| 1 | /* Assembler interface for targets using CGEN. -*- C -*- |
| 2 | CGEN: Cpu tools GENerator |
| 3 | |
| 4 | THIS FILE IS MACHINE GENERATED WITH CGEN. |
| 5 | - the resultant file is machine generated, cgen-asm.in isn't |
| 6 | |
| 7 | Copyright (C) 1996-2020 Free Software Foundation, Inc. |
| 8 | |
| 9 | This file is part of libopcodes. |
| 10 | |
| 11 | This library is free software; you can redistribute it and/or modify |
| 12 | it under the terms of the GNU General Public License as published by |
| 13 | the Free Software Foundation; either version 3, or (at your option) |
| 14 | any later version. |
| 15 | |
| 16 | It is distributed in the hope that it will be useful, but WITHOUT |
| 17 | ANY WARRANTY; without even the implied warranty of MERCHANTABILITY |
| 18 | or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public |
| 19 | License for more details. |
| 20 | |
| 21 | You should have received a copy of the GNU General Public License |
| 22 | along with this program; if not, write to the Free Software Foundation, Inc., |
| 23 | 51 Franklin Street - Fifth Floor, Boston, MA 02110-1301, USA. */ |
| 24 | |
| 25 | |
| 26 | /* ??? Eventually more and more of this stuff can go to cpu-independent files. |
| 27 | Keep that in mind. */ |
| 28 | |
| 29 | #include "sysdep.h" |
| 30 | #include <stdio.h> |
| 31 | #include "ansidecl.h" |
| 32 | #include "bfd.h" |
| 33 | #include "symcat.h" |
| 34 | #include "@prefix@-desc.h" |
| 35 | #include "@prefix@-opc.h" |
| 36 | #include "opintl.h" |
| 37 | #include "xregex.h" |
| 38 | #include "libiberty.h" |
| 39 | #include "safe-ctype.h" |
| 40 | |
| 41 | #undef min |
| 42 | #define min(a,b) ((a) < (b) ? (a) : (b)) |
| 43 | #undef max |
| 44 | #define max(a,b) ((a) > (b) ? (a) : (b)) |
| 45 | |
| 46 | static const char * parse_insn_normal |
| 47 | (CGEN_CPU_DESC, const CGEN_INSN *, const char **, CGEN_FIELDS *); |
| 48 | \f |
| 49 | /* -- assembler routines inserted here. */ |
| 50 | \f |
| 51 | |
| 52 | /* Regex construction routine. |
| 53 | |
| 54 | This translates an opcode syntax string into a regex string, |
| 55 | by replacing any non-character syntax element (such as an |
| 56 | opcode) with the pattern '.*' |
| 57 | |
| 58 | It then compiles the regex and stores it in the opcode, for |
| 59 | later use by @arch@_cgen_assemble_insn |
| 60 | |
| 61 | Returns NULL for success, an error message for failure. */ |
| 62 | |
| 63 | char * |
| 64 | @arch@_cgen_build_insn_regex (CGEN_INSN *insn) |
| 65 | { |
| 66 | CGEN_OPCODE *opc = (CGEN_OPCODE *) CGEN_INSN_OPCODE (insn); |
| 67 | const char *mnem = CGEN_INSN_MNEMONIC (insn); |
| 68 | char rxbuf[CGEN_MAX_RX_ELEMENTS]; |
| 69 | char *rx = rxbuf; |
| 70 | const CGEN_SYNTAX_CHAR_TYPE *syn; |
| 71 | int reg_err; |
| 72 | |
| 73 | syn = CGEN_SYNTAX_STRING (CGEN_OPCODE_SYNTAX (opc)); |
| 74 | |
| 75 | /* Mnemonics come first in the syntax string. */ |
| 76 | if (! CGEN_SYNTAX_MNEMONIC_P (* syn)) |
| 77 | return _("missing mnemonic in syntax string"); |
| 78 | ++syn; |
| 79 | |
