[PATCH] Numa-aware slab allocator V5
[linux-2.6/btrfs-unstable.git] / include / asm-x86_64 / pgtable.h
blob5e0f2fdab0d35885f75f1177d54d18b4117808ec
1 #ifndef _X86_64_PGTABLE_H
2 #define _X86_64_PGTABLE_H
4 /*
5 * This file contains the functions and defines necessary to modify and use
6 * the x86-64 page table tree.
7 */
8 #include <asm/processor.h>
9 #include <asm/fixmap.h>
10 #include <asm/bitops.h>
11 #include <linux/threads.h>
12 #include <asm/pda.h>
14 extern pud_t level3_kernel_pgt[512];
15 extern pud_t level3_physmem_pgt[512];
16 extern pud_t level3_ident_pgt[512];
17 extern pmd_t level2_kernel_pgt[512];
18 extern pgd_t init_level4_pgt[];
19 extern unsigned long __supported_pte_mask;
21 #define swapper_pg_dir init_level4_pgt
23 extern int nonx_setup(char *str);
24 extern void paging_init(void);
25 extern void clear_kernel_mapping(unsigned long addr, unsigned long size);
27 extern unsigned long pgkern_mask;
30 * ZERO_PAGE is a global shared page that is always zero: used
31 * for zero-mapped memory areas etc..
33 extern unsigned long empty_zero_page[PAGE_SIZE/sizeof(unsigned long)];
34 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
37 * PGDIR_SHIFT determines what a top-level page table entry can map
39 #define PGDIR_SHIFT 39
40 #define PTRS_PER_PGD 512
43 * 3rd level page
45 #define PUD_SHIFT 30
46 #define PTRS_PER_PUD 512
49 * PMD_SHIFT determines the size of the area a middle-level
50 * page table can map
52 #define PMD_SHIFT 21
53 #define PTRS_PER_PMD 512
56 * entries per page directory level
58 #define PTRS_PER_PTE 512
60 #define pte_ERROR(e) \
61 printk("%s:%d: bad pte %p(%016lx).\n", __FILE__, __LINE__, &(e), pte_val(e))
62 #define pmd_ERROR(e) \
63 printk("%s:%d: bad pmd %p(%016lx).\n", __FILE__, __LINE__, &(e), pmd_val(e))
64 #define pud_ERROR(e) \
65 printk("%s:%d: bad pud %p(%016lx).\n", __FILE__, __LINE__, &(e), pud_val(e))
66 #define pgd_ERROR(e) \
67 printk("%s:%d: bad pgd %p(%016lx).\n", __FILE__, __LINE__, &(e), pgd_val(e))
69 #define pgd_none(x) (!pgd_val(x))
70 #define pud_none(x) (!pud_val(x))
72 static inline void set_pte(pte_t *dst, pte_t val)
74 pte_val(*dst) = pte_val(val);
76 #define set_pte_at(mm,addr,ptep,pteval) set_pte(ptep,pteval)
78 static inline void set_pmd(pmd_t *dst, pmd_t val)
80 pmd_val(*dst) = pmd_val(val);
83 static inline void set_pud(pud_t *dst, pud_t val)
85 pud_val(*dst) = pud_val(val);
88 extern inline void pud_clear (pud_t *pud)
90 set_pud(pud, __pud(0));
93 static inline void set_pgd(pgd_t *dst, pgd_t val)
95 pgd_val(*dst) = pgd_val(val);
98 extern inline void pgd_clear (pgd_t * pgd)
100 set_pgd(pgd, __pgd(0));
103 #define pud_page(pud) \
104 ((unsigned long) __va(pud_val(pud) & PHYSICAL_PAGE_MASK))
106 #define ptep_get_and_clear(mm,addr,xp) __pte(xchg(&(xp)->pte, 0))
108 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm, unsigned long addr, pte_t *ptep, int full)
110 pte_t pte;
111 if (full) {
112 pte = *ptep;
113 *ptep = __pte(0);
114 } else {
115 pte = ptep_get_and_clear(mm, addr, ptep);
117 return pte;
120 #define pte_same(a, b) ((a).pte == (b).pte)
122 #define PMD_SIZE (1UL << PMD_SHIFT)
123 #define PMD_MASK (~(PMD_SIZE-1))
124 #define PUD_SIZE (1UL << PUD_SHIFT)
125 #define PUD_MASK (~(PUD_SIZE-1))
126 #define PGDIR_SIZE (1UL << PGDIR_SHIFT)
127 #define PGDIR_MASK (~(PGDIR_SIZE-1))
129 #define USER_PTRS_PER_PGD (TASK_SIZE/PGDIR_SIZE)
130 #define FIRST_USER_ADDRESS 0
132 #ifndef __ASSEMBLY__
133 #define MAXMEM 0x3fffffffffffUL
134 #define VMALLOC_START 0xffffc20000000000UL
135 #define VMALLOC_END 0xffffe1ffffffffffUL
136 #define MODULES_VADDR 0xffffffff88000000UL
137 #define MODULES_END 0xfffffffffff00000UL
138 #define MODULES_LEN (MODULES_END - MODULES_VADDR)
