aboutsummaryrefslogtreecommitdiffstats
path: root/include/asm-powerpc
diff options
context:
space:
mode:
authorDavid Gibson <david@gibson.dropbear.id.au>2007-04-30 02:30:56 -0400
committerPaul Mackerras <paulus@samba.org>2007-05-02 06:04:30 -0400
commitf88df14b1f15cdeffa060580a40c1ce3e13bb79e (patch)
tree0619f32c2be79a85792537ad4410cc8d729f4f75 /include/asm-powerpc
parent69d48b409cac747cc0707b05b769e38488a6ad35 (diff)
[POWERPC] Remove arch/powerpc's dependence on asm-ppc/pg{alloc,table}.h
Currently, all 32-bit powerpc platforms use asm-ppc/pgtable.h and asm-ppc/pgalloc.h, even when otherwise compiled with ARCH=powerpc. Those asm-ppc files are a fairly nasty tangle of #ifdefs including a bunch of things which shouldn't be necessary any more in arch/powerpc. Cleaning up that mess is going to take a while, but this patch is a first step. It separates the asm-powerpc/pg{alloc,table}.h into 64 bit and 32 bit versions in asm-powerpc, which the basic .h files in asm-powerpc select based on config. We make a few tiny tweaks to the innards of the files along the way, making the outermost ifdefs (double-inclusion protection and __KERNEL__) a little cleaner, and #including asm-generic/pgtable.h from the top-level asm-powerpc/pgtable.h (since both the old 32-bit and 64-bit versions ended with such an #include). Signed-off-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Paul Mackerras <paulus@samba.org>
Diffstat (limited to 'include/asm-powerpc')
-rw-r--r--include/asm-powerpc/pgalloc-32.h41
-rw-r--r--include/asm-powerpc/pgalloc-64.h152
-rw-r--r--include/asm-powerpc/pgalloc.h154
-rw-r--r--include/asm-powerpc/pgtable-4k.h3
-rw-r--r--include/asm-powerpc/pgtable-64k.h5
-rw-r--r--include/asm-powerpc/pgtable-ppc32.h838
-rw-r--r--include/asm-powerpc/pgtable-ppc64.h492
-rw-r--r--include/asm-powerpc/pgtable.h493
8 files changed, 1532 insertions, 646 deletions
diff --git a/include/asm-powerpc/pgalloc-32.h b/include/asm-powerpc/pgalloc-32.h
new file mode 100644
index 000000000000..235aef283edf
--- /dev/null
+++ b/include/asm-powerpc/pgalloc-32.h
@@ -0,0 +1,41 @@
1#ifndef _ASM_POWERPC_PGALLOC_32_H
2#define _ASM_POWERPC_PGALLOC_32_H
3
4#include <linux/threads.h>
5
6extern void __bad_pte(pmd_t *pmd);
7
8extern pgd_t *pgd_alloc(struct mm_struct *mm);
9extern void pgd_free(pgd_t *pgd);
10
11/*
12 * We don't have any real pmd's, and this code never triggers because
13 * the pgd will always be present..
14 */
15#define pmd_alloc_one(mm,address) ({ BUG(); ((pmd_t *)2); })
16#define pmd_free(x) do { } while (0)
17#define __pmd_free_tlb(tlb,x) do { } while (0)
18#define pgd_populate(mm, pmd, pte) BUG()
19
20#ifndef CONFIG_BOOKE
21#define pmd_populate_kernel(mm, pmd, pte) \
22 (pmd_val(*(pmd)) = __pa(pte) | _PMD_PRESENT)
23#define pmd_populate(mm, pmd, pte) \
24 (pmd_val(*(pmd)) = (page_to_pfn(pte) << PAGE_SHIFT) | _PMD_PRESENT)
25#else
26#define pmd_populate_kernel(mm, pmd, pte) \
27 (pmd_val(*(pmd)) = (unsigned long)pte | _PMD_PRESENT)
28#define pmd_populate(mm, pmd, pte) \
29 (pmd_val(*(pmd)) = (unsigned long)lowmem_page_address(pte) | _PMD_PRESENT)
30#endif
31
32extern pte_t *pte_alloc_one_kernel(struct mm_struct *mm, unsigned long addr);
33extern struct page *pte_alloc_one(struct mm_struct *mm, unsigned long addr);
34extern void pte_free_kernel(pte_t *pte);
35extern void pte_free(struct page *pte);
36
37#define __pte_free_tlb(tlb, pte) pte_free((pte))
38
39#define check_pgt_cache() do { } while (0)
40
41#endif /* _ASM_POWERPC_PGALLOC_32_H */
diff --git a/include/asm-powerpc/pgalloc-64.h b/include/asm-powerpc/pgalloc-64.h
new file mode 100644
index 000000000000..30b50cf56e2c
--- /dev/null
+++ b/include/asm-powerpc/pgalloc-64.h
@@ -0,0 +1,152 @@
1#ifndef _ASM_POWERPC_PGALLOC_64_H
2#define _ASM_POWERPC_PGALLOC_64_H
3/*
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License
6 * as published by the Free Software Foundation; either version
7 * 2 of the License, or (at your option) any later version.
8 */
9
10#include <linux/mm.h>
11#include <linux/slab.h>
12#include <linux/cpumask.h>
13#include <linux/percpu.h>
14
15extern struct kmem_cache *pgtable_cache[];
16
17#ifdef CONFIG_PPC_64K_PAGES
18#define PTE_CACHE_NUM 0
19#define PMD_CACHE_NUM 1
20#define PGD_CACHE_NUM 2
21#define HUGEPTE_CACHE_NUM 3
22#else
23#define PTE_CACHE_NUM 0
24#define PMD_CACHE_NUM 1
25#define PUD_CACHE_NUM 1
26#define PGD_CACHE_NUM 0
27#define HUGEPTE_CACHE_NUM 2
28#endif
29
30static inline pgd_t *pgd_alloc(struct mm_struct *mm)
31{
32 return kmem_cache_alloc(pgtable_cache[PGD_CACHE_NUM], GFP_KERNEL);
33}
34
35static inline void pgd_free(pgd_t *pgd)
36{
37 kmem_cache_free(pgtable_cache[PGD_CACHE_NUM], pgd);
38}
39
40#ifndef CONFIG_PPC_64K_PAGES
41
42#define pgd_populate(MM, PGD, PUD) pgd_set(PGD, PUD)
43
44static inline pud_t *pud_alloc_one(struct mm_struct *mm, unsigned long addr)
45{
46 return kmem_cache_alloc(pgtable_cache[PUD_CACHE_NUM],
47 GFP_KERNEL|__GFP_REPEAT);
48}
49
50static inline void pud_free(pud_t *pud)
51{
52 kmem_cache_free(pgtable_cache[PUD_CACHE_NUM], pud);
53}
54
55static inline void pud_populate(struct mm_struct *mm, pud_t *pud, pmd_t *pmd)
56{
57 pud_set(pud, (unsigned long)pmd);
58}
59
60#define pmd_populate(mm, pmd, pte_page) \
61 pmd_populate_kernel(mm, pmd, page_address(pte_page))
62#define pmd_populate_kernel(mm, pmd, pte) pmd_set(pmd, (unsigned long)(pte))
63
64
65#else /* CONFIG_PPC_64K_PAGES */
66
67#define pud_populate(mm, pud, pmd) pud_set(pud, (unsigned long)pmd)
68
69static inline void pmd_populate_kernel(struct mm_struct *mm, pmd_t *pmd,
70 pte_t *pte)
71{
72 pmd_set(pmd, (unsigned long)pte);
73}
74
75#define pmd_populate(mm, pmd, pte_page) \
76 pmd_populate_kernel(mm, pmd, page_address(pte_page))
77
78#endif /* CONFIG_PPC_64K_PAGES */
79
80static inline pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long addr)
81{
82 return kmem_cache_alloc(pgtable_cache[PMD_CACHE_NUM],
83 GFP_KERNEL|__GFP_REPEAT);
84}
85
86static inline void pmd_free(pmd_t *pmd)
87{
88 kmem_cache_free(pgtable_cache[PMD_CACHE_NUM], pmd);
89}
90
91static inline pte_t *pte_alloc_one_kernel(struct mm_struct *mm,
92 unsigned long address)
93{
94 return kmem_cache_alloc(pgtable_cache[PTE_CACHE_NUM],
95 GFP_KERNEL|__GFP_REPEAT);
96}
97
98static inline struct page *pte_alloc_one(struct mm_struct *mm,
99 unsigned long address)
100{
101 return virt_to_page(pte_alloc_one_kernel(mm, address));
102}
103
104static inline void pte_free_kernel(pte_t *pte)
105{
106 kmem_cache_free(pgtable_cache[PTE_CACHE_NUM], pte);
107}
108
109static inline void pte_free(struct page *ptepage)
110{
111 pte_free_kernel(page_address(ptepage));
112}
113
114#define PGF_CACHENUM_MASK 0x3
115
116typedef struct pgtable_free {
117 unsigned long val;
118} pgtable_free_t;
119
120static inline pgtable_free_t pgtable_free_cache(void *p, int cachenum,
121 unsigned long mask)
122{
123 BUG_ON(cachenum > PGF_CACHENUM_MASK);
124
125 return (pgtable_free_t){.val = ((unsigned long) p & ~mask) | cachenum};
126}
127
128static inline void pgtable_free(pgtable_free_t pgf)
129{
130 void *p = (void *)(pgf.val & ~PGF_CACHENUM_MASK);
131 int cachenum = pgf.val & PGF_CACHENUM_MASK;
132
133 kmem_cache_free(pgtable_cache[cachenum], p);
134}
135
136extern void pgtable_free_tlb(struct mmu_gather *tlb, pgtable_free_t pgf);
137
138#define __pte_free_tlb(tlb, ptepage) \
139 pgtable_free_tlb(tlb, pgtable_free_cache(page_address(ptepage), \
140 PTE_CACHE_NUM, PTE_TABLE_SIZE-1))
141#define __pmd_free_tlb(tlb, pmd) \
142 pgtable_free_tlb(tlb, pgtable_free_cache(pmd, \
143 PMD_CACHE_NUM, PMD_TABLE_SIZE-1))
144#ifndef CONFIG_PPC_64K_PAGES
145#define __pud_free_tlb(tlb, pud) \
146 pgtable_free_tlb(tlb, pgtable_free_cache(pud, \
147 PUD_CACHE_NUM, PUD_TABLE_SIZE-1))
148#endif /* CONFIG_PPC_64K_PAGES */
149
150#define check_pgt_cache() do { } while (0)
151
152#endif /* _ASM_POWERPC_PGALLOC_64_H */
diff --git a/include/asm-powerpc/pgalloc.h b/include/asm-powerpc/pgalloc.h
index b0830db68f8a..b4505ed0f0f2 100644
--- a/include/asm-powerpc/pgalloc.h
+++ b/include/asm-powerpc/pgalloc.h
@@ -2,159 +2,11 @@
2#define _ASM_POWERPC_PGALLOC_H 2#define _ASM_POWERPC_PGALLOC_H
3#ifdef __KERNEL__ 3#ifdef __KERNEL__
4 4
5#ifndef CONFIG_PPC64 5#ifdef CONFIG_PPC64
6#include <asm-ppc/pgalloc.h> 6#include <asm/pgalloc-64.h>
7#else 7#else
8 8#include <asm/pgalloc-32.h>
9#include <linux/mm.h>
10#include <linux/slab.h>
11#include <linux/cpumask.h>
12#include <linux/percpu.h>
13
14extern struct kmem_cache *pgtable_cache[];
15
16#ifdef CONFIG_PPC_64K_PAGES
17#define PTE_CACHE_NUM 0
18#define PMD_CACHE_NUM 1
19#define PGD_CACHE_NUM 2
20#define HUGEPTE_CACHE_NUM 3
21#else
22#define PTE_CACHE_NUM 0
23#define PMD_CACHE_NUM 1
24#define PUD_CACHE_NUM 1
25#define PGD_CACHE_NUM 0
26#define HUGEPTE_CACHE_NUM 2
27#endif 9#endif
28 10
29/*
30 * This program is free software; you can redistribute it and/or
31 * modify it under the terms of the GNU General Public License
32 * as published by the Free Software Foundation; either version
33 * 2 of the License, or (at your option) any later version.