| 80 | /* Generate a case sensitive regular expression that emulates case |
| 81 | insensitive matching in the "C" locale. We cannot generate a case |
| 82 | insensitive regular expression because in Turkish locales, 'i' and 'I' |
| 83 | are not equal modulo case conversion. */ |
| 84 | |
| 85 | /* Copy the literal mnemonic out of the insn. */ |
| 86 | for (; *mnem; mnem++) |
| 87 | { |
| 88 | char c = *mnem; |
| 89 | |
| 90 | if (ISALPHA (c)) |
| 91 | { |
| 92 | *rx++ = '['; |
| 93 | *rx++ = TOLOWER (c); |
| 94 | *rx++ = TOUPPER (c); |
| 95 | *rx++ = ']'; |
| 96 | } |
| 97 | else |
| 98 | *rx++ = c; |
| 99 | } |
| 100 | |
| 101 | /* Copy any remaining literals from the syntax string into the rx. */ |
| 102 | for(; * syn != 0 && rx <= rxbuf + (CGEN_MAX_RX_ELEMENTS - 7 - 4); ++syn) |
| 103 | { |
| 104 | if (CGEN_SYNTAX_CHAR_P (* syn)) |
| 105 | { |
| 106 | char c = CGEN_SYNTAX_CHAR (* syn); |
| 107 | |
| 108 | switch (c) |
| 109 | { |
| 110 | /* Escape any regex metacharacters in the syntax. */ |
| 111 | case '.': case '[': case '\\': |
| 112 | case '*': case '^': case '$': |
| 113 | |
| 114 | #ifdef CGEN_ESCAPE_EXTENDED_REGEX |
| 115 | case '?': case '{': case '}': |
| 116 | case '(': case ')': case '*': |
| 117 | case '|': case '+': case ']': |
| 118 | #endif |
| 119 | *rx++ = '\\'; |
| 120 | *rx++ = c; |
| 121 | break; |
| 122 | |
| 123 | default: |
| 124 | if (ISALPHA (c)) |
| 125 | { |
| 126 | *rx++ = '['; |
| 127 | *rx++ = TOLOWER (c); |
| 128 | *rx++ = TOUPPER (c); |
| 129 | *rx++ = ']'; |
| 130 | } |
| 131 | else |
| 132 | *rx++ = c; |
| 133 | break; |
| 134 | } |
| 135 | } |
| 136 | else |
| 137 | { |
| 138 | /* Replace non-syntax fields with globs. */ |
| 139 | *rx++ = '.'; |
| 140 | *rx++ = '*'; |
| 141 | } |
| 142 | } |
| 143 | |
| 144 | /* Trailing whitespace ok. */ |
| 145 | * rx++ = '['; |
| 146 | * rx++ = ' '; |
| 147 | * rx++ = '\t'; |
| 148 | * rx++ = ']'; |
| 149 | * rx++ = '*'; |
| 150 | |
| 151 | /* But anchor it after that. */ |
| 152 | * rx++ = '$'; |
| 153 | * rx = '\0'; |
| 154 | |
| 155 | CGEN_INSN_RX (insn) = xmalloc (sizeof (regex_t)); |
| 156 | reg_err = regcomp ((regex_t *) CGEN_INSN_RX (insn), rxbuf, REG_NOSUB); |
| 157 | |
| 158 | if (reg_err == 0) |
| 159 | return NULL; |
| 160 | else |
| 161 | { |
| 162 | static char msg[80]; |
| 163 | |
| 164 | regerror (reg_err, (regex_t *) CGEN_INSN_RX (insn), msg, 80); |
| 165 | regfree ((regex_t *) CGEN_INSN_RX (insn)); |
| 166 | free (CGEN_INSN_RX (insn)); |
| 167 | (CGEN_INSN_RX (insn)) = NULL; |
| 168 | return msg; |
| 169 | } |
| 170 | } |
| 171 | |
| 172 | \f |
| 173 | /* Default insn parser. |
| 174 | |
| 175 | The syntax string is scanned and operands are parsed and stored in FIELDS. |
| 176 | Relocs are queued as we go via other callbacks. |
| 177 | |
| 178 | ??? Note that this is currently an all-or-nothing parser. If we fail to |
| 179 | parse the instruction, we return 0 and the caller will start over from |
| 180 | the beginning. Backtracking will be necessary in parsing subexpressions, |
| 181 | but that can be handled there. Not handling backtracking here may get |
