Brian Silverman | 8649792 | 2018-02-10 19:28:39 -0500 | [diff] [blame] | 1 | /* Disassembler for x86. |
| 2 | Copyright (C) 2007, 2008, 2009, 2011 Red Hat, Inc. |
| 3 | This file is part of elfutils. |
| 4 | Written by Ulrich Drepper <drepper@redhat.com>, 2007. |
| 5 | |
| 6 | This file is free software; you can redistribute it and/or modify |
| 7 | it under the terms of either |
| 8 | |
| 9 | * the GNU Lesser General Public License as published by the Free |
| 10 | Software Foundation; either version 3 of the License, or (at |
| 11 | your option) any later version |
| 12 | |
| 13 | or |
| 14 | |
| 15 | * the GNU General Public License as published by the Free |
| 16 | Software Foundation; either version 2 of the License, or (at |
| 17 | your option) any later version |
| 18 | |
| 19 | or both in parallel, as here. |
| 20 | |
| 21 | elfutils is distributed in the hope that it will be useful, but |
| 22 | WITHOUT ANY WARRANTY; without even the implied warranty of |
| 23 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 24 | General Public License for more details. |
| 25 | |
| 26 | You should have received copies of the GNU General Public License and |
| 27 | the GNU Lesser General Public License along with this program. If |
| 28 | not, see <http://www.gnu.org/licenses/>. */ |
| 29 | |
| 30 | #ifdef HAVE_CONFIG_H |
| 31 | # include <config.h> |
| 32 | #endif |
| 33 | |
| 34 | #include <assert.h> |
| 35 | #include <config.h> |
| 36 | #include <ctype.h> |
| 37 | #include <endian.h> |
| 38 | #include <errno.h> |
| 39 | #include <gelf.h> |
| 40 | #include <stddef.h> |
| 41 | #include <stdint.h> |
| 42 | #include <stdlib.h> |
| 43 | #include <string.h> |
| 44 | |
| 45 | #include "../libebl/libeblP.h" |
| 46 | |
| 47 | #define MACHINE_ENCODING __LITTLE_ENDIAN |
| 48 | #include "memory-access.h" |
| 49 | |
| 50 | |
| 51 | #ifndef MNEFILE |
| 52 | # define MNEFILE "i386.mnemonics" |
| 53 | #endif |
| 54 | |
| 55 | #define MNESTRFIELD(line) MNESTRFIELD1 (line) |
| 56 | #define MNESTRFIELD1(line) str##line |
| 57 | static const union mnestr_t |
| 58 | { |
| 59 | struct |
| 60 | { |
| 61 | #define MNE(name) char MNESTRFIELD (__LINE__)[sizeof (#name)]; |
| 62 | #include MNEFILE |
| 63 | #undef MNE |
| 64 | }; |
| 65 | char str[0]; |
| 66 | } mnestr = |
| 67 | { |
| 68 | { |
| 69 | #define MNE(name) #name, |
| 70 | #include MNEFILE |
| 71 | #undef MNE |
| 72 | } |
| 73 | }; |
| 74 | |
| 75 | /* The index can be stored in the instrtab. */ |
| 76 | enum |
| 77 | { |
| 78 | #define MNE(name) MNE_##name, |
| 79 | #include MNEFILE |
| 80 | #undef MNE |
| 81 | MNE_INVALID |
| 82 | }; |
| 83 | |
| 84 | static const unsigned short int mneidx[] = |
| 85 | { |
| 86 | #define MNE(name) \ |
| 87 | [MNE_##name] = offsetof (union mnestr_t, MNESTRFIELD (__LINE__)), |
| 88 | #include MNEFILE |
| 89 | #undef MNE |
| 90 | }; |
| 91 | |
| 92 | |
| 93 | enum |
| 94 | { |
| 95 | idx_rex_b = 0, |
| 96 | idx_rex_x, |
| 97 | idx_rex_r, |
| 98 | idx_rex_w, |
| 99 | idx_rex, |
| 100 | idx_cs, |
| 101 | idx_ds, |
| 102 | idx_es, |
| 103 | idx_fs, |
| 104 | idx_gs, |
| 105 | idx_ss, |
| 106 | idx_data16, |
| 107 | idx_addr16, |
| 108 | idx_rep, |
| 109 | idx_repne, |
| 110 | idx_lock |
| 111 | }; |
| 112 | |
| 113 | enum |
| 114 | { |
| 115 | #define prefbit(pref) has_##pref = 1 << idx_##pref |
| 116 | prefbit (rex_b), |
| 117 | prefbit (rex_x), |
| 118 | prefbit (rex_r), |
| 119 | prefbit (rex_w), |
| 120 | prefbit (rex), |
| 121 | prefbit (cs), |
| 122 | prefbit (ds), |
| 123 | prefbit (es), |
| 124 | prefbit (fs), |
| 125 | prefbit (gs), |
| 126 | prefbit (ss), |
| 127 | prefbit (data16), |
| 128 | prefbit (addr16), |
| 129 | prefbit (rep), |
| 130 | prefbit (repne), |
| 131 | prefbit (lock) |
| 132 | #undef prefbit |
| 133 | }; |
| 134 | #define SEGMENT_PREFIXES \ |
| 135 | (has_cs | has_ds | has_es | has_fs | has_gs | has_ss) |
| 136 | |
| 137 | #define prefix_cs 0x2e |
| 138 | #define prefix_ds 0x3e |
| 139 | #define prefix_es 0x26 |
| 140 | #define prefix_fs 0x64 |
