Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | #ifndef _PPC64_PGTABLE_H |
| 2 | #define _PPC64_PGTABLE_H |
| 3 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4 | /* |
| 5 | * This file contains the functions and defines necessary to modify and use |
| 6 | * the ppc64 hashed page table. |
| 7 | */ |
| 8 | |
| 9 | #ifndef __ASSEMBLY__ |
| 10 | #include <linux/config.h> |
| 11 | #include <linux/stddef.h> |
| 12 | #include <asm/processor.h> /* For TASK_SIZE */ |
| 13 | #include <asm/mmu.h> |
| 14 | #include <asm/page.h> |
| 15 | #include <asm/tlbflush.h> |
| 16 | #endif /* __ASSEMBLY__ */ |
| 17 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 18 | /* |
| 19 | * Entries per page directory level. The PTE level must use a 64b record |
| 20 | * for each page table entry. The PMD and PGD level use a 32b record for |
| 21 | * each entry by assuming that each entry is page aligned. |
| 22 | */ |
| 23 | #define PTE_INDEX_SIZE 9 |
David Gibson | e28f7fa | 2005-08-05 19:39:06 +1000 | [diff] [blame] | 24 | #define PMD_INDEX_SIZE 7 |
| 25 | #define PUD_INDEX_SIZE 7 |
| 26 | #define PGD_INDEX_SIZE 9 |
| 27 | |
| 28 | #define PTE_TABLE_SIZE (sizeof(pte_t) << PTE_INDEX_SIZE) |
| 29 | #define PMD_TABLE_SIZE (sizeof(pmd_t) << PMD_INDEX_SIZE) |
| 30 | #define PUD_TABLE_SIZE (sizeof(pud_t) << PUD_INDEX_SIZE) |
| 31 | #define PGD_TABLE_SIZE (sizeof(pgd_t) << PGD_INDEX_SIZE) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 32 | |
| 33 | #define PTRS_PER_PTE (1 << PTE_INDEX_SIZE) |
| 34 | #define PTRS_PER_PMD (1 << PMD_INDEX_SIZE) |
David Gibson | e28f7fa | 2005-08-05 19:39:06 +1000 | [diff] [blame] | 35 | #define PTRS_PER_PUD (1 << PMD_INDEX_SIZE) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 36 | #define PTRS_PER_PGD (1 << PGD_INDEX_SIZE) |
| 37 | |
David Gibson | 1f8d419 | 2005-05-05 16:15:13 -0700 | [diff] [blame] | 38 | /* PMD_SHIFT determines what a second-level page table entry can map */ |
| 39 | #define PMD_SHIFT (PAGE_SHIFT + PTE_INDEX_SIZE) |
| 40 | #define PMD_SIZE (1UL << PMD_SHIFT) |
| 41 | #define PMD_MASK (~(PMD_SIZE-1)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 42 | |
David Gibson | e28f7fa | 2005-08-05 19:39:06 +1000 | [diff] [blame] | 43 | /* PUD_SHIFT determines what a third-level page table entry can map */ |
| 44 | #define PUD_SHIFT (PMD_SHIFT + PMD_INDEX_SIZE) |
| 45 | #define PUD_SIZE (1UL << PUD_SHIFT) |
| 46 | #define PUD_MASK (~(PUD_SIZE-1)) |
| 47 | |
| 48 | /* PGDIR_SHIFT determines what a fourth-level page table entry can map */ |
| 49 | #define PGDIR_SHIFT (PUD_SHIFT + PUD_INDEX_SIZE) |
David Gibson | 1f8d419 | 2005-05-05 16:15:13 -0700 | [diff] [blame] | 50 | #define PGDIR_SIZE (1UL << PGDIR_SHIFT) |
| 51 | #define PGDIR_MASK (~(PGDIR_SIZE-1)) |
| 52 | |
| 53 | #define FIRST_USER_ADDRESS 0 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 54 | |
| 55 | /* |
| 56 | * Size of EA range mapped by our pagetables. |
| 57 | */ |
David Gibson | e28f7fa | 2005-08-05 19:39:06 +1000 | [diff] [blame] | 58 | #define PGTABLE_EADDR_SIZE (PTE_INDEX_SIZE + PMD_INDEX_SIZE + \ |
| 59 | PUD_INDEX_SIZE + PGD_INDEX_SIZE + PAGE_SHIFT) |
| 60 | #define PGTABLE_RANGE (1UL << PGTABLE_EADDR_SIZE) |
| 61 | |
| 62 | #if TASK_SIZE_USER64 > PGTABLE_RANGE |
| 63 | #error TASK_SIZE_USER64 exceeds pagetable range |
| 64 | #endif |
| 65 | |
| 66 | #if TASK_SIZE_USER64 > (1UL << (USER_ESID_BITS + SID_SHIFT)) |
