aboutsummaryrefslogtreecommitdiffstats
path: root/include/asm-ppc64
diff options
context:
space:
mode:
Diffstat (limited to 'include/asm-ppc64')
-rw-r--r--include/asm-ppc64/mmu.h208
-rw-r--r--include/asm-ppc64/mmu_context.h15
-rw-r--r--include/asm-ppc64/paca.h13
-rw-r--r--include/asm-ppc64/page.h147
-rw-r--r--include/asm-ppc64/pgalloc.h47
-rw-r--r--include/asm-ppc64/pgtable-4k.h88
-rw-r--r--include/asm-ppc64/pgtable-64k.h87
-rw-r--r--include/asm-ppc64/pgtable.h160
-rw-r--r--include/asm-ppc64/prom.h8
-rw-r--r--include/asm-ppc64/system.h2
10 files changed, 532 insertions, 243 deletions
diff --git a/include/asm-ppc64/mmu.h b/include/asm-ppc64/mmu.h
index e0505acb77d9..4c18a5cb69f5 100644
--- a/include/asm-ppc64/mmu.h
+++ b/include/asm-ppc64/mmu.h
@@ -48,13 +48,21 @@ extern char initial_stab[];
48 48
49/* Bits in the SLB VSID word */ 49/* Bits in the SLB VSID word */
50#define SLB_VSID_SHIFT 12 50#define SLB_VSID_SHIFT 12
51#define SLB_VSID_B ASM_CONST(0xc000000000000000)
52#define SLB_VSID_B_256M ASM_CONST(0x0000000000000000)
53#define SLB_VSID_B_1T ASM_CONST(0x4000000000000000)
51#define SLB_VSID_KS ASM_CONST(0x0000000000000800) 54#define SLB_VSID_KS ASM_CONST(0x0000000000000800)
52#define SLB_VSID_KP ASM_CONST(0x0000000000000400) 55#define SLB_VSID_KP ASM_CONST(0x0000000000000400)
53#define SLB_VSID_N ASM_CONST(0x0000000000000200) /* no-execute */ 56#define SLB_VSID_N ASM_CONST(0x0000000000000200) /* no-execute */
54#define SLB_VSID_L ASM_CONST(0x0000000000000100) /* largepage */ 57#define SLB_VSID_L ASM_CONST(0x0000000000000100)
55#define SLB_VSID_C ASM_CONST(0x0000000000000080) /* class */ 58#define SLB_VSID_C ASM_CONST(0x0000000000000080) /* class */
56#define SLB_VSID_LS ASM_CONST(0x0000000000000070) /* size of largepage */ 59#define SLB_VSID_LP ASM_CONST(0x0000000000000030)
57 60#define SLB_VSID_LP_00 ASM_CONST(0x0000000000000000)
61#define SLB_VSID_LP_01 ASM_CONST(0x0000000000000010)
62#define SLB_VSID_LP_10 ASM_CONST(0x0000000000000020)
63#define SLB_VSID_LP_11 ASM_CONST(0x0000000000000030)
64#define SLB_VSID_LLP (SLB_VSID_L|SLB_VSID_LP)
65
58#define SLB_VSID_KERNEL (SLB_VSID_KP) 66#define SLB_VSID_KERNEL (SLB_VSID_KP)
59#define SLB_VSID_USER (SLB_VSID_KP|SLB_VSID_KS|SLB_VSID_C) 67#define SLB_VSID_USER (SLB_VSID_KP|SLB_VSID_KS|SLB_VSID_C)
60 68
@@ -69,6 +77,7 @@ extern char initial_stab[];
69#define HPTE_V_AVPN_SHIFT 7 77#define HPTE_V_AVPN_SHIFT 7
70#define HPTE_V_AVPN ASM_CONST(0xffffffffffffff80) 78#define HPTE_V_AVPN ASM_CONST(0xffffffffffffff80)
71#define HPTE_V_AVPN_VAL(x) (((x) & HPTE_V_AVPN) >> HPTE_V_AVPN_SHIFT) 79#define HPTE_V_AVPN_VAL(x) (((x) & HPTE_V_AVPN) >> HPTE_V_AVPN_SHIFT)
80#define HPTE_V_COMPARE(x,y) (!(((x) ^ (y)) & HPTE_V_AVPN))
72#define HPTE_V_BOLTED ASM_CONST(0x0000000000000010) 81#define HPTE_V_BOLTED ASM_CONST(0x0000000000000010)
73#define HPTE_V_LOCK ASM_CONST(0x0000000000000008) 82#define HPTE_V_LOCK ASM_CONST(0x0000000000000008)
74#define HPTE_V_LARGE ASM_CONST(0x0000000000000004) 83#define HPTE_V_LARGE ASM_CONST(0x0000000000000004)
@@ -81,6 +90,7 @@ extern char initial_stab[];
81#define HPTE_R_RPN ASM_CONST(0x3ffffffffffff000) 90#define HPTE_R_RPN ASM_CONST(0x3ffffffffffff000)
82#define HPTE_R_FLAGS ASM_CONST(0x00000000000003ff) 91#define HPTE_R_FLAGS ASM_CONST(0x00000000000003ff)
83#define HPTE_R_PP ASM_CONST(0x0000000000000003) 92#define HPTE_R_PP ASM_CONST(0x0000000000000003)
93#define HPTE_R_N ASM_CONST(0x0000000000000004)
84 94
85/* Values for PP (assumes Ks=0, Kp=1) */ 95/* Values for PP (assumes Ks=0, Kp=1) */
86/* pp0 will always be 0 for linux */ 96/* pp0 will always be 0 for linux */
@@ -99,100 +109,120 @@ typedef struct {
99extern hpte_t *htab_address; 109extern hpte_t *htab_address;
100extern unsigned long htab_hash_mask; 110extern unsigned long htab_hash_mask;
101 111
102static inline unsigned long hpt_hash(unsigned long vpn, int large) 112/*
113 * Page size definition
114 *
115 * shift : is the "PAGE_SHIFT" value for that page size
116 * sllp : is a bit mask with the value of SLB L || LP to be or'ed
117 * directly to a slbmte "vsid" value
118 * penc : is the HPTE encoding mask for the "LP" field:
119 *
120 */
121struct mmu_psize_def
103{ 122{
104 unsigned long vsid; 123 unsigned int shift; /* number of bits */
105 unsigned long page; 124 unsigned int penc; /* HPTE encoding */
106 125 unsigned int tlbiel; /* tlbiel supported for that page size */
107 if (large) { 126 unsigned long avpnm; /* bits to mask out in AVPN in the HPTE */
108 vsid = vpn >> 4; 127 unsigned long sllp; /* SLB L||LP (exact mask to use in slbmte) */
109 page = vpn & 0xf; 128};
110 } else {
111 vsid = vpn >> 16;
112 page = vpn & 0xffff;
113 }
114 129
115 return (vsid & 0x7fffffffffUL) ^ page; 130#endif /* __ASSEMBLY__ */
116}
117
118static inline void __tlbie(unsigned long va, int large)
119{
120 /* clear top 16 bits, non SLS segment */
121 va &= ~(0xffffULL << 48);
122
123 if (large) {
124 va &= HPAGE_MASK;
125 asm volatile("tlbie %0,1" : : "r"(va) : "memory");
126 } else {
127 va &= PAGE_MASK;
128 asm volatile("tlbie %0,0" : : "r"(va) : "memory");
129 }
130}
131 131
132static inline void tlbie(unsigned long va, int large) 132/*
133{ 133 * The kernel use the constants below to index in the page sizes array.
134 asm volatile("ptesync": : :"memory"); 134 * The use of fixed constants for this purpose is better for performances
135 __tlbie(va, large); 135 * of the low level hash refill handlers.
136 asm volatile("eieio; tlbsync; ptesync": : :"memory"); 136 *
137} 137 * A non supported page size has a "shift" field set to 0
138 *
139 * Any new page size being implemented can get a new entry in here. Whether
140 * the kernel will use it or not is a different matter though. The actual page
141 * size used by hugetlbfs is not defined here and may be made variable
142 */
138 143
139static inline void __tlbiel(unsigned long va) 144#define MMU_PAGE_4K 0 /* 4K */
140{ 145#define MMU_PAGE_64K 1 /* 64K */
141 /* clear top 16 bits, non SLS segment */ 146#define MMU_PAGE_64K_AP 2 /* 64K Admixed (in a 4K segment) */
142 va &= ~(0xffffULL << 48); 147#define MMU_PAGE_1M 3 /* 1M */
143 va &= PAGE_MASK; 148#define MMU_PAGE_16M 4 /* 16M */
144 149#define MMU_PAGE_16G 5 /* 16G */
145 /* 150#define MMU_PAGE_COUNT 6
146 * Thanks to Alan Modra we are now able to use machine specific
147 * assembly instructions (like tlbiel) by using the gas -many flag.
148 * However we have to support older toolchains so for the moment
149 * we hardwire it.