140 #define _PAGE_BIT_PRESENT 0
141 #define _PAGE_BIT_RW 1
142 #define _PAGE_BIT_USER 2
143 #define _PAGE_BIT_PWT 3
144 #define _PAGE_BIT_PCD 4
145 #define _PAGE_BIT_ACCESSED 5
146 #define _PAGE_BIT_DIRTY 6
147 #define _PAGE_BIT_PSE 7 /* 4 MB (or 2MB) page */
148 #define _PAGE_BIT_GLOBAL 8 /* Global TLB entry PPro+ */
149 #define _PAGE_BIT_NX 63 /* No execute: only valid after cpuid check */
151 #define _PAGE_PRESENT 0x001
152 #define _PAGE_RW 0x002
153 #define _PAGE_USER 0x004
154 #define _PAGE_PWT 0x008
155 #define _PAGE_PCD 0x010
156 #define _PAGE_ACCESSED 0x020
157 #define _PAGE_DIRTY 0x040
158 #define _PAGE_PSE 0x080 /* 2MB page */
159 #define _PAGE_FILE 0x040 /* nonlinear file mapping, saved PTE; unset:swap */
160 #define _PAGE_GLOBAL 0x100 /* Global TLB entry */
162 #define _PAGE_PROTNONE 0x080 /* If not present */
163 #define _PAGE_NX (1UL<<_PAGE_BIT_NX)
165 #define _PAGE_TABLE (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED | _PAGE_DIRTY)
166 #define _KERNPG_TABLE (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY)
168 #define _PAGE_CHG_MASK (PTE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY)
170 #define PAGE_NONE __pgprot(_PAGE_PROTNONE | _PAGE_ACCESSED)
171 #define PAGE_SHARED __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
172 #define PAGE_SHARED_EXEC __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED)
173 #define PAGE_COPY_NOEXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
174 #define PAGE_COPY PAGE_COPY_NOEXEC
175 #define PAGE_COPY_EXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
176 #define PAGE_READONLY __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
177 #define PAGE_READONLY_EXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
178 #define __PAGE_KERNEL \
179 (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_NX)
180 #define __PAGE_KERNEL_EXEC \
181 (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED)
182 #define __PAGE_KERNEL_NOCACHE \
183 (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_PCD | _PAGE_ACCESSED | _PAGE_NX)
184 #define __PAGE_KERNEL_RO \
185 (_PAGE_PRESENT | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_NX)
186 #define __PAGE_KERNEL_VSYSCALL \
187 (_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
188 #define __PAGE_KERNEL_VSYSCALL_NOCACHE \
189 (_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_PCD)
190 #define __PAGE_KERNEL_LARGE \
191 (__PAGE_KERNEL | _PAGE_PSE)
192 #define __PAGE_KERNEL_LARGE_EXEC \
193 (__PAGE_KERNEL_EXEC | _PAGE_PSE)
195 #define MAKE_GLOBAL(x) __pgprot((x) | _PAGE_GLOBAL)
197 #define PAGE_KERNEL MAKE_GLOBAL(__PAGE_KERNEL)
198 #define PAGE_KERNEL_EXEC MAKE_GLOBAL(__PAGE_KERNEL_EXEC)
199 #define PAGE_KERNEL_RO MAKE_GLOBAL(__PAGE_KERNEL_RO)
200 #define PAGE_KERNEL_NOCACHE MAKE_GLOBAL(__PAGE_KERNEL_NOCACHE)
201 #define PAGE_KERNEL_VSYSCALL32 __pgprot(__PAGE_KERNEL_VSYSCALL)
202 #define PAGE_KERNEL_VSYSCALL MAKE_GLOBAL(__PAGE_KERNEL_VSYSCALL)
203 #define PAGE_KERNEL_LARGE MAKE_GLOBAL(__PAGE_KERNEL_LARGE)
204 #define PAGE_KERNEL_VSYSCALL_NOCACHE MAKE_GLOBAL(__PAGE_KERNEL_VSYSCALL_NOCACHE)
206 /* xwr */
207 #define __P000 PAGE_NONE
208 #define __P001 PAGE_READONLY
209 #define __P010 PAGE_COPY
210 #define __P011 PAGE_COPY
211 #define __P100 PAGE_READONLY_EXEC
212 #define __P101 PAGE_READONLY_EXEC
213 #define __P110 PAGE_COPY_EXEC
214 #define __P111 PAGE_COPY_EXEC
216 #define __S000 PAGE_NONE
217 #define __S001 PAGE_READONLY
218 #define __S010 PAGE_SHARED
219 #define __S011 PAGE_SHARED
220 #define __S100 PAGE_READONLY_EXEC
221 #define __S101 PAGE_READONLY_EXEC
222 #define __S110 PAGE_SHARED_EXEC
223 #define __S111 PAGE_SHARED_EXEC
225 static inline unsigned long pgd_bad(pgd_t pgd)
