1 #ifndef _ASM_POWERPC_PGTABLE_H
2 #define _ASM_POWERPC_PGTABLE_H
6 #include <asm-ppc/pgtable.h>
10 * This file contains the functions and defines necessary to modify and use
11 * the ppc64 hashed page table.
15 #include <linux/stddef.h>
16 #include <asm/processor.h> /* For TASK_SIZE */
19 #include <asm/tlbflush.h>
21 #endif /* __ASSEMBLY__ */
23 #ifdef CONFIG_PPC_64K_PAGES
24 #include <asm/pgtable-64k.h>
26 #include <asm/pgtable-4k.h>
29 #define FIRST_USER_ADDRESS 0
32 * Size of EA range mapped by our pagetables.
34 #define PGTABLE_EADDR_SIZE (PTE_INDEX_SIZE + PMD_INDEX_SIZE + \
35 PUD_INDEX_SIZE + PGD_INDEX_SIZE + PAGE_SHIFT)
36 #define PGTABLE_RANGE (1UL << PGTABLE_EADDR_SIZE)
38 #if TASK_SIZE_USER64 > PGTABLE_RANGE
39 #error TASK_SIZE_USER64 exceeds pagetable range
42 #if TASK_SIZE_USER64 > (1UL << (USER_ESID_BITS + SID_SHIFT))
43 #error TASK_SIZE_USER64 exceeds user VSID range
47 * Define the address range of the vmalloc VM area.
49 #define VMALLOC_START ASM_CONST(0xD000000000000000)
50 #define VMALLOC_SIZE ASM_CONST(0x80000000000)
51 #define VMALLOC_END (VMALLOC_START + VMALLOC_SIZE)
54 * Define the address range of the imalloc VM area.
56 #define PHBS_IO_BASE VMALLOC_END
57 #define IMALLOC_BASE (PHBS_IO_BASE + 0x80000000ul) /* Reserve 2 gigs for PHBs */
58 #define IMALLOC_END (VMALLOC_START + PGTABLE_RANGE)
63 #define REGION_SHIFT 60UL
64 #define REGION_MASK (0xfUL << REGION_SHIFT)
65 #define REGION_ID(ea) (((unsigned long)(ea)) >> REGION_SHIFT)
67 #define VMALLOC_REGION_ID (REGION_ID(VMALLOC_START))
68 #define KERNEL_REGION_ID (REGION_ID(PAGE_OFFSET))
69 #define USER_REGION_ID (0UL)
72 * Common bits in a linux-style PTE. These match the bits in the
73 * (hardware-defined) PowerPC PTE as closely as possible. Additional
74 * bits may be defined in pgtable-*.h
76 #define _PAGE_PRESENT 0x0001 /* software: pte contains a translation */
77 #define _PAGE_USER 0x0002 /* matches one of the PP bits */
78 #define _PAGE_FILE 0x0002 /* (!present only) software: pte holds file offset */
79 #define _PAGE_EXEC 0x0004 /* No execute on POWER4 and newer (we invert) */
80 #define _PAGE_GUARDED 0x0008
81 #define _PAGE_COHERENT 0x0010 /* M: enforce memory coherence (SMP systems) */
82 #define _PAGE_NO_CACHE 0x0020 /* I: cache inhibit */
83 #define _PAGE_WRITETHRU 0x0040 /* W: cache write-through */
84 #define _PAGE_DIRTY 0x0080 /* C: page changed */
85 #define _PAGE_ACCESSED 0x0100 /* R: page referenced */
86 #define _PAGE_RW 0x0200 /* software: user write access allowed */
87 #define _PAGE_HASHPTE 0x0400 /* software: pte has an associated HPTE */
88 #define _PAGE_BUSY 0x0800 /* software: PTE & hash are busy */
90 #define _PAGE_BASE (_PAGE_PRESENT | _PAGE_ACCESSED | _PAGE_COHERENT)
92 #define _PAGE_WRENABLE (_PAGE_RW | _PAGE_DIRTY)
94 /* __pgprot defined in asm-powerpc/page.h */
95 #define PAGE_NONE __pgprot(_PAGE_PRESENT | _PAGE_ACCESSED)
97 #define PAGE_SHARED __pgprot(_PAGE_BASE | _PAGE_RW | _PAGE_USER)
98 #define PAGE_SHARED_X __pgprot(_PAGE_BASE | _PAGE_RW | _PAGE_USER | _PAGE_EXEC)
99 #define PAGE_COPY __pgprot(_PAGE_BASE | _PAGE_USER)
100 #define PAGE_COPY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC)
101 #define PAGE_READONLY __pgprot(_PAGE_BASE | _PAGE_USER)
102 #define PAGE_READONLY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC)
103 #define PAGE_KERNEL __pgprot(_PAGE_BASE | _PAGE_WRENABLE)
104 #define PAGE_KERNEL_CI __pgprot(_PAGE_PRESENT | _PAGE_ACCESSED | \
105 _PAGE_WRENABLE | _PAGE_NO_CACHE | _PAGE_GUARDED)
106 #define PAGE_KERNEL_EXEC __pgprot(_PAGE_BASE | _PAGE_WRENABLE | _PAGE_EXEC)
108 #define PAGE_AGP __pgprot(_PAGE_BASE | _PAGE_WRENABLE | _PAGE_NO_CACHE)
109 #define HAVE_PAGE_AGP
112 #define _PTEIDX_SECONDARY 0x8
113 #define _PTEIDX_GROUP_IX 0x7
117 * POWER4 and newer have per page execute protection, older chips can only
118 * do this on a segment (256MB) basis.
