Merge branch 'fixes' of git://git.linux-nfs.org/pub/linux/nfs-2.6
[linux-2.6/suspend2-2.6.18.git] / include / asm-arm / pgtable.h
blob8d3919c6458c6c9326ef35a35ee7c848d34ac467
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>
26 * Just any arbitrary offset to the start of the vmalloc VM area: the
27 * current 8MB value just means that there will be a 8MB "hole" after the
28 * physical memory until the kernel virtual memory starts. That means that
29 * any out-of-bounds memory accesses will hopefully be caught.
30 * The vmalloc() routines leaves a hole of 4kB between each vmalloced
31 * area for the same reason. ;)
33 * Note that platforms may override VMALLOC_START, but they must provide
34 * VMALLOC_END. VMALLOC_END defines the (exclusive) limit of this space,
35 * which may not overlap IO space.
37 #ifndef VMALLOC_START
38 #define VMALLOC_OFFSET (8*1024*1024)
39 #define VMALLOC_START (((unsigned long)high_memory + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1))
40 #endif
43 * Hardware-wise, we have a two level page table structure, where the first
44 * level has 4096 entries, and the second level has 256 entries. Each entry
45 * is one 32-bit word. Most of the bits in the second level entry are used
46 * by hardware, and there aren't any "accessed" and "dirty" bits.
48 * Linux on the other hand has a three level page table structure, which can
49 * be wrapped to fit a two level page table structure easily - using the PGD
50 * and PTE only. However, Linux also expects one "PTE" table per page, and
51 * at least a "dirty" bit.
53 * Therefore, we tweak the implementation slightly - we tell Linux that we
54 * have 2048 entries in the first level, each of which is 8 bytes (iow, two
55 * hardware pointers to the second level.) The second level contains two
56 * hardware PTE tables arranged contiguously, followed by Linux versions
57 * which contain the state information Linux needs. We, therefore, end up
58 * with 512 entries in the "PTE" level.
60 * This leads to the page tables having the following layout:
62 * pgd pte
63 * | |
64 * +--------+ +0
65 * | |-----> +------------+ +0
66 * +- - - - + +4 | h/w pt 0 |
67 * | |-----> +------------+ +1024
68 * +--------+ +8 | h/w pt 1 |
69 * | | +------------+ +2048
70 * +- - - - + | Linux pt 0 |
71 * | | +------------+ +3072
72 * +--------+ | Linux pt 1 |
73 * | | +------------+ +4096
75 * See L_PTE_xxx below for definitions of bits in the "Linux pt", and
76 * PTE_xxx for definitions of bits appearing in the "h/w pt".
78 * PMD_xxx definitions refer to bits in the first level page table.
80 * The "dirty" bit is emulated by only granting hardware write permission
81 * iff the page is marked "writable" and "dirty" in the Linux PTE. This
82 * means that a write to a clean page will cause a permission fault, and
83 * the Linux MM layer will mark the page dirty via handle_pte_fault().
84 * For the hardware to notice the permission change, the TLB entry must
85 * be flushed, and ptep_establish() does that for us.
87 * The "accessed" or "young" bit is emulated by a similar method; we only
88 * allow accesses to the page if the "young" bit is set. Accesses to the
89 * page will cause a fault, and handle_pte_fault() will set the young bit
90 * for us as long as the page is marked present in the corresponding Linux
91 * PTE entry. Again, ptep_establish() will ensure that the TLB is up to
92 * date.
94 * However, when the "young" bit is cleared, we deny access to the page
95 * by clearing the hardware PTE. Currently Linux does not flush the TLB
96 * for us in this case, which means the TLB will retain the transation
97 * until either the TLB entry is evicted under pressure, or a context
98 * switch which changes the user space mapping occurs.
