aboutsummaryrefslogtreecommitdiffstats
path: root/include/asm-powerpc/pgtable-ppc64.h
diff options
context:
space:
mode:
Diffstat (limited to 'include/asm-powerpc/pgtable-ppc64.h')
-rw-r--r--include/asm-powerpc/pgtable-ppc64.h492
1 files changed, 492 insertions, 0 deletions
diff --git a/include/asm-powerpc/pgtable-ppc64.h b/include/asm-powerpc/pgtable-ppc64.h
new file mode 100644
index 000000000000..704c4e669fe0
--- /dev/null
+++ b/include/asm-powerpc/pgtable-ppc64.h
@@ -0,0 +1,492 @@
1#ifndef _ASM_POWERPC_PGTABLE_PPC64_H_
2#define _ASM_POWERPC_PGTABLE_PPC64_H_
3/*
4 * This file contains the functions and defines necessary to modify and use
5 * the ppc64 hashed page table.
6 */
7
8#ifndef __ASSEMBLY__
9#include <linux/stddef.h>
10#include <asm/processor.h> /* For TASK_SIZE */
11#include <asm/mmu.h>
12#include <asm/page.h>
13#include <asm/tlbflush.h>
14struct mm_struct;
15#endif /* __ASSEMBLY__ */
16
17#ifdef CONFIG_PPC_64K_PAGES
18#include <asm/pgtable-64k.h>
19#else
20#include <asm/pgtable-4k.h>
21#endif
22
23#define FIRST_USER_ADDRESS 0
24
25/*
26 * Size of EA range mapped by our pagetables.
27 */
28#define PGTABLE_EADDR_SIZE (PTE_INDEX_SIZE + PMD_INDEX_SIZE + \
29 PUD_INDEX_SIZE + PGD_INDEX_SIZE + PAGE_SHIFT)
30#define PGTABLE_RANGE (1UL << PGTABLE_EADDR_SIZE)
31
32#if TASK_SIZE_USER64 > PGTABLE_RANGE
33#error TASK_SIZE_USER64 exceeds pagetable range
34#endif
35
36#if TASK_SIZE_USER64 > (1UL << (USER_ESID_BITS + SID_SHIFT))
37#error TASK_SIZE_USER64 exceeds user VSID range
38#endif
39
40/*
41 * Define the address range of the vmalloc VM area.
42 */
43#define VMALLOC_START ASM_CONST(0xD000000000000000)
44#define VMALLOC_SIZE ASM_CONST(0x80000000000)
45#define VMALLOC_END (VMALLOC_START + VMALLOC_SIZE)
46
47/*
48 * Define the address range of the imalloc VM area.
49 */
50#define PHBS_IO_BASE VMALLOC_END
51#define IMALLOC_BASE (PHBS_IO_BASE + 0x80000000ul) /* Reserve 2 gigs for PHBs */
52#define IMALLOC_END (VMALLOC_START + PGTABLE_RANGE)
53
54/*
55 * Region IDs
56 */
57#define REGION_SHIFT 60UL
58#define REGION_MASK (0xfUL << REGION_SHIFT)
59#define REGION_ID(ea) (((unsigned long)(ea)) >> REGION_SHIFT)
60
61#define VMALLOC_REGION_ID (REGION_ID(VMALLOC_START))
62#define KERNEL_REGION_ID (REGION_ID(PAGE_OFFSET))
63#define USER_REGION_ID (0UL)
64
65/*
66 * Common bits in a linux-style PTE. These match the bits in the
67 * (hardware-defined) PowerPC PTE as closely as possible. Additional
68 * bits may be defined in pgtable-*.h
69 */
70#define _PAGE_PRESENT 0x0001 /* software: pte contains a translation */
71#define _PAGE_USER 0x0002 /* matches one of the PP bits */
72#define _PAGE_FILE 0x0002 /* (!present only) software: pte holds file offset */
73#define _PAGE_EXEC 0x0004 /* No execute on POWER4 and newer (we invert) */
74#define _PAGE_GUARDED 0x0008
75#define _PAGE_COHERENT 0x0010 /* M: enforce memory coherence (SMP systems) */
76#define _PAGE_NO_CACHE 0x0020 /* I: cache inhibit */