120 * Also, write permissions imply read permissions.
121 * This is the closest we can get..
123 * Note due to the way vm flags are laid out, the bits are XWR
125 #define __P000 PAGE_NONE
126 #define __P001 PAGE_READONLY
127 #define __P010 PAGE_COPY
128 #define __P011 PAGE_COPY
129 #define __P100 PAGE_READONLY_X
130 #define __P101 PAGE_READONLY_X
131 #define __P110 PAGE_COPY_X
132 #define __P111 PAGE_COPY_X
134 #define __S000 PAGE_NONE
135 #define __S001 PAGE_READONLY
136 #define __S010 PAGE_SHARED
137 #define __S011 PAGE_SHARED
138 #define __S100 PAGE_READONLY_X
139 #define __S101 PAGE_READONLY_X
140 #define __S110 PAGE_SHARED_X
141 #define __S111 PAGE_SHARED_X
146 * ZERO_PAGE is a global shared page that is always zero: used
147 * for zero-mapped memory areas etc..
149 extern unsigned long empty_zero_page
[PAGE_SIZE
/sizeof(unsigned long)];
150 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
151 #endif /* __ASSEMBLY__ */
153 #ifdef CONFIG_HUGETLB_PAGE
155 #define HAVE_ARCH_UNMAPPED_AREA
156 #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
163 * Conversion functions: convert a page and protection to a page entry,
164 * and a page entry and page directory to the page they refer to.
166 * mk_pte takes a (struct page *) as input
168 #define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
170 static inline pte_t
pfn_pte(unsigned long pfn
, pgprot_t pgprot
)
175 pte_val(pte
) = (pfn
<< PTE_RPN_SHIFT
) | pgprot_val(pgprot
);
179 #define pte_modify(_pte, newprot) \
180 (__pte((pte_val(_pte) & _PAGE_CHG_MASK) | pgprot_val(newprot)))
182 #define pte_none(pte) ((pte_val(pte) & ~_PAGE_HPTEFLAGS) == 0)
183 #define pte_present(pte) (pte_val(pte) & _PAGE_PRESENT)
185 /* pte_clear moved to later in this file */
187 #define pte_pfn(x) ((unsigned long)((pte_val(x)>>PTE_RPN_SHIFT)))
188 #define pte_page(x) pfn_to_page(pte_pfn(x))
190 #define PMD_BAD_BITS (PTE_TABLE_SIZE-1)
191 #define PUD_BAD_BITS (PMD_TABLE_SIZE-1)
193 #define pmd_set(pmdp, pmdval) (pmd_val(*(pmdp)) = (pmdval))
194 #define pmd_none(pmd) (!pmd_val(pmd))
195 #define pmd_bad(pmd) (!is_kernel_addr(pmd_val(pmd)) \
196 || (pmd_val(pmd) & PMD_BAD_BITS))
197 #define pmd_present(pmd) (pmd_val(pmd) != 0)
198 #define pmd_clear(pmdp) (pmd_val(*(pmdp)) = 0)
199 #define pmd_page_vaddr(pmd) (pmd_val(pmd) & ~PMD_MASKED_BITS)
200 #define pmd_page(pmd) virt_to_page(pmd_page_vaddr(pmd))
202 #define pud_set(pudp, pudval) (pud_val(*(pudp)) = (pudval))
203 #define pud_none(pud) (!pud_val(pud))
204 #define pud_bad(pud) (!is_kernel_addr(pud_val(pud)) \
205 || (pud_val(pud) & PUD_BAD_BITS))
206 #define pud_present(pud) (pud_val(pud) != 0)
207 #define pud_clear(pudp) (pud_val(*(pudp)) = 0)
208 #define pud_page_vaddr(pud) (pud_val(pud) & ~PUD_MASKED_BITS)
209 #define pud_page(pud) virt_to_page(pud_page_vaddr(pud))
211 #define pgd_set(pgdp, pudp) ({pgd_val(*(pgdp)) = (unsigned long)(pudp);})
214 * Find an entry in a page-table-directory. We combine the address region
215 * (the high order N bits) and the pgd portion of the address.