100 #define PTRS_PER_PTE 512
101 #define PTRS_PER_PMD 1
102 #define PTRS_PER_PGD 2048
105 * PMD_SHIFT determines the size of the area a second-level page table can map
106 * PGDIR_SHIFT determines what a third-level page table entry can map
108 #define PMD_SHIFT 21
109 #define PGDIR_SHIFT 21
111 #define LIBRARY_TEXT_START 0x0c000000
113 #ifndef __ASSEMBLY__
114 extern void __pte_error(const char *file, int line, unsigned long val);
115 extern void __pmd_error(const char *file, int line, unsigned long val);
116 extern void __pgd_error(const char *file, int line, unsigned long val);
118 #define pte_ERROR(pte) __pte_error(__FILE__, __LINE__, pte_val(pte))
119 #define pmd_ERROR(pmd) __pmd_error(__FILE__, __LINE__, pmd_val(pmd))
120 #define pgd_ERROR(pgd) __pgd_error(__FILE__, __LINE__, pgd_val(pgd))
121 #endif /* !__ASSEMBLY__ */
123 #define PMD_SIZE (1UL << PMD_SHIFT)
124 #define PMD_MASK (~(PMD_SIZE-1))
125 #define PGDIR_SIZE (1UL << PGDIR_SHIFT)
126 #define PGDIR_MASK (~(PGDIR_SIZE-1))
129 * This is the lowest virtual address we can permit any user space
130 * mapping to be mapped at. This is particularly important for
131 * non-high vector CPUs.
133 #define FIRST_USER_ADDRESS PAGE_SIZE
135 #define FIRST_USER_PGD_NR 1
136 #define USER_PTRS_PER_PGD ((TASK_SIZE/PGDIR_SIZE) - FIRST_USER_PGD_NR)
139 * ARMv6 supersection address mask and size definitions.
141 #define SUPERSECTION_SHIFT 24
142 #define SUPERSECTION_SIZE (1UL << SUPERSECTION_SHIFT)
143 #define SUPERSECTION_MASK (~(SUPERSECTION_SIZE-1))
146 * "Linux" PTE definitions.
148 * We keep two sets of PTEs - the hardware and the linux version.
149 * This allows greater flexibility in the way we map the Linux bits
150 * onto the hardware tables, and allows us to have YOUNG and DIRTY
151 * bits.
153 * The PTE table pointer refers to the hardware entries; the "Linux"
154 * entries are stored 1024 bytes below.
156 #define L_PTE_PRESENT (1 << 0)
157 #define L_PTE_FILE (1 << 1) /* only when !PRESENT */
158 #define L_PTE_YOUNG (1 << 1)
159 #define L_PTE_BUFFERABLE (1 << 2) /* matches PTE */
160 #define L_PTE_CACHEABLE (1 << 3) /* matches PTE */
161 #define L_PTE_USER (1 << 4)
162 #define L_PTE_WRITE (1 << 5)
163 #define L_PTE_EXEC (1 << 6)
164 #define L_PTE_DIRTY (1 << 7)
165 #define L_PTE_COHERENT (1 << 9) /* I/O coherent (xsc3) */
166 #define L_PTE_SHARED (1 << 10) /* shared between CPUs (v6) */
167 #define L_PTE_ASID (1 << 11) /* non-global (use ASID, v6) */
169 #ifndef __ASSEMBLY__
172 * The following macros handle the cache and bufferable bits...
174 #define _L_PTE_DEFAULT L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_CACHEABLE | L_PTE_BUFFERABLE
175 #define _L_PTE_READ L_PTE_USER | L_PTE_EXEC
177 extern pgprot_t pgprot_kernel;
179 #define PAGE_NONE __pgprot(_L_PTE_DEFAULT)
180 #define PAGE_COPY __pgprot(_L_PTE_DEFAULT | _L_PTE_READ)
181 #define PAGE_SHARED __pgprot(_L_PTE_DEFAULT | _L_PTE_READ | L_PTE_WRITE)
182 #define PAGE_READONLY __pgprot(_L_PTE_DEFAULT | _L_PTE_READ)
183 #define PAGE_KERNEL pgprot_kernel
185 #endif /* __ASSEMBLY__ */
188 * The table below defines the page protection levels that we insert into our
189 * Linux page table version. These get translated into the best that the
190 * architecture can perform. Note that on most ARM hardware:
191 * 1) We cannot do execute protection
192 * 2) If we could do execute protection, then read is implied
193 * 3) write implies read permissions
195 #define __P000 PAGE_NONE
196 #define __P001 PAGE_READONLY
197 #define __P010 PAGE_COPY
198 #define __P011 PAGE_COPY
199 #define __P100 PAGE_READONLY
200 #define __P101 PAGE_READONLY
201 #define __P110 PAGE_COPY
202 #define __P111 PAGE_COPY
204 #define __S000 PAGE_NONE
205 #define __S001 PAGE_READONLY
206 #define __S010 PAGE_SHARED
207 #define __S011 PAGE_SHARED
208 #define __S100 PAGE_READONLY
209 #define __S101 PAGE_READONLY
210 #define __S110 PAGE_SHARED
211 #define __S111 PAGE_SHARED
213 #ifndef __ASSEMBLY__
215 * ZERO_PAGE is a global shared page that is always zero: used
216 * for zero-mapped memory areas etc..
