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) 1996-2014 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 (at) -desc.h" 35 #include "@prefix (at) -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 49 /* -- assembler routines inserted here. */ 51 52 54 /* Regex construction routine. 55 56 This translates an opcode syntax string into a regex string, 57 by replacing any non-character syntax element (such as an 58 opcode) with the pattern '.*' 59 60 It then compiles the regex and stores it in the opcode, for 61 later use by @arch@_cgen_assemble_insn 62 63 Returns NULL for success, an error message for failure. */ 64 65 char * 66 @arch@_cgen_build_insn_regex (CGEN_INSN *insn) 67 { 68 CGEN_OPCODE *opc = (CGEN_OPCODE *) CGEN_INSN_OPCODE (insn); 69 const char *mnem = CGEN_INSN_MNEMONIC (insn); 70 char rxbuf[CGEN_MAX_RX_ELEMENTS]; 71 char *rx = rxbuf; 72 const CGEN_SYNTAX_CHAR_TYPE *syn; 73 int reg_err; 74 75 syn = CGEN_SYNTAX_STRING (CGEN_OPCODE_SYNTAX (opc)); 76 77 /* Mnemonics come first in the syntax string. */ 78 if (! CGEN_SYNTAX_MNEMONIC_P (* syn)) 79 return _("missing mnemonic in syntax string"); 80 ++syn; 81 82 /* Generate a case sensitive regular expression that emulates case 83 insensitive matching in the "C" locale. We cannot generate a case 84 insensitive regular expression because in Turkish locales, 'i' and 'I' 85 are not equal modulo case conversion. */ 86 87 /* Copy the literal mnemonic out of the insn. */ 88 for (; *mnem; mnem++) 89 { 90 char c = *mnem; 91 92 if (ISALPHA (c)) 93 { 94 *rx++ = '['; 95 *rx++ = TOLOWER (c); 96 *rx++ = TOUPPER (c); 97 *rx++ = ']'; 98 } 99 else 100 *rx++ = c; 101 } 102 103 /* Copy any remaining literals from the syntax string into the rx. */ 104 for(; * syn != 0 && rx <= rxbuf + (CGEN_MAX_RX_ELEMENTS - 7 - 4); ++syn) 105 { 106 if (CGEN_SYNTAX_CHAR_P (* syn)) 107 { 108 char c = CGEN_SYNTAX_CHAR (* syn); 109 110 switch (c) 111 { 112 /* Escape any regex metacharacters in the syntax. */ 113 case '.': case '[': case '\\': 114 case '*': case '^': case '$': 115 116 #ifdef CGEN_ESCAPE_EXTENDED_REGEX 117 case '?': case '{': case '}': 118 case '(': case ')': case '*': 119 case '|': case '+': case ']': 120 #endif 121 *rx++ = '\\'; 122 *rx++ = c; 123 break; 124 125 default: 126 if (ISALPHA (c)) 127 { 128 *rx++ = '['; 129 *rx++ = TOLOWER (c); 130 *rx++ = TOUPPER (c); 131 *rx++ = ']'; 132 } 133 else 134 *rx++ = c; 135 break; 136 } 137 } 138 else 139 { 140 /* Replace non-syntax fields with globs. */ 141 *rx++ = '.'; 142 *rx++ = '*'; 143 } 144 } 145 146 /* Trailing whitespace ok. */ 147 * rx++ = '['; 148 * rx++ = ' '; 149 * rx++ = '\t'; 150 * rx++ = ']'; 151 * rx++ = '*'; 152 153 /* But anchor it after that. */ 154 * rx++ = '$'; 155 * rx = '\0'; 156 157 CGEN_INSN_RX (insn) = xmalloc (sizeof (regex_t)); 158 reg_err = regcomp ((regex_t *) CGEN_INSN_RX (insn), rxbuf, REG_NOSUB); 159 160 if (reg_err == 0) 161 return NULL; 162 else 163 { 164 static char msg[80]; 165 166 regerror (reg_err, (regex_t *) CGEN_INSN_RX (insn), msg, 80); 167 regfree ((regex_t *) CGEN_INSN_RX (insn)); 168 free (CGEN_INSN_RX (insn)); 169 (CGEN_INSN_RX (insn)) = NULL; 170 return msg; 171 } 172 } 173 174 175 /* Default insn parser. 177 178 The syntax string is scanned and operands are parsed and stored in FIELDS. 179 Relocs are queued as we go via other callbacks. 180 181 ??? Note that this is currently an all-or-nothing parser. If we fail to 182 parse the instruction, we return 0 and the caller will start over from 183 the beginning. Backtracking will be necessary in parsing subexpressions, 184 but that can be handled there. Not handling backtracking here may get 185 expensive in the case of the m68k. Deal with later. 