1 |
1275 |
phoenix |
#ifndef _I386_PGTABLE_H
|
2 |
|
|
#define _I386_PGTABLE_H
|
3 |
|
|
|
4 |
|
|
#include <linux/config.h>
|
5 |
|
|
|
6 |
|
|
/*
|
7 |
|
|
* The Linux memory management assumes a three-level page table setup. On
|
8 |
|
|
* the i386, we use that, but "fold" the mid level into the top-level page
|
9 |
|
|
* table, so that we physically have the same two-level page table as the
|
10 |
|
|
* i386 mmu expects.
|
11 |
|
|
*
|
12 |
|
|
* This file contains the functions and defines necessary to modify and use
|
13 |
|
|
* the i386 page table tree.
|
14 |
|
|
*/
|
15 |
|
|
#ifndef __ASSEMBLY__
|
16 |
|
|
#include <asm/processor.h>
|
17 |
|
|
#include <asm/fixmap.h>
|
18 |
|
|
#include <linux/threads.h>
|
19 |
|
|
|
20 |
|
|
#ifndef _I386_BITOPS_H
|
21 |
|
|
#include <asm/bitops.h>
|
22 |
|
|
#endif
|
23 |
|
|
|
24 |
|
|
extern pgd_t swapper_pg_dir[1024];
|
25 |
|
|
extern void paging_init(void);
|
26 |
|
|
|
27 |
|
|
/* Caches aren't brain-dead on the intel. */
|
28 |
|
|
#define flush_cache_all() do { } while (0)
|
29 |
|
|
#define flush_cache_mm(mm) do { } while (0)
|
30 |
|
|
#define flush_cache_range(mm, start, end) do { } while (0)
|
31 |
|
|
#define flush_cache_page(vma, vmaddr) do { } while (0)
|
32 |
|
|
#define flush_page_to_ram(page) do { } while (0)
|
33 |
|
|
#define flush_dcache_page(page) do { } while (0)
|
34 |
|
|
#define flush_icache_range(start, end) do { } while (0)
|
35 |
|
|
#define flush_icache_page(vma,pg) do { } while (0)
|
36 |
|
|
#define flush_icache_user_range(vma,pg,adr,len) do { } while (0)
|
37 |
|
|
|
38 |
|
|
#define __flush_tlb() \
|
39 |
|
|
do { \
|
40 |
|
|
unsigned int tmpreg; \
|
41 |
|
|
\
|
42 |
|
|
__asm__ __volatile__( \
|
43 |
|
|
"movl %%cr3, %0; # flush TLB \n" \
|
44 |
|
|
"movl %0, %%cr3; \n" \
|
45 |
|
|
: "=r" (tmpreg) \
|
46 |
|
|
:: "memory"); \
|
47 |
|
|
} while (0)
|
48 |
|
|
|
49 |
|
|
/*
|
50 |
|
|
* Global pages have to be flushed a bit differently. Not a real
|
51 |
|
|
* performance problem because this does not happen often.
|
52 |
|
|
*/
|
53 |
|
|
#define __flush_tlb_global() \
|
54 |
|
|
do { \
|
55 |
|
|
unsigned int tmpreg; \
|
56 |
|
|
\
|
57 |
|
|
__asm__ __volatile__( \
|
58 |
|
|
"movl %1, %%cr4; # turn off PGE \n" \
|
59 |
|
|
"movl %%cr3, %0; # flush TLB \n" \
|
60 |
|
|
"movl %0, %%cr3; \n" \
|
61 |
|
|
"movl %2, %%cr4; # turn PGE back on \n" \
|
62 |
|
|
: "=&r" (tmpreg) \
|
63 |
|
|
: "r" (mmu_cr4_features & ~X86_CR4_PGE), \
|
64 |
|
|
"r" (mmu_cr4_features) \
|
65 |
|
|
: "memory"); \
|
66 |
|
|
} while (0)
|
67 |
|
|
|
68 |
|
|
extern unsigned long pgkern_mask;
|
69 |
|
|
|
70 |
|
|
/*
|
71 |
|
|
* Do not check the PGE bit unnecesserily if this is a PPro+ kernel.
