1 #ifndef _X86_64_PGTABLE_H
2 #define _X86_64_PGTABLE_H
8 * This file contains the functions and defines necessary to modify and use
9 * the x86-64 page table tree.
11 #include <lwk/spinlock.h>
12 #include <arch/fixmap.h>
13 #include <arch/processor.h>
14 #include <arch/bitops.h>
17 extern pud_t level3_kernel_pgt[512];
18 extern pud_t level3_ident_pgt[512];
19 extern pmd_t level2_kernel_pgt[512];
20 extern pgd_t init_level4_pgt[];
21 extern unsigned long __supported_pte_mask;
23 #define swapper_pg_dir init_level4_pgt
25 extern void paging_init(void);
26 extern void clear_kernel_mapping(unsigned long addr, unsigned long size);
29 * ZERO_PAGE is a global shared page that is always zero: used
30 * for zero-mapped memory areas etc..
32 extern unsigned long empty_zero_page[PAGE_SIZE/sizeof(unsigned long)];
33 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
35 #endif /* !__ASSEMBLY__ */
38 * PGDIR_SHIFT determines what a top-level page table entry can map
40 #define PGDIR_SHIFT 39
41 #define PTRS_PER_PGD 512
47 #define PTRS_PER_PUD 512
50 * PMD_SHIFT determines the size of the area a middle-level
54 #define PTRS_PER_PMD 512
57 * entries per page directory level
59 #define PTRS_PER_PTE 512
63 #define pte_ERROR(e) \
64 printk("%s:%d: bad pte %p(%016lx).\n", __FILE__, __LINE__, &(e), pte_val(e))
65 #define pmd_ERROR(e) \
66 printk("%s:%d: bad pmd %p(%016lx).\n", __FILE__, __LINE__, &(e), pmd_val(e))
67 #define pud_ERROR(e) \
68 printk("%s:%d: bad pud %p(%016lx).\n", __FILE__, __LINE__, &(e), pud_val(e))
69 #define pgd_ERROR(e) \
70 printk("%s:%d: bad pgd %p(%016lx).\n", __FILE__, __LINE__, &(e), pgd_val(e))
72 #define pgd_none(x) (!pgd_val(x))
73 #define pud_none(x) (!pud_val(x))
75 static inline void set_pte(pte_t *dst, pte_t val)
77 pte_val(*dst) = pte_val(val);
79 #define set_pte_at(mm,addr,ptep,pteval) set_pte(ptep,pteval)
81 static inline void set_pmd(pmd_t *dst, pmd_t val)
83 pmd_val(*dst) = pmd_val(val);
86 static inline void set_pud(pud_t *dst, pud_t val)
88 pud_val(*dst) = pud_val(val);
91 static inline void pud_clear (pud_t *pud)
93 set_pud(pud, __pud(0));
96 static inline void set_pgd(pgd_t *dst, pgd_t val)
98 pgd_val(*dst) = pgd_val(val);
101 static inline void pgd_clear (pgd_t * pgd)
103 set_pgd(pgd, __pgd(0));
106 #define ptep_get_and_clear(mm,addr,xp) __pte(xchg(&(xp)->pte, 0))
110 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm, unsigned long addr, pte_t *ptep, int full)
117 pte = ptep_get_and_clear(mm, addr, ptep);
122 #define pte_same(a, b) ((a).pte == (b).pte)
124 #define pte_pgprot(a) (__pgprot((a).pte & ~PHYSICAL_PAGE_MASK))
126 #endif /* !__ASSEMBLY__ */
128 #define PMD_SIZE (_AC(1,UL) << PMD_SHIFT)
129 #define PMD_MASK (~(PMD_SIZE-1))
130 #define PUD_SIZE (_AC(1,UL) << PUD_SHIFT)
131 #define PUD_MASK (~(PUD_SIZE-1))
132 #define PGDIR_SIZE (_AC(1,UL) << PGDIR_SHIFT)
133 #define PGDIR_MASK (~(PGDIR_SIZE-1))
135 #define USER_PTRS_PER_PGD ((TASK_SIZE-1)/PGDIR_SIZE+1)
136 #define FIRST_USER_ADDRESS 0
138 #define MAXMEM _AC(0x3fffffffffff, UL)
139 #define VMALLOC_START _AC(0xffffc20000000000, UL)
140 #define VMALLOC_END _AC(0xffffe1ffffffffff, UL)
141 #define MODULES_VADDR _AC(0xffffffff88000000, UL)
142 #define MODULES_END _AC(0xfffffffffff00000, UL)
143 #define MODULES_LEN (MODULES_END - MODULES_VADDR)
145 #define _PAGE_BIT_PRESENT 0
146 #define _PAGE_BIT_RW 1
147 #define _PAGE_BIT_USER 2
148 #define _PAGE_BIT_PWT 3
149 #define _PAGE_BIT_PCD 4
150 #define _PAGE_BIT_ACCESSED 5
151 #define _PAGE_BIT_DIRTY 6
152 #define _PAGE_BIT_PSE 7 /* 4 MB (or 2MB) page */
153 #define _PAGE_BIT_GLOBAL 8 /* Global TLB entry PPro+ */
154 #define _PAGE_BIT_NX 63 /* No execute: only valid after cpuid check */
156 #define _PAGE_PRESENT 0x001
157 #define _PAGE_RW 0x002
158 #define _PAGE_USER 0x004
159 #define _PAGE_PWT 0x008
160 #define _PAGE_PCD 0x010
161 #define _PAGE_ACCESSED 0x020
162 #define _PAGE_DIRTY 0x040
163 #define _PAGE_PSE 0x080 /* 2MB page */
