1 #ifndef _X86_64_PGTABLE_H
2 #define _X86_64_PGTABLE_H
5 * This file contains the functions and defines necessary to modify and use
6 * the x86-64 page table tree.
8 * x86-64 has a 4 level table setup. Generic linux MM only supports
9 * three levels. The fourth level is currently a single static page that
10 * is shared by everybody and just contains a pointer to the current
11 * three level page setup on the beginning and some kernel mappings at
12 * the end. For more details see Documentation/x86_64/mm.txt
14 #include <asm/processor.h>
15 #include <asm/fixmap.h>
16 #include <asm/bitops.h>
17 #include <linux/threads.h>
20 extern pgd_t level3_kernel_pgt
[512];
21 extern pgd_t level3_physmem_pgt
[512];
22 extern pgd_t level3_ident_pgt
[512];
23 extern pmd_t level2_kernel_pgt
[512];
24 extern pml4_t init_level4_pgt
[];
25 extern pgd_t boot_vmalloc_pgt
[];
26 extern unsigned long __supported_pte_mask
;
28 #define swapper_pg_dir NULL
30 extern void paging_init(void);
31 extern void clear_kernel_mapping(unsigned long addr
, unsigned long size
);
33 extern unsigned long pgkern_mask
;
36 * ZERO_PAGE is a global shared page that is always zero: used
37 * for zero-mapped memory areas etc..
39 extern unsigned long empty_zero_page
[PAGE_SIZE
/sizeof(unsigned long)];
40 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
43 #define PTRS_PER_PML4 512
46 * PGDIR_SHIFT determines what a top-level page table entry can map
48 #define PGDIR_SHIFT 30
49 #define PTRS_PER_PGD 512
52 * PMD_SHIFT determines the size of the area a middle-level
56 #define PTRS_PER_PMD 512
59 * entries per page directory level
61 #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 pgd_ERROR(e) \
68 printk("%s:%d: bad pgd %p(%016lx).\n", __FILE__, __LINE__, &(e), pgd_val(e))
71 #define pml4_none(x) (!pml4_val(x))
72 #define pgd_none(x) (!pgd_val(x))
74 extern inline int pgd_present(pgd_t pgd
) { return !pgd_none(pgd
); }
76 static inline void set_pte(pte_t
*dst
, pte_t val
)
78 pte_val(*dst
) = pte_val(val
);
81 static inline void set_pmd(pmd_t
*dst
, pmd_t val
)
83 pmd_val(*dst
) = pmd_val(val
);
86 static inline void set_pgd(pgd_t
*dst
, pgd_t val
)
88 pgd_val(*dst
) = pgd_val(val
);
91 extern inline void pgd_clear (pgd_t
* pgd
)
93 set_pgd(pgd
, __pgd(0));
96 static inline void set_pml4(pml4_t
*dst
, pml4_t val
)
98 pml4_val(*dst
) = pml4_val(val
);
101 #define pgd_page(pgd) \
102 ((unsigned long) __va(pgd_val(pgd) & PHYSICAL_PAGE_MASK))
104 #define ptep_get_and_clear(xp) __pte(xchg(&(xp)->pte, 0))
105 #define pte_same(a, b) ((a).pte == (b).pte)
107 #define PML4_SIZE (1UL << PML4_SHIFT)
108 #define PML4_MASK (~(PML4_SIZE-1))
109 #define PMD_SIZE (1UL << PMD_SHIFT)
110 #define PMD_MASK (~(PMD_SIZE-1))
111 #define PGDIR_SIZE (1UL << PGDIR_SHIFT)
112 #define PGDIR_MASK (~(PGDIR_SIZE-1))
114 #define USER_PTRS_PER_PGD (TASK_SIZE/PGDIR_SIZE)
