diff options
Diffstat (limited to 'arch/sparc64/lib/copy_page.S')
-rw-r--r-- | arch/sparc64/lib/copy_page.S | 242 |
1 files changed, 242 insertions, 0 deletions
diff --git a/arch/sparc64/lib/copy_page.S b/arch/sparc64/lib/copy_page.S new file mode 100644 index 000000000000..23ebf2c970b7 --- /dev/null +++ b/arch/sparc64/lib/copy_page.S | |||
@@ -0,0 +1,242 @@ | |||
1 | /* clear_page.S: UltraSparc optimized copy page. | ||
2 | * | ||
3 | * Copyright (C) 1996, 1998, 1999, 2000, 2004 David S. Miller (davem@redhat.com) | ||
4 | * Copyright (C) 1997 Jakub Jelinek (jakub@redhat.com) | ||
5 | */ | ||
6 | |||
7 | #include <asm/visasm.h> | ||
8 | #include <asm/thread_info.h> | ||
9 | #include <asm/page.h> | ||
10 | #include <asm/pgtable.h> | ||
11 | #include <asm/spitfire.h> | ||
12 | #include <asm/head.h> | ||
13 | |||
14 | /* What we used to do was lock a TLB entry into a specific | ||
15 | * TLB slot, clear the page with interrupts disabled, then | ||
16 | * restore the original TLB entry. This was great for | ||
17 | * disturbing the TLB as little as possible, but it meant | ||
18 | * we had to keep interrupts disabled for a long time. | ||
19 | * | ||
20 | * Now, we simply use the normal TLB loading mechanism, | ||
21 | * and this makes the cpu choose a slot all by itself. | ||
22 | * Then we do a normal TLB flush on exit. We need only | ||
23 | * disable preemption during the clear. | ||
24 | */ | ||
25 | |||
26 | #define TTE_BITS_TOP (_PAGE_VALID | _PAGE_SZBITS) | ||
27 | #define TTE_BITS_BOTTOM (_PAGE_CP | _PAGE_CV | _PAGE_P | _PAGE_L | _PAGE_W) | ||
28 | #define DCACHE_SIZE (PAGE_SIZE * 2) | ||
29 | |||
30 | #if (PAGE_SHIFT == 13) || (PAGE_SHIFT == 19) | ||
31 | #define PAGE_SIZE_REM 0x80 | ||
32 | #elif (PAGE_SHIFT == 16) || (PAGE_SHIFT == 22) | ||
33 | #define PAGE_SIZE_REM 0x100 | ||
34 | #else | ||
35 | #error Wrong PAGE_SHIFT specified | ||
36 | #endif | ||
37 | |||
38 | #define TOUCH(reg0, reg1, reg2, reg3, reg4, reg5, reg6, reg7) \ | ||
39 | fmovd %reg0, %f48; fmovd %reg1, %f50; \ | ||
40 | fmovd %reg2, %f52; fmovd %reg3, %f54; \ | ||
41 | fmovd %reg4, %f56; fmovd %reg5, %f58; \ | ||
42 | fmovd %reg6, %f60; fmovd %reg7, %f62; | ||
43 | |||
44 | .text | ||
45 | |||
46 | .align 32 | ||
47 | .globl copy_user_page | ||
48 | .type copy_user_page,#function | ||
49 | copy_user_page: /* %o0=dest, %o1=src, %o2=vaddr */ | ||
50 | lduw [%g6 + TI_PRE_COUNT], %o4 | ||
51 | sethi %uhi(PAGE_OFFSET), %g2 | ||
52 | sethi %hi(PAGE_SIZE), %o3 | ||
53 | |||
54 | sllx %g2, 32, %g2 | ||
55 | sethi %uhi(TTE_BITS_TOP), %g3 | ||
56 | |||
57 | sllx %g3, 32, %g3 | ||
58 | sub %o0, %g2, %g1 ! dest paddr | ||
59 | |||
60 | sub %o1, %g2, %g2 ! src paddr | ||
61 | or %g3, TTE_BITS_BOTTOM, %g3 | ||
62 | |||
63 | and %o2, %o3, %o0 ! vaddr D-cache alias bit | ||
