x86: shrink __PAGE_KERNEL/__PAGE_KERNEL_EXEC on non PAE kernels
[linux-block.git] / include / asm-x86 / pgtable.h
CommitLineData
6c386655
JF
1#ifndef _ASM_X86_PGTABLE_H
2#define _ASM_X86_PGTABLE_H
3
4#define USER_PTRS_PER_PGD ((TASK_SIZE-1)/PGDIR_SIZE+1)
5#define FIRST_USER_ADDRESS 0
6
7#define _PAGE_BIT_PRESENT 0
8#define _PAGE_BIT_RW 1
9#define _PAGE_BIT_USER 2
10#define _PAGE_BIT_PWT 3
11#define _PAGE_BIT_PCD 4
12#define _PAGE_BIT_ACCESSED 5
13#define _PAGE_BIT_DIRTY 6
14#define _PAGE_BIT_FILE 6
15#define _PAGE_BIT_PSE 7 /* 4 MB (or 2MB) page */
16#define _PAGE_BIT_GLOBAL 8 /* Global TLB entry PPro+ */
17#define _PAGE_BIT_UNUSED1 9 /* available for programmer */
18#define _PAGE_BIT_UNUSED2 10
19#define _PAGE_BIT_UNUSED3 11
20#define _PAGE_BIT_NX 63 /* No execute: only valid after cpuid check */
21
f2919232
JF
22/*
23 * Note: we use _AC(1, L) instead of _AC(1, UL) so that we get a
24 * sign-extended value on 32-bit with all 1's in the upper word,
25 * which preserves the upper pte values on 64-bit ptes:
26 */
61f38226
IM
27#define _PAGE_PRESENT (_AC(1, L)<<_PAGE_BIT_PRESENT)
28#define _PAGE_RW (_AC(1, L)<<_PAGE_BIT_RW)
29#define _PAGE_USER (_AC(1, L)<<_PAGE_BIT_USER)
30#define _PAGE_PWT (_AC(1, L)<<_PAGE_BIT_PWT)
31#define _PAGE_PCD (_AC(1, L)<<_PAGE_BIT_PCD)
32#define _PAGE_ACCESSED (_AC(1, L)<<_PAGE_BIT_ACCESSED)
33#define _PAGE_DIRTY (_AC(1, L)<<_PAGE_BIT_DIRTY)
34#define _PAGE_PSE (_AC(1, L)<<_PAGE_BIT_PSE) /* 2MB page */
35#define _PAGE_GLOBAL (_AC(1, L)<<_PAGE_BIT_GLOBAL) /* Global TLB entry */
36#define _PAGE_UNUSED1 (_AC(1, L)<<_PAGE_BIT_UNUSED1)
37#define _PAGE_UNUSED2 (_AC(1, L)<<_PAGE_BIT_UNUSED2)
38#define _PAGE_UNUSED3 (_AC(1, L)<<_PAGE_BIT_UNUSED3)
6c386655
JF
39
40#if defined(CONFIG_X86_64) || defined(CONFIG_X86_PAE)
41#define _PAGE_NX (_AC(1, ULL) << _PAGE_BIT_NX)
42#else
43#define _PAGE_NX 0
44#endif
45
46/* If _PAGE_PRESENT is clear, we use these: */
47#define _PAGE_FILE _PAGE_DIRTY /* nonlinear file mapping, saved PTE; unset:swap */
48#define _PAGE_PROTNONE _PAGE_PSE /* if the user mapped it with PROT_NONE;
49 pte_present gives true */
50
51#define _PAGE_TABLE (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED | _PAGE_DIRTY)
52#define _KERNPG_TABLE (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY)
53
54#define _PAGE_CHG_MASK (PTE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY)
55
56#define PAGE_NONE __pgprot(_PAGE_PROTNONE | _PAGE_ACCESSED)
57#define PAGE_SHARED __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
58
59#define PAGE_SHARED_EXEC __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | _PAGE_ACCESSED)
60#define PAGE_COPY_NOEXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
61#define PAGE_COPY_EXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
62#define PAGE_COPY PAGE_COPY_NOEXEC
