blob: 5a0d19b466b0cd2242bbf6f1fca244ec90d12617 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/include/asm-arm/pgtable.h
3 *
4 * Copyright (C) 1995-2002 Russell King
5 *
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
12
13#include <asm-generic/4level-fixup.h>
14
15#include <asm/memory.h>
16#include <asm/proc-fns.h>
17#include <asm/arch/vmalloc.h>
18
19/*
Russell King5c3073e2005-05-03 12:20:29 +010020 * Just any arbitrary offset to the start of the vmalloc VM area: the
21 * current 8MB value just means that there will be a 8MB "hole" after the
22 * physical memory until the kernel virtual memory starts. That means that
23 * any out-of-bounds memory accesses will hopefully be caught.
24 * The vmalloc() routines leaves a hole of 4kB between each vmalloced
25 * area for the same reason. ;)
26 *
27 * Note that platforms may override VMALLOC_START, but they must provide
28 * VMALLOC_END. VMALLOC_END defines the (exclusive) limit of this space,
29 * which may not overlap IO space.
30 */
31#ifndef VMALLOC_START
32#define VMALLOC_OFFSET (8*1024*1024)
33#define VMALLOC_START (((unsigned long)high_memory + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1))
34#endif
35
36/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070037 * Hardware-wise, we have a two level page table structure, where the first
38 * level has 4096 entries, and the second level has 256 entries. Each entry
39 * is one 32-bit word. Most of the bits in the second level entry are used
40 * by hardware, and there aren't any "accessed" and "dirty" bits.
41 *
42 * Linux on the other hand has a three level page table structure, which can
43 * be wrapped to fit a two level page table structure easily - using the PGD
44 * and PTE only. However, Linux also expects one "PTE" table per page, and
45 * at least a "dirty" bit.
46 *
47 * Therefore, we tweak the implementation slightly - we tell Linux that we
48 * have 2048 entries in the first level, each of which is 8 bytes (iow, two
49 * hardware pointers to the second level.) The second level contains two
50 * hardware PTE tables arranged contiguously, followed by Linux versions
51 * which contain the state information Linux needs. We, therefore, end up
52 * with 512 entries in the "PTE" level.
53 *
54 * This leads to the page tables having the following layout:
55 *
56 * pgd pte
57 * | |
58 * +--------+ +0
59 * | |-----> +------------+ +0
60 * +- - - - + +4 | h/w pt 0 |
61 * | |-----> +------------+ +1024
62 * +--------+ +8 | h/w pt 1 |
63 * | | +------------+ +2048
64 * +- - - - + | Linux pt 0 |
65 * | | +------------+ +3072
66 * +--------+ | Linux pt 1 |
67 * | | +------------+ +4096
68 *
69 * See L_PTE_xxx below for definitions of bits in the "Linux pt", and
70 * PTE_xxx for definitions of bits appearing in the "h/w pt".
71 *
72 * PMD_xxx definitions refer to bits in the first level page table.
73 *
74 * The "dirty" bit is emulated by only granting hardware write permission
75 * iff the page is marked "writable" and "dirty" in the Linux PTE. This
76 * means that a write to a clean page will cause a permission fault, and
77 * the Linux MM layer will mark the page dirty via handle_pte_fault().
78 * For the hardware to notice the permission change, the TLB entry must
79 * be flushed, and ptep_establish() does that for us.
80 *
81 * The "accessed" or "young" bit is emulated by a similar method; we only
82 * allow accesses to the page if the "young" bit is set. Accesses to the
83 * page will cause a fault, and handle_pte_fault() will set the young bit
84 * for us as long as the page is marked present in the corresponding Linux
85 * PTE entry. Again, ptep_establish() will ensure that the TLB is up to
86 * date.
87 *
88 * However, when the "young" bit is cleared, we deny access to the page
89 * by clearing the hardware PTE. Currently Linux does not flush the TLB
90 * for us in this case, which means the TLB will retain the transation
91 * until either the TLB entry is evicted under pressure, or a context
92 * switch which changes the user space mapping occurs.
