1 %{ 2 /* 3 * CDDL HEADER START 4 * 5 * The contents of this file are subject to the terms of the 6 * Common Development and Distribution License (the "License"). 7 * You may not use this file except in compliance with the License. 8 * 9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 10 * or http://www.opensolaris.org/os/licensing. 11 * See the License for the specific language governing permissions 12 * and limitations under the License. 13 * 14 * When distributing Covered Code, include this CDDL HEADER in each 15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 16 * If applicable, add the following below this CDDL HEADER, with the 17 * fields enclosed by brackets "[]" replaced with your own identifying 18 * information: Portions Copyright [yyyy] [name of copyright owner] 19 * 20 * CDDL HEADER END 21 */ 22 23 /* 24 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved. 25 */ 26 27 #include <string.h> 28 #include <stdlib.h> 29 #include <stdio.h> 30 #include <assert.h> 31 #include <ctype.h> 32 #include <errno.h> 33 34 #include <dt_impl.h> 35 #include <dt_grammar.h> 36 #include <dt_parser.h> 37 #include <dt_string.h> 38 39 /* 40 * We need to undefine lex's input and unput macros so that references to these 41 * call the functions provided at the end of this source file. 42 */ 43 #if defined(sun) 44 #undef input 45 #undef unput 46 #else 47 /* 48 * Define YY_INPUT for flex since input() can't be re-defined. 49 */ 50 #define YY_INPUT(buf,result,max_size) \ 51 if (yypcb->pcb_fileptr != NULL) { \ 52 if (((result = fread(buf, 1, max_size, yypcb->pcb_fileptr)) == 0) \ 53 && ferror(yypcb->pcb_fileptr)) \ 54 longjmp(yypcb->pcb_jmpbuf, EDT_FIO); \ 55 } else { \ 56 int n; \ 57 for (n = 0; n < max_size && \ 58 yypcb->pcb_strptr < yypcb->pcb_string + yypcb->pcb_strlen; n++) \ 59 buf[n] = *yypcb->pcb_strptr++; \ 60 result = n; \ 61 } 62 /* 63 * Do not EOF let tokens to be put back. This does not work with flex. 64 * On the other hand, leaving current buffer in same state it was when 65 * last EOF was received guarantees that input() will keep returning EOF 66 * for all subsequent invocations, which is the effect desired. 67 */ 68 #undef unput 69 #define unput(c) \ 70 do { \ 71 int _c = c; \ 72 if (_c != EOF) \ 73 yyunput(_c, yytext_ptr); \ 74 } while(0) 75 #endif 76 77 static int id_or_type(const char *); 78 #if defined(sun) 79 static int input(void); 80 static void unput(int); 81 #endif 82 83 /* 84 * We first define a set of labeled states for use in the D lexer and then a 85 * set of regular expressions to simplify things below. The lexer states are: 86 * 87 * S0 - D program clause and expression lexing 88 * S1 - D comments (i.e. skip everything until end of comment) 89 * S2 - D program outer scope (probe specifiers and declarations) 90 * S3 - D control line parsing (i.e. after ^# is seen but before \n) 91 * S4 - D control line scan (locate control directives only and invoke S3) 92 */ 93 %} 94 95 %e 1500 /* maximum nodes */ 96 %p 3700 /* maximum positions */ 97 %n 600 /* maximum states */ 98 99 %s S0 S1 S2 S3 S4 100 101 RGX_AGG "@"[a-zA-Z_][0-9a-zA-Z_]* 102 RGX_PSPEC [-$:a-zA-Z_.?*\\\[\]!][-$:0-9a-zA-Z_.`?*\\\[\]!]