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 16extern pud_t level3_kernel_pgt[512]; 17extern pud_t level3_ident_pgt[512]; 18extern pmd_t level2_kernel_pgt[512]; 19extern pgd_t init_level4_pgt[]; 20extern unsigned long __supported_pte_mask; 21 22#define swapper_pg_dir init_level4_pgt 23 24extern void paging_init(void); 25extern 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 */ 31extern 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 74static 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 80static inline void set_pmd(pmd_t *dst, pmd_t val) 81{ 82 pmd_val(*dst) = pmd_val(val); 83} 84 85static inline void set_pud(pud_t *dst, pud_t val) 86{ 87 pud_val(*dst) = pud_val(val); 88} 89 90static inline void pud_clear (pud_t *pud) 91{ 92 set_pud(pud, __pud(0)); 93} 94 95static inline void set_pgd(pgd_t *dst, pgd_t val) 96{ 97 pgd_val(*dst) = pgd_val(val); 98} 99 100static 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 107struct mm_struct; 108 109static 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 231static inline unsigned long pgd_bad(pgd_t pgd) 232{ 233 return pgd_val(pgd) & ~(PTE_MASK | _KERNPG_TABLE | _PAGE_USER); 234} 235 236static inline unsigned long pud_bad(pud_t pud) 237{ 238 return pud_val(pud) & ~(PTE_MASK | _KERNPG_TABLE | _PAGE_USER); 239} 240 241static 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)) 251#define pte_page(x) pfn_to_page(pte_pfn(x)) 252#define pte_pfn(x) ((pte_val(x) & __PHYSICAL_MASK) >> PAGE_SHIFT) 253 254static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot) 255{ 256 pte_t pte; 257 pte_val(pte) = (page_nr << PAGE_SHIFT); 258 pte_val(pte) |= pgprot_val(pgprot); 259 pte_val(pte) &= __supported_pte_mask; 260 return pte; 261} 262 263/* 264 * The following only work if pte_present() is true. 265 * Undefined behaviour if not.. 266 */ 267#define __LARGE_PTE (_PAGE_PSE|_PAGE_PRESENT) 268static inline int pte_user(pte_t pte) { return pte_val(pte) & _PAGE_USER; } 269static inline int pte_read(pte_t pte) { return pte_val(pte) & _PAGE_USER; } 270static inline int pte_exec(pte_t pte) { return !(pte_val(pte) & _PAGE_NX); } 271static inline int pte_dirty(pte_t pte) { return pte_val(pte) & _PAGE_DIRTY; } 272static inline int pte_young(pte_t pte) { return pte_val(pte) & _PAGE_ACCESSED; } 273static inline int pte_write(pte_t pte) { return pte_val(pte) & _PAGE_RW; } 274static inline int pte_file(pte_t pte) { return pte_val(pte) & _PAGE_FILE; } 275static inline int pte_huge(pte_t pte) { return pte_val(pte) & _PAGE_PSE; } 276 277static inline pte_t pte_rdprotect(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_USER)); return pte; } 278static inline pte_t pte_exprotect(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_USER)); return pte; } 279static inline pte_t pte_mkclean(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_DIRTY)); return pte; } 280static inline pte_t pte_mkold(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_ACCESSED)); return pte; } 281static inline pte_t pte_wrprotect(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_RW)); return pte; } 282static inline pte_t pte_mkread(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) | _PAGE_USER)); return pte; } 283static inline pte_t pte_mkexec(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_NX)); return pte; } 284static inline pte_t pte_mkdirty(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) | _PAGE_DIRTY)); return pte; } 285static inline pte_t pte_mkyoung(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) | _PAGE_ACCESSED)); return pte; } 286static inline pte_t pte_mkwrite(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) | _PAGE_RW)); return pte; } 287static inline pte_t pte_mkhuge(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) | _PAGE_PSE)); return pte; } 288static inline pte_t pte_clrhuge(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_PSE)); return pte; } 289 290struct vm_area_struct; 291 292static inline int ptep_test_and_clear_dirty(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep) 293{ 294 if (!pte_dirty(*ptep)) 295 return 0; 296 return test_and_clear_bit(_PAGE_BIT_DIRTY, &ptep->pte); 297} 298 299static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep) 300{ 301 if (!pte_young(*ptep)) 302 return 0; 303 return test_and_clear_bit(_PAGE_BIT_ACCESSED, &ptep->pte); 304} 305 306static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep) 307{ 308 clear_bit(_PAGE_BIT_RW, &ptep->pte); 309} 310 311/* 312 * Macro to mark a page protection value as "uncacheable". 313 */ 314#define pgprot_noncached(prot) (__pgprot(pgprot_val(prot) | _PAGE_PCD | _PAGE_PWT)) 315 316static inline int pmd_large(pmd_t pte) { 317 return (pmd_val(pte) & __LARGE_PTE) == __LARGE_PTE; 318} 319 320 321/* 322 * Conversion functions: convert a page and protection to a page entry, 323 * and a page entry and page directory to the page they refer to. 324 */ 325 326/* 327 * Level 4 access. 