186 187 Returns NULL for success, an error message for failure. */ 188 189 static const char * 190 parse_insn_normal (CGEN_CPU_DESC cd, 191 const CGEN_INSN *insn, 192 const char **strp, 193 CGEN_FIELDS *fields) 194 { 195 /* ??? Runtime added insns not handled yet. */ 196 const CGEN_SYNTAX *syntax = CGEN_INSN_SYNTAX (insn); 197 const char *str = *strp; 198 const char *errmsg; 199 const char *p; 200 const CGEN_SYNTAX_CHAR_TYPE * syn; 201 #ifdef CGEN_MNEMONIC_OPERANDS 202 /* FIXME: wip */ 203 int past_opcode_p; 204 #endif 205 206 /* For now we assume the mnemonic is first (there are no leading operands). 207 We can parse it without needing to set up operand parsing. 208 GAS's input scrubber will ensure mnemonics are lowercase, but we may 209 not be called from GAS. */ 210 p = CGEN_INSN_MNEMONIC (insn); 211 while (*p && TOLOWER (*p) == TOLOWER (*str)) 212 ++p, ++str; 213 214 if (* p) 215 return _("unrecognized instruction"); 216 217 #ifndef CGEN_MNEMONIC_OPERANDS 218 if (* str && ! ISSPACE (* str)) 219 return _("unrecognized instruction"); 220 #endif 221 222 CGEN_INIT_PARSE (cd); 223 cgen_init_parse_operand (cd); 224 #ifdef CGEN_MNEMONIC_OPERANDS 225 past_opcode_p = 0; 226 #endif 227 228 /* We don't check for (*str != '\0') here because we want to parse 229 any trailing fake arguments in the syntax string. */ 230 syn = CGEN_SYNTAX_STRING (syntax); 231 232 /* Mnemonics come first for now, ensure valid string. */ 233 if (! CGEN_SYNTAX_MNEMONIC_P (* syn)) 234 abort (); 235 236 ++syn; 237 238 while (* syn != 0) 239 { 240 /* Non operand chars must match exactly. */ 241 if (CGEN_SYNTAX_CHAR_P (* syn)) 242 { 243 /* FIXME: While we allow for non-GAS callers above, we assume the 244 first char after the mnemonic part is a space. */ 245 /* FIXME: We also take inappropriate advantage of the fact that 246 GAS's input scrubber will remove extraneous blanks. */ 247 if (TOLOWER (*str) == TOLOWER (CGEN_SYNTAX_CHAR (* syn))) 248 { 249 #ifdef CGEN_MNEMONIC_OPERANDS 250 if (CGEN_SYNTAX_CHAR(* syn) == ' ') 251 past_opcode_p = 1; 252 #endif 253 ++ syn; 254 ++ str; 255 } 256 else if (*str) 257 { 258 /* Syntax char didn't match. Can't be this insn. */ 259 static char msg [80]; 260 261 /* xgettext:c-format */ 262 sprintf (msg, _("syntax error (expected char `%c', found `%c')"), 263 CGEN_SYNTAX_CHAR(*syn), *str); 264 return msg; 265 } 266 else 267 { 268 /* Ran out of input. */ 269 static char msg [80]; 270 271 /* xgettext:c-format */ 272 sprintf (msg, _("syntax error (expected char `%c', found end of instruction)"), 273 CGEN_SYNTAX_CHAR(*syn)); 274 return msg; 275 } 276 continue; 277 } 278 279 #ifdef CGEN_MNEMONIC_OPERANDS 280 (void) past_opcode_p; 281 #endif 282 /* We have an operand of some sort. */ 283 errmsg = cd->parse_operand (cd, CGEN_SYNTAX_FIELD (*syn), &str, fields); 284 if (errmsg) 285 return errmsg; 286 287 /* Done with this operand, continue with next one. */ 288 ++ syn; 289 } 290 291 /* If we're at the end of the syntax string, we're done. */ 292 if (* syn == 0) 293 { 294 /* FIXME: For the moment we assume a valid `str' can only contain 295 blanks now. IE: We needn't try again with a longer version of 296 the insn and it is assumed that longer versions of insns appear 297 before shorter ones (eg: lsr r2,r3,1 vs lsr r2,r3). */ 298 while (ISSPACE (* str)) 299 ++ str; 300 301 if (* str != '\0') 302 return _("junk at end of line"); /* FIXME: would like to include `str' */ 303 304 return NULL; 305 } 306 307 /* We couldn't parse it. */ 308 return _("unrecognized instruction"); 309 } 310 311 /* Main entry point. 313 This routine is called for each instruction to be assembled. 314 STR points to the insn to be assembled. 315 We assume all necessary tables have been initialized. 