77#define _PAGE_WRITETHRU 0x0040 /* W: cache write-through */
78#define _PAGE_DIRTY 0x0080 /* C: page changed */
79#define _PAGE_ACCESSED 0x0100 /* R: page referenced */
80#define _PAGE_RW 0x0200 /* software: user write access allowed */
81#define _PAGE_HASHPTE 0x0400 /* software: pte has an associated HPTE */
82#define _PAGE_BUSY 0x0800 /* software: PTE & hash are busy */
83
84#define _PAGE_BASE (_PAGE_PRESENT | _PAGE_ACCESSED | _PAGE_COHERENT)
85
86#define _PAGE_WRENABLE (_PAGE_RW | _PAGE_DIRTY)
87
88/* __pgprot defined in asm-powerpc/page.h */
89#define PAGE_NONE __pgprot(_PAGE_PRESENT | _PAGE_ACCESSED)
90
91#define PAGE_SHARED __pgprot(_PAGE_BASE | _PAGE_RW | _PAGE_USER)
92#define PAGE_SHARED_X __pgprot(_PAGE_BASE | _PAGE_RW | _PAGE_USER | _PAGE_EXEC)
93#define PAGE_COPY __pgprot(_PAGE_BASE | _PAGE_USER)
94#define PAGE_COPY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC)
95#define PAGE_READONLY __pgprot(_PAGE_BASE | _PAGE_USER)
96#define PAGE_READONLY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC)
97#define PAGE_KERNEL __pgprot(_PAGE_BASE | _PAGE_WRENABLE)
98#define PAGE_KERNEL_CI __pgprot(_PAGE_PRESENT | _PAGE_ACCESSED | \
99 _PAGE_WRENABLE | _PAGE_NO_CACHE | _PAGE_GUARDED)
100#define PAGE_KERNEL_EXEC __pgprot(_PAGE_BASE | _PAGE_WRENABLE | _PAGE_EXEC)
101
102#define PAGE_AGP __pgprot(_PAGE_BASE | _PAGE_WRENABLE | _PAGE_NO_CACHE)
103#define HAVE_PAGE_AGP
104
105/* PTEIDX nibble */
106#define _PTEIDX_SECONDARY 0x8
107#define _PTEIDX_GROUP_IX 0x7
108
109
110/*
111 * POWER4 and newer have per page execute protection, older chips can only
112 * do this on a segment (256MB) basis.
113 *
114 * Also, write permissions imply read permissions.
115 * This is the closest we can get..
116 *
117 * Note due to the way vm flags are laid out, the bits are XWR
118 */
119#define __P000 PAGE_NONE
120#define __P001 PAGE_READONLY
121#define __P010 PAGE_COPY
122#define __P011 PAGE_COPY
123#define __P100 PAGE_READONLY_X
124#define __P101 PAGE_READONLY_X
125#define __P110 PAGE_COPY_X
126#define __P111 PAGE_COPY_X
127
128#define __S000 PAGE_NONE
129#define __S001 PAGE_READONLY
130#define __S010 PAGE_SHARED
131#define __S011 PAGE_SHARED
132#define __S100 PAGE_READONLY_X
133#define __S101 PAGE_READONLY_X
134#define __S110 PAGE_SHARED_X
135#define __S111 PAGE_SHARED_X
136
137#ifndef __ASSEMBLY__
138
139/*
140 * ZERO_PAGE is a global shared page that is always zero: used
141 * for zero-mapped memory areas etc..
142 */
143extern unsigned long empty_zero_page[PAGE_SIZE/sizeof(unsigned long)];
144#define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
145#endif /* __ASSEMBLY__ */
146
147#ifdef CONFIG_HUGETLB_PAGE
148
149#define HAVE_ARCH_UNMAPPED_AREA
150#define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
151
152#endif
153
154#ifndef __ASSEMBLY__
155
156/*
157 * Conversion functions: convert a page and protection to a page entry,
158 * and a page entry and page directory to the page they refer to.