63#define PAGE_READONLY __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED | _PAGE_NX)
64#define PAGE_READONLY_EXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_ACCESSED)
65
66#ifdef CONFIG_X86_32
67#define _PAGE_KERNEL_EXEC \
68 (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED)
69#define _PAGE_KERNEL (_PAGE_KERNEL_EXEC | _PAGE_NX)
70
71#ifndef __ASSEMBLY__
c93c82bb 72extern pteval_t __PAGE_KERNEL, __PAGE_KERNEL_EXEC;
6c386655
JF
73#endif /* __ASSEMBLY__ */
74#else
75#define __PAGE_KERNEL_EXEC \
76 (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED)
77#define __PAGE_KERNEL (__PAGE_KERNEL_EXEC | _PAGE_NX)
78#endif
79
80#define __PAGE_KERNEL_RO (__PAGE_KERNEL & ~_PAGE_RW)
81#define __PAGE_KERNEL_RX (__PAGE_KERNEL_EXEC & ~_PAGE_RW)
82#define __PAGE_KERNEL_NOCACHE (__PAGE_KERNEL | _PAGE_PCD | _PAGE_PWT)
83#define __PAGE_KERNEL_VSYSCALL (__PAGE_KERNEL_RX | _PAGE_USER)
84#define __PAGE_KERNEL_VSYSCALL_NOCACHE (__PAGE_KERNEL_VSYSCALL | _PAGE_PCD | _PAGE_PWT)
85#define __PAGE_KERNEL_LARGE (__PAGE_KERNEL | _PAGE_PSE)
86#define __PAGE_KERNEL_LARGE_EXEC (__PAGE_KERNEL_EXEC | _PAGE_PSE)
87
88#ifdef CONFIG_X86_32
89# define MAKE_GLOBAL(x) __pgprot((x))
90#else
91# define MAKE_GLOBAL(x) __pgprot((x) | _PAGE_GLOBAL)
92#endif
93
94#define PAGE_KERNEL MAKE_GLOBAL(__PAGE_KERNEL)
95#define PAGE_KERNEL_RO MAKE_GLOBAL(__PAGE_KERNEL_RO)
96#define PAGE_KERNEL_EXEC MAKE_GLOBAL(__PAGE_KERNEL_EXEC)
97#define PAGE_KERNEL_RX MAKE_GLOBAL(__PAGE_KERNEL_RX)
98#define PAGE_KERNEL_NOCACHE MAKE_GLOBAL(__PAGE_KERNEL_NOCACHE)
99#define PAGE_KERNEL_LARGE MAKE_GLOBAL(__PAGE_KERNEL_LARGE)
100#define PAGE_KERNEL_LARGE_EXEC MAKE_GLOBAL(__PAGE_KERNEL_LARGE_EXEC)
101#define PAGE_KERNEL_VSYSCALL MAKE_GLOBAL(__PAGE_KERNEL_VSYSCALL)
102#define PAGE_KERNEL_VSYSCALL_NOCACHE MAKE_GLOBAL(__PAGE_KERNEL_VSYSCALL_NOCACHE)
103
104/* xwr */
105#define __P000 PAGE_NONE
106#define __P001 PAGE_READONLY
107#define __P010 PAGE_COPY
108#define __P011 PAGE_COPY
109#define __P100 PAGE_READONLY_EXEC
110#define __P101 PAGE_READONLY_EXEC
111#define __P110 PAGE_COPY_EXEC
112#define __P111 PAGE_COPY_EXEC
113
114#define __S000 PAGE_NONE
115#define __S001 PAGE_READONLY
116#define __S010 PAGE_SHARED
117#define __S011 PAGE_SHARED
118#define __S100 PAGE_READONLY_EXEC
119#define __S101 PAGE_READONLY_EXEC
120#define __S110 PAGE_SHARED_EXEC
121#define __S111 PAGE_SHARED_EXEC
122
4614139c 123#ifndef __ASSEMBLY__
195466dc 124
8405b122
JF
125/*
126 * ZERO_PAGE is a global shared page that is always zero: used
127 * for zero-mapped memory areas etc..
128 */
129extern unsigned long empty_zero_page[PAGE_SIZE/sizeof(unsigned long)];
130#define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
131
132
4614139c
JF
133/*
134 * The following only work if pte_present() is true.
135 * Undefined behaviour if not..
