fs: use kmem_cache_zalloc instead
[pv_ops_mirror.git] / include / asm-arm / pgtable.h
blobd2e8171d1d4e9eb93f5f107c0cca421593aa7bca
1 /*
2 * linux/include/asm-arm/pgtable.h
4 * Copyright (C) 1995-2002 Russell King
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
9 */
10 #ifndef _ASMARM_PGTABLE_H
11 #define _ASMARM_PGTABLE_H
13 #include <asm-generic/4level-fixup.h>
14 #include <asm/proc-fns.h>
16 #ifndef CONFIG_MMU
18 #include "pgtable-nommu.h"
20 #else
22 #include <asm/memory.h>
23 #include <asm/arch/vmalloc.h>
24 #include <asm/pgtable-hwdef.h>
27 * Just any arbitrary offset to the start of the vmalloc VM area: the
28 * current 8MB value just means that there will be a 8MB "hole" after the
29 * physical memory until the kernel virtual memory starts. That means that
30 * any out-of-bounds memory accesses will hopefully be caught.
31 * The vmalloc() routines leaves a hole of 4kB between each vmalloced
32 * area for the same reason. ;)
34 * Note that platforms may override VMALLOC_START, but they must provide
35 * VMALLOC_END. VMALLOC_END defines the (exclusive) limit of this space,
36 * which may not overlap IO space.
38 #ifndef VMALLOC_START
39 #define VMALLOC_OFFSET (8*1024*1024)
40 #define VMALLOC_START (((unsigned long)high_memory + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1))
41 #endif
44 * Hardware-wise, we have a two level page table structure, where the first
45 * level has 4096 entries, and the second level has 256 entries. Each entry
46 * is one 32-bit word. Most of the bits in the second level entry are used
47 * by hardware, and there aren't any "accessed" and "dirty" bits.
49 * Linux on the other hand has a three level page table structure, which can
50 * be wrapped to fit a two level page table structure easily - using the PGD
51 * and PTE only. However, Linux also expects one "PTE" table per page, and
52 * at least a "dirty" bit.
54 * Therefore, we tweak the implementation slightly - we tell Linux that we
55 * have 2048 entries in the first level, each of which is 8 bytes (iow, two
56 * hardware pointers to the second level.) The second level contains two
57 * hardware PTE tables arranged contiguously, followed by Linux versions
58 * which contain the state information Linux needs. We, therefore, end up
59 * with 512 entries in the "PTE" level.
61 * This leads to the page tables having the following layout:
63 * pgd pte
64 * | |
65 * +--------+ +0
66 * | |-----> +------------+ +0
67 * +- - - - + +4 | h/w pt 0 |
68 * | |-----> +------------+ +1024
69 * +--------+ +8 | h/w pt 1 |
70 * | | +------------+ +2048
71 * +- - - - + | Linux pt 0 |
72 * | | +------------+ +3072
73 * +--------+ | Linux pt 1 |
74 * | | +------------+ +4096
76 * See L_PTE_xxx below for definitions of bits in the "Linux pt", and
77 * PTE_xxx for definitions of bits appearing in the "h/w pt".
79 * PMD_xxx definitions refer to bits in the first level page table.
81 * The "dirty" bit is emulated by only granting hardware write permission
82 * iff the page is marked "writable" and "dirty" in the Linux PTE. This
83 * means that a write to a clean page will cause a permission fault, and
84 * the Linux MM layer will mark the page dirty via handle_pte_fault().
85 * For the hardware to notice the permission change, the TLB entry must
86 * be flushed, and ptep_set_access_flags() does that for us.
88 * The "accessed" or "young" bit is emulated by a similar method; we only
89 * allow accesses to the page if the "young" bit is set. Accesses to the
90 * page will cause a fault, and handle_pte_fault() will set the young bit
91 * for us as long as the page is marked present in the corresponding Linux
92 * PTE entry. Again, ptep_set_access_flags() will ensure that the TLB is
93 * up to date.
95 * However, when the "young" bit is cleared, we deny access to the page
96 * by clearing the hardware PTE. Currently Linux does not flush the TLB
97 * for us in this case, which means the TLB will retain the transation
98 * until either the TLB entry is evicted under pressure, or a context
99 * switch which changes the user space mapping occurs.