316 The assembled instruction, less any fixups, is stored in BUF. 317 Remember that if CGEN_INT_INSN_P then BUF is an int and thus the value 318 still needs to be converted to target byte order, otherwise BUF is an array 319 of bytes in target byte order. 320 The result is a pointer to the insn's entry in the opcode table, 321 or NULL if an error occured (an error message will have already been 322 printed). 323 324 Note that when processing (non-alias) macro-insns, 325 this function recurses. 326 327 ??? It's possible to make this cpu-independent. 328 One would have to deal with a few minor things. 329 At this point in time doing so would be more of a curiosity than useful 330 [for example this file isn't _that_ big], but keeping the possibility in 331 mind helps keep the design clean. */ 332 333 const CGEN_INSN * 334 @arch@_cgen_assemble_insn (CGEN_CPU_DESC cd, 335 const char *str, 336 CGEN_FIELDS *fields, 337 CGEN_INSN_BYTES_PTR buf, 338 char **errmsg) 339 { 340 const char *start; 341 CGEN_INSN_LIST *ilist; 342 const char *parse_errmsg = NULL; 343 const char *insert_errmsg = NULL; 344 int recognized_mnemonic = 0; 345 346 /* Skip leading white space. */ 347 while (ISSPACE (* str)) 348 ++ str; 349 350 /* The instructions are stored in hashed lists. 351 Get the first in the list. */ 352 ilist = CGEN_ASM_LOOKUP_INSN (cd, str); 353 354 /* Keep looking until we find a match. */ 355 start = str; 356 for ( ; ilist != NULL ; ilist = CGEN_ASM_NEXT_INSN (ilist)) 357 { 358 const CGEN_INSN *insn = ilist->insn; 359 recognized_mnemonic = 1; 360 361 #ifdef CGEN_VALIDATE_INSN_SUPPORTED 362 /* Not usually needed as unsupported opcodes 363 shouldn't be in the hash lists. */ 364 /* Is this insn supported by the selected cpu? */ 365 if (! @arch@_cgen_insn_supported (cd, insn)) 366 continue; 367 #endif 368 /* If the RELAXED attribute is set, this is an insn that shouldn't be 369 chosen immediately. Instead, it is used during assembler/linker 370 relaxation if possible. */ 371 if (CGEN_INSN_ATTR_VALUE (insn, CGEN_INSN_RELAXED) != 0) 372 continue; 373 374 str = start; 375 376 /* Skip this insn if str doesn't look right lexically. */ 377 if (CGEN_INSN_RX (insn) != NULL && 378 regexec ((regex_t *) CGEN_INSN_RX (insn), str, 0, NULL, 0) == REG_NOMATCH) 379 continue; 380 381 /* Allow parse/insert handlers to obtain length of insn. */ 382 CGEN_FIELDS_BITSIZE (fields) = CGEN_INSN_BITSIZE (insn); 383 384 parse_errmsg = CGEN_PARSE_FN (cd, insn) (cd, insn, & str, fields); 385 if (parse_errmsg != NULL) 386 continue; 387 388 /* ??? 0 is passed for `pc'. */ 389 insert_errmsg = CGEN_INSERT_FN (cd, insn) (cd, insn, fields, buf, 390 (bfd_vma) 0); 391 if (insert_errmsg != NULL) 392 continue; 393 394 /* It is up to the caller to actually output the insn and any 395 queued relocs. */ 396 return insn; 397 } 398 399 { 400 static char errbuf[150]; 401 const char *tmp_errmsg; 402 #ifdef CGEN_VERBOSE_ASSEMBLER_ERRORS 403 #define be_verbose 1 404 #else 405 #define be_verbose 0 406 #endif 407 408 if (be_verbose) 409 { 410 /* If requesting verbose error messages, use insert_errmsg. 411 Failing that, use parse_errmsg. */ 412 tmp_errmsg = (insert_errmsg ? insert_errmsg : 413 parse_errmsg ? parse_errmsg : 414 recognized_mnemonic ? 415 _("unrecognized form of instruction") : 416 _("unrecognized instruction")); 417 418 if (strlen (start) > 50) 419 /* xgettext:c-format */ 420 sprintf (errbuf, "%s `%.50s...'", tmp_errmsg, start); 421 else 422 /* xgettext:c-format */ 423 sprintf (errbuf, "%s `%.50s'", tmp_errmsg, start); 424 } 425 else 426 { 427 if (strlen (start) > 50) 428 /* xgettext:c-format */ 429 sprintf (errbuf, _("bad instruction `%.50s...'"), start); 430 else 431 /* xgettext:c-format */ 432 sprintf (errbuf, _("bad instruction `%.50s'"), start); 433 } 434 435 *errmsg = errbuf; 436 return NULL; 437 } 438 } 439