1/* checksum.S: Sparc optimized checksum code. 2 * 3 * Copyright(C) 1995 Linus Torvalds 4 * Copyright(C) 1995 Miguel de Icaza 5 * Copyright(C) 1996 David S. Miller 6 * Copyright(C) 1997 Jakub Jelinek 7 * 8 * derived from: 9 * Linux/Alpha checksum c-code 10 * Linux/ix86 inline checksum assembly 11 * RFC1071 Computing the Internet Checksum (esp. Jacobsons m68k code) 12 * David Mosberger-Tang for optimized reference c-code 13 * BSD4.4 portable checksum routine 14 */ 15 16#include <asm/errno.h> 17 18#define CSUM_BIGCHUNK(buf, offset, sum, t0, t1, t2, t3, t4, t5) \ 19 ldd [buf + offset + 0x00], t0; \ 20 ldd [buf + offset + 0x08], t2; \ 21 addxcc t0, sum, sum; \ 22 addxcc t1, sum, sum; \ 23 ldd [buf + offset + 0x10], t4; \ 24 addxcc t2, sum, sum; \ 25 addxcc t3, sum, sum; \ 26 ldd [buf + offset + 0x18], t0; \ 27 addxcc t4, sum, sum; \ 28 addxcc t5, sum, sum; \ 29 addxcc t0, sum, sum; \ 30 addxcc t1, sum, sum; 31 32#define CSUM_LASTCHUNK(buf, offset, sum, t0, t1, t2, t3) \ 33 ldd [buf - offset - 0x08], t0; \ 34 ldd [buf - offset - 0x00], t2; \ 35 addxcc t0, sum, sum; \ 36 addxcc t1, sum, sum; \ 37 addxcc t2, sum, sum; \ 38 addxcc t3, sum, sum; 39 40 /* Do end cruft out of band to get better cache patterns. */ 41csum_partial_end_cruft: 42 be 1f ! caller asks %o1 & 0x8 43 andcc %o1, 4, %g0 ! nope, check for word remaining 44 ldd [%o0], %g2 ! load two 45 addcc %g2, %o2, %o2 ! add first word to sum 46 addxcc %g3, %o2, %o2 ! add second word as well 47 add %o0, 8, %o0 ! advance buf ptr 48 addx %g0, %o2, %o2 ! add in final carry 49 andcc %o1, 4, %g0 ! check again for word remaining 501: be 1f ! nope, skip this code 51 andcc %o1, 3, %o1 ! check for trailing bytes 52 ld [%o0], %g2 ! load it 53 addcc %g2, %o2, %o2 ! add to sum 54 add %o0, 4, %o0 ! advance buf ptr 55 addx %g0, %o2, %o2 ! add in final carry 56 andcc %o1, 3, %g0 ! check again for trailing bytes 571: be 1f ! no trailing bytes, return 58 addcc %o1, -1, %g0 ! only one byte remains? 59 bne 2f ! at least two bytes more 60 subcc %o1, 2, %o1 ! only two bytes more? 61 b 4f ! only one byte remains 62 or %g0, %g0, %o4 ! clear fake hword value 632: lduh [%o0], %o4 ! get hword 64 be 6f ! jmp if only hword remains 65 add %o0, 2, %o0 ! advance buf ptr either way 66 sll %o4, 16, %o4 ! create upper hword 674: ldub [%o0], %o5 ! get final byte 68 sll %o5, 8, %o5 ! put into place 69 or %o5, %o4, %o4 ! coalese with hword (if any) 706: addcc %o4, %o2, %o2 ! add to sum 711: retl ! get outta here 72 addx %g0, %o2, %o0 ! add final carry into retval 73 74 /* Also do alignment out of band to get better cache patterns. */ 75csum_partial_fix_alignment: 76 cmp %o1, 6 77 bl cpte - 0x4 78 andcc %o0, 0x2, %g0 79 be 1f 80 andcc %o0, 0x4, %g0 81 lduh [%o0 + 0x00], %g2 82 sub %o1, 2, %o1 83 add %o0, 2, %o0 84 sll %g2, 16, %g2 85 addcc %g2, %o2, %o2 86 srl %o2, 16, %g3 87 addx %g0, %g3, %g2 88 sll %o2, 16, %o2 89 sll %g2, 16, %g3 90 srl %o2, 16, %o2 91 andcc %o0, 0x4, %g0 92 or %g3, %o2, %o2 931: be cpa 94 andcc %o1, 0xffffff80, %o3 95 ld [%o0 + 0x00], %g2 96 sub %o1, 4, %o1 97 addcc %g2, %o2, %o2 98 add %o0, 4, %o0 99 addx %g0, %o2, %o2 100 b cpa 101 andcc %o1, 0xffffff80, %o3 102 103 /* The common case is to get called with a nicely aligned 104 * buffer of size 0x20. Follow the code path for that case. 105 */ 106 .globl csum_partial 107csum_partial: /* %o0=buf, %o1=len, %o2=sum */ 108 andcc %o0, 0x7, %g0 ! alignment problems? 109 bne csum_partial_fix_alignment ! yep, handle it 110 sethi %hi(cpte - 8), %g7 ! prepare table jmp ptr 111 andcc %o1, 0xffffff80, %o3 ! num loop iterations 112cpa: be 3f ! none to do 113 andcc %o1, 0x70, %g1 ! clears carry flag too 1145: CSUM_BIGCHUNK(%o0, 0x00, %o2, %o4, %o5, %g2, %g3, %g4, %g5) 115 CSUM_BIGCHUNK(%o0, 0x20, %o2, %o4, %o5, %g2, %g3, %g4, %g5) 116 CSUM_BIGCHUNK(%o0, 0x40, %o2, %o4, %o5, %g2, %g3, %g4, %g5) 117 CSUM_BIGCHUNK(%o0, 0x60, %o2, %o4, %o5, %g2, %g3, %g4, %g5) 118 addx %g0, %o2, %o2 ! sink in final carry 119 subcc %o3, 128, %o3 ! detract from loop iters 120 bne 5b ! more to do 121 add %o0, 128, %o0 ! advance buf ptr 122 andcc %o1, 0x70, %g1 ! clears carry flag too 1233: be cpte ! nope 124 andcc %o1, 0xf, %g0 ! anything left at all? 125 srl %g1, 1, %o4 ! compute offset 126 sub %g7, %g1, %g7 ! adjust jmp ptr 127 sub %g7, %o4, %g7 ! final jmp ptr adjust 128 jmp %g7 + %lo(cpte - 8) ! enter the table 129 add %o0, %g1, %o0 ! advance buf ptr 130cptbl: CSUM_LASTCHUNK(%o0, 0x68, %o2, %g2, %g3, %g4, %g5) 131 CSUM_LASTCHUNK(%o0, 0x58, %o2, %g2, %g3, %g4, %g5) 132 CSUM_LASTCHUNK(%o0, 0x48, %o2, %g2, %g3, %g4, %g5) 133 CSUM_LASTCHUNK(%o0, 0x38, %o2, %g2, %g3, %g4, %g5) 134 CSUM_LASTCHUNK(%o0, 0x28, %o2, %g2, %g3, %g4, %g5) 135 CSUM_LASTCHUNK(%o0, 0x18, %o2, %g2, %g3, %g4, %g5) 136 CSUM_LASTCHUNK(%o0, 0x08, %o2, %g2, %g3, %g4, %g5) 137 addx %g0, %o2, %o2 ! fetch final carry 138 andcc %o1, 0xf, %g0 ! anything left at all? 139cpte: bne csum_partial_end_cruft ! yep, handle it 140 andcc %o1, 8, %g0 ! check