* 103 RGX_IDENT [a-zA-Z_`][0-9a-zA-Z_`]* 104 RGX_INT ([0-9]+|0[xX][0-9A-Fa-f]+)[uU]?[lL]?[lL]? 105 RGX_FP ([0-9]+("."?)[0-9]*|"."[0-9]+)((e|E)("+"|-)?[0-9]+)?[fFlL]? 106 RGX_WS [\f\n\r\t\v ] 107 RGX_STR ([^"\\\n]|\\[^"\n]|\\\")* 108 RGX_CHR ([^'\\\n]|\\[^'\n]|\\')* 109 RGX_INTERP ^[\f\t\v ]*#!.* 110 RGX_CTL ^[\f\t\v ]*# 111 112 %% 113 114 %{ 115 116 /* 117 * We insert a special prologue into yylex() itself: if the pcb contains a 118 * context token, we return that prior to running the normal lexer. This 119 * allows libdtrace to force yacc into one of our three parsing contexts: D 120 * expression (DT_CTX_DEXPR), D program (DT_CTX_DPROG) or D type (DT_CTX_DTYPE). 121 * Once the token is returned, we clear it so this only happens once. 122 */ 123 if (yypcb->pcb_token != 0) { 124 int tok = yypcb->pcb_token; 125 yypcb->pcb_token = 0; 126 return (tok); 127 } 128 129 %} 130 131 <S0>auto return (DT_KEY_AUTO); 132 <S0>break return (DT_KEY_BREAK); 133 <S0>case return (DT_KEY_CASE); 134 <S0>char return (DT_KEY_CHAR); 135 <S0>const return (DT_KEY_CONST); 136 <S0>continue return (DT_KEY_CONTINUE); 137 <S0>counter return (DT_KEY_COUNTER); 138 <S0>default return (DT_KEY_DEFAULT); 139 <S0>do return (DT_KEY_DO); 140 <S0>double return (DT_KEY_DOUBLE); 141 <S0>else return (DT_KEY_ELSE); 142 <S0>enum return (DT_KEY_ENUM); 143 <S0>extern return (DT_KEY_EXTERN); 144 <S0>float return (DT_KEY_FLOAT); 145 <S0>for return (DT_KEY_FOR); 146 <S0>goto return (DT_KEY_GOTO); 147 <S0>if return (DT_KEY_IF); 148 <S0>import return (DT_KEY_IMPORT); 149 <S0>inline return (DT_KEY_INLINE); 150 <S0>int return (DT_KEY_INT); 151 <S0>long return (DT_KEY_LONG); 152 <S0>offsetof return (DT_TOK_OFFSETOF); 153 <S0>probe return (DT_KEY_PROBE); 154 <S0>provider return (DT_KEY_PROVIDER); 155 <S0>register return (DT_KEY_REGISTER); 156 <S0>restrict return (DT_KEY_RESTRICT); 157 <S0>return return (DT_KEY_RETURN); 158 <S0>self return (DT_KEY_SELF); 159 <S0>short return (DT_KEY_SHORT); 160 <S0>signed return (DT_KEY_SIGNED); 161 <S0>sizeof return (DT_TOK_SIZEOF); 162 <S0>static return (DT_KEY_STATIC); 163 <S0>string return (DT_KEY_STRING); 164 <S0>stringof return (DT_TOK_STRINGOF); 165 <S0>struct return (DT_KEY_STRUCT); 166 <S0>switch return (DT_KEY_SWITCH); 167 <S0>this return (DT_KEY_THIS); 168 <S0>translator return (DT_KEY_XLATOR); 169 <S0>typedef return (DT_KEY_TYPEDEF); 170 <S0>union return (DT_KEY_UNION); 171 <S0>unsigned return (DT_KEY_UNSIGNED); 172 <S0>void return (DT_KEY_VOID); 173 <S0>volatile return (DT_KEY_VOLATILE); 174 <S0>while return (DT_KEY_WHILE); 175 <S0>xlate return (DT_TOK_XLATE); 176 177 <S2>auto { yybegin(YYS_EXPR); return (DT_KEY_AUTO); } 178 <S2>char { yybegin(YYS_EXPR); return (DT_KEY_CHAR); } 179 <S2>const { yybegin(YYS_EXPR); return (DT_KEY_CONST); } 180 <S2>counter { yybegin(YYS_DEFINE); return (DT_KEY_COUNTER); } 181 <S2>double { yybegin(YYS_EXPR); return (DT_KEY_DOUBLE); } 182 <S2>enum { yybegin(YYS_EXPR); return (DT_KEY_ENUM); } 183 <S2>extern { yybegin(YYS_EXPR); return (DT_KEY_EXTERN); } 184 <S2>float { yybegin(YYS_EXPR); return (DT_KEY_FLOAT); } 185 <S2>import { yybegin(YYS_EXPR); return (DT_KEY_IMPORT); } 186 <S2>inline { yybegin(YYS_DEFINE); return (DT_KEY_INLINE); } 187 <S2>int { yybegin(YYS_EXPR); return (DT_KEY_INT); } 188 <S2>long { yybegin(YYS_EXPR); return (DT_KEY_LONG); } 189 <S2>provider { yybegin(YYS_DEFINE); return (DT_KEY_PROVIDER); } 190 <S2>register { yybegin(YYS_EXPR); return (DT_KEY_REGISTER); } 191 <S2>restrict { yybegin(YYS_EXPR); return (DT_KEY_RESTRICT); } 192 <S2>self { yybegin(YYS_EXPR); return (DT_KEY_SELF); } 193 <S2>short { yybegin(YYS_EXPR); return (DT_KEY_SHORT); } 194 <S2>signed { yybegin(YYS_EXPR); return (DT_KEY_SIGNED); } 195 <S2>static { yybegin(YYS_EXPR); return (DT_KEY_STATIC); } 196 <S2>string { yybegin(YYS_EXPR); return (DT_KEY_STRING); } 197 <S2>struct { yybegin(YYS_EXPR); return (DT_KEY_STRUCT); } 198 <S2>this { yybegin(YYS_EXPR); return (DT_KEY_THIS); } 199 <S2>translator { yybegin(YYS_DEFINE); return (DT_KEY_XLATOR); } 200 <S2>typedef { yybegin(YYS_EXPR); return (DT_KEY_TYPEDEF); } 201 <S2>union { yybegin(YYS_EXPR); return (DT_KEY_UNION); } 202 <S2>unsigned { yybegin(YYS_EXPR); return (DT_KEY_UNSIGNED); } 203 <S2>void { yybegin(YYS_EXPR); return (DT_KEY_VOID); } 204 <S2>volatile { yybegin(YYS_EXPR); return (DT_KEY_VOLATILE); } 205 206 <S0>"$$"[0-9]+ { 207 int i = atoi(yytext + 2); 208 char *v = ""; 209 210 /* 211 * A macro argument reference substitutes the text of 212 * an argument in place of the current token. When we 213 * see $$<d> we fetch the saved string from pcb_sargv 214 * (or use the default argument if the option has been 215 * set and the argument hasn't been specified) and 216 * return a token corresponding to this string. 217 */ 218 if (i < 0 || (i >= yypcb->pcb_sargc && 219 !(yypcb->pcb_cflags & DTRACE_C_DEFARG))) { 220 xyerror(D_MACRO_UNDEF, "macro argument %s is " 221 "not defined\n", yytext); 222 } 223 224 if (i < yypcb->pcb_sargc) { 225 v = yypcb->pcb_sargv[i]; /* get val from pcb */ 226 yypcb->pcb_sflagv[i] |= DT_IDFLG_REF; 227 } 228 229 if ((yylval.l_str = strdup(v)) == NULL) 230 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 231 232 (void) stresc2chr(yylval.l_str); 233 return (DT_TOK_STRING); 234 } 235 236 <S0>"$"[0-9]+ { 237 int i = atoi(yytext + 1); 238 char *p, *v = "0"; 239 240 /* 241 * A macro argument reference substitutes the text of 242 * one identifier or integer pattern for another. When 243 * we see $<d> we fetch the saved string from pcb_sargv 244 * (or use the default argument if the option has been 245 * set and the argument hasn't been specified) and 246 * return a token corresponding to this string. 247 */ 248 if (i < 0 || (i >= yypcb->pcb_sargc && 249 !(yypcb->pcb_cflags & DTRACE_C_DEFARG))) { 250 xyerror(D_MACRO_UNDEF, "macro argument %s is " 251 "not defined\n", yytext); 252 } 253 254 if (i < yypcb->pcb_sargc) { 255 v = yypcb->pcb_sargv[i]; /* get val from pcb */ 256 yypcb->pcb_sflagv[i] |= DT_IDFLG_REF; 257 } 258 259 /* 260 * If the macro text is not a valid integer or ident, 261 * then we treat it as a string. The string may be 262 * optionally enclosed in quotes, which we strip. 263 */ 264 if (strbadidnum(v)) { 265 size_t len = strlen(v); 266 267 if (len != 1 && *v == '"' && v[len - 1] == '"') 268 yylval.l_str = strndup(v + 1, len - 2); 269 else 270 yylval.l_str = strndup(v, len); 271 272 if (yylval.l_str == NULL) 273 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 274 275 (void) stresc2chr(yylval.l_str); 276 return (DT_TOK_STRING); 277 } 278 279 /* 280 * If the macro text is not a string an begins with a 281 * digit or a +/- sign, process it as an integer token. 