64 | or %g1, %g3, %g1 ! dest TTE data | ||
65 | |||
66 | or %g2, %g3, %g2 ! src TTE data | ||
67 | sethi %hi(TLBTEMP_BASE), %o3 | ||
68 | |||
69 | sethi %hi(DCACHE_SIZE), %o1 | ||
70 | add %o0, %o3, %o0 ! dest TTE vaddr | ||
71 | |||
72 | add %o4, 1, %o2 | ||
73 | add %o0, %o1, %o1 ! src TTE vaddr | ||
74 | |||
75 | /* Disable preemption. */ | ||
76 | mov TLB_TAG_ACCESS, %g3 | ||
77 | stw %o2, [%g6 + TI_PRE_COUNT] | ||
78 | |||
79 | /* Load TLB entries. */ | ||
80 | rdpr %pstate, %o2 | ||
81 | wrpr %o2, PSTATE_IE, %pstate | ||
82 | stxa %o0, [%g3] ASI_DMMU | ||
83 | stxa %g1, [%g0] ASI_DTLB_DATA_IN | ||
84 | membar #Sync | ||
85 | stxa %o1, [%g3] ASI_DMMU | ||
86 | stxa %g2, [%g0] ASI_DTLB_DATA_IN | ||
87 | membar #Sync | ||
88 | wrpr %o2, 0x0, %pstate | ||
89 | |||
90 | BRANCH_IF_ANY_CHEETAH(g3,o2,1f) | ||
91 | ba,pt %xcc, 9f | ||
92 | nop | ||
93 | |||
94 | 1: | ||
95 | VISEntryHalf | ||
96 | membar #StoreLoad | #StoreStore | #LoadStore | ||
97 | sethi %hi((PAGE_SIZE/64)-2), %o2 | ||
98 | mov %o0, %g1 | ||
99 | prefetch [%o1 + 0x000], #one_read | ||
100 | or %o2, %lo((PAGE_SIZE/64)-2), %o2 | ||
101 | prefetch [%o1 + 0x040], #one_read | ||
102 | prefetch [%o1 + 0x080], #one_read | ||
103 | prefetch [%o1 + 0x0c0], #one_read | ||
104 | ldd [%o1 + 0x000], %f0 | ||
105 | prefetch [%o1 + 0x100], #one_read | ||
106 | ldd [%o1 + 0x008], %f2 | ||
107 | prefetch [%o1 + 0x140], #one_read | ||
108 | ldd [%o1 + 0x010], %f4 | ||
109 | prefetch [%o1 + 0x180], #one_read | ||
110 | fmovd %f0, %f16 | ||
111 | ldd [%o1 + 0x018], %f6 | ||
112 | fmovd %f2, %f18 | ||
113 | ldd [%o1 + 0x020], %f8 | ||
114 | fmovd %f4, %f20 | ||
115 | ldd [%o1 + 0x028], %f10 | ||
116 | fmovd %f6, %f22 | ||
117 | ldd [%o1 + 0x030], %f12 | ||
118 | fmovd %f8, %f24 | ||
119 | ldd [%o1 + 0x038], %f14 | ||
120 | fmovd %f10, %f26 | ||
121 | ldd [%o1 + 0x040], %f0 | ||
122 | 1: ldd [%o1 + 0x048], %f2 | ||
123 | fmovd %f12, %f28 | ||
124 | ldd [%o1 + 0x050], %f4 | ||
125 | fmovd %f14, %f30 | ||
126 | stda %f16, [%o0] ASI_BLK_P | ||
127 | ldd [%o1 + 0x058], %f6 | ||
128 | fmovd %f0, %f16 | ||
129 | ldd [%o1 + 0x060], %f8 | ||
130 | fmovd %f2, %f18 | ||
131 | ldd [%o1 + 0x068], %f10 | ||
132 | fmovd %f4, %f20 | ||
133 | ldd [%o1 + 0x070], %f12 | ||
134 | fmovd %f6, %f22 | ||
135 | ldd [%o1 + 0x078], %f14 | ||
136 | fmovd %f8, %f24 | ||
137 | ldd [%o1 + 0x080], %f0 | ||
138 | prefetch [%o1 + 0x180], #one_read | ||
139 | fmovd %f10, %f26 | ||
140 | subcc %o2, 1, %o2 | ||
141 | add %o0, 0x40, %o0 | ||
142 | bne,pt %xcc, 1b | ||
143 | add %o1, 0x40, %o1 | ||
144 | |||
145 | ldd [%o1 + 0x048], %f2 | ||
146 | fmovd %f12, %f28 | ||
147 | ldd [%o1 + 0x050], %f4 | ||
148 | fmovd %f14, %f30 | ||
149 | stda %f16, [%o0] ASI_BLK_P | ||
150 | ldd [%o1 + 0x058], %f6 | ||
151 | fmovd %f0, %f16 | ||