|
72 |
|
|
*/
|
73 |
|
|
#ifdef CONFIG_X86_PGE
|
74 |
|
|
# define __flush_tlb_all() __flush_tlb_global()
|
75 |
|
|
#else
|
76 |
|
|
# define __flush_tlb_all() \
|
77 |
|
|
do { \
|
78 |
|
|
if (cpu_has_pge) \
|
79 |
|
|
__flush_tlb_global(); \
|
80 |
|
|
else \
|
81 |
|
|
__flush_tlb(); \
|
82 |
|
|
} while (0)
|
83 |
|
|
#endif
|
84 |
|
|
|
85 |
|
|
#define __flush_tlb_single(addr) \
|
86 |
|
|
__asm__ __volatile__("invlpg %0": :"m" (*(char *) addr))
|
87 |
|
|
|
88 |
|
|
#ifdef CONFIG_X86_INVLPG
|
89 |
|
|
# define __flush_tlb_one(addr) __flush_tlb_single(addr)
|
90 |
|
|
#else
|
91 |
|
|
# define __flush_tlb_one(addr) \
|
92 |
|
|
do { \
|
93 |
|
|
if (cpu_has_pge) \
|
94 |
|
|
__flush_tlb_single(addr); \
|
95 |
|
|
else \
|
96 |
|
|
__flush_tlb(); \
|
97 |
|
|
} while (0)
|
98 |
|
|
#endif
|
99 |
|
|
|
100 |
|
|
/*
|
101 |
|
|
* ZERO_PAGE is a global shared page that is always zero: used
|
102 |
|
|
* for zero-mapped memory areas etc..
|
103 |
|
|
*/
|
104 |
|
|
extern unsigned long empty_zero_page[1024];
|
105 |
|
|
#define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
|
106 |
|
|
|
107 |
|
|
#endif /* !__ASSEMBLY__ */
|
108 |
|
|
|
109 |
|
|
/*
|
110 |
|
|
* The Linux x86 paging architecture is 'compile-time dual-mode', it
|
111 |
|
|
* implements both the traditional 2-level x86 page tables and the
|
112 |
|
|
* newer 3-level PAE-mode page tables.
|
113 |
|
|
*/
|
114 |
|
|
#ifndef __ASSEMBLY__
|
115 |
|
|
#if CONFIG_X86_PAE
|
116 |
|
|
# include <asm/pgtable-3level.h>
|
117 |
|
|
|
118 |
|
|
/*
|
119 |
|
|
* Need to initialise the X86 PAE caches
|
120 |
|
|
*/
|
121 |
|
|
extern void pgtable_cache_init(void);
|
122 |
|
|
|
123 |
|
|
#else
|
124 |
|
|
# include <asm/pgtable-2level.h>
|
125 |
|
|
|
126 |
|
|
/*
|
127 |
|
|
* No page table caches to initialise
|
128 |
|
|
*/
|
129 |
|
|
#define pgtable_cache_init() do { } while (0)
|
130 |
|
|
|
131 |
|
|
#endif
|
132 |
|
|
#endif
|
133 |
|
|
|
134 |
|
|
#define __beep() asm("movb $0x3,%al; outb %al,$0x61")
|
135 |
|
|
|
136 |
|
|
#define PMD_SIZE (1UL << PMD_SHIFT)
|
137 |
|
|
#define PMD_MASK (~(PMD_SIZE-1))
|
138 |
|
|
#define PGDIR_SIZE (1UL << PGDIR_SHIFT)
|
139 |
|
|
#define PGDIR_MASK (~(PGDIR_SIZE-1))
|
140 |
|
|
|
141 |
|
|
#define USER_PTRS_PER_PGD (TASK_SIZE/PGDIR_SIZE)
|
142 |
|
|
#define FIRST_USER_PGD_NR 0
|
143 |
|
|
|
144 |
|
|
#define USER_PGD_PTRS (PAGE_OFFSET >> PGDIR_SHIFT)
|
145 |
|
|
#define KERNEL_PGD_PTRS (PTRS_PER_PGD-USER_PGD_PTRS)
|
146 |
|
|
|
147 |
|
|
#define TWOLEVEL_PGDIR_SHIFT 22
|
148 |
|
|
#define BOOT_USER_PGD_PTRS (__PAGE_OFFSET >> TWOLEVEL_PGDIR_SHIFT)
|
149 |
|
|
#define BOOT_KERNEL_PGD_PTRS (1024-BOOT_USER_PGD_PTRS)
|
150 |
|
|
|
151 |
|
|
|
152 |
|
|
#ifndef __ASSEMBLY__
|
153 |
|
|
/* Just any arbitrary offset to the start of the vmalloc VM area: the
|
154 |
|
|
* current 8MB value just means that there will be a 8MB "hole" after the
|
155 |
|
|
* physical memory until the kernel virtual memory starts. That means that
|
156 |
|
|
* any out-of-bounds memory accesses will hopefully be caught.