34 */
35
36static inline pgd_t *pgd_alloc(struct mm_struct *mm)
37{
38 return kmem_cache_alloc(pgtable_cache[PGD_CACHE_NUM], GFP_KERNEL);
39}
40
41static inline void pgd_free(pgd_t *pgd)
42{
43 kmem_cache_free(pgtable_cache[PGD_CACHE_NUM], pgd);
44}
45
46#ifndef CONFIG_PPC_64K_PAGES
47
48#define pgd_populate(MM, PGD, PUD) pgd_set(PGD, PUD)
49
50static inline pud_t *pud_alloc_one(struct mm_struct *mm, unsigned long addr)
51{
52 return kmem_cache_alloc(pgtable_cache[PUD_CACHE_NUM],
53 GFP_KERNEL|__GFP_REPEAT);
54}
55
56static inline void pud_free(pud_t *pud)
57{
58 kmem_cache_free(pgtable_cache[PUD_CACHE_NUM], pud);
59}
60
61static inline void pud_populate(struct mm_struct *mm, pud_t *pud, pmd_t *pmd)
62{
63 pud_set(pud, (unsigned long)pmd);
64}
65
66#define pmd_populate(mm, pmd, pte_page) \
67 pmd_populate_kernel(mm, pmd, page_address(pte_page))
68#define pmd_populate_kernel(mm, pmd, pte) pmd_set(pmd, (unsigned long)(pte))
69
70
71#else /* CONFIG_PPC_64K_PAGES */
72
73#define pud_populate(mm, pud, pmd) pud_set(pud, (unsigned long)pmd)
74
75static inline void pmd_populate_kernel(struct mm_struct *mm, pmd_t *pmd,
76 pte_t *pte)
77{
78 pmd_set(pmd, (unsigned long)pte);
79}
80
81#define pmd_populate(mm, pmd, pte_page) \
82 pmd_populate_kernel(mm, pmd, page_address(pte_page))
83
84#endif /* CONFIG_PPC_64K_PAGES */
85
86static inline pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long addr)
87{
88 return kmem_cache_alloc(pgtable_cache[PMD_CACHE_NUM],
89 GFP_KERNEL|__GFP_REPEAT);
90}
91
92static inline void pmd_free(pmd_t *pmd)
93{
94 kmem_cache_free(pgtable_cache[PMD_CACHE_NUM], pmd);
95}
96
97static inline pte_t *pte_alloc_one_kernel(struct mm_struct *mm,
98 unsigned long address)
99{
100 return kmem_cache_alloc(pgtable_cache[PTE_CACHE_NUM],
101 GFP_KERNEL|__GFP_REPEAT);
102}
103
104static inline struct page *pte_alloc_one(struct mm_struct *mm,
105 unsigned long address)
106{
107 return virt_to_page(pte_alloc_one_kernel(mm, address));
108}
109
110static inline void pte_free_kernel(pte_t *pte)
111{
112 kmem_cache_free(pgtable_cache[PTE_CACHE_NUM], pte);
113}
114
115static inline void pte_free(struct page *ptepage)
116{
117 pte_free_kernel(page_address(ptepage));
118}
119
120#define PGF_CACHENUM_MASK 0x3
121
122typedef struct pgtable_free {
123 unsigned long val;
124} pgtable_free_t;
125
126static inline pgtable_free_t pgtable_free_cache(void *p, int cachenum,
127 unsigned long mask)
128{
129 BUG_ON(cachenum > PGF_CACHENUM_MASK);
130
131 return (pgtable_free_t){.val = ((unsigned long) p & ~mask) | cachenum};
132}
133
134static inline void pgtable_free(pgtable_free_t pgf)
135{
136 void *p = (void *)(pgf.val & ~PGF_CACHENUM_MASK);
137 int cachenum = pgf.val & PGF_CACHENUM_MASK;
138
139 kmem_cache_free(pgtable_cache[cachenum], p);
140}
141
142extern void pgtable_free_tlb(struct mmu_gather *tlb, pgtable_free_t pgf);
143
144#define __pte_free_tlb(tlb, ptepage) \
145 pgtable_free_tlb(tlb, pgtable_free_cache(page_address(ptepage), \
146 PTE_CACHE_NUM, PTE_TABLE_SIZE-1))
147#define __pmd_free_tlb(tlb, pmd) \
148 pgtable_free_tlb(tlb, pgtable_free_cache(pmd, \
149 PMD_CACHE_NUM, PMD_TABLE_SIZE-1))
150#ifndef CONFIG_PPC_64K_PAGES
151#define __pud_free_tlb(tlb, pud) \
152 pgtable_free_tlb(tlb, pgtable_free_cache(pud, \
153 PUD_CACHE_NUM, PUD_TABLE_SIZE-1))
154#endif /* CONFIG_PPC_64K_PAGES */
155
156#define check_pgt_cache() do { } while (0)
157
158#endif /* CONFIG_PPC64 */
159#endif /* __KERNEL__ */ 11#endif /* __KERNEL__ */
160#endif /* _ASM_POWERPC_PGALLOC_H */ 12#endif /* _ASM_POWERPC_PGALLOC_H */
diff --git a/include/asm-powerpc/pgtable-4k.h b/include/asm-powerpc/pgtable-4k.h
index a28fa8bc01da..1744d6ac12a2 100644
--- a/include/asm-powerpc/pgtable-4k.h
+++ b/include/asm-powerpc/pgtable-4k.h
@@ -1,3 +1,5 @@
1#ifndef _ASM_POWERPC_PGTABLE_4K_H
2#define _ASM_POWERPC_PGTABLE_4K_H
1/* 3/*
2 * Entries per page directory level. The PTE level must use a 64b record 4 * Entries per page directory level. The PTE level must use a 64b record
3 * for each page table entry. The PMD and PGD level use a 32b record for 5 * for each page table entry. The PMD and PGD level use a 32b record for
@@ -100,3 +102,4 @@
100 102
101#define remap_4k_pfn(vma, addr, pfn, prot) \ 103#define remap_4k_pfn(vma, addr, pfn, prot) \
102 remap_pfn_range((vma), (addr), (pfn), PAGE_SIZE, (prot)) 104 remap_pfn_range((vma), (addr), (pfn), PAGE_SIZE, (prot))
105#endif /* _ASM_POWERPC_PGTABLE_4K_H */
diff --git a/include/asm-powerpc/pgtable-64k.h b/include/asm-powerpc/pgtable-64k.h
index 5e84f070eaf7..16ef4978520d 100644
--- a/include/asm-powerpc/pgtable-64k.h
+++ b/include/asm-powerpc/pgtable-64k.h
@@ -1,6 +1,5 @@
1#ifndef _ASM_POWERPC_PGTABLE_64K_H 1#ifndef _ASM_POWERPC_PGTABLE_64K_H
2#define _ASM_POWERPC_PGTABLE_64K_H 2#define _ASM_POWERPC_PGTABLE_64K_H
3#ifdef __KERNEL__
4 3
5#include <asm-generic/pgtable-nopud.h> 4#include <asm-generic/pgtable-nopud.h>
6 5
@@ -65,8 +64,6 @@
65/* Bits to mask out from a PGD/PUD to get to the PMD page */ 64/* Bits to mask out from a PGD/PUD to get to the PMD page */
66#define PUD_MASKED_BITS 0x1ff 65#define PUD_MASKED_BITS 0x1ff
67 66
68#ifndef __ASSEMBLY__
69
70/* Manipulate "rpte" values */ 67/* Manipulate "rpte" values */
71#define __real_pte(e,p) ((real_pte_t) { \ 68#define __real_pte(e,p) ((real_pte_t) { \
72 (e), pte_val(*((p) + PTRS_PER_PTE)) }) 69 (e), pte_val(*((p) + PTRS_PER_PTE)) })
@@ -98,6 +95,4 @@
98 remap_pfn_range((vma), (addr), (pfn), PAGE_SIZE, \ 95 remap_pfn_range((vma), (addr), (pfn), PAGE_SIZE, \
99 __pgprot(pgprot_val((prot)) | _PAGE_4K_PFN)) 96 __pgprot(pgprot_val((prot)) | _PAGE_4K_PFN))
100 97
101#endif /* __ASSEMBLY__ */
102#endif /* __KERNEL__ */
103#endif /* _ASM_POWERPC_PGTABLE_64K_H */ 98#endif /* _ASM_POWERPC_PGTABLE_64K_H */
diff --git a/include/asm-powerpc/pgtable-ppc32.h b/include/asm-powerpc/pgtable-ppc32.h
new file mode 100644
index 000000000000..80c75474c65b
--- /dev/null
+++ b/include/asm-powerpc/pgtable-ppc32.h
@@ -0,0 +1,838 @@
1#ifndef _ASM_POWERPC_PGTABLE_PPC32_H
2#define _ASM_POWERPC_PGTABLE_PPC32_H
3
4#include <asm-generic/4level-fixup.h>
5
6#ifndef __ASSEMBLY__
7#include <linux/sched.h>
8#include <linux/threads.h>
9#include <asm/processor.h> /* For TASK_SIZE */
10#include <asm/mmu.h>
11#include <asm/page.h>
12#include <asm/io.h> /* For sub-arch specific PPC_PIN_SIZE */
13struct mm_struct;
14
15extern unsigned long va_to_phys(unsigned long address);
16extern pte_t *va_to_pte(unsigned long address);
17extern unsigned long ioremap_bot, ioremap_base;
18#endif /* __ASSEMBLY__ */
19
20/*
21 * The PowerPC MMU uses a hash table containing PTEs, together with
22 * a set of 16 segment registers (on 32-bit implementations), to define
23 * the virtual to physical address mapping.