136 */
137static inline int pte_dirty(pte_t pte) { return pte_val(pte) & _PAGE_DIRTY; }
138static inline int pte_young(pte_t pte) { return pte_val(pte) & _PAGE_ACCESSED; }
139static inline int pte_write(pte_t pte) { return pte_val(pte) & _PAGE_RW; }
140static inline int pte_file(pte_t pte) { return pte_val(pte) & _PAGE_FILE; }
141static inline int pte_huge(pte_t pte) { return pte_val(pte) & _PAGE_PSE; }
a5a5dc31 142static inline int pte_global(pte_t pte) { return pte_val(pte) & _PAGE_GLOBAL; }
4c3c4b45 143static inline int pte_exec(pte_t pte) { return !(pte_val(pte) & _PAGE_NX); }
4614139c
JF
144
145static inline int pmd_large(pmd_t pte) {
146 return (pmd_val(pte) & (_PAGE_PSE|_PAGE_PRESENT)) ==
147 (_PAGE_PSE|_PAGE_PRESENT);
148}
149
150static inline pte_t pte_mkclean(pte_t pte) { return __pte(pte_val(pte) & ~_PAGE_DIRTY); }
151static inline pte_t pte_mkold(pte_t pte) { return __pte(pte_val(pte) & ~_PAGE_ACCESSED); }
152static inline pte_t pte_wrprotect(pte_t pte) { return __pte(pte_val(pte) & ~_PAGE_RW); }
153static inline pte_t pte_mkexec(pte_t pte) { return __pte(pte_val(pte) & ~_PAGE_NX); }
154static inline pte_t pte_mkdirty(pte_t pte) { return __pte(pte_val(pte) | _PAGE_DIRTY); }
155static inline pte_t pte_mkyoung(pte_t pte) { return __pte(pte_val(pte) | _PAGE_ACCESSED); }
156static inline pte_t pte_mkwrite(pte_t pte) { return __pte(pte_val(pte) | _PAGE_RW); }
157static inline pte_t pte_mkhuge(pte_t pte) { return __pte(pte_val(pte) | _PAGE_PSE); }
158static inline pte_t pte_clrhuge(pte_t pte) { return __pte(pte_val(pte) & ~_PAGE_PSE); }
a5a5dc31
AK
159static inline pte_t pte_mkglobal(pte_t pte) { return __pte(pte_val(pte) | _PAGE_GLOBAL); }
160static inline pte_t pte_clrglobal(pte_t pte) { return __pte(pte_val(pte) & ~_PAGE_GLOBAL); }
4614139c 161
6fdc05d4
JF
162extern pteval_t __supported_pte_mask;
163
164static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
165{
166 return __pte((((phys_addr_t)page_nr << PAGE_SHIFT) |
167 pgprot_val(pgprot)) & __supported_pte_mask);
168}
169
170static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
171{
172 return __pmd((((phys_addr_t)page_nr << PAGE_SHIFT) |
173 pgprot_val(pgprot)) & __supported_pte_mask);
174}
175
38472311
IM
176static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
177{
178 pteval_t val = pte_val(pte);
179
180 /*
181 * Chop off the NX bit (if present), and add the NX portion of
182 * the newprot (if present):
183 */
184 val &= _PAGE_CHG_MASK & ~_PAGE_NX;
185 val |= pgprot_val(newprot) & __supported_pte_mask;
186
187 return __pte(val);
188}
189
4891645e
JF
190#ifdef CONFIG_PARAVIRT
191#include <asm/paravirt.h>
192#else /* !CONFIG_PARAVIRT */
193#define set_pte(ptep, pte) native_set_pte(ptep, pte)
194#define set_pte_at(mm, addr, ptep, pte) native_set_pte_at(mm, addr, ptep, pte)
195
196#define set_pte_present(mm, addr, ptep, pte) \
197 native_set_pte_present(mm, addr, ptep, pte)
198#define set_pte_atomic(ptep, pte) \
199 native_set_pte_atomic(ptep, pte)
200
201#define set_pmd(pmdp, pmd) native_set_pmd(pmdp, pmd)
202
203#ifndef __PAGETABLE_PUD_FOLDED
204#define set_pgd(pgdp, pgd) native_set_pgd(pgdp, pgd)
205#define pgd_clear(pgd) native_pgd_clear(pgd)
206#endif
207
208#ifndef set_pud
209# define set_pud(pudp, pud) native_set_pud(pudp, pud)
210#endif
211
212#ifndef __PAGETABLE_PMD_FOLDED
213#define pud_clear(pud) native_pud_clear(pud)
214#endif
215
216#define pte_clear(mm, addr, ptep) native_pte_clear(mm, addr, ptep)
217#define pmd_clear(pmd) native_pmd_clear(pmd)
218
219#define pte_update(mm, addr, ptep) do { } while (0)
220#define pte_update_defer(mm, addr, ptep) do { } while (0)
221#endif /* CONFIG_PARAVIRT */
222
4614139c
JF
223#endif /* __ASSEMBLY__ */
224
96a388de
TG
225#ifdef CONFIG_X86_32
226# include "pgtable_32.h"
227#else
228# include "pgtable_64.h"
229#endif