164 #define _PAGE_FILE 0x040 /* nonlinear file mapping, saved PTE; unset:swap */
165 #define _PAGE_GLOBAL 0x100 /* Global TLB entry */
167 #define _PAGE_PROTNONE 0x080 /* If not present */
168 #define _PAGE_NX (_AC(1,UL)<<_PAGE_BIT_NX)
170 #define _PAGE_TABLE (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED | _PAGE_DIRTY)
171 #define _KERNPG_TABLE (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY)
173 #define _PAGE_CHG_MASK (PTE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY)
175 #define PAGE_NONE __pgprot(_PAGE_PROTNONE | _PAGE_ACCESSED)
176 #define PAGE_SHARED __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
177 #define PAGE_SHARED_EXEC __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED)
178 #define PAGE_COPY_NOEXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
179 #define PAGE_COPY PAGE_COPY_NOEXEC
180 #define PAGE_COPY_EXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
181 #define PAGE_READONLY __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
182 #define PAGE_READONLY_EXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
183 #define __PAGE_KERNEL \
184 (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_NX)
185 #define __PAGE_KERNEL_EXEC \
186 (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED)
187 #define __PAGE_KERNEL_NOCACHE \
188 (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_PCD | _PAGE_ACCESSED | _PAGE_NX)
189 #define __PAGE_KERNEL_RO \
190 (_PAGE_PRESENT | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_NX)
191 #define __PAGE_KERNEL_VSYSCALL \
192 (_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
193 #define __PAGE_KERNEL_VSYSCALL_NOCACHE \
194 (_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_PCD)
195 #define __PAGE_KERNEL_LARGE \
196 (__PAGE_KERNEL | _PAGE_PSE)
197 #define __PAGE_KERNEL_LARGE_EXEC \
198 (__PAGE_KERNEL_EXEC | _PAGE_PSE)
200 #define MAKE_GLOBAL(x) __pgprot((x) | _PAGE_GLOBAL)
202 #define PAGE_KERNEL MAKE_GLOBAL(__PAGE_KERNEL)
203 #define PAGE_KERNEL_EXEC MAKE_GLOBAL(__PAGE_KERNEL_EXEC)
204 #define PAGE_KERNEL_RO MAKE_GLOBAL(__PAGE_KERNEL_RO)
205 #define PAGE_KERNEL_NOCACHE MAKE_GLOBAL(__PAGE_KERNEL_NOCACHE)
206 #define PAGE_KERNEL_VSYSCALL32 __pgprot(__PAGE_KERNEL_VSYSCALL)
207 #define PAGE_KERNEL_VSYSCALL MAKE_GLOBAL(__PAGE_KERNEL_VSYSCALL)
208 #define PAGE_KERNEL_LARGE MAKE_GLOBAL(__PAGE_KERNEL_LARGE)
209 #define PAGE_KERNEL_VSYSCALL_NOCACHE MAKE_GLOBAL(__PAGE_KERNEL_VSYSCALL_NOCACHE)
212 #define __P000 PAGE_NONE
213 #define __P001 PAGE_READONLY
214 #define __P010 PAGE_COPY
215 #define __P011 PAGE_COPY
216 #define __P100 PAGE_READONLY_EXEC
217 #define __P101 PAGE_READONLY_EXEC
218 #define __P110 PAGE_COPY_EXEC
219 #define __P111 PAGE_COPY_EXEC
221 #define __S000 PAGE_NONE
222 #define __S001 PAGE_READONLY
223 #define __S010 PAGE_SHARED
224 #define __S011 PAGE_SHARED
225 #define __S100 PAGE_READONLY_EXEC
226 #define __S101 PAGE_READONLY_EXEC
227 #define __S110 PAGE_SHARED_EXEC
228 #define __S111 PAGE_SHARED_EXEC
232 static inline unsigned long pgd_bad(pgd_t pgd)
234 return pgd_val(pgd) & ~(PTE_MASK | _KERNPG_TABLE | _PAGE_USER);
237 static inline unsigned long pud_bad(pud_t pud)
239 return pud_val(pud) & ~(PTE_MASK | _KERNPG_TABLE | _PAGE_USER);
242 static inline unsigned long pmd_bad(pmd_t pmd)
244 return pmd_val(pmd) & ~(PTE_MASK | _KERNPG_TABLE | _PAGE_USER);
247 #define pte_none(x) (!pte_val(x))
248 #define pte_present(x) (pte_val(x) & (_PAGE_PRESENT | _PAGE_PROTNONE))
249 #define pte_clear(mm,addr,xp) do { set_pte_at(mm, addr, xp, __pte(0)); } while (0)
251 #define pages_to_mb(x) ((x) >> (20-PAGE_SHIFT)) /* FIXME: is this
253 #define pte_page(x) pfn_to_page(pte_pfn(x))
254 #define pte_pfn(x) ((pte_val(x) & __PHYSICAL_MASK) >> PAGE_SHIFT)
256 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
259 pte_val(pte) = (page_nr << PAGE_SHIFT);
260 pte_val(pte) |= pgprot_val(pgprot);
261 pte_val(pte) &= __supported_pte_mask;
266 * The following only work if pte_present() is true.