| 141 | #define prefix_gs 0x65 |
| 142 | #define prefix_ss 0x36 |
| 143 | #define prefix_data16 0x66 |
| 144 | #define prefix_addr16 0x67 |
| 145 | #define prefix_rep 0xf3 |
| 146 | #define prefix_repne 0xf2 |
| 147 | #define prefix_lock 0xf0 |
| 148 | |
| 149 | |
| 150 | static const uint8_t known_prefixes[] = |
| 151 | { |
| 152 | #define newpref(pref) [idx_##pref] = prefix_##pref |
| 153 | newpref (cs), |
| 154 | newpref (ds), |
| 155 | newpref (es), |
| 156 | newpref (fs), |
| 157 | newpref (gs), |
| 158 | newpref (ss), |
| 159 | newpref (data16), |
| 160 | newpref (addr16), |
| 161 | newpref (rep), |
| 162 | newpref (repne), |
| 163 | newpref (lock) |
| 164 | #undef newpref |
| 165 | }; |
| 166 | #define nknown_prefixes (sizeof (known_prefixes) / sizeof (known_prefixes[0])) |
| 167 | |
| 168 | |
| 169 | #if 0 |
| 170 | static const char *prefix_str[] = |
| 171 | { |
| 172 | #define newpref(pref) [idx_##pref] = #pref |
| 173 | newpref (cs), |
| 174 | newpref (ds), |
| 175 | newpref (es), |
| 176 | newpref (fs), |
| 177 | newpref (gs), |
| 178 | newpref (ss), |
| 179 | newpref (data16), |
| 180 | newpref (addr16), |
| 181 | newpref (rep), |
| 182 | newpref (repne), |
| 183 | newpref (lock) |
| 184 | #undef newpref |
| 185 | }; |
| 186 | #endif |
| 187 | |
| 188 | |
| 189 | static const char amd3dnowstr[] = |
| 190 | #define MNE_3DNOW_PAVGUSB 1 |
| 191 | "pavgusb\0" |
| 192 | #define MNE_3DNOW_PFADD (MNE_3DNOW_PAVGUSB + 8) |
| 193 | "pfadd\0" |
| 194 | #define MNE_3DNOW_PFSUB (MNE_3DNOW_PFADD + 6) |
| 195 | "pfsub\0" |
| 196 | #define MNE_3DNOW_PFSUBR (MNE_3DNOW_PFSUB + 6) |
| 197 | "pfsubr\0" |
| 198 | #define MNE_3DNOW_PFACC (MNE_3DNOW_PFSUBR + 7) |
| 199 | "pfacc\0" |
| 200 | #define MNE_3DNOW_PFCMPGE (MNE_3DNOW_PFACC + 6) |
| 201 | "pfcmpge\0" |
| 202 | #define MNE_3DNOW_PFCMPGT (MNE_3DNOW_PFCMPGE + 8) |
| 203 | "pfcmpgt\0" |
| 204 | #define MNE_3DNOW_PFCMPEQ (MNE_3DNOW_PFCMPGT + 8) |
| 205 | "pfcmpeq\0" |
| 206 | #define MNE_3DNOW_PFMIN (MNE_3DNOW_PFCMPEQ + 8) |
| 207 | "pfmin\0" |
| 208 | #define MNE_3DNOW_PFMAX (MNE_3DNOW_PFMIN + 6) |
| 209 | "pfmax\0" |
| 210 | #define MNE_3DNOW_PI2FD (MNE_3DNOW_PFMAX + 6) |
| 211 | "pi2fd\0" |
| 212 | #define MNE_3DNOW_PF2ID (MNE_3DNOW_PI2FD + 6) |
| 213 | "pf2id\0" |
| 214 | #define MNE_3DNOW_PFRCP (MNE_3DNOW_PF2ID + 6) |
| 215 | "pfrcp\0" |
| 216 | #define MNE_3DNOW_PFRSQRT (MNE_3DNOW_PFRCP + 6) |
| 217 | "pfrsqrt\0" |
| 218 | #define MNE_3DNOW_PFMUL (MNE_3DNOW_PFRSQRT + 8) |
| 219 | "pfmul\0" |
| 220 | #define MNE_3DNOW_PFRCPIT1 (MNE_3DNOW_PFMUL + 6) |
| 221 | "pfrcpit1\0" |
| 222 | #define MNE_3DNOW_PFRSQIT1 (MNE_3DNOW_PFRCPIT1 + 9) |
| 223 | "pfrsqit1\0" |
| 224 | #define MNE_3DNOW_PFRCPIT2 (MNE_3DNOW_PFRSQIT1 + 9) |
| 225 | "pfrcpit2\0" |
| 226 | #define MNE_3DNOW_PMULHRW (MNE_3DNOW_PFRCPIT2 + 9) |
| 227 | "pmulhrw"; |
| 228 | |
| 229 | #define AMD3DNOW_LOW_IDX 0x0d |
| 230 | #define AMD3DNOW_HIGH_IDX (sizeof (amd3dnow) + AMD3DNOW_LOW_IDX - 1) |
| 231 | #define AMD3DNOW_IDX(val) ((val) - AMD3DNOW_LOW_IDX) |
| 232 | static const unsigned char amd3dnow[] = |
| 233 | { |
| 234 | [AMD3DNOW_IDX (0xbf)] = MNE_3DNOW_PAVGUSB, |
| 235 | [AMD3DNOW_IDX (0x9e)] = MNE_3DNOW_PFADD, |
| 236 | [AMD3DNOW_IDX (0x9a)] = MNE_3DNOW_PFSUB, |
| 237 | [AMD3DNOW_IDX (0xaa)] = MNE_3DNOW_PFSUBR, |
| 238 | [AMD3DNOW_IDX (0xae)] = MNE_3DNOW_PFACC, |
| 239 | [AMD3DNOW_IDX (0x90)] = MNE_3DNOW_PFCMPGE, |
| 240 | [AMD3DNOW_IDX (0xa0)] = MNE_3DNOW_PFCMPGT, |
| 241 | [AMD3DNOW_IDX (0xb0)] = MNE_3DNOW_PFCMPEQ, |
| 242 | [AMD3DNOW_IDX (0x94)] = MNE_3DNOW_PFMIN, |
| 243 | [AMD3DNOW_IDX (0xa4)] = MNE_3DNOW_PFMAX, |
| 244 | [AMD3DNOW_IDX (0x0d)] = MNE_3DNOW_PI2FD, |
| 245 | [AMD3DNOW_IDX (0x1d)] = MNE_3DNOW_PF2ID, |
| 246 | [AMD3DNOW_IDX (0x96)] = MNE_3DNOW_PFRCP, |
| 247 | [AMD3DNOW_IDX (0x97)] = MNE_3DNOW_PFRSQRT, |
| 248 | [AMD3DNOW_IDX (0xb4)] = MNE_3DNOW_PFMUL, |
| 249 | [AMD3DNOW_IDX (0xa6)] = MNE_3DNOW_PFRCPIT1, |
| 250 | [AMD3DNOW_IDX (0xa7)] = MNE_3DNOW_PFRSQIT1, |
| 251 | [AMD3DNOW_IDX (0xb6)] = MNE_3DNOW_PFRCPIT2, |
