIR_SIZE (__IA64_UL(1) << PGDIR_SHIFT) |
| 93 | #define PGDIR_MASK (~(PGDIR_SIZE-1)) |
| 94 | #define PTRS_PER_PGD (1UL << (PAGE_SHIFT-3)) |
| 95 | #define USER_PTRS_PER_PGD (5*PTRS_PER_PGD/8) /* regions 0-4 are user regions */ |
Hugh Dickins | d455a36 | 2005-04-19 13:29:23 -0700 | [diff] [blame] | 96 | #define FIRST_USER_ADDRESS 0 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 97 | |
| 98 | /* |
| 99 | * Definitions for second level: |
| 100 | * |
| 101 | * PMD_SHIFT determines the size of the area a second-level page table |
| 102 | * can map. |
| 103 | */ |
| 104 | #define PMD_SHIFT (PAGE_SHIFT + (PAGE_SHIFT-3)) |
| 105 | #define PMD_SIZE (1UL << PMD_SHIFT) |
| 106 | #define PMD_MASK (~(PMD_SIZE-1)) |
| 107 | #define PTRS_PER_PMD (1UL << (PAGE_SHIFT-3)) |
| 108 | |
| 109 | /* |
| 110 | * Definitions for third level: |
| 111 | */ |
| 112 | #define PTRS_PER_PTE (__IA64_UL(1) << (PAGE_SHIFT-3)) |
| 113 | |
| 114 | /* |
| 115 | * All the normal masks have the "page accessed" bits on, as any time |
| 116 | * they are used, the page is accessed. They are cleared only by the |
| 117 | * page-out routines. |
| 118 | */ |
| 119 | #define PAGE_NONE __pgprot(_PAGE_PROTNONE | _PAGE_A) |
| 120 | #define PAGE_SHARED __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RW) |
| 121 | #define PAGE_READONLY __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_R) |
| 122 | #define PAGE_COPY __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_R) |
| 123 | #define PAGE_COPY_EXEC __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RX) |
| 124 | #define PAGE_GATE __pgprot(__ACCESS_BITS | _PAGE_PL_0 | _PAGE_AR_X_RX) |
| 125 | #define PAGE_KERNEL __pgprot(__DIRTY_BITS | _PAGE_PL_0 | _PAGE_AR_RWX) |
| 126 | #define PAGE_KERNELRX __pgprot(__ACCESS_BITS | _PAGE_PL_0 | _PAGE_AR_RX) |
| 127 | |
| 128 | # ifndef __ASSEMBLY__ |
| 129 | |
| 130 | #include <asm/bitops.h> |
| 131 | #include <asm/cacheflush.h> |
| 132 | #include <asm/mmu_context.h> |
| 133 | #include <asm/processor.h> |
| 134 | |
| 135 | /* |
| 136 | * Next come the mappings that determine how mmap() protection bits |
| 137 | * (PROT_EXEC, PROT_READ, PROT_WRITE, PROT_NONE) get implemented. The |
| 138 | * _P version gets used for a private shared memory segment, the _S |
| 139 | * version gets used for a shared memory segment with MAP_SHARED on. |
| 140 | * In a private shared memory segment, we do a copy-on-write if a task |
| 141 | * attempts to write to the page. |
| 142 | */ |
| 143 | /* xwr */ |
| 144 | #define __P000 PAGE_NONE |
| 145 | #define __P001 PAGE_READONLY |
| 146 | #define __P010 PAGE_READONLY /* write to priv pg -> copy & make writable */ |
| 147 | #define __P011 PAGE_READONLY /* ditto */ |
| 148 | #define __P100 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_X_RX) |
| 149 | #define __P101 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RX) |
| 150 | #define __P110 PAGE_COPY_EXEC |
| 151 | #define __P111 PAGE_COPY_EXEC |
| 152 | |
| 153 | #define __S000 PAGE_NONE |
| 154 | #define __S001 PAGE_READONLY |
| 155 | #define __S010 PAGE_SHARED /* we don't have (and don't need) write-only */ |
| 156 | #define __S011 PAGE_SHARED |