282 */ 283 if (isdigit(v[0]) || v[0] == '-' || v[0] == '+') { 284 if (isdigit(v[0])) 285 yyintprefix = 0; 286 else 287 yyintprefix = *v++; 288 289 errno = 0; 290 yylval.l_int = strtoull(v, &p, 0); 291 (void) strncpy(yyintsuffix, p, 292 sizeof (yyintsuffix)); 293 yyintdecimal = *v != '0'; 294 295 if (errno == ERANGE) { 296 xyerror(D_MACRO_OFLOW, "macro argument" 297 " %s constant %s results in integer" 298 " overflow\n", yytext, v); 299 } 300 301 return (DT_TOK_INT); 302 } 303 304 return (id_or_type(v)); 305 } 306 307 <S0>"$$"{RGX_IDENT} { 308 dt_ident_t *idp = dt_idhash_lookup( 309 yypcb->pcb_hdl->dt_macros, yytext + 2); 310 311 char s[16]; /* enough for UINT_MAX + \0 */ 312 313 if (idp == NULL) { 314 xyerror(D_MACRO_UNDEF, "macro variable %s " 315 "is not defined\n", yytext); 316 } 317 318 /* 319 * For the moment, all current macro variables are of 320 * type id_t (refer to dtrace_update() for details). 321 */ 322 (void) snprintf(s, sizeof (s), "%u", idp->di_id); 323 if ((yylval.l_str = strdup(s)) == NULL) 324 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 325 326 return (DT_TOK_STRING); 327 } 328 329 <S0>"$"{RGX_IDENT} { 330 dt_ident_t *idp = dt_idhash_lookup( 331 yypcb->pcb_hdl->dt_macros, yytext + 1); 332 333 if (idp == NULL) { 334 xyerror(D_MACRO_UNDEF, "macro variable %s " 335 "is not defined\n", yytext); 336 } 337 338 /* 339 * For the moment, all current macro variables are of 340 * type id_t (refer to dtrace_update() for details). 341 */ 342 yylval.l_int = (intmax_t)(int)idp->di_id; 343 yyintprefix = 0; 344 yyintsuffix[0] = '\0'; 345 yyintdecimal = 1; 346 347 return (DT_TOK_INT); 348 } 349 350 <S0>{RGX_IDENT} { 351 return (id_or_type(yytext)); 352 } 353 354 <S0>{RGX_AGG} { 355 if ((yylval.l_str = strdup(yytext)) == NULL) 356 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 357 return (DT_TOK_AGG); 358 } 359 360 <S0>"@" { 361 if ((yylval.l_str = strdup("@_")) == NULL) 362 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 363 return (DT_TOK_AGG); 364 } 365 366 <S0>{RGX_INT} | 367 <S2>{RGX_INT} | 368 <S3>{RGX_INT} { 369 char *p; 370 371 errno = 0; 372 yylval.l_int = strtoull(yytext, &p, 0); 373 yyintprefix = 0; 374 (void) strncpy(yyintsuffix, p, sizeof (yyintsuffix)); 375 yyintdecimal = yytext[0] != '0'; 376 377 if (errno == ERANGE) { 378 xyerror(D_INT_OFLOW, "constant %s results in " 379 "integer overflow\n", yytext); 380 } 381 382 if (*p != '\0' && strchr("uUlL", *p) == NULL) { 383 xyerror(D_INT_DIGIT, "constant %s contains " 384 "invalid digit %c\n", yytext, *p); 385 } 386 387 if ((YYSTATE) != S3) 388 return (DT_TOK_INT); 389 390 yypragma = dt_node_link(yypragma, 391 dt_node_int(yylval.l_int)); 392 } 393 394 <S0>{RGX_FP} yyerror("floating-point constants are not permitted\n"); 395 396 <S0>\"{RGX_STR}$ | 397 <S3>\"{RGX_STR}$ xyerror(D_STR_NL, "newline encountered in string literal"); 398 399 <S0>\"{RGX_STR}\" | 400 <S3>\"{RGX_STR}\" { 401 /* 402 * Quoted string -- convert C escape sequences and 403 * return the string as a token. 