24 *
25 * We use the hash table as an extended TLB, i.e. a cache of currently
26 * active mappings. We maintain a two-level page table tree, much
27 * like that used by the i386, for the sake of the Linux memory
28 * management code. Low-level assembler code in hashtable.S
29 * (procedure hash_page) is responsible for extracting ptes from the
30 * tree and putting them into the hash table when necessary, and
31 * updating the accessed and modified bits in the page table tree.
32 */
33
34/*
35 * The PowerPC MPC8xx uses a TLB with hardware assisted, software tablewalk.
36 * We also use the two level tables, but we can put the real bits in them
37 * needed for the TLB and tablewalk. These definitions require Mx_CTR.PPM = 0,
38 * Mx_CTR.PPCS = 0, and MD_CTR.TWAM = 1. The level 2 descriptor has
39 * additional page protection (when Mx_CTR.PPCS = 1) that allows TLB hit
40 * based upon user/super access. The TLB does not have accessed nor write
41 * protect. We assume that if the TLB get loaded with an entry it is
42 * accessed, and overload the changed bit for write protect. We use
43 * two bits in the software pte that are supposed to be set to zero in
44 * the TLB entry (24 and 25) for these indicators. Although the level 1
45 * descriptor contains the guarded and writethrough/copyback bits, we can
46 * set these at the page level since they get copied from the Mx_TWC
47 * register when the TLB entry is loaded. We will use bit 27 for guard, since
48 * that is where it exists in the MD_TWC, and bit 26 for writethrough.
49 * These will get masked from the level 2 descriptor at TLB load time, and
50 * copied to the MD_TWC before it gets loaded.
51 * Large page sizes added. We currently support two sizes, 4K and 8M.
52 * This also allows a TLB hander optimization because we can directly
53 * load the PMD into MD_TWC. The 8M pages are only used for kernel
54 * mapping of well known areas. The PMD (PGD) entries contain control
55 * flags in addition to the address, so care must be taken that the
56 * software no longer assumes these are only pointers.
57 */
58
59/*
60 * At present, all PowerPC 400-class processors share a similar TLB
61 * architecture. The instruction and data sides share a unified,
62 * 64-entry, fully-associative TLB which is maintained totally under
63 * software control. In addition, the instruction side has a
64 * hardware-managed, 4-entry, fully-associative TLB which serves as a
65 * first level to the shared TLB. These two TLBs are known as the UTLB
66 * and ITLB, respectively (see "mmu.h" for definitions).
67 */
68
69/*
70 * The normal case is that PTEs are 32-bits and we have a 1-page
71 * 1024-entry pgdir pointing to 1-page 1024-entry PTE pages. -- paulus
72 *
73 * For any >32-bit physical address platform, we can use the following
74 * two level page table layout where the pgdir is 8KB and the MS 13 bits
75 * are an index to the second level table. The combined pgdir/pmd first
76 * level has 2048 entries and the second level has 512 64-bit PTE entries.
77 * -Matt
78 */
79/* PMD_SHIFT determines the size of the area mapped by the PTE pages */
80#define PMD_SHIFT (PAGE_SHIFT + PTE_SHIFT)
81#define PMD_SIZE (1UL << PMD_SHIFT)
82#define PMD_MASK (~(PMD_SIZE-1))
83
84/* PGDIR_SHIFT determines what a top-level page table entry can map */
85#define PGDIR_SHIFT PMD_SHIFT
86#define PGDIR_SIZE (1UL << PGDIR_SHIFT)
87#define PGDIR_MASK (~(PGDIR_SIZE-1))
88
89/*
90 * entries per page directory level: our page-table tree is two-level, so
91 * we don't really have any PMD directory.
92 */
93#define PTRS_PER_PTE (1 << PTE_SHIFT)
94#define PTRS_PER_PMD 1
95#define PTRS_PER_PGD (1 << (32 - PGDIR_SHIFT))
96
97#define USER_PTRS_PER_PGD (TASK_SIZE / PGDIR_SIZE)
98#define FIRST_USER_ADDRESS 0
99
100#define USER_PGD_PTRS (PAGE_OFFSET >> PGDIR_SHIFT)
101#define KERNEL_PGD_PTRS (PTRS_PER_PGD-USER_PGD_PTRS)
102
103#define pte_ERROR(e) \
104 printk("%s:%d: bad pte "PTE_FMT".\n", __FILE__, __LINE__, pte_val(e))
105#define pmd_ERROR(e) \
106 printk("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pmd_val(e))
107#define pgd_ERROR(e) \
108 printk("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e))
109
110/*
111 * Just any arbitrary offset to the start of the vmalloc VM area: the
112 * current 64MB value just means that there will be a 64MB "hole" after the
113 * physical memory until the kernel virtual memory starts. That means that
114 * any out-of-bounds memory accesses will hopefully be caught.
115 * The vmalloc() routines leaves a hole of 4kB between each vmalloced
116 * area for the same reason. ;)
117 *
118 * We no longer map larger than phys RAM with the BATs so we don't have
119 * to worry about the VMALLOC_OFFSET causing problems. We do have to worry
120 * about clashes between our early calls to ioremap() that start growing down
121 * from ioremap_base being run into the VM area allocations (growing upwards
122 * from VMALLOC_START). For this reason we have ioremap_bot to check when
123 * we actually run into our mappings setup in the early boot with the VM
124 * system. This really does become a problem for machines with good amounts
125 * of RAM. -- Cort
126 */
127#define VMALLOC_OFFSET (0x1000000) /* 16M */
128#ifdef PPC_PIN_SIZE
129#define VMALLOC_START (((_ALIGN((long)high_memory, PPC_PIN_SIZE) + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1)))
130#else
131#define VMALLOC_START ((((long)high_memory + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1)))
132#endif
133#define VMALLOC_END ioremap_bot
134
135/*
136 * Bits in a linux-style PTE. These match the bits in the
137 * (hardware-defined) PowerPC PTE as closely as possible.
138 */
139
140#if defined(CONFIG_40x)
141
142/* There are several potential gotchas here. The 40x hardware TLBLO
143 field looks like this:
144
145 0 1 2 3 4 ... 18 19 20 21 22 23 24 25 26 27 28 29 30 31
146 RPN..................... 0 0 EX WR ZSEL....... W I M G
147
148 Where possible we make the Linux PTE bits match up with this
149
150 - bits 20 and 21 must be cleared, because we use 4k pages (40x can
151 support down to 1k pages), this is done in the TLBMiss exception
152 handler.
153 - We use only zones 0 (for kernel pages) and 1 (for user pages)
154 of the 16 available. Bit 24-26 of the TLB are cleared in the TLB
155 miss handler. Bit 27 is PAGE_USER, thus selecting the correct
156 zone.
157 - PRESENT *must* be in the bottom two bits because swap cache
158 entries use the top 30 bits. Because 40x doesn't support SMP
159 anyway, M is irrelevant so we borrow it for PAGE_PRESENT. Bit 30
160 is cleared in the TLB miss handler before the TLB entry is loaded.
