Merge git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / include / asm-powerpc / pgtable.h
blobe9f1f4627e6bab35cd34910e3bf4bb05763b85ff
1 #ifndef _ASM_POWERPC_PGTABLE_H
2 #define _ASM_POWERPC_PGTABLE_H
3 #ifdef __KERNEL__
5 #ifndef CONFIG_PPC64
6 #include <asm-ppc/pgtable.h>
7 #else
9 /*
10 * This file contains the functions and defines necessary to modify and use
11 * the ppc64 hashed page table.
14 #ifndef __ASSEMBLY__
15 #include <linux/config.h>
16 #include <linux/stddef.h>
17 #include <asm/processor.h> /* For TASK_SIZE */
18 #include <asm/mmu.h>
19 #include <asm/page.h>
20 #include <asm/tlbflush.h>
21 struct mm_struct;
22 #endif /* __ASSEMBLY__ */
24 #ifdef CONFIG_PPC_64K_PAGES
25 #include <asm/pgtable-64k.h>
26 #else
27 #include <asm/pgtable-4k.h>
28 #endif
30 #define FIRST_USER_ADDRESS 0
33 * Size of EA range mapped by our pagetables.
35 #define PGTABLE_EADDR_SIZE (PTE_INDEX_SIZE + PMD_INDEX_SIZE + \
36 PUD_INDEX_SIZE + PGD_INDEX_SIZE + PAGE_SHIFT)
37 #define PGTABLE_RANGE (1UL << PGTABLE_EADDR_SIZE)
39 #if TASK_SIZE_USER64 > PGTABLE_RANGE
40 #error TASK_SIZE_USER64 exceeds pagetable range
41 #endif
43 #if TASK_SIZE_USER64 > (1UL << (USER_ESID_BITS + SID_SHIFT))
44 #error TASK_SIZE_USER64 exceeds user VSID range
45 #endif
48 * Define the address range of the vmalloc VM area.
50 #define VMALLOC_START (0xD000000000000000ul)
51 #define VMALLOC_SIZE (0x80000000000UL)
52 #define VMALLOC_END (VMALLOC_START + VMALLOC_SIZE)
55 * Define the address range of the imalloc VM area.
57 #define PHBS_IO_BASE VMALLOC_END
58 #define IMALLOC_BASE (PHBS_IO_BASE + 0x80000000ul) /* Reserve 2 gigs for PHBs */
59 #define IMALLOC_END (VMALLOC_START + PGTABLE_RANGE)
62 * Region IDs
64 #define REGION_SHIFT 60UL
65 #define REGION_MASK (0xfUL << REGION_SHIFT)
66 #define REGION_ID(ea) (((unsigned long)(ea)) >> REGION_SHIFT)
68 #define VMALLOC_REGION_ID (REGION_ID(VMALLOC_START))
69 #define KERNEL_REGION_ID (REGION_ID(PAGE_OFFSET))
70 #define USER_REGION_ID (0UL)
73 * Common bits in a linux-style PTE. These match the bits in the
74 * (hardware-defined) PowerPC PTE as closely as possible. Additional
75 * bits may be defined in pgtable-*.h
77 #define _PAGE_PRESENT 0x0001 /* software: pte contains a translation */
78 #define _PAGE_USER 0x0002 /* matches one of the PP bits */
79 #define _PAGE_FILE 0x0002 /* (!present only) software: pte holds file offset */
80 #define _PAGE_EXEC 0x0004 /* No execute on POWER4 and newer (we invert) */
81 #define _PAGE_GUARDED 0x0008
82 #define _PAGE_COHERENT 0x0010 /* M: enforce memory coherence (SMP systems) */
83 #define _PAGE_NO_CACHE 0x0020 /* I: cache inhibit */
84 #define _PAGE_WRITETHRU 0x0040 /* W: cache write-through */
85 #define _PAGE_DIRTY 0x0080 /* C: page changed */
