diff options
Diffstat (limited to 'arch/arm/include/asm/pgtable.h')
-rw-r--r-- | arch/arm/include/asm/pgtable.h | 141 |
1 files changed, 5 insertions, 136 deletions
diff --git a/arch/arm/include/asm/pgtable.h b/arch/arm/include/asm/pgtable.h index f1956b27ae5a..9451dce3a553 100644 --- a/arch/arm/include/asm/pgtable.h +++ b/arch/arm/include/asm/pgtable.h | |||
@@ -24,6 +24,8 @@ | |||
24 | #include <mach/vmalloc.h> | 24 | #include <mach/vmalloc.h> |
25 | #include <asm/pgtable-hwdef.h> | 25 | #include <asm/pgtable-hwdef.h> |
26 | 26 | ||
27 | #include <asm/pgtable-2level.h> | ||
28 | |||
27 | /* | 29 | /* |
28 | * Just any arbitrary offset to the start of the vmalloc VM area: the | 30 | * Just any arbitrary offset to the start of the vmalloc VM area: the |
29 | * current 8MB value just means that there will be a 8MB "hole" after the | 31 | * current 8MB value just means that there will be a 8MB "hole" after the |
@@ -41,79 +43,6 @@ | |||
41 | #define VMALLOC_START (((unsigned long)high_memory + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1)) | 43 | #define VMALLOC_START (((unsigned long)high_memory + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1)) |
42 | #endif | 44 | #endif |
43 | 45 | ||
44 | /* | ||
45 | * Hardware-wise, we have a two level page table structure, where the first | ||
46 | * level has 4096 entries, and the second level has 256 entries. Each entry | ||
47 | * is one 32-bit word. Most of the bits in the second level entry are used | ||
48 | * by hardware, and there aren't any "accessed" and "dirty" bits. | ||
49 | * | ||
50 | * Linux on the other hand has a three level page table structure, which can | ||
51 | * be wrapped to fit a two level page table structure easily - using the PGD | ||
52 | * and PTE only. However, Linux also expects one "PTE" table per page, and | ||
53 | * at least a "dirty" bit. | ||
54 | * | ||
55 | * Therefore, we tweak the implementation slightly - we tell Linux that we | ||
56 | * have 2048 entries in the first level, each of which is 8 bytes (iow, two | ||
57 | * hardware pointers to the second level.) The second level contains two | ||
58 | * hardware PTE tables arranged contiguously, preceded by Linux versions | ||
59 | * which contain the state information Linux needs. We, therefore, end up | ||
60 | * with 512 entries in the "PTE" level. | ||
61 | * | ||
62 | * This leads to the page tables having the following layout: | ||
63 | * | ||
64 | * pgd pte | ||
65 | * | | | ||
66 | * +--------+ | ||
67 | * | | +------------+ +0 | ||
68 | * +- - - - + | Linux pt 0 | | ||
69 | * | | +------------+ +1024 | ||
70 | * +--------+ +0 | Linux pt 1 | | ||
71 | * | |-----> +------------+ +2048 | ||
72 | * +- - - - + +4 | h/w pt 0 | | ||
73 | * | |-----> +------------+ +3072 | ||
74 | * +--------+ +8 | h/w pt 1 | | ||
75 | * | | +------------+ +4096 | ||
76 | * | ||
77 | * See L_PTE_xxx below for definitions of bits in the "Linux pt", and | ||
78 | * PTE_xxx for definitions of bits appearing in the "h/w pt". | ||
79 | * | ||
80 | * PMD_xxx definitions refer to bits in the first level page table. | ||
81 | * | ||
82 | * The "dirty" bit is emulated by only granting hardware write permission | ||
83 | * iff the page is marked "writable" and "dirty" in the Linux PTE. This | ||
84 | * means that a write to a clean page will cause a permission fault, and | ||
85 | * the Linux MM layer will mark the page dirty via handle_pte_fault(). | ||
86 | * For the hardware to notice the permission change, the TLB entry must | ||
87 | * be flushed, and ptep_set_access_flags() does that for us. | ||
88 | * | ||