161 - All other bits of the PTE are loaded into TLBLO without
162 modification, leaving us only the bits 20, 21, 24, 25, 26, 30 for
163 software PTE bits. We actually use use bits 21, 24, 25, and
164 30 respectively for the software bits: ACCESSED, DIRTY, RW, and
165 PRESENT.
166*/
167
168/* Definitions for 40x embedded chips. */
169#define _PAGE_GUARDED 0x001 /* G: page is guarded from prefetch */
170#define _PAGE_FILE 0x001 /* when !present: nonlinear file mapping */
171#define _PAGE_PRESENT 0x002 /* software: PTE contains a translation */
172#define _PAGE_NO_CACHE 0x004 /* I: caching is inhibited */
173#define _PAGE_WRITETHRU 0x008 /* W: caching is write-through */
174#define _PAGE_USER 0x010 /* matches one of the zone permission bits */
175#define _PAGE_RW 0x040 /* software: Writes permitted */
176#define _PAGE_DIRTY 0x080 /* software: dirty page */
177#define _PAGE_HWWRITE 0x100 /* hardware: Dirty & RW, set in exception */
178#define _PAGE_HWEXEC 0x200 /* hardware: EX permission */
179#define _PAGE_ACCESSED 0x400 /* software: R: page referenced */
180
181#define _PMD_PRESENT 0x400 /* PMD points to page of PTEs */
182#define _PMD_BAD 0x802
183#define _PMD_SIZE 0x0e0 /* size field, != 0 for large-page PMD entry */
184#define _PMD_SIZE_4M 0x0c0
185#define _PMD_SIZE_16M 0x0e0
186#define PMD_PAGE_SIZE(pmdval) (1024 << (((pmdval) & _PMD_SIZE) >> 4))
187
188#elif defined(CONFIG_44x)
189/*
190 * Definitions for PPC440
191 *
192 * Because of the 3 word TLB entries to support 36-bit addressing,
193 * the attribute are difficult to map in such a fashion that they
194 * are easily loaded during exception processing. I decided to
195 * organize the entry so the ERPN is the only portion in the
196 * upper word of the PTE and the attribute bits below are packed
197 * in as sensibly as they can be in the area below a 4KB page size
198 * oriented RPN. This at least makes it easy to load the RPN and
199 * ERPN fields in the TLB. -Matt
200 *
201 * Note that these bits preclude future use of a page size
202 * less than 4KB.
203 *
204 *
205 * PPC 440 core has following TLB attribute fields;
206 *
207 * TLB1:
208 * 0 1 2 3 4 ... 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
209 * RPN................................. - - - - - - ERPN.......
210 *
211 * TLB2:
212 * 0 1 2 3 4 ... 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
213 * - - - - - - U0 U1 U2 U3 W I M G E - UX UW UR SX SW SR
214 *
215 * There are some constrains and options, to decide mapping software bits
216 * into TLB entry.
217 *
218 * - PRESENT *must* be in the bottom three bits because swap cache
219 * entries use the top 29 bits for TLB2.
220 *
221 * - FILE *must* be in the bottom three bits because swap cache
222 * entries use the top 29 bits for TLB2.
223 *
224 * - CACHE COHERENT bit (M) has no effect on PPC440 core, because it
225 * doesn't support SMP. So we can use this as software bit, like
226 * DIRTY.
227 *
228 * With the PPC 44x Linux implementation, the 0-11th LSBs of the PTE are used
229 * for memory protection related functions (see PTE structure in
230 * include/asm-ppc/mmu.h). The _PAGE_XXX definitions in this file map to the
231 * above bits. Note that the bit values are CPU specific, not architecture
232 * specific.
233 *
234 * The kernel PTE entry holds an arch-dependent swp_entry structure under
235 * certain situations. In other words, in such situations some portion of
236 * the PTE bits are used as a swp_entry. In the PPC implementation, the
237 * 3-24th LSB are shared with swp_entry, however the 0-2nd three LSB still
238 * hold protection values. That means the three protection bits are
239 * reserved for both PTE and SWAP entry at the most significant three
240 * LSBs.
241 *
242 * There are three protection bits available for SWAP entry:
243 * _PAGE_PRESENT
244 * _PAGE_FILE
245 * _PAGE_HASHPTE (if HW has)
246 *
247 * So those three bits have to be inside of 0-2nd LSB of PTE.
248 *
249 */
250
251#define _PAGE_PRESENT 0x00000001 /* S: PTE valid */
252#define _PAGE_RW 0x00000002 /* S: Write permission */
253#define _PAGE_FILE 0x00000004 /* S: nonlinear file mapping */
254#define _PAGE_ACCESSED 0x00000008 /* S: Page referenced */
255#define _PAGE_HWWRITE 0x00000010 /* H: Dirty & RW */
256#define _PAGE_HWEXEC 0x00000020 /* H: Execute permission */
257#define _PAGE_USER 0x00000040 /* S: User page */
258#define _PAGE_ENDIAN 0x00000080 /* H: E bit */
259#define _PAGE_GUARDED 0x00000100 /* H: G bit */
260#define _PAGE_DIRTY 0x00000200 /* S: Page dirty */
261#define _PAGE_NO_CACHE 0x00000400 /* H: I bit */
262#define _PAGE_WRITETHRU 0x00000800 /* H: W bit */
263
264/* TODO: Add large page lowmem mapping support */
265#define _PMD_PRESENT 0
266#define _PMD_PRESENT_MASK (PAGE_MASK)
267#define _PMD_BAD (~PAGE_MASK)
268
269/* ERPN in a PTE never gets cleared, ignore it */
270#define _PTE_NONE_MASK 0xffffffff00000000ULL
271
272#elif defined(CONFIG_FSL_BOOKE)
273/*
274 MMU Assist Register 3:
275
276 32 33 34 35 36 ... 50 51 52 53 54 55 56 57 58 59 60 61 62 63
277 RPN...................... 0 0 U0 U1 U2 U3 UX SX UW SW UR SR
278
279 - PRESENT *must* be in the bottom three bits because swap cache
280 entries use the top 29 bits.
281
282 - FILE *must* be in the bottom three bits because swap cache
283 entries use the top 29 bits.
284*/
285
286/* Definitions for FSL Book-E Cores */
287#define _PAGE_PRESENT 0x00001 /* S: PTE contains a translation */
288#define _PAGE_USER 0x00002 /* S: User page (maps to UR) */
289#define _PAGE_FILE 0x00002 /* S: when !present: nonlinear file mapping */
290#define _PAGE_ACCESSED 0x00004 /* S: Page referenced */
291#define _PAGE_HWWRITE 0x00008 /* H: Dirty & RW, set in exception */
292#define _PAGE_RW 0x00010 /* S: Write permission */
293#define _PAGE_HWEXEC 0x00020 /* H: UX permission */
294
295#define _PAGE_ENDIAN 0x00040 /* H: E bit */
296#define _PAGE_GUARDED 0x00080 /* H: G bit */
297#define _PAGE_COHERENT 0x00100 /* H: M bit */
298#define _PAGE_NO_CACHE 0x00200 /* H: I bit */
299#define _PAGE_WRITETHRU 0x00400 /* H: W bit */
300
301#ifdef CONFIG_PTE_64BIT
302#define _PAGE_DIRTY 0x08000 /* S: Page dirty */
303
304/* ERPN in a PTE never gets cleared, ignore it */
305#define _PTE_NONE_MASK 0xffffffffffff0000ULL
306#else
307#define _PAGE_DIRTY 0x00800 /* S: Page dirty */
308#endif
309
310#define _PMD_PRESENT 0
311#define _PMD_PRESENT_MASK (PAGE_MASK)
312#define _PMD_BAD (~PAGE_MASK)
313
314#elif defined(CONFIG_8xx)
315/* Definitions for 8xx embedded chips. */
316#define _PAGE_PRESENT 0x0001 /* Page is valid */
317#define _PAGE_FILE 0x0002 /* when !present: nonlinear file mapping */
318#define _PAGE_NO_CACHE 0x0002 /* I: cache inhibit */
319#define _PAGE_SHARED 0x0004 /* No ASID (context) compare */
320
321/* These five software bits must be masked out when the entry is loaded
322 * into the TLB.
323 */
324#define _PAGE_EXEC 0x0008 /* software: i-cache coherency required */
325#define _PAGE_GUARDED 0x0010 /* software: guarded access */
326#define _PAGE_DIRTY 0x0020 /* software: page changed */
327#define _PAGE_RW 0x0040 /* software: user write access allowed */
328#define _PAGE_ACCESSED 0x0080 /* software: page referenced */
329
330/* Setting any bits in the nibble with the follow two controls will
331 * require a TLB exception handler change. It is assumed unused bits
332 * are always zero.
333 */
334#define _PAGE_HWWRITE 0x0100 /* h/w write enable: never set in Linux PTE */
335#define _PAGE_USER 0x0800 /* One of the PP bits, the other is USER&~RW */
336
337#define _PMD_PRESENT 0x0001
338#define _PMD_BAD 0x0ff0
339#define _PMD_PAGE_MASK 0x000c
340#define _PMD_PAGE_8M 0x000c
341
342/*
343 * The 8xx TLB miss handler allegedly sets _PAGE_ACCESSED in the PTE
344 * for an address even if _PAGE_PRESENT is not set, as a performance
345 * optimization. This is a bug if you ever want to use swap unless
346 * _PAGE_ACCESSED is 2, which it isn't, or unless you have 8xx-specific
347 * definitions for __swp_entry etc. below, which would be gross.