150 */
151#if 0
152 asm volatile("tlbiel %0" : : "r"(va) : "memory");
153#else
154 asm volatile(".long 0x7c000224 | (%0 << 11)" : : "r"(va) : "memory");
155#endif
156}
157 151
158static inline void tlbiel(unsigned long va) 152#ifndef __ASSEMBLY__
159{
160 asm volatile("ptesync": : :"memory");
161 __tlbiel(va);
162 asm volatile("ptesync": : :"memory");
163}
164 153
165static inline unsigned long slot2va(unsigned long hpte_v, unsigned long slot) 154/*
166{ 155 * The current system page sizes
167 unsigned long avpn = HPTE_V_AVPN_VAL(hpte_v); 156 */
168 unsigned long va; 157extern struct mmu_psize_def mmu_psize_defs[MMU_PAGE_COUNT];
158extern int mmu_linear_psize;
159extern int mmu_virtual_psize;
169 160
170 va = avpn << 23; 161#ifdef CONFIG_HUGETLB_PAGE
162/*
163 * The page size index of the huge pages for use by hugetlbfs
164 */
165extern int mmu_huge_psize;
171 166
172 if (! (hpte_v & HPTE_V_LARGE)) { 167#endif /* CONFIG_HUGETLB_PAGE */
173 unsigned long vpi, pteg;
174 168
175 pteg = slot / HPTES_PER_GROUP; 169/*
176 if (hpte_v & HPTE_V_SECONDARY) 170 * This function sets the AVPN and L fields of the HPTE appropriately
177 pteg = ~pteg; 171 * for the page size
172 */
173static inline unsigned long hpte_encode_v(unsigned long va, int psize)
174{
175 unsigned long v =
176 v = (va >> 23) & ~(mmu_psize_defs[psize].avpnm);
177 v <<= HPTE_V_AVPN_SHIFT;
178 if (psize != MMU_PAGE_4K)
179 v |= HPTE_V_LARGE;
180 return v;
181}
178 182
179 vpi = ((va >> 28) ^ pteg) & htab_hash_mask; 183/*
184 * This function sets the ARPN, and LP fields of the HPTE appropriately
185 * for the page size. We assume the pa is already "clean" that is properly
186 * aligned for the requested page size
187 */
188static inline unsigned long hpte_encode_r(unsigned long pa, int psize)
189{
190 unsigned long r;
180 191
181 va |= vpi << PAGE_SHIFT; 192 /* A 4K page needs no special encoding */
193 if (psize == MMU_PAGE_4K)
194 return pa & HPTE_R_RPN;
195 else {
196 unsigned int penc = mmu_psize_defs[psize].penc;
197 unsigned int shift = mmu_psize_defs[psize].shift;
198 return (pa & ~((1ul << shift) - 1)) | (penc << 12);
182 } 199 }
183 200 return r;
184 return va;
185} 201}
186 202
187/* 203/*
188 * Handle a fault by adding an HPTE. If the address can't be determined 204 * This hashes a virtual address for a 256Mb segment only for now
189 * to be valid via Linux page tables, return 1. If handled return 0
190 */ 205 */
191extern int __hash_page(unsigned long ea, unsigned long access, 206
192 unsigned long vsid, pte_t *ptep, unsigned long trap, 207static inline unsigned long hpt_hash(unsigned long va, unsigned int shift)
193 int local); 208{
209 return ((va >> 28) & 0x7fffffffffUL) ^ ((va & 0x0fffffffUL) >> shift);
210}
211
212extern int __hash_page_4K(unsigned long ea, unsigned long access,
213 unsigned long vsid, pte_t *ptep, unsigned long trap,
214 unsigned int local);
215extern int __hash_page_64K(unsigned long ea, unsigned long access,
216 unsigned long vsid, pte_t *ptep, unsigned long trap,
217 unsigned int local);
218struct mm_struct;
219extern int hash_huge_page(struct mm_struct *mm, unsigned long access,
220 unsigned long ea, unsigned long vsid, int local);
194 221
195extern void htab_finish_init(void); 222extern void htab_finish_init(void);
223extern int htab_bolt_mapping(unsigned long vstart, unsigned long vend,
224 unsigned long pstart, unsigned long mode,
225 int psize);
196 226
197extern void hpte_init_native(void); 227extern void hpte_init_native(void);
198extern void hpte_init_lpar(void); 228extern void hpte_init_lpar(void);
@@ -200,17 +230,21 @@ extern void hpte_init_iSeries(void);
200 230
201extern long pSeries_lpar_hpte_insert(unsigned long hpte_group, 231extern long pSeries_lpar_hpte_insert(unsigned long hpte_group,
202 unsigned long va, unsigned long prpn, 232 unsigned long va, unsigned long prpn,
203 unsigned long vflags, 233 unsigned long rflags,
204 unsigned long rflags); 234 unsigned long vflags, int psize);
205extern long native_hpte_insert(unsigned long hpte_group, unsigned long va, 235
206 unsigned long prpn, 236extern long native_hpte_insert(unsigned long hpte_group,
207 unsigned long vflags, unsigned long rflags); 237 unsigned long va, unsigned long prpn,
238 unsigned long rflags,
239 unsigned long vflags, int psize);
208 240
209extern long iSeries_hpte_bolt_or_insert(unsigned long hpte_group, 241extern long iSeries_hpte_insert(unsigned long hpte_group,
210 unsigned long va, unsigned long prpn, 242 unsigned long va, unsigned long prpn,
211 unsigned long vflags, unsigned long rflags); 243 unsigned long rflags,
244 unsigned long vflags, int psize);
212 245
213extern void stabs_alloc(void); 246extern void stabs_alloc(void);
247extern void slb_initialize(void);
214 248
215#endif /* __ASSEMBLY__ */ 249#endif /* __ASSEMBLY__ */
216 250
diff --git a/include/asm-ppc64/mmu_context.h b/include/asm-ppc64/mmu_context.h
index 820dd729b895..4f512e9fa6b8 100644
--- a/include/asm-ppc64/mmu_context.h
+++ b/include/asm-ppc64/mmu_context.h
@@ -16,8 +16,16 @@
16 * 2 of the License, or (at your option) any later version. 16 * 2 of the License, or (at your option) any later version.
17 */ 17 */
18 18
19static inline void enter_lazy_tlb(struct mm_struct *mm, struct task_struct *tsk) 19/*
20 * Getting into a kernel thread, there is no valid user segment, mark
21 * paca->pgdir NULL so that SLB miss on user addresses will fault
22 */
23static inline void enter_lazy_tlb(struct mm_struct *mm,
24 struct task_struct *tsk)
20{ 25{
26#ifdef CONFIG_PPC_64K_PAGES
27 get_paca()->pgdir = NULL;
28#endif /* CONFIG_PPC_64K_PAGES */
21} 29}
22 30
23#define NO_CONTEXT 0 31#define NO_CONTEXT 0
@@ -40,8 +48,13 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
40 cpu_set(smp_processor_id(), next->cpu_vm_mask); 48 cpu_set(smp_processor_id(), next->cpu_vm_mask);
41 49
42 /* No need to flush userspace segments if the mm doesnt change */ 50 /* No need to flush userspace segments if the mm doesnt change */
51#ifdef CONFIG_PPC_64K_PAGES
52 if (prev == next && get_paca()->pgdir == next->pgd)
53 return;
54#else
43 if (prev == next) 55 if (prev == next)
44 return; 56 return;
57#endif /* CONFIG_PPC_64K_PAGES */
45 58
46#ifdef CONFIG_ALTIVEC 59#ifdef CONFIG_ALTIVEC
47 if (cpu_has_feature(CPU_FTR_ALTIVEC)) 60 if (cpu_has_feature(CPU_FTR_ALTIVEC))
diff --git a/include/asm-ppc64/paca.h b/include/asm-ppc64/paca.h
index f68fe91debaf..bccacd6aa93a 100644
--- a/include/asm-ppc64/paca.h
+++ b/include/asm-ppc64/paca.h
@@ -72,10 +72,15 @@ struct paca_struct {
72 /* 72 /*
73 * Now, starting in cacheline 2, the exception save areas 73 * Now, starting in cacheline 2, the exception save areas
74 */ 74 */
75 u64 exgen[8] __attribute__((aligned(0x80))); /* used for most interrupts/exceptions */ 75 /* used for most interrupts/exceptions */
76 u64 exmc[8]; /* used for machine checks */ 76 u64 exgen[10] __attribute__((aligned(0x80)));
77 u64 exslb[8]; /* used for SLB/segment table misses 77 u64 exmc[10]; /* used for machine checks */
78 * on the linear mapping */ 78 u64 exslb[10]; /* used for SLB/segment table misses
79 * on the linear mapping */
80#ifdef CONFIG_PPC_64K_PAGES
81 pgd_t *pgdir;
82#endif /* CONFIG_PPC_64K_PAGES */
83
79 mm_context_t context; 84 mm_context_t context;
80 u16 slb_cache[SLB_CACHE_ENTRIES]; 85 u16 slb_cache[SLB_CACHE_ENTRIES];
81 u16 slb_cache_ptr; 86 u16 slb_cache_ptr;
diff --git a/include/asm-ppc64/page.h b/include/asm-ppc64/page.h
index d404431f0a9a..82ce187e5be8 100644
--- a/include/asm-ppc64/page.h
+++ b/include/asm-ppc64/page.h
@@ -13,32 +13,59 @@
13#include <linux/config.h> 13#include <linux/config.h>
14#include <asm/ppc_asm.h> /* for ASM_CONST */ 14#include <asm/ppc_asm.h> /* for ASM_CONST */
15 15
16/* PAGE_SHIFT determines the page size */ 16/*
17#define PAGE_SHIFT 12 17 * We support either 4k or 64k software page size. When using 64k pages
18#define PAGE_SIZE (ASM_CONST(1) << PAGE_SHIFT) 18 * however, wether we are really supporting 64k pages in HW or not is
19#define PAGE_MASK (~(PAGE_SIZE-1)) 19 * irrelevant to those definitions. We always define HW_PAGE_SHIFT to 12
20 * as use of 64k pages remains a linux kernel specific, every notion of
21 * page number shared with the firmware, TCEs, iommu, etc... still assumes
22 * a page size of 4096.