|
157 |
|
|
* The vmalloc() routines leaves a hole of 4kB between each vmalloced
|
158 |
|
|
* area for the same reason. ;)
|
159 |
|
|
*/
|
160 |
|
|
#define VMALLOC_OFFSET (8*1024*1024)
|
161 |
|
|
#define VMALLOC_START (((unsigned long) high_memory + 2*VMALLOC_OFFSET-1) & \
|
162 |
|
|
~(VMALLOC_OFFSET-1))
|
163 |
|
|
#define VMALLOC_VMADDR(x) ((unsigned long)(x))
|
164 |
|
|
#if CONFIG_HIGHMEM
|
165 |
|
|
# define VMALLOC_END (PKMAP_BASE-2*PAGE_SIZE)
|
166 |
|
|
#else
|
167 |
|
|
# define VMALLOC_END (FIXADDR_START-2*PAGE_SIZE)
|
168 |
|
|
#endif
|
169 |
|
|
|
170 |
|
|
/*
|
171 |
|
|
* The 4MB page is guessing.. Detailed in the infamous "Chapter H"
|
172 |
|
|
* of the Pentium details, but assuming intel did the straightforward
|
173 |
|
|
* thing, this bit set in the page directory entry just means that
|
174 |
|
|
* the page directory entry points directly to a 4MB-aligned block of
|
175 |
|
|
* memory.
|
176 |
|
|
*/
|
177 |
|
|
#define _PAGE_BIT_PRESENT 0
|
178 |
|
|
#define _PAGE_BIT_RW 1
|
179 |
|
|
#define _PAGE_BIT_USER 2
|
180 |
|
|
#define _PAGE_BIT_PWT 3
|
181 |
|
|
#define _PAGE_BIT_PCD 4
|
182 |
|
|
#define _PAGE_BIT_ACCESSED 5
|
183 |
|
|
#define _PAGE_BIT_DIRTY 6
|
184 |
|
|
#define _PAGE_BIT_PSE 7 /* 4 MB (or 2MB) page, Pentium+, if present.. */
|
185 |
|
|
#define _PAGE_BIT_GLOBAL 8 /* Global TLB entry PPro+ */
|
186 |
|
|
|
187 |
|
|
#define _PAGE_PRESENT 0x001
|
188 |
|
|
#define _PAGE_RW 0x002
|
189 |
|
|
#define _PAGE_USER 0x004
|
190 |
|
|
#define _PAGE_PWT 0x008
|
191 |
|
|
#define _PAGE_PCD 0x010
|
192 |
|
|
#define _PAGE_ACCESSED 0x020
|
193 |
|
|
#define _PAGE_DIRTY 0x040
|
194 |
|
|
#define _PAGE_PSE 0x080 /* 4 MB (or 2MB) page, Pentium+, if present.. */
|
195 |
|
|
#define _PAGE_GLOBAL 0x100 /* Global TLB entry PPro+ */
|
196 |
|
|
|
197 |
|
|
#define _PAGE_PROTNONE 0x080 /* If not present */
|
198 |
|
|
|
199 |
|
|
#define _PAGE_TABLE (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED | _PAGE_DIRTY)
|
200 |
|
|
#define _KERNPG_TABLE (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY)
|
201 |
|
|
#define _PAGE_CHG_MASK (PTE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY)
|
202 |
|
|
|
203 |
|
|
#define PAGE_NONE __pgprot(_PAGE_PROTNONE | _PAGE_ACCESSED)
|
204 |
|
|
#define PAGE_SHARED __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED)