218 extern struct page *empty_zero_page;
219 #define ZERO_PAGE(vaddr) (empty_zero_page)
221 #define pte_pfn(pte) (pte_val(pte) >> PAGE_SHIFT)
222 #define pfn_pte(pfn,prot) (__pte(((pfn) << PAGE_SHIFT) | pgprot_val(prot)))
224 #define pte_none(pte) (!pte_val(pte))
225 #define pte_clear(mm,addr,ptep) set_pte_at((mm),(addr),(ptep), __pte(0))
226 #define pte_page(pte) (pfn_to_page(pte_pfn(pte)))
227 #define pte_offset_kernel(dir,addr) (pmd_page_kernel(*(dir)) + __pte_index(addr))
228 #define pte_offset_map(dir,addr) (pmd_page_kernel(*(dir)) + __pte_index(addr))
229 #define pte_offset_map_nested(dir,addr) (pmd_page_kernel(*(dir)) + __pte_index(addr))
230 #define pte_unmap(pte) do { } while (0)
231 #define pte_unmap_nested(pte) do { } while (0)
233 #define set_pte(ptep, pte) cpu_set_pte(ptep,pte)
234 #define set_pte_at(mm,addr,ptep,pteval) set_pte(ptep,pteval)
237 * The following only work if pte_present() is true.
238 * Undefined behaviour if not..
240 #define pte_present(pte) (pte_val(pte) & L_PTE_PRESENT)
241 #define pte_read(pte) (pte_val(pte) & L_PTE_USER)
242 #define pte_write(pte) (pte_val(pte) & L_PTE_WRITE)
243 #define pte_exec(pte) (pte_val(pte) & L_PTE_EXEC)
244 #define pte_dirty(pte) (pte_val(pte) & L_PTE_DIRTY)
245 #define pte_young(pte) (pte_val(pte) & L_PTE_YOUNG)
248 * The following only works if pte_present() is not true.
250 #define pte_file(pte) (pte_val(pte) & L_PTE_FILE)
251 #define pte_to_pgoff(x) (pte_val(x) >> 2)
252 #define pgoff_to_pte(x) __pte(((x) << 2) | L_PTE_FILE)
254 #define PTE_FILE_MAX_BITS 30
256 #define PTE_BIT_FUNC(fn,op) \
257 static inline pte_t pte_##fn(pte_t pte) { pte_val(pte) op; return pte; }
259 /*PTE_BIT_FUNC(rdprotect, &= ~L_PTE_USER);*/
260 /*PTE_BIT_FUNC(mkread, |= L_PTE_USER);*/
261 PTE_BIT_FUNC(wrprotect, &= ~L_PTE_WRITE);
262 PTE_BIT_FUNC(mkwrite, |= L_PTE_WRITE);
263 PTE_BIT_FUNC(exprotect, &= ~L_PTE_EXEC);
264 PTE_BIT_FUNC(mkexec, |= L_PTE_EXEC);
265 PTE_BIT_FUNC(mkclean, &= ~L_PTE_DIRTY);
266 PTE_BIT_FUNC(mkdirty, |= L_PTE_DIRTY);
267 PTE_BIT_FUNC(mkold, &= ~L_PTE_YOUNG);
268 PTE_BIT_FUNC(mkyoung, |= L_PTE_YOUNG);
271 * Mark the prot value as uncacheable and unbufferable.
