1 #ifndef _ASM_POWERPC_PGTABLE_PPC64_H_
2 #define _ASM_POWERPC_PGTABLE_PPC64_H_
4 * This file contains the functions and defines necessary to modify and use
5 * the ppc64 hashed page table.
8 #ifdef CONFIG_PPC_64K_PAGES
9 #include <asm/pgtable-ppc64-64k.h>
11 #include <asm/pgtable-ppc64-4k.h>
14 #define FIRST_USER_ADDRESS 0
17 * Size of EA range mapped by our pagetables.
19 #define PGTABLE_EADDR_SIZE (PTE_INDEX_SIZE + PMD_INDEX_SIZE + \
20 PUD_INDEX_SIZE + PGD_INDEX_SIZE + PAGE_SHIFT)
21 #define PGTABLE_RANGE (ASM_CONST(1) << PGTABLE_EADDR_SIZE)
24 /* Some sanity checking */
25 #if TASK_SIZE_USER64 > PGTABLE_RANGE
26 #error TASK_SIZE_USER64 exceeds pagetable range
29 #ifdef CONFIG_PPC_STD_MMU_64
30 #if TASK_SIZE_USER64 > (1UL << (USER_ESID_BITS + SID_SHIFT))
31 #error TASK_SIZE_USER64 exceeds user VSID range
36 * Define the address range of the kernel non-linear virtual area
39 #ifdef CONFIG_PPC_BOOK3E
40 #define KERN_VIRT_START ASM_CONST(0x8000000000000000)
42 #define KERN_VIRT_START ASM_CONST(0xD000000000000000)
44 #define KERN_VIRT_SIZE PGTABLE_RANGE
47 * The vmalloc space starts at the beginning of that region, and
48 * occupies half of it on hash CPUs and a quarter of it on Book3E
50 #define VMALLOC_START KERN_VIRT_START
51 #ifdef CONFIG_PPC_BOOK3E
52 #define VMALLOC_SIZE (KERN_VIRT_SIZE >> 2)
54 #define VMALLOC_SIZE (KERN_VIRT_SIZE >> 1)
56 #define VMALLOC_END (VMALLOC_START + VMALLOC_SIZE)
59 * The second half of the kernel virtual space is used for IO mappings,
60 * it's itself carved into the PIO region (ISA and PHB IO space) and
63 * ISA_IO_BASE = KERN_IO_START, 64K reserved area
64 * PHB_IO_BASE = ISA_IO_BASE + 64K to ISA_IO_BASE + 2G, PHB IO spaces
65 * IOREMAP_BASE = ISA_IO_BASE + 2G to VMALLOC_START + PGTABLE_RANGE
67 #define KERN_IO_START (KERN_VIRT_START + (KERN_VIRT_SIZE >> 1))
68 #define FULL_IO_SIZE 0x80000000ul
69 #define ISA_IO_BASE (KERN_IO_START)
70 #define ISA_IO_END (KERN_IO_START + 0x10000ul)
71 #define PHB_IO_BASE (ISA_IO_END)
72 #define PHB_IO_END (KERN_IO_START + FULL_IO_SIZE)
73 #define IOREMAP_BASE (PHB_IO_END)
74 #define IOREMAP_END (KERN_VIRT_START + KERN_VIRT_SIZE)
80 #define REGION_SHIFT 60UL
81 #define REGION_MASK (0xfUL << REGION_SHIFT)
82 #define REGION_ID(ea) (((unsigned long)(ea)) >> REGION_SHIFT)
84 #define VMALLOC_REGION_ID (REGION_ID(VMALLOC_START))
85 #define KERNEL_REGION_ID (REGION_ID(PAGE_OFFSET))
86 #define VMEMMAP_REGION_ID (0xfUL)
87 #define USER_REGION_ID (0UL)
90 * Defines the address of the vmemap area, in its own region on
91 * hash table CPUs and after the vmalloc space on Book3E
93 #ifdef CONFIG_PPC_BOOK3E
94 #define VMEMMAP_BASE VMALLOC_END
95 #define VMEMMAP_END KERN_IO_START
97 #define VMEMMAP_BASE (VMEMMAP_REGION_ID << REGION_SHIFT)
99 #define vmemmap ((struct page *)VMEMMAP_BASE)
103 * Include the PTE bits definitions
105 #ifdef CONFIG_PPC_BOOK3S
106 #include <asm/pte-hash64.h>
108 #include <asm/pte-book3e.h>
110 #include <asm/pte-common.h>
112 #ifdef CONFIG_PPC_MM_SLICES
113 #define HAVE_ARCH_UNMAPPED_AREA
114 #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
115 #endif /* CONFIG_PPC_MM_SLICES */
119 #include <linux/stddef.h>
120 #include <asm/tlbflush.h>
123 * This is the default implementation of various PTE accessors, it's
124 * used in all cases except Book3S with 64K pages where we have a
125 * concept of sub-pages
129 #ifdef STRICT_MM_TYPECHECKS
130 #define __real_pte(e,p) ((real_pte_t){(e)})
131 #define __rpte_to_pte(r) ((r).pte)
133 #define __real_pte(e,p) (e)
134 #define __rpte_to_pte(r) (__pte(r))
136 #define __rpte_to_hidx(r,index) (pte_val(__rpte_to_pte(r)) >> 12)
138 #define pte_iterate_hashed_subpages(rpte, psize, va, index, shift) \
141 shift = mmu_psize_defs[psize].shift; \
143 #define pte_iterate_hashed_end() } while(0)
145 #ifdef CONFIG_PPC_HAS_HASH_64K
146 #define pte_pagesize_index(mm, addr, pte) get_slice_psize(mm, addr)