227 unsigned long val = pgd_val(pgd);
228 val &= ~PTE_MASK;
229 val &= ~(_PAGE_USER | _PAGE_DIRTY);
230 return val & ~(_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED);
233 static inline unsigned long pud_bad(pud_t pud)
235 unsigned long val = pud_val(pud);
236 val &= ~PTE_MASK;
237 val &= ~(_PAGE_USER | _PAGE_DIRTY);
238 return val & ~(_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED);
241 #define pte_none(x) (!pte_val(x))
242 #define pte_present(x) (pte_val(x) & (_PAGE_PRESENT | _PAGE_PROTNONE))
243 #define pte_clear(mm,addr,xp) do { set_pte_at(mm, addr, xp, __pte(0)); } while (0)
245 #define pages_to_mb(x) ((x) >> (20-PAGE_SHIFT)) /* FIXME: is this
246 right? */
247 #define pte_page(x) pfn_to_page(pte_pfn(x))
248 #define pte_pfn(x) ((pte_val(x) >> PAGE_SHIFT) & __PHYSICAL_MASK)
250 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
252 pte_t pte;
253 pte_val(pte) = (page_nr << PAGE_SHIFT);
254 pte_val(pte) |= pgprot_val(pgprot);
255 pte_val(pte) &= __supported_pte_mask;
256 return pte;
260 * The following only work if pte_present() is true.
261 * Undefined behaviour if not..
263 #define __LARGE_PTE (_PAGE_PSE|_PAGE_PRESENT)
264 static inline int pte_user(pte_t pte) { return pte_val(pte) & _PAGE_USER; }
265 extern inline int pte_read(pte_t pte) { return pte_val(pte) & _PAGE_USER; }
266 extern inline int pte_exec(pte_t pte) { return pte_val(pte) & _PAGE_USER; }
267 extern inline int pte_dirty(pte_t pte) { return pte_val(pte) & _PAGE_DIRTY; }
268 extern inline int pte_young(pte_t pte) { return pte_val(pte) & _PAGE_ACCESSED; }
269 extern inline int pte_write(pte_t pte) { return pte_val(pte) & _PAGE_RW; }
270 static inline int pte_file(pte_t pte) { return pte_val(pte) & _PAGE_FILE; }
271 static inline int pte_huge(pte_t pte) { return (pte_val(pte) & __LARGE_PTE) == __LARGE_PTE; }
273 extern inline pte_t pte_rdprotect(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_USER)); return pte; }
274 extern inline pte_t pte_exprotect(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_USER)); return pte; }
275 extern inline pte_t pte_mkclean(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_DIRTY)); return pte; }
276 extern inline pte_t pte_mkold(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_ACCESSED)); return pte; }
277 extern inline pte_t pte_wrprotect(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) & ~_PAGE_RW)); return pte; }
278 extern inline pte_t pte_mkread(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) | _PAGE_USER)); return pte; }
279 extern inline pte_t pte_mkexec(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) | _PAGE_USER)); return pte; }
280 extern inline pte_t pte_mkdirty(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) | _PAGE_DIRTY)); return pte; }
281 extern inline pte_t pte_mkyoung(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) | _PAGE_ACCESSED)); return pte; }
282 extern inline pte_t pte_mkwrite(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) | _PAGE_RW)); return pte; }
283 extern inline pte_t pte_mkhuge(pte_t pte) { set_pte(&pte, __pte(pte_val(pte) | __LARGE_PTE)); return pte; }
285 struct vm_area_struct;
287 static inline int ptep_test_and_clear_dirty(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
289 if (!pte_dirty(*ptep))
290 return 0;
291 return test_and_clear_bit(_PAGE_BIT_DIRTY, ptep);
294 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
296 if (!pte_young(*ptep))
297 return 0;
298 return test_and_clear_bit(_PAGE_BIT_ACCESSED, ptep);
301 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
303 clear_bit(_PAGE_BIT_RW, ptep);
307 * Macro to mark a page protection value as "uncacheable".