89 | * The "accessed" or "young" bit is emulated by a similar method; we only | ||
90 | * allow accesses to the page if the "young" bit is set. Accesses to the | ||
91 | * page will cause a fault, and handle_pte_fault() will set the young bit | ||
92 | * for us as long as the page is marked present in the corresponding Linux | ||
93 | * PTE entry. Again, ptep_set_access_flags() will ensure that the TLB is | ||
94 | * up to date. | ||
95 | * | ||
96 | * However, when the "young" bit is cleared, we deny access to the page | ||
97 | * by clearing the hardware PTE. Currently Linux does not flush the TLB | ||
98 | * for us in this case, which means the TLB will retain the transation | ||
99 | * until either the TLB entry is evicted under pressure, or a context | ||
100 | * switch which changes the user space mapping occurs. | ||
101 | */ | ||
102 | #define PTRS_PER_PTE 512 | ||
103 | #define PTRS_PER_PMD 1 | ||
104 | #define PTRS_PER_PGD 2048 | ||
105 | |||
106 | #define PTE_HWTABLE_PTRS (PTRS_PER_PTE) | ||
107 | #define PTE_HWTABLE_OFF (PTE_HWTABLE_PTRS * sizeof(pte_t)) | ||
108 | #define PTE_HWTABLE_SIZE (PTRS_PER_PTE * sizeof(u32)) | ||
109 | |||
110 | /* | ||
111 | * PMD_SHIFT determines the size of the area a second-level page table can map | ||
112 | * PGDIR_SHIFT determines what a third-level page table entry can map | ||
113 | */ | ||
114 | #define PMD_SHIFT 21 | ||
115 | #define PGDIR_SHIFT 21 | ||
116 | |||
117 | #define LIBRARY_TEXT_START 0x0c000000 | 46 | #define LIBRARY_TEXT_START 0x0c000000 |
118 | 47 | ||
119 | #ifndef __ASSEMBLY__ | 48 | #ifndef __ASSEMBLY__ |
@@ -124,12 +53,6 @@ extern void __pgd_error(const char *file, int line, pgd_t); | |||
124 | #define pte_ERROR(pte) __pte_error(__FILE__, __LINE__, pte) | 53 | #define pte_ERROR(pte) __pte_error(__FILE__, __LINE__, pte) |
125 | #define pmd_ERROR(pmd) __pmd_error(__FILE__, __LINE__, pmd) | 54 | #define pmd_ERROR(pmd) __pmd_error(__FILE__, __LINE__, pmd) |
126 | #define pgd_ERROR(pgd) __pgd_error(__FILE__, __LINE__, pgd) | 55 | #define pgd_ERROR(pgd) __pgd_error(__FILE__, __LINE__, pgd) |
127 | #endif /* !__ASSEMBLY__ */ | ||
128 | |||
129 | #define PMD_SIZE (1UL << PMD_SHIFT) | ||
130 | #define PMD_MASK (~(PMD_SIZE-1)) | ||
131 | #define PGDIR_SIZE (1UL << PGDIR_SHIFT) | ||
132 | #define PGDIR_MASK (~(PGDIR_SIZE-1)) | ||
133 | 56 | ||
134 | /* | 57 | /* |
135 | * This is the lowest virtual address we can permit any user space | 58 | * This is the lowest virtual address we can permit any user space |
@@ -138,60 +61,6 @@ extern void __pgd_error(const char *file, int line, pgd_t); | |||
138 | */ | 61 | */ |
139 | #define FIRST_USER_ADDRESS PAGE_SIZE | 62 | #define FIRST_USER_ADDRESS PAGE_SIZE |
140 | 63 | ||
141 | #define USER_PTRS_PER_PGD (TASK_SIZE / PGDIR_SIZE) | ||
142 | |||
143 | /* | ||
144 | * section address mask and size definitions. | ||
145 | */ | ||
146 | #define SECTION_SHIFT 20 | ||
147 | #define SECTION_SIZE (1UL << SECTION_SHIFT) | ||
148 | #define SECTION_MASK (~(SECTION_SIZE-1)) | ||
149 | |||
150 | /* | ||
151 | * ARMv6 supersection address mask and size definitions. | ||
152 | */ | ||
153 | #define SUPERSECTION_SHIFT 24 | ||
154 | #define SUPERSECTION_SIZE (1UL << SUPERSECTION_SHIFT) | ||
155 | #define SUPERSECTION_MASK (~(SUPERSECTION_SIZE-1)) | ||
156 | |||
157 | /* | ||
158 | * "Linux" PTE definitions. | ||
159 | * | ||
160 | * We keep two sets of PTEs - the hardware and the linux version. | ||
161 | * This allows greater flexibility in the way we map the Linux bits | ||
162 | * onto the hardware tables, and allows us to have YOUNG and DIRTY | ||
163 | * bits. | ||
