asm-generic: introduce __ARCH_USE_5LEVEL_HACK
[linux-block.git] / arch / powerpc / include / asm / nohash / 32 / pgtable.h
CommitLineData
17ed9e31
AK
1#ifndef _ASM_POWERPC_NOHASH_32_PGTABLE_H
2#define _ASM_POWERPC_NOHASH_32_PGTABLE_H
f88df14b 3
d1953c88 4#include <asm-generic/pgtable-nopmd.h>
f88df14b
DG
5
6#ifndef __ASSEMBLY__
7#include <linux/sched.h>
8#include <linux/threads.h>
f88df14b 9#include <asm/io.h> /* For sub-arch specific PPC_PIN_SIZE */
f88df14b 10
f637a49e 11extern unsigned long ioremap_bot;
b98ac05d
BH
12
13#ifdef CONFIG_44x
14extern int icache_44x_need_flush;
15#endif
16
f88df14b
DG
17#endif /* __ASSEMBLY__ */
18
9b081e10
CL
19#define PTE_INDEX_SIZE PTE_SHIFT
20#define PMD_INDEX_SIZE 0
21#define PUD_INDEX_SIZE 0
22#define PGD_INDEX_SIZE (32 - PGDIR_SHIFT)
23
24#define PMD_CACHE_INDEX PMD_INDEX_SIZE
25
26#ifndef __ASSEMBLY__
27#define PTE_TABLE_SIZE (sizeof(pte_t) << PTE_INDEX_SIZE)
28#define PMD_TABLE_SIZE 0
29#define PUD_TABLE_SIZE 0
30#define PGD_TABLE_SIZE (sizeof(pgd_t) << PGD_INDEX_SIZE)
31#endif /* __ASSEMBLY__ */
32
33#define PTRS_PER_PTE (1 << PTE_INDEX_SIZE)
34#define PTRS_PER_PGD (1 << PGD_INDEX_SIZE)
35
f88df14b
DG
36/*
37 * The normal case is that PTEs are 32-bits and we have a 1-page
38 * 1024-entry pgdir pointing to 1-page 1024-entry PTE pages. -- paulus
39 *
40 * For any >32-bit physical address platform, we can use the following
41 * two level page table layout where the pgdir is 8KB and the MS 13 bits
42 * are an index to the second level table. The combined pgdir/pmd first
43 * level has 2048 entries and the second level has 512 64-bit PTE entries.
44 * -Matt
45 */
f88df14b 46/* PGDIR_SHIFT determines what a top-level page table entry can map */
9b081e10 47#define PGDIR_SHIFT (PAGE_SHIFT + PTE_INDEX_SIZE)
f88df14b
DG
48#define PGDIR_SIZE (1UL << PGDIR_SHIFT)
49#define PGDIR_MASK (~(PGDIR_SIZE-1))
50
9b081e10
CL
51/* Bits to mask out from a PGD to get to the PUD page */
52#define PGD_MASKED_BITS 0
f88df14b
DG
53
54#define USER_PTRS_PER_PGD (TASK_SIZE / PGDIR_SIZE)
d016bf7e 55#define FIRST_USER_ADDRESS 0UL
f88df14b 56
f88df14b 57#define pte_ERROR(e) \
a7696b36 58 pr_err("%s:%d: bad pte %llx.\n", __FILE__, __LINE__, \
0aeafb0c 59 (unsigned long long)pte_val(e))
f88df14b 60#define pgd_ERROR(e) \
a7696b36 61 pr_err("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e))
f88df14b 62
f637a49e
BH
63/*
64 * This is the bottom of the PKMAP area with HIGHMEM or an arbitrary
65 * value (for now) on others, from where we can start layout kernel
66 * virtual space that goes below PKMAP and FIXMAP
67 */
68#ifdef CONFIG_HIGHMEM
69#define KVIRT_TOP PKMAP_BASE
70#else
71#define KVIRT_TOP (0xfe000000UL) /* for now, could be FIXMAP_BASE ? */
72#endif
73
74/*
75 * ioremap_bot starts at that address. Early ioremaps move down from there,
76 * until mem_init() at which point this becomes the top of the vmalloc
77 * and ioremap space
78 */
8b31e49d
BH
79#ifdef CONFIG_NOT_COHERENT_CACHE
80#define IOREMAP_TOP ((KVIRT_TOP - CONFIG_CONSISTENT_SIZE) & PAGE_MASK)
81#else
f637a49e 82#define IOREMAP_TOP KVIRT_TOP
8b31e49d 83#endif
f637a49e 84
f88df14b
DG
85/*
86 * Just any arbitrary offset to the start of the vmalloc VM area: the
f637a49e 87 * current 16MB value just means that there will be a 64MB "hole" after the
f88df14b
DG
88 * physical memory until the kernel virtual memory starts. That means that
89 * any out-of-bounds memory accesses will hopefully be caught.