217 /* to avoid overflow in free_pgtables we don't use PTRS_PER_PGD here */
218 #define pgd_index(address) (((address) >> (PGDIR_SHIFT)) & 0x1ff)
220 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
222 #define pmd_offset(pudp,addr) \
223 (((pmd_t *) pud_page_vaddr(*(pudp))) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1)))
225 #define pte_offset_kernel(dir,addr) \
226 (((pte_t *) pmd_page_vaddr(*(dir))) + (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)))
228 #define pte_offset_map(dir,addr) pte_offset_kernel((dir), (addr))
229 #define pte_offset_map_nested(dir,addr) pte_offset_kernel((dir), (addr))
230 #define pte_unmap(pte) do { } while(0)
231 #define pte_unmap_nested(pte) do { } while(0)
233 /* to find an entry in a kernel page-table-directory */
234 /* This now only contains the vmalloc pages */
235 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
238 * The following only work if pte_present() is true.
239 * Undefined behaviour if not..
241 static inline int pte_read(pte_t pte
) { return pte_val(pte
) & _PAGE_USER
;}
242 static inline int pte_write(pte_t pte
) { return pte_val(pte
) & _PAGE_RW
;}
243 static inline int pte_exec(pte_t pte
) { return pte_val(pte
) & _PAGE_EXEC
;}
244 static inline int pte_dirty(pte_t pte
) { return pte_val(pte
) & _PAGE_DIRTY
;}
245 static inline int pte_young(pte_t pte
) { return pte_val(pte
) & _PAGE_ACCESSED
;}
246 static inline int pte_file(pte_t pte
) { return pte_val(pte
) & _PAGE_FILE
;}
248 static inline void pte_uncache(pte_t pte
) { pte_val(pte
) |= _PAGE_NO_CACHE
; }
249 static inline void pte_cache(pte_t pte
) { pte_val(pte
) &= ~_PAGE_NO_CACHE
; }
251 static inline pte_t
pte_rdprotect(pte_t pte
) {
252 pte_val(pte
) &= ~_PAGE_USER
; return pte
; }
253 static inline pte_t
pte_exprotect(pte_t pte
) {
254 pte_val(pte
) &= ~_PAGE_EXEC
; return pte
; }
255 static inline pte_t
pte_wrprotect(pte_t pte
) {
256 pte_val(pte
) &= ~(_PAGE_RW
); return pte
; }
257 static inline pte_t
pte_mkclean(pte_t pte
) {
258 pte_val(pte
) &= ~(_PAGE_DIRTY
); return pte
; }
259 static inline pte_t
pte_mkold(pte_t pte
) {
260 pte_val(pte
) &= ~_PAGE_ACCESSED
; return pte
; }
261 static inline pte_t
pte_mkread(pte_t pte
) {
262 pte_val(pte
) |= _PAGE_USER
; return pte
; }
263 static inline pte_t
pte_mkexec(pte_t pte
) {
264 pte_val(pte
) |= _PAGE_USER
| _PAGE_EXEC
; return pte
; }
265 static inline pte_t
pte_mkwrite(pte_t pte
) {
266 pte_val(pte
) |= _PAGE_RW
; return pte
; }
267 static inline pte_t
pte_mkdirty(pte_t pte
) {
268 pte_val(pte
) |= _PAGE_DIRTY
; return pte
; }
269 static inline pte_t
pte_mkyoung(pte_t pte
) {
270 pte_val(pte
) |= _PAGE_ACCESSED
; return pte
; }
271 static inline pte_t
pte_mkhuge(pte_t pte
) {
274 /* Atomic PTE updates */
275 static inline unsigned long pte_update(pte_t
*p
, unsigned long clr
)
277 unsigned long old
, tmp
;
279 __asm__
__volatile__(
280 "1: ldarx %0,0,%3 # pte_update\n\
286 : "=&r" (old
), "=&r" (tmp
), "=m" (*p
)
287 : "r" (p
), "r" (clr
), "m" (*p
), "i" (_PAGE_BUSY
)
292 /* PTE updating functions, this function puts the PTE in the
293 * batch, doesn't actually triggers the hash flush immediately,
294 * you need to call flush_tlb_pending() to do that.