86 #define _PAGE_ACCESSED 0x0100 /* R: page referenced */
87 #define _PAGE_RW 0x0200 /* software: user write access allowed */
88 #define _PAGE_HASHPTE 0x0400 /* software: pte has an associated HPTE */
89 #define _PAGE_BUSY 0x0800 /* software: PTE & hash are busy */
91 #define _PAGE_BASE (_PAGE_PRESENT | _PAGE_ACCESSED | _PAGE_COHERENT)
93 #define _PAGE_WRENABLE (_PAGE_RW | _PAGE_DIRTY)
95 /* __pgprot defined in asm-powerpc/page.h */
96 #define PAGE_NONE __pgprot(_PAGE_PRESENT | _PAGE_ACCESSED)
98 #define PAGE_SHARED __pgprot(_PAGE_BASE | _PAGE_RW | _PAGE_USER)
99 #define PAGE_SHARED_X __pgprot(_PAGE_BASE | _PAGE_RW | _PAGE_USER | _PAGE_EXEC)
100 #define PAGE_COPY __pgprot(_PAGE_BASE | _PAGE_USER)
101 #define PAGE_COPY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC)
102 #define PAGE_READONLY __pgprot(_PAGE_BASE | _PAGE_USER)
103 #define PAGE_READONLY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC)
104 #define PAGE_KERNEL __pgprot(_PAGE_BASE | _PAGE_WRENABLE)
105 #define PAGE_KERNEL_CI __pgprot(_PAGE_PRESENT | _PAGE_ACCESSED | \
106 _PAGE_WRENABLE | _PAGE_NO_CACHE | _PAGE_GUARDED)
107 #define PAGE_KERNEL_EXEC __pgprot(_PAGE_BASE | _PAGE_WRENABLE | _PAGE_EXEC)
109 #define PAGE_AGP __pgprot(_PAGE_BASE | _PAGE_WRENABLE | _PAGE_NO_CACHE)
110 #define HAVE_PAGE_AGP
112 /* PTEIDX nibble */
113 #define _PTEIDX_SECONDARY 0x8
114 #define _PTEIDX_GROUP_IX 0x7
118 * POWER4 and newer have per page execute protection, older chips can only
119 * do this on a segment (256MB) basis.
121 * Also, write permissions imply read permissions.
122 * This is the closest we can get..
124 * Note due to the way vm flags are laid out, the bits are XWR
126 #define __P000 PAGE_NONE
127 #define __P001 PAGE_READONLY
128 #define __P010 PAGE_COPY
129 #define __P011 PAGE_COPY
130 #define __P100 PAGE_READONLY_X
131 #define __P101 PAGE_READONLY_X
132 #define __P110 PAGE_COPY_X
133 #define __P111 PAGE_COPY_X
135 #define __S000 PAGE_NONE
136 #define __S001 PAGE_READONLY
137 #define __S010 PAGE_SHARED
138 #define __S011 PAGE_SHARED
139 #define __S100 PAGE_READONLY_X
140 #define __S101 PAGE_READONLY_X
141 #define __S110 PAGE_SHARED_X
142 #define __S111 PAGE_SHARED_X
144 #ifndef __ASSEMBLY__
147 * ZERO_PAGE is a global shared page that is always zero: used
148 * for zero-mapped memory areas etc..
150 extern unsigned long empty_zero_page[PAGE_SIZE/sizeof(unsigned long)];
151 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
152 #endif /* __ASSEMBLY__ */
154 #ifdef CONFIG_HUGETLB_PAGE
156 #define HAVE_ARCH_UNMAPPED_AREA
157 #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
159 #endif
161 #ifndef __ASSEMBLY__
164 * Conversion functions: convert a page and protection to a page entry,
165 * and a page entry and page directory to the page they refer to.