|
205 |
|
|
#define PAGE_COPY __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
|
206 |
|
|
#define PAGE_READONLY __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
|
207 |
|
|
|
208 |
|
|
#define __PAGE_KERNEL \
|
209 |
|
|
(_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED)
|
210 |
|
|
#define __PAGE_KERNEL_NOCACHE \
|
211 |
|
|
(_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_PCD | _PAGE_ACCESSED)
|
212 |
|
|
#define __PAGE_KERNEL_RO \
|
213 |
|
|
(_PAGE_PRESENT | _PAGE_DIRTY | _PAGE_ACCESSED)
|
214 |
|
|
|
215 |
|
|
#ifdef CONFIG_X86_PGE
|
216 |
|
|
# define MAKE_GLOBAL(x) __pgprot((x) | _PAGE_GLOBAL)
|
217 |
|
|
#else
|
218 |
|
|
# define MAKE_GLOBAL(x) \
|
219 |
|
|
({ \
|
220 |
|
|
pgprot_t __ret; \
|
221 |
|
|
\
|
222 |
|
|
if (cpu_has_pge) \
|
223 |
|
|
__ret = __pgprot((x) | _PAGE_GLOBAL); \
|
224 |
|
|
else \
|
225 |
|
|
__ret = __pgprot(x); \
|
226 |
|
|
__ret; \
|
227 |
|
|
})
|
228 |
|
|
#endif
|
229 |
|
|
|
230 |
|
|
#define PAGE_KERNEL MAKE_GLOBAL(__PAGE_KERNEL)
|
231 |
|
|
#define PAGE_KERNEL_RO MAKE_GLOBAL(__PAGE_KERNEL_RO)
|
232 |
|
|
#define PAGE_KERNEL_NOCACHE MAKE_GLOBAL(__PAGE_KERNEL_NOCACHE)
|
233 |
|
|
|
234 |
|
|
/*
|
235 |
|
|
* The i386 can't do page protection for execute, and considers that
|
236 |
|
|
* the same are read. Also, write permissions imply read permissions.
|
237 |
|
|
* This is the closest we can get..
|
238 |
|
|
*/
|
239 |
|
|
#define __P000 PAGE_NONE
|
240 |
|
|
#define __P001 PAGE_READONLY
|
241 |
|
|
#define __P010 PAGE_COPY
|
242 |
|
|
#define __P011 PAGE_COPY
|
243 |
|
|
#define __P100 PAGE_READONLY
|
244 |
|
|
#define __P101 PAGE_READONLY
|
245 |
|
|
#define __P110 PAGE_COPY
|
246 |
|
|
#define __P111 PAGE_COPY
|
247 |
|
|
|
248 |
|
|
#define __S000 PAGE_NONE
|
249 |
|
|
#define __S001 PAGE_READONLY
|
250 |
|
|
#define __S010 PAGE_SHARED
|
251 |
|
|
#define __S011 PAGE_SHARED
|
252 |
|
|
#define __S100 PAGE_READONLY
|
253 |
|
|
#define __S101 PAGE_READONLY
|
254 |
|
|
#define __S110 PAGE_SHARED
|
255 |
|
|
#define __S111 PAGE_SHARED
|
256 |
|
|
|
257 |
|
|
/*
|
258 |
|
|
* Define this if things work differently on an i386 and an i486:
|
259 |
|
|
* it will (on an i486) warn about kernel memory accesses that are
|
260 |
|
|
* done without a 'verify_area(VERIFY_WRITE,..)'