| 67 | #error TASK_SIZE_USER64 exceeds user VSID range |
| 68 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 69 | |
| 70 | /* |
| 71 | * Define the address range of the vmalloc VM area. |
| 72 | */ |
| 73 | #define VMALLOC_START (0xD000000000000000ul) |
David Gibson | e28f7fa | 2005-08-05 19:39:06 +1000 | [diff] [blame] | 74 | #define VMALLOC_SIZE (0x80000000000UL) |
David Gibson | 20cee16 | 2005-06-21 17:15:31 -0700 | [diff] [blame] | 75 | #define VMALLOC_END (VMALLOC_START + VMALLOC_SIZE) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 76 | |
| 77 | /* |
| 78 | * Bits in a linux-style PTE. These match the bits in the |
| 79 | * (hardware-defined) PowerPC PTE as closely as possible. |
| 80 | */ |
| 81 | #define _PAGE_PRESENT 0x0001 /* software: pte contains a translation */ |
| 82 | #define _PAGE_USER 0x0002 /* matches one of the PP bits */ |
| 83 | #define _PAGE_FILE 0x0002 /* (!present only) software: pte holds file offset */ |
| 84 | #define _PAGE_EXEC 0x0004 /* No execute on POWER4 and newer (we invert) */ |
| 85 | #define _PAGE_GUARDED 0x0008 |
| 86 | #define _PAGE_COHERENT 0x0010 /* M: enforce memory coherence (SMP systems) */ |
| 87 | #define _PAGE_NO_CACHE 0x0020 /* I: cache inhibit */ |
| 88 | #define _PAGE_WRITETHRU 0x0040 /* W: cache write-through */ |
| 89 | #define _PAGE_DIRTY 0x0080 /* C: page changed */ |
| 90 | #define _PAGE_ACCESSED 0x0100 /* R: page referenced */ |
| 91 | #define _PAGE_RW 0x0200 /* software: user write access allowed */ |
| 92 | #define _PAGE_HASHPTE 0x0400 /* software: pte has an associated HPTE */ |
| 93 | #define _PAGE_BUSY 0x0800 /* software: PTE & hash are busy */ |
| 94 | #define _PAGE_SECONDARY 0x8000 /* software: HPTE is in secondary group */ |
| 95 | #define _PAGE_GROUP_IX 0x7000 /* software: HPTE index within group */ |
| 96 | #define _PAGE_HUGE 0x10000 /* 16MB page */ |
| 97 | /* Bits 0x7000 identify the index within an HPT Group */ |
| 98 | #define _PAGE_HPTEFLAGS (_PAGE_BUSY | _PAGE_HASHPTE | _PAGE_SECONDARY | _PAGE_GROUP_IX) |
| 99 | /* PAGE_MASK gives the right answer below, but only by accident */ |
| 100 | /* It should be preserving the high 48 bits and then specifically */ |
| 101 | /* preserving _PAGE_SECONDARY | _PAGE_GROUP_IX */ |
| 102 | #define _PAGE_CHG_MASK (PAGE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY | _PAGE_HPTEFLAGS) |
| 103 | |
| 104 | #define _PAGE_BASE (_PAGE_PRESENT | _PAGE_ACCESSED | _PAGE_COHERENT) |
| 105 | |
| 106 | #define _PAGE_WRENABLE (_PAGE_RW | _PAGE_DIRTY) |
| 107 | |
| 108 | /* __pgprot defined in asm-ppc64/page.h */ |
| 109 | #define PAGE_NONE __pgprot(_PAGE_PRESENT | _PAGE_ACCESSED) |
| 110 | |
| 111 | #define PAGE_SHARED __pgprot(_PAGE_BASE | _PAGE_RW | _PAGE_USER) |
| 112 | #define PAGE_SHARED_X __pgprot(_PAGE_BASE | _PAGE_RW | _PAGE_USER | _PAGE_EXEC) |
| 113 | #define PAGE_COPY __pgprot(_PAGE_BASE | _PAGE_USER) |
| 114 | #define PAGE_COPY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC) |
| 115 | #define PAGE_READONLY __pgprot(_PAGE_BASE | _PAGE_USER) |
| 116 | #define PAGE_READONLY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC) |
| 117 | #define PAGE_KERNEL __pgprot(_PAGE_BASE | _PAGE_WRENABLE) |
| 118 | #define PAGE_KERNEL_CI __pgprot(_PAGE_PRESENT | _PAGE_ACCESSED | \ |
| 119 | _PAGE_WRENABLE | _PAGE_NO_CACHE | _PAGE_GUARDED) |
| 120 | #define PAGE_KERNEL_EXEC __pgprot(_PAGE_BASE | _PAGE_WRENABLE | _PAGE_EXEC) |