101 #define PTRS_PER_PTE 512
102 #define PTRS_PER_PMD 1
103 #define PTRS_PER_PGD 2048
106 * PMD_SHIFT determines the size of the area a second-level page table can map
107 * PGDIR_SHIFT determines what a third-level page table entry can map
109 #define PMD_SHIFT 21
110 #define PGDIR_SHIFT 21
112 #define LIBRARY_TEXT_START 0x0c000000
114 #ifndef __ASSEMBLY__
115 extern void __pte_error(const char *file, int line, unsigned long val);
116 extern void __pmd_error(const char *file, int line, unsigned long val);
117 extern void __pgd_error(const char *file, int line, unsigned long val);
119 #define pte_ERROR(pte) __pte_error(__FILE__, __LINE__, pte_val(pte))
120 #define pmd_ERROR(pmd) __pmd_error(__FILE__, __LINE__, pmd_val(pmd))
121 #define pgd_ERROR(pgd) __pgd_error(__FILE__, __LINE__, pgd_val(pgd))
122 #endif /* !__ASSEMBLY__ */
124 #define PMD_SIZE (1UL << PMD_SHIFT)
125 #define PMD_MASK (~(PMD_SIZE-1))
126 #define PGDIR_SIZE (1UL << PGDIR_SHIFT)
127 #define PGDIR_MASK (~(PGDIR_SIZE-1))
130 * This is the lowest virtual address we can permit any user space
131 * mapping to be mapped at. This is particularly important for
132 * non-high vector CPUs.
134 #define FIRST_USER_ADDRESS PAGE_SIZE
136 #define FIRST_USER_PGD_NR 1
137 #define USER_PTRS_PER_PGD ((TASK_SIZE/PGDIR_SIZE) - FIRST_USER_PGD_NR)
140 * section address mask and size definitions.
142 #define SECTION_SHIFT 20
143 #define SECTION_SIZE (1UL << SECTION_SHIFT)
144 #define SECTION_MASK (~(SECTION_SIZE-1))
147 * ARMv6 supersection address mask and size definitions.
149 #define SUPERSECTION_SHIFT 24
150 #define SUPERSECTION_SIZE (1UL << SUPERSECTION_SHIFT)
151 #define SUPERSECTION_MASK (~(SUPERSECTION_SIZE-1))
154 * "Linux" PTE definitions.
156 * We keep two sets of PTEs - the hardware and the linux version.
157 * This allows greater flexibility in the way we map the Linux bits
158 * onto the hardware tables, and allows us to have YOUNG and DIRTY
159 * bits.
161 * The PTE table pointer refers to the hardware entries; the "Linux"
162 * entries are stored 1024 bytes below.
164 #define L_PTE_PRESENT (1 << 0)
165 #define L_PTE_FILE (1 << 1) /* only when !PRESENT */
166 #define L_PTE_YOUNG (1 << 1)
167 #define L_PTE_BUFFERABLE (1 << 2) /* matches PTE */
168 #define L_PTE_CACHEABLE (1 << 3) /* matches PTE */
169 #define L_PTE_USER (1 << 4)
170 #define L_PTE_WRITE (1 << 5)
171 #define L_PTE_EXEC (1 << 6)
172 #define L_PTE_DIRTY (1 << 7)
173 #define L_PTE_SHARED (1 << 10) /* shared(v6), coherent(xsc3) */
175 #ifndef __ASSEMBLY__
178 * The pgprot_* and protection_map entries will be fixed up in runtime
179 * to include the cachable and bufferable bits based on memory policy,
180 * as well as any architecture dependent bits like global/ASID and SMP
181 * shared mapping bits.
