1 | /* SPDX-License-Identifier: GPL-2.0 */ |
2 | /* copy_user.S: Sparc optimized copy_from_user and copy_to_user code. |
3 | * |
4 | * Copyright(C) 1995 Linus Torvalds |
5 | * Copyright(C) 1996 David S. Miller |
6 | * Copyright(C) 1996 Eddie C. Dost |
7 | * Copyright(C) 1996,1998 Jakub Jelinek |
8 | * |
9 | * derived from: |
10 | * e-mail between David and Eddie. |
11 | * |
12 | * Returns 0 if successful, otherwise count of bytes not copied yet |
13 | */ |
14 | |
15 | #include <linux/export.h> |
16 | #include <asm/ptrace.h> |
17 | #include <asm/asmmacro.h> |
18 | #include <asm/page.h> |
19 | #include <asm/thread_info.h> |
20 | |
21 | /* Work around cpp -rob */ |
22 | #define ALLOC #alloc |
23 | #define EXECINSTR #execinstr |
24 | |
25 | #define EX_ENTRY(l1, l2) \ |
26 | .section __ex_table,ALLOC; \ |
27 | .align 4; \ |
28 | .word l1, l2; \ |
29 | .text; |
30 | |
31 | #define EX(x,y,a,b) \ |
32 | 98: x,y; \ |
33 | .section .fixup,ALLOC,EXECINSTR; \ |
34 | .align 4; \ |
35 | 99: retl; \ |
36 | a, b, %o0; \ |
37 | EX_ENTRY(98b, 99b) |
38 | |
39 | #define EX2(x,y,c,d,e,a,b) \ |
40 | 98: x,y; \ |
41 | .section .fixup,ALLOC,EXECINSTR; \ |
42 | .align 4; \ |
43 | 99: c, d, e; \ |
44 | retl; \ |
45 | a, b, %o0; \ |
46 | EX_ENTRY(98b, 99b) |
47 | |
48 | #define EXO2(x,y) \ |
49 | 98: x, y; \ |
50 | EX_ENTRY(98b, 97f) |
51 | |
52 | #define LD(insn, src, offset, reg, label) \ |
53 | 98: insn [%src + (offset)], %reg; \ |
54 | .section .fixup,ALLOC,EXECINSTR; \ |
55 | 99: ba label; \ |
56 | mov offset, %g5; \ |
57 | EX_ENTRY(98b, 99b) |
58 | |
59 | #define ST(insn, dst, offset, reg, label) \ |
60 | 98: insn %reg, [%dst + (offset)]; \ |
61 | .section .fixup,ALLOC,EXECINSTR; \ |
62 | 99: ba label; \ |
63 | mov offset, %g5; \ |
64 | EX_ENTRY(98b, 99b) |
65 | |
66 | /* Both these macros have to start with exactly the same insn */ |
67 | /* left: g7 + (g1 % 128) - offset */ |
68 | #define MOVE_BIGCHUNK(src, dst, offset, t0, t1, t2, t3, t4, t5, t6, t7) \ |
69 | LD(ldd, src, offset + 0x00, t0, bigchunk_fault) \ |
70 | LD(ldd, src, offset + 0x08, t2, bigchunk_fault) \ |
71 | LD(ldd, src, offset + 0x10, t4, bigchunk_fault) \ |
72 | LD(ldd, src, offset + 0x18, t6, bigchunk_fault) \ |
73 | ST(st, dst, offset + 0x00, t0, bigchunk_fault) \ |
74 | ST(st, dst, offset + 0x04, t1, bigchunk_fault) \ |
75 | ST(st, dst, offset + 0x08, t2, bigchunk_fault) \ |
76 | ST(st, dst, offset + 0x0c, t3, bigchunk_fault) \ |