how much 141cpout: retl ! get outta here 142 mov %o2, %o0 ! return computed csum 143 144 .globl __csum_partial_copy_start, __csum_partial_copy_end 145__csum_partial_copy_start: 146 147#define ALLOC #alloc 148#define EXECINSTR #execinstr 149#define EX(x,y,a,b) \ 15098: x,y; \ 151 .section .fixup,ALLOC,EXECINSTR; \ 152 .align 4; \ 15399: ba 30f; \ 154 a, b, %o3; \ 155 .section __ex_table,ALLOC; \ 156 .align 4; \ 157 .word 98b, 99b; \ 158 .text; \ 159 .align 4 160 161#define EX2(x,y) \ 16298: x,y; \ 163 .section __ex_table,ALLOC; \ 164 .align 4; \ 165 .word 98b, 30f; \ 166 .text; \ 167 .align 4 168 169#define EX3(x,y) \ 17098: x,y; \ 171 .section __ex_table,ALLOC; \ 172 .align 4; \ 173 .word 98b, 96f; \ 174 .text; \ 175 .align 4 176 177#define EXT(start,end,handler) \ 178 .section __ex_table,ALLOC; \ 179 .align 4; \ 180 .word start, 0, end, handler; \ 181 .text; \ 182 .align 4 183 184 /* This aligned version executes typically in 8.5 superscalar cycles, this 185 * is the best I can do. I say 8.5 because the final add will pair with 186 * the next ldd in the main unrolled loop. Thus the pipe is always full. 187 * If you change these macros (including order of instructions), 188 * please check the fixup code below as well. 189 */ 190#define CSUMCOPY_BIGCHUNK_ALIGNED(src, dst, sum, off, t0, t1, t2, t3, t4, t5, t6, t7) \ 191 ldd [src + off + 0x00], t0; \ 192 ldd [src + off + 0x08], t2; \ 193 addxcc t0, sum, sum; \ 194 ldd [src + off + 0x10], t4; \ 195 addxcc t1, sum, sum; \ 196 ldd [src + off + 0x18], t6; \ 197 addxcc t2, sum, sum; \ 198 std t0, [dst + off + 0x00]; \ 199 addxcc t3, sum, sum; \ 200 std t2, [dst + off + 0x08]; \ 201 addxcc t4, sum, sum; \ 202 std t4, [dst + off + 0x10]; \ 203 addxcc t5, sum, sum; \ 204 std t6, [dst + off + 0x18]; \ 205 addxcc t6, sum, sum; \ 206 addxcc t7, sum, sum; 207 208 /* 12 superscalar cycles seems to be the limit for this case, 209 * because of this we thus do all the ldd's together to get 210 * Viking MXCC into streaming mode. Ho hum... 211 */ 212#define CSUMCOPY_BIGCHUNK(src, dst, sum, off, t0, t1, t2, t3, t4, t5, t6, t7) \ 213 ldd [src + off + 0x00], t0; \ 214 ldd [src + off + 0x08], t2; \ 215 ldd [src + off + 0x10], t4; \ 216 ldd [src + off + 0x18], t6; \ 217 st t0, [dst + off + 0x00]; \ 218 addxcc t0, sum, sum; \ 219 st t1, [dst + off + 0x04]; \ 220 addxcc t1, sum, sum; \ 221 st t2, [dst + off + 0x08]; \ 222 addxcc t2, sum, sum; \ 223 st t3, [dst + off + 0x0c]; \ 224 addxcc t3, sum, sum; \ 