| 121 | |
| 122 | #define PAGE_AGP __pgprot(_PAGE_BASE | _PAGE_WRENABLE | _PAGE_NO_CACHE) |
| 123 | #define HAVE_PAGE_AGP |
| 124 | |
| 125 | /* |
| 126 | * This bit in a hardware PTE indicates that the page is *not* executable. |
| 127 | */ |
| 128 | #define HW_NO_EXEC _PAGE_EXEC |
| 129 | |
| 130 | /* |
| 131 | * POWER4 and newer have per page execute protection, older chips can only |
| 132 | * do this on a segment (256MB) basis. |
| 133 | * |
| 134 | * Also, write permissions imply read permissions. |
| 135 | * This is the closest we can get.. |
| 136 | * |
| 137 | * Note due to the way vm flags are laid out, the bits are XWR |
| 138 | */ |
| 139 | #define __P000 PAGE_NONE |
| 140 | #define __P001 PAGE_READONLY |
| 141 | #define __P010 PAGE_COPY |
| 142 | #define __P011 PAGE_COPY |
| 143 | #define __P100 PAGE_READONLY_X |
| 144 | #define __P101 PAGE_READONLY_X |
| 145 | #define __P110 PAGE_COPY_X |
| 146 | #define __P111 PAGE_COPY_X |
| 147 | |
| 148 | #define __S000 PAGE_NONE |
| 149 | #define __S001 PAGE_READONLY |
| 150 | #define __S010 PAGE_SHARED |
| 151 | #define __S011 PAGE_SHARED |
| 152 | #define __S100 PAGE_READONLY_X |
| 153 | #define __S101 PAGE_READONLY_X |
| 154 | #define __S110 PAGE_SHARED_X |
| 155 | #define __S111 PAGE_SHARED_X |
| 156 | |
| 157 | #ifndef __ASSEMBLY__ |
| 158 | |
| 159 | /* |
| 160 | * ZERO_PAGE is a global shared page that is always zero: used |
| 161 | * for zero-mapped memory areas etc.. |
| 162 | */ |
| 163 | extern unsigned long empty_zero_page[PAGE_SIZE/sizeof(unsigned long)]; |
| 164 | #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page)) |
| 165 | #endif /* __ASSEMBLY__ */ |
| 166 | |
| 167 | /* shift to put page number into pte */ |
| 168 | #define PTE_SHIFT (17) |
| 169 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 170 | #ifdef CONFIG_HUGETLB_PAGE |
| 171 | |
| 172 | #ifndef __ASSEMBLY__ |
| 173 | int hash_huge_page(struct mm_struct *mm, unsigned long access, |
| 174 | unsigned long ea, unsigned long vsid, int local); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 175 | #endif /* __ASSEMBLY__ */ |
| 176 | |
| 177 | #define HAVE_ARCH_UNMAPPED_AREA |
| 178 | #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN |
| 179 | #else |
| 180 | |
| 181 | #define hash_huge_page(mm,a,ea,vsid,local) -1 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 182 | |
| 183 | #endif |
| 184 | |
| 185 | #ifndef __ASSEMBLY__ |
| 186 | |
| 187 | /* |
| 188 | * Conversion functions: convert a page and protection to a page entry, |
| 189 | * and a page entry and page directory to the page they refer to. |
| 190 | * |
| 191 | * mk_pte takes a (struct page *) as input |
| 192 | */ |
| 193 | #define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot)) |
| 194 | |
David Gibson | 1f8d419 | 2005-05-05 16:15:13 -0700 | [diff] [blame] | 195 | static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot) |
| 196 | { |
| 197 | pte_t pte; |
| 198 | |
| 199 | |
| 200 | pte_val(pte) = (pfn << PTE_SHIFT) | pgprot_val(pgprot); |
| 201 | return pte; |
| 202 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 203 | |
| 204 | #define pte_modify(_pte, newprot) \ |
| 205 | (__pte((pte_val(_pte) & _PAGE_CHG_MASK) | pgprot_val(newprot))) |
| 206 | |
| 207 | #define pte_none(pte) ((pte_val(pte) & ~_PAGE_HPTEFLAGS) == 0) |
| 208 | #define pte_present(pte) (pte_val(pte) & _PAGE_PRESENT) |
| 209 | |