404 */ 405 yylval.l_str = strndup(yytext + 1, yyleng - 2); 406 407 if (yylval.l_str == NULL) 408 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 409 410 (void) stresc2chr(yylval.l_str); 411 if ((YYSTATE) != S3) 412 return (DT_TOK_STRING); 413 414 yypragma = dt_node_link(yypragma, 415 dt_node_string(yylval.l_str)); 416 } 417 418 <S0>'{RGX_CHR}$ xyerror(D_CHR_NL, "newline encountered in character constant"); 419 420 <S0>'{RGX_CHR}' { 421 char *s, *p, *q; 422 size_t nbytes; 423 424 /* 425 * Character constant -- convert C escape sequences and 426 * return the character as an integer immediate value. 427 */ 428 if (yyleng == 2) 429 xyerror(D_CHR_NULL, "empty character constant"); 430 431 s = yytext + 1; 432 yytext[yyleng - 1] = '\0'; 433 nbytes = stresc2chr(s); 434 yylval.l_int = 0; 435 yyintprefix = 0; 436 yyintsuffix[0] = '\0'; 437 yyintdecimal = 1; 438 439 if (nbytes > sizeof (yylval.l_int)) { 440 xyerror(D_CHR_OFLOW, "character constant is " 441 "too long"); 442 } 443 #if BYTE_ORDER == _LITTLE_ENDIAN 444 p = ((char *)&yylval.l_int) + nbytes - 1; 445 for (q = s; nbytes != 0; nbytes--) 446 *p-- = *q++; 447 #else 448 bcopy(s, ((char *)&yylval.l_int) + 449 sizeof (yylval.l_int) - nbytes, nbytes); 450 #endif 451 return (DT_TOK_INT); 452 } 453 454 <S0>"/*" | 455 <S2>"/*" { 456 yypcb->pcb_cstate = (YYSTATE); 457 BEGIN(S1); 458 } 459 460 <S0>{RGX_INTERP} | 461 <S2>{RGX_INTERP} ; /* discard any #! lines */ 462 463 <S0>{RGX_CTL} | 464 <S2>{RGX_CTL} | 465 <S4>{RGX_CTL} { 466 assert(yypragma == NULL); 467 yypcb->pcb_cstate = (YYSTATE); 468 BEGIN(S3); 469 } 470 471 <S4>. ; /* discard */ 472 <S4>"\n" ; /* discard */ 473 474 <S0>"/" { 475 int c, tok; 476 477 /* 478 * The use of "/" as the predicate delimiter and as the 479 * integer division symbol requires special lookahead 480 * to avoid a shift/reduce conflict in the D grammar. 481 * We look ahead to the next non-whitespace character. 482 * If we encounter EOF, ";", "{", or "/", then this "/" 483 * closes the predicate and we return DT_TOK_EPRED. 484 * If we encounter anything else, it's DT_TOK_DIV. 485 */ 486 while ((c = input()) != 0) { 487 if (strchr("\f\n\r\t\v ", c) == NULL) 488 break; 489 } 490 491 if (c == 0 || c == ';' || c == '{' || c == '/') { 492 if (yypcb->pcb_parens != 0) { 493 yyerror("closing ) expected in " 494 "predicate before /\n"); 495 } 496 if (yypcb->pcb_brackets != 0) { 497 yyerror("closing ] expected in " 498 "predicate before /\n"); 499 } 500 tok = DT_TOK_EPRED; 501 } else 502 tok = DT_TOK_DIV; 503 504 unput(c); 505 return (tok); 506 } 507 508 <S0>"(" { 509 yypcb->pcb_parens++; 510 return (DT_TOK_LPAR); 511 } 512 513 <S0>")" { 514 if (--yypcb->pcb_parens < 0) 515 yyerror("extra ) in input stream\n"); 516 return (DT_TOK_RPAR); 517 } 518 519 <S0>"[" { 520 yypcb->pcb_brackets++; 521 return (DT_TOK_LBRAC); 522 } 523 524 <S0>"]" { 525 if (--yypcb->pcb_brackets < 0) 526 yyerror("extra ] in input stream\n"); 527 return (DT_TOK_RBRAC); 528 } 529 530 <S0>"{" | 531 <S2>"{" { 532 yypcb->pcb_braces++; 533 return ('{'); 534 } 535 536 <S0>"}" { 537 if (--yypcb->pcb_braces < 0) 538 yyerror("extra } in input stream\n"); 539 return ('}'); 540 } 541 542 <S0>"|" return (DT_TOK_BOR); 543 <S0>"^" return (DT_TOK_XOR); 544 <S0>"&" return (DT_TOK_BAND); 545 <S0>"&&" return (DT_TOK_LAND); 546 <S0>"^^" return (DT_TOK_LXOR); 547 <S0>"||" return (DT_TOK_LOR); 548 <S0>"==" return (DT_TOK_EQU); 549 <S0>"!