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 * Copyright (c) 2013 by Delphix. All rights reserved. 28 * Copyright (c) 2013, Joyent, Inc. All rights reserved. 29 */ 30 31 #include <string.h> 32 #include <stdlib.h> 33 #include <stdio.h> 34 #include <assert.h> 35 #include <ctype.h> 36 #include <errno.h> 37 38 #include <dt_impl.h> 39 #include <dt_grammar.h> 40 #include <dt_parser.h> 41 #include <dt_string.h> 42 43 /* 44 * We need to undefine lex's input and unput macros so that references to these 45 * call the functions provided at the end of this source file. 46 */ 47 #ifdef illumos 48 #undef input 49 #undef unput 50 #else 51 /* 52 * Define YY_INPUT for flex since input() can't be re-defined. 53 */ 54 #define YY_INPUT(buf, result, max_size) \ 55 if (yypcb->pcb_fileptr != NULL) { \ 56 if (((result = fread(buf, 1, max_size, yypcb->pcb_fileptr)) == \ 57 0) && ferror(yypcb->pcb_fileptr)) \ 58 longjmp(yypcb->pcb_jmpbuf, EDT_FIO); \ 59 } else { \ 60 int n; \ 61 for (n = 0; n < max_size && \ 62 yypcb->pcb_strptr < yypcb->pcb_string + yypcb->pcb_strlen; \ 63 n++) \ 64 buf[n] = *yypcb->pcb_strptr++; \ 65 result = n; \ 66 } 67 /* 68 * Do not EOF let tokens to be put back. This does not work with flex. 69 * On the other hand, leaving current buffer in same state it was when 70 * last EOF was received guarantees that input() will keep returning EOF 71 * for all subsequent invocations, which is the effect desired. 72 */ 73 #undef unput 74 #define unput(c) \ 75 do { \ 76 int _c = c; \ 77 if (_c != EOF) \ 78 yyunput(_c, yytext_ptr); \ 79 } while(0) 80 #endif 81 82 static int id_or_type(const char *); 83 #ifdef illumos 84 static int input(void); 85 static void unput(int); 86 #endif 87 88 /* 89 * We first define a set of labeled states for use in the D lexer and then a 90 * set of regular expressions to simplify things below. The lexer states are: 91 * 92 * S0 - D program clause and expression lexing 93 * S1 - D comments (i.e. skip everything until end of comment) 94 * S2 - D program outer scope (probe specifiers and declarations) 95 * S3 - D control line parsing (i.e. after ^# is seen but before \n) 96 * S4 - D control line scan (locate control directives only and invoke S3) 97 */ 98 %} 99 100 %e 1500 /* maximum nodes */ 101 %p 4900 /* maximum positions */ 102 %n 600 /* maximum states */ 103 %a 3000 /* maximum transitions */ 104 105 %s S0 S1 S2 S3 S4 106 107 RGX_AGG "@"[a-zA-Z_][0-9a-zA-Z_]* 108 RGX_PSPEC [-$:a-zA-Z_.?*\\\[\]!][-$:0-9a-zA-Z_.`?*\\\[\]!]* 109 RGX_ALTIDENT [a-zA-Z_][0-9a-zA-Z_]* 110 RGX_LMID LM[0-9a-fA-F]+` 111 RGX_MOD_IDENT [a-zA-Z_`][0-9a-z.A-Z_`]*` 112 RGX_IDENT [a-zA-Z_`][0-9a-zA-Z_`]* 113 RGX_INT ([0-9]+|0[xX][0-9A-Fa-f]+)[uU]?[lL]?[lL]? 114 RGX_FP ([0-9]+("."?)[0-9]*|"."[0-9]+)((e|E)("+"|-)?[0-9]+)?[fFlL]? 115 RGX_WS [\f\n\r\t\v ] 116 RGX_STR ([^"\\\n]|\\[^"\n]|\\\")* 117 RGX_CHR ([^'\\\n]|\\[^'\n]|\\')* 118 RGX_INTERP ^[\f\t\v ]*#!.* 119 RGX_CTL ^[\f\t\v ]*# 120 121 %% 122 123 %{ 124 125 /* 126 * We insert a special prologue into yylex() itself: if the pcb contains a 127 * context token, we return that prior to running the normal lexer. This 128 * allows libdtrace to force yacc into one of our three parsing contexts: D 129 * expression (DT_CTX_DEXPR), D program (DT_CTX_DPROG) or D type (DT_CTX_DTYPE). 