|
261 |
|
|
*/
|
262 |
|
|
#undef TEST_VERIFY_AREA
|
263 |
|
|
|
264 |
|
|
/* page table for 0-4MB for everybody */
|
265 |
|
|
extern unsigned long pg0[1024];
|
266 |
|
|
|
267 |
|
|
#define pte_present(x) ((x).pte_low & (_PAGE_PRESENT | _PAGE_PROTNONE))
|
268 |
|
|
#define pte_clear(xp) do { set_pte(xp, __pte(0)); } while (0)
|
269 |
|
|
|
270 |
|
|
#define pmd_none(x) (!pmd_val(x))
|
271 |
|
|
#define pmd_present(x) (pmd_val(x) & _PAGE_PRESENT)
|
272 |
|
|
#define pmd_clear(xp) do { set_pmd(xp, __pmd(0)); } while (0)
|
273 |
|
|
#define pmd_bad(x) ((pmd_val(x) & (~PAGE_MASK & ~_PAGE_USER)) != _KERNPG_TABLE)
|
274 |
|
|
|
275 |
|
|
|
276 |
|
|
#define pages_to_mb(x) ((x) >> (20-PAGE_SHIFT))
|
277 |
|
|
|
278 |
|
|
/*
|
279 |
|
|
* The following only work if pte_present() is true.
|
280 |
|
|
* Undefined behaviour if not..
|
281 |
|
|
*/
|
282 |
|
|
static inline int pte_read(pte_t pte) { return (pte).pte_low & _PAGE_USER; }
|
283 |
|
|
static inline int pte_exec(pte_t pte) { return (pte).pte_low & _PAGE_USER; }
|
284 |
|
|
static inline int pte_dirty(pte_t pte) { return (pte).pte_low & _PAGE_DIRTY; }
|
285 |
|
|
static inline int pte_young(pte_t pte) { return (pte).pte_low & _PAGE_ACCESSED; }
|
286 |
|
|
static inline int pte_write(pte_t pte) { return (pte).pte_low & _PAGE_RW; }
|
287 |
|
|
|
288 |
|
|
static inline pte_t pte_rdprotect(pte_t pte) { (pte).pte_low &= ~_PAGE_USER; return pte; }
|
289 |
|
|
static inline pte_t pte_exprotect(pte_t pte) { (pte).pte_low &= ~_PAGE_USER; return pte; }
|
290 |
|
|
static inline pte_t pte_mkclean(pte_t pte) { (pte).pte_low &= ~_PAGE_DIRTY; return pte; }
|
291 |
|
|
static inline pte_t pte_mkold(pte_t pte) { (pte).pte_low &= ~_PAGE_ACCESSED; return pte; }
|
292 |
|
|
static inline pte_t pte_wrprotect(pte_t pte) { (pte).pte_low &= ~_PAGE_RW; return pte; }
|
293 |
|
|
static inline pte_t pte_mkread(pte_t pte) { (pte).pte_low |= _PAGE_USER; return pte; }
|
294 |
|
|
static inline pte_t pte_mkexec(pte_t pte) { (pte).pte_low |= _PAGE_USER; return pte; }
|
295 |
|
|
static inline pte_t pte_mkdirty(pte_t pte) { (pte).pte_low |= _PAGE_DIRTY; return pte; }
|
296 |
|
|
static inline pte_t pte_mkyoung(pte_t pte) { (pte).pte_low |= _PAGE_ACCESSED; return pte; }
|
297 |
|
|
static inline pte_t pte_mkwrite(pte_t pte) { (pte).pte_low |= _PAGE_RW; return pte; }
|
298 |
|
|
|
299 |
|
|
static inline int ptep_test_and_clear_dirty(pte_t *ptep) { return test_and_clear_bit(_PAGE_BIT_DIRTY, ptep); }
|
300 |
|
|
static inline int ptep_test_and_clear_young(pte_t *ptep) { return test_and_clear_bit(_PAGE_BIT_ACCESSED, ptep); }
|
301 |
|
|
static inline void ptep_set_wrprotect(pte_t *ptep) { clear_bit(_PAGE_BIT_RW, ptep); }
|
302 |
|
|
static inline void ptep_mkdirty(pte_t *ptep) { set_bit(_PAGE_BIT_DIRTY, ptep); }
|
303 |
|
|
|
304 |
|
|
/*
|
305 |
|
|
* Conversion functions: convert a page and protection to a page entry,
|
306 |
|
|
* and a page entry and page directory to the page they refer to.