| 157 | #define __S100 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_X_RX) |
| 158 | #define __S101 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RX) |
| 159 | #define __S110 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RWX) |
| 160 | #define __S111 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RWX) |
| 161 | |
| 162 | #define pgd_ERROR(e) printk("%s:%d: bad pgd %016lx.\n", __FILE__, __LINE__, pgd_val(e)) |
| 163 | #define pmd_ERROR(e) printk("%s:%d: bad pmd %016lx.\n", __FILE__, __LINE__, pmd_val(e)) |
| 164 | #define pte_ERROR(e) printk("%s:%d: bad pte %016lx.\n", __FILE__, __LINE__, pte_val(e)) |
| 165 | |
| 166 | |
| 167 | /* |
| 168 | * Some definitions to translate between mem_map, PTEs, and page addresses: |
| 169 | */ |
| 170 | |
| 171 | |
| 172 | /* Quick test to see if ADDR is a (potentially) valid physical address. */ |
| 173 | static inline long |
| 174 | ia64_phys_addr_valid (unsigned long addr) |
| 175 | { |
| 176 | return (addr & (local_cpu_data->unimpl_pa_mask)) == 0; |
| 177 | } |
| 178 | |
| 179 | /* |
| 180 | * kern_addr_valid(ADDR) tests if ADDR is pointing to valid kernel |
| 181 | * memory. For the return value to be meaningful, ADDR must be >= |
| 182 | * PAGE_OFFSET. This operation can be relatively expensive (e.g., |
| 183 | * require a hash-, or multi-level tree-lookup or something of that |
| 184 | * sort) but it guarantees to return TRUE only if accessing the page |
| 185 | * at that address does not cause an error. Note that there may be |
| 186 | * addresses for which kern_addr_valid() returns FALSE even though an |
| 187 | * access would not cause an error (e.g., this is typically true for |
| 188 | * memory mapped I/O regions. |
| 189 | * |
| 190 | * XXX Need to implement this for IA-64. |
| 191 | */ |
| 192 | #define kern_addr_valid(addr) (1) |
| 193 | |
| 194 | |
| 195 | /* |
| 196 | * Now come the defines and routines to manage and access the three-level |
| 197 | * page table. |
| 198 | */ |
| 199 | |
| 200 | /* |
| 201 | * On some architectures, special things need to be done when setting |
| 202 | * the PTE in a page table. Nothing special needs to be on IA-64. |
| 203 | */ |
| 204 | #define set_pte(ptep, pteval) (*(ptep) = (pteval)) |
| 205 | #define set_pte_at(mm,addr,ptep,pteval) set_pte(ptep,pteval) |
| 206 | |
| 207 | #define RGN_SIZE (1UL << 61) |
| 208 | #define RGN_KERNEL 7 |
| 209 | |
| 210 | #define VMALLOC_START 0xa000000200000000UL |
| 211 | #ifdef CONFIG_VIRTUAL_MEM_MAP |
| 212 | # define VMALLOC_END_INIT (0xa000000000000000UL + (1UL << (4*PAGE_SHIFT - 9))) |
| 213 | # define VMALLOC_END vmalloc_end |
| 214 | extern unsigned long vmalloc_end; |
| 215 | #else |
| 216 | # define VMALLOC_END (0xa000000000000000UL + (1UL << (4*PAGE_SHIFT - 9))) |
| 217 | #endif |
| 218 | |
| 219 | /* fs/proc/kcore.c */ |
| 220 | #define kc_vaddr_to_offset(v) ((v) - 0xa000000000000000UL) |
| 221 | #define kc_offset_to_vaddr(o) ((o) + 0xa000000000000000UL) |
| 222 | |
| 223 | /* |
| 224 | * Conversion functions: convert page frame number (pfn) and a protection value to a page |
| 225 | * table entry (pte). |
| 226 | */ |