90 * The vmalloc() routines leaves a hole of 4kB between each vmalloced
91 * area for the same reason. ;)
92 *
93 * We no longer map larger than phys RAM with the BATs so we don't have
94 * to worry about the VMALLOC_OFFSET causing problems. We do have to worry
95 * about clashes between our early calls to ioremap() that start growing down
e974cd4b 96 * from IOREMAP_TOP being run into the VM area allocations (growing upwards
f88df14b
DG
97 * from VMALLOC_START). For this reason we have ioremap_bot to check when
98 * we actually run into our mappings setup in the early boot with the VM
99 * system. This really does become a problem for machines with good amounts
100 * of RAM. -- Cort
101 */
102#define VMALLOC_OFFSET (0x1000000) /* 16M */
103#ifdef PPC_PIN_SIZE
104#define VMALLOC_START (((_ALIGN((long)high_memory, PPC_PIN_SIZE) + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1)))
105#else
106#define VMALLOC_START ((((long)high_memory + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1)))
107#endif
108#define VMALLOC_END ioremap_bot
109
110/*
111 * Bits in a linux-style PTE. These match the bits in the
112 * (hardware-defined) PowerPC PTE as closely as possible.
113 */
114
115#if defined(CONFIG_40x)
17ed9e31 116#include <asm/nohash/32/pte-40x.h>
f88df14b 117#elif defined(CONFIG_44x)
17ed9e31 118#include <asm/nohash/32/pte-44x.h>
76acc2c1 119#elif defined(CONFIG_FSL_BOOKE) && defined(CONFIG_PTE_64BIT)
17ed9e31 120#include <asm/nohash/pte-book3e.h>
f88df14b 121#elif defined(CONFIG_FSL_BOOKE)
17ed9e31 122#include <asm/nohash/32/pte-fsl-booke.h>
f88df14b 123#elif defined(CONFIG_8xx)
17ed9e31 124#include <asm/nohash/32/pte-8xx.h>
4ee7084e 125#endif
f88df14b 126
71087002
BH
127/* And here we include common definitions */
128#include <asm/pte-common.h>
f88df14b
DG
129
130#ifndef __ASSEMBLY__
f88df14b 131
9bf2b5cd
KG
132#define pte_clear(mm, addr, ptep) \
133 do { pte_update(ptep, ~_PAGE_HASHPTE, 0); } while (0)
f88df14b
DG
134
135#define pmd_none(pmd) (!pmd_val(pmd))
136#define pmd_bad(pmd) (pmd_val(pmd) & _PMD_BAD)
137#define pmd_present(pmd) (pmd_val(pmd) & _PMD_PRESENT_MASK)
f281b5d5
AK
138static inline void pmd_clear(pmd_t *pmdp)
139{
140 *pmdp = __pmd(0);
141}
142
143
f88df14b 144
f88df14b
DG
145/*
146 * When flushing the tlb entry for a page, we also need to flush the hash
147 * table entry. flush_hash_pages is assembler (for speed) in hashtable.S.
148 */
149extern int flush_hash_pages(unsigned context, unsigned long va,
150 unsigned long pmdval, int count);
151
152/* Add an HPTE to the hash table */
153extern void add_hash_page(unsigned context, unsigned long va,
154 unsigned long pmdval);
155
4ee7084e
BB
156/* Flush an entry from the TLB/hash table */
157extern void flush_hash_entry(struct mm_struct *mm, pte_t *ptep,
158 unsigned long address);
159
f88df14b 160/*
c605782b
BH
161 * PTE updates. This function is called whenever an existing
162 * valid PTE is updated. This does -not- include set_pte_at()
163 * which nowadays only sets a new PTE.