|
307 |
|
|
*/
|
308 |
|
|
|
309 |
|
|
#define mk_pte(page, pgprot) __mk_pte((page) - mem_map, (pgprot))
|
310 |
|
|
|
311 |
|
|
/* This takes a physical page address that is used by the remapping functions */
|
312 |
|
|
#define mk_pte_phys(physpage, pgprot) __mk_pte((physpage) >> PAGE_SHIFT, pgprot)
|
313 |
|
|
|
314 |
|
|
static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
|
315 |
|
|
{
|
316 |
|
|
pte.pte_low &= _PAGE_CHG_MASK;
|
317 |
|
|
pte.pte_low |= pgprot_val(newprot);
|
318 |
|
|
return pte;
|
319 |
|
|
}
|
320 |
|
|
|
321 |
|
|
#define page_pte(page) page_pte_prot(page, __pgprot(0))
|
322 |
|
|
|
323 |
|
|
#define pmd_page(pmd) \
|
324 |
|
|
((unsigned long) __va(pmd_val(pmd) & PAGE_MASK))
|
325 |
|
|
|
326 |
|
|
/* to find an entry in a page-table-directory. */
|
327 |
|
|
#define pgd_index(address) ((address >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
|
328 |
|
|
|
329 |
|
|
#define __pgd_offset(address) pgd_index(address)
|
330 |
|
|
|
331 |
|
|
#define pgd_offset(mm, address) ((mm)->pgd+pgd_index(address))
|
332 |
|
|
|
333 |
|
|
/* to find an entry in a kernel page-table-directory */
|
334 |
|
|
#define pgd_offset_k(address) pgd_offset(&init_mm, address)
|
335 |
|
|
|
336 |
|
|
#define __pmd_offset(address) \
|
337 |
|
|
(((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
|
338 |
|
|
|
339 |
|
|
/* Find an entry in the third-level page table.. */
|
340 |
|
|
#define __pte_offset(address) \
|
341 |
|
|
((address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
|
342 |
|
|
#define pte_offset(dir, address) ((pte_t *) pmd_page(*(dir)) + \
|
343 |
|
|
__pte_offset(address))
|
344 |
|
|
|
345 |
|
|
/*
|
346 |
|
|
* The i386 doesn't have any external MMU info: the kernel page
|
347 |
|
|
* tables contain all the necessary information.
|
348 |
|
|
*/
|
349 |
|
|
#define update_mmu_cache(vma,address,pte) do { } while (0)
|
350 |
|
|
|
351 |
|
|
/* Encode and de-code a swap entry */
|
352 |
|
|
#define SWP_TYPE(x) (((x).val >> 1) & 0x3f)
|
353 |
|
|
#define SWP_OFFSET(x) ((x).val >> 8)
|
354 |
|
|
#define SWP_ENTRY(type, offset) ((swp_entry_t) { ((type) << 1) | ((offset) << 8) })
|
355 |
|
|
#define pte_to_swp_entry(pte) ((swp_entry_t) { (pte).pte_low })
|
356 |
|
|
#define swp_entry_to_pte(x) ((pte_t) { (x).val })
|
357 |
|
|
|
358 |
|
|
struct page;
|
359 |
|
|
int change_page_attr(struct page *, int, pgprot_t prot);
|
360 |
|
|
|
361 |
|
|
#endif /* !__ASSEMBLY__ */
|
362 |
|
|
|
363 |
|
|
/* Needs to be defined here and not in linux/mm.h, as it is arch dependent */
|
364 |
|
|
#define PageSkip(page) (0)
|
365 |
|
|
#define kern_addr_valid(addr) (1)
|
366 |
|
|
|
367 |
|
|
#define io_remap_page_range remap_page_range
|
368 |
|
|
|
369 |
|
|
#endif /* _I386_PGTABLE_H */
|