93 */
94#define PTRS_PER_PTE 512
95#define PTRS_PER_PMD 1
96#define PTRS_PER_PGD 2048
97
98/*
99 * PMD_SHIFT determines the size of the area a second-level page table can map
100 * PGDIR_SHIFT determines what a third-level page table entry can map
101 */
102#define PMD_SHIFT 21
103#define PGDIR_SHIFT 21
104
105#define LIBRARY_TEXT_START 0x0c000000
106
107#ifndef __ASSEMBLY__
108extern void __pte_error(const char *file, int line, unsigned long val);
109extern void __pmd_error(const char *file, int line, unsigned long val);
110extern void __pgd_error(const char *file, int line, unsigned long val);
111
112#define pte_ERROR(pte) __pte_error(__FILE__, __LINE__, pte_val(pte))
113#define pmd_ERROR(pmd) __pmd_error(__FILE__, __LINE__, pmd_val(pmd))
114#define pgd_ERROR(pgd) __pgd_error(__FILE__, __LINE__, pgd_val(pgd))
115#endif /* !__ASSEMBLY__ */
116
117#define PMD_SIZE (1UL << PMD_SHIFT)
118#define PMD_MASK (~(PMD_SIZE-1))
119#define PGDIR_SIZE (1UL << PGDIR_SHIFT)
120#define PGDIR_MASK (~(PGDIR_SIZE-1))
121
Hugh Dickins6119be02005-04-19 13:29:21 -0700122/*
123 * This is the lowest virtual address we can permit any user space
124 * mapping to be mapped at. This is particularly important for
125 * non-high vector CPUs.
126 */
127#define FIRST_USER_ADDRESS PAGE_SIZE
128
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129#define FIRST_USER_PGD_NR 1
130#define USER_PTRS_PER_PGD ((TASK_SIZE/PGDIR_SIZE) - FIRST_USER_PGD_NR)
131
132/*
133 * ARMv6 supersection address mask and size definitions.
134 */
135#define SUPERSECTION_SHIFT 24
136#define SUPERSECTION_SIZE (1UL << SUPERSECTION_SHIFT)
137#define SUPERSECTION_MASK (~(SUPERSECTION_SIZE-1))
138
139/*
140 * Hardware page table definitions.
141 *
142 * + Level 1 descriptor (PMD)
143 * - common
144 */
145#define PMD_TYPE_MASK (3 << 0)
146#define PMD_TYPE_FAULT (0 << 0)
147#define PMD_TYPE_TABLE (1 << 0)
148#define PMD_TYPE_SECT (2 << 0)
149#define PMD_BIT4 (1 << 4)
150#define PMD_DOMAIN(x) ((x) << 5)
151#define PMD_PROTECTION (1 << 9) /* v5 */
152/*
153 * - section
154 */
155#define PMD_SECT_BUFFERABLE (1 << 2)
156#define PMD_SECT_CACHEABLE (1 << 3)
157#define PMD_SECT_AP_WRITE (1 << 10)
158#define PMD_SECT_AP_READ (1 << 11)
159#define PMD_SECT_TEX(x) ((x) << 12) /* v5 */
160#define PMD_SECT_APX (1 << 15) /* v6 */
161#define PMD_SECT_S (1 << 16) /* v6 */
162#define PMD_SECT_nG (1 << 17) /* v6 */
163#define PMD_SECT_SUPER (1 << 18) /* v6 */
164
165#define PMD_SECT_UNCACHED (0)
166#define PMD_SECT_BUFFERED (PMD_SECT_BUFFERABLE)
167#define PMD_SECT_WT (PMD_SECT_CACHEABLE)
168#define PMD_SECT_WB (PMD_SECT_CACHEABLE | PMD_SECT_BUFFERABLE)
169#define PMD_SECT_MINICACHE (PMD_SECT_TEX(1) | PMD_SECT_CACHEABLE)
170#define PMD_SECT_WBWA (PMD_SECT_TEX(1) | PMD_SECT_CACHEABLE | PMD_SECT_BUFFERABLE)
171
172/*
173 * - coarse table (not used)
174 */
175
176/*
177 * + Level 2 descriptor (PTE)
178 * - common
179 */
180#define PTE_TYPE_MASK (3 << 0)
181#define PTE_TYPE_FAULT (0 << 0)
182#define PTE_TYPE_LARGE (1 << 0)
183#define PTE_TYPE_SMALL (2 << 0)
184#define PTE_TYPE_EXT (3 << 0) /* v5 */
185#define PTE_BUFFERABLE (1 << 2)
186#define PTE_CACHEABLE (1 << 3)
187
188/*
189 * - extended small page/tiny page
190 */
Russell King1b9749e2005-08-10 16:15:32 +0100191#define PTE_EXT_XN (1 << 0) /* v6 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192#define PTE_EXT_AP_MASK (3 << 4)
Russell King1b9749e2005-08-10 16:15:32 +0100193#define PTE_EXT_AP0 (1 << 4)
194#define PTE_EXT_AP1 (2 << 4)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195#define PTE_EXT_AP_UNO_SRO (0 << 4)