77 | ST(st, dst, offset + 0x10, t4, bigchunk_fault) \ |
78 | ST(st, dst, offset + 0x14, t5, bigchunk_fault) \ |
79 | ST(st, dst, offset + 0x18, t6, bigchunk_fault) \ |
80 | ST(st, dst, offset + 0x1c, t7, bigchunk_fault) |
81 | |
82 | /* left: g7 + (g1 % 128) - offset */ |
83 | #define MOVE_BIGALIGNCHUNK(src, dst, offset, t0, t1, t2, t3, t4, t5, t6, t7) \ |
84 | LD(ldd, src, offset + 0x00, t0, bigchunk_fault) \ |
85 | LD(ldd, src, offset + 0x08, t2, bigchunk_fault) \ |
86 | LD(ldd, src, offset + 0x10, t4, bigchunk_fault) \ |
87 | LD(ldd, src, offset + 0x18, t6, bigchunk_fault) \ |
88 | ST(std, dst, offset + 0x00, t0, bigchunk_fault) \ |
89 | ST(std, dst, offset + 0x08, t2, bigchunk_fault) \ |
90 | ST(std, dst, offset + 0x10, t4, bigchunk_fault) \ |
91 | ST(std, dst, offset + 0x18, t6, bigchunk_fault) |
92 | |
93 | .section .fixup,#alloc,#execinstr |
94 | bigchunk_fault: |
95 | sub %g7, %g5, %o0 |
96 | and %g1, 127, %g1 |
97 | retl |
98 | add %o0, %g1, %o0 |
99 | |
100 | /* left: offset + 16 + (g1 % 16) */ |
101 | #define MOVE_LASTCHUNK(src, dst, offset, t0, t1, t2, t3) \ |
102 | LD(ldd, src, -(offset + 0x10), t0, lastchunk_fault) \ |
103 | LD(ldd, src, -(offset + 0x08), t2, lastchunk_fault) \ |
104 | ST(st, dst, -(offset + 0x10), t0, lastchunk_fault) \ |
105 | ST(st, dst, -(offset + 0x0c), t1, lastchunk_fault) \ |
106 | ST(st, dst, -(offset + 0x08), t2, lastchunk_fault) \ |
107 | ST(st, dst, -(offset + 0x04), t3, lastchunk_fault) |
108 | |
109 | .section .fixup,#alloc,#execinstr |
110 | lastchunk_fault: |
111 | and %g1, 15, %g1 |
112 | retl |
113 | sub %g1, %g5, %o0 |
114 | |
115 | /* left: o3 + (o2 % 16) - offset */ |
116 | #define MOVE_HALFCHUNK(src, dst, offset, t0, t1, t2, t3) \ |
117 | LD(lduh, src, offset + 0x00, t0, halfchunk_fault) \ |
118 | LD(lduh, src, offset + 0x02, t1, halfchunk_fault) \ |
119 | LD(lduh, src, offset + 0x04, t2, halfchunk_fault) \ |
120 | LD(lduh, src, offset + 0x06, t3, halfchunk_fault) \ |
121 | ST(sth, dst, offset + 0x00, t0, halfchunk_fault) \ |
122 | ST(sth, dst, offset + 0x02, t1, halfchunk_fault) \ |
123 | ST(sth, dst, offset + 0x04, t2, halfchunk_fault) \ |
124 | ST(sth, dst, offset + 0x06, t3, halfchunk_fault) |
125 | |
126 | /* left: o3 + (o2 % 16) + offset + 2 */ |
127 | #define MOVE_SHORTCHUNK(src, dst, offset, t0, t1) \ |
128 | LD(ldub, src, -(offset + 0x02), t0, halfchunk_fault) \ |
129 | LD(ldub, src, -(offset + 0x01), t1, halfchunk_fault) \ |
130 | ST(stb, dst, -(offset + 0x02), t0, halfchunk_fault) \ |