23 */
24#ifdef CONFIG_PPC_64K_PAGES
25#define PAGE_SHIFT 16
26#else
27#define PAGE_SHIFT 12
28#endif
20 29
21#define SID_SHIFT 28 30#define PAGE_SIZE (ASM_CONST(1) << PAGE_SHIFT)
22#define SID_MASK 0xfffffffffUL 31#define PAGE_MASK (~(PAGE_SIZE-1))
23#define ESID_MASK 0xfffffffff0000000UL
24#define GET_ESID(x) (((x) >> SID_SHIFT) & SID_MASK)
25 32
26#define HPAGE_SHIFT 24 33/* HW_PAGE_SHIFT is always 4k pages */
27#define HPAGE_SIZE ((1UL) << HPAGE_SHIFT) 34#define HW_PAGE_SHIFT 12
28#define HPAGE_MASK (~(HPAGE_SIZE - 1)) 35#define HW_PAGE_SIZE (ASM_CONST(1) << HW_PAGE_SHIFT)
36#define HW_PAGE_MASK (~(HW_PAGE_SIZE-1))
29 37
30#ifdef CONFIG_HUGETLB_PAGE 38/* PAGE_FACTOR is the number of bits factor between PAGE_SHIFT and
39 * HW_PAGE_SHIFT, that is 4k pages
40 */
41#define PAGE_FACTOR (PAGE_SHIFT - HW_PAGE_SHIFT)
42
43/* Segment size */
44#define SID_SHIFT 28
45#define SID_MASK 0xfffffffffUL
46#define ESID_MASK 0xfffffffff0000000UL
47#define GET_ESID(x) (((x) >> SID_SHIFT) & SID_MASK)
31 48
49/* Large pages size */
50
51#ifndef __ASSEMBLY__
52extern unsigned int HPAGE_SHIFT;
53#define HPAGE_SIZE ((1UL) << HPAGE_SHIFT)
54#define HPAGE_MASK (~(HPAGE_SIZE - 1))
32#define HUGETLB_PAGE_ORDER (HPAGE_SHIFT - PAGE_SHIFT) 55#define HUGETLB_PAGE_ORDER (HPAGE_SHIFT - PAGE_SHIFT)
56#endif /* __ASSEMBLY__ */
57
58#ifdef CONFIG_HUGETLB_PAGE
59
33 60
34#define HTLB_AREA_SHIFT 40 61#define HTLB_AREA_SHIFT 40
35#define HTLB_AREA_SIZE (1UL << HTLB_AREA_SHIFT) 62#define HTLB_AREA_SIZE (1UL << HTLB_AREA_SHIFT)
36#define GET_HTLB_AREA(x) ((x) >> HTLB_AREA_SHIFT) 63#define GET_HTLB_AREA(x) ((x) >> HTLB_AREA_SHIFT)
37 64
38#define LOW_ESID_MASK(addr, len) (((1U << (GET_ESID(addr+len-1)+1)) \ 65#define LOW_ESID_MASK(addr, len) (((1U << (GET_ESID(addr+len-1)+1)) \
39 - (1U << GET_ESID(addr))) & 0xffff) 66 - (1U << GET_ESID(addr))) & 0xffff)
40#define HTLB_AREA_MASK(addr, len) (((1U << (GET_HTLB_AREA(addr+len-1)+1)) \ 67#define HTLB_AREA_MASK(addr, len) (((1U << (GET_HTLB_AREA(addr+len-1)+1)) \
41 - (1U << GET_HTLB_AREA(addr))) & 0xffff) 68 - (1U << GET_HTLB_AREA(addr))) & 0xffff)
42 69
43#define ARCH_HAS_HUGEPAGE_ONLY_RANGE 70#define ARCH_HAS_HUGEPAGE_ONLY_RANGE
44#define ARCH_HAS_PREPARE_HUGEPAGE_RANGE 71#define ARCH_HAS_PREPARE_HUGEPAGE_RANGE
@@ -114,7 +141,25 @@ static __inline__ void clear_page(void *addr)
114 : "ctr", "memory"); 141 : "ctr", "memory");
115} 142}
116 143
117extern void copy_page(void *to, void *from); 144extern void copy_4K_page(void *to, void *from);
145
146#ifdef CONFIG_PPC_64K_PAGES
147static inline void copy_page(void *to, void *from)
148{
149 unsigned int i;
150 for (i=0; i < (1 << (PAGE_SHIFT - 12)); i++) {
151 copy_4K_page(to, from);
152 to += 4096;
153 from += 4096;
154 }
155}
156#else /* CONFIG_PPC_64K_PAGES */
157static inline void copy_page(void *to, void *from)
158{
159 copy_4K_page(to, from);
160}
161#endif /* CONFIG_PPC_64K_PAGES */
162
118struct page; 163struct page;
119extern void clear_user_page(void *page, unsigned long vaddr, struct page *pg); 164extern void clear_user_page(void *page, unsigned long vaddr, struct page *pg);
120extern void copy_user_page(void *to, void *from, unsigned long vaddr, struct page *p); 165extern void copy_user_page(void *to, void *from, unsigned long vaddr, struct page *p);
@@ -124,43 +169,75 @@ extern void copy_user_page(void *to, void *from, unsigned long vaddr, struct pag
124 * These are used to make use of C type-checking. 169 * These are used to make use of C type-checking.
125 * Entries in the pte table are 64b, while entries in the pgd & pmd are 32b. 170 * Entries in the pte table are 64b, while entries in the pgd & pmd are 32b.