| 210 | /* pte_clear moved to later in this file */ |
| 211 | |
| 212 | #define pte_pfn(x) ((unsigned long)((pte_val(x) >> PTE_SHIFT))) |
| 213 | #define pte_page(x) pfn_to_page(pte_pfn(x)) |
| 214 | |
David Gibson | e28f7fa | 2005-08-05 19:39:06 +1000 | [diff] [blame] | 215 | #define pmd_set(pmdp, ptep) ({BUG_ON((u64)ptep < KERNELBASE); pmd_val(*(pmdp)) = (unsigned long)(ptep);}) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 216 | #define pmd_none(pmd) (!pmd_val(pmd)) |
| 217 | #define pmd_bad(pmd) (pmd_val(pmd) == 0) |
| 218 | #define pmd_present(pmd) (pmd_val(pmd) != 0) |
| 219 | #define pmd_clear(pmdp) (pmd_val(*(pmdp)) = 0) |
David Gibson | e28f7fa | 2005-08-05 19:39:06 +1000 | [diff] [blame] | 220 | #define pmd_page_kernel(pmd) (pmd_val(pmd)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 221 | #define pmd_page(pmd) virt_to_page(pmd_page_kernel(pmd)) |
Benjamin Herrenschmidt | 58366af | 2005-05-01 08:58:44 -0700 | [diff] [blame] | 222 | |
David Gibson | e28f7fa | 2005-08-05 19:39:06 +1000 | [diff] [blame] | 223 | #define pud_set(pudp, pmdp) (pud_val(*(pudp)) = (unsigned long)(pmdp)) |
Benjamin Herrenschmidt | 58366af | 2005-05-01 08:58:44 -0700 | [diff] [blame] | 224 | #define pud_none(pud) (!pud_val(pud)) |
David Gibson | e28f7fa | 2005-08-05 19:39:06 +1000 | [diff] [blame] | 225 | #define pud_bad(pud) ((pud_val(pud)) == 0) |
| 226 | #define pud_present(pud) (pud_val(pud) != 0) |
| 227 | #define pud_clear(pudp) (pud_val(*(pudp)) = 0) |
| 228 | #define pud_page(pud) (pud_val(pud)) |
| 229 | |
| 230 | #define pgd_set(pgdp, pudp) ({pgd_val(*(pgdp)) = (unsigned long)(pudp);}) |
| 231 | #define pgd_none(pgd) (!pgd_val(pgd)) |
| 232 | #define pgd_bad(pgd) (pgd_val(pgd) == 0) |
| 233 | #define pgd_present(pgd) (pgd_val(pgd) != 0) |
| 234 | #define pgd_clear(pgdp) (pgd_val(*(pgdp)) = 0) |
| 235 | #define pgd_page(pgd) (pgd_val(pgd)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 236 | |
| 237 | /* |
| 238 | * Find an entry in a page-table-directory. We combine the address region |
| 239 | * (the high order N bits) and the pgd portion of the address. |
| 240 | */ |
| 241 | /* to avoid overflow in free_pgtables we don't use PTRS_PER_PGD here */ |
David Gibson | e28f7fa | 2005-08-05 19:39:06 +1000 | [diff] [blame] | 242 | #define pgd_index(address) (((address) >> (PGDIR_SHIFT)) & 0x1ff) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 243 | |
| 244 | #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address)) |
| 245 | |
David Gibson | e28f7fa | 2005-08-05 19:39:06 +1000 | [diff] [blame] | 246 | #define pud_offset(pgdp, addr) \ |
| 247 | (((pud_t *) pgd_page(*(pgdp))) + (((addr) >> PUD_SHIFT) & (PTRS_PER_PUD - 1))) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 248 | |
David Gibson | e28f7fa | 2005-08-05 19:39:06 +1000 | [diff] [blame] | 249 | #define pmd_offset(pudp,addr) \ |
| 250 | (((pmd_t *) pud_page(*(pudp))) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1))) |
| 251 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 252 | #define pte_offset_kernel(dir,addr) \ |
David Gibson | e28f7fa | 2005-08-05 19:39:06 +1000 | [diff] [blame] | 253 | (((pte_t *) pmd_page_kernel(*(dir))) + (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 254 | |
| 255 | #define pte_offset_map(dir,addr) pte_offset_kernel((dir), (addr)) |
| 256 | #define pte_offset_map_nested(dir,addr) pte_offset_kernel((dir), (addr)) |