115 #define FIRST_USER_PGD_NR 0
117 #define USER_PGD_PTRS (PAGE_OFFSET >> PGDIR_SHIFT)
118 #define KERNEL_PGD_PTRS (PTRS_PER_PGD-USER_PGD_PTRS)
120 #define TWOLEVEL_PGDIR_SHIFT 20
121 #define BOOT_USER_L4_PTRS 1
122 #define BOOT_KERNEL_L4_PTRS 511 /* But we will do it in 4rd level */
127 #define VMALLOC_START 0xffffff0000000000UL
128 #define VMALLOC_END 0xffffff7fffffffffUL
129 #define MODULES_VADDR 0xffffffffa0000000UL
130 #define MODULES_END 0xffffffffafffffffUL
131 #define MODULES_LEN (MODULES_END - MODULES_VADDR)
133 #define IOMAP_START 0xfffffe8000000000UL
135 #define _PAGE_BIT_PRESENT 0
136 #define _PAGE_BIT_RW 1
137 #define _PAGE_BIT_USER 2
138 #define _PAGE_BIT_PWT 3
139 #define _PAGE_BIT_PCD 4
140 #define _PAGE_BIT_ACCESSED 5
141 #define _PAGE_BIT_DIRTY 6
142 #define _PAGE_BIT_PSE 7 /* 4 MB (or 2MB) page */
143 #define _PAGE_BIT_GLOBAL 8 /* Global TLB entry PPro+ */
144 #define _PAGE_BIT_NX 63 /* No execute: only valid after cpuid check */
146 #define _PAGE_PRESENT 0x001
147 #define _PAGE_RW 0x002
148 #define _PAGE_USER 0x004
149 #define _PAGE_PWT 0x008
150 #define _PAGE_PCD 0x010
151 #define _PAGE_ACCESSED 0x020
152 #define _PAGE_DIRTY 0x040
153 #define _PAGE_PSE 0x080 /* 2MB page */
154 #define _PAGE_FILE 0x040 /* set:pagecache, unset:swap */
155 #define _PAGE_GLOBAL 0x100 /* Global TLB entry */
157 #define _PAGE_PROTNONE 0x080 /* If not present */
158 #define _PAGE_NX (1UL<<_PAGE_BIT_NX)
160 #define _PAGE_TABLE (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED | _PAGE_DIRTY)
161 #define _KERNPG_TABLE (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY)
163 #define _PAGE_CHG_MASK (PTE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY)
165 #define PAGE_NONE __pgprot(_PAGE_PROTNONE | _PAGE_ACCESSED)
166 #define PAGE_SHARED __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
167 #define PAGE_SHARED_EXEC __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED)
168 #define PAGE_COPY_NOEXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
169 #define PAGE_COPY PAGE_COPY_NOEXEC
170 #define PAGE_COPY_EXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
171 #define PAGE_READONLY __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
172 #define PAGE_READONLY_EXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
173 #define __PAGE_KERNEL \
174 (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_NX)
175 #define __PAGE_KERNEL_EXEC \
176 (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED)
177 #define __PAGE_KERNEL_NOCACHE \
178 (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_PCD | _PAGE_ACCESSED | _PAGE_NX)
179 #define __PAGE_KERNEL_RO \
180 (_PAGE_PRESENT | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_NX)
181 #define __PAGE_KERNEL_VSYSCALL \
182 (_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
183 #define __PAGE_KERNEL_VSYSCALL_NOCACHE \