| 252 | [AMD3DNOW_IDX (0xb7)] = MNE_3DNOW_PMULHRW |
| 253 | }; |
| 254 | |
| 255 | |
| 256 | struct output_data |
| 257 | { |
| 258 | GElf_Addr addr; |
| 259 | int *prefixes; |
| 260 | size_t opoff1; |
| 261 | size_t opoff2; |
| 262 | size_t opoff3; |
| 263 | char *bufp; |
| 264 | size_t *bufcntp; |
| 265 | size_t bufsize; |
| 266 | const uint8_t *data; |
| 267 | const uint8_t **param_start; |
| 268 | const uint8_t *end; |
| 269 | char *labelbuf; |
| 270 | size_t labelbufsize; |
| 271 | enum |
| 272 | { |
| 273 | addr_none = 0, |
| 274 | addr_abs_symbolic, |
| 275 | addr_abs_always, |
| 276 | addr_rel_symbolic, |
| 277 | addr_rel_always |
| 278 | } symaddr_use; |
| 279 | GElf_Addr symaddr; |
| 280 | }; |
| 281 | |
| 282 | |
| 283 | #ifndef DISFILE |
| 284 | # define DISFILE "i386_dis.h" |
| 285 | #endif |
| 286 | #include DISFILE |
| 287 | |
| 288 | |
| 289 | #define ADD_CHAR(ch) \ |
| 290 | do { \ |
| 291 | if (unlikely (bufcnt == bufsize)) \ |
| 292 | goto enomem; \ |
| 293 | buf[bufcnt++] = (ch); \ |
| 294 | } while (0) |
| 295 | |
| 296 | #define ADD_STRING(str) \ |
| 297 | do { \ |
| 298 | const char *_str0 = (str); \ |
| 299 | size_t _len0 = strlen (_str0); \ |
| 300 | ADD_NSTRING (_str0, _len0); \ |
| 301 | } while (0) |
| 302 | |
| 303 | #define ADD_NSTRING(str, len) \ |
| 304 | do { \ |
| 305 | const char *_str = (str); \ |
| 306 | size_t _len = (len); \ |
| 307 | if (unlikely (bufcnt + _len > bufsize)) \ |
| 308 | goto enomem; \ |
| 309 | memcpy (buf + bufcnt, _str, _len); \ |
| 310 | bufcnt += _len; \ |
| 311 | } while (0) |
| 312 | |
| 313 | |
| 314 | int |
| 315 | i386_disasm (Ebl *ebl __attribute__((unused)), |
| 316 | const uint8_t **startp, const uint8_t *end, GElf_Addr addr, |
| 317 | const char *fmt, DisasmOutputCB_t outcb, DisasmGetSymCB_t symcb, |
| 318 | void *outcbarg, void *symcbarg) |
| 319 | { |
| 320 | const char *save_fmt = fmt; |
| 321 | |
| 322 | #define BUFSIZE 512 |
| 323 | char initbuf[BUFSIZE]; |
| 324 | int prefixes; |
| 325 | size_t bufcnt; |
| 326 | size_t bufsize = BUFSIZE; |
| 327 | char *buf = initbuf; |
| 328 | const uint8_t *param_start; |
| 329 | |
| 330 | struct output_data output_data = |
| 331 | { |
| 332 | .prefixes = &prefixes, |
| 333 | .bufp = buf, |
| 334 | .bufsize = bufsize, |
| 335 | .bufcntp = &bufcnt, |
| 336 | .param_start = ¶m_start, |
| 337 | .end = end |
| 338 | }; |
| 339 | |
| 340 | int retval = 0; |
| 341 | while (1) |
| 342 | { |
| 343 | prefixes = 0; |
| 344 | |
| 345 | const uint8_t *data = *startp; |
| 346 | const uint8_t *begin = data; |
| 347 | |
| 348 | /* Recognize all prefixes. */ |
| 349 | int last_prefix_bit = 0; |
| 350 | while (data < end) |
| 351 | { |
| 352 | unsigned int i; |
| 353 | for (i = idx_cs; i < nknown_prefixes; ++i) |
| 354 | if (known_prefixes[i] == *data) |
| 355 | break; |
| 356 | if (i == nknown_prefixes) |
| 357 | break; |
| 358 | |
| 359 | prefixes |= last_prefix_bit = 1 << i; |
| 360 | |
| 361 | ++data; |
| 362 | } |
| 363 | |
| 364 | #ifdef X86_64 |
| 365 | if (data < end && (*data & 0xf0) == 0x40) |
| 366 | prefixes |= ((*data++) & 0xf) | has_rex; |
| 367 | #endif |
| 368 | |
| 369 | bufcnt = 0; |
| 370 | size_t cnt = 0; |
| 371 | |
| 372 | const uint8_t *curr = match_data; |
| 373 | const uint8_t *const match_end = match_data + sizeof (match_data); |
| 374 | |
| 375 | assert (data <= end); |
| 376 | if (data == end) |
| 377 | { |
| 378 | if (prefixes != 0) |
| 379 | goto print_prefix; |
| 380 | |
| 381 | retval = -1; |
| 382 | goto do_ret; |
| 383 | } |
| 384 | |
| 385 | next_match: |
| 386 | while (curr < match_end) |
| 387 | { |
| 388 | uint_fast8_t len = *curr++; |
| 389 | uint_fast8_t clen = len >> 4; |
| 390 | len &= 0xf; |
| 391 | const uint8_t *next_curr = curr + clen + (len - clen) * 2; |
| 392 | |
| 393 | assert (len > 0); |
| 394 | assert (curr + clen + 2 * (len - clen) <= match_end); |
| 395 | |
| 396 | const uint8_t *codep = data; |
| 397 | int correct_prefix = 0; |
| 398 | int opoff = 0; |
| 399 | |
