blob: c04b3d0f484e78043467443b372caf32588c0e12 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#ifndef _I386_PGTABLE_H
2#define _I386_PGTABLE_H
3
Linus Torvalds1da177e2005-04-16 15:20:36 -07004
5/*
6 * The Linux memory management assumes a three-level page table setup. On
7 * the i386, we use that, but "fold" the mid level into the top-level page
8 * table, so that we physically have the same two-level page table as the
9 * i386 mmu expects.
10 *
11 * This file contains the functions and defines necessary to modify and use
12 * the i386 page table tree.
13 */
14#ifndef __ASSEMBLY__
15#include <asm/processor.h>
16#include <asm/fixmap.h>
17#include <linux/threads.h>
18
19#ifndef _I386_BITOPS_H
20#include <asm/bitops.h>
21#endif
22
23#include <linux/slab.h>
24#include <linux/list.h>
25#include <linux/spinlock.h>
26
Tim Schmielau8c65b4a2005-11-07 00:59:43 -080027struct mm_struct;
28struct vm_area_struct;
29
Linus Torvalds1da177e2005-04-16 15:20:36 -070030/*
31 * ZERO_PAGE is a global shared page that is always zero: used
32 * for zero-mapped memory areas etc..
33 */
34#define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
35extern unsigned long empty_zero_page[1024];
36extern pgd_t swapper_pg_dir[1024];
37extern kmem_cache_t *pgd_cache;
38extern kmem_cache_t *pmd_cache;
39extern spinlock_t pgd_lock;
40extern struct page *pgd_list;
41
42void pmd_ctor(void *, kmem_cache_t *, unsigned long);
43void pgd_ctor(void *, kmem_cache_t *, unsigned long);
44void pgd_dtor(void *, kmem_cache_t *, unsigned long);
45void pgtable_cache_init(void);
46void paging_init(void);
47
48/*
49 * The Linux x86 paging architecture is 'compile-time dual-mode', it
50 * implements both the traditional 2-level x86 page tables and the
51 * newer 3-level PAE-mode page tables.
52 */
53#ifdef CONFIG_X86_PAE
54# include <asm/pgtable-3level-defs.h>
55# define PMD_SIZE (1UL << PMD_SHIFT)
56# define PMD_MASK (~(PMD_SIZE-1))
57#else
58# include <asm/pgtable-2level-defs.h>
59#endif
60
61#define PGDIR_SIZE (1UL << PGDIR_SHIFT)
62#define PGDIR_MASK (~(PGDIR_SIZE-1))
63
64#define USER_PTRS_PER_PGD (TASK_SIZE/PGDIR_SIZE)
Hugh Dickinsd455a362005-04-19 13:29:23 -070065#define FIRST_USER_ADDRESS 0
Linus Torvalds1da177e2005-04-16 15:20:36 -070066
67#define USER_PGD_PTRS (PAGE_OFFSET >> PGDIR_SHIFT)
68#define KERNEL_PGD_PTRS (PTRS_PER_PGD-USER_PGD_PTRS)
69
70#define TWOLEVEL_PGDIR_SHIFT 22
71#define BOOT_USER_PGD_PTRS (__PAGE_OFFSET >> TWOLEVEL_PGDIR_SHIFT)
72#define BOOT_KERNEL_PGD_PTRS (1024-BOOT_USER_PGD_PTRS)
73
74/* Just any arbitrary offset to the start of the vmalloc VM area: the
75 * current 8MB value just means that there will be a 8MB "hole" after the
76 * physical memory until the kernel virtual memory starts. That means that
77 * any out-of-bounds memory accesses will hopefully be caught.
78 * The vmalloc() routines leaves a hole of 4kB between each vmalloced
79 * area for the same reason. ;)
80 */
81#define VMALLOC_OFFSET (8*1024*1024)
82#define VMALLOC_START (((unsigned long) high_memory + vmalloc_earlyreserve + \
83 2*VMALLOC_OFFSET-1) & ~(VMALLOC_OFFSET-1))
84#ifdef CONFIG_HIGHMEM
85# define VMALLOC_END (PKMAP_BASE-2*PAGE_SIZE)
86#else
87# define VMALLOC_END (FIXADDR_START-2*PAGE_SIZE)
88#endif
89
90/*
Paolo 'Blaisorblade' Giarrusso9b4ee402005-09-03 15:54:57 -070091 * _PAGE_PSE set in the page directory entry just means that
Linus Torvalds1da177e2005-04-16 15:20:36 -070092 * the page directory entry points directly to a 4MB-aligned block of
93 * memory.