225 st t4, [dst + off + 0x10]; \ 226 addxcc t4, sum, sum; \ 227 st t5, [dst + off + 0x14]; \ 228 addxcc t5, sum, sum; \ 229 st t6, [dst + off + 0x18]; \ 230 addxcc t6, sum, sum; \ 231 st t7, [dst + off + 0x1c]; \ 232 addxcc t7, sum, sum; 233 234 /* Yuck, 6 superscalar cycles... */ 235#define CSUMCOPY_LASTCHUNK(src, dst, sum, off, t0, t1, t2, t3) \ 236 ldd [src - off - 0x08], t0; \ 237 ldd [src - off - 0x00], t2; \ 238 addxcc t0, sum, sum; \ 239 st t0, [dst - off - 0x08]; \ 240 addxcc t1, sum, sum; \ 241 st t1, [dst - off - 0x04]; \ 242 addxcc t2, sum, sum; \ 243 st t2, [dst - off - 0x00]; \ 244 addxcc t3, sum, sum; \ 245 st t3, [dst - off + 0x04]; 246 247 /* Handle the end cruft code out of band for better cache patterns. */ 248cc_end_cruft: 249 be 1f 250 andcc %o3, 4, %g0 251 EX(ldd [%o0 + 0x00], %g2, and %o3, 0xf) 252 add %o1, 8, %o1 253 addcc %g2, %g7, %g7 254 add %o0, 8, %o0 255 addxcc %g3, %g7, %g7 256 EX2(st %g2, [%o1 - 0x08]) 257 addx %g0, %g7, %g7 258 andcc %o3, 4, %g0 259 EX2(st %g3, [%o1 - 0x04]) 2601: be 1f 261 andcc %o3, 3, %o3 262 EX(ld [%o0 + 0x00], %g2, add %o3, 4) 263 add %o1, 4, %o1 264 addcc %g2, %g7, %g7 265 EX2(st %g2, [%o1 - 0x04]) 266 addx %g0, %g7, %g7 267 andcc %o3, 3, %g0 268 add %o0, 4, %o0 2691: be 1f 270 addcc %o3, -1, %g0 271 bne 2f 272 subcc %o3, 2, %o3 273 b 4f 274 or %g0, %g0, %o4 2752: EX(lduh [%o0 + 0x00], %o4, add %o3, 2) 276 add %o0, 2, %o0 277 EX2(sth %o4, [%o1 + 0x00]) 278 be 6f 279 add %o1, 2, %o1 280 sll %o4, 16, %o4 2814: EX(ldub [%o0 + 0x00], %o5, add %g0, 1) 282 EX2(stb %o5, [%o1 + 0x00]) 283 sll %o5, 8, %o5 284 or %o5, %o4, %o4 2856: addcc %o4, %g7, %g7 2861: retl 287 addx %g0, %g7, %o0 288 289 /* Also, handle the alignment code out of band. */ 290cc_dword_align: 291 cmp %g1, 6 292 bl,a ccte 293 andcc %g1, 0xf, %o3 294 andcc %o0, 0x1, %g0 295 bne ccslow 296 andcc %o0, 0x2, %g0 297 be 1f 298 andcc %o0, 0x4, %g0 299 EX(lduh [%o0 + 0x00], %g4, add %g1, 0) 300 sub %g1, 2, %g1 301 EX2(sth %g4, [%o1 + 0x00]) 302 add %o0, 2, %o0 303 sll %g4, 16, %g4 304 addcc %g4, %g7, %g7 305 add %o1, 2, %o1 306 srl %g7, 16, %g3 307 addx %g0, %g3, %g4 308 sll %g7, 16, %g7 309 sll %g4, 16, %g3 310 srl %g7, 16, %g7 311 andcc %o0, 0x4, %g0 312 or %g3, %g7, %g7 3131: be 3f 314 andcc %g1, 0xffffff80, %g0 315 EX(ld [%o0 + 0x00], %g4, add %g1, 0) 316 sub %g1, 4, %g1 317 EX2(st %g4, [%o1 + 0x00]) 318 add %o0, 4, %o0 319 addcc %g4, %g7, %g7 320 add %o1, 4, %o1 321 addx %g0, %g7, %g7 322 b 3f 323 andcc %g1, 0xffffff80, %g0 