159 *
160 * mk_pte takes a (struct page *) as input
161 */
162#define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
163
164static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot)
165{
166 pte_t pte;
167
168
169 pte_val(pte) = (pfn << PTE_RPN_SHIFT) | pgprot_val(pgprot);
170 return pte;
171}
172
173#define pte_modify(_pte, newprot) \
174 (__pte((pte_val(_pte) & _PAGE_CHG_MASK) | pgprot_val(newprot)))
175
176#define pte_none(pte) ((pte_val(pte) & ~_PAGE_HPTEFLAGS) == 0)
177#define pte_present(pte) (pte_val(pte) & _PAGE_PRESENT)
178
179/* pte_clear moved to later in this file */
180
181#define pte_pfn(x) ((unsigned long)((pte_val(x)>>PTE_RPN_SHIFT)))
182#define pte_page(x) pfn_to_page(pte_pfn(x))
183
184#define PMD_BAD_BITS (PTE_TABLE_SIZE-1)
185#define PUD_BAD_BITS (PMD_TABLE_SIZE-1)
186
187#define pmd_set(pmdp, pmdval) (pmd_val(*(pmdp)) = (pmdval))
188#define pmd_none(pmd) (!pmd_val(pmd))
189#define pmd_bad(pmd) (!is_kernel_addr(pmd_val(pmd)) \
190 || (pmd_val(pmd) & PMD_BAD_BITS))
191#define pmd_present(pmd) (pmd_val(pmd) != 0)
192#define pmd_clear(pmdp) (pmd_val(*(pmdp)) = 0)
193#define pmd_page_vaddr(pmd) (pmd_val(pmd) & ~PMD_MASKED_BITS)
194#define pmd_page(pmd) virt_to_page(pmd_page_vaddr(pmd))
195
196#define pud_set(pudp, pudval) (pud_val(*(pudp)) = (pudval))
197#define pud_none(pud) (!pud_val(pud))
198#define pud_bad(pud) (!is_kernel_addr(pud_val(pud)) \
199 || (pud_val(pud) & PUD_BAD_BITS))
200#define pud_present(pud) (pud_val(pud) != 0)
201#define pud_clear(pudp) (pud_val(*(pudp)) = 0)
202#define pud_page_vaddr(pud) (pud_val(pud) & ~PUD_MASKED_BITS)
203#define pud_page(pud) virt_to_page(pud_page_vaddr(pud))
204
205#define pgd_set(pgdp, pudp) ({pgd_val(*(pgdp)) = (unsigned long)(pudp);})
206
207/*
208 * Find an entry in a page-table-directory. We combine the address region
209 * (the high order N bits) and the pgd portion of the address.
210 */
211/* to avoid overflow in free_pgtables we don't use PTRS_PER_PGD here */
212#define pgd_index(address) (((address) >> (PGDIR_SHIFT)) & 0x1ff)
213
214#define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
215
216#define pmd_offset(pudp,addr) \
217 (((pmd_t *) pud_page_vaddr(*(pudp))) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1)))
218
219#define pte_offset_kernel(dir,addr) \
220 (((pte_t *) pmd_page_vaddr(*(dir))) + (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)))
221
222#define pte_offset_map(dir,addr) pte_offset_kernel((dir), (addr))
223#define pte_offset_map_nested(dir,addr) pte_offset_kernel((dir), (addr))
224#define pte_unmap(pte) do { } while(0)
225#define pte_unmap_nested(pte) do { } while(0)
226
227/* to find an entry in a kernel page-table-directory */
228/* This now only contains the vmalloc pages */
229#define pgd_offset_k(address) pgd_offset(&init_mm, address)
230
231/*
232 * The following only work if pte_present() is true.
233 * Undefined behaviour if not..
234 */
235static inline int pte_read(pte_t pte) { return pte_val(pte) & _PAGE_USER;}
236static inline int pte_write(pte_t pte) { return pte_val(pte) & _PAGE_RW;}
237static inline int pte_exec(pte_t pte) { return pte_val(pte) & _PAGE_EXEC;}
238static inline int pte_dirty(pte_t pte) { return pte_val(pte) & _PAGE_DIRTY;}
239static inline int pte_young(pte_t pte) { return pte_val(pte) & _PAGE_ACCESSED;}
240static inline int pte_file(pte_t pte) { return pte_val(pte) & _PAGE_FILE;}
241
242static inline void pte_uncache(pte_t pte) { pte_val(pte) |= _PAGE_NO_CACHE; }
243static inline void pte_cache(pte_t pte) { pte_val(pte) &= ~_PAGE_NO_CACHE; }
244
245static inline pte_t pte_rdprotect(pte_t pte) {
246 pte_val(pte) &= ~_PAGE_USER; return pte; }
247static inline pte_t pte_exprotect(pte_t pte) {
248 pte_val(pte) &= ~_PAGE_EXEC; return pte; }
249static inline pte_t pte_wrprotect(pte_t pte) {
250 pte_val(pte) &= ~(_PAGE_RW); return pte; }
251static inline pte_t pte_mkclean(pte_t pte) {
252 pte_val(pte) &= ~(_PAGE_DIRTY); return pte; }
253static inline pte_t pte_mkold(pte_t pte) {
254 pte_val(pte) &= ~_PAGE_ACCESSED; return pte; }