94 */
95#define _PAGE_BIT_PRESENT 0
96#define _PAGE_BIT_RW 1
97#define _PAGE_BIT_USER 2
98#define _PAGE_BIT_PWT 3
99#define _PAGE_BIT_PCD 4
100#define _PAGE_BIT_ACCESSED 5
101#define _PAGE_BIT_DIRTY 6
102#define _PAGE_BIT_PSE 7 /* 4 MB (or 2MB) page, Pentium+, if present.. */
103#define _PAGE_BIT_GLOBAL 8 /* Global TLB entry PPro+ */
104#define _PAGE_BIT_UNUSED1 9 /* available for programmer */
105#define _PAGE_BIT_UNUSED2 10
106#define _PAGE_BIT_UNUSED3 11
107#define _PAGE_BIT_NX 63
108
109#define _PAGE_PRESENT 0x001
110#define _PAGE_RW 0x002
111#define _PAGE_USER 0x004
112#define _PAGE_PWT 0x008
113#define _PAGE_PCD 0x010
114#define _PAGE_ACCESSED 0x020
115#define _PAGE_DIRTY 0x040
116#define _PAGE_PSE 0x080 /* 4 MB (or 2MB) page, Pentium+, if present.. */
117#define _PAGE_GLOBAL 0x100 /* Global TLB entry PPro+ */
118#define _PAGE_UNUSED1 0x200 /* available for programmer */
119#define _PAGE_UNUSED2 0x400
120#define _PAGE_UNUSED3 0x800
121
Paolo 'Blaisorblade' Giarrusso9b4ee402005-09-03 15:54:57 -0700122/* If _PAGE_PRESENT is clear, we use these: */
123#define _PAGE_FILE 0x040 /* nonlinear file mapping, saved PTE; unset:swap */
124#define _PAGE_PROTNONE 0x080 /* if the user mapped it with PROT_NONE;
125 pte_present gives true */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126#ifdef CONFIG_X86_PAE
127#define _PAGE_NX (1ULL<<_PAGE_BIT_NX)
128#else
129#define _PAGE_NX 0
130#endif
131
132#define _PAGE_TABLE (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED | _PAGE_DIRTY)
133#define _KERNPG_TABLE (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY)
134#define _PAGE_CHG_MASK (PTE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY)
135
136#define PAGE_NONE \
137 __pgprot(_PAGE_PROTNONE | _PAGE_ACCESSED)
138#define PAGE_SHARED \
139 __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED)
140
141#define PAGE_SHARED_EXEC \
142 __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED)
143#define PAGE_COPY_NOEXEC \
144 __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
145#define PAGE_COPY_EXEC \
146 __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
147#define PAGE_COPY \
148 PAGE_COPY_NOEXEC
149#define PAGE_READONLY \
150 __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
151#define PAGE_READONLY_EXEC \
152 __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
153
154#define _PAGE_KERNEL \
155 (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_NX)
156#define _PAGE_KERNEL_EXEC \
157 (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED)
158
159extern unsigned long long __PAGE_KERNEL, __PAGE_KERNEL_EXEC;
160#define __PAGE_KERNEL_RO (__PAGE_KERNEL & ~_PAGE_RW)
161#define __PAGE_KERNEL_NOCACHE (__PAGE_KERNEL | _PAGE_PCD)
162#define __PAGE_KERNEL_LARGE (__PAGE_KERNEL | _PAGE_PSE)
163#define __PAGE_KERNEL_LARGE_EXEC (__PAGE_KERNEL_EXEC | _PAGE_PSE)
164
165#define PAGE_KERNEL __pgprot(__PAGE_KERNEL)
166#define PAGE_KERNEL_RO __pgprot(__PAGE_KERNEL_RO)
167#define PAGE_KERNEL_EXEC __pgprot(__PAGE_KERNEL_EXEC)
168#define PAGE_KERNEL_NOCACHE __pgprot(__PAGE_KERNEL_NOCACHE)
169#define PAGE_KERNEL_LARGE __pgprot(__PAGE_KERNEL_LARGE)
170#define PAGE_KERNEL_LARGE_EXEC __pgprot(__PAGE_KERNEL_LARGE_EXEC)
171
172/*
173 * The i386 can't do page protection for execute, and considers that
174 * the same are read. Also, write permissions imply read permissions.