Russell King1b9749e2005-08-10 16:15:32 +0100196#define PTE_EXT_AP_UNO_SRW (PTE_EXT_AP0)
197#define PTE_EXT_AP_URO_SRW (PTE_EXT_AP1)
198#define PTE_EXT_AP_URW_SRW (PTE_EXT_AP1|PTE_EXT_AP0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199#define PTE_EXT_TEX(x) ((x) << 6) /* v5 */
Russell King1b9749e2005-08-10 16:15:32 +0100200#define PTE_EXT_APX (1 << 9) /* v6 */
201#define PTE_EXT_SHARED (1 << 10) /* v6 */
202#define PTE_EXT_NG (1 << 11) /* v6 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203
204/*
205 * - small page
206 */
207#define PTE_SMALL_AP_MASK (0xff << 4)
208#define PTE_SMALL_AP_UNO_SRO (0x00 << 4)
209#define PTE_SMALL_AP_UNO_SRW (0x55 << 4)
210#define PTE_SMALL_AP_URO_SRW (0xaa << 4)
211#define PTE_SMALL_AP_URW_SRW (0xff << 4)
212
213/*
214 * "Linux" PTE definitions.
215 *
216 * We keep two sets of PTEs - the hardware and the linux version.
217 * This allows greater flexibility in the way we map the Linux bits
218 * onto the hardware tables, and allows us to have YOUNG and DIRTY
219 * bits.
220 *
221 * The PTE table pointer refers to the hardware entries; the "Linux"
222 * entries are stored 1024 bytes below.
223 */
224#define L_PTE_PRESENT (1 << 0)
225#define L_PTE_FILE (1 << 1) /* only when !PRESENT */
226#define L_PTE_YOUNG (1 << 1)
227#define L_PTE_BUFFERABLE (1 << 2) /* matches PTE */
228#define L_PTE_CACHEABLE (1 << 3) /* matches PTE */
229#define L_PTE_USER (1 << 4)
230#define L_PTE_WRITE (1 << 5)
231#define L_PTE_EXEC (1 << 6)
232#define L_PTE_DIRTY (1 << 7)
Russell King6626a702005-08-10 16:18:35 +0100233#define L_PTE_SHARED (1 << 10) /* shared between CPUs (v6) */
234#define L_PTE_ASID (1 << 11) /* non-global (use ASID, v6) */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235
236#ifndef __ASSEMBLY__
237
238#include <asm/domain.h>
239
240#define _PAGE_USER_TABLE (PMD_TYPE_TABLE | PMD_BIT4 | PMD_DOMAIN(DOMAIN_USER))
241#define _PAGE_KERNEL_TABLE (PMD_TYPE_TABLE | PMD_BIT4 | PMD_DOMAIN(DOMAIN_KERNEL))
242
243/*
244 * The following macros handle the cache and bufferable bits...
245 */
246#define _L_PTE_DEFAULT L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_CACHEABLE | L_PTE_BUFFERABLE
247#define _L_PTE_READ L_PTE_USER | L_PTE_EXEC
248
249extern pgprot_t pgprot_kernel;
250
251#define PAGE_NONE __pgprot(_L_PTE_DEFAULT)
252#define PAGE_COPY __pgprot(_L_PTE_DEFAULT | _L_PTE_READ)
253#define PAGE_SHARED __pgprot(_L_PTE_DEFAULT | _L_PTE_READ | L_PTE_WRITE)
254#define PAGE_READONLY __pgprot(_L_PTE_DEFAULT | _L_PTE_READ)
255#define PAGE_KERNEL pgprot_kernel
256
257#endif /* __ASSEMBLY__ */
258
259/*
260 * The table below defines the page protection levels that we insert into our
261 * Linux page table version. These get translated into the best that the
262 * architecture can perform. Note that on most ARM hardware:
263 * 1) We cannot do execute protection
264 * 2) If we could do execute protection, then read is implied
265 * 3) write implies read permissions
266 */
267#define __P000 PAGE_NONE
268#define __P001 PAGE_READONLY
269#define __P010 PAGE_COPY
270#define __P011 PAGE_COPY
271#define __P100 PAGE_READONLY
272#define __P101 PAGE_READONLY
273#define __P110 PAGE_COPY
274#define __P111 PAGE_COPY
275
276#define __S000 PAGE_NONE
277#define __S001 PAGE_READONLY
278#define __S010 PAGE_SHARED
279#define __S011 PAGE_SHARED
280#define __S100 PAGE_READONLY
281#define __S101 PAGE_READONLY
282#define __S110 PAGE_SHARED
283#define __S111 PAGE_SHARED
284
285#ifndef __ASSEMBLY__
286/*
287 * ZERO_PAGE is a global shared page that is always zero: used
288 * for zero-mapped memory areas etc..