309 #define pgprot_noncached(prot) (__pgprot(pgprot_val(prot) | _PAGE_PCD | _PAGE_PWT))
311 static inline int pmd_large(pmd_t pte) {
312 return (pmd_val(pte) & __LARGE_PTE) == __LARGE_PTE;
317 * Conversion functions: convert a page and protection to a page entry,
318 * and a page entry and page directory to the page they refer to.
321 #define page_pte(page) page_pte_prot(page, __pgprot(0))
324 * Level 4 access.
326 #define pgd_page(pgd) ((unsigned long) __va((unsigned long)pgd_val(pgd) & PTE_MASK))
327 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
328 #define pgd_offset(mm, addr) ((mm)->pgd + pgd_index(addr))
329 #define pgd_offset_k(address) (init_level4_pgt + pgd_index(address))
330 #define pgd_present(pgd) (pgd_val(pgd) & _PAGE_PRESENT)
331 #define mk_kernel_pgd(address) ((pgd_t){ (address) | _KERNPG_TABLE })
333 /* PUD - Level3 access */
334 /* to find an entry in a page-table-directory. */
335 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
336 #define pud_offset(pgd, address) ((pud_t *) pgd_page(*(pgd)) + pud_index(address))
337 #define pud_offset_k(pgd, addr) pud_offset(pgd, addr)
338 #define pud_present(pud) (pud_val(pud) & _PAGE_PRESENT)
340 static inline pud_t *__pud_offset_k(pud_t *pud, unsigned long address)
342 return pud + pud_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 *) pud_page(*(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 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)
372 pte_t pte;
373 pte_val(pte) = physpage | pgprot_val(pgprot);
374 return pte;
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;
383 return pte;
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)) + \
389 pte_index(address))
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) \
406 do { \
407 if (__dirty) { \
408 set_pte(__ptep, __entry); \
409 flush_tlb_page(__vma, __address); \
411 } while (0)
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(vma, vaddr, paddr, size, prot) \
425 remap_pfn_range(vma, vaddr, (paddr) >> PAGE_SHIFT, size, prot)
427 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
428 remap_pfn_range(vma, vaddr, pfn, size, prot)
430 #define MK_IOSPACE_PFN(space, pfn) (pfn)
431 #define GET_IOSPACE(pfn) 0
432 #define GET_PFN(pfn) (pfn)
434 #define HAVE_ARCH_UNMAPPED_AREA
436 #define pgtable_cache_init() do { } while (0)
437 #define check_pgt_cache() do { } while (0)
439 #define PAGE_AGP PAGE_KERNEL_NOCACHE
440 #define HAVE_PAGE_AGP 1
442 /* fs/proc/kcore.c */
443 #define kc_vaddr_to_offset(v) ((v) & __VIRTUAL_MASK)
444 #define kc_offset_to_vaddr(o) \
445 (((o) & (1UL << (__VIRTUAL_MASK_SHIFT-1))) ? ((o) | (~__VIRTUAL_MASK)) : (o))
447 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
448 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
449 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
450 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
451 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
452 #define __HAVE_ARCH_PTE_SAME
453 #include <asm-generic/pgtable.h>
455 #endif /* _X86_64_PGTABLE_H */