130 * Once the token is returned, we clear it so this only happens once. 131 */ 132 if (yypcb->pcb_token != 0) { 133 int tok = yypcb->pcb_token; 134 yypcb->pcb_token = 0; 135 return (tok); 136 } 137 138 %} 139 140 <S0>auto return (DT_KEY_AUTO); 141 <S0>break return (DT_KEY_BREAK); 142 <S0>case return (DT_KEY_CASE); 143 <S0>char return (DT_KEY_CHAR); 144 <S0>const return (DT_KEY_CONST); 145 <S0>continue return (DT_KEY_CONTINUE); 146 <S0>counter return (DT_KEY_COUNTER); 147 <S0>default return (DT_KEY_DEFAULT); 148 <S0>do return (DT_KEY_DO); 149 <S0>double return (DT_KEY_DOUBLE); 150 <S0>else return (DT_KEY_ELSE); 151 <S0>enum return (DT_KEY_ENUM); 152 <S0>extern return (DT_KEY_EXTERN); 153 <S0>float return (DT_KEY_FLOAT); 154 <S0>for return (DT_KEY_FOR); 155 <S0>goto return (DT_KEY_GOTO); 156 <S0>if return (DT_KEY_IF); 157 <S0>import return (DT_KEY_IMPORT); 158 <S0>inline return (DT_KEY_INLINE); 159 <S0>int return (DT_KEY_INT); 160 <S0>long return (DT_KEY_LONG); 161 <S0>offsetof return (DT_TOK_OFFSETOF); 162 <S0>probe return (DT_KEY_PROBE); 163 <S0>provider return (DT_KEY_PROVIDER); 164 <S0>register return (DT_KEY_REGISTER); 165 <S0>restrict return (DT_KEY_RESTRICT); 166 <S0>return return (DT_KEY_RETURN); 167 <S0>self return (DT_KEY_SELF); 168 <S0>short return (DT_KEY_SHORT); 169 <S0>signed return (DT_KEY_SIGNED); 170 <S0>sizeof return (DT_TOK_SIZEOF); 171 <S0>static return (DT_KEY_STATIC); 172 <S0>string return (DT_KEY_STRING); 173 <S0>stringof return (DT_TOK_STRINGOF); 174 <S0>struct return (DT_KEY_STRUCT); 175 <S0>switch return (DT_KEY_SWITCH); 176 <S0>this return (DT_KEY_THIS); 177 <S0>translator return (DT_KEY_XLATOR); 178 <S0>typedef return (DT_KEY_TYPEDEF); 179 <S0>union return (DT_KEY_UNION); 180 <S0>unsigned return (DT_KEY_UNSIGNED); 181 <S0>userland return (DT_KEY_USERLAND); 182 <S0>void return (DT_KEY_VOID); 183 <S0>volatile return (DT_KEY_VOLATILE); 184 <S0>while return (DT_KEY_WHILE); 185 <S0>xlate return (DT_TOK_XLATE); 186 187 <S2>auto { yybegin(YYS_EXPR); return (DT_KEY_AUTO); } 188 <S2>char { yybegin(YYS_EXPR); return (DT_KEY_CHAR); } 189 <S2>const { yybegin(YYS_EXPR); return (DT_KEY_CONST); } 190 <S2>counter { yybegin(YYS_DEFINE); return (DT_KEY_COUNTER); } 191 <S2>double { yybegin(YYS_EXPR); return (DT_KEY_DOUBLE); } 192 <S2>enum { yybegin(YYS_EXPR); return (DT_KEY_ENUM); } 193 <S2>extern { yybegin(YYS_EXPR); return (DT_KEY_EXTERN); } 194 <S2>float { yybegin(YYS_EXPR); return (DT_KEY_FLOAT); } 195 <S2>import { yybegin(YYS_EXPR); return (DT_KEY_IMPORT); } 196 <S2>inline { yybegin(YYS_DEFINE); return (DT_KEY_INLINE); } 197 <S2>int { yybegin(YYS_EXPR); return (DT_KEY_INT); } 198 <S2>long { yybegin(YYS_EXPR); return (DT_KEY_LONG); } 199 <S2>provider { yybegin(YYS_DEFINE); return (DT_KEY_PROVIDER); } 200 <S2>register { yybegin(YYS_EXPR); return (DT_KEY_REGISTER); } 201 <S2>restrict { yybegin(YYS_EXPR); return (DT_KEY_RESTRICT); } 202 <S2>self { yybegin(YYS_EXPR); return (DT_KEY_SELF); } 203 <S2>short { yybegin(YYS_EXPR); return (DT_KEY_SHORT); } 204 <S2>signed { yybegin(YYS_EXPR); return (DT_KEY_SIGNED); } 205 <S2>static { yybegin(YYS_EXPR); return (DT_KEY_STATIC); } 206 <S2>string { yybegin(YYS_EXPR); return (DT_KEY_STRING); } 207 <S2>struct { yybegin(YYS_EXPR); return (DT_KEY_STRUCT); } 208 <S2>this { yybegin(YYS_EXPR); return (DT_KEY_THIS); } 209 <S2>translator { yybegin(YYS_DEFINE); return (DT_KEY_XLATOR); } 210 <S2>typedef { yybegin(YYS_EXPR); return (DT_KEY_TYPEDEF); } 211 <S2>union { yybegin(YYS_EXPR); return (DT_KEY_UNION); } 212 <S2>unsigned { yybegin(YYS_EXPR); return (DT_KEY_UNSIGNED); } 213 <S2>void { yybegin(YYS_EXPR); return (DT_KEY_VOID); } 214 <S2>volatile { yybegin(YYS_EXPR); return (DT_KEY_VOLATILE); } 215 216 <S0>"$$"[0-9]+ { 217 int i = atoi(yytext + 2); 218 char *v = ""; 219 220 /* 221 * A macro argument reference substitutes the text of 222 * an argument in place of the current token. When we 223 * see $$<d> we fetch the saved string from pcb_sargv 224 * (or use the default argument if the option has been 225 * set and the argument hasn't been specified) and 226 * return a token corresponding to this string. 227 */ 228 if (i < 0 || (i >= yypcb->pcb_sargc && 229 !(yypcb->pcb_cflags & DTRACE_C_DEFARG))) { 230 xyerror(D_MACRO_UNDEF, "macro argument %s is " 231 "not defined\n", yytext); 232 } 233 234 if (i < yypcb->pcb_sargc) { 235 v = yypcb->pcb_sargv[i]; /* get val from pcb */ 236 yypcb->pcb_sflagv[i] |= DT_IDFLG_REF; 237 } 238 239 if ((yylval.l_str = strdup(v)) == NULL) 240 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 241 242 (void) stresc2chr(yylval.l_str); 243 return (DT_TOK_STRING); 244 } 245 246 <S0>"$"[0-9]+ { 247 int i = atoi(yytext + 1); 248 char *p, *v = "0"; 249 250 /* 251 * A macro argument reference substitutes the text of 252 * one identifier or integer pattern for another. When 253 * we see $<d> we fetch the saved string from pcb_sargv 254 * (or use the default argument if the option has been 255 * set and the argument hasn't been specified) and 256 * return a token corresponding to this string. 257 */ 258 if (i < 0 || (i >= yypcb->pcb_sargc && 259 !(yypcb->pcb_cflags & DTRACE_C_DEFARG))) { 260 xyerror(D_MACRO_UNDEF, "macro argument %s is " 261 "not defined\n", yytext); 262 } 263 264 if (i < yypcb->pcb_sargc) { 265 v = yypcb->pcb_sargv[i]; /* get val from pcb */ 266 yypcb->pcb_sflagv[i] |= DT_IDFLG_REF; 267 } 268 269 /* 270 * If the macro text is not a valid integer or ident, 271 * then we treat it as a string. The string may be 272 * optionally enclosed in quotes, which we strip. 273 */ 274 if (strbadidnum(v)) { 275 size_t len = strlen(v); 276 277 if (len != 1 && *v == '"' && v[len - 1] == '"') 278 yylval.l_str = strndup(v + 1, len - 2); 279 else 280 yylval.l_str = strndup(v, len); 281 282 if (yylval.l_str == NULL) 283 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 284 285 (void) stresc2chr(yylval.l_str); 286 return (DT_TOK_STRING); 287 } 288 289 /* 290 * If the macro text is not a string an begins with a 291 * digit or a +/- sign, process it as an integer token. 292 */ 293 if (isdigit(v[0]) || v[0] == '-' || v[0] == '+') { 294 if (isdigit(v[0])) 295 yyintprefix = 0; 296 else 297 yyintprefix = *v++; 298 299 errno = 0; 300 yylval.l_int = strtoull(v, &p, 0); 301 (void) strncpy(yyintsuffix, p, 302 sizeof (yyintsuffix)); 303 yyintdecimal = *v != '0'; 304 305 if (errno == ERANGE) { 306 xyerror(D_MACRO_OFLOW, "macro argument" 307 " %s constant %s results in integer" 308 " overflow\n", yytext, v); 309 } 310 311 return (DT_TOK_INT); 312 } 313 314 return (id_or_type(v)); 315 } 316 317 <S0>"$$"{RGX_IDENT} { 318 dt_ident_t *idp = dt_idhash_lookup( 319 yypcb->pcb_hdl->dt_macros, yytext + 2); 320 321 char s[16]; /* enough for UINT_MAX + \0 */ 322 323 if (idp == NULL) { 324 xyerror(D_MACRO_UNDEF, "macro variable %s " 325 "is not defined\n", yytext); 326 } 327 328 /* 329 * For the moment, all current macro variables are of 330 * type id_t (refer to dtrace_update() for details). 