=" return (DT_TOK_NEQ); 550 <S0>"<" return (DT_TOK_LT); 551 <S0>"<=" return (DT_TOK_LE); 552 <S0>">" return (DT_TOK_GT); 553 <S0>">=" return (DT_TOK_GE); 554 <S0>"<<" return (DT_TOK_LSH); 555 <S0>">>" return (DT_TOK_RSH); 556 <S0>"+" return (DT_TOK_ADD); 557 <S0>"-" return (DT_TOK_SUB); 558 <S0>"*" return (DT_TOK_MUL); 559 <S0>"%" return (DT_TOK_MOD); 560 <S0>"~" return (DT_TOK_BNEG); 561 <S0>"!" return (DT_TOK_LNEG); 562 <S0>"?" return (DT_TOK_QUESTION); 563 <S0>":" return (DT_TOK_COLON); 564 <S0>"." return (DT_TOK_DOT); 565 <S0>"->" return (DT_TOK_PTR); 566 <S0>"=" return (DT_TOK_ASGN); 567 <S0>"+=" return (DT_TOK_ADD_EQ); 568 <S0>"-=" return (DT_TOK_SUB_EQ); 569 <S0>"*=" return (DT_TOK_MUL_EQ); 570 <S0>"/=" return (DT_TOK_DIV_EQ); 571 <S0>"%=" return (DT_TOK_MOD_EQ); 572 <S0>"&=" return (DT_TOK_AND_EQ); 573 <S0>"^=" return (DT_TOK_XOR_EQ); 574 <S0>"|=" return (DT_TOK_OR_EQ); 575 <S0>"<<=" return (DT_TOK_LSH_EQ); 576 <S0>">>=" return (DT_TOK_RSH_EQ); 577 <S0>"++" return (DT_TOK_ADDADD); 578 <S0>"--" return (DT_TOK_SUBSUB); 579 <S0>"..." return (DT_TOK_ELLIPSIS); 580 <S0>"," return (DT_TOK_COMMA); 581 <S0>";" return (';'); 582 <S0>{RGX_WS} ; /* discard */ 583 <S0>"\\"\n ; /* discard */ 584 <S0>. yyerror("syntax error near \"%c\"\n", yytext[0]); 585 586 <S1>"/*" yyerror("/* encountered inside a comment\n"); 587 <S1>"*/" BEGIN(yypcb->pcb_cstate); 588 <S1>.|\n ; /* discard */ 589 590 <S2>{RGX_PSPEC} { 591 /* 592 * S2 has an ambiguity because RGX_PSPEC includes '*' 593 * as a glob character and '*' also can be DT_TOK_STAR. 594 * Since lex always matches the longest token, this 595 * rule can be matched by an input string like "int*", 596 * which could begin a global variable declaration such 597 * as "int*x;" or could begin a RGX_PSPEC with globbing 598 * such as "int* { trace(timestamp); }". If C_PSPEC is 599 * not set, we must resolve the ambiguity in favor of 600 * the type and perform lexer pushback if the fragment 601 * before '*' or entire fragment matches a type name. 602 * If C_PSPEC is set, we always return a PSPEC token. 603 * If C_PSPEC is off, the user can avoid ambiguity by 604 * including a ':' delimiter in the specifier, which 605 * they should be doing anyway to specify the provider. 606 */ 607 if (!(yypcb->pcb_cflags & DTRACE_C_PSPEC) && 608 strchr(yytext, ':') == NULL) { 609 610 char *p = strchr(yytext, '*'); 611 char *q = yytext + yyleng - 1; 612 613 if (p != NULL && p > yytext) 614 *p = '\0'; /* prune yytext */ 615 616 if (dt_type_lookup(yytext, NULL) == 0) { 617 yylval.l_str = strdup(yytext); 618 619 if (yylval.l_str == NULL) { 620 longjmp(yypcb->pcb_jmpbuf, 621 EDT_NOMEM); 622 } 623 624 if (p != NULL && p > yytext) { 625 for (*p = '*'; q >= p; q--) 626 unput(*q); 627 } 628 629 yybegin(YYS_EXPR); 630 return (DT_TOK_TNAME); 631 } 632 633 if (p != NULL && p > yytext) 634 *p = '*'; /* restore yytext */ 635 } 636 637 if ((yylval.l_str = strdup(yytext)) == NULL) 638 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 639 640 