255static inline pte_t pte_mkread(pte_t pte) {
256 pte_val(pte) |= _PAGE_USER; return pte; }
257static inline pte_t pte_mkexec(pte_t pte) {
258 pte_val(pte) |= _PAGE_USER | _PAGE_EXEC; return pte; }
259static inline pte_t pte_mkwrite(pte_t pte) {
260 pte_val(pte) |= _PAGE_RW; return pte; }
261static inline pte_t pte_mkdirty(pte_t pte) {
262 pte_val(pte) |= _PAGE_DIRTY; return pte; }
263static inline pte_t pte_mkyoung(pte_t pte) {
264 pte_val(pte) |= _PAGE_ACCESSED; return pte; }
265static inline pte_t pte_mkhuge(pte_t pte) {
266 return pte; }
267
268/* Atomic PTE updates */
269static inline unsigned long pte_update(struct mm_struct *mm,
270 unsigned long addr,
271 pte_t *ptep, unsigned long clr,
272 int huge)
273{
274 unsigned long old, tmp;
275
276 __asm__ __volatile__(
277 "1: ldarx %0,0,%3 # pte_update\n\
278 andi. %1,%0,%6\n\
279 bne- 1b \n\
280 andc %1,%0,%4 \n\
281 stdcx. %1,0,%3 \n\
282 bne- 1b"
283 : "=&r" (old), "=&r" (tmp), "=m" (*ptep)
284 : "r" (ptep), "r" (clr), "m" (*ptep), "i" (_PAGE_BUSY)
285 : "cc" );
286
287 if (old & _PAGE_HASHPTE)
288 hpte_need_flush(mm, addr, ptep, old, huge);
289 return old;
290}
291
292static inline int __ptep_test_and_clear_young(struct mm_struct *mm,
293 unsigned long addr, pte_t *ptep)
294{
295 unsigned long old;
296
297 if ((pte_val(*ptep) & (_PAGE_ACCESSED | _PAGE_HASHPTE)) == 0)
298 return 0;
299 old = pte_update(mm, addr, ptep, _PAGE_ACCESSED, 0);
300 return (old & _PAGE_ACCESSED) != 0;
301}
302#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
303#define ptep_test_and_clear_young(__vma, __addr, __ptep) \
304({ \
305 int __r; \
306 __r = __ptep_test_and_clear_young((__vma)->vm_mm, __addr, __ptep); \
307 __r; \
308})
309
310/*
311 * On RW/DIRTY bit transitions we can avoid flushing the hpte. For the
312 * moment we always flush but we need to fix hpte_update and test if the
313 * optimisation is worth it.
314 */
315static inline int __ptep_test_and_clear_dirty(struct mm_struct *mm,
316 unsigned long addr, pte_t *ptep)
317{
318 unsigned long old;
319
320 if ((pte_val(*ptep) & _PAGE_DIRTY) == 0)
321 return 0;
322 old = pte_update(mm, addr, ptep, _PAGE_DIRTY, 0);
323 return (old & _PAGE_DIRTY) != 0;
324}
325#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
326#define ptep_test_and_clear_dirty(__vma, __addr, __ptep) \
327({ \
328 int __r; \
329 __r = __ptep_test_and_clear_dirty((__vma)->vm_mm, __addr, __ptep); \
330 __r; \
331})
332
333#define __HAVE_ARCH_PTEP_SET_WRPROTECT
334static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
335 pte_t *ptep)
336{
337 unsigned long old;
338
339 if ((pte_val(*ptep) & _PAGE_RW) == 0)
340 return;
341 old = pte_update(mm, addr, ptep, _PAGE_RW, 0);
342}
343
344/*
345 * We currently remove entries from the hashtable regardless of whether
346 * the entry was young or dirty. The generic routines only flush if the
347 * entry was young or dirty which is not good enough.
348 *
349 * We should be more intelligent about this but for the moment we override
350 * these functions and force a tlb flush unconditionally
351 */
352#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
353#define ptep_clear_flush_young(__vma, __address, __ptep) \
354({ \
355 int __young = __ptep_test_and_clear_young((__vma)->vm_mm, __address, \
356 __ptep); \
357 __young; \
358})
359
360#define __HAVE_ARCH_PTEP_CLEAR_DIRTY_FLUSH
361#define ptep_clear_flush_dirty(__vma, __address, __ptep) \
362({ \
363 int __dirty = __ptep_test_and_clear_dirty((__vma)->vm_mm, __address, \
364 __ptep); \
365 __dirty; \
366})
367
368#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
369static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
370 unsigned long addr, pte_t *ptep)
371{
372 unsigned long old = pte_update(mm, addr, ptep, ~0UL, 0);
373 return __pte(old);
374}
375
376static inline void pte_clear(struct mm_struct *mm, unsigned long addr,
377 pte_t * ptep)
378{
379 pte_update(mm, addr, ptep, ~0UL, 0);
380}
381
382/*
383 * set_pte stores a linux PTE into the linux page table.