126 */ 171 */
127typedef struct { unsigned long pte; } pte_t;
128typedef struct { unsigned long pmd; } pmd_t;
129typedef struct { unsigned long pud; } pud_t;
130typedef struct { unsigned long pgd; } pgd_t;
131typedef struct { unsigned long pgprot; } pgprot_t;
132 172
173/* PTE level */
174typedef struct { unsigned long pte; } pte_t;
133#define pte_val(x) ((x).pte) 175#define pte_val(x) ((x).pte)
134#define pmd_val(x) ((x).pmd)
135#define pud_val(x) ((x).pud)
136#define pgd_val(x) ((x).pgd)
137#define pgprot_val(x) ((x).pgprot)
138
139#define __pte(x) ((pte_t) { (x) }) 176#define __pte(x) ((pte_t) { (x) })
177
178/* 64k pages additionally define a bigger "real PTE" type that gathers
179 * the "second half" part of the PTE for pseudo 64k pages
180 */
181#ifdef CONFIG_PPC_64K_PAGES
182typedef struct { pte_t pte; unsigned long hidx; } real_pte_t;
183#else
184typedef struct { pte_t pte; } real_pte_t;
185#endif
186
187/* PMD level */
188typedef struct { unsigned long pmd; } pmd_t;
189#define pmd_val(x) ((x).pmd)
140#define __pmd(x) ((pmd_t) { (x) }) 190#define __pmd(x) ((pmd_t) { (x) })
191
192/* PUD level exusts only on 4k pages */
193#ifndef CONFIG_PPC_64K_PAGES
194typedef struct { unsigned long pud; } pud_t;
195#define pud_val(x) ((x).pud)
141#define __pud(x) ((pud_t) { (x) }) 196#define __pud(x) ((pud_t) { (x) })
197#endif
198
199/* PGD level */
200typedef struct { unsigned long pgd; } pgd_t;
201#define pgd_val(x) ((x).pgd)
142#define __pgd(x) ((pgd_t) { (x) }) 202#define __pgd(x) ((pgd_t) { (x) })
203
204/* Page protection bits */
205typedef struct { unsigned long pgprot; } pgprot_t;
206#define pgprot_val(x) ((x).pgprot)
143#define __pgprot(x) ((pgprot_t) { (x) }) 207#define __pgprot(x) ((pgprot_t) { (x) })
144 208
145#else 209#else
210
146/* 211/*
147 * .. while these make it easier on the compiler 212 * .. while these make it easier on the compiler
148 */ 213 */
149typedef unsigned long pte_t;
150typedef unsigned long pmd_t;
151typedef unsigned long pud_t;
152typedef unsigned long pgd_t;
153typedef unsigned long pgprot_t;
154 214
215typedef unsigned long pte_t;
155#define pte_val(x) (x) 216#define pte_val(x) (x)
217#define __pte(x) (x)
218
219#ifdef CONFIG_PPC_64K_PAGES
220typedef struct { pte_t pte; unsigned long hidx; } real_pte_t;
221#else
222typedef unsigned long real_pte_t;
223#endif
224
225
226typedef unsigned long pmd_t;
156#define pmd_val(x) (x) 227#define pmd_val(x) (x)
228#define __pmd(x) (x)
229
230#ifndef CONFIG_PPC_64K_PAGES
231typedef unsigned long pud_t;
157#define pud_val(x) (x) 232#define pud_val(x) (x)
233#define __pud(x) (x)
234#endif
235
236typedef unsigned long pgd_t;
158#define pgd_val(x) (x) 237#define pgd_val(x) (x)
159#define pgprot_val(x) (x) 238#define pgprot_val(x) (x)
160 239
161#define __pte(x) (x) 240typedef unsigned long pgprot_t;
162#define __pmd(x) (x)
163#define __pud(x) (x)
164#define __pgd(x) (x) 241#define __pgd(x) (x)
165#define __pgprot(x) (x) 242#define __pgprot(x) (x)
166 243
diff --git a/include/asm-ppc64/pgalloc.h b/include/asm-ppc64/pgalloc.h
index 26bc49c1108d..98da0e4262bd 100644
--- a/include/asm-ppc64/pgalloc.h
+++ b/include/asm-ppc64/pgalloc.h
@@ -8,10 +8,16 @@
8 8
9extern kmem_cache_t *pgtable_cache[]; 9extern kmem_cache_t *pgtable_cache[];
10 10
11#ifdef CONFIG_PPC_64K_PAGES
12#define PTE_CACHE_NUM 0
13#define PMD_CACHE_NUM 0
14#define PGD_CACHE_NUM 1
15#else
11#define PTE_CACHE_NUM 0 16#define PTE_CACHE_NUM 0
12#define PMD_CACHE_NUM 1 17#define PMD_CACHE_NUM 1
13#define PUD_CACHE_NUM 1 18#define PUD_CACHE_NUM 1
14#define PGD_CACHE_NUM 0 19#define PGD_CACHE_NUM 0
20#endif
15 21
16/* 22/*
17 * This program is free software; you can redistribute it and/or 23 * This program is free software; you can redistribute it and/or
@@ -30,6 +36,8 @@ static inline void pgd_free(pgd_t *pgd)
30 kmem_cache_free(pgtable_cache[PGD_CACHE_NUM], pgd); 36 kmem_cache_free(pgtable_cache[PGD_CACHE_NUM], pgd);
31} 37}
32 38
39#ifndef CONFIG_PPC_64K_PAGES
40
33#define pgd_populate(MM, PGD, PUD) pgd_set(PGD, PUD) 41#define pgd_populate(MM, PGD, PUD) pgd_set(PGD, PUD)
34 42
35static inline pud_t *pud_alloc_one(struct mm_struct *mm, unsigned long addr) 43static inline pud_t *pud_alloc_one(struct mm_struct *mm, unsigned long addr)
@@ -43,7 +51,30 @@ static inline void pud_free(pud_t *pud)
43 kmem_cache_free(pgtable_cache[PUD_CACHE_NUM], pud); 51 kmem_cache_free(pgtable_cache[PUD_CACHE_NUM], pud);
44} 52}
45 53
46#define pud_populate(MM, PUD, PMD) pud_set(PUD, PMD) 54static inline void pud_populate(struct mm_struct *mm, pud_t *pud, pmd_t *pmd)
55{
56 pud_set(pud, (unsigned long)pmd);
57}
58
59#define pmd_populate(mm, pmd, pte_page) \
60 pmd_populate_kernel(mm, pmd, page_address(pte_page))
61#define pmd_populate_kernel(mm, pmd, pte) pmd_set(pmd, (unsigned long)(pte))
62
63
64#else /* CONFIG_PPC_64K_PAGES */
65
66#define pud_populate(mm, pud, pmd) pud_set(pud, (unsigned long)pmd)
67
68static inline void pmd_populate_kernel(struct mm_struct *mm, pmd_t *pmd,
69 pte_t *pte)
70{
71 pmd_set(pmd, (unsigned long)pte);
72}
73
74#define pmd_populate(mm, pmd, pte_page) \
75 pmd_populate_kernel(mm, pmd, page_address(pte_page))
76
77#endif /* CONFIG_PPC_64K_PAGES */
47 78
48static inline pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long addr) 79static inline pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long addr)
49{ 80{
@@ -56,17 +87,15 @@ static inline void pmd_free(pmd_t *pmd)
56 kmem_cache_free(pgtable_cache[PMD_CACHE_NUM], pmd); 87 kmem_cache_free(pgtable_cache[PMD_CACHE_NUM], pmd);
57} 88}
58 89
59#define pmd_populate_kernel(mm, pmd, pte) pmd_set(pmd, pte) 90static inline pte_t *pte_alloc_one_kernel(struct mm_struct *mm,
60#define pmd_populate(mm, pmd, pte_page) \ 91 unsigned long address)
61 pmd_populate_kernel(mm, pmd, page_address(pte_page))
62
63static inline pte_t *pte_alloc_one_kernel(struct mm_struct *mm, unsigned long address)
64{ 92{
65 return kmem_cache_alloc(pgtable_cache[PTE_CACHE_NUM], 93 return kmem_cache_alloc(pgtable_cache[PTE_CACHE_NUM],
66 GFP_KERNEL|__GFP_REPEAT); 94 GFP_KERNEL|__GFP_REPEAT);
67} 95}
68 96
69static inline struct page *pte_alloc_one(struct mm_struct *mm, unsigned long address) 97static inline struct page *pte_alloc_one(struct mm_struct *mm,
98 unsigned long address)
70{ 99{
71 return virt_to_page(pte_alloc_one_kernel(mm, address)); 100 return virt_to_page(pte_alloc_one_kernel(mm, address));
72} 101}
@@ -103,7 +132,7 @@ static inline void pgtable_free(pgtable_free_t pgf)
103 kmem_cache_free(pgtable_cache[cachenum], p); 132 kmem_cache_free(pgtable_cache[cachenum], p);
104} 133}
105 134
106void pgtable_free_tlb(struct mmu_gather *tlb, pgtable_free_t pgf); 135extern void pgtable_free_tlb(struct mmu_gather *tlb, pgtable_free_t pgf);
107 136
108#define __pte_free_tlb(tlb, ptepage) \ 137#define __pte_free_tlb(tlb, ptepage) \
109 pgtable_free_tlb(tlb, pgtable_free_cache(page_address(ptepage), \ 138 pgtable_free_tlb(tlb, pgtable_free_cache(page_address(ptepage), \
@@ -111,9 +140,11 @@ void pgtable_free_tlb(struct mmu_gather *tlb, pgtable_free_t pgf);
111#define __pmd_free_tlb(tlb, pmd) \ 140#define __pmd_free_tlb(tlb, pmd) \
112 pgtable_free_tlb(tlb, pgtable_free_cache(pmd, \ 141 pgtable_free_tlb(tlb, pgtable_free_cache(pmd, \
113 PMD_CACHE_NUM, PMD_TABLE_SIZE-1)) 142 PMD_CACHE_NUM, PMD_TABLE_SIZE-1))
143#ifndef CONFIG_PPC_64K_PAGES
114#define __pud_free_tlb(tlb, pmd) \ 144#define __pud_free_tlb(tlb, pmd) \
115 pgtable_free_tlb(tlb, pgtable_free_cache(pud, \ 145 pgtable_free_tlb(tlb, pgtable_free_cache(pud, \
116 PUD_CACHE_NUM, PUD_TABLE_SIZE-1)) 146 PUD_CACHE_NUM, PUD_TABLE_SIZE-1))
147#endif /* CONFIG_PPC_64K_PAGES */
117 148
118#define check_pgt_cache() do { } while (0) 149#define check_pgt_cache() do { } while (0)
119 150
diff --git a/include/asm-ppc64/pgtable-4k.h b/include/asm-ppc64/pgtable-4k.h
new file mode 100644
index 000000000000..c883a2748558
--- /dev/null
+++ b/include/asm-ppc64/pgtable-4k.h
@@ -0,0 +1,88 @@
1/*
2 * 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
4 * each entry by assuming that each entry is page aligned.