328 */ 329#define pgd_page_vaddr(pgd) ((unsigned long) __va((unsigned long)pgd_val(pgd) & PTE_MASK)) 330#define pgd_page(pgd) (pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)) 331#define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1)) 332#define pgd_offset(mm, addr) ((mm)->pgd + pgd_index(addr)) 333#define pgd_offset_k(address) (init_level4_pgt + pgd_index(address)) 334#define pgd_present(pgd) (pgd_val(pgd) & _PAGE_PRESENT) 335#define mk_kernel_pgd(address) ((pgd_t){ (address) | _KERNPG_TABLE }) 336 337/* PUD - Level3 access */ 338/* to find an entry in a page-table-directory. */ 339#define pud_page_vaddr(pud) ((unsigned long) __va(pud_val(pud) & PHYSICAL_PAGE_MASK)) 340#define pud_page(pud) (pfn_to_page(pud_val(pud) >> PAGE_SHIFT)) 341#define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1)) 342#define pud_offset(pgd, address) ((pud_t *) pgd_page_vaddr(*(pgd)) + pud_index(address)) 343#define pud_present(pud) (pud_val(pud) & _PAGE_PRESENT) 344 345/* PMD - Level 2 access */ 346#define pmd_page_vaddr(pmd) ((unsigned long) __va(pmd_val(pmd) & PTE_MASK)) 347#define pmd_page(pmd) (pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)) 348 349#define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1)) 350#define pmd_offset(dir, address) ((pmd_t *) pud_page_vaddr(*(dir)) + \ 351 pmd_index(address)) 352#define pmd_none(x) (!pmd_val(x)) 353#define pmd_present(x) (pmd_val(x) & _PAGE_PRESENT) 354#define pmd_clear(xp) do { set_pmd(xp, __pmd(0)); } while (0) 355#define pfn_pmd(nr,prot) (__pmd(((nr) << PAGE_SHIFT) | pgprot_val(prot))) 356#define pmd_pfn(x) ((pmd_val(x) & __PHYSICAL_MASK) >> PAGE_SHIFT) 357 358#define pte_to_pgoff(pte) ((pte_val(pte) & PHYSICAL_PAGE_MASK) >> PAGE_SHIFT) 359#define pgoff_to_pte(off) ((pte_t) { ((off) << PAGE_SHIFT) | _PAGE_FILE }) 360#define PTE_FILE_MAX_BITS __PHYSICAL_MASK_SHIFT 361 362/* PTE - Level 1 access. */ 363 364/* page, protection -> pte */ 365#define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot)) 366#define mk_pte_huge(entry) (pte_val(entry) |= _PAGE_PRESENT | _PAGE_PSE) 367 368/* Change flags of a PTE */ 369static inline pte_t pte_modify(pte_t pte, pgprot_t newprot) 370{ 371 pte_val(pte) &= _PAGE_CHG_MASK; 372 pte_val(pte) |= pgprot_val(newprot); 373 pte_val(pte) &= __supported_pte_mask; 374 return pte; 375} 376 377#define pte_index(address) \ 378 (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) 379#define pte_offset_kernel(dir, address) ((pte_t *) pmd_page_vaddr(*(dir)) + \ 380 pte_index(address)) 381 382/* x86-64 always has all page tables mapped. */ 383#define pte_offset_map(dir,address) pte_offset_kernel(dir,address) 384#define pte_offset_map_nested(dir,address) pte_offset_kernel(dir,address) 385#define pte_unmap(pte) /* NOP */ 386#define pte_unmap_nested(pte) /* NOP */ 387 388#define update_mmu_cache(vma,address,pte) do { } while (0) 389 390/* We only update the dirty/accessed state if we set 391 * the dirty bit by hand in the kernel, since the hardware 392 * will do the accessed bit for us, and we don't want to 393 * race with other CPU's that might be updating the dirty 394 * bit at the same time. */ 395#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS 396#define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \ 397({ \ 398 int __changed = !pte_same(*(__ptep), __entry); \ 399 if (__changed && __dirty) { \ 400 set_pte(__ptep, __entry); \ 401 flush_tlb_page(__vma, __address); \ 402 } \ 403 __changed; \ 404}) 405 406/* Encode and de-code a swap entry */ 407#define __swp_type(x) (((x).val >> 1) & 0x3f) 408#define __swp_offset(x) ((x).val >> 8) 409#define __swp_entry(type, offset) ((swp_entry_t) { ((type) << 1) | ((offset) << 8) }) 410#define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) }) 411#define __swp_entry_to_pte(x) ((pte_t) { (x).val }) 412 413extern spinlock_t pgd_lock; 414extern struct list_head pgd_list; 415 416extern int kern_addr_valid(unsigned long addr); 417 418#define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \ 419 remap_pfn_range(vma, vaddr, pfn, size, prot) 420 421#define HAVE_ARCH_UNMAPPED_AREA 422 423#define pgtable_cache_init() do { } while (0) 424#define check_pgt_cache() do { } while (0) 425 426#define PAGE_AGP PAGE_KERNEL_NOCACHE 427#define HAVE_PAGE_AGP 1 428 429/* fs/proc/kcore.c */ 430#define kc_vaddr_to_offset(v) ((v) & __VIRTUAL_MASK) 431#define kc_offset_to_vaddr(o) \ 432 (((o) & (1UL << (__VIRTUAL_MASK_SHIFT-1))) ? ((o) | (~__VIRTUAL_MASK)) : (o)) 433 434#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG 435#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY 436#define __HAVE_ARCH_PTEP_GET_AND_CLEAR 437#define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL 438#define __HAVE_ARCH_PTEP_SET_WRPROTECT 439#define __HAVE_ARCH_PTE_SAME 440#include <asm-generic/pgtable.h> 441#endif /* !__ASSEMBLY__ */ 442 443#endif /* _X86_64_PGTABLE_H */ 444