184 (_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_PCD)
185 #define __PAGE_KERNEL_LARGE \
186 (__PAGE_KERNEL | _PAGE_PSE)
188 #define MAKE_GLOBAL(x) __pgprot((x) | _PAGE_GLOBAL)
190 #define PAGE_KERNEL MAKE_GLOBAL(__PAGE_KERNEL)
191 #define PAGE_KERNEL_EXEC MAKE_GLOBAL(__PAGE_KERNEL_EXEC)
192 #define PAGE_KERNEL_RO MAKE_GLOBAL(__PAGE_KERNEL_RO)
193 #define PAGE_KERNEL_NOCACHE MAKE_GLOBAL(__PAGE_KERNEL_NOCACHE)
194 #define PAGE_KERNEL_VSYSCALL MAKE_GLOBAL(__PAGE_KERNEL_VSYSCALL)
195 #define PAGE_KERNEL_LARGE MAKE_GLOBAL(__PAGE_KERNEL_LARGE)
196 #define PAGE_KERNEL_VSYSCALL_NOCACHE MAKE_GLOBAL(__PAGE_KERNEL_VSYSCALL_NOCACHE)
199 #define __P000 PAGE_NONE
200 #define __P001 PAGE_READONLY
201 #define __P010 PAGE_COPY
202 #define __P011 PAGE_COPY
203 #define __P100 PAGE_READONLY_EXEC
204 #define __P101 PAGE_READONLY_EXEC
205 #define __P110 PAGE_COPY_EXEC
206 #define __P111 PAGE_COPY_EXEC
208 #define __S000 PAGE_NONE
209 #define __S001 PAGE_READONLY
210 #define __S010 PAGE_SHARED
211 #define __S011 PAGE_SHARED
212 #define __S100 PAGE_READONLY_EXEC
213 #define __S101 PAGE_READONLY_EXEC
214 #define __S110 PAGE_SHARED_EXEC
215 #define __S111 PAGE_SHARED_EXEC
217 static inline unsigned long pgd_bad(pgd_t pgd
)
219 unsigned long val
= pgd_val(pgd
);
221 val
&= ~(_PAGE_USER
| _PAGE_DIRTY
);
222 return val
& ~(_PAGE_PRESENT
| _PAGE_RW
| _PAGE_ACCESSED
);
225 #define pte_none(x) (!pte_val(x))
226 #define pte_present(x) (pte_val(x) & (_PAGE_PRESENT | _PAGE_PROTNONE))
227 #define pte_clear(xp) do { set_pte(xp, __pte(0)); } while (0)
229 #define pages_to_mb(x) ((x) >> (20-PAGE_SHIFT)) /* FIXME: is this
231 #define pte_page(x) pfn_to_page(pte_pfn(x))
232 #define pte_pfn(x) ((pte_val(x) >> PAGE_SHIFT) & __PHYSICAL_MASK)
234 static inline pte_t
pfn_pte(unsigned long page_nr
, pgprot_t pgprot
)
237 pte_val(pte
) = (page_nr
<< PAGE_SHIFT
);
238 pte_val(pte
) |= pgprot_val(pgprot
);
239 pte_val(pte
) &= __supported_pte_mask
;
244 * The following only work if pte_present() is true.
245 * Undefined behaviour if not..
247 static inline int pte_user(pte_t pte
) { return pte_val(pte
) & _PAGE_USER
; }
248 extern inline int pte_read(pte_t pte
) { return pte_val(pte
) & _PAGE_USER
; }
249 extern inline int pte_exec(pte_t pte
) { return pte_val(pte
) & _PAGE_USER
; }
250 extern inline int pte_dirty(pte_t pte
) { return pte_val(pte
) & _PAGE_DIRTY
; }
251 extern inline int pte_young(pte_t pte
) { return pte_val(pte
) & _PAGE_ACCESSED
; }
252 extern inline int pte_write(pte_t pte
) { return pte_val(pte
) & _PAGE_RW
; }
253 static inline int pte_file(pte_t pte
) { return pte_val(pte
) & _PAGE_FILE
; }
255 extern inline pte_t
pte_rdprotect(pte_t pte
) { set_pte(&pte
, __pte(pte_val(pte
) & ~_PAGE_USER
)); return pte
; }
256 extern inline pte_t
pte_exprotect(pte_t pte
) { set_pte(&pte
, __pte(pte_val(pte
) & ~_PAGE_USER
)); return pte
; }