| 182 | expensive in the case of the m68k. Deal with later. |
| 183 | |
| 184 | Returns NULL for success, an error message for failure. */ |
| 185 | |
| 186 | static const char * |
| 187 | parse_insn_normal (CGEN_CPU_DESC cd, |
| 188 | const CGEN_INSN *insn, |
| 189 | const char **strp, |
| 190 | CGEN_FIELDS *fields) |
| 191 | { |
| 192 | /* ??? Runtime added insns not handled yet. */ |
| 193 | const CGEN_SYNTAX *syntax = CGEN_INSN_SYNTAX (insn); |
| 194 | const char *str = *strp; |
| 195 | const char *errmsg; |
| 196 | const char *p; |
| 197 | const CGEN_SYNTAX_CHAR_TYPE * syn; |
| 198 | #ifdef CGEN_MNEMONIC_OPERANDS |
| 199 | /* FIXME: wip */ |
| 200 | int past_opcode_p; |
| 201 | #endif |
| 202 | |
| 203 | /* For now we assume the mnemonic is first (there are no leading operands). |
| 204 | We can parse it without needing to set up operand parsing. |
| 205 | GAS's input scrubber will ensure mnemonics are lowercase, but we may |
| 206 | not be called from GAS. */ |
| 207 | p = CGEN_INSN_MNEMONIC (insn); |
| 208 | while (*p && TOLOWER (*p) == TOLOWER (*str)) |
| 209 | ++p, ++str; |
| 210 | |
| 211 | if (* p) |
| 212 | return _("unrecognized instruction"); |
| 213 | |
| 214 | #ifndef CGEN_MNEMONIC_OPERANDS |
| 215 | if (* str && ! ISSPACE (* str)) |
| 216 | return _("unrecognized instruction"); |
| 217 | #endif |
| 218 | |
| 219 | CGEN_INIT_PARSE (cd); |
| 220 | cgen_init_parse_operand (cd); |
| 221 | #ifdef CGEN_MNEMONIC_OPERANDS |
| 222 | past_opcode_p = 0; |
| 223 | #endif |
| 224 | |
| 225 | /* We don't check for (*str != '\0') here because we want to parse |
| 226 | any trailing fake arguments in the syntax string. */ |
| 227 | syn = CGEN_SYNTAX_STRING (syntax); |
| 228 | |
| 229 | /* Mnemonics come first for now, ensure valid string. */ |
| 230 | if (! CGEN_SYNTAX_MNEMONIC_P (* syn)) |
| 231 | abort (); |
| 232 | |
| 233 | ++syn; |
| 234 | |
| 235 | while (* syn != 0) |
| 236 | { |
| 237 | /* Non operand chars must match exactly. */ |
| 238 | if (CGEN_SYNTAX_CHAR_P (* syn)) |
| 239 | { |
| 240 | /* FIXME: While we allow for non-GAS callers above, we assume the |
| 241 | first char after the mnemonic part is a space. */ |
| 242 | /* FIXME: We also take inappropriate advantage of the fact that |
| 243 | GAS's input scrubber will remove extraneous blanks. */ |
| 244 | if (TOLOWER (*str) == TOLOWER (CGEN_SYNTAX_CHAR (* syn))) |
| 245 | { |
| 246 | #ifdef CGEN_MNEMONIC_OPERANDS |
| 247 | if (CGEN_SYNTAX_CHAR(* syn) == ' ') |
| 248 | past_opcode_p = 1; |
| 249 | #endif |
| 250 | ++ syn; |
| 251 | ++ str; |
| 252 | } |
| 253 | else if (*str) |
| 254 | { |
| 255 | /* Syntax char didn't match. Can't be this insn. */ |
| 256 | static char msg [80]; |
| 257 | |
| 258 | /* xgettext:c-format */ |
| 259 | sprintf (msg, _("syntax error (expected char `%c', found `%c')"), |
| 260 | CGEN_SYNTAX_CHAR(*syn), *str); |
| 261 | return msg; |
| 262 | } |
| 263 | else |
| 264 | { |
| 265 | /* Ran out of input. */ |
| 266 | static char msg [80]; |
| 267 | |
| 268 | /* xgettext:c-format */ |