| 257 | #define pte_unmap(pte) do { } while(0) |
| 258 | #define pte_unmap_nested(pte) do { } while(0) |
| 259 | |
| 260 | /* to find an entry in a kernel page-table-directory */ |
| 261 | /* This now only contains the vmalloc pages */ |
| 262 | #define pgd_offset_k(address) pgd_offset(&init_mm, address) |
| 263 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 264 | /* |
| 265 | * The following only work if pte_present() is true. |
| 266 | * Undefined behaviour if not.. |
| 267 | */ |
| 268 | static inline int pte_read(pte_t pte) { return pte_val(pte) & _PAGE_USER;} |
| 269 | static inline int pte_write(pte_t pte) { return pte_val(pte) & _PAGE_RW;} |
| 270 | static inline int pte_exec(pte_t pte) { return pte_val(pte) & _PAGE_EXEC;} |
| 271 | static inline int pte_dirty(pte_t pte) { return pte_val(pte) & _PAGE_DIRTY;} |
| 272 | static inline int pte_young(pte_t pte) { return pte_val(pte) & _PAGE_ACCESSED;} |
| 273 | static inline int pte_file(pte_t pte) { return pte_val(pte) & _PAGE_FILE;} |
| 274 | static inline int pte_huge(pte_t pte) { return pte_val(pte) & _PAGE_HUGE;} |
| 275 | |
| 276 | static inline void pte_uncache(pte_t pte) { pte_val(pte) |= _PAGE_NO_CACHE; } |
| 277 | static inline void pte_cache(pte_t pte) { pte_val(pte) &= ~_PAGE_NO_CACHE; } |
| 278 | |
| 279 | static inline pte_t pte_rdprotect(pte_t pte) { |
| 280 | pte_val(pte) &= ~_PAGE_USER; return pte; } |
| 281 | static inline pte_t pte_exprotect(pte_t pte) { |
| 282 | pte_val(pte) &= ~_PAGE_EXEC; return pte; } |
| 283 | static inline pte_t pte_wrprotect(pte_t pte) { |
| 284 | pte_val(pte) &= ~(_PAGE_RW); return pte; } |
| 285 | static inline pte_t pte_mkclean(pte_t pte) { |
| 286 | pte_val(pte) &= ~(_PAGE_DIRTY); return pte; } |
| 287 | static inline pte_t pte_mkold(pte_t pte) { |
| 288 | pte_val(pte) &= ~_PAGE_ACCESSED; return pte; } |
| 289 | |
| 290 | static inline pte_t pte_mkread(pte_t pte) { |
| 291 | pte_val(pte) |= _PAGE_USER; return pte; } |
| 292 | static inline pte_t pte_mkexec(pte_t pte) { |
| 293 | pte_val(pte) |= _PAGE_USER | _PAGE_EXEC; return pte; } |
| 294 | static inline pte_t pte_mkwrite(pte_t pte) { |
| 295 | pte_val(pte) |= _PAGE_RW; return pte; } |
| 296 | static inline pte_t pte_mkdirty(pte_t pte) { |
| 297 | pte_val(pte) |= _PAGE_DIRTY; return pte; } |
| 298 | static inline pte_t pte_mkyoung(pte_t pte) { |
| 299 | pte_val(pte) |= _PAGE_ACCESSED; return pte; } |
| 300 | static inline pte_t pte_mkhuge(pte_t pte) { |
| 301 | pte_val(pte) |= _PAGE_HUGE; return pte; } |
| 302 | |
| 303 | /* Atomic PTE updates */ |
| 304 | static inline unsigned long pte_update(pte_t *p, unsigned long clr) |
| 305 | { |
| 306 | unsigned long old, tmp; |
| 307 | |
| 308 | __asm__ __volatile__( |
| 309 | "1: ldarx %0,0,%3 # pte_update\n\ |
| 310 | andi. %1,%0,%6\n\ |
| 311 | bne- 1b \n\ |
| 312 | andc %1,%0,%4 \n\ |
| 313 | stdcx. %1,0,%3 \n\ |
| 314 | bne- 1b" |
| 315 | : "=&r" (old), "=&r" (tmp), "=m" (*p) |
| 316 | : "r" (p), "r" (clr), "m" (*p), "i" (_PAGE_BUSY) |
| 317 | : "cc" ); |
| 318 | return old; |
| 319 | } |
| 320 | |
| 321 | /* PTE updating functions, this function puts the PTE in the |
| 322 | * batch, doesn't actually triggers the hash flush immediately, |
| 323 | * you need to call flush_tlb_pending() to do that. |
| 324 | */ |
| 325 | extern void hpte_update(struct mm_struct *mm, unsigned long addr, unsigned long pte, |
| 326 | int wrprot); |