257 extern inline pte_t
pte_mkclean(pte_t pte
) { set_pte(&pte
, __pte(pte_val(pte
) & ~_PAGE_DIRTY
)); return pte
; }
258 extern inline pte_t
pte_mkold(pte_t pte
) { set_pte(&pte
, __pte(pte_val(pte
) & ~_PAGE_ACCESSED
)); return pte
; }
259 extern inline pte_t
pte_wrprotect(pte_t pte
) { set_pte(&pte
, __pte(pte_val(pte
) & ~_PAGE_RW
)); return pte
; }
260 extern inline pte_t
pte_mkread(pte_t pte
) { set_pte(&pte
, __pte(pte_val(pte
) | _PAGE_USER
)); return pte
; }
261 extern inline pte_t
pte_mkexec(pte_t pte
) { set_pte(&pte
, __pte(pte_val(pte
) | _PAGE_USER
)); return pte
; }
262 extern inline pte_t
pte_mkdirty(pte_t pte
) { set_pte(&pte
, __pte(pte_val(pte
) | _PAGE_DIRTY
)); return pte
; }
263 extern inline pte_t
pte_mkyoung(pte_t pte
) { set_pte(&pte
, __pte(pte_val(pte
) | _PAGE_ACCESSED
)); return pte
; }
264 extern inline pte_t
pte_mkwrite(pte_t pte
) { set_pte(&pte
, __pte(pte_val(pte
) | _PAGE_RW
)); return pte
; }
266 static inline int ptep_test_and_clear_dirty(pte_t
*ptep
)
268 if (!pte_dirty(*ptep
))
270 return test_and_clear_bit(_PAGE_BIT_DIRTY
, ptep
);
273 static inline int ptep_test_and_clear_young(pte_t
*ptep
)
275 if (!pte_young(*ptep
))
277 return test_and_clear_bit(_PAGE_BIT_ACCESSED
, ptep
);
280 static inline void ptep_set_wrprotect(pte_t
*ptep
) { clear_bit(_PAGE_BIT_RW
, ptep
); }
281 static inline void ptep_mkdirty(pte_t
*ptep
) { set_bit(_PAGE_BIT_DIRTY
, ptep
); }
284 * Macro to mark a page protection value as "uncacheable".
286 #define pgprot_noncached(prot) (__pgprot(pgprot_val(prot) | _PAGE_PCD | _PAGE_PWT))
288 #define __LARGE_PTE (_PAGE_PSE|_PAGE_PRESENT)
289 static inline int pmd_large(pmd_t pte
) {
290 return (pmd_val(pte
) & __LARGE_PTE
) == __LARGE_PTE
;
295 * Conversion functions: convert a page and protection to a page entry,
296 * and a page entry and page directory to the page they refer to.
299 #define page_pte(page) page_pte_prot(page, __pgprot(0))
303 * Never use these in the common code.
305 #define pml4_page(pml4) ((unsigned long) __va(pml4_val(pml4) & PTE_MASK))
306 #define pml4_index(address) ((address >> PML4_SHIFT) & (PTRS_PER_PML4-1))
307 #define pml4_offset_k(address) (init_level4_pgt + pml4_index(address))
308 #define pml4_present(pml4) (pml4_val(pml4) & _PAGE_PRESENT)
309 #define mk_kernel_pml4(address) ((pml4_t){ (address) | _KERNPG_TABLE })
310 #define level3_offset_k(dir, address) ((pgd_t *) pml4_page(*(dir)) + pgd_index(address))
312 /* PGD - Level3 access */
313 /* to find an entry in a page-table-directory. */
314 #define pgd_index(address) ((address >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
315 static inline pgd_t
*__pgd_offset_k(pgd_t
*pgd
, unsigned long address
)
317 return pgd
+ pgd_index(address
);
320 /* Find correct pgd via the hidden fourth level page level: */
322 /* This accesses the reference page table of the boot cpu.