6c386655 230
195466dc
JF
231#ifndef __ASSEMBLY__
232
4891645e
JF
233/* local pte updates need not use xchg for locking */
234static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
235{
236 pte_t res = *ptep;
237
238 /* Pure native function needs no input for mm, addr */
239 native_pte_clear(NULL, 0, ptep);
240 return res;
241}
242
243static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
244 pte_t *ptep , pte_t pte)
245{
246 native_set_pte(ptep, pte);
247}
248
195466dc
JF
249#ifndef CONFIG_PARAVIRT
250/*
251 * Rules for using pte_update - it must be called after any PTE update which
252 * has not been done using the set_pte / clear_pte interfaces. It is used by
253 * shadow mode hypervisors to resynchronize the shadow page tables. Kernel PTE
254 * updates should either be sets, clears, or set_pte_atomic for P->P
255 * transitions, which means this hook should only be called for user PTEs.
256 * This hook implies a P->P protection or access change has taken place, which
257 * requires a subsequent TLB flush. The notification can optionally be delayed
258 * until the TLB flush event by using the pte_update_defer form of the
259 * interface, but care must be taken to assure that the flush happens while
260 * still holding the same page table lock so that the shadow and primary pages
261 * do not become out of sync on SMP.
262 */
263#define pte_update(mm, addr, ptep) do { } while (0)
264#define pte_update_defer(mm, addr, ptep) do { } while (0)
265#endif
266
195466dc
JF
267/*
268 * We only update the dirty/accessed state if we set
269 * the dirty bit by hand in the kernel, since the hardware
270 * will do the accessed bit for us, and we don't want to
271 * race with other CPU's that might be updating the dirty
272 * bit at the same time.
273 */
274#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
275#define ptep_set_access_flags(vma, address, ptep, entry, dirty) \
276({ \
277 int __changed = !pte_same(*(ptep), entry); \
278 if (__changed && dirty) { \
279 *ptep = entry; \
280 pte_update_defer((vma)->vm_mm, (address), (ptep)); \
281 flush_tlb_page(vma, address); \
282 } \
283 __changed; \
284})
285
286#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
287#define ptep_test_and_clear_young(vma, addr, ptep) ({ \
288 int __ret = 0; \
289 if (pte_young(*(ptep))) \
290 __ret = test_and_clear_bit(_PAGE_BIT_ACCESSED, \
291 &(ptep)->pte); \
292 if (__ret) \
293 pte_update((vma)->vm_mm, addr, ptep); \
294 __ret; \
295})
296
297#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
298#define ptep_clear_flush_young(vma, address, ptep) \
299({ \
300 int __young; \
301 __young = ptep_test_and_clear_young((vma), (address), (ptep)); \
302 if (__young) \
303 flush_tlb_page(vma, address); \
304 __young; \
305})
306
307#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
308static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
309{
310 pte_t pte = native_ptep_get_and_clear(ptep);
311 pte_update(mm, addr, ptep);
312 return pte;
313}
314
315#define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
316static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm, unsigned long addr, pte_t *ptep, int full)
317{
318 pte_t pte;
319 if (full) {
320 /*
321 * Full address destruction in progress; paravirt does not
322 * care about updates and native needs no locking
323 */
324 pte = native_local_ptep_get_and_clear(ptep);
325 } else {
326 pte = ptep_get_and_clear(mm, addr, ptep);
327 }
328 return pte;
329}
330
331#define __HAVE_ARCH_PTEP_SET_WRPROTECT
332static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
333{
d8d89827 334 clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
195466dc
JF
335 pte_update(mm, addr, ptep);
336}
337
195466dc
JF
338#include <asm-generic/pgtable.h>
339#endif /* __ASSEMBLY__ */
340
6c386655 341#endif /* _ASM_X86_PGTABLE_H */