| 400 | if (data > begin && codep[-1] == *curr && clen > 0) |
| 401 | { |
| 402 | /* We match a prefix byte. This is exactly one byte and |
| 403 | is matched exactly, without a mask. */ |
| 404 | --len; |
| 405 | --clen; |
| 406 | opoff = 8; |
| 407 | |
| 408 | ++curr; |
| 409 | |
| 410 | assert (last_prefix_bit != 0); |
| 411 | correct_prefix = last_prefix_bit; |
| 412 | } |
| 413 | |
| 414 | size_t avail = len; |
| 415 | while (clen > 0) |
| 416 | { |
| 417 | if (*codep++ != *curr++) |
| 418 | goto not; |
| 419 | --avail; |
| 420 | --clen; |
| 421 | if (codep == end && avail > 0) |
| 422 | goto do_ret; |
| 423 | } |
| 424 | |
| 425 | while (avail > 0) |
| 426 | { |
| 427 | uint_fast8_t masked = *codep++ & *curr++; |
| 428 | if (masked != *curr++) |
| 429 | { |
| 430 | not: |
| 431 | curr = next_curr; |
| 432 | ++cnt; |
| 433 | bufcnt = 0; |
| 434 | goto next_match; |
| 435 | } |
| 436 | |
| 437 | --avail; |
| 438 | if (codep == end && avail > 0) |
| 439 | goto do_ret; |
| 440 | } |
| 441 | |
| 442 | if (len > end - data) |
| 443 | /* There is not enough data for the entire instruction. The |
| 444 | caller can figure this out by looking at the pointer into |
| 445 | the input data. */ |
| 446 | goto do_ret; |
| 447 | |
| 448 | assert (correct_prefix == 0 |
| 449 | || (prefixes & correct_prefix) != 0); |
| 450 | prefixes ^= correct_prefix; |
| 451 | |
| 452 | if (0) |
| 453 | { |
| 454 | /* Resize the buffer. */ |
| 455 | char *oldbuf; |
| 456 | enomem: |
| 457 | oldbuf = buf; |
| 458 | if (buf == initbuf) |
| 459 | buf = malloc (2 * bufsize); |
| 460 | else |
| 461 | buf = realloc (buf, 2 * bufsize); |
| 462 | if (buf == NULL) |
| 463 | { |
| 464 | buf = oldbuf; |
| 465 | retval = ENOMEM; |
| 466 | goto do_ret; |
| 467 | } |
| 468 | bufsize *= 2; |
| 469 | |
| 470 | output_data.bufp = buf; |
| 471 | output_data.bufsize = bufsize; |
| 472 | bufcnt = 0; |
| 473 | |
| 474 | if (data == end) |
| 475 | { |
| 476 | assert (prefixes != 0); |
| 477 | goto print_prefix; |
| 478 | } |
| 479 | |
| 480 | /* gcc is not clever enough to see the following variables |
| 481 | are not used uninitialized. */ |
| 482 | asm ("" |
| 483 | : "=mr" (opoff), "=mr" (correct_prefix), "=mr" (codep), |
| 484 | "=mr" (next_curr), "=mr" (len)); |
| 485 | } |
| 486 | |
| 487 | size_t prefix_size = 0; |
| 488 | |
| 489 | // XXXonly print as prefix if valid? |
| 490 | if ((prefixes & has_lock) != 0) |
| 491 | { |
| 492 | ADD_STRING ("lock "); |
| 493 | prefix_size += 5; |
| 494 | } |
| 495 | |
| 496 | if (instrtab[cnt].rep) |
| 497 | { |
| 498 | if ((prefixes & has_rep) != 0) |
| 499 | { |
| 500 | ADD_STRING ("rep "); |
| 501 | prefix_size += 4; |
| 502 | } |
| 503 | } |
| 504 | else if (instrtab[cnt].repe |
| 505 | && (prefixes & (has_rep | has_repne)) != 0) |
| 506 | { |
| 507 | if ((prefixes & has_repne) != 0) |
| 508 | { |
| 509 | ADD_STRING ("repne "); |
| 510 | prefix_size += 6; |
| 511 | } |
| 512 | else if ((prefixes & has_rep) != 0) |
| 513 | { |
| 514 | ADD_STRING ("repe "); |
| 515 | prefix_size += 5; |
| 516 | } |
| 517 | } |
| 518 | else if ((prefixes & (has_rep | has_repne)) != 0) |
| 519 | { |
| 520 | uint_fast8_t byte; |
| 521 | print_prefix: |
| 522 | bufcnt = 0; |
| 523 | byte = *begin; |
| 524 | /* This is a prefix byte. Print it. */ |
| 525 | switch (byte) |
| 526 | { |
| 527 | case prefix_rep: |
| 528 | ADD_STRING ("rep"); |
| 529 | break; |
| 530 | case prefix_repne: |
| 531 | ADD_STRING ("repne"); |
| 532 | break; |
| 533 | case prefix_cs: |
| 534 | ADD_STRING ("cs"); |
| 535 | break; |
| 536 | case prefix_ds: |
| 537 | ADD_STRING ("ds"); |
| 538 | break; |
| 539 | case prefix_es: |
| 540 | ADD_STRING ("es"); |
| 541 | break; |
| 542 | case prefix_fs: |
| 543 | ADD_STRING ("fs"); |
| 544 | break; |
| 545 | case prefix_gs: |
| 546 | ADD_STRING ("gs"); |
| 547 | break; |
| 548 | case prefix_ss: |
| 549 | ADD_STRING ("ss"); |
| 550 | break; |
| 551 | case prefix_data16: |
| 552 | ADD_STRING ("data16"); |
| 553 | break; |
| 554 | case prefix_addr16: |
| 555 | ADD_STRING ("addr16"); |