324 325 /* Sun, you just can't beat me, you just can't. Stop trying, 326 * give up. I'm serious, I am going to kick the living shit 327 * out of you, game over, lights out. 328 */ 329 .align 8 330 .globl __csum_partial_copy_sparc_generic 331__csum_partial_copy_sparc_generic: 332 /* %o0=src, %o1=dest, %g1=len, %g7=sum */ 333 xor %o0, %o1, %o4 ! get changing bits 334 andcc %o4, 3, %g0 ! check for mismatched alignment 335 bne ccslow ! better this than unaligned/fixups 336 andcc %o0, 7, %g0 ! need to align things? 337 bne cc_dword_align ! yes, we check for short lengths there 338 andcc %g1, 0xffffff80, %g0 ! can we use unrolled loop? 3393: be 3f ! nope, less than one loop remains 340 andcc %o1, 4, %g0 ! dest aligned on 4 or 8 byte boundary? 341 be ccdbl + 4 ! 8 byte aligned, kick ass 3425: CSUMCOPY_BIGCHUNK(%o0,%o1,%g7,0x00,%o4,%o5,%g2,%g3,%g4,%g5,%o2,%o3) 343 CSUMCOPY_BIGCHUNK(%o0,%o1,%g7,0x20,%o4,%o5,%g2,%g3,%g4,%g5,%o2,%o3) 344 CSUMCOPY_BIGCHUNK(%o0,%o1,%g7,0x40,%o4,%o5,%g2,%g3,%g4,%g5,%o2,%o3) 345 CSUMCOPY_BIGCHUNK(%o0,%o1,%g7,0x60,%o4,%o5,%g2,%g3,%g4,%g5,%o2,%o3) 34610: EXT(5b, 10b, 20f) ! note for exception handling 347 sub %g1, 128, %g1 ! detract from length 348 addx %g0, %g7, %g7 ! add in last carry bit 349 andcc %g1, 0xffffff80, %g0 ! more to csum? 350 add %o0, 128, %o0 ! advance src ptr 351 bne 5b ! we did not go negative, continue looping 352 add %o1, 128, %o1 ! advance dest ptr 3533: andcc %g1, 0x70, %o2 ! can use table? 354ccmerge:be ccte ! nope, go and check for end cruft 355 andcc %g1, 0xf, %o3 ! get low bits of length (clears carry btw) 356 srl %o2, 1, %o4 ! begin negative offset computation 357 sethi %hi(12f), %o5 ! set up table ptr end 358 add %o0, %o2, %o0 ! advance src ptr 359 sub %o5, %o4, %o5 ! continue table calculation 360 sll %o2, 1, %g2 ! constant multiplies are fun... 361 sub %o5, %g2, %o5 ! some more adjustments 362 jmp %o5 + %lo(12f) ! jump into it, duff style, wheee... 363 add %o1, %o2, %o1 ! advance dest ptr (carry is clear btw) 364cctbl: CSUMCOPY_LASTCHUNK(%o0,%o1,%g7,0x68,%g2,%g3,%g4,%g5) 365 CSUMCOPY_LASTCHUNK(%o0,%o1,%g7,0x58,%g2,%g3,%g4,%g5) 366 CSUMCOPY_LASTCHUNK(%o0,%o1,%g7,0x48,%g2,%g3,%g4,%g5) 367 CSUMCOPY_LASTCHUNK(%o0,%o1,%g7,0x38,%g2,%g3,%g4,%g5) 368 CSUMCOPY_LASTCHUNK(%o0,%o1,%g7,0x28,%g2,%g3,%g4,%g5) 369 CSUMCOPY_LASTCHUNK(%o0,%o1,%g7,0x18,%g2,%g3,%g4,%g5) 370 CSUMCOPY_LASTCHUNK(%o0,%o1,%g7,0x08,%g2,%g3,%g4,%g5) 37112: EXT(cctbl, 12b, 22f) ! note for exception table handling 372 addx %g0, %g7, %g7 373 andcc %o3, 0xf, %g0 ! check for low bits set 374ccte: bne cc_end_cruft ! something left, handle it out of band 375 andcc %o3, 8, %g0 ! begin checks for that code 376 retl ! return 377 mov %g7, %o0 ! give em the computed checksum 378ccdbl: CSUMCOPY_BIGCHUNK_ALIGNED(%o0,%o1,%g7,0x00,%o4,%o5,%g2,%g3,%g4,%g5,%o2,%o3) 379 CSUMCOPY_BIGCHUNK_ALIGNED(%o0,%o1,%g7,0x20,%o4,%o5,%g2,%g3,%g4,%g5,%o2,%o3) 380 CSUMCOPY_BIGCHUNK_ALIGNED(%o0,%o1,%g7,0x40,%o4,%o5,%g2,%g3,%g4,%g5,%o2,%o3) 381 CSUMCOPY_BIGCHUNK_ALIGNED(%o0,%o1,%g7,0x60,%o4,%o5,%g2,%g3,%g4,%g5,%o2,%o3) 38211: EXT(ccdbl, 11b, 21f) ! note for exception table handling 383 sub %g1, 128, %g1 ! detract from length 384 addx %g0, %g7, %g7 ! add in last carry bit 385 andcc %g1, 0xffffff80, %g0 ! more to csum? 386 add %o0, 128, %o0 ! advance src ptr 387 bne ccdbl ! we did not go negative, continue looping 388 add %o1, 128, %o1 ! advance dest ptr 389 b ccmerge ! finish it off, above 390 andcc %g1, 0x70, %o2 ! can use table? (clears carry btw) 391 392ccslow: cmp %g1, 0 393 mov 0, %g5 394 bleu 4f 395 andcc %o0, 1, %o5 396 be,a 1f 397 srl %g1, 1, %g4 398 sub %g1, 1, %g1 399 EX(ldub [%o0], %g5, add %g1, 1) 400 add %o0, 1, %o0 401 EX2(stb %g5, [%o1]) 402 srl %g1, 1, %g4 403 add %o1, 1, %o1 4041: cmp %g4, 0 405 be,a 3f 406 andcc %g1, 1, %g0 407 andcc %o0, 2, %g0 408 be,a 1f 409 srl %g4, 1, %g4 410 EX(lduh [%o0], %o4, add %g1, 0) 411 sub %g1, 2, %g1 412 srl %o4, 8, %g2 413 sub %g4, 1, %g4 414 EX2(stb %g2, [%o1]) 415 add %o4, %g5, %g5 416 EX2(stb %o4, [%o1 + 1]) 417 add %o0, 2, %o0 418 srl %g4, 1, %g4 419 add %o1, 2, %o1 4201: cmp %g4, 0 421 be,a 2f 422 andcc %g1, 2, %g0 423 EX3(ld [%o0], %o4) 4245: srl %o4, 24, %g2 425 srl %o4, 16, %g3 426 EX2(stb %g2, [%o1]) 427 srl %o4, 8, %g2 428 EX2(stb %g3, [%o1 + 1]) 429 add %o0, 4, %o0 430 EX2(stb %g2, [%o1 + 2]) 431 addcc %o4, %g5, %g5 432 EX2(stb %o4, [%o1 + 3]) 433 addx %g5, %g0, %g5 ! I am now to lazy to optimize this (question it 434 add %o1, 4, %o1 ! is worthy). Maybe some day - with the sll/srl 435 subcc %g4, 1, %g4 ! tricks 436 bne,a 5b 437 EX3(ld [%o0], %o4) 438 sll %g5, 16, %g2 439 srl %g5, 16, %g5 440 srl %g2, 16, %g2 441 andcc %g1, 2, %g0 442 add %g2, %g5, %g5 4432: be,a 3f 444 andcc %g1, 1, %g0 445 EX(lduh [%o0], %o4, and %g1, 3) 446 andcc %g1, 1, %g0 447 srl %o4, 8, %g2 448 add %o0, 2, %o0 449 EX2(stb %g2, [%o1]) 450 add %g5, %o4, %g5 451 EX2(stb %o4, [%o1 + 1]) 452 add %o1, 2, %o1 4533: be,a 1f 454 sll %g5, 16, %o4 455 EX(ldub [%o0], %g2, add %g0, 1) 456 sll %g2, 8, %o4 457 EX2(stb %g2, [%o1]) 458 add %g5, %o4, %g5 459 sll %g5, 16, %o4 4601: addcc %o4, %g5, %g5 461 srl %g5, 16, %o4 462 addx %g0, %o4, %g5 463 orcc %o5, %g0, %g0 464 be 4f 465 srl %g5, 8, %o4 466 and %g5, 0xff, %g2 467 and %o4, 0xff, %o4 468 sll %g2, 8, %g2 469 or %g2, %o4, %g5 4704: addcc %g7, %g5, %g7 471 retl 472 addx %g0, %g7, %o0 473__csum_partial_copy_end: 474 475/* We do these strange calculations for the csum_*_from_user case only, ie. 476 * we only bother with faults on loads... */ 477 478/* o2 = ((g2%20)&3)*8 479 * o3 = g1 - (g2/20)*32 - o2 */ 48020: 481 cmp %g2, 20 482 blu,a 1f 483 and %g2, 3, %o2 484 sub %g1, 32, %g1 485 b 20b 486 sub %g2, 20, %g2 4871: 488 sll %o2, 3, %o2 489 b 31f 490 sub %g1, %o2, %o3 491 492/* o2 = (!(g2 & 15) ? 0 : (((g2 & 15) + 1) & ~1)*8) 493 * o3 = g1 - (g2/16)*32 - o2 */ 49421: 495 andcc %g2, 15, %o3 496 srl %g2, 4, %g2 497 be,a 1f 498 clr %o2 499 add %o3, 1, %o3 500 and %o3, 14, %o3 501 sll %o3, 3, %o2 5021: 503 sll %g2, 5, %g2 504 sub %g1, %g2, %o3 505 b 31f 506 sub %o3, %o2, %o3 507 508/* o0 += (g2/10)*16 - 0x70 509 * 01 += (g2/10)*16 - 0x70 510 * o2 = (g2 % 10) ? 8 : 0 511 * o3 += 0x70 - (g2/10)*16 - o2 */ 51222: 513 cmp %g2, 10 514 blu,a 1f 515 sub %o0, 0x70, %o0 516 add %o0, 16, %o0 517 add %o1, 16, %o1 518 sub %o3, 16, %o3 519 b 22b 520 sub %g2, 10, %g2 5211: 522 sub %o1, 0x70, %o1 523 add %o3, 0x70, %o3 524 clr %o2 525 tst %g2 526 bne,a 1f 527 mov 8, %o2 5281: 529 b 31f 530 sub %o3, %o2, %o3 53196: 532 and %g1, 3, %g1 533 sll %g4, 2, %g4 534 add %g1, %g4, %o3 53530: 536/* %o1 is dst 537 * %o3 is # bytes to zero out 538 * %o4 is faulting address 539 * %o5 is %pc where fault occurred */ 540 clr %o2 54131: 542/* %o0 is src 543 * %o1 is dst 544 * %o2 is # of bytes to copy from src to dst 545 * %o3 is # bytes to zero out 546 * %o4 is faulting address 547 * %o5 is %pc where fault occurred */ 548 save %sp, -104, %sp 549 mov %i5, %o0 550 mov %i7, %o1 551 mov %i4, %o2 552 call lookup_fault 553 mov %g7, %i4 554 cmp %o0, 2 555 bne 1f 556 add %g0, -EFAULT, %i5 557 tst %i2 558 be 2f 559 mov %i0, %o1 560 mov %i1, %o0 5615: 562 call memcpy 563 mov %i2, %o2 564 tst %o0 565 bne,a 2f 566 add %i3, %i2, %i3 567 add %i1, %i2, %i1 5682: 569 mov %i1, %o0 5706: 571 call __bzero 572 mov %i3, %o1 5731: 574 ld [%sp + 168], %o2 ! struct_ptr of parent 575 st %i5, [%o2] 576 ret 577 restore 578 579 .section __ex_table,#alloc 580 .align 4 581 .word 5b,2 582 .word 6b,2 583