| 327 | |
| 328 | static inline int __ptep_test_and_clear_young(struct mm_struct *mm, unsigned long addr, pte_t *ptep) |
| 329 | { |
| 330 | unsigned long old; |
| 331 | |
| 332 | if ((pte_val(*ptep) & (_PAGE_ACCESSED | _PAGE_HASHPTE)) == 0) |
| 333 | return 0; |
| 334 | old = pte_update(ptep, _PAGE_ACCESSED); |
| 335 | if (old & _PAGE_HASHPTE) { |
| 336 | hpte_update(mm, addr, old, 0); |
| 337 | flush_tlb_pending(); |
| 338 | } |
| 339 | return (old & _PAGE_ACCESSED) != 0; |
| 340 | } |
| 341 | #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG |
| 342 | #define ptep_test_and_clear_young(__vma, __addr, __ptep) \ |
| 343 | ({ \ |
| 344 | int __r; \ |
| 345 | __r = __ptep_test_and_clear_young((__vma)->vm_mm, __addr, __ptep); \ |
| 346 | __r; \ |
| 347 | }) |
| 348 | |
| 349 | /* |
| 350 | * On RW/DIRTY bit transitions we can avoid flushing the hpte. For the |
| 351 | * moment we always flush but we need to fix hpte_update and test if the |
| 352 | * optimisation is worth it. |
| 353 | */ |
| 354 | static inline int __ptep_test_and_clear_dirty(struct mm_struct *mm, unsigned long addr, pte_t *ptep) |
| 355 | { |
| 356 | unsigned long old; |
| 357 | |
| 358 | if ((pte_val(*ptep) & _PAGE_DIRTY) == 0) |
| 359 | return 0; |
| 360 | old = pte_update(ptep, _PAGE_DIRTY); |
| 361 | if (old & _PAGE_HASHPTE) |
| 362 | hpte_update(mm, addr, old, 0); |
| 363 | return (old & _PAGE_DIRTY) != 0; |
| 364 | } |
| 365 | #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY |
| 366 | #define ptep_test_and_clear_dirty(__vma, __addr, __ptep) \ |
| 367 | ({ \ |
| 368 | int __r; \ |
| 369 | __r = __ptep_test_and_clear_dirty((__vma)->vm_mm, __addr, __ptep); \ |
| 370 | __r; \ |
| 371 | }) |
| 372 | |
| 373 | #define __HAVE_ARCH_PTEP_SET_WRPROTECT |
| 374 | static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep) |
| 375 | { |
| 376 | unsigned long old; |
| 377 | |
| 378 | if ((pte_val(*ptep) & _PAGE_RW) == 0) |
| 379 | return; |
| 380 | old = pte_update(ptep, _PAGE_RW); |
| 381 | if (old & _PAGE_HASHPTE) |
| 382 | hpte_update(mm, addr, old, 0); |
| 383 | } |
| 384 | |
| 385 | /* |
| 386 | * We currently remove entries from the hashtable regardless of whether |
| 387 | * the entry was young or dirty. The generic routines only flush if the |
| 388 | * entry was young or dirty which is not good enough. |
| 389 | * |
| 390 | * We should be more intelligent about this but for the moment we override |
| 391 | * these functions and force a tlb flush unconditionally |
| 392 | */ |
| 393 | #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH |
| 394 | #define ptep_clear_flush_young(__vma, __address, __ptep) \ |
| 395 | ({ \ |
| 396 | int __young = __ptep_test_and_clear_young((__vma)->vm_mm, __address, \ |
| 397 | __ptep); \ |
| 398 | __young; \ |
| 399 | }) |
| 400 | |
| 401 | #define __HAVE_ARCH_PTEP_CLEAR_DIRTY_FLUSH |
| 402 | #define ptep_clear_flush_dirty(__vma, __address, __ptep) \ |
| 403 | ({ \ |
| 404 | int __dirty = __ptep_test_and_clear_dirty((__vma)->vm_mm, __address, \ |
| 405 | __ptep); \ |
| 406 | flush_tlb_page(__vma, __address); \ |
| 407 | __dirty; \ |
| 408 | }) |
| 409 | |
| 410 | #define __HAVE_ARCH_PTEP_GET_AND_CLEAR |
| 411 | static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep) |
| 412 | { |
| 413 | unsigned long old = pte_update(ptep, ~0UL); |
| 414 | |
| 415 | if (old & _PAGE_HASHPTE) |
| 416 | hpte_update(mm, addr, old, 0); |