384 */
385static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
386 pte_t *ptep, pte_t pte)
387{
388 if (pte_present(*ptep))
389 pte_clear(mm, addr, ptep);
390 pte = __pte(pte_val(pte) & ~_PAGE_HPTEFLAGS);
391 *ptep = pte;
392}
393
394/* Set the dirty and/or accessed bits atomically in a linux PTE, this
395 * function doesn't need to flush the hash entry
396 */
397#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
398static inline void __ptep_set_access_flags(pte_t *ptep, pte_t entry, int dirty)
399{
400 unsigned long bits = pte_val(entry) &
401 (_PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_RW | _PAGE_EXEC);
402 unsigned long old, tmp;
403
404 __asm__ __volatile__(
405 "1: ldarx %0,0,%4\n\
406 andi. %1,%0,%6\n\
407 bne- 1b \n\
408 or %0,%3,%0\n\
409 stdcx. %0,0,%4\n\
410 bne- 1b"
411 :"=&r" (old), "=&r" (tmp), "=m" (*ptep)
412 :"r" (bits), "r" (ptep), "m" (*ptep), "i" (_PAGE_BUSY)
413 :"cc");
414}
415#define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
416 do { \
417 __ptep_set_access_flags(__ptep, __entry, __dirty); \
418 flush_tlb_page_nohash(__vma, __address); \
419 } while(0)
420
421/*
422 * Macro to mark a page protection value as "uncacheable".
423 */
424#define pgprot_noncached(prot) (__pgprot(pgprot_val(prot) | _PAGE_NO_CACHE | _PAGE_GUARDED))
425
426struct file;
427extern pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
428 unsigned long size, pgprot_t vma_prot);
429#define __HAVE_PHYS_MEM_ACCESS_PROT
430
431#define __HAVE_ARCH_PTE_SAME
432#define pte_same(A,B) (((pte_val(A) ^ pte_val(B)) & ~_PAGE_HPTEFLAGS) == 0)
433
434#define pte_ERROR(e) \
435 printk("%s:%d: bad pte %08lx.\n", __FILE__, __LINE__, pte_val(e))
436#define pmd_ERROR(e) \
437 printk("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pmd_val(e))
438#define pgd_ERROR(e) \
439 printk("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e))
440
441extern pgd_t swapper_pg_dir[];
442
443extern void paging_init(void);
444
445/* Encode and de-code a swap entry */
446#define __swp_type(entry) (((entry).val >> 1) & 0x3f)
447#define __swp_offset(entry) ((entry).val >> 8)
448#define __swp_entry(type, offset) ((swp_entry_t){((type)<< 1)|((offset)<<8)})
449#define __pte_to_swp_entry(pte) ((swp_entry_t){pte_val(pte) >> PTE_RPN_SHIFT})
450#define __swp_entry_to_pte(x) ((pte_t) { (x).val << PTE_RPN_SHIFT })
451#define pte_to_pgoff(pte) (pte_val(pte) >> PTE_RPN_SHIFT)
452#define pgoff_to_pte(off) ((pte_t) {((off) << PTE_RPN_SHIFT)|_PAGE_FILE})
453#define PTE_FILE_MAX_BITS (BITS_PER_LONG - PTE_RPN_SHIFT)
454
455/*
456 * kern_addr_valid is intended to indicate whether an address is a valid
457 * kernel address. Most 32-bit archs define it as always true (like this)
458 * but most 64-bit archs actually perform a test. What should we do here?
459 * The only use is in fs/ncpfs/dir.c
460 */
461#define kern_addr_valid(addr) (1)
462
463#define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
464 remap_pfn_range(vma, vaddr, pfn, size, prot)
465
466void pgtable_cache_init(void);
467
468/*
469 * find_linux_pte returns the address of a linux pte for a given
470 * effective address and directory. If not found, it returns zero.
471 */static inline pte_t *find_linux_pte(pgd_t *pgdir, unsigned long ea)
472{
473 pgd_t *pg;
474 pud_t *pu;
475 pmd_t *pm;
476 pte_t *pt = NULL;
477
478 pg = pgdir + pgd_index(ea);
479 if (!pgd_none(*pg)) {
480 pu = pud_offset(pg, ea);
481 if (!pud_none(*pu)) {
482 pm = pmd_offset(pu, ea);
483 if (pmd_present(*pm))
484 pt = pte_offset_kernel(pm, ea);
485 }
486 }
487 return pt;
488}
489
490#endif /* __ASSEMBLY__ */
491
492#endif /* _ASM_POWERPC_PGTABLE_PPC64_H_ */