295 * Pass -1 for "normal" size (4K or 64K)
297 extern void hpte_update(struct mm_struct
*mm
, unsigned long addr
,
298 pte_t
*ptep
, unsigned long pte
, int huge
);
300 static inline int __ptep_test_and_clear_young(struct mm_struct
*mm
,
301 unsigned long addr
, pte_t
*ptep
)
305 if ((pte_val(*ptep
) & (_PAGE_ACCESSED
| _PAGE_HASHPTE
)) == 0)
307 old
= pte_update(ptep
, _PAGE_ACCESSED
);
308 if (old
& _PAGE_HASHPTE
) {
309 hpte_update(mm
, addr
, ptep
, old
, 0);
312 return (old
& _PAGE_ACCESSED
) != 0;
314 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
315 #define ptep_test_and_clear_young(__vma, __addr, __ptep) \
318 __r = __ptep_test_and_clear_young((__vma)->vm_mm, __addr, __ptep); \
323 * On RW/DIRTY bit transitions we can avoid flushing the hpte. For the
324 * moment we always flush but we need to fix hpte_update and test if the
325 * optimisation is worth it.
327 static inline int __ptep_test_and_clear_dirty(struct mm_struct
*mm
,
328 unsigned long addr
, pte_t
*ptep
)
332 if ((pte_val(*ptep
) & _PAGE_DIRTY
) == 0)
334 old
= pte_update(ptep
, _PAGE_DIRTY
);
335 if (old
& _PAGE_HASHPTE
)
336 hpte_update(mm
, addr
, ptep
, old
, 0);
337 return (old
& _PAGE_DIRTY
) != 0;
339 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
340 #define ptep_test_and_clear_dirty(__vma, __addr, __ptep) \
343 __r = __ptep_test_and_clear_dirty((__vma)->vm_mm, __addr, __ptep); \
347 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
348 static inline void ptep_set_wrprotect(struct mm_struct
*mm
, unsigned long addr
,
353 if ((pte_val(*ptep
) & _PAGE_RW
) == 0)
355 old
= pte_update(ptep
, _PAGE_RW
);
356 if (old
& _PAGE_HASHPTE
)
357 hpte_update(mm
, addr
, ptep
, old
, 0);
361 * We currently remove entries from the hashtable regardless of whether
362 * the entry was young or dirty. The generic routines only flush if the
363 * entry was young or dirty which is not good enough.
365 * We should be more intelligent about this but for the moment we override
366 * these functions and force a tlb flush unconditionally
368 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
369 #define ptep_clear_flush_young(__vma, __address, __ptep) \
371 int __young = __ptep_test_and_clear_young((__vma)->vm_mm, __address, \
376 #define __HAVE_ARCH_PTEP_CLEAR_DIRTY_FLUSH
377 #define ptep_clear_flush_dirty(__vma, __address, __ptep) \
379 int __dirty = __ptep_test_and_clear_dirty((__vma)->vm_mm, __address, \
381 flush_tlb_page(__vma, __address); \
385 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
386 static inline pte_t
ptep_get_and_clear(struct mm_struct
*mm
,
387 unsigned long addr
, pte_t
*ptep
)
389 unsigned long old
= pte_update(ptep
, ~0UL);
391 if (old
& _PAGE_HASHPTE
)
392 hpte_update(mm
, addr
, ptep
, old
, 0);
396 static inline void pte_clear(struct mm_struct
*mm
, unsigned long addr
,
399 unsigned long old
= pte_update(ptep
, ~0UL);
401 if (old
& _PAGE_HASHPTE
)
402 hpte_update(mm
, addr
, ptep
, old
, 0);
406 * set_pte stores a linux PTE into the linux page table.