| 227 | #define pfn_pte(pfn, pgprot) \ |
| 228 | ({ pte_t __pte; pte_val(__pte) = ((pfn) << PAGE_SHIFT) | pgprot_val(pgprot); __pte; }) |
| 229 | |
| 230 | /* Extract pfn from pte. */ |
| 231 | #define pte_pfn(_pte) ((pte_val(_pte) & _PFN_MASK) >> PAGE_SHIFT) |
| 232 | |
| 233 | #define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot)) |
| 234 | |
| 235 | /* This takes a physical page address that is used by the remapping functions */ |
| 236 | #define mk_pte_phys(physpage, pgprot) \ |
| 237 | ({ pte_t __pte; pte_val(__pte) = physpage + pgprot_val(pgprot); __pte; }) |
| 238 | |
| 239 | #define pte_modify(_pte, newprot) \ |
| 240 | (__pte((pte_val(_pte) & ~_PAGE_CHG_MASK) | (pgprot_val(newprot) & _PAGE_CHG_MASK))) |
| 241 | |
| 242 | #define page_pte_prot(page,prot) mk_pte(page, prot) |
| 243 | #define page_pte(page) page_pte_prot(page, __pgprot(0)) |
| 244 | |
| 245 | #define pte_none(pte) (!pte_val(pte)) |
| 246 | #define pte_present(pte) (pte_val(pte) & (_PAGE_P | _PAGE_PROTNONE)) |
| 247 | #define pte_clear(mm,addr,pte) (pte_val(*(pte)) = 0UL) |
| 248 | /* pte_page() returns the "struct page *" corresponding to the PTE: */ |
| 249 | #define pte_page(pte) virt_to_page(((pte_val(pte) & _PFN_MASK) + PAGE_OFFSET)) |
| 250 | |
| 251 | #define pmd_none(pmd) (!pmd_val(pmd)) |
| 252 | #define pmd_bad(pmd) (!ia64_phys_addr_valid(pmd_val(pmd))) |
| 253 | #define pmd_present(pmd) (pmd_val(pmd) != 0UL) |
| 254 | #define pmd_clear(pmdp) (pmd_val(*(pmdp)) = 0UL) |
| 255 | #define pmd_page_kernel(pmd) ((unsigned long) __va(pmd_val(pmd) & _PFN_MASK)) |
| 256 | #define pmd_page(pmd) virt_to_page((pmd_val(pmd) + PAGE_OFFSET)) |
| 257 | |
| 258 | #define pud_none(pud) (!pud_val(pud)) |
| 259 | #define pud_bad(pud) (!ia64_phys_addr_valid(pud_val(pud))) |
| 260 | #define pud_present(pud) (pud_val(pud) != 0UL) |
| 261 | #define pud_clear(pudp) (pud_val(*(pudp)) = 0UL) |
| 262 | |
| 263 | #define pud_page(pud) ((unsigned long) __va(pud_val(pud) & _PFN_MASK)) |
| 264 | |
| 265 | /* |
| 266 | * The following have defined behavior only work if pte_present() is true. |
| 267 | */ |
| 268 | #define pte_user(pte) ((pte_val(pte) & _PAGE_PL_MASK) == _PAGE_PL_3) |
| 269 | #define pte_read(pte) (((pte_val(pte) & _PAGE_AR_MASK) >> _PAGE_AR_SHIFT) < 6) |
| 270 | #define pte_write(pte) ((unsigned) (((pte_val(pte) & _PAGE_AR_MASK) >> _PAGE_AR_SHIFT) - 2) <= 4) |
| 271 | #define pte_exec(pte) ((pte_val(pte) & _PAGE_AR_RX) != 0) |
| 272 | #define pte_dirty(pte) ((pte_val(pte) & _PAGE_D) != 0) |
| 273 | #define pte_young(pte) ((pte_val(pte) & _PAGE_A) != 0) |
| 274 | #define pte_file(pte) ((pte_val(pte) & _PAGE_FILE) != 0) |
| 275 | /* |
| 276 | * Note: we convert AR_RWX to AR_RX and AR_RW to AR_R by clearing the 2nd bit in the |
| 277 | * access rights: |
| 278 | */ |
| 279 | #define pte_wrprotect(pte) (__pte(pte_val(pte) & ~_PAGE_AR_RW)) |
| 280 | #define pte_mkwrite(pte) (__pte(pte_val(pte) | _PAGE_AR_RW)) |
| 281 | #define pte_mkexec(pte) (__pte(pte_val(pte) | _PAGE_AR_RX)) |
| 282 | #define pte_mkold(pte) (__pte(pte_val(pte) & ~_PAGE_A)) |
| 283 | #define pte_mkyoung(pte) (__pte(pte_val(pte) | _PAGE_A)) |