273 #define pgprot_noncached(prot) __pgprot(pgprot_val(prot) & ~(L_PTE_CACHEABLE | L_PTE_BUFFERABLE))
274 #define pgprot_writecombine(prot) __pgprot(pgprot_val(prot) & ~L_PTE_CACHEABLE)
276 #define pmd_none(pmd) (!pmd_val(pmd))
277 #define pmd_present(pmd) (pmd_val(pmd))
278 #define pmd_bad(pmd) (pmd_val(pmd) & 2)
280 #define copy_pmd(pmdpd,pmdps) \
281 do { \
282 pmdpd[0] = pmdps[0]; \
283 pmdpd[1] = pmdps[1]; \
284 flush_pmd_entry(pmdpd); \
285 } while (0)
287 #define pmd_clear(pmdp) \
288 do { \
289 pmdp[0] = __pmd(0); \
290 pmdp[1] = __pmd(0); \
291 clean_pmd_entry(pmdp); \
292 } while (0)
294 static inline pte_t *pmd_page_kernel(pmd_t pmd)
296 unsigned long ptr;
298 ptr = pmd_val(pmd) & ~(PTRS_PER_PTE * sizeof(void *) - 1);
299 ptr += PTRS_PER_PTE * sizeof(void *);
301 return __va(ptr);
304 #define pmd_page(pmd) virt_to_page(__va(pmd_val(pmd)))
307 * Permanent address of a page. We never have highmem, so this is trivial.
309 #define pages_to_mb(x) ((x) >> (20 - PAGE_SHIFT))
312 * Conversion functions: convert a page and protection to a page entry,
313 * and a page entry and page directory to the page they refer to.
315 #define mk_pte(page,prot) pfn_pte(page_to_pfn(page),prot)
318 * The "pgd_xxx()" functions here are trivial for a folded two-level
319 * setup: the pgd is never bad, and a pmd always exists (as it's folded
320 * into the pgd entry)
322 #define pgd_none(pgd) (0)
323 #define pgd_bad(pgd) (0)
324 #define pgd_present(pgd) (1)
325 #define pgd_clear(pgdp) do { } while (0)
326 #define set_pgd(pgd,pgdp) do { } while (0)
328 /* to find an entry in a page-table-directory */
329 #define pgd_index(addr) ((addr) >> PGDIR_SHIFT)
331 #define pgd_offset(mm, addr) ((mm)->pgd+pgd_index(addr))
333 /* to find an entry in a kernel page-table-directory */
334 #define pgd_offset_k(addr) pgd_offset(&init_mm, addr)
336 /* Find an entry in the second-level page table.. */
337 #define pmd_offset(dir, addr) ((pmd_t *)(dir))
339 /* Find an entry in the third-level page table.. */
340 #define __pte_index(addr) (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
342 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
344 const unsigned long mask = L_PTE_EXEC | L_PTE_WRITE | L_PTE_USER;
345 pte_val(pte) = (pte_val(pte) & ~mask) | (pgprot_val(newprot) & mask);
346 return pte;
349 extern pgd_t swapper_pg_dir[PTRS_PER_PGD];
351 /* Encode and decode a swap entry.
353 * We support up to 32GB of swap on 4k machines
355 #define __swp_type(x) (((x).val >> 2) & 0x7f)
356 #define __swp_offset(x) ((x).val >> 9)
357 #define __swp_entry(type,offset) ((swp_entry_t) { ((type) << 2) | ((offset) << 9) })
358 #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
359 #define __swp_entry_to_pte(swp) ((pte_t) { (swp).val })
361 /* Needs to be defined here and not in linux/mm.h, as it is arch dependent */
362 /* FIXME: this is not correct */
363 #define kern_addr_valid(addr) (1)
365 #include <asm-generic/pgtable.h>
368 * We provide our own arch_get_unmapped_area to cope with VIPT caches.
370 #define HAVE_ARCH_UNMAPPED_AREA
373 * remap a physical page `pfn' of size `size' with page protection `prot'
374 * into virtual address `from'
376 #define io_remap_pfn_range(vma,from,pfn,size,prot) \
377 remap_pfn_range(vma, from, pfn, size, prot)
379 #define MK_IOSPACE_PFN(space, pfn) (pfn)
380 #define GET_IOSPACE(pfn) 0
381 #define GET_PFN(pfn) (pfn)
383 #define pgtable_cache_init() do { } while (0)
385 #endif /* !__ASSEMBLY__ */
387 #endif /* CONFIG_MMU */
389 #endif /* _ASMARM_PGTABLE_H */