return (DT_TOK_PSPEC); 641 } 642 643 <S2>"/" return (DT_TOK_DIV); 644 <S2>"," return (DT_TOK_COMMA); 645 646 <S2>{RGX_WS} ; /* discard */ 647 <S2>. yyerror("syntax error near \"%c\"\n", yytext[0]); 648 649 <S3>\n { 650 dt_pragma(yypragma); 651 yypragma = NULL; 652 BEGIN(yypcb->pcb_cstate); 653 } 654 655 <S3>[\f\t\v ]+ ; /* discard */ 656 657 <S3>[^\f\n\t\v "]+ { 658 dt_node_t *dnp; 659 660 if ((yylval.l_str = strdup(yytext)) == NULL) 661 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 662 663 /* 664 * We want to call dt_node_ident() here, but we can't 665 * because it will expand inlined identifiers, which we 666 * don't want to do from #pragma context in order to 667 * support pragmas that apply to the ident itself. We 668 * call dt_node_string() and then reset dn_op instead. 669 */ 670 dnp = dt_node_string(yylval.l_str); 671 dnp->dn_kind = DT_NODE_IDENT; 672 dnp->dn_op = DT_TOK_IDENT; 673 yypragma = dt_node_link(yypragma, dnp); 674 } 675 676 <S3>. yyerror("syntax error near \"%c\"\n", yytext[0]); 677 678 %% 679 680 /* 681 * yybegin provides a wrapper for use from C code around the lex BEGIN() macro. 682 * We use two main states for lexing because probe descriptions use a syntax 683 * that is incompatible with the normal D tokens (e.g. names can contain "-"). 684 * yybegin also handles the job of switching between two lists of dt_nodes 685 * as we allocate persistent definitions, like inlines, and transient nodes 686 * that will be freed once we are done parsing the current program file. 687 */ 688 void 689 yybegin(yystate_t state) 690 { 691 #ifdef YYDEBUG 692 yydebug = _dtrace_debug; 693 #endif 694 if (yypcb->pcb_yystate == state) 695 return; /* nothing to do if we're in the state already */ 696 697 if (yypcb->pcb_yystate == YYS_DEFINE) { 698 yypcb->pcb_list = yypcb->pcb_hold; 699 yypcb->pcb_hold = NULL; 700 } 701 702 switch (state) { 703 case YYS_CLAUSE: 704 BEGIN(S2); 705 break; 706 case YYS_DEFINE: 707 assert(yypcb->pcb_hold == NULL); 708 yypcb->pcb_hold = yypcb->pcb_list; 709 yypcb->pcb_list = NULL; 710 /*FALLTHRU*/ 711 case YYS_EXPR: 712 BEGIN(S0); 713 break; 714 case YYS_DONE: 715 break; 716 case YYS_CONTROL: 717 BEGIN(S4); 718 break; 719 default: 720 xyerror(D_UNKNOWN, "internal error -- bad yystate %d\n", state); 721 } 722 723 yypcb->pcb_yystate = state; 724 } 725 726 void 727 yyinit(dt_pcb_t *pcb) 728 { 729 yypcb = pcb; 730 yylineno = 1; 731 yypragma = NULL; 732 #if defined(sun) 733 yysptr = yysbuf; 734 #endif 735 } 736 737 /* 738 * Given a lexeme 's' (typically yytext), set yylval and return an appropriate 739 * token to the parser indicating either an identifier or a typedef name. 740 * User-defined global variables always take precedence over types, but we do 741 * use some heuristics because D programs can look at an ever-changing set of 742 * kernel types and also can implicitly instantiate variables by assignment, 743 * unlike in C. The code here is ordered carefully as lookups are not cheap. 744 */ 745 static int 746 id_or_type(const char *s) 747 { 748 dtrace_hdl_t *dtp = yypcb->pcb_hdl; 749 dt_decl_t *ddp = yypcb->pcb_dstack.ds_decl; 750 int c0, c1, ttok = DT_TOK_TNAME; 751 dt_ident_t *idp; 752 753 if ((s = yylval.l_str = strdup(s)) == NULL) 754 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 755 756 /* 757 * If the lexeme is a global variable or likely identifier or *not* a 758 * type_name, then it is an identifier token. 