131 | ST(stb, dst, -(offset + 0x01), t1, halfchunk_fault) |
132 | |
133 | .section .fixup,#alloc,#execinstr |
134 | halfchunk_fault: |
135 | and %o2, 15, %o2 |
136 | sub %o3, %g5, %o3 |
137 | retl |
138 | add %o2, %o3, %o0 |
139 | |
140 | /* left: offset + 2 + (o2 % 2) */ |
141 | #define MOVE_LAST_SHORTCHUNK(src, dst, offset, t0, t1) \ |
142 | LD(ldub, src, -(offset + 0x02), t0, last_shortchunk_fault) \ |
143 | LD(ldub, src, -(offset + 0x01), t1, last_shortchunk_fault) \ |
144 | ST(stb, dst, -(offset + 0x02), t0, last_shortchunk_fault) \ |
145 | ST(stb, dst, -(offset + 0x01), t1, last_shortchunk_fault) |
146 | |
147 | .section .fixup,#alloc,#execinstr |
148 | last_shortchunk_fault: |
149 | and %o2, 1, %o2 |
150 | retl |
151 | sub %o2, %g5, %o0 |
152 | |
153 | .text |
154 | .align 4 |
155 | |
156 | .globl __copy_user_begin |
157 | __copy_user_begin: |
158 | |
159 | .globl __copy_user |
160 | EXPORT_SYMBOL(__copy_user) |
161 | dword_align: |
162 | andcc %o1, 1, %g0 |
163 | be 4f |
164 | andcc %o1, 2, %g0 |
165 | |
166 | EXO2(ldub [%o1], %g2) |
167 | add %o1, 1, %o1 |
168 | EXO2(stb %g2, [%o0]) |
169 | sub %o2, 1, %o2 |
170 | bne 3f |
171 | add %o0, 1, %o0 |
172 | |
173 | EXO2(lduh [%o1], %g2) |
174 | add %o1, 2, %o1 |
175 | EXO2(sth %g2, [%o0]) |
176 | sub %o2, 2, %o2 |
177 | b 3f |
178 | add %o0, 2, %o0 |
179 | 4: |
180 | EXO2(lduh [%o1], %g2) |
181 | add %o1, 2, %o1 |
182 | EXO2(sth %g2, [%o0]) |
183 | sub %o2, 2, %o2 |
184 | b 3f |
185 | add %o0, 2, %o0 |
186 | |
187 | __copy_user: /* %o0=dst %o1=src %o2=len */ |
188 | xor %o0, %o1, %o4 |
189 | 1: |
190 | andcc %o4, 3, %o5 |
191 | 2: |
192 | bne cannot_optimize |
193 | cmp %o2, 15 |
194 | |
195 | bleu short_aligned_end |
196 | andcc %o1, 3, %g0 |
197 | |
198 | bne dword_align |
199 | 3: |
200 | andcc %o1, 4, %g0 |
201 | |
202 | be 2f |
203 | mov %o2, %g1 |
204 | |
205 | EXO2(ld [%o1], %o4) |
206 | sub %g1, 4, %g1 |
207 | EXO2(st %o4, [%o0]) |
208 | add %o1, 4, %o1 |
209 | add %o0, 4, %o0 |
210 | 2: |
211 | andcc %g1, 0xffffff80, %g7 |
212 | be 3f |
213 | andcc %o0, 4, %g0 |
214 | |
215 | be ldd_std + 4 |
216 | 5: |
217 | MOVE_BIGCHUNK(o1, o0, 0x00, o2, o3, o4, o5, g2, g3, g4, g5) |
218 | MOVE_BIGCHUNK(o1, o0, 0x20, o2, o3, o4, o5, g2, g3, g4, g5) |
219 | MOVE_BIGCHUNK(o1, o0, 0x40, o2, o3, o4, o5, g2, g3, g4, g5) |
220 | MOVE_BIGCHUNK(o1, o0, 0x60, o2, o3, o4, o5, g2, g3, g4, g5) |
221 | subcc %g7, 128, %g7 |