| 556 | break; |
| 557 | case prefix_lock: |
| 558 | ADD_STRING ("lock"); |
| 559 | break; |
| 560 | #ifdef X86_64 |
| 561 | case 0x40 ... 0x4f: |
| 562 | ADD_STRING ("rex"); |
| 563 | if (byte != 0x40) |
| 564 | { |
| 565 | ADD_CHAR ('.'); |
| 566 | if (byte & 0x8) |
| 567 | ADD_CHAR ('w'); |
| 568 | if (byte & 0x4) |
| 569 | ADD_CHAR ('r'); |
| 570 | if (byte & 0x3) |
| 571 | ADD_CHAR ('x'); |
| 572 | if (byte & 0x1) |
| 573 | ADD_CHAR ('b'); |
| 574 | } |
| 575 | break; |
| 576 | #endif |
| 577 | default: |
| 578 | /* Cannot happen. */ |
| 579 | puts ("unknown prefix"); |
| 580 | abort (); |
| 581 | } |
| 582 | data = begin + 1; |
| 583 | ++addr; |
| 584 | |
| 585 | goto out; |
| 586 | } |
| 587 | |
| 588 | /* We have a match. First determine how many bytes are |
| 589 | needed for the adressing mode. */ |
| 590 | param_start = codep; |
| 591 | if (instrtab[cnt].modrm) |
| 592 | { |
| 593 | uint_fast8_t modrm = codep[-1]; |
| 594 | |
| 595 | #ifndef X86_64 |
| 596 | if (likely ((prefixes & has_addr16) != 0)) |
| 597 | { |
| 598 | /* Account for displacement. */ |
| 599 | if ((modrm & 0xc7) == 6 || (modrm & 0xc0) == 0x80) |
| 600 | param_start += 2; |
| 601 | else if ((modrm & 0xc0) == 0x40) |
| 602 | param_start += 1; |
| 603 | } |
| 604 | else |
| 605 | #endif |
| 606 | { |
| 607 | /* Account for SIB. */ |
| 608 | if ((modrm & 0xc0) != 0xc0 && (modrm & 0x7) == 0x4) |
| 609 | param_start += 1; |
| 610 | |
| 611 | /* Account for displacement. */ |
| 612 | if ((modrm & 0xc7) == 5 || (modrm & 0xc0) == 0x80 |
| 613 | || ((modrm & 0xc7) == 0x4 && (codep[0] & 0x7) == 0x5)) |
| 614 | param_start += 4; |
| 615 | else if ((modrm & 0xc0) == 0x40) |
| 616 | param_start += 1; |
| 617 | } |
| 618 | |
| 619 | if (unlikely (param_start > end)) |
| 620 | goto not; |
| 621 | } |
| 622 | |
| 623 | output_data.addr = addr + (data - begin); |
| 624 | output_data.data = data; |
| 625 | |
| 626 | unsigned long string_end_idx = 0; |
| 627 | fmt = save_fmt; |
| 628 | const char *deferred_start = NULL; |
| 629 | size_t deferred_len = 0; |
| 630 | // XXX Can we get this from color.c? |
| 631 | static const char color_off[] = "\e[0m"; |
| 632 | while (*fmt != '\0') |
| 633 | { |
| 634 | if (*fmt != '%') |
| 635 | { |
| 636 | char ch = *fmt++; |
| 637 | if (ch == '\\') |
| 638 | { |
| 639 | switch ((ch = *fmt++)) |
| 640 | { |
| 641 | case '0' ... '7': |
| 642 | { |
| 643 | int val = ch - '0'; |
| 644 | ch = *fmt; |
| 645 | if (ch >= '0' && ch <= '7') |
| 646 | { |
| 647 | val *= 8; |
| 648 | val += ch - '0'; |
| 649 | ch = *++fmt; |
| 650 | if (ch >= '0' && ch <= '7' && val < 32) |
| 651 | { |
| 652 | val *= 8; |
| 653 | val += ch - '0'; |
| 654 | ++fmt; |
| 655 | } |
| 656 | } |
| 657 | ch = val; |
| 658 | } |
| 659 | break; |
| 660 | |
| 661 | case 'n': |
| 662 | ch = '\n'; |
| 663 | break; |
| 664 | |
| 665 | case 't': |
| 666 | ch = '\t'; |
| 667 | break; |
| 668 | |
| 669 | default: |
| 670 | retval = EINVAL; |
| 671 | goto do_ret; |
| 672 | } |
| 673 | } |
| 674 | else if (ch == '\e' && *fmt == '[') |
| 675 | { |
| 676 | deferred_start = fmt - 1; |
| 677 | do |
| 678 | ++fmt; |
| 679 | while (*fmt != 'm' && *fmt != '\0'); |
| 680 | |
| 681 | if (*fmt == 'm') |
| 682 | { |
| 683 | deferred_len = ++fmt - deferred_start; |
| 684 | continue; |
| 685 | } |
| 686 | |
| 687 | fmt = deferred_start + 1; |
| 688 | deferred_start = NULL; |
| 689 | } |
| 690 | ADD_CHAR (ch); |
| 691 | continue; |
| 692 | } |
| 693 | ++fmt; |
| 694 | |
| 695 | int width = 0; |
| 696 | while (isdigit (*fmt)) |
| 697 | width = width * 10 + (*fmt++ - '0'); |
| 698 | |
| 699 | int prec = 0; |
| 700 | if (*fmt == '.') |
| 701 | while (isdigit (*++fmt)) |
| 702 | prec = prec * 10 + (*fmt - '0'); |
| 703 | |
| 704 | size_t start_idx = bufcnt; |
| 705 | size_t non_printing = 0; |
| 706 | switch (*fmt++) |
| 707 | { |
| 708 | char mnebuf[16]; |
| 709 | const char *str; |
| 710 | |
| 711 | case 'm': |
| 712 | /* Mnemonic. */ |
| 713 | |
| 714 | if (unlikely (instrtab[cnt].mnemonic == MNE_INVALID)) |