289 */
290extern struct page *empty_zero_page;
291#define ZERO_PAGE(vaddr) (empty_zero_page)
292
293#define pte_pfn(pte) (pte_val(pte) >> PAGE_SHIFT)
294#define pfn_pte(pfn,prot) (__pte(((pfn) << PAGE_SHIFT) | pgprot_val(prot)))
295
296#define pte_none(pte) (!pte_val(pte))
297#define pte_clear(mm,addr,ptep) set_pte_at((mm),(addr),(ptep), __pte(0))
298#define pte_page(pte) (pfn_to_page(pte_pfn(pte)))
299#define pte_offset_kernel(dir,addr) (pmd_page_kernel(*(dir)) + __pte_index(addr))
300#define pte_offset_map(dir,addr) (pmd_page_kernel(*(dir)) + __pte_index(addr))
301#define pte_offset_map_nested(dir,addr) (pmd_page_kernel(*(dir)) + __pte_index(addr))
302#define pte_unmap(pte) do { } while (0)
303#define pte_unmap_nested(pte) do { } while (0)
304
305#define set_pte(ptep, pte) cpu_set_pte(ptep,pte)
306#define set_pte_at(mm,addr,ptep,pteval) set_pte(ptep,pteval)
307
308/*
309 * The following only work if pte_present() is true.
310 * Undefined behaviour if not..
311 */
312#define pte_present(pte) (pte_val(pte) & L_PTE_PRESENT)
313#define pte_read(pte) (pte_val(pte) & L_PTE_USER)
314#define pte_write(pte) (pte_val(pte) & L_PTE_WRITE)
315#define pte_exec(pte) (pte_val(pte) & L_PTE_EXEC)
316#define pte_dirty(pte) (pte_val(pte) & L_PTE_DIRTY)
317#define pte_young(pte) (pte_val(pte) & L_PTE_YOUNG)
318
319/*
320 * The following only works if pte_present() is not true.
321 */
322#define pte_file(pte) (pte_val(pte) & L_PTE_FILE)
323#define pte_to_pgoff(x) (pte_val(x) >> 2)
324#define pgoff_to_pte(x) __pte(((x) << 2) | L_PTE_FILE)
325
326#define PTE_FILE_MAX_BITS 30
327
328#define PTE_BIT_FUNC(fn,op) \
329static inline pte_t pte_##fn(pte_t pte) { pte_val(pte) op; return pte; }
330
331/*PTE_BIT_FUNC(rdprotect, &= ~L_PTE_USER);*/
332/*PTE_BIT_FUNC(mkread, |= L_PTE_USER);*/
333PTE_BIT_FUNC(wrprotect, &= ~L_PTE_WRITE);
334PTE_BIT_FUNC(mkwrite, |= L_PTE_WRITE);
335PTE_BIT_FUNC(exprotect, &= ~L_PTE_EXEC);
336PTE_BIT_FUNC(mkexec, |= L_PTE_EXEC);
337PTE_BIT_FUNC(mkclean, &= ~L_PTE_DIRTY);
338PTE_BIT_FUNC(mkdirty, |= L_PTE_DIRTY);
339PTE_BIT_FUNC(mkold, &= ~L_PTE_YOUNG);
340PTE_BIT_FUNC(mkyoung, |= L_PTE_YOUNG);
341
342/*
343 * Mark the prot value as uncacheable and unbufferable.