222 | add %o1, 128, %o1 |
223 | bne 5b |
224 | add %o0, 128, %o0 |
225 | 3: |
226 | andcc %g1, 0x70, %g7 |
227 | be copy_user_table_end |
228 | andcc %g1, 8, %g0 |
229 | |
230 | sethi %hi(copy_user_table_end), %o5 |
231 | srl %g7, 1, %o4 |
232 | add %g7, %o4, %o4 |
233 | add %o1, %g7, %o1 |
234 | sub %o5, %o4, %o5 |
235 | jmpl %o5 + %lo(copy_user_table_end), %g0 |
236 | add %o0, %g7, %o0 |
237 | |
238 | MOVE_LASTCHUNK(o1, o0, 0x60, g2, g3, g4, g5) |
239 | MOVE_LASTCHUNK(o1, o0, 0x50, g2, g3, g4, g5) |
240 | MOVE_LASTCHUNK(o1, o0, 0x40, g2, g3, g4, g5) |
241 | MOVE_LASTCHUNK(o1, o0, 0x30, g2, g3, g4, g5) |
242 | MOVE_LASTCHUNK(o1, o0, 0x20, g2, g3, g4, g5) |
243 | MOVE_LASTCHUNK(o1, o0, 0x10, g2, g3, g4, g5) |
244 | MOVE_LASTCHUNK(o1, o0, 0x00, g2, g3, g4, g5) |
245 | copy_user_table_end: |
246 | be copy_user_last7 |
247 | andcc %g1, 4, %g0 |
248 | |
249 | EX(ldd [%o1], %g2, and %g1, 0xf) |
250 | add %o0, 8, %o0 |
251 | add %o1, 8, %o1 |
252 | EX(st %g2, [%o0 - 0x08], and %g1, 0xf) |
253 | EX2(st %g3, [%o0 - 0x04], and %g1, 0xf, %g1, sub %g1, 4) |
254 | copy_user_last7: |
255 | be 1f |
256 | andcc %g1, 2, %g0 |
257 | |
258 | EX(ld [%o1], %g2, and %g1, 7) |
259 | add %o1, 4, %o1 |
260 | EX(st %g2, [%o0], and %g1, 7) |
261 | add %o0, 4, %o0 |
262 | 1: |
263 | be 1f |
264 | andcc %g1, 1, %g0 |
265 | |
266 | EX(lduh [%o1], %g2, and %g1, 3) |
267 | add %o1, 2, %o1 |
268 | EX(sth %g2, [%o0], and %g1, 3) |
269 | add %o0, 2, %o0 |
270 | 1: |
271 | be 1f |
272 | nop |
273 | |
274 | EX(ldub [%o1], %g2, add %g0, 1) |
275 | EX(stb %g2, [%o0], add %g0, 1) |
276 | 1: |
277 | retl |
278 | clr %o0 |
279 | |
280 | ldd_std: |
281 | MOVE_BIGALIGNCHUNK(o1, o0, 0x00, o2, o3, o4, o5, g2, g3, g4, g5) |
282 | MOVE_BIGALIGNCHUNK(o1, o0, 0x20, o2, o3, o4, o5, g2, g3, g4, g5) |
283 | MOVE_BIGALIGNCHUNK(o1, o0, 0x40, o2, o3, o4, o5, g2, g3, g4, g5) |
284 | MOVE_BIGALIGNCHUNK(o1, o0, 0x60, o2, o3, o4, o5, g2, g3, g4, g5) |
285 | subcc %g7, 128, %g7 |
286 | add %o1, 128, %o1 |
287 | bne ldd_std |
288 | add %o0, 128, %o0 |
289 | |
290 | andcc %g1, 0x70, %g7 |
291 | be copy_user_table_end |
292 | andcc %g1, 8, %g0 |
293 | |
294 | sethi %hi(copy_user_table_end), %o5 |
295 | srl %g7, 1, %o4 |
296 | add %g7, %o4, %o4 |
297 | add %o1, %g7, %o1 |
298 | sub %o5, %o4, %o5 |
299 | jmpl %o5 + %lo(copy_user_table_end), %g0 |
300 | add %o0, %g7, %o0 |
301 | |
302 | cannot_optimize: |
303 | bleu short_end |
304 | cmp %o5, 2 |
305 | |
306 | bne byte_chunk |