| 284 | #define pte_mkclean(pte) (__pte(pte_val(pte) & ~_PAGE_D)) |
| 285 | #define pte_mkdirty(pte) (__pte(pte_val(pte) | _PAGE_D)) |
| 286 | |
| 287 | /* |
| 288 | * Macro to a page protection value as "uncacheable". Note that "protection" is really a |
| 289 | * misnomer here as the protection value contains the memory attribute bits, dirty bits, |
| 290 | * and various other bits as well. |
| 291 | */ |
| 292 | #define pgprot_noncached(prot) __pgprot((pgprot_val(prot) & ~_PAGE_MA_MASK) | _PAGE_MA_UC) |
| 293 | |
| 294 | /* |
| 295 | * Macro to make mark a page protection value as "write-combining". |
| 296 | * Note that "protection" is really a misnomer here as the protection |
| 297 | * value contains the memory attribute bits, dirty bits, and various |
| 298 | * other bits as well. Accesses through a write-combining translation |
| 299 | * works bypasses the caches, but does allow for consecutive writes to |
| 300 | * be combined into single (but larger) write transactions. |
| 301 | */ |
| 302 | #define pgprot_writecombine(prot) __pgprot((pgprot_val(prot) & ~_PAGE_MA_MASK) | _PAGE_MA_WC) |
| 303 | |
| 304 | static inline unsigned long |
| 305 | pgd_index (unsigned long address) |
| 306 | { |
| 307 | unsigned long region = address >> 61; |
| 308 | unsigned long l1index = (address >> PGDIR_SHIFT) & ((PTRS_PER_PGD >> 3) - 1); |
| 309 | |
| 310 | return (region << (PAGE_SHIFT - 6)) | l1index; |
| 311 | } |
| 312 | |
| 313 | /* The offset in the 1-level directory is given by the 3 region bits |
| 314 | (61..63) and the level-1 bits. */ |
| 315 | static inline pgd_t* |
| 316 | pgd_offset (struct mm_struct *mm, unsigned long address) |
| 317 | { |
| 318 | return mm->pgd + pgd_index(address); |
| 319 | } |
| 320 | |
| 321 | /* In the kernel's mapped region we completely ignore the region number |
| 322 | (since we know it's in region number 5). */ |
| 323 | #define pgd_offset_k(addr) \ |
| 324 | (init_mm.pgd + (((addr) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))) |
| 325 | |
| 326 | /* Look up a pgd entry in the gate area. On IA-64, the gate-area |
| 327 | resides in the kernel-mapped segment, hence we use pgd_offset_k() |
| 328 | here. */ |
| 329 | #define pgd_offset_gate(mm, addr) pgd_offset_k(addr) |
| 330 | |
| 331 | /* Find an entry in the second-level page table.. */ |
| 332 | #define pmd_offset(dir,addr) \ |
| 333 | ((pmd_t *) pud_page(*(dir)) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1))) |
| 334 | |
| 335 | /* |
| 336 | * Find an entry in the third-level page table. This looks more complicated than it |
| 337 | * should be because some platforms place page tables in high memory. |
| 338 | */ |
| 339 | #define pte_index(addr) (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) |
| 340 | #define pte_offset_kernel(dir,addr) ((pte_t *) pmd_page_kernel(*(dir)) + pte_index(addr)) |
| 341 | #define pte_offset_map(dir,addr) pte_offset_kernel(dir, addr) |
| 342 | #define pte_offset_map_nested(dir,addr) pte_offset_map(dir, addr) |
| 343 | #define pte_unmap(pte) do { } while (0) |
| 344 | #define pte_unmap_nested(pte) do { } while (0) |
| 345 | |
| 346 | /* atomic versions of the some PTE manipulations: */ |