408 static inline void set_pte_at(struct mm_struct
*mm
, unsigned long addr
,
409 pte_t
*ptep
, pte_t pte
)
411 if (pte_present(*ptep
)) {
412 pte_clear(mm
, addr
, ptep
);
415 pte
= __pte(pte_val(pte
) & ~_PAGE_HPTEFLAGS
);
419 /* Set the dirty and/or accessed bits atomically in a linux PTE, this
420 * function doesn't need to flush the hash entry
422 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
423 static inline void __ptep_set_access_flags(pte_t
*ptep
, pte_t entry
, int dirty
)
425 unsigned long bits
= pte_val(entry
) &
426 (_PAGE_DIRTY
| _PAGE_ACCESSED
| _PAGE_RW
| _PAGE_EXEC
);
427 unsigned long old
, tmp
;
429 __asm__
__volatile__(
436 :"=&r" (old
), "=&r" (tmp
), "=m" (*ptep
)
437 :"r" (bits
), "r" (ptep
), "m" (*ptep
), "i" (_PAGE_BUSY
)
440 #define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
442 __ptep_set_access_flags(__ptep, __entry, __dirty); \
443 flush_tlb_page_nohash(__vma, __address); \
447 * Macro to mark a page protection value as "uncacheable".
449 #define pgprot_noncached(prot) (__pgprot(pgprot_val(prot) | _PAGE_NO_CACHE | _PAGE_GUARDED))
452 extern pgprot_t
phys_mem_access_prot(struct file
*file
, unsigned long pfn
,
453 unsigned long size
, pgprot_t vma_prot
);
454 #define __HAVE_PHYS_MEM_ACCESS_PROT
456 #define __HAVE_ARCH_PTE_SAME
457 #define pte_same(A,B) (((pte_val(A) ^ pte_val(B)) & ~_PAGE_HPTEFLAGS) == 0)
459 #define pte_ERROR(e) \
460 printk("%s:%d: bad pte %08lx.\n", __FILE__, __LINE__, pte_val(e))
461 #define pmd_ERROR(e) \
462 printk("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pmd_val(e))
463 #define pgd_ERROR(e) \
464 printk("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e))
466 extern pgd_t swapper_pg_dir
[];
468 extern void paging_init(void);
471 * This gets called at the end of handling a page fault, when
472 * the kernel has put a new PTE into the page table for the process.
473 * We use it to put a corresponding HPTE into the hash table
474 * ahead of time, instead of waiting for the inevitable extra
475 * hash-table miss exception.
477 struct vm_area_struct
;
478 extern void update_mmu_cache(struct vm_area_struct
*, unsigned long, pte_t
);
480 /* Encode and de-code a swap entry */
481 #define __swp_type(entry) (((entry).val >> 1) & 0x3f)
482 #define __swp_offset(entry) ((entry).val >> 8)
483 #define __swp_entry(type, offset) ((swp_entry_t){((type)<< 1)|((offset)<<8)})
484 #define __pte_to_swp_entry(pte) ((swp_entry_t){pte_val(pte) >> PTE_RPN_SHIFT})
485 #define __swp_entry_to_pte(x) ((pte_t) { (x).val << PTE_RPN_SHIFT })
486 #define pte_to_pgoff(pte) (pte_val(pte) >> PTE_RPN_SHIFT)
487 #define pgoff_to_pte(off) ((pte_t) {((off) << PTE_RPN_SHIFT)|_PAGE_FILE})
488 #define PTE_FILE_MAX_BITS (BITS_PER_LONG - PTE_RPN_SHIFT)
491 * kern_addr_valid is intended to indicate whether an address is a valid
492 * kernel address. Most 32-bit archs define it as always true (like this)
493 * but most 64-bit archs actually perform a test. What should we do here?
494 * The only use is in fs/ncpfs/dir.c
496 #define kern_addr_valid(addr) (1)
498 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
499 remap_pfn_range(vma, vaddr, pfn, size, prot)
501 void pgtable_cache_init(void);
504 * find_linux_pte returns the address of a linux pte for a given
505 * effective address and directory. If not found, it returns zero.
506 */static inline pte_t
*find_linux_pte(pgd_t
*pgdir
, unsigned long ea
)
513 pg
= pgdir
+ pgd_index(ea
);
514 if (!pgd_none(*pg
)) {
515 pu
= pud_offset(pg
, ea
);
516 if (!pud_none(*pu
)) {
517 pm
= pmd_offset(pu
, ea
);
518 if (pmd_present(*pm
))
519 pt
= pte_offset_kernel(pm
, ea
);
525 #include <asm-generic/pgtable.h>
527 #endif /* __ASSEMBLY__ */
529 #endif /* CONFIG_PPC64 */
530 #endif /* __KERNEL__ */
531 #endif /* _ASM_POWERPC_PGTABLE_H */