| 269 | sprintf (msg, _("syntax error (expected char `%c', found end of instruction)"), |
| 270 | CGEN_SYNTAX_CHAR(*syn)); |
| 271 | return msg; |
| 272 | } |
| 273 | continue; |
| 274 | } |
| 275 | |
| 276 | #ifdef CGEN_MNEMONIC_OPERANDS |
| 277 | (void) past_opcode_p; |
| 278 | #endif |
| 279 | /* We have an operand of some sort. */ |
| 280 | errmsg = cd->parse_operand (cd, CGEN_SYNTAX_FIELD (*syn), &str, fields); |
| 281 | if (errmsg) |
| 282 | return errmsg; |
| 283 | |
| 284 | /* Done with this operand, continue with next one. */ |
| 285 | ++ syn; |
| 286 | } |
| 287 | |
| 288 | /* If we're at the end of the syntax string, we're done. */ |
| 289 | if (* syn == 0) |
| 290 | { |
| 291 | /* FIXME: For the moment we assume a valid `str' can only contain |
| 292 | blanks now. IE: We needn't try again with a longer version of |
| 293 | the insn and it is assumed that longer versions of insns appear |
| 294 | before shorter ones (eg: lsr r2,r3,1 vs lsr r2,r3). */ |
| 295 | while (ISSPACE (* str)) |
| 296 | ++ str; |
| 297 | |
| 298 | if (* str != '\0') |
| 299 | return _("junk at end of line"); /* FIXME: would like to include `str' */ |
| 300 | |
| 301 | return NULL; |
| 302 | } |
| 303 | |
| 304 | /* We couldn't parse it. */ |
| 305 | return _("unrecognized instruction"); |
| 306 | } |
| 307 | \f |
| 308 | /* Main entry point. |
| 309 | This routine is called for each instruction to be assembled. |
| 310 | STR points to the insn to be assembled. |
| 311 | We assume all necessary tables have been initialized. |
| 312 | The assembled instruction, less any fixups, is stored in BUF. |
| 313 | Remember that if CGEN_INT_INSN_P then BUF is an int and thus the value |
| 314 | still needs to be converted to target byte order, otherwise BUF is an array |
| 315 | of bytes in target byte order. |
| 316 | The result is a pointer to the insn's entry in the opcode table, |
| 317 | or NULL if an error occured (an error message will have already been |
| 318 | printed). |
| 319 | |
| 320 | Note that when processing (non-alias) macro-insns, |
| 321 | this function recurses. |
| 322 | |
| 323 | ??? It's possible to make this cpu-independent. |
| 324 | One would have to deal with a few minor things. |
| 325 | At this point in time doing so would be more of a curiosity than useful |
| 326 | [for example this file isn't _that_ big], but keeping the possibility in |
| 327 | mind helps keep the design clean. */ |
| 328 | |
| 329 | const CGEN_INSN * |
| 330 | @arch@_cgen_assemble_insn (CGEN_CPU_DESC cd, |
| 331 | const char *str, |
| 332 | CGEN_FIELDS *fields, |
| 333 | CGEN_INSN_BYTES_PTR buf, |
| 334 | char **errmsg) |
| 335 | { |
| 336 | const char *start; |
| 337 | CGEN_INSN_LIST *ilist; |
| 338 | const char *parse_errmsg = NULL; |
| 339 | const char *insert_errmsg = NULL; |
| 340 | int recognized_mnemonic = 0; |
| 341 | |
| 342 | /* Skip leading white space. */ |
| 343 | while (ISSPACE (* str)) |
| 344 | ++ str; |
| 345 | |
| 346 | /* The instructions are stored in hashed lists. |
| 347 | Get the first in the list. */ |