152 | ldd [%o1 + 0x060], %f8 | ||
153 | fmovd %f2, %f18 | ||
154 | ldd [%o1 + 0x068], %f10 | ||
155 | fmovd %f4, %f20 | ||
156 | ldd [%o1 + 0x070], %f12 | ||
157 | fmovd %f6, %f22 | ||
158 | add %o0, 0x40, %o0 | ||
159 | ldd [%o1 + 0x078], %f14 | ||
160 | fmovd %f8, %f24 | ||
161 | fmovd %f10, %f26 | ||
162 | fmovd %f12, %f28 | ||
163 | fmovd %f14, %f30 | ||
164 | stda %f16, [%o0] ASI_BLK_P | ||
165 | membar #Sync | ||
166 | VISExitHalf | ||
167 | ba,pt %xcc, 5f | ||
168 | nop | ||
169 | |||
170 | 9: | ||
171 | VISEntry | ||
172 | ldub [%g6 + TI_FAULT_CODE], %g3 | ||
173 | mov %o0, %g1 | ||
174 | cmp %g3, 0 | ||
175 | rd %asi, %g3 | ||
176 | be,a,pt %icc, 1f | ||
177 | wr %g0, ASI_BLK_P, %asi | ||
178 | wr %g0, ASI_BLK_COMMIT_P, %asi | ||
179 | 1: ldda [%o1] ASI_BLK_P, %f0 | ||
180 | add %o1, 0x40, %o1 | ||
181 | ldda [%o1] ASI_BLK_P, %f16 | ||
182 | add %o1, 0x40, %o1 | ||
183 | sethi %hi(PAGE_SIZE), %o2 | ||
184 | 1: TOUCH(f0, f2, f4, f6, f8, f10, f12, f14) | ||
185 | ldda [%o1] ASI_BLK_P, %f32 | ||
186 | stda %f48, [%o0] %asi | ||
187 | add %o1, 0x40, %o1 | ||
188 | sub %o2, 0x40, %o2 | ||
189 | add %o0, 0x40, %o0 | ||
190 | TOUCH(f16, f18, f20, f22, f24, f26, f28, f30) | ||
191 | ldda [%o1] ASI_BLK_P, %f0 | ||
192 | stda %f48, [%o0] %asi | ||
193 | add %o1, 0x40, %o1 | ||
194 | sub %o2, 0x40, %o2 | ||
195 | add %o0, 0x40, %o0 | ||
196 | TOUCH(f32, f34, f36, f38, f40, f42, f44, f46) | ||
197 | ldda [%o1] ASI_BLK_P, %f16 | ||
198 | stda %f48, [%o0] %asi | ||
199 | sub %o2, 0x40, %o2 | ||
200 | add %o1, 0x40, %o1 | ||
201 | cmp %o2, PAGE_SIZE_REM | ||
202 | bne,pt %xcc, 1b | ||
203 | add %o0, 0x40, %o0 | ||
204 | #if (PAGE_SHIFT == 16) || (PAGE_SHIFT == 22) | ||
205 | TOUCH(f0, f2, f4, f6, f8, f10, f12, f14) | ||
206 | ldda [%o1] ASI_BLK_P, %f32 | ||
207 | stda %f48, [%o0] %asi | ||
208 | add %o1, 0x40, %o1 | ||
209 | sub %o2, 0x40, %o2 | ||
210 | add %o0, 0x40, %o0 | ||
211 | TOUCH(f16, f18, f20, f22, f24, f26, f28, f30) | ||
212 | ldda [%o1] ASI_BLK_P, %f0 | ||
213 | stda %f48, [%o0] %asi | ||
214 | add %o1, 0x40, %o1 | ||
215 | sub %o2, 0x40, %o2 | ||
216 | add %o0, 0x40, %o0 | ||
217 | membar #Sync | ||
218 | stda %f32, [%o0] %asi | ||
219 | add %o0, 0x40, %o0 | ||
220 | stda %f0, [%o0] %asi | ||
221 | #else | ||
222 | membar #Sync | ||
223 | stda %f0, [%o0] %asi | ||
224 | add %o0, 0x40, %o0 | ||
225 | stda %f16, [%o0] %asi | ||
226 | #endif | ||
227 | membar #Sync | ||
228 | wr %g3, 0x0, %asi | ||
229 | VISExit | ||
230 | |||
231 | 5: | ||
232 | stxa %g0, [%g1] ASI_DMMU_DEMAP | ||
233 | membar #Sync | ||
234 | |||
235 | sethi %hi(DCACHE_SIZE), %g2 | ||
236 | stxa %g0, [%g1 + %g2] ASI_DMMU_DEMAP | ||
237 | membar #Sync | ||
238 | |||
239 | retl | ||
240 | stw %o4, [%g6 + TI_PRE_COUNT] | ||
241 | |||
242 | .size copy_user_page, .-copy_user_page | ||