| 417 | return __pte(old); |
| 418 | } |
| 419 | |
| 420 | static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t * ptep) |
| 421 | { |
| 422 | unsigned long old = pte_update(ptep, ~0UL); |
| 423 | |
| 424 | if (old & _PAGE_HASHPTE) |
| 425 | hpte_update(mm, addr, old, 0); |
| 426 | } |
| 427 | |
| 428 | /* |
| 429 | * set_pte stores a linux PTE into the linux page table. |
| 430 | */ |
| 431 | static inline void set_pte_at(struct mm_struct *mm, unsigned long addr, |
| 432 | pte_t *ptep, pte_t pte) |
| 433 | { |
| 434 | if (pte_present(*ptep)) { |
| 435 | pte_clear(mm, addr, ptep); |
| 436 | flush_tlb_pending(); |
| 437 | } |
David Gibson | 1f8d419 | 2005-05-05 16:15:13 -0700 | [diff] [blame] | 438 | *ptep = __pte(pte_val(pte) & ~_PAGE_HPTEFLAGS); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 439 | } |
| 440 | |
| 441 | /* Set the dirty and/or accessed bits atomically in a linux PTE, this |
| 442 | * function doesn't need to flush the hash entry |
| 443 | */ |
| 444 | #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS |
| 445 | static inline void __ptep_set_access_flags(pte_t *ptep, pte_t entry, int dirty) |
| 446 | { |
| 447 | unsigned long bits = pte_val(entry) & |
| 448 | (_PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_RW | _PAGE_EXEC); |
| 449 | unsigned long old, tmp; |
| 450 | |
| 451 | __asm__ __volatile__( |
| 452 | "1: ldarx %0,0,%4\n\ |
| 453 | andi. %1,%0,%6\n\ |
| 454 | bne- 1b \n\ |
| 455 | or %0,%3,%0\n\ |
| 456 | stdcx. %0,0,%4\n\ |
| 457 | bne- 1b" |
| 458 | :"=&r" (old), "=&r" (tmp), "=m" (*ptep) |
| 459 | :"r" (bits), "r" (ptep), "m" (*ptep), "i" (_PAGE_BUSY) |
| 460 | :"cc"); |
| 461 | } |
| 462 | #define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \ |
| 463 | do { \ |
| 464 | __ptep_set_access_flags(__ptep, __entry, __dirty); \ |
| 465 | flush_tlb_page_nohash(__vma, __address); \ |
| 466 | } while(0) |
| 467 | |
| 468 | /* |
| 469 | * Macro to mark a page protection value as "uncacheable". |
| 470 | */ |
| 471 | #define pgprot_noncached(prot) (__pgprot(pgprot_val(prot) | _PAGE_NO_CACHE | _PAGE_GUARDED)) |
| 472 | |
| 473 | struct file; |
| 474 | extern pgprot_t phys_mem_access_prot(struct file *file, unsigned long addr, |
| 475 | unsigned long size, pgprot_t vma_prot); |
| 476 | #define __HAVE_PHYS_MEM_ACCESS_PROT |
| 477 | |
| 478 | #define __HAVE_ARCH_PTE_SAME |
| 479 | #define pte_same(A,B) (((pte_val(A) ^ pte_val(B)) & ~_PAGE_HPTEFLAGS) == 0) |
| 480 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 481 | #define pmd_ERROR(e) \ |
David Gibson | e28f7fa | 2005-08-05 19:39:06 +1000 | [diff] [blame] | 482 | printk("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pmd_val(e)) |
| 483 | #define pud_ERROR(e) \ |
| 484 | printk("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pud_val(e)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 485 | #define pgd_ERROR(e) \ |
David Gibson | e28f7fa | 2005-08-05 19:39:06 +1000 | [diff] [blame] | 486 | printk("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 487 | |
David Gibson | 1f8d419 | 2005-05-05 16:15:13 -0700 | [diff] [blame] | 488 | extern pgd_t swapper_pg_dir[]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 489 | |
| 490 | extern void paging_init(void); |
| 491 | |
Andrew Morton | b74d0bd | 2005-08-10 13:53:41 +1000 | [diff] [blame^] | 492 | #ifdef CONFIG_HUGETLB_PAGE |