175 * This is the closest we can get..
176 */
177#define __P000 PAGE_NONE
178#define __P001 PAGE_READONLY
179#define __P010 PAGE_COPY
180#define __P011 PAGE_COPY
181#define __P100 PAGE_READONLY_EXEC
182#define __P101 PAGE_READONLY_EXEC
183#define __P110 PAGE_COPY_EXEC
184#define __P111 PAGE_COPY_EXEC
185
186#define __S000 PAGE_NONE
187#define __S001 PAGE_READONLY
188#define __S010 PAGE_SHARED
189#define __S011 PAGE_SHARED
190#define __S100 PAGE_READONLY_EXEC
191#define __S101 PAGE_READONLY_EXEC
192#define __S110 PAGE_SHARED_EXEC
193#define __S111 PAGE_SHARED_EXEC
194
195/*
196 * Define this if things work differently on an i386 and an i486:
197 * it will (on an i486) warn about kernel memory accesses that are
Jesper Juhle49332b2005-05-01 08:59:08 -0700198 * done without a 'access_ok(VERIFY_WRITE,..)'
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 */
Jesper Juhle49332b2005-05-01 08:59:08 -0700200#undef TEST_ACCESS_OK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201
202/* The boot page tables (all created as a single array) */
203extern unsigned long pg0[];
204
205#define pte_present(x) ((x).pte_low & (_PAGE_PRESENT | _PAGE_PROTNONE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206
Hugh Dickins705e87c2005-10-29 18:16:27 -0700207/* To avoid harmful races, pmd_none(x) should check only the lower when PAE */
208#define pmd_none(x) (!(unsigned long)pmd_val(x))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209#define pmd_present(x) (pmd_val(x) & _PAGE_PRESENT)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210#define pmd_bad(x) ((pmd_val(x) & (~PAGE_MASK & ~_PAGE_USER)) != _KERNPG_TABLE)
211
212
213#define pages_to_mb(x) ((x) >> (20-PAGE_SHIFT))
214
215/*
216 * The following only work if pte_present() is true.
217 * Undefined behaviour if not..
218 */
219static inline int pte_user(pte_t pte) { return (pte).pte_low & _PAGE_USER; }
220static inline int pte_read(pte_t pte) { return (pte).pte_low & _PAGE_USER; }
221static inline int pte_dirty(pte_t pte) { return (pte).pte_low & _PAGE_DIRTY; }
222static inline int pte_young(pte_t pte) { return (pte).pte_low & _PAGE_ACCESSED; }
223static inline int pte_write(pte_t pte) { return (pte).pte_low & _PAGE_RW; }
Zhang, Yanmin8f860592006-03-22 00:08:50 -0800224static inline int pte_huge(pte_t pte) { return (pte).pte_low & _PAGE_PSE; }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225
226/*
227 * The following only works if pte_present() is not true.