5 */
6#define PTE_INDEX_SIZE 9
7#define PMD_INDEX_SIZE 7
8#define PUD_INDEX_SIZE 7
9#define PGD_INDEX_SIZE 9
10
11#define PTE_TABLE_SIZE (sizeof(pte_t) << PTE_INDEX_SIZE)
12#define PMD_TABLE_SIZE (sizeof(pmd_t) << PMD_INDEX_SIZE)
13#define PUD_TABLE_SIZE (sizeof(pud_t) << PUD_INDEX_SIZE)
14#define PGD_TABLE_SIZE (sizeof(pgd_t) << PGD_INDEX_SIZE)
15
16#define PTRS_PER_PTE (1 << PTE_INDEX_SIZE)
17#define PTRS_PER_PMD (1 << PMD_INDEX_SIZE)
18#define PTRS_PER_PUD (1 << PMD_INDEX_SIZE)
19#define PTRS_PER_PGD (1 << PGD_INDEX_SIZE)
20
21/* PMD_SHIFT determines what a second-level page table entry can map */
22#define PMD_SHIFT (PAGE_SHIFT + PTE_INDEX_SIZE)
23#define PMD_SIZE (1UL << PMD_SHIFT)
24#define PMD_MASK (~(PMD_SIZE-1))
25
26/* PUD_SHIFT determines what a third-level page table entry can map */
27#define PUD_SHIFT (PMD_SHIFT + PMD_INDEX_SIZE)
28#define PUD_SIZE (1UL << PUD_SHIFT)
29#define PUD_MASK (~(PUD_SIZE-1))
30
31/* PGDIR_SHIFT determines what a fourth-level page table entry can map */
32#define PGDIR_SHIFT (PUD_SHIFT + PUD_INDEX_SIZE)
33#define PGDIR_SIZE (1UL << PGDIR_SHIFT)
34#define PGDIR_MASK (~(PGDIR_SIZE-1))
35
36/* PTE bits */
37#define _PAGE_SECONDARY 0x8000 /* software: HPTE is in secondary group */
38#define _PAGE_GROUP_IX 0x7000 /* software: HPTE index within group */
39#define _PAGE_F_SECOND _PAGE_SECONDARY
40#define _PAGE_F_GIX _PAGE_GROUP_IX
41
42/* PTE flags to conserve for HPTE identification */
43#define _PAGE_HPTEFLAGS (_PAGE_BUSY | _PAGE_HASHPTE | \
44 _PAGE_SECONDARY | _PAGE_GROUP_IX)
45
46/* PAGE_MASK gives the right answer below, but only by accident */
47/* It should be preserving the high 48 bits and then specifically */
48/* preserving _PAGE_SECONDARY | _PAGE_GROUP_IX */
49#define _PAGE_CHG_MASK (PAGE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY | \
50 _PAGE_HPTEFLAGS)
51
52/* Bits to mask out from a PMD to get to the PTE page */
53#define PMD_MASKED_BITS 0
54/* Bits to mask out from a PUD to get to the PMD page */
55#define PUD_MASKED_BITS 0
56/* Bits to mask out from a PGD to get to the PUD page */
57#define PGD_MASKED_BITS 0
58
59/* shift to put page number into pte */
60#define PTE_RPN_SHIFT (17)
61
62#define __real_pte(e,p) ((real_pte_t)(e))
63#define __rpte_to_pte(r) (r)
64#define __rpte_to_hidx(r,index) (pte_val((r)) >> 12)
65
66#define pte_iterate_hashed_subpages(rpte, psize, va, index, shift) \
67 do { \
68 index = 0; \
69 shift = mmu_psize_defs[psize].shift; \
70
71#define pte_iterate_hashed_end() } while(0)
72
73/*
74 * 4-level page tables related bits
75 */
76
77#define pgd_none(pgd) (!pgd_val(pgd))
78#define pgd_bad(pgd) (pgd_val(pgd) == 0)
79#define pgd_present(pgd) (pgd_val(pgd) != 0)
80#define pgd_clear(pgdp) (pgd_val(*(pgdp)) = 0)
81#define pgd_page(pgd) (pgd_val(pgd) & ~PGD_MASKED_BITS)
82
83#define pud_offset(pgdp, addr) \
84 (((pud_t *) pgd_page(*(pgdp))) + \
85 (((addr) >> PUD_SHIFT) & (PTRS_PER_PUD - 1)))
86
87#define pud_ERROR(e) \
88 printk("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pud_val(e))
diff --git a/include/asm-ppc64/pgtable-64k.h b/include/asm-ppc64/pgtable-64k.h
new file mode 100644
index 000000000000..c5f437c86b3c
--- /dev/null
+++ b/include/asm-ppc64/pgtable-64k.h
@@ -0,0 +1,87 @@
1#include <asm-generic/pgtable-nopud.h>
2
3
4#define PTE_INDEX_SIZE 12
5#define PMD_INDEX_SIZE 12
6#define PUD_INDEX_SIZE 0
7#define PGD_INDEX_SIZE 4
8
9#define PTE_TABLE_SIZE (sizeof(real_pte_t) << PTE_INDEX_SIZE)
10#define PMD_TABLE_SIZE (sizeof(pmd_t) << PMD_INDEX_SIZE)
11#define PGD_TABLE_SIZE (sizeof(pgd_t) << PGD_INDEX_SIZE)
12
13#define PTRS_PER_PTE (1 << PTE_INDEX_SIZE)
14#define PTRS_PER_PMD (1 << PMD_INDEX_SIZE)
15#define PTRS_PER_PGD (1 << PGD_INDEX_SIZE)
16
17/* PMD_SHIFT determines what a second-level page table entry can map */
18#define PMD_SHIFT (PAGE_SHIFT + PTE_INDEX_SIZE)
19#define PMD_SIZE (1UL << PMD_SHIFT)
20#define PMD_MASK (~(PMD_SIZE-1))
21
22/* PGDIR_SHIFT determines what a third-level page table entry can map */
23#define PGDIR_SHIFT (PMD_SHIFT + PMD_INDEX_SIZE)
24#define PGDIR_SIZE (1UL << PGDIR_SHIFT)
25#define PGDIR_MASK (~(PGDIR_SIZE-1))
26
27/* Additional PTE bits (don't change without checking asm in hash_low.S) */
28#define _PAGE_HPTE_SUB 0x0ffff000 /* combo only: sub pages HPTE bits */
29#define _PAGE_HPTE_SUB0 0x08000000 /* combo only: first sub page */
30#define _PAGE_COMBO 0x10000000 /* this is a combo 4k page */
31#define _PAGE_F_SECOND 0x00008000 /* full page: hidx bits */
32#define _PAGE_F_GIX 0x00007000 /* full page: hidx bits */
33
34/* PTE flags to conserve for HPTE identification */
35#define _PAGE_HPTEFLAGS (_PAGE_BUSY | _PAGE_HASHPTE | _PAGE_HPTE_SUB |\
36 _PAGE_COMBO)
37
38/* Shift to put page number into pte.
39 *
40 * That gives us a max RPN of 32 bits, which means a max of 48 bits
41 * of addressable physical space.