348 * -- paulus
349 */
350#define _PTE_NONE_MASK _PAGE_ACCESSED
351
352#else /* CONFIG_6xx */
353/* Definitions for 60x, 740/750, etc. */
354#define _PAGE_PRESENT 0x001 /* software: pte contains a translation */
355#define _PAGE_HASHPTE 0x002 /* hash_page has made an HPTE for this pte */
356#define _PAGE_FILE 0x004 /* when !present: nonlinear file mapping */
357#define _PAGE_USER 0x004 /* usermode access allowed */
358#define _PAGE_GUARDED 0x008 /* G: prohibit speculative access */
359#define _PAGE_COHERENT 0x010 /* M: enforce memory coherence (SMP systems) */
360#define _PAGE_NO_CACHE 0x020 /* I: cache inhibit */
361#define _PAGE_WRITETHRU 0x040 /* W: cache write-through */
362#define _PAGE_DIRTY 0x080 /* C: page changed */
363#define _PAGE_ACCESSED 0x100 /* R: page referenced */
364#define _PAGE_EXEC 0x200 /* software: i-cache coherency required */
365#define _PAGE_RW 0x400 /* software: user write access allowed */
366
367#define _PTE_NONE_MASK _PAGE_HASHPTE
368
369#define _PMD_PRESENT 0
370#define _PMD_PRESENT_MASK (PAGE_MASK)
371#define _PMD_BAD (~PAGE_MASK)
372#endif
373
374/*
375 * Some bits are only used on some cpu families...
376 */
377#ifndef _PAGE_HASHPTE
378#define _PAGE_HASHPTE 0
379#endif
380#ifndef _PTE_NONE_MASK
381#define _PTE_NONE_MASK 0
382#endif
383#ifndef _PAGE_SHARED
384#define _PAGE_SHARED 0
385#endif
386#ifndef _PAGE_HWWRITE
387#define _PAGE_HWWRITE 0
388#endif
389#ifndef _PAGE_HWEXEC
390#define _PAGE_HWEXEC 0
391#endif
392#ifndef _PAGE_EXEC
393#define _PAGE_EXEC 0
394#endif
395#ifndef _PMD_PRESENT_MASK
396#define _PMD_PRESENT_MASK _PMD_PRESENT
397#endif
398#ifndef _PMD_SIZE
399#define _PMD_SIZE 0
400#define PMD_PAGE_SIZE(pmd) bad_call_to_PMD_PAGE_SIZE()
401#endif
402
403#define _PAGE_CHG_MASK (PAGE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY)
404
405/*
406 * Note: the _PAGE_COHERENT bit automatically gets set in the hardware
407 * PTE if CONFIG_SMP is defined (hash_page does this); there is no need
408 * to have it in the Linux PTE, and in fact the bit could be reused for
409 * another purpose. -- paulus.
410 */
411
412#ifdef CONFIG_44x
413#define _PAGE_BASE (_PAGE_PRESENT | _PAGE_ACCESSED | _PAGE_GUARDED)
414#else
415#define _PAGE_BASE (_PAGE_PRESENT | _PAGE_ACCESSED)
416#endif
417#define _PAGE_WRENABLE (_PAGE_RW | _PAGE_DIRTY | _PAGE_HWWRITE)
418#define _PAGE_KERNEL (_PAGE_BASE | _PAGE_SHARED | _PAGE_WRENABLE)
419
420#ifdef CONFIG_PPC_STD_MMU
421/* On standard PPC MMU, no user access implies kernel read/write access,
422 * so to write-protect kernel memory we must turn on user access */
423#define _PAGE_KERNEL_RO (_PAGE_BASE | _PAGE_SHARED | _PAGE_USER)
424#else
425#define _PAGE_KERNEL_RO (_PAGE_BASE | _PAGE_SHARED)
426#endif
427
428#define _PAGE_IO (_PAGE_KERNEL | _PAGE_NO_CACHE | _PAGE_GUARDED)
429#define _PAGE_RAM (_PAGE_KERNEL | _PAGE_HWEXEC)
430
431#if defined(CONFIG_KGDB) || defined(CONFIG_XMON) || defined(CONFIG_BDI_SWITCH)
432/* We want the debuggers to be able to set breakpoints anywhere, so
433 * don't write protect the kernel text */
434#define _PAGE_RAM_TEXT _PAGE_RAM
435#else
436#define _PAGE_RAM_TEXT (_PAGE_KERNEL_RO | _PAGE_HWEXEC)
437#endif
438
439#define PAGE_NONE __pgprot(_PAGE_BASE)
440#define PAGE_READONLY __pgprot(_PAGE_BASE | _PAGE_USER)
441#define PAGE_READONLY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC)
442#define PAGE_SHARED __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_RW)
443#define PAGE_SHARED_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_RW | _PAGE_EXEC)
444#define PAGE_COPY __pgprot(_PAGE_BASE | _PAGE_USER)
445#define PAGE_COPY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC)
446
447#define PAGE_KERNEL __pgprot(_PAGE_RAM)
448#define PAGE_KERNEL_NOCACHE __pgprot(_PAGE_IO)
449
450/*
451 * The PowerPC can only do execute protection on a segment (256MB) basis,
452 * not on a page basis. So we consider execute permission the same as read.
453 * Also, write permissions imply read permissions.
454 * This is the closest we can get..
455 */
456#define __P000 PAGE_NONE
457#define __P001 PAGE_READONLY_X
458#define __P010 PAGE_COPY
459#define __P011 PAGE_COPY_X
460#define __P100 PAGE_READONLY
461#define __P101 PAGE_READONLY_X
462#define __P110 PAGE_COPY
463#define __P111 PAGE_COPY_X
464
465#define __S000 PAGE_NONE
466#define __S001 PAGE_READONLY_X
467#define __S010 PAGE_SHARED
468#define __S011 PAGE_SHARED_X
469#define __S100 PAGE_READONLY
470#define __S101 PAGE_READONLY_X
471#define __S110 PAGE_SHARED
472#define __S111 PAGE_SHARED_X
473
474#ifndef __ASSEMBLY__
475/* Make sure we get a link error if PMD_PAGE_SIZE is ever called on a
476 * kernel without large page PMD support */
477extern unsigned long bad_call_to_PMD_PAGE_SIZE(void);
478
479/*
480 * Conversions between PTE values and page frame numbers.
481 */
482
483/* in some case we want to additionaly adjust where the pfn is in the pte to
484 * allow room for more flags */
485#if defined(CONFIG_FSL_BOOKE) && defined(CONFIG_PTE_64BIT)
486#define PFN_SHIFT_OFFSET (PAGE_SHIFT + 8)
487#else
488#define PFN_SHIFT_OFFSET (PAGE_SHIFT)
489#endif
490
491#define pte_pfn(x) (pte_val(x) >> PFN_SHIFT_OFFSET)
492#define pte_page(x) pfn_to_page(pte_pfn(x))
493
494#define pfn_pte(pfn, prot) __pte(((pte_basic_t)(pfn) << PFN_SHIFT_OFFSET) |\
495 pgprot_val(prot))
496#define mk_pte(page, prot) pfn_pte(page_to_pfn(page), prot)
497
498/*
499 * ZERO_PAGE is a global shared page that is always zero: used
500 * for zero-mapped memory areas etc..
501 */
502extern unsigned long empty_zero_page[1024];
503#define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
504
505#endif /* __ASSEMBLY__ */
506
507#define pte_none(pte) ((pte_val(pte) & ~_PTE_NONE_MASK) == 0)
508#define pte_present(pte) (pte_val(pte) & _PAGE_PRESENT)
509#define pte_clear(mm,addr,ptep) do { set_pte_at((mm), (addr), (ptep), __pte(0)); } while (0)
510
511#define pmd_none(pmd) (!pmd_val(pmd))
512#define pmd_bad(pmd) (pmd_val(pmd) & _PMD_BAD)
513#define pmd_present(pmd) (pmd_val(pmd) & _PMD_PRESENT_MASK)
514#define pmd_clear(pmdp) do { pmd_val(*(pmdp)) = 0; } while (0)
515
516#ifndef __ASSEMBLY__
517/*
518 * The "pgd_xxx()" functions here are trivial for a folded two-level
519 * setup: the pgd is never bad, and a pmd always exists (as it's folded
520 * into the pgd entry)
521 */
522static inline int pgd_none(pgd_t pgd) { return 0; }
523static inline int pgd_bad(pgd_t pgd) { return 0; }
524static inline int pgd_present(pgd_t pgd) { return 1; }
525#define pgd_clear(xp) do { } while (0)
526
527#define pgd_page_vaddr(pgd) \
528 ((unsigned long) __va(pgd_val(pgd) & PAGE_MASK))
529
530/*
531 * The following only work if pte_present() is true.
532 * Undefined behaviour if not..
533 */
534static inline int pte_read(pte_t pte) { return pte_val(pte) & _PAGE_USER; }
535static inline int pte_write(pte_t pte) { return pte_val(pte) & _PAGE_RW; }
536static inline int pte_exec(pte_t pte) { return pte_val(pte) & _PAGE_EXEC; }
537static inline int pte_dirty(pte_t pte) { return pte_val(pte) & _PAGE_DIRTY; }
538static inline int pte_young(pte_t pte) { return pte_val(pte) & _PAGE_ACCESSED; }
539static inline int pte_file(pte_t pte) { return pte_val(pte) & _PAGE_FILE; }
540
541static inline void pte_uncache(pte_t pte) { pte_val(pte) |= _PAGE_NO_CACHE; }
542static inline void pte_cache(pte_t pte) { pte_val(pte) &= ~_PAGE_NO_CACHE; }
543
544static inline pte_t pte_rdprotect(pte_t pte) {
545 pte_val(pte) &= ~_PAGE_USER; return pte; }
546static inline pte_t pte_wrprotect(pte_t pte) {
547 pte_val(pte) &= ~(_PAGE_RW | _PAGE_HWWRITE); return pte; }
548static inline pte_t pte_exprotect(pte_t pte) {
549 pte_val(pte) &= ~_PAGE_EXEC; return pte; }
550static inline pte_t pte_mkclean(pte_t pte) {
551 pte_val(pte) &= ~(_PAGE_DIRTY | _PAGE_HWWRITE); return pte; }
552static inline pte_t pte_mkold(pte_t pte) {
553 pte_val(pte) &= ~_PAGE_ACCESSED; return pte; }
554
555static inline pte_t pte_mkread(pte_t pte) {
556 pte_val(pte) |= _PAGE_USER; return pte; }
557static inline pte_t pte_mkexec(pte_t pte) {
558 pte_val(pte) |= _PAGE_USER | _PAGE_EXEC; return pte; }
559static inline pte_t pte_mkwrite(pte_t pte) {
560 pte_val(pte) |= _PAGE_RW; return pte; }
561static inline pte_t pte_mkdirty(pte_t pte) {
562 pte_val(pte) |= _PAGE_DIRTY; return pte; }
563static inline pte_t pte_mkyoung(pte_t pte) {
564 pte_val(pte) |= _PAGE_ACCESSED; return pte; }
565
566static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
567{
568 pte_val(pte) = (pte_val(pte) & _PAGE_CHG_MASK) | pgprot_val(newprot);
569 return pte;
570}
571
572/*
573 * When flushing the tlb entry for a page, we also need to flush the hash
574 * table entry. flush_hash_pages is assembler (for speed) in hashtable.S.