228 */
229static inline int pte_file(pte_t pte) { return (pte).pte_low & _PAGE_FILE; }
230
231static inline pte_t pte_rdprotect(pte_t pte) { (pte).pte_low &= ~_PAGE_USER; return pte; }
232static inline pte_t pte_exprotect(pte_t pte) { (pte).pte_low &= ~_PAGE_USER; return pte; }
233static inline pte_t pte_mkclean(pte_t pte) { (pte).pte_low &= ~_PAGE_DIRTY; return pte; }
234static inline pte_t pte_mkold(pte_t pte) { (pte).pte_low &= ~_PAGE_ACCESSED; return pte; }
235static inline pte_t pte_wrprotect(pte_t pte) { (pte).pte_low &= ~_PAGE_RW; return pte; }
236static inline pte_t pte_mkread(pte_t pte) { (pte).pte_low |= _PAGE_USER; return pte; }
237static inline pte_t pte_mkexec(pte_t pte) { (pte).pte_low |= _PAGE_USER; return pte; }
238static inline pte_t pte_mkdirty(pte_t pte) { (pte).pte_low |= _PAGE_DIRTY; return pte; }
239static inline pte_t pte_mkyoung(pte_t pte) { (pte).pte_low |= _PAGE_ACCESSED; return pte; }
240static inline pte_t pte_mkwrite(pte_t pte) { (pte).pte_low |= _PAGE_RW; return pte; }
Zhang, Yanmin8f860592006-03-22 00:08:50 -0800241static inline pte_t pte_mkhuge(pte_t pte) { (pte).pte_low |= _PAGE_PSE; return pte; }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242
243#ifdef CONFIG_X86_PAE
244# include <asm/pgtable-3level.h>
245#else
246# include <asm/pgtable-2level.h>
247#endif
248
249static inline int ptep_test_and_clear_dirty(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
250{
251 if (!pte_dirty(*ptep))
252 return 0;
253 return test_and_clear_bit(_PAGE_BIT_DIRTY, &ptep->pte_low);
254}
255
256static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
257{
258 if (!pte_young(*ptep))
259 return 0;
260 return test_and_clear_bit(_PAGE_BIT_ACCESSED, &ptep->pte_low);
261}
262
Zachary Amsdena6003882005-09-03 15:55:04 -0700263static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm, unsigned long addr, pte_t *ptep, int full)
264{
265 pte_t pte;
266 if (full) {
267 pte = *ptep;
Zachary Amsden6e5882c2006-04-27 11:32:29 -0700268 pte_clear(mm, addr, ptep);
Zachary Amsdena6003882005-09-03 15:55:04 -0700269 } else {
270 pte = ptep_get_and_clear(mm, addr, ptep);
271 }
272 return pte;
273}
274
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
276{
277 clear_bit(_PAGE_BIT_RW, &ptep->pte_low);
278}
279
280/*
Zachary Amsdend7271b12005-09-03 15:56:50 -0700281 * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
282 *
283 * dst - pointer to pgd range anwhere on a pgd page
284 * src - ""
285 * count - the number of pgds to copy.
286 *
287 * dst and src can be on the same page, but the range must not overlap,
288 * and must not cross a page boundary.
289 */
290static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
291{
292 memcpy(dst, src, count * sizeof(pgd_t));
293}
294
295/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296 * Macro to mark a page protection value as "uncacheable". On processors which do not support
297 * it, this is a no-op.
298 */
299#define pgprot_noncached(prot) ((boot_cpu_data.x86 > 3) \
300 ? (__pgprot(pgprot_val(prot) | _PAGE_PCD | _PAGE_PWT)) : (prot))
301
302/*
303 * Conversion functions: convert a page and protection to a page entry,
304 * and a page entry and page directory to the page they refer to.
305 */
306
307#define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308
309static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
310{
311 pte.pte_low &= _PAGE_CHG_MASK;
312 pte.pte_low |= pgprot_val(newprot);
313#ifdef CONFIG_X86_PAE
314 /*
315 * Chop off the NX bit (if present), and add the NX portion of
316 * the newprot (if present):
317 */
318 pte.pte_high &= ~(1 << (_PAGE_BIT_NX - 32));
319 pte.pte_high |= (pgprot_val(newprot) >> 32) & \
320 (__supported_pte_mask >> 32);
321#endif
322 return pte;
323}
324
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325#define pmd_large(pmd) \
326((pmd_val(pmd) & (_PAGE_PSE|_PAGE_PRESENT)) == (_PAGE_PSE|_PAGE_PRESENT))
327
328/*
329 * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
330 *
331 * this macro returns the index of the entry in the pgd page which would
332 * control the given virtual address
333 */
334#define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
335#define pgd_index_k(addr) pgd_index(addr)
336
337/*
338 * pgd_offset() returns a (pgd_t *)
339 * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
340 */
341#define pgd_offset(mm, address) ((mm)->pgd+pgd_index(address))
342
343/*
344 * a shortcut which implies the use of the kernel's pgd, instead
345 * of a process's
346 */
347#define pgd_offset_k(address) pgd_offset(&init_mm, address)
348
349/*
350 * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
351 *
352 * this macro returns the index of the entry in the pmd page which would
353 * control the given virtual address
354 */
355#define pmd_index(address) \
356 (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
357
358/*
359 * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
360 *
361 * this macro returns the index of the entry in the pte page which would
362 * control the given virtual address
363 */
364#define pte_index(address) \
365 (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
366#define pte_offset_kernel(dir, address) \
Dave McCracken46a82b22006-09-25 23:31:48 -0700367 ((pte_t *) pmd_page_vaddr(*(dir)) + pte_index(address))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368
Paolo 'Blaisorblade' Giarrussoca140fd2005-10-30 14:59:31 -0800369#define pmd_page(pmd) (pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT))
370
Dave McCracken46a82b22006-09-25 23:31:48 -0700371#define pmd_page_vaddr(pmd) \
Paolo 'Blaisorblade' Giarrussoca140fd2005-10-30 14:59:31 -0800372 ((unsigned long) __va(pmd_val(pmd) & PAGE_MASK))
373
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374/*
375 * Helper function that returns the kernel pagetable entry controlling
376 * the virtual address 'address'. NULL means no pagetable entry present.