167 * mk_pte takes a (struct page *) as input
169 #define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
171 static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot)
173 pte_t pte;
176 pte_val(pte) = (pfn << PTE_RPN_SHIFT) | pgprot_val(pgprot);
177 return pte;
180 #define pte_modify(_pte, newprot) \
181 (__pte((pte_val(_pte) & _PAGE_CHG_MASK) | pgprot_val(newprot)))
183 #define pte_none(pte) ((pte_val(pte) & ~_PAGE_HPTEFLAGS) == 0)
184 #define pte_present(pte) (pte_val(pte) & _PAGE_PRESENT)
186 /* pte_clear moved to later in this file */
188 #define pte_pfn(x) ((unsigned long)((pte_val(x)>>PTE_RPN_SHIFT)))
189 #define pte_page(x) pfn_to_page(pte_pfn(x))
191 #define PMD_BAD_BITS (PTE_TABLE_SIZE-1)
192 #define PUD_BAD_BITS (PMD_TABLE_SIZE-1)
194 #define pmd_set(pmdp, pmdval) (pmd_val(*(pmdp)) = (pmdval))
195 #define pmd_none(pmd) (!pmd_val(pmd))
196 #define pmd_bad(pmd) (!is_kernel_addr(pmd_val(pmd)) \
197 || (pmd_val(pmd) & PMD_BAD_BITS))
198 #define pmd_present(pmd) (pmd_val(pmd) != 0)
199 #define pmd_clear(pmdp) (pmd_val(*(pmdp)) = 0)
200 #define pmd_page_kernel(pmd) (pmd_val(pmd) & ~PMD_MASKED_BITS)
201 #define pmd_page(pmd) virt_to_page(pmd_page_kernel(pmd))
203 #define pud_set(pudp, pudval) (pud_val(*(pudp)) = (pudval))
204 #define pud_none(pud) (!pud_val(pud))
205 #define pud_bad(pud) (!is_kernel_addr(pud_val(pud)) \
206 || (pud_val(pud) & PUD_BAD_BITS))
207 #define pud_present(pud) (pud_val(pud) != 0)
208 #define pud_clear(pudp) (pud_val(*(pudp)) = 0)
209 #define pud_page(pud) (pud_val(pud) & ~PUD_MASKED_BITS)
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(*(pudp))) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1)))
225 #define pte_offset_kernel(dir,addr) \
226 (((pte_t *) pmd_page_kernel(*(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) {
272 return 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\
281 andi. %1,%0,%6\n\
282 bne- 1b \n\
283 andc %1,%0,%4 \n\
284 stdcx. %1,0,%3 \n\
285 bne- 1b"
286 : "=&r" (old), "=&r" (tmp), "=m" (*p)
287 : "r" (p), "r" (clr), "m" (*p), "i" (_PAGE_BUSY)
288 : "cc" );
289 return old;
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)
303 unsigned long old;
305 if ((pte_val(*ptep) & (_PAGE_ACCESSED | _PAGE_HASHPTE)) == 0)
306 return 0;
307 old = pte_update(ptep, _PAGE_ACCESSED);
308 if (old & _PAGE_HASHPTE) {
309 hpte_update(mm, addr, ptep, old, 0);
310 flush_tlb_pending();
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) \
316 ({ \
317 int __r; \
318 __r = __ptep_test_and_clear_young((__vma)->vm_mm, __addr, __ptep); \
319 __r; \
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)
330 unsigned long old;
332 if ((pte_val(*ptep) & _PAGE_DIRTY) == 0)
333 return 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) \
341 ({ \
342 int __r; \
343 __r = __ptep_test_and_clear_dirty((__vma)->vm_mm, __addr, __ptep); \
344 __r; \
347 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
348 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
349 pte_t *ptep)
351 unsigned long old;
353 if ((pte_val(*ptep) & _PAGE_RW) == 0)
354 return;
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) \
370 ({ \
371 int __young = __ptep_test_and_clear_young((__vma)->vm_mm, __address, \
372 __ptep); \
373 __young; \
376 #define __HAVE_ARCH_PTEP_CLEAR_DIRTY_FLUSH
377 #define ptep_clear_flush_dirty(__vma, __address, __ptep) \
378 ({ \
379 int __dirty = __ptep_test_and_clear_dirty((__vma)->vm_mm, __address, \
380 __ptep); \
381 flush_tlb_page(__vma, __address); \
382 __dirty; \
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);
393 return __pte(old);
396 static inline void pte_clear(struct mm_struct *mm, unsigned long addr,
397 pte_t * ptep)
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);
413 flush_tlb_pending();
415 pte = __pte(pte_val(pte) & ~_PAGE_HPTEFLAGS);
417 #ifdef CONFIG_PPC_64K_PAGES
418 if (mmu_virtual_psize != MMU_PAGE_64K)
419 pte = __pte(pte_val(pte) | _PAGE_COMBO);
420 #endif /* CONFIG_PPC_64K_PAGES */
422 *ptep = pte;
425 /* Set the dirty and/or accessed bits atomically in a linux PTE, this
426 * function doesn't need to flush the hash entry
428 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
429 static inline void __ptep_set_access_flags(pte_t *ptep, pte_t entry, int dirty)
431 unsigned long bits = pte_val(entry) &
432 (_PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_RW | _PAGE_EXEC);
433 unsigned long old, tmp;
435 __asm__ __volatile__(
436 "1: ldarx %0,0,%4\n\
437 andi. %1,%0,%6\n\
438 bne- 1b \n\
439 or %0,%3,%0\n\
440 stdcx. %0,0,%4\n\
441 bne- 1b"
442 :"=&r" (old), "=&r" (tmp), "=m" (*ptep)
443 :"r" (bits), "r" (ptep), "m" (*ptep), "i" (_PAGE_BUSY)
444 :"cc");
446 #define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
447 do { \
448 __ptep_set_access_flags(__ptep, __entry, __dirty); \
449 flush_tlb_page_nohash(__vma, __address); \
450 } while(0)
453 * Macro to mark a page protection value as "uncacheable".