| 715 | { |
| 716 | switch (*data) |
| 717 | { |
| 718 | #ifdef X86_64 |
| 719 | case 0x90: |
| 720 | if (prefixes & has_rex_b) |
| 721 | goto not; |
| 722 | str = "nop"; |
| 723 | break; |
| 724 | #endif |
| 725 | |
| 726 | case 0x98: |
| 727 | #ifdef X86_64 |
| 728 | if (prefixes == (has_rex_w | has_rex)) |
| 729 | { |
| 730 | str = "cltq"; |
| 731 | break; |
| 732 | } |
| 733 | #endif |
| 734 | if (prefixes & ~has_data16) |
| 735 | goto print_prefix; |
| 736 | str = prefixes & has_data16 ? "cbtw" : "cwtl"; |
| 737 | break; |
| 738 | |
| 739 | case 0x99: |
| 740 | #ifdef X86_64 |
| 741 | if (prefixes == (has_rex_w | has_rex)) |
| 742 | { |
| 743 | str = "cqto"; |
| 744 | break; |
| 745 | } |
| 746 | #endif |
| 747 | if (prefixes & ~has_data16) |
| 748 | goto print_prefix; |
| 749 | str = prefixes & has_data16 ? "cwtd" : "cltd"; |
| 750 | break; |
| 751 | |
| 752 | case 0xe3: |
| 753 | if (prefixes & ~has_addr16) |
| 754 | goto print_prefix; |
| 755 | #ifdef X86_64 |
| 756 | str = prefixes & has_addr16 ? "jecxz" : "jrcxz"; |
| 757 | #else |
| 758 | str = prefixes & has_addr16 ? "jcxz" : "jecxz"; |
| 759 | #endif |
| 760 | break; |
| 761 | |
| 762 | case 0x0f: |
| 763 | if (data[1] == 0x0f) |
| 764 | { |
| 765 | /* AMD 3DNOW. We need one more byte. */ |
| 766 | if (param_start >= end) |
| 767 | goto not; |
| 768 | if (*param_start < AMD3DNOW_LOW_IDX |
| 769 | || *param_start > AMD3DNOW_HIGH_IDX) |
| 770 | goto not; |
| 771 | unsigned int idx |
| 772 | = amd3dnow[AMD3DNOW_IDX (*param_start)]; |
| 773 | if (idx == 0) |
| 774 | goto not; |
| 775 | str = amd3dnowstr + idx - 1; |
| 776 | /* Eat the immediate byte indicating the |
| 777 | operation. */ |
| 778 | ++param_start; |
| 779 | break; |
| 780 | } |
| 781 | #ifdef X86_64 |
| 782 | if (data[1] == 0xc7) |
| 783 | { |
| 784 | str = ((prefixes & has_rex_w) |
| 785 | ? "cmpxchg16b" : "cmpxchg8b"); |
| 786 | break; |
| 787 | } |
| 788 | #endif |
| 789 | if (data[1] == 0xc2) |
| 790 | { |
| 791 | if (param_start >= end) |
| 792 | goto not; |
| 793 | if (*param_start > 7) |
| 794 | goto not; |
| 795 | static const char cmpops[][9] = |
| 796 | { |
| 797 | [0] = "cmpeq", |
| 798 | [1] = "cmplt", |
| 799 | [2] = "cmple", |
| 800 | [3] = "cmpunord", |
| 801 | [4] = "cmpneq", |
| 802 | [5] = "cmpnlt", |
| 803 | [6] = "cmpnle", |
| 804 | [7] = "cmpord" |
| 805 | }; |
| 806 | char *cp = stpcpy (mnebuf, cmpops[*param_start]); |
| 807 | if (correct_prefix & (has_rep | has_repne)) |
| 808 | *cp++ = 's'; |
| 809 | else |
| 810 | *cp++ = 'p'; |
| 811 | if (correct_prefix & (has_data16 | has_repne)) |
| 812 | *cp++ = 'd'; |
| 813 | else |
| 814 | *cp++ = 's'; |
| 815 | *cp = '\0'; |
| 816 | str = mnebuf; |
| 817 | /* Eat the immediate byte indicating the |
| 818 | operation. */ |
| 819 | ++param_start; |
| 820 | break; |
| 821 | } |
| 822 | FALLTHROUGH; |
| 823 | default: |
| 824 | assert (! "INVALID not handled"); |
| 825 | } |
| 826 | } |
| 827 | else |
| 828 | str = mnestr.str + mneidx[instrtab[cnt].mnemonic]; |
| 829 | |
| 830 | if (deferred_start != NULL) |
| 831 | { |
| 832 | ADD_NSTRING (deferred_start, deferred_len); |
| 833 | non_printing += deferred_len; |
| 834 | } |
| 835 | |
| 836 | ADD_STRING (str); |
| 837 | |
| 838 | switch (instrtab[cnt].suffix) |
| 839 | { |
| 840 | case suffix_none: |
| 841 | break; |
| 842 | |
| 843 | case suffix_w: |
| 844 | if ((codep[-1] & 0xc0) != 0xc0) |
| 845 | { |
| 846 | char ch; |
| 847 | |
| 848 | if (data[0] & 1) |
| 849 | { |
| 850 | if (prefixes & has_data16) |
| 851 | ch = 'w'; |
| 852 | #ifdef X86_64 |
| 853 | else if (prefixes & has_rex_w) |
| 854 | ch = 'q'; |
| 855 | #endif |
| 856 | else |
| 857 | ch = 'l'; |
| 858 | } |
| 859 | else |
| 860 | ch = 'b'; |
| 861 | |
| 862 | ADD_CHAR (ch); |
| 863 | } |
| 864 | break; |
| 865 | |
| 866 | case suffix_w0: |
| 867 | if ((codep[-1] & 0xc0) != 0xc0) |
| 868 | ADD_CHAR ('l'); |
| 869 | break; |
| 870 | |
| 871 | case suffix_w1: |
| 872 | if ((data[0] & 0x4) == 0) |
| 873 | ADD_CHAR ('l'); |