164 *
165 * Depending on the type of MMU, we may need to use atomic updates
166 * and the PTE may be either 32 or 64 bit wide. In the later case,
167 * when using atomic updates, only the low part of the PTE is
168 * accessed atomically.
f88df14b 169 *
c605782b
BH
170 * In addition, on 44x, we also maintain a global flag indicating
171 * that an executable user mapping was modified, which is needed
172 * to properly flush the virtually tagged instruction cache of
173 * those implementations.
f88df14b
DG
174 */
175#ifndef CONFIG_PTE_64BIT
1bc54c03
BH
176static inline unsigned long pte_update(pte_t *p,
177 unsigned long clr,
f88df14b
DG
178 unsigned long set)
179{
1bc54c03 180#ifdef PTE_ATOMIC_UPDATES
f88df14b
DG
181 unsigned long old, tmp;
182
183 __asm__ __volatile__("\
1841: lwarx %0,0,%3\n\
185 andc %1,%0,%4\n\
186 or %1,%1,%5\n"
187 PPC405_ERR77(0,%3)
188" stwcx. %1,0,%3\n\
189 bne- 1b"
190 : "=&r" (old), "=&r" (tmp), "=m" (*p)
191 : "r" (p), "r" (clr), "r" (set), "m" (*p)
192 : "cc" );
1bc54c03
BH
193#else /* PTE_ATOMIC_UPDATES */
194 unsigned long old = pte_val(*p);
195 *p = __pte((old & ~clr) | set);
196#endif /* !PTE_ATOMIC_UPDATES */
197
b98ac05d 198#ifdef CONFIG_44x
ea3cc330 199 if ((old & _PAGE_USER) && (old & _PAGE_EXEC))
b98ac05d
BH
200 icache_44x_need_flush = 1;
201#endif
f88df14b
DG
202 return old;
203}
1bc54c03 204#else /* CONFIG_PTE_64BIT */
1bc54c03
BH
205static inline unsigned long long pte_update(pte_t *p,
206 unsigned long clr,
207 unsigned long set)
f88df14b 208{
1bc54c03 209#ifdef PTE_ATOMIC_UPDATES
f88df14b
DG
210 unsigned long long old;
211 unsigned long tmp;
212
213 __asm__ __volatile__("\
2141: lwarx %L0,0,%4\n\
215 lwzx %0,0,%3\n\
216 andc %1,%L0,%5\n\
217 or %1,%1,%6\n"
218 PPC405_ERR77(0,%3)
219" stwcx. %1,0,%4\n\
220 bne- 1b"
221 : "=&r" (old), "=&r" (tmp), "=m" (*p)
222 : "r" (p), "r" ((unsigned long)(p) + 4), "r" (clr), "r" (set), "m" (*p)
223 : "cc" );
1bc54c03
BH
224#else /* PTE_ATOMIC_UPDATES */
225 unsigned long long old = pte_val(*p);
585583d9 226 *p = __pte((old & ~(unsigned long long)clr) | set);
1bc54c03
BH
227#endif /* !PTE_ATOMIC_UPDATES */
228
b98ac05d 229#ifdef CONFIG_44x
ea3cc330 230 if ((old & _PAGE_USER) && (old & _PAGE_EXEC))
b98ac05d
BH
231 icache_44x_need_flush = 1;
232#endif
f88df14b
DG
233 return old;
234}
1bc54c03 235#endif /* CONFIG_PTE_64BIT */
f88df14b 236
f88df14b 237/*
bf2737f7
BB
238 * 2.6 calls this without flushing the TLB entry; this is wrong
239 * for our hash-based implementation, we fix that up here.