455 #define pgprot_noncached(prot) (__pgprot(pgprot_val(prot) | _PAGE_NO_CACHE | _PAGE_GUARDED))
457 struct file;
458 extern pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
459 unsigned long size, pgprot_t vma_prot);
460 #define __HAVE_PHYS_MEM_ACCESS_PROT
462 #define __HAVE_ARCH_PTE_SAME
463 #define pte_same(A,B) (((pte_val(A) ^ pte_val(B)) & ~_PAGE_HPTEFLAGS) == 0)
465 #define pte_ERROR(e) \
466 printk("%s:%d: bad pte %08lx.\n", __FILE__, __LINE__, pte_val(e))
467 #define pmd_ERROR(e) \
468 printk("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pmd_val(e))
469 #define pgd_ERROR(e) \
470 printk("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e))
472 extern pgd_t swapper_pg_dir[];
474 extern void paging_init(void);
477 * This gets called at the end of handling a page fault, when
478 * the kernel has put a new PTE into the page table for the process.
479 * We use it to put a corresponding HPTE into the hash table
480 * ahead of time, instead of waiting for the inevitable extra
481 * hash-table miss exception.
483 struct vm_area_struct;
484 extern void update_mmu_cache(struct vm_area_struct *, unsigned long, pte_t);
486 /* Encode and de-code a swap entry */
487 #define __swp_type(entry) (((entry).val >> 1) & 0x3f)
488 #define __swp_offset(entry) ((entry).val >> 8)
489 #define __swp_entry(type, offset) ((swp_entry_t){((type)<< 1)|((offset)<<8)})
490 #define __pte_to_swp_entry(pte) ((swp_entry_t){pte_val(pte) >> PTE_RPN_SHIFT})
491 #define __swp_entry_to_pte(x) ((pte_t) { (x).val << PTE_RPN_SHIFT })
492 #define pte_to_pgoff(pte) (pte_val(pte) >> PTE_RPN_SHIFT)
493 #define pgoff_to_pte(off) ((pte_t) {((off) << PTE_RPN_SHIFT)|_PAGE_FILE})
494 #define PTE_FILE_MAX_BITS (BITS_PER_LONG - PTE_RPN_SHIFT)
497 * kern_addr_valid is intended to indicate whether an address is a valid
498 * kernel address. Most 32-bit archs define it as always true (like this)
499 * but most 64-bit archs actually perform a test. What should we do here?
500 * The only use is in fs/ncpfs/dir.c
502 #define kern_addr_valid(addr) (1)
504 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
505 remap_pfn_range(vma, vaddr, pfn, size, prot)
507 void pgtable_cache_init(void);
510 * find_linux_pte returns the address of a linux pte for a given
511 * effective address and directory. If not found, it returns zero.
512 */static inline pte_t *find_linux_pte(pgd_t *pgdir, unsigned long ea)
514 pgd_t *pg;
515 pud_t *pu;
516 pmd_t *pm;
517 pte_t *pt = NULL;
519 pg = pgdir + pgd_index(ea);
520 if (!pgd_none(*pg)) {
521 pu = pud_offset(pg, ea);
522 if (!pud_none(*pu)) {
523 pm = pmd_offset(pu, ea);
524 if (pmd_present(*pm))
525 pt = pte_offset_kernel(pm, ea);
528 return pt;
531 #include <asm-generic/pgtable.h>
533 #endif /* __ASSEMBLY__ */
535 #endif /* CONFIG_PPC64 */
536 #endif /* __KERNEL__ */
537 #endif /* _ASM_POWERPC_PGTABLE_H */