42 * We could get 3 more bits here by setting PTE_RPN_SHIFT to 29 but
43 * 32 makes PTEs more readable for debugging for now :)
44 */
45#define PTE_RPN_SHIFT (32)
46#define PTE_RPN_MAX (1UL << (64 - PTE_RPN_SHIFT))
47#define PTE_RPN_MASK (~((1UL<<PTE_RPN_SHIFT)-1))
48
49/* _PAGE_CHG_MASK masks of bits that are to be preserved accross
50 * pgprot changes
51 */
52#define _PAGE_CHG_MASK (PTE_RPN_MASK | _PAGE_HPTEFLAGS | _PAGE_DIRTY | \
53 _PAGE_ACCESSED)
54
55/* Bits to mask out from a PMD to get to the PTE page */
56#define PMD_MASKED_BITS 0x1ff
57/* Bits to mask out from a PGD/PUD to get to the PMD page */
58#define PUD_MASKED_BITS 0x1ff
59
60#ifndef __ASSEMBLY__
61
62/* Manipulate "rpte" values */
63#define __real_pte(e,p) ((real_pte_t) { \
64 (e), pte_val(*((p) + PTRS_PER_PTE)) })
65#define __rpte_to_hidx(r,index) ((pte_val((r).pte) & _PAGE_COMBO) ? \
66 (((r).hidx >> ((index)<<2)) & 0xf) : ((pte_val((r).pte) >> 12) & 0xf))
67#define __rpte_to_pte(r) ((r).pte)
68#define __rpte_sub_valid(rpte, index) \
69 (pte_val(rpte.pte) & (_PAGE_HPTE_SUB0 >> (index)))
70
71
72/* Trick: we set __end to va + 64k, which happens works for
73 * a 16M page as well as we want only one iteration
74 */
75#define pte_iterate_hashed_subpages(rpte, psize, va, index, shift) \
76 do { \
77 unsigned long __end = va + PAGE_SIZE; \
78 unsigned __split = (psize == MMU_PAGE_4K || \
79 psize == MMU_PAGE_64K_AP); \
80 shift = mmu_psize_defs[psize].shift; \
81 for (index = 0; va < __end; index++, va += (1 << shift)) { \
82 if (!__split || __rpte_sub_valid(rpte, index)) do { \
83
84#define pte_iterate_hashed_end() } while(0); } } while(0)
85
86
87#endif /* __ASSEMBLY__ */
diff --git a/include/asm-ppc64/pgtable.h b/include/asm-ppc64/pgtable.h
index 8c3f574046b6..fde93ec36abc 100644
--- a/include/asm-ppc64/pgtable.h
+++ b/include/asm-ppc64/pgtable.h
@@ -15,40 +15,11 @@
15#include <asm/tlbflush.h> 15#include <asm/tlbflush.h>
16#endif /* __ASSEMBLY__ */ 16#endif /* __ASSEMBLY__ */
17 17
18/* 18#ifdef CONFIG_PPC_64K_PAGES
19 * Entries per page directory level. The PTE level must use a 64b record 19#include <asm/pgtable-64k.h>
20 * for each page table entry. The PMD and PGD level use a 32b record for 20#else
21 * each entry by assuming that each entry is page aligned. 21#include <asm/pgtable-4k.h>
22 */ 22#endif
23#define PTE_INDEX_SIZE 9
24#define PMD_INDEX_SIZE 7
25#define PUD_INDEX_SIZE 7
26#define PGD_INDEX_SIZE 9
27
28#define PTE_TABLE_SIZE (sizeof(pte_t) << PTE_INDEX_SIZE)
29#define PMD_TABLE_SIZE (sizeof(pmd_t) << PMD_INDEX_SIZE)
30#define PUD_TABLE_SIZE (sizeof(pud_t) << PUD_INDEX_SIZE)
31#define PGD_TABLE_SIZE (sizeof(pgd_t) << PGD_INDEX_SIZE)
32
33#define PTRS_PER_PTE (1 << PTE_INDEX_SIZE)
34#define PTRS_PER_PMD (1 << PMD_INDEX_SIZE)
35#define PTRS_PER_PUD (1 << PMD_INDEX_SIZE)
36#define PTRS_PER_PGD (1 << PGD_INDEX_SIZE)
37
38/* PMD_SHIFT determines what a second-level page table entry can map */
39#define PMD_SHIFT (PAGE_SHIFT + PTE_INDEX_SIZE)
40#define PMD_SIZE (1UL << PMD_SHIFT)
41#define PMD_MASK (~(PMD_SIZE-1))
42
43/* PUD_SHIFT determines what a third-level page table entry can map */
44#define PUD_SHIFT (PMD_SHIFT + PMD_INDEX_SIZE)
45#define PUD_SIZE (1UL << PUD_SHIFT)
46#define PUD_MASK (~(PUD_SIZE-1))
47
48/* PGDIR_SHIFT determines what a fourth-level page table entry can map */
49#define PGDIR_SHIFT (PUD_SHIFT + PUD_INDEX_SIZE)
50#define PGDIR_SIZE (1UL << PGDIR_SHIFT)
51#define PGDIR_MASK (~(PGDIR_SIZE-1))
52 23
53#define FIRST_USER_ADDRESS 0 24#define FIRST_USER_ADDRESS 0
54 25
@@ -75,8 +46,9 @@
75#define VMALLOC_END (VMALLOC_START + VMALLOC_SIZE) 46#define VMALLOC_END (VMALLOC_START + VMALLOC_SIZE)
76 47
77/* 48/*
78 * Bits in a linux-style PTE. These match the bits in the 49 * Common bits in a linux-style PTE. These match the bits in the
79 * (hardware-defined) PowerPC PTE as closely as possible. 50 * (hardware-defined) PowerPC PTE as closely as possible. Additional
51 * bits may be defined in pgtable-*.h
80 */ 52 */
81#define _PAGE_PRESENT 0x0001 /* software: pte contains a translation */ 53#define _PAGE_PRESENT 0x0001 /* software: pte contains a translation */
82#define _PAGE_USER 0x0002 /* matches one of the PP bits */ 54#define _PAGE_USER 0x0002 /* matches one of the PP bits */
@@ -91,15 +63,6 @@
91#define _PAGE_RW 0x0200 /* software: user write access allowed */ 63#define _PAGE_RW 0x0200 /* software: user write access allowed */
92#define _PAGE_HASHPTE 0x0400 /* software: pte has an associated HPTE */ 64#define _PAGE_HASHPTE 0x0400 /* software: pte has an associated HPTE */
93#define _PAGE_BUSY 0x0800 /* software: PTE & hash are busy */ 65#define _PAGE_BUSY 0x0800 /* software: PTE & hash are busy */
94#define _PAGE_SECONDARY 0x8000 /* software: HPTE is in secondary group */
95#define _PAGE_GROUP_IX 0x7000 /* software: HPTE index within group */
96#define _PAGE_HUGE 0x10000 /* 16MB page */
97/* Bits 0x7000 identify the index within an HPT Group */
98#define _PAGE_HPTEFLAGS (_PAGE_BUSY | _PAGE_HASHPTE | _PAGE_SECONDARY | _PAGE_GROUP_IX)
99/* PAGE_MASK gives the right answer below, but only by accident */
100/* It should be preserving the high 48 bits and then specifically */
101/* preserving _PAGE_SECONDARY | _PAGE_GROUP_IX */
102#define _PAGE_CHG_MASK (PAGE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY | _PAGE_HPTEFLAGS)
103 66
104#define _PAGE_BASE (_PAGE_PRESENT | _PAGE_ACCESSED | _PAGE_COHERENT) 67#define _PAGE_BASE (_PAGE_PRESENT | _PAGE_ACCESSED | _PAGE_COHERENT)
105 68
@@ -122,10 +85,10 @@
122#define PAGE_AGP __pgprot(_PAGE_BASE | _PAGE_WRENABLE | _PAGE_NO_CACHE) 85#define PAGE_AGP __pgprot(_PAGE_BASE | _PAGE_WRENABLE | _PAGE_NO_CACHE)
123#define HAVE_PAGE_AGP 86#define HAVE_PAGE_AGP
124 87
125/* 88/* PTEIDX nibble */
126 * This bit in a hardware PTE indicates that the page is *not* executable. 89#define _PTEIDX_SECONDARY 0x8
127 */ 90#define _PTEIDX_GROUP_IX 0x7
128#define HW_NO_EXEC _PAGE_EXEC 91
129 92
130/* 93/*
131 * POWER4 and newer have per page execute protection, older chips can only 94 * POWER4 and newer have per page execute protection, older chips can only
@@ -164,21 +127,10 @@ extern unsigned long empty_zero_page[PAGE_SIZE/sizeof(unsigned long)];
164#define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page)) 127#define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
165#endif /* __ASSEMBLY__ */ 128#endif /* __ASSEMBLY__ */
166 129
167/* shift to put page number into pte */
168#define PTE_SHIFT (17)
169
170#ifdef CONFIG_HUGETLB_PAGE 130#ifdef CONFIG_HUGETLB_PAGE
171 131
172#ifndef __ASSEMBLY__
173int hash_huge_page(struct mm_struct *mm, unsigned long access,
174 unsigned long ea, unsigned long vsid, int local);
175#endif /* __ASSEMBLY__ */
176
177#define HAVE_ARCH_UNMAPPED_AREA 132#define HAVE_ARCH_UNMAPPED_AREA
178#define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN 133#define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
179#else
180
181#define hash_huge_page(mm,a,ea,vsid,local) -1
182 134
183#endif 135#endif
184 136
@@ -197,7 +149,7 @@ static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot)
197 pte_t pte; 149 pte_t pte;
198 150
199 151
200 pte_val(pte) = (pfn << PTE_SHIFT) | pgprot_val(pgprot); 152 pte_val(pte) = (pfn << PTE_RPN_SHIFT) | pgprot_val(pgprot);