f88df14b
DG
240 */
241#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
242static inline int __ptep_test_and_clear_young(unsigned int context, unsigned long addr, pte_t *ptep)
243{
244 unsigned long old;
245 old = pte_update(ptep, _PAGE_ACCESSED, 0);
246#if _PAGE_HASHPTE != 0
247 if (old & _PAGE_HASHPTE) {
248 unsigned long ptephys = __pa(ptep) & PAGE_MASK;
249 flush_hash_pages(context, addr, ptephys, 1);
250 }
251#endif
252 return (old & _PAGE_ACCESSED) != 0;
253}
254#define ptep_test_and_clear_young(__vma, __addr, __ptep) \
255 __ptep_test_and_clear_young((__vma)->vm_mm->context.id, __addr, __ptep)
256
f88df14b
DG
257#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
258static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
259 pte_t *ptep)
260{
261 return __pte(pte_update(ptep, ~_PAGE_HASHPTE, 0));
262}
263
264#define __HAVE_ARCH_PTEP_SET_WRPROTECT
265static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
266 pte_t *ptep)
267{
a7b9f671 268 pte_update(ptep, (_PAGE_RW | _PAGE_HWWRITE), _PAGE_RO);
f88df14b 269}
016b33c4
AW
270static inline void huge_ptep_set_wrprotect(struct mm_struct *mm,
271 unsigned long addr, pte_t *ptep)
272{
273 ptep_set_wrprotect(mm, addr, ptep);
274}
275
f88df14b 276
c6d1a767 277static inline void __ptep_set_access_flags(struct mm_struct *mm,
b3603e17
AK
278 pte_t *ptep, pte_t entry,
279 unsigned long address)
f88df14b 280{
a7b9f671 281 unsigned long set = pte_val(entry) &
ea3cc330 282 (_PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_RW | _PAGE_EXEC);
a7b9f671
LC
283 unsigned long clr = ~pte_val(entry) & _PAGE_RO;
284
285 pte_update(ptep, clr, set);
f88df14b
DG
286}
287
f88df14b
DG
288#define __HAVE_ARCH_PTE_SAME
289#define pte_same(A,B) (((pte_val(A) ^ pte_val(B)) & ~_PAGE_HASHPTE) == 0)
290
291/*
292 * Note that on Book E processors, the pmd contains the kernel virtual
293 * (lowmem) address of the pte page. The physical address is less useful
294 * because everything runs with translation enabled (even the TLB miss
295 * handler). On everything else the pmd contains the physical address
296 * of the pte page. -- paulus
297 */
298#ifndef CONFIG_BOOKE
299#define pmd_page_vaddr(pmd) \
300 ((unsigned long) __va(pmd_val(pmd) & PAGE_MASK))
301#define pmd_page(pmd) \
43b5fefc 302 pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
f88df14b
DG
303#else
304#define pmd_page_vaddr(pmd) \
305 ((unsigned long) (pmd_val(pmd) & PAGE_MASK))
306#define pmd_page(pmd) \
af892e0f 307 pfn_to_page((__pa(pmd_val(pmd)) >> PAGE_SHIFT))
f88df14b
DG
308#endif
309
310/* to find an entry in a kernel page-table-directory */
311#define pgd_offset_k(address) pgd_offset(&init_mm, address)
312
313/* to find an entry in a page-table-directory */
314#define pgd_index(address) ((address) >> PGDIR_SHIFT)
315#define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
316
f88df14b
DG
317/* Find an entry in the third-level page table.. */
318#define pte_index(address) \
319 (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
320#define pte_offset_kernel(dir, addr) \
be00ed72
CL
321 (pmd_bad(*(dir)) ? NULL : (pte_t *)pmd_page_vaddr(*(dir)) + \
322 pte_index(addr))
f88df14b 323#define pte_offset_map(dir, addr) \
ece0e2b6
PZ
324 ((pte_t *) kmap_atomic(pmd_page(*(dir))) + pte_index(addr))
325#define pte_unmap(pte) kunmap_atomic(pte)
f88df14b 326
f88df14b
DG
327/*
328 * Encode and decode a swap entry.
329 * Note that the bits we use in a PTE for representing a swap entry
780fc564
KS
330 * must not include the _PAGE_PRESENT bit or the _PAGE_HASHPTE bit (if used).
331 * -- paulus
f88df14b
DG
332 */
333#define __swp_type(entry) ((entry).val & 0x1f)
334#define __swp_offset(entry) ((entry).val >> 5)
335#define __swp_entry(type, offset) ((swp_entry_t) { (type) | ((offset) << 5) })
336#define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) >> 3 })
337#define __swp_entry_to_pte(x) ((pte_t) { (x).val << 3 })
338
f88df14b
DG
339extern int get_pteptr(struct mm_struct *mm, unsigned long addr, pte_t **ptep,
340 pmd_t **pmdp);
341
342#endif /* !__ASSEMBLY__ */
343
17ed9e31 344#endif /* __ASM_POWERPC_NOHASH_32_PGTABLE_H */