183 #define _L_PTE_DEFAULT L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_CACHEABLE | L_PTE_BUFFERABLE
184 #define _L_PTE_READ L_PTE_USER | L_PTE_EXEC
186 extern pgprot_t pgprot_user;
187 extern pgprot_t pgprot_kernel;
189 #define PAGE_NONE pgprot_user
190 #define PAGE_COPY __pgprot(pgprot_val(pgprot_user) | _L_PTE_READ)
191 #define PAGE_SHARED __pgprot(pgprot_val(pgprot_user) | _L_PTE_READ | \
192 L_PTE_WRITE)
193 #define PAGE_READONLY __pgprot(pgprot_val(pgprot_user) | _L_PTE_READ)
194 #define PAGE_KERNEL pgprot_kernel
196 #define __PAGE_NONE __pgprot(_L_PTE_DEFAULT)
197 #define __PAGE_COPY __pgprot(_L_PTE_DEFAULT | _L_PTE_READ)
198 #define __PAGE_SHARED __pgprot(_L_PTE_DEFAULT | _L_PTE_READ | L_PTE_WRITE)
199 #define __PAGE_READONLY __pgprot(_L_PTE_DEFAULT | _L_PTE_READ)
201 #endif /* __ASSEMBLY__ */
204 * The table below defines the page protection levels that we insert into our
205 * Linux page table version. These get translated into the best that the
206 * architecture can perform. Note that on most ARM hardware:
207 * 1) We cannot do execute protection
208 * 2) If we could do execute protection, then read is implied
209 * 3) write implies read permissions
211 #define __P000 __PAGE_NONE
212 #define __P001 __PAGE_READONLY
213 #define __P010 __PAGE_COPY
214 #define __P011 __PAGE_COPY
215 #define __P100 __PAGE_READONLY
216 #define __P101 __PAGE_READONLY
217 #define __P110 __PAGE_COPY
218 #define __P111 __PAGE_COPY
220 #define __S000 __PAGE_NONE
221 #define __S001 __PAGE_READONLY
222 #define __S010 __PAGE_SHARED
223 #define __S011 __PAGE_SHARED
224 #define __S100 __PAGE_READONLY
225 #define __S101 __PAGE_READONLY
226 #define __S110 __PAGE_SHARED
227 #define __S111 __PAGE_SHARED
229 #ifndef __ASSEMBLY__
231 * ZERO_PAGE is a global shared page that is always zero: used
232 * for zero-mapped memory areas etc..
234 extern struct page *empty_zero_page;
235 #define ZERO_PAGE(vaddr) (empty_zero_page)
237 #define pte_pfn(pte) (pte_val(pte) >> PAGE_SHIFT)
238 #define pfn_pte(pfn,prot) (__pte(((pfn) << PAGE_SHIFT) | pgprot_val(prot)))
240 #define pte_none(pte) (!pte_val(pte))
241 #define pte_clear(mm,addr,ptep) set_pte_ext(ptep, __pte(0), 0)
242 #define pte_page(pte) (pfn_to_page(pte_pfn(pte)))
243 #define pte_offset_kernel(dir,addr) (pmd_page_vaddr(*(dir)) + __pte_index(addr))
244 #define pte_offset_map(dir,addr) (pmd_page_vaddr(*(dir)) + __pte_index(addr))
245 #define pte_offset_map_nested(dir,addr) (pmd_page_vaddr(*(dir)) + __pte_index(addr))
246 #define pte_unmap(pte) do { } while (0)
247 #define pte_unmap_nested(pte) do { } while (0)
249 #define set_pte_ext(ptep,pte,ext) cpu_set_pte_ext(ptep,pte,ext)
251 #define set_pte_at(mm,addr,ptep,pteval) do { \
252 set_pte_ext(ptep, pteval, (addr) >= PAGE_OFFSET ? 0 : PTE_EXT_NG); \
253 } while (0)
256 * The following only work if pte_present() is true.
257 * Undefined behaviour if not..
259 #define pte_present(pte) (pte_val(pte) & L_PTE_PRESENT)
260 #define pte_write(pte) (pte_val(pte) & L_PTE_WRITE)
261 #define pte_dirty(pte) (pte_val(pte) & L_PTE_DIRTY)
262 #define pte_young(pte) (pte_val(pte) & L_PTE_YOUNG)
265 * The following only works if pte_present() is not true.
267 #define pte_file(pte) (pte_val(pte) & L_PTE_FILE)
268 #define pte_to_pgoff(x) (pte_val(x) >> 2)
269 #define pgoff_to_pte(x) __pte(((x) << 2) | L_PTE_FILE)
271 #define PTE_FILE_MAX_BITS 30
273 #define PTE_BIT_FUNC(fn,op) \
274 static inline pte_t pte_##fn(pte_t pte) { pte_val(pte) op; return pte; }
276 PTE_BIT_FUNC(wrprotect, &= ~L_PTE_WRITE);
277 PTE_BIT_FUNC(mkwrite, |= L_PTE_WRITE);
278 PTE_BIT_FUNC(mkclean, &= ~L_PTE_DIRTY);
279 PTE_BIT_FUNC(mkdirty, |= L_PTE_DIRTY);
280 PTE_BIT_FUNC(mkold, &= ~L_PTE_YOUNG);
281 PTE_BIT_FUNC(mkyoung, |= L_PTE_YOUNG);
284 * Mark the prot value as uncacheable and unbufferable.