148 #define pte_pagesize_index(mm, addr, pte) MMU_PAGE_4K
151 #endif /* __real_pte */
154 /* pte_clear moved to later in this file */
156 #define PMD_BAD_BITS (PTE_TABLE_SIZE-1)
157 #define PUD_BAD_BITS (PMD_TABLE_SIZE-1)
159 #define pmd_set(pmdp, pmdval) (pmd_val(*(pmdp)) = (pmdval))
160 #define pmd_none(pmd) (!pmd_val(pmd))
161 #define pmd_bad(pmd) (!is_kernel_addr(pmd_val(pmd)) \
162 || (pmd_val(pmd) & PMD_BAD_BITS))
163 #define pmd_present(pmd) (pmd_val(pmd) != 0)
164 #define pmd_clear(pmdp) (pmd_val(*(pmdp)) = 0)
165 #define pmd_page_vaddr(pmd) (pmd_val(pmd) & ~PMD_MASKED_BITS)
166 #define pmd_page(pmd) virt_to_page(pmd_page_vaddr(pmd))
168 #define pud_set(pudp, pudval) (pud_val(*(pudp)) = (pudval))
169 #define pud_none(pud) (!pud_val(pud))
170 #define pud_bad(pud) (!is_kernel_addr(pud_val(pud)) \
171 || (pud_val(pud) & PUD_BAD_BITS))
172 #define pud_present(pud) (pud_val(pud) != 0)
173 #define pud_clear(pudp) (pud_val(*(pudp)) = 0)
174 #define pud_page_vaddr(pud) (pud_val(pud) & ~PUD_MASKED_BITS)
175 #define pud_page(pud) virt_to_page(pud_page_vaddr(pud))
177 #define pgd_set(pgdp, pudp) ({pgd_val(*(pgdp)) = (unsigned long)(pudp);})
180 * Find an entry in a page-table-directory. We combine the address region
181 * (the high order N bits) and the pgd portion of the address.
183 /* to avoid overflow in free_pgtables we don't use PTRS_PER_PGD here */
184 #define pgd_index(address) (((address) >> (PGDIR_SHIFT)) & 0x1ff)
186 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
188 #define pmd_offset(pudp,addr) \
189 (((pmd_t *) pud_page_vaddr(*(pudp))) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1)))
191 #define pte_offset_kernel(dir,addr) \
192 (((pte_t *) pmd_page_vaddr(*(dir))) + (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)))
194 #define pte_offset_map(dir,addr) pte_offset_kernel((dir), (addr))
195 #define pte_offset_map_nested(dir,addr) pte_offset_kernel((dir), (addr))
196 #define pte_unmap(pte) do { } while(0)
197 #define pte_unmap_nested(pte) do { } while(0)
199 /* to find an entry in a kernel page-table-directory */
200 /* This now only contains the vmalloc pages */
201 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
204 /* Atomic PTE updates */
205 static inline unsigned long pte_update(struct mm_struct *mm,
207 pte_t *ptep, unsigned long clr,
210 #ifdef PTE_ATOMIC_UPDATES
211 unsigned long old, tmp;
213 __asm__ __volatile__(
214 "1: ldarx %0,0,%3 # pte_update\n\
220 : "=&r" (old), "=&r" (tmp), "=m" (*ptep)
221 : "r" (ptep), "r" (clr), "m" (*ptep), "i" (_PAGE_BUSY)
224 unsigned long old = pte_val(*ptep);
225 *ptep = __pte(old & ~clr);
227 /* huge pages use the old page table lock */
229 assert_pte_locked(mm, addr);
231 #ifdef CONFIG_PPC_STD_MMU_64
232 if (old & _PAGE_HASHPTE)
233 hpte_need_flush(mm, addr, ptep, old, huge);
239 static inline int __ptep_test_and_clear_young(struct mm_struct *mm,
240 unsigned long addr, pte_t *ptep)
244 if ((pte_val(*ptep) & (_PAGE_ACCESSED | _PAGE_HASHPTE)) == 0)
246 old = pte_update(mm, addr, ptep, _PAGE_ACCESSED, 0);
247 return (old & _PAGE_ACCESSED) != 0;
249 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
250 #define ptep_test_and_clear_young(__vma, __addr, __ptep) \
253 __r = __ptep_test_and_clear_young((__vma)->vm_mm, __addr, __ptep); \
257 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
258 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
263 if ((pte_val(*ptep) & _PAGE_RW) == 0)
265 old = pte_update(mm, addr, ptep, _PAGE_RW, 0);
268 static inline void huge_ptep_set_wrprotect(struct mm_struct *mm,
269 unsigned long addr, pte_t *ptep)
273 if ((pte_val(*ptep) & _PAGE_RW) == 0)
275 old = pte_update(mm, addr, ptep, _PAGE_RW, 1);
279 * We currently remove entries from the hashtable regardless of whether
280 * the entry was young or dirty. The generic routines only flush if the
281 * entry was young or dirty which is not good enough.