| 347 | |
| 348 | static inline int |
| 349 | ptep_test_and_clear_young (struct vm_area_struct *vma, unsigned long addr, pte_t *ptep) |
| 350 | { |
| 351 | #ifdef CONFIG_SMP |
| 352 | if (!pte_young(*ptep)) |
| 353 | return 0; |
| 354 | return test_and_clear_bit(_PAGE_A_BIT, ptep); |
| 355 | #else |
| 356 | pte_t pte = *ptep; |
| 357 | if (!pte_young(pte)) |
| 358 | return 0; |
| 359 | set_pte_at(vma->vm_mm, addr, ptep, pte_mkold(pte)); |
| 360 | return 1; |
| 361 | #endif |
| 362 | } |
| 363 | |
| 364 | static inline int |
| 365 | ptep_test_and_clear_dirty (struct vm_area_struct *vma, unsigned long addr, pte_t *ptep) |
| 366 | { |
| 367 | #ifdef CONFIG_SMP |
| 368 | if (!pte_dirty(*ptep)) |
| 369 | return 0; |
| 370 | return test_and_clear_bit(_PAGE_D_BIT, ptep); |
| 371 | #else |
| 372 | pte_t pte = *ptep; |
| 373 | if (!pte_dirty(pte)) |
| 374 | return 0; |
| 375 | set_pte_at(vma->vm_mm, addr, ptep, pte_mkclean(pte)); |
| 376 | return 1; |
| 377 | #endif |
| 378 | } |
| 379 | |
| 380 | static inline pte_t |
| 381 | ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep) |
| 382 | { |
| 383 | #ifdef CONFIG_SMP |
| 384 | return __pte(xchg((long *) ptep, 0)); |
| 385 | #else |
| 386 | pte_t pte = *ptep; |
| 387 | pte_clear(mm, addr, ptep); |
| 388 | return pte; |
| 389 | #endif |
| 390 | } |
| 391 | |
| 392 | static inline void |
| 393 | ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep) |
| 394 | { |
| 395 | #ifdef CONFIG_SMP |
| 396 | unsigned long new, old; |
| 397 | |
| 398 | do { |
| 399 | old = pte_val(*ptep); |
| 400 | new = pte_val(pte_wrprotect(__pte (old))); |
| 401 | } while (cmpxchg((unsigned long *) ptep, old, new) != old); |
| 402 | #else |
| 403 | pte_t old_pte = *ptep; |
| 404 | set_pte_at(mm, addr, ptep, pte_wrprotect(old_pte)); |
| 405 | #endif |
| 406 | } |
| 407 | |
| 408 | static inline int |
| 409 | pte_same (pte_t a, pte_t b) |
| 410 | { |
| 411 | return pte_val(a) == pte_val(b); |
| 412 | } |
| 413 | |
| 414 | #define update_mmu_cache(vma, address, pte) do { } while (0) |
| 415 | |
| 416 | extern pgd_t swapper_pg_dir[PTRS_PER_PGD]; |
| 417 | extern void paging_init (void); |
| 418 | |
| 419 | /* |
| 420 | * Note: The macros below rely on the fact that MAX_SWAPFILES_SHIFT <= number of |
| 421 | * bits in the swap-type field of the swap pte. It would be nice to |
| 422 | * enforce that, but we can't easily include <linux/swap.h> here. |
| 423 | * (Of course, better still would be to define MAX_SWAPFILES_SHIFT here...). |
| 424 | * |
| 425 | * Format of swap pte: |
| 426 | * bit 0 : present bit (must be zero) |
| 427 | * bit 1 : _PAGE_FILE (must be zero) |
| 428 | * bits 2- 8: swap-type |
| 429 | * bits 9-62: swap offset |
| 430 | * bit 63 : _PAGE_PROTNONE bit |
| 431 | * |
| 432 | * Format of file pte: |
| 433 | * bit 0 : present bit (must be zero) |
| 434 | * bit 1 : _PAGE_FILE (must be one) |
| 435 | * bits 2-62: file_offset/PAGE_SIZE |
| 436 | * bit 63 : _PAGE_PROTNONE bit |
| 437 | */ |
| 438 | #define __swp_type(entry) (((entry).val >> 2) & 0x7f) |
| 439 | #define __swp_offset(entry) (((entry).val << 1) >> 10) |