575 */
576extern int flush_hash_pages(unsigned context, unsigned long va,
577 unsigned long pmdval, int count);
578
579/* Add an HPTE to the hash table */
580extern void add_hash_page(unsigned context, unsigned long va,
581 unsigned long pmdval);
582
583/*
584 * Atomic PTE updates.
585 *
586 * pte_update clears and sets bit atomically, and returns
587 * the old pte value. In the 64-bit PTE case we lock around the
588 * low PTE word since we expect ALL flag bits to be there
589 */
590#ifndef CONFIG_PTE_64BIT
591static inline unsigned long pte_update(pte_t *p, unsigned long clr,
592 unsigned long set)
593{
594 unsigned long old, tmp;
595
596 __asm__ __volatile__("\
5971: lwarx %0,0,%3\n\
598 andc %1,%0,%4\n\
599 or %1,%1,%5\n"
600 PPC405_ERR77(0,%3)
601" stwcx. %1,0,%3\n\
602 bne- 1b"
603 : "=&r" (old), "=&r" (tmp), "=m" (*p)
604 : "r" (p), "r" (clr), "r" (set), "m" (*p)
605 : "cc" );
606 return old;
607}
608#else
609static inline unsigned long long pte_update(pte_t *p, unsigned long clr,
610 unsigned long set)
611{
612 unsigned long long old;
613 unsigned long tmp;
614
615 __asm__ __volatile__("\
6161: lwarx %L0,0,%4\n\
617 lwzx %0,0,%3\n\
618 andc %1,%L0,%5\n\
619 or %1,%1,%6\n"
620 PPC405_ERR77(0,%3)
621" stwcx. %1,0,%4\n\
622 bne- 1b"
623 : "=&r" (old), "=&r" (tmp), "=m" (*p)
624 : "r" (p), "r" ((unsigned long)(p) + 4), "r" (clr), "r" (set), "m" (*p)
625 : "cc" );
626 return old;
627}
628#endif
629
630/*
631 * set_pte stores a linux PTE into the linux page table.
632 * On machines which use an MMU hash table we avoid changing the
633 * _PAGE_HASHPTE bit.
634 */
635static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
636 pte_t *ptep, pte_t pte)
637{
638#if _PAGE_HASHPTE != 0
639 pte_update(ptep, ~_PAGE_HASHPTE, pte_val(pte) & ~_PAGE_HASHPTE);
640#else
641 *ptep = pte;
642#endif
643}
644
645/*
646 * 2.6 calles this without flushing the TLB entry, this is wrong
647 * for our hash-based implementation, we fix that up here
648 */
649#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
650static inline int __ptep_test_and_clear_young(unsigned int context, unsigned long addr, pte_t *ptep)
651{
652 unsigned long old;
653 old = pte_update(ptep, _PAGE_ACCESSED, 0);
654#if _PAGE_HASHPTE != 0
655 if (old & _PAGE_HASHPTE) {
656 unsigned long ptephys = __pa(ptep) & PAGE_MASK;
657 flush_hash_pages(context, addr, ptephys, 1);
658 }
659#endif
660 return (old & _PAGE_ACCESSED) != 0;
661}
662#define ptep_test_and_clear_young(__vma, __addr, __ptep) \
663 __ptep_test_and_clear_young((__vma)->vm_mm->context.id, __addr, __ptep)
664
665#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
666static inline int ptep_test_and_clear_dirty(struct vm_area_struct *vma,
667 unsigned long addr, pte_t *ptep)
668{
669 return (pte_update(ptep, (_PAGE_DIRTY | _PAGE_HWWRITE), 0) & _PAGE_DIRTY) != 0;
670}
671
672#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
673static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
674 pte_t *ptep)
675{
676 return __pte(pte_update(ptep, ~_PAGE_HASHPTE, 0));
677}
678
679#define __HAVE_ARCH_PTEP_SET_WRPROTECT
680static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
681 pte_t *ptep)
682{
683 pte_update(ptep, (_PAGE_RW | _PAGE_HWWRITE), 0);
684}
685
686#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
687static inline void __ptep_set_access_flags(pte_t *ptep, pte_t entry, int dirty)
688{
689 unsigned long bits = pte_val(entry) &
690 (_PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_RW);
691 pte_update(ptep, 0, bits);
692}
693
694#define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
695 do { \
696 __ptep_set_access_flags(__ptep, __entry, __dirty); \
697 flush_tlb_page_nohash(__vma, __address); \
698 } while(0)
699
700/*
701 * Macro to mark a page protection value as "uncacheable".
702 */
703#define pgprot_noncached(prot) (__pgprot(pgprot_val(prot) | _PAGE_NO_CACHE | _PAGE_GUARDED))
704
705struct file;
706extern pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
707 unsigned long size, pgprot_t vma_prot);
708#define __HAVE_PHYS_MEM_ACCESS_PROT
709
710#define __HAVE_ARCH_PTE_SAME
711#define pte_same(A,B) (((pte_val(A) ^ pte_val(B)) & ~_PAGE_HASHPTE) == 0)
712
713/*
714 * Note that on Book E processors, the pmd contains the kernel virtual
715 * (lowmem) address of the pte page. The physical address is less useful
716 * because everything runs with translation enabled (even the TLB miss
717 * handler). On everything else the pmd contains the physical address
718 * of the pte page. -- paulus
719 */
720#ifndef CONFIG_BOOKE
721#define pmd_page_vaddr(pmd) \
722 ((unsigned long) __va(pmd_val(pmd) & PAGE_MASK))
723#define pmd_page(pmd) \
724 (mem_map + (pmd_val(pmd) >> PAGE_SHIFT))
725#else
726#define pmd_page_vaddr(pmd) \
727 ((unsigned long) (pmd_val(pmd) & PAGE_MASK))
728#define pmd_page(pmd) \
729 (mem_map + (__pa(pmd_val(pmd)) >> PAGE_SHIFT))
730#endif
731
732/* to find an entry in a kernel page-table-directory */
733#define pgd_offset_k(address) pgd_offset(&init_mm, address)
734
735/* to find an entry in a page-table-directory */
736#define pgd_index(address) ((address) >> PGDIR_SHIFT)
737#define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
738
739/* Find an entry in the second-level page table.. */
740static inline pmd_t * pmd_offset(pgd_t * dir, unsigned long address)
741{
742 return (pmd_t *) dir;
743}
744
745/* Find an entry in the third-level page table.. */
746#define pte_index(address) \
747 (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
748#define pte_offset_kernel(dir, addr) \
749 ((pte_t *) pmd_page_vaddr(*(dir)) + pte_index(addr))
750#define pte_offset_map(dir, addr) \
751 ((pte_t *) kmap_atomic(pmd_page(*(dir)), KM_PTE0) + pte_index(addr))
752#define pte_offset_map_nested(dir, addr) \
753 ((pte_t *) kmap_atomic(pmd_page(*(dir)), KM_PTE1) + pte_index(addr))
754
755#define pte_unmap(pte) kunmap_atomic(pte, KM_PTE0)
756#define pte_unmap_nested(pte) kunmap_atomic(pte, KM_PTE1)
757
758extern pgd_t swapper_pg_dir[PTRS_PER_PGD];
759
760extern void paging_init(void);
761
762/*
763 * Encode and decode a swap entry.
764 * Note that the bits we use in a PTE for representing a swap entry
765 * must not include the _PAGE_PRESENT bit, the _PAGE_FILE bit, or the
766 *_PAGE_HASHPTE bit (if used). -- paulus
767 */
768#define __swp_type(entry) ((entry).val & 0x1f)
769#define __swp_offset(entry) ((entry).val >> 5)
770#define __swp_entry(type, offset) ((swp_entry_t) { (type) | ((offset) << 5) })
771#define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) >> 3 })
772#define __swp_entry_to_pte(x) ((pte_t) { (x).val << 3 })
773
774/* Encode and decode a nonlinear file mapping entry */
775#define PTE_FILE_MAX_BITS 29
776#define pte_to_pgoff(pte) (pte_val(pte) >> 3)
777#define pgoff_to_pte(off) ((pte_t) { ((off) << 3) | _PAGE_FILE })
778
779/* CONFIG_APUS */
780/* For virtual address to physical address conversion */
781extern void cache_clear(__u32 addr, int length);
782extern void cache_push(__u32 addr, int length);
783extern int mm_end_of_chunk (unsigned long addr, int len);
784extern unsigned long iopa(unsigned long addr);
785extern unsigned long mm_ptov(unsigned long addr) __attribute_const__;
786
787/* Values for nocacheflag and cmode */
788/* These are not used by the APUS kernel_map, but prevents
789 compilation errors. */
790#define KERNELMAP_FULL_CACHING 0
791#define KERNELMAP_NOCACHE_SER 1
792#define KERNELMAP_NOCACHE_NONSER 2
793#define KERNELMAP_NO_COPYBACK 3
794
795/*
796 * Map some physical address range into the kernel address space.