344 */
345#define pgprot_noncached(prot) __pgprot(pgprot_val(prot) & ~(L_PTE_CACHEABLE | L_PTE_BUFFERABLE))
346#define pgprot_writecombine(prot) __pgprot(pgprot_val(prot) & ~L_PTE_CACHEABLE)
347
348#define pmd_none(pmd) (!pmd_val(pmd))
349#define pmd_present(pmd) (pmd_val(pmd))
350#define pmd_bad(pmd) (pmd_val(pmd) & 2)
351
352#define copy_pmd(pmdpd,pmdps) \
353 do { \
354 pmdpd[0] = pmdps[0]; \
355 pmdpd[1] = pmdps[1]; \
356 flush_pmd_entry(pmdpd); \
357 } while (0)
358
359#define pmd_clear(pmdp) \
360 do { \
361 pmdp[0] = __pmd(0); \
362 pmdp[1] = __pmd(0); \
363 clean_pmd_entry(pmdp); \
364 } while (0)
365
366static inline pte_t *pmd_page_kernel(pmd_t pmd)
367{
368 unsigned long ptr;
369
370 ptr = pmd_val(pmd) & ~(PTRS_PER_PTE * sizeof(void *) - 1);
371 ptr += PTRS_PER_PTE * sizeof(void *);
372
373 return __va(ptr);
374}
375
376#define pmd_page(pmd) virt_to_page(__va(pmd_val(pmd)))
377
378/*
379 * Permanent address of a page. We never have highmem, so this is trivial.
380 */
381#define pages_to_mb(x) ((x) >> (20 - PAGE_SHIFT))
382
383/*
384 * Conversion functions: convert a page and protection to a page entry,
385 * and a page entry and page directory to the page they refer to.
386 */
387#define mk_pte(page,prot) pfn_pte(page_to_pfn(page),prot)
388
389/*
390 * The "pgd_xxx()" functions here are trivial for a folded two-level
391 * setup: the pgd is never bad, and a pmd always exists (as it's folded
392 * into the pgd entry)
393 */
394#define pgd_none(pgd) (0)
395#define pgd_bad(pgd) (0)
396#define pgd_present(pgd) (1)
397#define pgd_clear(pgdp) do { } while (0)
398#define set_pgd(pgd,pgdp) do { } while (0)
399
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400/* to find an entry in a page-table-directory */
401#define pgd_index(addr) ((addr) >> PGDIR_SHIFT)
402
403#define pgd_offset(mm, addr) ((mm)->pgd+pgd_index(addr))
404
405/* to find an entry in a kernel page-table-directory */
406#define pgd_offset_k(addr) pgd_offset(&init_mm, addr)
407
408/* Find an entry in the second-level page table.. */
409#define pmd_offset(dir, addr) ((pmd_t *)(dir))
410
411/* Find an entry in the third-level page table.. */
412#define __pte_index(addr) (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
413
414static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
415{
416 const unsigned long mask = L_PTE_EXEC | L_PTE_WRITE | L_PTE_USER;
417 pte_val(pte) = (pte_val(pte) & ~mask) | (pgprot_val(newprot) & mask);
418 return pte;
419}
420
421extern pgd_t swapper_pg_dir[PTRS_PER_PGD];
422
423/* Encode and decode a swap entry.
424 *
425 * We support up to 32GB of swap on 4k machines
426 */
427#define __swp_type(x) (((x).val >> 2) & 0x7f)
428#define __swp_offset(x) ((x).val >> 9)
429#define __swp_entry(type,offset) ((swp_entry_t) { ((type) << 2) | ((offset) << 9) })
430#define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
431#define __swp_entry_to_pte(swp) ((pte_t) { (swp).val })
432
433/* Needs to be defined here and not in linux/mm.h, as it is arch dependent */
434/* FIXME: this is not correct */
435#define kern_addr_valid(addr) (1)
436
437#include <asm-generic/pgtable.h>
438
439/*
440 * We provide our own arch_get_unmapped_area to cope with VIPT caches.
441 */
442#define HAVE_ARCH_UNMAPPED_AREA
443
444/*
Randy Dunlap33bf5612005-09-13 01:25:50 -0700445 * remap a physical page `pfn' of size `size' with page protection `prot'
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 * into virtual address `from'
447 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448#define io_remap_pfn_range(vma,from,pfn,size,prot) \
449 remap_pfn_range(vma, from, pfn, size, prot)
450
451#define MK_IOSPACE_PFN(space, pfn) (pfn)
452#define GET_IOSPACE(pfn) 0
453#define GET_PFN(pfn) (pfn)
454
455#define pgtable_cache_init() do { } while (0)
456
457#endif /* !__ASSEMBLY__ */
458
459#endif /* _ASMARM_PGTABLE_H */