307 | and %o2, 0xfffffff0, %o3 |
308 | |
309 | andcc %o1, 1, %g0 |
310 | be 10f |
311 | nop |
312 | |
313 | EXO2(ldub [%o1], %g2) |
314 | add %o1, 1, %o1 |
315 | EXO2(stb %g2, [%o0]) |
316 | sub %o2, 1, %o2 |
317 | andcc %o2, 0xfffffff0, %o3 |
318 | be short_end |
319 | add %o0, 1, %o0 |
320 | 10: |
321 | MOVE_HALFCHUNK(o1, o0, 0x00, g2, g3, g4, g5) |
322 | MOVE_HALFCHUNK(o1, o0, 0x08, g2, g3, g4, g5) |
323 | subcc %o3, 0x10, %o3 |
324 | add %o1, 0x10, %o1 |
325 | bne 10b |
326 | add %o0, 0x10, %o0 |
327 | b 2f |
328 | and %o2, 0xe, %o3 |
329 | |
330 | byte_chunk: |
331 | MOVE_SHORTCHUNK(o1, o0, -0x02, g2, g3) |
332 | MOVE_SHORTCHUNK(o1, o0, -0x04, g2, g3) |
333 | MOVE_SHORTCHUNK(o1, o0, -0x06, g2, g3) |
334 | MOVE_SHORTCHUNK(o1, o0, -0x08, g2, g3) |
335 | MOVE_SHORTCHUNK(o1, o0, -0x0a, g2, g3) |
336 | MOVE_SHORTCHUNK(o1, o0, -0x0c, g2, g3) |
337 | MOVE_SHORTCHUNK(o1, o0, -0x0e, g2, g3) |
338 | MOVE_SHORTCHUNK(o1, o0, -0x10, g2, g3) |
339 | subcc %o3, 0x10, %o3 |
340 | add %o1, 0x10, %o1 |
341 | bne byte_chunk |
342 | add %o0, 0x10, %o0 |
343 | |
344 | short_end: |
345 | and %o2, 0xe, %o3 |
346 | 2: |
347 | sethi %hi(short_table_end), %o5 |
348 | sll %o3, 3, %o4 |
349 | add %o0, %o3, %o0 |
350 | sub %o5, %o4, %o5 |
351 | add %o1, %o3, %o1 |
352 | jmpl %o5 + %lo(short_table_end), %g0 |
353 | andcc %o2, 1, %g0 |
354 | MOVE_LAST_SHORTCHUNK(o1, o0, 0x0c, g2, g3) |
355 | MOVE_LAST_SHORTCHUNK(o1, o0, 0x0a, g2, g3) |
356 | MOVE_LAST_SHORTCHUNK(o1, o0, 0x08, g2, g3) |
357 | MOVE_LAST_SHORTCHUNK(o1, o0, 0x06, g2, g3) |
358 | MOVE_LAST_SHORTCHUNK(o1, o0, 0x04, g2, g3) |
359 | MOVE_LAST_SHORTCHUNK(o1, o0, 0x02, g2, g3) |
360 | MOVE_LAST_SHORTCHUNK(o1, o0, 0x00, g2, g3) |
361 | short_table_end: |
362 | be 1f |
363 | nop |
364 | EX(ldub [%o1], %g2, add %g0, 1) |
365 | EX(stb %g2, [%o0], add %g0, 1) |
366 | 1: |
367 | retl |
368 | clr %o0 |
369 | |
370 | short_aligned_end: |
371 | bne short_end |
372 | andcc %o2, 8, %g0 |
373 | |
374 | be 1f |
375 | andcc %o2, 4, %g0 |
376 | |
377 | EXO2(ld [%o1 + 0x00], %g2) |
378 | EXO2(ld [%o1 + 0x04], %g3) |
379 | add %o1, 8, %o1 |
380 | EXO2(st %g2, [%o0 + 0x00]) |
381 | EX(st %g3, [%o0 + 0x04], sub %o2, 4) |
382 | add %o0, 8, %o0 |
383 | 1: |
384 | b copy_user_last7 |
385 | mov %o2, %g1 |
386 | |
387 | .section .fixup,#alloc,#execinstr |
388 | .align 4 |
389 | 97: |
390 | retl |
391 | mov %o2, %o0 |
392 | |
393 | .globl __copy_user_end |
394 | __copy_user_end: |
395 | |