267 * Undefined behaviour if not..
269 #define __LARGE_PTE (_PAGE_PSE|_PAGE_PRESENT)
270 static inline int pte_user(pte_t pte) { return pte_val(pte) & _PAGE_USER; }
271 static inline int pte_read(pte_t pte) { return pte_val(pte) & _PAGE_USER; }
272 static inline int pte_exec(pte_t pte) { return !(pte_val(pte) & _PAGE_NX); }
273 static inline int pte_dirty(pte_t pte) { return pte_val(pte) & _PAGE_DIRTY; }
274 static inline int pte_young(pte_t pte) { return pte_val(pte) & _PAGE_ACCESSED; }
275 static inline int pte_write(pte_t pte) { return pte_val(pte) & _PAGE_RW; }
276 static inline int pte_file(pte_t pte) { return pte_val(pte) & _PAGE_FILE; }
277 static inline int pte_huge(pte_t pte) { return pte_val(pte) & _PAGE_PSE; }
279 static inline pte_t pte_rdprotect(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_USER)); return pte; }
280 static inline pte_t pte_exprotect(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_USER)); return pte; }
281 static inline pte_t pte_mkclean(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_DIRTY)); return pte; }
282 static inline pte_t pte_mkold(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_ACCESSED)); return pte; }
283 static inline pte_t pte_wrprotect(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_RW)); return pte; }
284 static inline pte_t pte_mkread(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) | _PAGE_USER)); return pte; }
285 static inline pte_t pte_mkexec(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_NX)); return pte; }
286 static inline pte_t pte_mkdirty(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) | _PAGE_DIRTY)); return pte; }
287 static inline pte_t pte_mkyoung(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) | _PAGE_ACCESSED)); return pte; }
288 static inline pte_t pte_mkwrite(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) | _PAGE_RW)); return pte; }
289 static inline pte_t pte_mkhuge(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) | _PAGE_PSE)); return pte; }
290 static inline pte_t pte_clrhuge(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_PSE)); return pte; }
292 struct vm_area_struct;
294 static inline int ptep_test_and_clear_dirty(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
296 if (!pte_dirty(*ptep))
298 return test_and_clear_bit(_PAGE_BIT_DIRTY, &ptep->pte);
301 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
303 if (!pte_young(*ptep))
305 return test_and_clear_bit(_PAGE_BIT_ACCESSED, &ptep->pte);
308 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
310 clear_bit(_PAGE_BIT_RW, &ptep->pte);
314 * Macro to mark a page protection value as "uncacheable".
316 #define pgprot_noncached(prot) (__pgprot(pgprot_val(prot) | _PAGE_PCD | _PAGE_PWT))
318 static inline int pmd_large(pmd_t pte) {
319 return (pmd_val(pte) & __LARGE_PTE) == __LARGE_PTE;
324 * Conversion functions: convert a page and protection to a page entry,
325 * and a page entry and page directory to the page they refer to.