164 | * | ||
165 | * The PTE table pointer refers to the hardware entries; the "Linux" | ||
166 | * entries are stored 1024 bytes below. | ||
167 | */ | ||
168 | #define L_PTE_PRESENT (_AT(pteval_t, 1) << 0) | ||
169 | #define L_PTE_YOUNG (_AT(pteval_t, 1) << 1) | ||
170 | #define L_PTE_FILE (_AT(pteval_t, 1) << 2) /* only when !PRESENT */ | ||
171 | #define L_PTE_DIRTY (_AT(pteval_t, 1) << 6) | ||
172 | #define L_PTE_RDONLY (_AT(pteval_t, 1) << 7) | ||
173 | #define L_PTE_USER (_AT(pteval_t, 1) << 8) | ||
174 | #define L_PTE_XN (_AT(pteval_t, 1) << 9) | ||
175 | #define L_PTE_SHARED (_AT(pteval_t, 1) << 10) /* shared(v6), coherent(xsc3) */ | ||
176 | |||
177 | /* | ||
178 | * These are the memory types, defined to be compatible with | ||
179 | * pre-ARMv6 CPUs cacheable and bufferable bits: XXCB | ||
180 | */ | ||
181 | #define L_PTE_MT_UNCACHED (_AT(pteval_t, 0x00) << 2) /* 0000 */ | ||
182 | #define L_PTE_MT_BUFFERABLE (_AT(pteval_t, 0x01) << 2) /* 0001 */ | ||
183 | #define L_PTE_MT_WRITETHROUGH (_AT(pteval_t, 0x02) << 2) /* 0010 */ | ||
184 | #define L_PTE_MT_WRITEBACK (_AT(pteval_t, 0x03) << 2) /* 0011 */ | ||
185 | #define L_PTE_MT_MINICACHE (_AT(pteval_t, 0x06) << 2) /* 0110 (sa1100, xscale) */ | ||
186 | #define L_PTE_MT_WRITEALLOC (_AT(pteval_t, 0x07) << 2) /* 0111 */ | ||
187 | #define L_PTE_MT_DEV_SHARED (_AT(pteval_t, 0x04) << 2) /* 0100 */ | ||
188 | #define L_PTE_MT_DEV_NONSHARED (_AT(pteval_t, 0x0c) << 2) /* 1100 */ | ||
189 | #define L_PTE_MT_DEV_WC (_AT(pteval_t, 0x09) << 2) /* 1001 */ | ||
190 | #define L_PTE_MT_DEV_CACHED (_AT(pteval_t, 0x0b) << 2) /* 1011 */ | ||
191 | #define L_PTE_MT_MASK (_AT(pteval_t, 0x0f) << 2) | ||
192 | |||
193 | #ifndef __ASSEMBLY__ | ||
194 | |||
195 | /* | 64 | /* |
196 | * The pgprot_* and protection_map entries will be fixed up in runtime | 65 | * The pgprot_* and protection_map entries will be fixed up in runtime |
197 | * to include the cachable and bufferable bits based on memory policy, | 66 | * to include the cachable and bufferable bits based on memory policy, |
@@ -330,10 +199,10 @@ extern pgd_t swapper_pg_dir[PTRS_PER_PGD]; | |||
330 | 199 | ||
331 | static inline pte_t *pmd_page_vaddr(pmd_t pmd) | 200 | static inline pte_t *pmd_page_vaddr(pmd_t pmd) |
332 | { | 201 | { |
333 | return __va(pmd_val(pmd) & PAGE_MASK); | 202 | return __va(pmd_val(pmd) & PHYS_MASK & (s32)PAGE_MASK); |
334 | } | 203 | } |
335 | 204 | ||
336 | #define pmd_page(pmd) pfn_to_page(__phys_to_pfn(pmd_val(pmd))) | 205 | #define pmd_page(pmd) pfn_to_page(__phys_to_pfn(pmd_val(pmd) & PHYS_MASK)) |
337 | 206 | ||
338 | /* we don't need complex calculations here as the pmd is folded into the pgd */ | 207 | /* we don't need complex calculations here as the pmd is folded into the pgd */ |
339 | #define pmd_addr_end(addr,end) (end) | 208 | #define pmd_addr_end(addr,end) (end) |
@@ -354,7 +223,7 @@ static inline pte_t *pmd_page_vaddr(pmd_t pmd) | |||
354 | #define pte_offset_map(pmd,addr) (__pte_map(pmd) + pte_index(addr)) | 223 | #define pte_offset_map(pmd,addr) (__pte_map(pmd) + pte_index(addr)) |
355 | #define pte_unmap(pte) __pte_unmap(pte) | 224 | #define pte_unmap(pte) __pte_unmap(pte) |
356 | 225 | ||
357 | #define pte_pfn(pte) (pte_val(pte) >> PAGE_SHIFT) | 226 | #define pte_pfn(pte) ((pte_val(pte) & PHYS_MASK) >> PAGE_SHIFT) |
358 | #define pfn_pte(pfn,prot) __pte(__pfn_to_phys(pfn) | pgprot_val(prot)) | 227 | #define pfn_pte(pfn,prot) __pte(__pfn_to_phys(pfn) | pgprot_val(prot)) |
359 | 228 | ||
360 | #define pte_page(pte) pfn_to_page(pte_pfn(pte)) | 229 | #define pte_page(pte) pfn_to_page(pte_pfn(pte)) |