| 440 | #define __swp_entry(type,offset) ((swp_entry_t) { ((type) << 2) | ((long) (offset) << 9) }) |
| 441 | #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) }) |
| 442 | #define __swp_entry_to_pte(x) ((pte_t) { (x).val }) |
| 443 | |
| 444 | #define PTE_FILE_MAX_BITS 61 |
| 445 | #define pte_to_pgoff(pte) ((pte_val(pte) << 1) >> 3) |
| 446 | #define pgoff_to_pte(off) ((pte_t) { ((off) << 2) | _PAGE_FILE }) |
| 447 | |
| 448 | /* XXX is this right? */ |
| 449 | #define io_remap_page_range(vma, vaddr, paddr, size, prot) \ |
| 450 | remap_pfn_range(vma, vaddr, (paddr) >> PAGE_SHIFT, size, prot) |
| 451 | |
| 452 | #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \ |
| 453 | remap_pfn_range(vma, vaddr, pfn, size, prot) |
| 454 | |
| 455 | #define MK_IOSPACE_PFN(space, pfn) (pfn) |
| 456 | #define GET_IOSPACE(pfn) 0 |
| 457 | #define GET_PFN(pfn) (pfn) |
| 458 | |
| 459 | /* |
| 460 | * ZERO_PAGE is a global shared page that is always zero: used |
| 461 | * for zero-mapped memory areas etc.. |
| 462 | */ |
| 463 | extern unsigned long empty_zero_page[PAGE_SIZE/sizeof(unsigned long)]; |
| 464 | extern struct page *zero_page_memmap_ptr; |
| 465 | #define ZERO_PAGE(vaddr) (zero_page_memmap_ptr) |
| 466 | |
| 467 | /* We provide our own get_unmapped_area to cope with VA holes for userland */ |
| 468 | #define HAVE_ARCH_UNMAPPED_AREA |
| 469 | |
| 470 | #ifdef CONFIG_HUGETLB_PAGE |
| 471 | #define HUGETLB_PGDIR_SHIFT (HPAGE_SHIFT + 2*(PAGE_SHIFT-3)) |
| 472 | #define HUGETLB_PGDIR_SIZE (__IA64_UL(1) << HUGETLB_PGDIR_SHIFT) |
| 473 | #define HUGETLB_PGDIR_MASK (~(HUGETLB_PGDIR_SIZE-1)) |
| 474 | struct mmu_gather; |
Hugh Dickins | 3bf5ee9 | 2005-04-19 13:29:16 -0700 | [diff] [blame] | 475 | void hugetlb_free_pgd_range(struct mmu_gather **tlb, unsigned long addr, |
| 476 | unsigned long end, unsigned long floor, unsigned long ceiling); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 477 | #endif |
| 478 | |
| 479 | /* |
| 480 | * IA-64 doesn't have any external MMU info: the page tables contain all the necessary |
| 481 | * information. However, we use this routine to take care of any (delayed) i-cache |
| 482 | * flushing that may be necessary. |
| 483 | */ |
| 484 | extern void lazy_mmu_prot_update (pte_t pte); |
| 485 | |
| 486 | #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS |
| 487 | /* |
| 488 | * Update PTEP with ENTRY, which is guaranteed to be a less |
| 489 | * restrictive PTE. That is, ENTRY may have the ACCESSED, DIRTY, and |
| 490 | * WRITABLE bits turned on, when the value at PTEP did not. The |
| 491 | * WRITABLE bit may only be turned if SAFELY_WRITABLE is TRUE. |
| 492 | * |
| 493 | * SAFELY_WRITABLE is TRUE if we can update the value at PTEP without |
| 494 | * having to worry about races. On SMP machines, there are only two |
| 495 | * cases where this is true: |
| 496 | * |
| 497 | * (1) *PTEP has the PRESENT bit turned OFF |
| 498 | * (2) ENTRY has the DIRTY bit turned ON |
| 499 | * |
| 500 | * On ia64, we could implement this routine with a cmpxchg()-loop |
| 501 | * which ORs in the _PAGE_A/_PAGE_D bit if they're set in ENTRY. |
| 502 | * However, like on x86, we can get a more streamlined version by |