331 #define pgd_page_vaddr(pgd) ((unsigned long) __va((unsigned long)pgd_val(pgd) & PTE_MASK))
332 #define pgd_page(pgd) (pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT))
333 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
334 #define pgd_offset(mm, addr) ((mm)->pgd + pgd_index(addr))
335 #define pgd_offset_k(address) (init_level4_pgt + pgd_index(address))
336 #define pgd_present(pgd) (pgd_val(pgd) & _PAGE_PRESENT)
337 #define mk_kernel_pgd(address) ((pgd_t){ (address) | _KERNPG_TABLE })
339 /* PUD - Level3 access */
340 /* to find an entry in a page-table-directory. */
341 #define pud_page_vaddr(pud) ((unsigned long) __va(pud_val(pud) & PHYSICAL_PAGE_MASK))
342 #define pud_page(pud) (pfn_to_page(pud_val(pud) >> PAGE_SHIFT))
343 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
344 #define pud_offset(pgd, address) ((pud_t *) pgd_page_vaddr(*(pgd)) + pud_index(address))
345 #define pud_present(pud) (pud_val(pud) & _PAGE_PRESENT)
347 /* PMD - Level 2 access */
348 #define pmd_page_vaddr(pmd) ((unsigned long) __va(pmd_val(pmd) & PTE_MASK))
349 #define pmd_page(pmd) (pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT))
351 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
352 #define pmd_offset(dir, address) ((pmd_t *) pud_page_vaddr(*(dir)) + \
354 #define pmd_none(x) (!pmd_val(x))
355 #define pmd_present(x) (pmd_val(x) & _PAGE_PRESENT)
356 #define pmd_clear(xp) do { set_pmd(xp, __pmd(0)); } while (0)
357 #define pfn_pmd(nr,prot) (__pmd(((nr) << PAGE_SHIFT) | pgprot_val(prot)))
358 #define pmd_pfn(x) ((pmd_val(x) & __PHYSICAL_MASK) >> PAGE_SHIFT)
360 #define pte_to_pgoff(pte) ((pte_val(pte) & PHYSICAL_PAGE_MASK) >> PAGE_SHIFT)
361 #define pgoff_to_pte(off) ((pte_t) { ((off) << PAGE_SHIFT) | _PAGE_FILE })
362 #define PTE_FILE_MAX_BITS __PHYSICAL_MASK_SHIFT
364 /* PTE - Level 1 access. */
366 /* page, protection -> pte */
367 #define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
368 #define mk_pte_huge(entry) (pte_val(entry) |= _PAGE_PRESENT | _PAGE_PSE)
370 /* Change flags of a PTE */
371 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
373 pte_val(pte) &= _PAGE_CHG_MASK;
374 pte_val(pte) |= pgprot_val(newprot);
375 pte_val(pte) &= __supported_pte_mask;
379 #define pte_index(address) \
380 (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
381 #define pte_offset_kernel(dir, address) ((pte_t *) pmd_page_vaddr(*(dir)) + \
384 /* x86-64 always has all page tables mapped. */
385 #define pte_offset_map(dir,address) pte_offset_kernel(dir,address)
386 #define pte_offset_map_nested(dir,address) pte_offset_kernel(dir,address)
387 #define pte_unmap(pte) /* NOP */
388 #define pte_unmap_nested(pte) /* NOP */
390 #define update_mmu_cache(vma,address,pte) do { } while (0)
392 /* We only update the dirty/accessed state if we set
393 * the dirty bit by hand in the kernel, since the hardware
394 * will do the accessed bit for us, and we don't want to
395 * race with other CPU's that might be updating the dirty
396 * bit at the same time. */
397 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
398 #define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
400 int __changed = !pte_same(*(__ptep), __entry); \
401 if (__changed && __dirty) { \
402 set_pte(__ptep, __entry); \
403 flush_tlb_page(__vma, __address); \
408 /* Encode and de-code a swap entry */
409 #define __swp_type(x) (((x).val >> 1) & 0x3f)
410 #define __swp_offset(x) ((x).val >> 8)
411 #define __swp_entry(type, offset) ((swp_entry_t) { ((type) << 1) | ((offset) << 8) })
412 #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
413 #define __swp_entry_to_pte(x) ((pte_t) { (x).val })
415 extern spinlock_t pgd_lock;
416 extern struct list_head pgd_list;
418 extern int kern_addr_valid(unsigned long addr);
420 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
421 remap_pfn_range(vma, vaddr, pfn, size, prot)
423 #define HAVE_ARCH_UNMAPPED_AREA
425 #define pgtable_cache_init() do { } while (0)
426 #define check_pgt_cache() do { } while (0)
428 #define PAGE_AGP PAGE_KERNEL_NOCACHE
429 #define HAVE_PAGE_AGP 1
431 /* fs/proc/kcore.c */
432 #define kc_vaddr_to_offset(v) ((v) & __VIRTUAL_MASK)
433 #define kc_offset_to_vaddr(o) \
434 (((o) & (1UL << (__VIRTUAL_MASK_SHIFT-1))) ? ((o) | (~__VIRTUAL_MASK)) : (o))
436 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
437 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
438 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
439 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
440 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
441 #define __HAVE_ARCH_PTE_SAME
442 #include <arch-generic/pgtable.h>
443 #endif /* !__ASSEMBLY__ */
445 #endif /* _X86_64_PGTABLE_H */