286 #define pgprot_noncached(prot) __pgprot(pgprot_val(prot) & ~(L_PTE_CACHEABLE | L_PTE_BUFFERABLE))
287 #define pgprot_writecombine(prot) __pgprot(pgprot_val(prot) & ~L_PTE_CACHEABLE)
289 #define pmd_none(pmd) (!pmd_val(pmd))
290 #define pmd_present(pmd) (pmd_val(pmd))
291 #define pmd_bad(pmd) (pmd_val(pmd) & 2)
293 #define copy_pmd(pmdpd,pmdps) \
294 do { \
295 pmdpd[0] = pmdps[0]; \
296 pmdpd[1] = pmdps[1]; \
297 flush_pmd_entry(pmdpd); \
298 } while (0)
300 #define pmd_clear(pmdp) \
301 do { \
302 pmdp[0] = __pmd(0); \
303 pmdp[1] = __pmd(0); \
304 clean_pmd_entry(pmdp); \
305 } while (0)
307 static inline pte_t *pmd_page_vaddr(pmd_t pmd)
309 unsigned long ptr;
311 ptr = pmd_val(pmd) & ~(PTRS_PER_PTE * sizeof(void *) - 1);
312 ptr += PTRS_PER_PTE * sizeof(void *);
314 return __va(ptr);
317 #define pmd_page(pmd) virt_to_page(__va(pmd_val(pmd)))
320 * Permanent address of a page. We never have highmem, so this is trivial.
322 #define pages_to_mb(x) ((x) >> (20 - PAGE_SHIFT))
325 * Conversion functions: convert a page and protection to a page entry,
326 * and a page entry and page directory to the page they refer to.
328 #define mk_pte(page,prot) pfn_pte(page_to_pfn(page),prot)
331 * The "pgd_xxx()" functions here are trivial for a folded two-level
332 * setup: the pgd is never bad, and a pmd always exists (as it's folded
333 * into the pgd entry)
335 #define pgd_none(pgd) (0)
336 #define pgd_bad(pgd) (0)
337 #define pgd_present(pgd) (1)
338 #define pgd_clear(pgdp) do { } while (0)
339 #define set_pgd(pgd,pgdp) do { } while (0)
341 /* to find an entry in a page-table-directory */
342 #define pgd_index(addr) ((addr) >> PGDIR_SHIFT)
344 #define pgd_offset(mm, addr) ((mm)->pgd+pgd_index(addr))
346 /* to find an entry in a kernel page-table-directory */
347 #define pgd_offset_k(addr) pgd_offset(&init_mm, addr)
349 /* Find an entry in the second-level page table.. */
350 #define pmd_offset(dir, addr) ((pmd_t *)(dir))
352 /* Find an entry in the third-level page table.. */
353 #define __pte_index(addr) (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
355 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
357 const unsigned long mask = L_PTE_EXEC | L_PTE_WRITE | L_PTE_USER;
358 pte_val(pte) = (pte_val(pte) & ~mask) | (pgprot_val(newprot) & mask);
359 return pte;
362 extern pgd_t swapper_pg_dir[PTRS_PER_PGD];
364 /* Encode and decode a swap entry.
366 * We support up to 32GB of swap on 4k machines
368 #define __swp_type(x) (((x).val >> 2) & 0x7f)
369 #define __swp_offset(x) ((x).val >> 9)
370 #define __swp_entry(type,offset) ((swp_entry_t) { ((type) << 2) | ((offset) << 9) })
371 #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
372 #define __swp_entry_to_pte(swp) ((pte_t) { (swp).val })
374 /* Needs to be defined here and not in linux/mm.h, as it is arch dependent */
375 /* FIXME: this is not correct */
376 #define kern_addr_valid(addr) (1)
378 #include <asm-generic/pgtable.h>
381 * We provide our own arch_get_unmapped_area to cope with VIPT caches.
383 #define HAVE_ARCH_UNMAPPED_AREA
386 * remap a physical page `pfn' of size `size' with page protection `prot'
387 * into virtual address `from'
389 #define io_remap_pfn_range(vma,from,pfn,size,prot) \
390 remap_pfn_range(vma, from, pfn, size, prot)
392 #define pgtable_cache_init() do { } while (0)
394 #endif /* !__ASSEMBLY__ */
396 #endif /* CONFIG_MMU */
398 #endif /* _ASMARM_PGTABLE_H */