283 * We should be more intelligent about this but for the moment we override
284 * these functions and force a tlb flush unconditionally
286 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
287 #define ptep_clear_flush_young(__vma, __address, __ptep) \
289 int __young = __ptep_test_and_clear_young((__vma)->vm_mm, __address, \
294 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
295 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
296 unsigned long addr, pte_t *ptep)
298 unsigned long old = pte_update(mm, addr, ptep, ~0UL, 0);
302 static inline void pte_clear(struct mm_struct *mm, unsigned long addr,
305 pte_update(mm, addr, ptep, ~0UL, 0);
309 /* Set the dirty and/or accessed bits atomically in a linux PTE, this
310 * function doesn't need to flush the hash entry
312 static inline void __ptep_set_access_flags(pte_t *ptep, pte_t entry)
314 unsigned long bits = pte_val(entry) &
315 (_PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_RW |
316 _PAGE_EXEC | _PAGE_HWEXEC);
318 #ifdef PTE_ATOMIC_UPDATES
319 unsigned long old, tmp;
321 __asm__ __volatile__(
328 :"=&r" (old), "=&r" (tmp), "=m" (*ptep)
329 :"r" (bits), "r" (ptep), "m" (*ptep), "i" (_PAGE_BUSY)
332 unsigned long old = pte_val(*ptep);
333 *ptep = __pte(old | bits);
337 #define __HAVE_ARCH_PTE_SAME
338 #define pte_same(A,B) (((pte_val(A) ^ pte_val(B)) & ~_PAGE_HPTEFLAGS) == 0)
340 #define pte_ERROR(e) \
341 printk("%s:%d: bad pte %08lx.\n", __FILE__, __LINE__, pte_val(e))
342 #define pmd_ERROR(e) \
343 printk("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pmd_val(e))
344 #define pgd_ERROR(e) \
345 printk("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e))
347 /* Encode and de-code a swap entry */
348 #define __swp_type(entry) (((entry).val >> 1) & 0x3f)
349 #define __swp_offset(entry) ((entry).val >> 8)
350 #define __swp_entry(type, offset) ((swp_entry_t){((type)<< 1)|((offset)<<8)})
351 #define __pte_to_swp_entry(pte) ((swp_entry_t){pte_val(pte) >> PTE_RPN_SHIFT})
352 #define __swp_entry_to_pte(x) ((pte_t) { (x).val << PTE_RPN_SHIFT })
353 #define pte_to_pgoff(pte) (pte_val(pte) >> PTE_RPN_SHIFT)
354 #define pgoff_to_pte(off) ((pte_t) {((off) << PTE_RPN_SHIFT)|_PAGE_FILE})
355 #define PTE_FILE_MAX_BITS (BITS_PER_LONG - PTE_RPN_SHIFT)
357 void pgtable_cache_init(void);
360 * find_linux_pte returns the address of a linux pte for a given
361 * effective address and directory. If not found, it returns zero.
362 */static inline pte_t *find_linux_pte(pgd_t *pgdir, unsigned long ea)
369 pg = pgdir + pgd_index(ea);
370 if (!pgd_none(*pg)) {
371 pu = pud_offset(pg, ea);
372 if (!pud_none(*pu)) {
373 pm = pmd_offset(pu, ea);
374 if (pmd_present(*pm))
375 pt = pte_offset_kernel(pm, ea);
381 pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long address);
383 #endif /* __ASSEMBLY__ */
385 #endif /* _ASM_POWERPC_PGTABLE_PPC64_H_ */