201 return pte; 153 return pte;
202} 154}
203 155
@@ -209,30 +161,25 @@ static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot)
209 161
210/* pte_clear moved to later in this file */ 162/* pte_clear moved to later in this file */
211 163
212#define pte_pfn(x) ((unsigned long)((pte_val(x) >> PTE_SHIFT))) 164#define pte_pfn(x) ((unsigned long)((pte_val(x)>>PTE_RPN_SHIFT)))
213#define pte_page(x) pfn_to_page(pte_pfn(x)) 165#define pte_page(x) pfn_to_page(pte_pfn(x))
214 166
215#define pmd_set(pmdp, ptep) ({BUG_ON((u64)ptep < KERNELBASE); pmd_val(*(pmdp)) = (unsigned long)(ptep);}) 167#define pmd_set(pmdp, pmdval) (pmd_val(*(pmdp)) = (pmdval))
216#define pmd_none(pmd) (!pmd_val(pmd)) 168#define pmd_none(pmd) (!pmd_val(pmd))
217#define pmd_bad(pmd) (pmd_val(pmd) == 0) 169#define pmd_bad(pmd) (pmd_val(pmd) == 0)
218#define pmd_present(pmd) (pmd_val(pmd) != 0) 170#define pmd_present(pmd) (pmd_val(pmd) != 0)
219#define pmd_clear(pmdp) (pmd_val(*(pmdp)) = 0) 171#define pmd_clear(pmdp) (pmd_val(*(pmdp)) = 0)
220#define pmd_page_kernel(pmd) (pmd_val(pmd)) 172#define pmd_page_kernel(pmd) (pmd_val(pmd) & ~PMD_MASKED_BITS)
221#define pmd_page(pmd) virt_to_page(pmd_page_kernel(pmd)) 173#define pmd_page(pmd) virt_to_page(pmd_page_kernel(pmd))
222 174
223#define pud_set(pudp, pmdp) (pud_val(*(pudp)) = (unsigned long)(pmdp)) 175#define pud_set(pudp, pudval) (pud_val(*(pudp)) = (pudval))
224#define pud_none(pud) (!pud_val(pud)) 176#define pud_none(pud) (!pud_val(pud))
225#define pud_bad(pud) ((pud_val(pud)) == 0) 177#define pud_bad(pud) ((pud_val(pud)) == 0)
226#define pud_present(pud) (pud_val(pud) != 0) 178#define pud_present(pud) (pud_val(pud) != 0)
227#define pud_clear(pudp) (pud_val(*(pudp)) = 0) 179#define pud_clear(pudp) (pud_val(*(pudp)) = 0)
228#define pud_page(pud) (pud_val(pud)) 180#define pud_page(pud) (pud_val(pud) & ~PUD_MASKED_BITS)
229 181
230#define pgd_set(pgdp, pudp) ({pgd_val(*(pgdp)) = (unsigned long)(pudp);}) 182#define pgd_set(pgdp, pudp) ({pgd_val(*(pgdp)) = (unsigned long)(pudp);})
231#define pgd_none(pgd) (!pgd_val(pgd))
232#define pgd_bad(pgd) (pgd_val(pgd) == 0)
233#define pgd_present(pgd) (pgd_val(pgd) != 0)
234#define pgd_clear(pgdp) (pgd_val(*(pgdp)) = 0)
235#define pgd_page(pgd) (pgd_val(pgd))
236 183
237/* 184/*
238 * Find an entry in a page-table-directory. We combine the address region 185 * Find an entry in a page-table-directory. We combine the address region
@@ -243,9 +190,6 @@ static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot)
243 190
244#define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address)) 191#define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
245 192
246#define pud_offset(pgdp, addr) \
247 (((pud_t *) pgd_page(*(pgdp))) + (((addr) >> PUD_SHIFT) & (PTRS_PER_PUD - 1)))
248
249#define pmd_offset(pudp,addr) \ 193#define pmd_offset(pudp,addr) \
250 (((pmd_t *) pud_page(*(pudp))) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1))) 194 (((pmd_t *) pud_page(*(pudp))) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1)))
251 195
@@ -271,7 +215,6 @@ static inline int pte_exec(pte_t pte) { return pte_val(pte) & _PAGE_EXEC;}
271static inline int pte_dirty(pte_t pte) { return pte_val(pte) & _PAGE_DIRTY;} 215static inline int pte_dirty(pte_t pte) { return pte_val(pte) & _PAGE_DIRTY;}
272static inline int pte_young(pte_t pte) { return pte_val(pte) & _PAGE_ACCESSED;} 216static inline int pte_young(pte_t pte) { return pte_val(pte) & _PAGE_ACCESSED;}
273static inline int pte_file(pte_t pte) { return pte_val(pte) & _PAGE_FILE;} 217static inline int pte_file(pte_t pte) { return pte_val(pte) & _PAGE_FILE;}
274static inline int pte_huge(pte_t pte) { return pte_val(pte) & _PAGE_HUGE;}
275 218
276static inline void pte_uncache(pte_t pte) { pte_val(pte) |= _PAGE_NO_CACHE; } 219static inline void pte_uncache(pte_t pte) { pte_val(pte) |= _PAGE_NO_CACHE; }
277static inline void pte_cache(pte_t pte) { pte_val(pte) &= ~_PAGE_NO_CACHE; } 220static inline void pte_cache(pte_t pte) { pte_val(pte) &= ~_PAGE_NO_CACHE; }
@@ -286,7 +229,6 @@ static inline pte_t pte_mkclean(pte_t pte) {
286 pte_val(pte) &= ~(_PAGE_DIRTY); return pte; } 229 pte_val(pte) &= ~(_PAGE_DIRTY); return pte; }
287static inline pte_t pte_mkold(pte_t pte) { 230static inline pte_t pte_mkold(pte_t pte) {
288 pte_val(pte) &= ~_PAGE_ACCESSED; return pte; } 231 pte_val(pte) &= ~_PAGE_ACCESSED; return pte; }
289
290static inline pte_t pte_mkread(pte_t pte) { 232static inline pte_t pte_mkread(pte_t pte) {
291 pte_val(pte) |= _PAGE_USER; return pte; } 233 pte_val(pte) |= _PAGE_USER; return pte; }
292static inline pte_t pte_mkexec(pte_t pte) { 234static inline pte_t pte_mkexec(pte_t pte) {
@@ -298,7 +240,7 @@ static inline pte_t pte_mkdirty(pte_t pte) {
298static inline pte_t pte_mkyoung(pte_t pte) { 240static inline pte_t pte_mkyoung(pte_t pte) {
299 pte_val(pte) |= _PAGE_ACCESSED; return pte; } 241 pte_val(pte) |= _PAGE_ACCESSED; return pte; }
300static inline pte_t pte_mkhuge(pte_t pte) { 242static inline pte_t pte_mkhuge(pte_t pte) {
301 pte_val(pte) |= _PAGE_HUGE; return pte; } 243 return pte; }
302 244
303/* Atomic PTE updates */ 245/* Atomic PTE updates */
304static inline unsigned long pte_update(pte_t *p, unsigned long clr) 246static inline unsigned long pte_update(pte_t *p, unsigned long clr)
@@ -321,11 +263,13 @@ static inline unsigned long pte_update(pte_t *p, unsigned long clr)
321/* PTE updating functions, this function puts the PTE in the 263/* PTE updating functions, this function puts the PTE in the
322 * batch, doesn't actually triggers the hash flush immediately, 264 * batch, doesn't actually triggers the hash flush immediately,
323 * you need to call flush_tlb_pending() to do that. 265 * you need to call flush_tlb_pending() to do that.
266 * Pass -1 for "normal" size (4K or 64K)
324 */ 267 */
325extern void hpte_update(struct mm_struct *mm, unsigned long addr, unsigned long pte, 268extern void hpte_update(struct mm_struct *mm, unsigned long addr,
326 int wrprot); 269 pte_t *ptep, unsigned long pte, int huge);
327 270
328static inline int __ptep_test_and_clear_young(struct mm_struct *mm, unsigned long addr, pte_t *ptep) 271static inline int __ptep_test_and_clear_young(struct mm_struct *mm,
272 unsigned long addr, pte_t *ptep)
329{ 273{
330 unsigned long old; 274 unsigned long old;
331 275
@@ -333,7 +277,7 @@ static inline int __ptep_test_and_clear_young(struct mm_struct *mm, unsigned lon
333 return 0; 277 return 0;
334 old = pte_update(ptep, _PAGE_ACCESSED); 278 old = pte_update(ptep, _PAGE_ACCESSED);
335 if (old & _PAGE_HASHPTE) { 279 if (old & _PAGE_HASHPTE) {
336 hpte_update(mm, addr, old, 0); 280 hpte_update(mm, addr, ptep, old, 0);
337 flush_tlb_pending(); 281 flush_tlb_pending();
338 } 282 }
339 return (old & _PAGE_ACCESSED) != 0; 283 return (old & _PAGE_ACCESSED) != 0;
@@ -351,7 +295,8 @@ static inline int __ptep_test_and_clear_young(struct mm_struct *mm, unsigned lon
351 * moment we always flush but we need to fix hpte_update and test if the 295 * moment we always flush but we need to fix hpte_update and test if the
352 * optimisation is worth it. 296 * optimisation is worth it.