323 Other CPUs get synced lazily via the page fault handler. */
324 static inline pgd_t
*pgd_offset_k(unsigned long address
)
328 addr
= pml4_val(init_level4_pgt
[pml4_index(address
)]);
329 addr
&= PHYSICAL_PAGE_MASK
;
330 return __pgd_offset_k((pgd_t
*)__va(addr
), address
);
333 /* Access the pgd of the page table as seen by the current CPU. */
334 static inline pgd_t
*current_pgd_offset_k(unsigned long address
)
338 addr
= read_pda(level4_pgt
)[pml4_index(address
)];
339 addr
&= PHYSICAL_PAGE_MASK
;
340 return __pgd_offset_k((pgd_t
*)__va(addr
), address
);
343 #define pgd_offset(mm, address) ((mm)->pgd+pgd_index(address))
345 /* PMD - Level 2 access */
346 #define pmd_page_kernel(pmd) ((unsigned long) __va(pmd_val(pmd) & PTE_MASK))
347 #define pmd_page(pmd) (pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT))
349 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
350 #define pmd_offset(dir, address) ((pmd_t *) pgd_page(*(dir)) + \
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 pmd_bad(x) ((pmd_val(x) & (~PTE_MASK & ~_PAGE_USER)) != _KERNPG_TABLE )
356 #define pfn_pmd(nr,prot) (__pmd(((nr) << PAGE_SHIFT) | pgprot_val(prot)))
357 #define pmd_pfn(x) ((pmd_val(x) >> PAGE_SHIFT) & __PHYSICAL_MASK)
359 #define pte_to_pgoff(pte) ((pte_val(pte) & PHYSICAL_PAGE_MASK) >> PAGE_SHIFT)
360 #define pgoff_to_pte(off) ((pte_t) { ((off) << PAGE_SHIFT) | _PAGE_FILE })
361 #define PTE_FILE_MAX_BITS __PHYSICAL_MASK_SHIFT
363 /* PTE - Level 1 access. */
365 /* page, protection -> pte */
366 #define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
367 #define mk_pte_huge(entry) (pte_val(entry) |= _PAGE_PRESENT | _PAGE_PSE)
369 /* physical address -> PTE */
370 static inline pte_t
mk_pte_phys(unsigned long physpage
, pgprot_t pgprot
)
373 pte_val(pte
) = physpage
| pgprot_val(pgprot
);
377 /* Change flags of a PTE */
378 extern inline pte_t
pte_modify(pte_t pte
, pgprot_t newprot
)
380 pte_val(pte
) &= _PAGE_CHG_MASK
;
381 pte_val(pte
) |= pgprot_val(newprot
);
382 pte_val(pte
) &= __supported_pte_mask
;
386 #define pte_index(address) \
387 ((address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
388 #define pte_offset_kernel(dir, address) ((pte_t *) pmd_page_kernel(*(dir)) + \
391 /* x86-64 always has all page tables mapped. */
392 #define pte_offset_map(dir,address) pte_offset_kernel(dir,address)
393 #define pte_offset_map_nested(dir,address) pte_offset_kernel(dir,address)
394 #define pte_unmap(pte) /* NOP */
395 #define pte_unmap_nested(pte) /* NOP */
397 #define update_mmu_cache(vma,address,pte) do { } while (0)
399 /* We only update the dirty/accessed state if we set
400 * the dirty bit by hand in the kernel, since the hardware
401 * will do the accessed bit for us, and we don't want to
402 * race with other CPU's that might be updating the dirty
403 * bit at the same time. */
404 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
405 #define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
408 set_pte(__ptep, __entry); \
409 flush_tlb_page(__vma, __address); \
413 /* Encode and de-code a swap entry */
414 #define __swp_type(x) (((x).val >> 1) & 0x3f)
415 #define __swp_offset(x) ((x).val >> 8)
416 #define __swp_entry(type, offset) ((swp_entry_t) { ((type) << 1) | ((offset) << 8) })
417 #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
418 #define __swp_entry_to_pte(x) ((pte_t) { (x).val })
420 #endif /* !__ASSEMBLY__ */
422 extern int kern_addr_valid(unsigned long addr
);
424 #define io_remap_page_range remap_page_range
426 #define HAVE_ARCH_UNMAPPED_AREA
428 #define pgtable_cache_init() do { } while (0)
429 #define check_pgt_cache() do { } while (0)
431 #define PAGE_AGP PAGE_KERNEL_NOCACHE
432 #define HAVE_PAGE_AGP 1
434 /* fs/proc/kcore.c */
435 #define kc_vaddr_to_offset(v) ((v) & __VIRTUAL_MASK)
436 #define kc_offset_to_vaddr(o) \
437 (((o) & (1UL << (__VIRTUAL_MASK_SHIFT-1))) ? ((o) | (~__VIRTUAL_MASK)) : (o))
439 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
440 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
441 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
442 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
443 #define __HAVE_ARCH_PTEP_MKDIRTY
444 #define __HAVE_ARCH_PTE_SAME
445 #include <asm-generic/pgtable.h>
447 #endif /* _X86_64_PGTABLE_H */