| 874 | break; |
| 875 | |
| 876 | case suffix_W: |
| 877 | if (prefixes & has_data16) |
| 878 | { |
| 879 | ADD_CHAR ('w'); |
| 880 | prefixes &= ~has_data16; |
| 881 | } |
| 882 | #ifdef X86_64 |
| 883 | else |
| 884 | ADD_CHAR ('q'); |
| 885 | #endif |
| 886 | break; |
| 887 | |
| 888 | case suffix_W1: |
| 889 | if (prefixes & has_data16) |
| 890 | { |
| 891 | ADD_CHAR ('w'); |
| 892 | prefixes &= ~has_data16; |
| 893 | } |
| 894 | #ifdef X86_64 |
| 895 | else if (prefixes & has_rex_w) |
| 896 | ADD_CHAR ('q'); |
| 897 | #endif |
| 898 | break; |
| 899 | |
| 900 | case suffix_tttn:; |
| 901 | static const char tttn[16][3] = |
| 902 | { |
| 903 | "o", "no", "b", "ae", "e", "ne", "be", "a", |
| 904 | "s", "ns", "p", "np", "l", "ge", "le", "g" |
| 905 | }; |
| 906 | ADD_STRING (tttn[codep[-1 - instrtab[cnt].modrm] & 0x0f]); |
| 907 | break; |
| 908 | |
| 909 | case suffix_D: |
| 910 | if ((codep[-1] & 0xc0) != 0xc0) |
| 911 | ADD_CHAR ((data[0] & 0x04) == 0 ? 's' : 'l'); |
| 912 | break; |
| 913 | |
| 914 | default: |
| 915 | printf("unknown suffix %d\n", instrtab[cnt].suffix); |
| 916 | abort (); |
| 917 | } |
| 918 | |
| 919 | if (deferred_start != NULL) |
| 920 | { |
| 921 | ADD_STRING (color_off); |
| 922 | non_printing += strlen (color_off); |
| 923 | } |
| 924 | |
| 925 | string_end_idx = bufcnt; |
| 926 | break; |
| 927 | |
| 928 | case 'o': |
| 929 | if (prec == 1 && instrtab[cnt].fct1 != 0) |
| 930 | { |
| 931 | /* First parameter. */ |
| 932 | if (deferred_start != NULL) |
| 933 | { |
| 934 | ADD_NSTRING (deferred_start, deferred_len); |
| 935 | non_printing += deferred_len; |
| 936 | } |
| 937 | |
| 938 | if (instrtab[cnt].str1 != 0) |
| 939 | ADD_STRING (op1_str |
| 940 | + op1_str_idx[instrtab[cnt].str1 - 1]); |
| 941 | |
| 942 | output_data.opoff1 = (instrtab[cnt].off1_1 |
| 943 | + OFF1_1_BIAS - opoff); |
| 944 | output_data.opoff2 = (instrtab[cnt].off1_2 |
| 945 | + OFF1_2_BIAS - opoff); |
| 946 | output_data.opoff3 = (instrtab[cnt].off1_3 |
| 947 | + OFF1_3_BIAS - opoff); |
| 948 | int r = op1_fct[instrtab[cnt].fct1] (&output_data); |
| 949 | if (r < 0) |
| 950 | goto not; |
| 951 | if (r > 0) |
| 952 | goto enomem; |
| 953 | |
| 954 | if (deferred_start != NULL) |
| 955 | { |
| 956 | ADD_STRING (color_off); |
| 957 | non_printing += strlen (color_off); |
| 958 | } |
| 959 | |
| 960 | string_end_idx = bufcnt; |
| 961 | } |
| 962 | else if (prec == 2 && instrtab[cnt].fct2 != 0) |
| 963 | { |
| 964 | /* Second parameter. */ |
| 965 | if (deferred_start != NULL) |
| 966 | { |
| 967 | ADD_NSTRING (deferred_start, deferred_len); |
| 968 | non_printing += deferred_len; |
| 969 | } |
| 970 | |
| 971 | if (instrtab[cnt].str2 != 0) |
| 972 | ADD_STRING (op2_str |
| 973 | + op2_str_idx[instrtab[cnt].str2 - 1]); |
| 974 | |
| 975 | output_data.opoff1 = (instrtab[cnt].off2_1 |
| 976 | + OFF2_1_BIAS - opoff); |
| 977 | output_data.opoff2 = (instrtab[cnt].off2_2 |
| 978 | + OFF2_2_BIAS - opoff); |
| 979 | output_data.opoff3 = (instrtab[cnt].off2_3 |
| 980 | + OFF2_3_BIAS - opoff); |
| 981 | int r = op2_fct[instrtab[cnt].fct2] (&output_data); |
| 982 | if (r < 0) |
| 983 | goto not; |
| 984 | if (r > 0) |
| 985 | goto enomem; |
| 986 | |
| 987 | if (deferred_start != NULL) |
| 988 | { |
| 989 | ADD_STRING (color_off); |
| 990 | non_printing += strlen (color_off); |
| 991 | } |
| 992 | |
| 993 | string_end_idx = bufcnt; |
| 994 | } |
| 995 | else if (prec == 3 && instrtab[cnt].fct3 != 0) |
| 996 | { |
| 997 | /* Third parameter. */ |
| 998 | if (deferred_start != NULL) |
| 999 | { |
| 1000 | ADD_NSTRING (deferred_start, deferred_len); |
| 1001 | non_printing += deferred_len; |
| 1002 | } |
| 1003 | |
| 1004 | if (instrtab[cnt].str3 != 0) |
| 1005 | ADD_STRING (op3_str |
| 1006 | + op3_str_idx[instrtab[cnt].str3 - 1]); |
| 1007 | |
| 1008 | output_data.opoff1 = (instrtab[cnt].off3_1 |
| 1009 | + OFF3_1_BIAS - opoff); |
| 1010 | output_data.opoff2 = (instrtab[cnt].off3_2 |
| 1011 | + OFF3_2_BIAS - opoff); |
| 1012 | #ifdef OFF3_3_BITS |