| 503 | * observing that it is OK to drop ACCESSED bit updates when |
| 504 | * SAFELY_WRITABLE is FALSE. Besides being rare, all that would do is |
| 505 | * result in an extra Access-bit fault, which would then turn on the |
| 506 | * ACCESSED bit in the low-level fault handler (iaccess_bit or |
| 507 | * daccess_bit in ivt.S). |
| 508 | */ |
| 509 | #ifdef CONFIG_SMP |
| 510 | # define ptep_set_access_flags(__vma, __addr, __ptep, __entry, __safely_writable) \ |
| 511 | do { \ |
| 512 | if (__safely_writable) { \ |
| 513 | set_pte(__ptep, __entry); \ |
| 514 | flush_tlb_page(__vma, __addr); \ |
| 515 | } \ |
| 516 | } while (0) |
| 517 | #else |
| 518 | # define ptep_set_access_flags(__vma, __addr, __ptep, __entry, __safely_writable) \ |
| 519 | ptep_establish(__vma, __addr, __ptep, __entry) |
| 520 | #endif |
| 521 | |
| 522 | # ifdef CONFIG_VIRTUAL_MEM_MAP |
| 523 | /* arch mem_map init routine is needed due to holes in a virtual mem_map */ |
| 524 | # define __HAVE_ARCH_MEMMAP_INIT |
| 525 | extern void memmap_init (unsigned long size, int nid, unsigned long zone, |
| 526 | unsigned long start_pfn); |
| 527 | # endif /* CONFIG_VIRTUAL_MEM_MAP */ |
| 528 | # endif /* !__ASSEMBLY__ */ |
| 529 | |
| 530 | /* |
| 531 | * Identity-mapped regions use a large page size. We'll call such large pages |
| 532 | * "granules". If you can think of a better name that's unambiguous, let me |
| 533 | * know... |
| 534 | */ |
| 535 | #if defined(CONFIG_IA64_GRANULE_64MB) |
| 536 | # define IA64_GRANULE_SHIFT _PAGE_SIZE_64M |
| 537 | #elif defined(CONFIG_IA64_GRANULE_16MB) |
| 538 | # define IA64_GRANULE_SHIFT _PAGE_SIZE_16M |
| 539 | #endif |
| 540 | #define IA64_GRANULE_SIZE (1 << IA64_GRANULE_SHIFT) |
| 541 | /* |
| 542 | * log2() of the page size we use to map the kernel image (IA64_TR_KERNEL): |
| 543 | */ |
| 544 | #define KERNEL_TR_PAGE_SHIFT _PAGE_SIZE_64M |
| 545 | #define KERNEL_TR_PAGE_SIZE (1 << KERNEL_TR_PAGE_SHIFT) |
| 546 | |
| 547 | /* |
| 548 | * No page table caches to initialise |
| 549 | */ |
| 550 | #define pgtable_cache_init() do { } while (0) |
| 551 | |
| 552 | /* These tell get_user_pages() that the first gate page is accessible from user-level. */ |
| 553 | #define FIXADDR_USER_START GATE_ADDR |
David Mosberger-Tang | ad597bd | 2005-06-08 10:45:00 -0700 | [diff] [blame] | 554 | #ifdef HAVE_BUGGY_SEGREL |
| 555 | # define FIXADDR_USER_END (GATE_ADDR + 2*PAGE_SIZE) |
| 556 | #else |
| 557 | # define FIXADDR_USER_END (GATE_ADDR + 2*PERCPU_PAGE_SIZE) |
| 558 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 559 | |
| 560 | #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG |
| 561 | #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY |
| 562 | #define __HAVE_ARCH_PTEP_GET_AND_CLEAR |
| 563 | #define __HAVE_ARCH_PTEP_SET_WRPROTECT |
| 564 | #define __HAVE_ARCH_PTE_SAME |
| 565 | #define __HAVE_ARCH_PGD_OFFSET_GATE |
| 566 | #define __HAVE_ARCH_LAZY_MMU_PROT_UPDATE |
| 567 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 568 | #include <asm-generic/pgtable-nopud.h> |
| 569 | #include <asm-generic/pgtable.h> |
| 570 | |
| 571 | #endif /* _ASM_IA64_PGTABLE_H */ |