| 348 | ilist = CGEN_ASM_LOOKUP_INSN (cd, str); |
| 349 | |
| 350 | /* Keep looking until we find a match. */ |
| 351 | start = str; |
| 352 | for ( ; ilist != NULL ; ilist = CGEN_ASM_NEXT_INSN (ilist)) |
| 353 | { |
| 354 | const CGEN_INSN *insn = ilist->insn; |
| 355 | recognized_mnemonic = 1; |
| 356 | |
| 357 | #ifdef CGEN_VALIDATE_INSN_SUPPORTED |
| 358 | /* Not usually needed as unsupported opcodes |
| 359 | shouldn't be in the hash lists. */ |
| 360 | /* Is this insn supported by the selected cpu? */ |
| 361 | if (! @arch@_cgen_insn_supported (cd, insn)) |
| 362 | continue; |
| 363 | #endif |
| 364 | /* If the RELAXED attribute is set, this is an insn that shouldn't be |
| 365 | chosen immediately. Instead, it is used during assembler/linker |
| 366 | relaxation if possible. */ |
| 367 | if (CGEN_INSN_ATTR_VALUE (insn, CGEN_INSN_RELAXED) != 0) |
| 368 | continue; |
| 369 | |
| 370 | str = start; |
| 371 | |
| 372 | /* Skip this insn if str doesn't look right lexically. */ |
| 373 | if (CGEN_INSN_RX (insn) != NULL && |
| 374 | regexec ((regex_t *) CGEN_INSN_RX (insn), str, 0, NULL, 0) == REG_NOMATCH) |
| 375 | continue; |
| 376 | |
| 377 | /* Allow parse/insert handlers to obtain length of insn. */ |
| 378 | CGEN_FIELDS_BITSIZE (fields) = CGEN_INSN_BITSIZE (insn); |
| 379 | |
| 380 | parse_errmsg = CGEN_PARSE_FN (cd, insn) (cd, insn, & str, fields); |
| 381 | if (parse_errmsg != NULL) |
| 382 | continue; |
| 383 | |
| 384 | /* ??? 0 is passed for `pc'. */ |
| 385 | insert_errmsg = CGEN_INSERT_FN (cd, insn) (cd, insn, fields, buf, |
| 386 | (bfd_vma) 0); |
| 387 | if (insert_errmsg != NULL) |
| 388 | continue; |
| 389 | |
| 390 | /* It is up to the caller to actually output the insn and any |
| 391 | queued relocs. */ |
| 392 | return insn; |
| 393 | } |
| 394 | |
| 395 | { |
| 396 | static char errbuf[150]; |
| 397 | const char *tmp_errmsg; |
| 398 | #ifdef CGEN_VERBOSE_ASSEMBLER_ERRORS |
| 399 | #define be_verbose 1 |
| 400 | #else |
| 401 | #define be_verbose 0 |
| 402 | #endif |
| 403 | |
| 404 | if (be_verbose) |
| 405 | { |
| 406 | /* If requesting verbose error messages, use insert_errmsg. |
| 407 | Failing that, use parse_errmsg. */ |
| 408 | tmp_errmsg = (insert_errmsg ? insert_errmsg : |
| 409 | parse_errmsg ? parse_errmsg : |
| 410 | recognized_mnemonic ? |
| 411 | _("unrecognized form of instruction") : |
| 412 | _("unrecognized instruction")); |
| 413 | |
| 414 | if (strlen (start) > 50) |
| 415 | /* xgettext:c-format */ |
| 416 | sprintf (errbuf, "%s `%.50s...'", tmp_errmsg, start); |
| 417 | else |
| 418 | /* xgettext:c-format */ |
| 419 | sprintf (errbuf, "%s `%.50s'", tmp_errmsg, start); |
| 420 | } |
| 421 | else |
| 422 | { |
| 423 | if (strlen (start) > 50) |
| 424 | /* xgettext:c-format */ |
| 425 | sprintf (errbuf, _("bad instruction `%.50s...'"), start); |
| 426 | else |
| 427 | /* xgettext:c-format */ |
| 428 | sprintf (errbuf, _("bad instruction `%.50s'"), start); |
| 429 | } |
| 430 | |
| 431 | *errmsg = errbuf; |
| 432 | return NULL; |
| 433 | } |
| 434 | } |