| 1013 | output_data.opoff3 = (instrtab[cnt].off3_3 |
| 1014 | + OFF3_3_BIAS - opoff); |
| 1015 | #else |
| 1016 | output_data.opoff3 = 0; |
| 1017 | #endif |
| 1018 | int r = op3_fct[instrtab[cnt].fct3] (&output_data); |
| 1019 | if (r < 0) |
| 1020 | goto not; |
| 1021 | if (r > 0) |
| 1022 | goto enomem; |
| 1023 | |
| 1024 | if (deferred_start != NULL) |
| 1025 | { |
| 1026 | ADD_STRING (color_off); |
| 1027 | non_printing += strlen (color_off); |
| 1028 | } |
| 1029 | |
| 1030 | string_end_idx = bufcnt; |
| 1031 | } |
| 1032 | else |
| 1033 | bufcnt = string_end_idx; |
| 1034 | break; |
| 1035 | |
| 1036 | case 'e': |
| 1037 | string_end_idx = bufcnt; |
| 1038 | break; |
| 1039 | |
| 1040 | case 'a': |
| 1041 | /* Pad to requested column. */ |
| 1042 | while (bufcnt - non_printing < (size_t) width) |
| 1043 | ADD_CHAR (' '); |
| 1044 | width = 0; |
| 1045 | break; |
| 1046 | |
| 1047 | case 'l': |
| 1048 | if (deferred_start != NULL) |
| 1049 | { |
| 1050 | ADD_NSTRING (deferred_start, deferred_len); |
| 1051 | non_printing += deferred_len; |
| 1052 | } |
| 1053 | |
| 1054 | if (output_data.labelbuf != NULL |
| 1055 | && output_data.labelbuf[0] != '\0') |
| 1056 | { |
| 1057 | ADD_STRING (output_data.labelbuf); |
| 1058 | output_data.labelbuf[0] = '\0'; |
| 1059 | string_end_idx = bufcnt; |
| 1060 | } |
| 1061 | else if (output_data.symaddr_use != addr_none) |
| 1062 | { |
| 1063 | GElf_Addr symaddr = output_data.symaddr; |
| 1064 | if (output_data.symaddr_use >= addr_rel_symbolic) |
| 1065 | symaddr += addr + param_start - begin; |
| 1066 | |
| 1067 | // XXX Lookup symbol based on symaddr |
| 1068 | const char *symstr = NULL; |
| 1069 | if (symcb != NULL |
| 1070 | && symcb (0 /* XXX */, 0 /* XXX */, symaddr, |
| 1071 | &output_data.labelbuf, |
| 1072 | &output_data.labelbufsize, symcbarg) == 0) |
| 1073 | symstr = output_data.labelbuf; |
| 1074 | |
| 1075 | size_t bufavail = bufsize - bufcnt; |
| 1076 | int r = 0; |
| 1077 | if (symstr != NULL) |
| 1078 | r = snprintf (&buf[bufcnt], bufavail, "# <%s>", |
| 1079 | symstr); |
| 1080 | else if (output_data.symaddr_use == addr_abs_always |
| 1081 | || output_data.symaddr_use == addr_rel_always) |
| 1082 | r = snprintf (&buf[bufcnt], bufavail, "# %#" PRIx64, |
| 1083 | (uint64_t) symaddr); |
| 1084 | |
| 1085 | assert (r >= 0); |
| 1086 | if ((size_t) r >= bufavail) |
| 1087 | goto enomem; |
| 1088 | bufcnt += r; |
| 1089 | string_end_idx = bufcnt; |
| 1090 | |
| 1091 | output_data.symaddr_use = addr_none; |
| 1092 | } |
| 1093 | if (deferred_start != NULL) |
| 1094 | { |
| 1095 | ADD_STRING (color_off); |
| 1096 | non_printing += strlen (color_off); |
| 1097 | } |
| 1098 | break; |
| 1099 | |
| 1100 | default: |
| 1101 | abort (); |
| 1102 | } |
| 1103 | |
| 1104 | deferred_start = NULL; |
| 1105 | |
| 1106 | /* Pad according to the specified width. */ |
| 1107 | while (bufcnt + prefix_size - non_printing < start_idx + width) |
| 1108 | ADD_CHAR (' '); |
| 1109 | prefix_size = 0; |
| 1110 | } |
| 1111 | |
| 1112 | if ((prefixes & SEGMENT_PREFIXES) != 0) |
| 1113 | goto print_prefix; |
| 1114 | |
| 1115 | assert (string_end_idx != ~0ul); |
| 1116 | bufcnt = string_end_idx; |
| 1117 | |
| 1118 | addr += param_start - begin; |
| 1119 | data = param_start; |
| 1120 | |
| 1121 | goto out; |
| 1122 | } |
| 1123 | |
| 1124 | /* Invalid (or at least unhandled) opcode. */ |
| 1125 | if (prefixes != 0) |
| 1126 | goto print_prefix; |
| 1127 | assert (*startp == data); |
| 1128 | ++data; |
| 1129 | ADD_STRING ("(bad)"); |
| 1130 | addr += data - begin; |
| 1131 | |
| 1132 | out: |
| 1133 | if (bufcnt == bufsize) |
| 1134 | goto enomem; |
| 1135 | buf[bufcnt] = '\0'; |
| 1136 | |
| 1137 | *startp = data; |
| 1138 | retval = outcb (buf, bufcnt, outcbarg); |
| 1139 | if (retval != 0) |
| 1140 | goto do_ret; |
| 1141 | } |
| 1142 | |
| 1143 | do_ret: |
| 1144 | free (output_data.labelbuf); |
| 1145 | if (buf != initbuf) |
| 1146 | free (buf); |
| 1147 | |
| 1148 | return retval; |
| 1149 | } |