Hugh Dickins | 3bf5ee9 | 2005-04-19 13:29:16 -0700 | [diff] [blame] | 493 | #define hugetlb_free_pgd_range(tlb, addr, end, floor, ceiling) \ |
David Gibson | e28f7fa | 2005-08-05 19:39:06 +1000 | [diff] [blame] | 494 | free_pgd_range(tlb, addr, end, floor, ceiling) |
Andrew Morton | b74d0bd | 2005-08-10 13:53:41 +1000 | [diff] [blame^] | 495 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 496 | |
| 497 | /* |
| 498 | * This gets called at the end of handling a page fault, when |
| 499 | * the kernel has put a new PTE into the page table for the process. |
| 500 | * We use it to put a corresponding HPTE into the hash table |
| 501 | * ahead of time, instead of waiting for the inevitable extra |
| 502 | * hash-table miss exception. |
| 503 | */ |
| 504 | struct vm_area_struct; |
| 505 | extern void update_mmu_cache(struct vm_area_struct *, unsigned long, pte_t); |
| 506 | |
| 507 | /* Encode and de-code a swap entry */ |
| 508 | #define __swp_type(entry) (((entry).val >> 1) & 0x3f) |
| 509 | #define __swp_offset(entry) ((entry).val >> 8) |
| 510 | #define __swp_entry(type, offset) ((swp_entry_t) { ((type) << 1) | ((offset) << 8) }) |
| 511 | #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) >> PTE_SHIFT }) |
| 512 | #define __swp_entry_to_pte(x) ((pte_t) { (x).val << PTE_SHIFT }) |
| 513 | #define pte_to_pgoff(pte) (pte_val(pte) >> PTE_SHIFT) |
| 514 | #define pgoff_to_pte(off) ((pte_t) {((off) << PTE_SHIFT)|_PAGE_FILE}) |
| 515 | #define PTE_FILE_MAX_BITS (BITS_PER_LONG - PTE_SHIFT) |
| 516 | |
| 517 | /* |
| 518 | * kern_addr_valid is intended to indicate whether an address is a valid |
| 519 | * kernel address. Most 32-bit archs define it as always true (like this) |
| 520 | * but most 64-bit archs actually perform a test. What should we do here? |
| 521 | * The only use is in fs/ncpfs/dir.c |
| 522 | */ |
| 523 | #define kern_addr_valid(addr) (1) |
| 524 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 525 | #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \ |
| 526 | remap_pfn_range(vma, vaddr, pfn, size, prot) |
| 527 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 528 | void pgtable_cache_init(void); |
| 529 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 530 | /* |
| 531 | * find_linux_pte returns the address of a linux pte for a given |
| 532 | * effective address and directory. If not found, it returns zero. |
| 533 | */ |
| 534 | static inline pte_t *find_linux_pte(pgd_t *pgdir, unsigned long ea) |
| 535 | { |
| 536 | pgd_t *pg; |
Benjamin Herrenschmidt | 58366af | 2005-05-01 08:58:44 -0700 | [diff] [blame] | 537 | pud_t *pu; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 538 | pmd_t *pm; |
| 539 | pte_t *pt = NULL; |
| 540 | pte_t pte; |
| 541 | |
| 542 | pg = pgdir + pgd_index(ea); |
| 543 | if (!pgd_none(*pg)) { |
Benjamin Herrenschmidt | 58366af | 2005-05-01 08:58:44 -0700 | [diff] [blame] | 544 | pu = pud_offset(pg, ea); |
| 545 | if (!pud_none(*pu)) { |
| 546 | pm = pmd_offset(pu, ea); |
| 547 | if (pmd_present(*pm)) { |
| 548 | pt = pte_offset_kernel(pm, ea); |
| 549 | pte = *pt; |
| 550 | if (!pte_present(pte)) |
| 551 | pt = NULL; |
| 552 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 553 | } |
| 554 | } |
| 555 | |
| 556 | return pt; |
| 557 | } |
| 558 | |
| 559 | #include <asm-generic/pgtable.h> |
| 560 | |
| 561 | #endif /* __ASSEMBLY__ */ |
| 562 | |
| 563 | #endif /* _PPC64_PGTABLE_H */ |