353 */ 297 */
354static inline int __ptep_test_and_clear_dirty(struct mm_struct *mm, unsigned long addr, pte_t *ptep) 298static inline int __ptep_test_and_clear_dirty(struct mm_struct *mm,
299 unsigned long addr, pte_t *ptep)
355{ 300{
356 unsigned long old; 301 unsigned long old;
357 302
@@ -359,7 +304,7 @@ static inline int __ptep_test_and_clear_dirty(struct mm_struct *mm, unsigned lon
359 return 0; 304 return 0;
360 old = pte_update(ptep, _PAGE_DIRTY); 305 old = pte_update(ptep, _PAGE_DIRTY);
361 if (old & _PAGE_HASHPTE) 306 if (old & _PAGE_HASHPTE)
362 hpte_update(mm, addr, old, 0); 307 hpte_update(mm, addr, ptep, old, 0);
363 return (old & _PAGE_DIRTY) != 0; 308 return (old & _PAGE_DIRTY) != 0;
364} 309}
365#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY 310#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
@@ -371,7 +316,8 @@ static inline int __ptep_test_and_clear_dirty(struct mm_struct *mm, unsigned lon
371}) 316})
372 317
373#define __HAVE_ARCH_PTEP_SET_WRPROTECT 318#define __HAVE_ARCH_PTEP_SET_WRPROTECT
374static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep) 319static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
320 pte_t *ptep)
375{ 321{
376 unsigned long old; 322 unsigned long old;
377 323
@@ -379,7 +325,7 @@ static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
379 return; 325 return;
380 old = pte_update(ptep, _PAGE_RW); 326 old = pte_update(ptep, _PAGE_RW);
381 if (old & _PAGE_HASHPTE) 327 if (old & _PAGE_HASHPTE)
382 hpte_update(mm, addr, old, 0); 328 hpte_update(mm, addr, ptep, old, 0);
383} 329}
384 330
385/* 331/*
@@ -408,21 +354,23 @@ static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
408}) 354})
409 355
410#define __HAVE_ARCH_PTEP_GET_AND_CLEAR 356#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
411static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep) 357static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
358 unsigned long addr, pte_t *ptep)
412{ 359{
413 unsigned long old = pte_update(ptep, ~0UL); 360 unsigned long old = pte_update(ptep, ~0UL);
414 361
415 if (old & _PAGE_HASHPTE) 362 if (old & _PAGE_HASHPTE)
416 hpte_update(mm, addr, old, 0); 363 hpte_update(mm, addr, ptep, old, 0);
417 return __pte(old); 364 return __pte(old);
418} 365}
419 366
420static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t * ptep) 367static inline void pte_clear(struct mm_struct *mm, unsigned long addr,
368 pte_t * ptep)
421{ 369{
422 unsigned long old = pte_update(ptep, ~0UL); 370 unsigned long old = pte_update(ptep, ~0UL);
423 371
424 if (old & _PAGE_HASHPTE) 372 if (old & _PAGE_HASHPTE)
425 hpte_update(mm, addr, old, 0); 373 hpte_update(mm, addr, ptep, old, 0);
426} 374}
427 375
428/* 376/*
@@ -435,7 +383,14 @@ static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
435 pte_clear(mm, addr, ptep); 383 pte_clear(mm, addr, ptep);
436 flush_tlb_pending(); 384 flush_tlb_pending();
437 } 385 }
438 *ptep = __pte(pte_val(pte) & ~_PAGE_HPTEFLAGS); 386 pte = __pte(pte_val(pte) & ~_PAGE_HPTEFLAGS);
387
388#ifdef CONFIG_PPC_64K_PAGES
389 if (mmu_virtual_psize != MMU_PAGE_64K)
390 pte = __pte(pte_val(pte) | _PAGE_COMBO);
391#endif /* CONFIG_PPC_64K_PAGES */
392
393 *ptep = pte;
439} 394}
440 395
441/* Set the dirty and/or accessed bits atomically in a linux PTE, this 396/* Set the dirty and/or accessed bits atomically in a linux PTE, this
@@ -482,8 +437,6 @@ extern pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
482 printk("%s:%d: bad pte %08lx.\n", __FILE__, __LINE__, pte_val(e)) 437 printk("%s:%d: bad pte %08lx.\n", __FILE__, __LINE__, pte_val(e))
483#define pmd_ERROR(e) \ 438#define pmd_ERROR(e) \
484 printk("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pmd_val(e)) 439 printk("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pmd_val(e))
485#define pud_ERROR(e) \
486 printk("%s:%d: bad pud %08lx.\n", __FILE__, __LINE__, pud_val(e))
487#define pgd_ERROR(e) \ 440#define pgd_ERROR(e) \
488 printk("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e)) 441 printk("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e))
489 442
@@ -509,12 +462,12 @@ extern void update_mmu_cache(struct vm_area_struct *, unsigned long, pte_t);
509/* Encode and de-code a swap entry */ 462/* Encode and de-code a swap entry */
510#define __swp_type(entry) (((entry).val >> 1) & 0x3f) 463#define __swp_type(entry) (((entry).val >> 1) & 0x3f)
511#define __swp_offset(entry) ((entry).val >> 8) 464#define __swp_offset(entry) ((entry).val >> 8)
512#define __swp_entry(type, offset) ((swp_entry_t) { ((type) << 1) | ((offset) << 8) }) 465#define __swp_entry(type, offset) ((swp_entry_t){((type)<< 1)|((offset)<<8)})
513#define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) >> PTE_SHIFT }) 466#define __pte_to_swp_entry(pte) ((swp_entry_t){pte_val(pte) >> PTE_RPN_SHIFT})
514#define __swp_entry_to_pte(x) ((pte_t) { (x).val << PTE_SHIFT }) 467#define __swp_entry_to_pte(x) ((pte_t) { (x).val << PTE_RPN_SHIFT })
515#define pte_to_pgoff(pte) (pte_val(pte) >> PTE_SHIFT) 468#define pte_to_pgoff(pte) (pte_val(pte) >> PTE_RPN_SHIFT)
516#define pgoff_to_pte(off) ((pte_t) {((off) << PTE_SHIFT)|_PAGE_FILE}) 469#define pgoff_to_pte(off) ((pte_t) {((off) << PTE_RPN_SHIFT)|_PAGE_FILE})
517#define PTE_FILE_MAX_BITS (BITS_PER_LONG - PTE_SHIFT) 470#define PTE_FILE_MAX_BITS (BITS_PER_LONG - PTE_RPN_SHIFT)
518 471
519/* 472/*
520 * kern_addr_valid is intended to indicate whether an address is a valid 473 * kern_addr_valid is intended to indicate whether an address is a valid
@@ -532,29 +485,22 @@ void pgtable_cache_init(void);
532/* 485/*
533 * find_linux_pte returns the address of a linux pte for a given 486 * find_linux_pte returns the address of a linux pte for a given
534 * effective address and directory. If not found, it returns zero. 487 * effective address and directory. If not found, it returns zero.
535 */ 488 */static inline pte_t *find_linux_pte(pgd_t *pgdir, unsigned long ea)
536static inline pte_t *find_linux_pte(pgd_t *pgdir, unsigned long ea)
537{ 489{
538 pgd_t *pg; 490 pgd_t *pg;
539 pud_t *pu; 491 pud_t *pu;
540 pmd_t *pm; 492 pmd_t *pm;
541 pte_t *pt = NULL; 493 pte_t *pt = NULL;
542 pte_t pte;
543 494
544 pg = pgdir + pgd_index(ea); 495 pg = pgdir + pgd_index(ea);
545 if (!pgd_none(*pg)) { 496 if (!pgd_none(*pg)) {
546 pu = pud_offset(pg, ea); 497 pu = pud_offset(pg, ea);
547 if (!pud_none(*pu)) { 498 if (!pud_none(*pu)) {
548 pm = pmd_offset(pu, ea); 499 pm = pmd_offset(pu, ea);
549 if (pmd_present(*pm)) { 500 if (pmd_present(*pm))
550 pt = pte_offset_kernel(pm, ea); 501 pt = pte_offset_kernel(pm, ea);
551 pte = *pt;
552 if (!pte_present(pte))
553 pt = NULL;
554 }
555 } 502 }
556 } 503 }
557
558 return pt; 504 return pt;
559} 505}
560 506
diff --git a/include/asm-ppc64/prom.h b/include/asm-ppc64/prom.h
index e8d0d2ab4c0f..bdb47174ff0e 100644
--- a/include/asm-ppc64/prom.h
+++ b/include/asm-ppc64/prom.h
@@ -188,6 +188,14 @@ extern struct device_node *of_get_next_child(const struct device_node *node,
188extern struct device_node *of_node_get(struct device_node *node); 188extern struct device_node *of_node_get(struct device_node *node);
189extern void of_node_put(struct device_node *node); 189extern void of_node_put(struct device_node *node);
190 190
191/* For scanning the flat device-tree at boot time */
192int __init of_scan_flat_dt(int (*it)(unsigned long node,
193 const char *uname, int depth,
194 void *data),
195 void *data);
196void* __init of_get_flat_dt_prop(unsigned long node, const char *name,
197 unsigned long *size);
198
191/* For updating the device tree at runtime */ 199/* For updating the device tree at runtime */
192extern void of_attach_node(struct device_node *); 200extern void of_attach_node(struct device_node *);
193extern void of_detach_node(const struct device_node *); 201extern void of_detach_node(const struct device_node *);
diff --git a/include/asm-ppc64/system.h b/include/asm-ppc64/system.h
index 99b8ca52f101..0cdd66c9f4b7 100644
--- a/include/asm-ppc64/system.h
+++ b/include/asm-ppc64/system.h
@@ -248,7 +248,7 @@ __cmpxchg_u32(volatile unsigned int *p, unsigned long old, unsigned long new)
248} 248}
249 249
250static __inline__ unsigned long 250static __inline__ unsigned long
251__cmpxchg_u64(volatile long *p, unsigned long old, unsigned long new) 251__cmpxchg_u64(volatile unsigned long *p, unsigned long old, unsigned long new)
252{ 252{
253 unsigned long prev; 253 unsigned long prev;
254 254