377 * NOTE: the return type is pte_t but if the pmd is PSE then we return it
378 * as a pte too.
379 */
380extern pte_t *lookup_address(unsigned long address);
381
382/*
383 * Make a given kernel text page executable/non-executable.
384 * Returns the previous executability setting of that page (which
385 * is used to restore the previous state). Used by the SMP bootup code.
386 * NOTE: this is an __init function for security reasons.
387 */
388#ifdef CONFIG_X86_PAE
389 extern int set_kernel_exec(unsigned long vaddr, int enable);
390#else
391 static inline int set_kernel_exec(unsigned long vaddr, int enable) { return 0;}
392#endif
393
394extern void noexec_setup(const char *str);
395
396#if defined(CONFIG_HIGHPTE)
397#define pte_offset_map(dir, address) \
398 ((pte_t *)kmap_atomic(pmd_page(*(dir)),KM_PTE0) + pte_index(address))
399#define pte_offset_map_nested(dir, address) \
400 ((pte_t *)kmap_atomic(pmd_page(*(dir)),KM_PTE1) + pte_index(address))
401#define pte_unmap(pte) kunmap_atomic(pte, KM_PTE0)
402#define pte_unmap_nested(pte) kunmap_atomic(pte, KM_PTE1)
403#else
404#define pte_offset_map(dir, address) \
405 ((pte_t *)page_address(pmd_page(*(dir))) + pte_index(address))
406#define pte_offset_map_nested(dir, address) pte_offset_map(dir, address)
407#define pte_unmap(pte) do { } while (0)
408#define pte_unmap_nested(pte) do { } while (0)
409#endif
410
411/*
412 * The i386 doesn't have any external MMU info: the kernel page
413 * tables contain all the necessary information.
414 *
415 * Also, we only update the dirty/accessed state if we set
416 * the dirty bit by hand in the kernel, since the hardware
417 * will do the accessed bit for us, and we don't want to
418 * race with other CPU's that might be updating the dirty
419 * bit at the same time.
420 */
421#define update_mmu_cache(vma,address,pte) do { } while (0)
422#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
423#define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
424 do { \
425 if (__dirty) { \
426 (__ptep)->pte_low = (__entry).pte_low; \
427 flush_tlb_page(__vma, __address); \
428 } \
429 } while (0)
430
431#endif /* !__ASSEMBLY__ */
432
Andy Whitcroft05b79bd2005-06-23 00:07:57 -0700433#ifdef CONFIG_FLATMEM
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434#define kern_addr_valid(addr) (1)
Andy Whitcroft05b79bd2005-06-23 00:07:57 -0700435#endif /* CONFIG_FLATMEM */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437#define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
438 remap_pfn_range(vma, vaddr, pfn, size, prot)
439
440#define MK_IOSPACE_PFN(space, pfn) (pfn)
441#define GET_IOSPACE(pfn) 0
442#define GET_PFN(pfn) (pfn)
443
444#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
445#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
446#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
Zachary Amsdena6003882005-09-03 15:55:04 -0700447#define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448#define __HAVE_ARCH_PTEP_SET_WRPROTECT
449#define __HAVE_ARCH_PTE_SAME
450#include <asm-generic/pgtable.h>
451
452#endif /* _I386_PGTABLE_H */