797 */
798extern unsigned long kernel_map(unsigned long paddr, unsigned long size,
799 int nocacheflag, unsigned long *memavailp );
800
801/*
802 * Set cache mode of (kernel space) address range.
803 */
804extern void kernel_set_cachemode (unsigned long address, unsigned long size,
805 unsigned int cmode);
806
807/* Needs to be defined here and not in linux/mm.h, as it is arch dependent */
808#define kern_addr_valid(addr) (1)
809
810#ifdef CONFIG_PHYS_64BIT
811extern int remap_pfn_range(struct vm_area_struct *vma, unsigned long from,
812 unsigned long paddr, unsigned long size, pgprot_t prot);
813
814static inline int io_remap_pfn_range(struct vm_area_struct *vma,
815 unsigned long vaddr,
816 unsigned long pfn,
817 unsigned long size,
818 pgprot_t prot)
819{
820 phys_addr_t paddr64 = fixup_bigphys_addr(pfn << PAGE_SHIFT, size);
821 return remap_pfn_range(vma, vaddr, paddr64 >> PAGE_SHIFT, size, prot);
822}
823#else
824#define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
825 remap_pfn_range(vma, vaddr, pfn, size, prot)
826#endif
827
828/*
829 * No page table caches to initialise
830 */
831#define pgtable_cache_init() do { } while (0)
832
833extern int get_pteptr(struct mm_struct *mm, unsigned long addr, pte_t **ptep,
834 pmd_t **pmdp);
835
836#endif /* !__ASSEMBLY__ */
837
838#endif /* _ASM_POWERPC_PGTABLE_PPC32_H */
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_ */
diff --git a/include/asm-powerpc/pgtable.h b/include/asm-powerpc/pgtable.h
index 19edb6982b81..78bf4ae712a6 100644
--- a/include/asm-powerpc/pgtable.h
+++ b/include/asm-powerpc/pgtable.h
@@ -2,502 +2,15 @@
2#define _ASM_POWERPC_PGTABLE_H 2#define _ASM_POWERPC_PGTABLE_H
3#ifdef __KERNEL__ 3#ifdef __KERNEL__
4 4
5#ifndef CONFIG_PPC64 5#if defined(CONFIG_PPC64)
6#include <asm-ppc/pgtable.h> 6# include <asm/pgtable-ppc64.h>
7#else 7#else
8 8# include <asm/pgtable-ppc32.h>
9/*
10 * This file contains the functions and defines necessary to modify and use
11 * the ppc64 hashed page table.
12 */
13
14#ifndef __ASSEMBLY__
15#include <linux/stddef.h>
16#include <asm/processor.h> /* For TASK_SIZE */
17#include <asm/mmu.h>
18#include <asm/page.h>
19#include <asm/tlbflush.h>
20struct mm_struct;
21#endif /* __ASSEMBLY__ */
22
23#ifdef CONFIG_PPC_64K_PAGES
24#include <asm/pgtable-64k.h>
25#else
26#include <asm/pgtable-4k.h>
27#endif
28
29#define FIRST_USER_ADDRESS 0
30
31/*
32 * Size of EA range mapped by our pagetables.
33 */
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)
37
38#if TASK_SIZE_USER64 > PGTABLE_RANGE
39#error TASK_SIZE_USER64 exceeds pagetable range
40#endif
41
42#if TASK_SIZE_USER64 > (1UL << (USER_ESID_BITS + SID_SHIFT))
43#error TASK_SIZE_USER64 exceeds user VSID range
44#endif 9#endif
45 10
46/*
47 * Define the address range of the vmalloc VM area.
48 */
49#define VMALLOC_START ASM_CONST(0xD000000000000000)
50#define VMALLOC_SIZE ASM_CONST(0x80000000000)
51#define VMALLOC_END (VMALLOC_START + VMALLOC_SIZE)
52
53/*
54 * Define the address range of the imalloc VM area.
55 */
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)
59
60/*
61 * Region IDs
62 */
63#define REGION_SHIFT 60UL
64#define REGION_MASK (0xfUL << REGION_SHIFT)
65#define REGION_ID(ea) (((unsigned long)(ea)) >> REGION_SHIFT)
66
67#define VMALLOC_REGION_ID (REGION_ID(VMALLOC_START))
68#define KERNEL_REGION_ID (REGION_ID(PAGE_OFFSET))
69#define USER_REGION_ID (0UL)
70
71/*
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
75 */
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 */
89
90#define _PAGE_BASE (_PAGE_PRESENT | _PAGE_ACCESSED | _PAGE_COHERENT)
91
92#define _PAGE_WRENABLE (_PAGE_RW | _PAGE_DIRTY)
93
94/* __pgprot defined in asm-powerpc/page.h */
95#define PAGE_NONE __pgprot(_PAGE_PRESENT | _PAGE_ACCESSED)
96
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)
107
108#define PAGE_AGP __pgprot(_PAGE_BASE | _PAGE_WRENABLE | _PAGE_NO_CACHE)
109#define HAVE_PAGE_AGP
110
111/* PTEIDX nibble */
112#define _PTEIDX_SECONDARY 0x8
113#define _PTEIDX_GROUP_IX 0x7
114
115
116/*
117 * POWER4 and newer have per page execute protection, older chips can only
118 * do this on a segment (256MB) basis.
119 *
120 * Also, write permissions imply read permissions.
121 * This is the closest we can get..
122 *
123 * Note due to the way vm flags are laid out, the bits are XWR
124 */
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
133
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
142
143#ifndef __ASSEMBLY__ 11#ifndef __ASSEMBLY__
144
145/*
146 * ZERO_PAGE is a global shared page that is always zero: used
147 * for zero-mapped memory areas etc..
148 */
149extern unsigned long empty_zero_page[PAGE_SIZE/sizeof(unsigned long)];
150#define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
151#endif /* __ASSEMBLY__ */
152
153#ifdef CONFIG_HUGETLB_PAGE
154
155#define HAVE_ARCH_UNMAPPED_AREA
156#define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
157
158#endif
159
160#ifndef __ASSEMBLY__
161
162/*
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.
165 *
166 * mk_pte takes a (struct page *) as input
167 */
168#define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
169
170static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot)
171{
172 pte_t pte;
173
174
175 pte_val(pte) = (pfn << PTE_RPN_SHIFT) | pgprot_val(pgprot);
176 return pte;
177}
178
179#define pte_modify(_pte, newprot) \
180 (__pte((pte_val(_pte) & _PAGE_CHG_MASK) | pgprot_val(newprot)))
181
182#define pte_none(pte) ((pte_val(pte) & ~_PAGE_HPTEFLAGS) == 0)
183#define pte_present(pte) (pte_val(pte) & _PAGE_PRESENT)
184
185/* pte_clear moved to later in this file */
186
187#define pte_pfn(x) ((unsigned long)((pte_val(x)>>PTE_RPN_SHIFT)))
188#define pte_page(x) pfn_to_page(pte_pfn(x))
189
190#define PMD_BAD_BITS (PTE_TABLE_SIZE-1)
191#define PUD_BAD_BITS (PMD_TABLE_SIZE-1)
192
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))
201
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))
210
211#define pgd_set(pgdp, pudp) ({pgd_val(*(pgdp)) = (unsigned long)(pudp);})
212
213/*
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.
216 */
217/* to avoid overflow in free_pgtables we don't use PTRS_PER_PGD here */
218#define pgd_index(address) (((address) >> (PGDIR_SHIFT)) & 0x1ff)
219
220#define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
221
222#define pmd_offset(pudp,addr) \
223 (((pmd_t *) pud_page_vaddr(*(pudp))) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1)))
224
225#define pte_offset_kernel(dir,addr) \
226 (((pte_t *) pmd_page_vaddr(*(dir))) + (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)))
227
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)
232
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)
236
237/*
238 * The following only work if pte_present() is true.
239 * Undefined behaviour if not..