759 */ 760 if (dt_idstack_lookup(&yypcb->pcb_globals, s) != NULL || 761 dt_idhash_lookup(yypcb->pcb_idents, s) != NULL || 762 dt_type_lookup(s, NULL) != 0) 763 return (DT_TOK_IDENT); 764 765 /* 766 * If we're in the midst of parsing a declaration and a type_specifier 767 * has already been shifted, then return DT_TOK_IDENT instead of TNAME. 768 * This semantic is necessary to permit valid ISO C code such as: 769 * 770 * typedef int foo; 771 * struct s { foo foo; }; 772 * 773 * without causing shift/reduce conflicts in the direct_declarator part 774 * of the grammar. The result is that we must check for conflicting 775 * redeclarations of the same identifier as part of dt_node_decl(). 776 */ 777 if (ddp != NULL && ddp->dd_name != NULL) 778 return (DT_TOK_IDENT); 779 780 /* 781 * If the lexeme is a type name and we are not in a program clause, 782 * then always interpret it as a type and return DT_TOK_TNAME. 783 */ 784 if ((YYSTATE) != S0) 785 return (DT_TOK_TNAME); 786 787 /* 788 * If the lexeme matches a type name but is in a program clause, then 789 * it could be a type or it could be an undefined variable. Peek at 790 * the next token to decide. If we see ++, --, [, or =, we know there 791 * might be an assignment that is trying to create a global variable, 792 * so we optimistically return DT_TOK_IDENT. There is no harm in being 793 * wrong: a type_name followed by ++, --, [, or = is a syntax error. 794 */ 795 while ((c0 = input()) != 0) { 796 if (strchr("\f\n\r\t\v ", c0) == NULL) 797 break; 798 } 799 800 switch (c0) { 801 case '+': 802 case '-': 803 if ((c1 = input()) == c0) 804 ttok = DT_TOK_IDENT; 805 unput(c1); 806 break; 807 808 case '=': 809 if ((c1 = input()) != c0) 810 ttok = DT_TOK_IDENT; 811 unput(c1); 812 break; 813 case '[': 814 ttok = DT_TOK_IDENT; 815 break; 816 } 817 818 if (ttok == DT_TOK_IDENT) { 819 idp = dt_idhash_insert(yypcb->pcb_idents, s, DT_IDENT_SCALAR, 0, 820 0, _dtrace_defattr, 0, &dt_idops_thaw, NULL, dtp->dt_gen); 821 822 if (idp == NULL) 823 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 824 } 825 826 unput(c0); 827 return (ttok); 828 } 829 830 #if defined(sun) 831 static int 832 input(void) 833 { 834 int c; 835 836 if (yysptr > yysbuf) 837 c = *--yysptr; 838 else if (yypcb->pcb_fileptr != NULL) 839 c = fgetc(yypcb->pcb_fileptr); 840 else if (yypcb->pcb_strptr < yypcb->pcb_string + yypcb->pcb_strlen) 841 c = *(unsigned char *)(yypcb->pcb_strptr++); 842 else 843 c = EOF; 844 845 if (c == '\n') 846 yylineno++; 847 848 if (c != EOF) 849 return (c); 850 851 if ((YYSTATE) == S1) 852 yyerror("end-of-file encountered before matching */\n"); 853 854 if ((YYSTATE) == S3) 855 yyerror("end-of-file encountered before end of control line\n"); 856 857 if (yypcb->pcb_fileptr != NULL && ferror(yypcb->pcb_fileptr)) 858 longjmp(yypcb->pcb_jmpbuf, EDT_FIO); 859 860 return (0); /* EOF */ 861 } 862 863 static void 864 unput(int c) 865 { 866 if (c == '\n') 867 yylineno--; 868 869 *yysptr++ = c; 870 yytchar = c; 871 } 872 #endif 873