240 */
241static inline int pte_read(pte_t pte) { return pte_val(pte) & _PAGE_USER;}
242static inline int pte_write(pte_t pte) { return pte_val(pte) & _PAGE_RW;}
243static inline int pte_exec(pte_t pte) { return pte_val(pte) & _PAGE_EXEC;}
244static inline int pte_dirty(pte_t pte) { return pte_val(pte) & _PAGE_DIRTY;}
245static inline int pte_young(pte_t pte) { return pte_val(pte) & _PAGE_ACCESSED;}
246static inline int pte_file(pte_t pte) { return pte_val(pte) & _PAGE_FILE;}
247
248static inline void pte_uncache(pte_t pte) { pte_val(pte) |= _PAGE_NO_CACHE; }
249static inline void pte_cache(pte_t pte) { pte_val(pte) &= ~_PAGE_NO_CACHE; }
250
251static inline pte_t pte_rdprotect(pte_t pte) {
252 pte_val(pte) &= ~_PAGE_USER; return pte; }
253static inline pte_t pte_exprotect(pte_t pte) {
254 pte_val(pte) &= ~_PAGE_EXEC; return pte; }
255static inline pte_t pte_wrprotect(pte_t pte) {
256 pte_val(pte) &= ~(_PAGE_RW); return pte; }
257static inline pte_t pte_mkclean(pte_t pte) {
258 pte_val(pte) &= ~(_PAGE_DIRTY); return pte; }
259static inline pte_t pte_mkold(pte_t pte) {
260 pte_val(pte) &= ~_PAGE_ACCESSED; return pte; }
261static inline pte_t pte_mkread(pte_t pte) {
262 pte_val(pte) |= _PAGE_USER; return pte; }
263static inline pte_t pte_mkexec(pte_t pte) {
264 pte_val(pte) |= _PAGE_USER | _PAGE_EXEC; return pte; }
265static inline pte_t pte_mkwrite(pte_t pte) {
266 pte_val(pte) |= _PAGE_RW; return pte; }
267static inline pte_t pte_mkdirty(pte_t pte) {
268 pte_val(pte) |= _PAGE_DIRTY; return pte; }
269static inline pte_t pte_mkyoung(pte_t pte) {
270 pte_val(pte) |= _PAGE_ACCESSED; return pte; }
271static inline pte_t pte_mkhuge(pte_t pte) {
272 return pte; }
273
274/* Atomic PTE updates */
275static inline unsigned long pte_update(struct mm_struct *mm,
276 unsigned long addr,
277 pte_t *ptep, unsigned long clr,
278 int huge)
279{
280 unsigned long old, tmp;
281
282 __asm__ __volatile__(
283 "1: ldarx %0,0,%3 # pte_update\n\
284 andi. %1,%0,%6\n\
285 bne- 1b \n\
286 andc %1,%0,%4 \n\
287 stdcx. %1,0,%3 \n\
288 bne- 1b"
289 : "=&r" (old), "=&r" (tmp), "=m" (*ptep)
290 : "r" (ptep), "r" (clr), "m" (*ptep), "i" (_PAGE_BUSY)
291 : "cc" );
292
293 if (old & _PAGE_HASHPTE)
294 hpte_need_flush(mm, addr, ptep, old, huge);
295 return old;
296}
297
298static inline int __ptep_test_and_clear_young(struct mm_struct *mm,
299 unsigned long addr, pte_t *ptep)
300{
301 unsigned long old;
302
303 if ((pte_val(*ptep) & (_PAGE_ACCESSED | _PAGE_HASHPTE)) == 0)
304 return 0;
305 old = pte_update(mm, addr, ptep, _PAGE_ACCESSED, 0);
306 return (old & _PAGE_ACCESSED) != 0;
307}
308#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
309#define ptep_test_and_clear_young(__vma, __addr, __ptep) \
310({ \
311 int __r; \
312 __r = __ptep_test_and_clear_young((__vma)->vm_mm, __addr, __ptep); \
313 __r; \
314})
315
316/*
317 * On RW/DIRTY bit transitions we can avoid flushing the hpte. For the
318 * moment we always flush but we need to fix hpte_update and test if the
319 * optimisation is worth it.
320 */
321static inline int __ptep_test_and_clear_dirty(struct mm_struct *mm,
322 unsigned long addr, pte_t *ptep)
323{
324 unsigned long old;
325
326 if ((pte_val(*ptep) & _PAGE_DIRTY) == 0)
327 return 0;
328 old = pte_update(mm, addr, ptep, _PAGE_DIRTY, 0);
329 return (old & _PAGE_DIRTY) != 0;
330}
331#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
332#define ptep_test_and_clear_dirty(__vma, __addr, __ptep) \
333({ \
334 int __r; \
335 __r = __ptep_test_and_clear_dirty((__vma)->vm_mm, __addr, __ptep); \
336 __r; \
337})
338
339#define __HAVE_ARCH_PTEP_SET_WRPROTECT
340static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
341 pte_t *ptep)
342{
343 unsigned long old;
344
345 if ((pte_val(*ptep) & _PAGE_RW) == 0)
346 return;
347 old = pte_update(mm, addr, ptep, _PAGE_RW, 0);
348}
349
350/*
351 * We currently remove entries from the hashtable regardless of whether
352 * the entry was young or dirty. The generic routines only flush if the
353 * entry was young or dirty which is not good enough.
354 *
355 * We should be more intelligent about this but for the moment we override
356 * these functions and force a tlb flush unconditionally
357 */
358#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
359#define ptep_clear_flush_young(__vma, __address, __ptep) \
360({ \
361 int __young = __ptep_test_and_clear_young((__vma)->vm_mm, __address, \
362 __ptep); \
363 __young; \
364})
365
366#define __HAVE_ARCH_PTEP_CLEAR_DIRTY_FLUSH
367#define ptep_clear_flush_dirty(__vma, __address, __ptep) \
368({ \
369 int __dirty = __ptep_test_and_clear_dirty((__vma)->vm_mm, __address, \
370 __ptep); \
371 __dirty; \
372})
373
374#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
375static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
376 unsigned long addr, pte_t *ptep)
377{
378 unsigned long old = pte_update(mm, addr, ptep, ~0UL, 0);
379 return __pte(old);
380}
381
382static inline void pte_clear(struct mm_struct *mm, unsigned long addr,
383 pte_t * ptep)
384{
385 pte_update(mm, addr, ptep, ~0UL, 0);
386}
387
388/*
389 * set_pte stores a linux PTE into the linux page table.
390 */
391static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
392 pte_t *ptep, pte_t pte)
393{
394 if (pte_present(*ptep))
395 pte_clear(mm, addr, ptep);
396 pte = __pte(pte_val(pte) & ~_PAGE_HPTEFLAGS);
397 *ptep = pte;
398}
399
400/* Set the dirty and/or accessed bits atomically in a linux PTE, this
401 * function doesn't need to flush the hash entry
402 */
403#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
404static inline void __ptep_set_access_flags(pte_t *ptep, pte_t entry, int dirty)
405{
406 unsigned long bits = pte_val(entry) &
407 (_PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_RW | _PAGE_EXEC);
408 unsigned long old, tmp;
409
410 __asm__ __volatile__(
411 "1: ldarx %0,0,%4\n\
412 andi. %1,%0,%6\n\
413 bne- 1b \n\
414 or %0,%3,%0\n\
415 stdcx. %0,0,%4\n\
416 bne- 1b"
417 :"=&r" (old), "=&r" (tmp), "=m" (*ptep)
418 :"r" (bits), "r" (ptep), "m" (*ptep), "i" (_PAGE_BUSY)
419 :"cc");
420}
421#define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
422 do { \
423 __ptep_set_access_flags(__ptep, __entry, __dirty); \
424 flush_tlb_page_nohash(__vma, __address); \
425 } while(0)
426
427/*
428 * Macro to mark a page protection value as "uncacheable".
429 */
430#define pgprot_noncached(prot) (__pgprot(pgprot_val(prot) | _PAGE_NO_CACHE | _PAGE_GUARDED))
431
432struct file;
433extern pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
434 unsigned long size, pgprot_t vma_prot);
435#define __HAVE_PHYS_MEM_ACCESS_PROT
436
437#define __HAVE_ARCH_PTE_SAME
438#define pte_same(A,B) (((pte_val(A) ^ pte_val(B)) & ~_PAGE_HPTEFLAGS) == 0)
439
440#define pte_ERROR(e) \
441 printk("%s:%d: bad pte %08lx.\n", __FILE__, __LINE__, pte_val(e))
442#define pmd_ERROR(e) \
443 printk("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pmd_val(e))
444#define pgd_ERROR(e) \
445 printk("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e))
446
447extern pgd_t swapper_pg_dir[];
448
449extern void paging_init(void);
450
451/* Encode and de-code a swap entry */
452#define __swp_type(entry) (((entry).val >> 1) & 0x3f)
453#define __swp_offset(entry) ((entry).val >> 8)
454#define __swp_entry(type, offset) ((swp_entry_t){((type)<< 1)|((offset)<<8)})
455#define __pte_to_swp_entry(pte) ((swp_entry_t){pte_val(pte) >> PTE_RPN_SHIFT})
456#define __swp_entry_to_pte(x) ((pte_t) { (x).val << PTE_RPN_SHIFT })
457#define pte_to_pgoff(pte) (pte_val(pte) >> PTE_RPN_SHIFT)
458#define pgoff_to_pte(off) ((pte_t) {((off) << PTE_RPN_SHIFT)|_PAGE_FILE})
459#define PTE_FILE_MAX_BITS (BITS_PER_LONG - PTE_RPN_SHIFT)
460
461/*
462 * kern_addr_valid is intended to indicate whether an address is a valid
463 * kernel address. Most 32-bit archs define it as always true (like this)
464 * but most 64-bit archs actually perform a test. What should we do here?
465 * The only use is in fs/ncpfs/dir.c
466 */
467#define kern_addr_valid(addr) (1)
468
469#define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
470 remap_pfn_range(vma, vaddr, pfn, size, prot)
471
472void pgtable_cache_init(void);
473
474/*
475 * find_linux_pte returns the address of a linux pte for a given
476 * effective address and directory. If not found, it returns zero.
477 */static inline pte_t *find_linux_pte(pgd_t *pgdir, unsigned long ea)
478{
479 pgd_t *pg;
480 pud_t *pu;
481 pmd_t *pm;
482 pte_t *pt = NULL;
483
484 pg = pgdir + pgd_index(ea);
485 if (!pgd_none(*pg)) {
486 pu = pud_offset(pg, ea);
487 if (!pud_none(*pu)) {
488 pm = pmd_offset(pu, ea);
489 if (pmd_present(*pm))
490 pt = pte_offset_kernel(pm, ea);
491 }
492 }
493 return pt;
494}
495
496
497#include <asm-generic/pgtable.h> 12#include <asm-generic/pgtable.h>
498
499#endif /* __ASSEMBLY__ */ 13#endif /* __ASSEMBLY__ */
500 14
501#endif /* CONFIG_PPC64 */
502#endif /* __KERNEL__ */ 15#endif /* __KERNEL__ */
503#endif /* _ASM_POWERPC_PGTABLE_H */ 16#endif /* _ASM_POWERPC_PGTABLE_H */
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-12 15:05:19 -0500