331 */ 332 (void) snprintf(s, sizeof (s), "%u", idp->di_id); 333 if ((yylval.l_str = strdup(s)) == NULL) 334 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 335 336 return (DT_TOK_STRING); 337 } 338 339 <S0>"$"{RGX_IDENT} { 340 dt_ident_t *idp = dt_idhash_lookup( 341 yypcb->pcb_hdl->dt_macros, yytext + 1); 342 343 if (idp == NULL) { 344 xyerror(D_MACRO_UNDEF, "macro variable %s " 345 "is not defined\n", yytext); 346 } 347 348 /* 349 * For the moment, all current macro variables are of 350 * type id_t (refer to dtrace_update() for details). 351 */ 352 yylval.l_int = (intmax_t)(int)idp->di_id; 353 yyintprefix = 0; 354 yyintsuffix[0] = '\0'; 355 yyintdecimal = 1; 356 357 return (DT_TOK_INT); 358 } 359 360 <S0>{RGX_IDENT} | 361 <S0>{RGX_MOD_IDENT}{RGX_IDENT} | 362 <S0>{RGX_MOD_IDENT} { 363 return (id_or_type(yytext)); 364 } 365 366 <S0>{RGX_AGG} { 367 if ((yylval.l_str = strdup(yytext)) == NULL) 368 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 369 return (DT_TOK_AGG); 370 } 371 372 <S0>"@" { 373 if ((yylval.l_str = strdup("@_")) == NULL) 374 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 375 return (DT_TOK_AGG); 376 } 377 378 <S0>{RGX_INT} | 379 <S2>{RGX_INT} | 380 <S3>{RGX_INT} { 381 char *p; 382 383 errno = 0; 384 yylval.l_int = strtoull(yytext, &p, 0); 385 yyintprefix = 0; 386 (void) strncpy(yyintsuffix, p, sizeof (yyintsuffix)); 387 yyintdecimal = yytext[0] != '0'; 388 389 if (errno == ERANGE) { 390 xyerror(D_INT_OFLOW, "constant %s results in " 391 "integer overflow\n", yytext); 392 } 393 394 if (*p != '\0' && strchr("uUlL", *p) == NULL) { 395 xyerror(D_INT_DIGIT, "constant %s contains " 396 "invalid digit %c\n", yytext, *p); 397 } 398 399 if ((YYSTATE) != S3) 400 return (DT_TOK_INT); 401 402 yypragma = dt_node_link(yypragma, 403 dt_node_int(yylval.l_int)); 404 } 405 406 <S0>{RGX_FP} yyerror("floating-point constants are not permitted\n"); 407 408 <S0>\"{RGX_STR}$ | 409 <S3>\"{RGX_STR}$ xyerror(D_STR_NL, "newline encountered in string literal"); 410 411 <S0>\"{RGX_STR}\" | 412 <S3>\"{RGX_STR}\" { 413 /* 414 * Quoted string -- convert C escape sequences and 415 * return the string as a token. 416 */ 417 yylval.l_str = strndup(yytext + 1, yyleng - 2); 418 419 if (yylval.l_str == NULL) 420 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 421 422 (void) stresc2chr(yylval.l_str); 423 if ((YYSTATE) != S3) 424 return (DT_TOK_STRING); 425 426 yypragma = dt_node_link(yypragma, 427 dt_node_string(yylval.l_str)); 428 } 429 430 <S0>'{RGX_CHR}$ xyerror(D_CHR_NL, "newline encountered in character constant"); 431 432 <S0>'{RGX_CHR}' { 433 char *s, *p, *q; 434 size_t nbytes; 435 436 /* 437 * Character constant -- convert C escape sequences and 438 * return the character as an integer immediate value. 439 */ 440 if (yyleng == 2) 441 xyerror(D_CHR_NULL, "empty character constant"); 442 443 s = yytext + 1; 444 yytext[yyleng - 1] = '\0'; 445 nbytes = stresc2chr(s); 446 yylval.l_int = 0; 447 yyintprefix = 0; 448 yyintsuffix[0] = '\0'; 449 yyintdecimal = 1; 450 451 if (nbytes > sizeof (yylval.l_int)) { 452 xyerror(D_CHR_OFLOW, "character constant is " 453 "too long"); 454 } 455 #if BYTE_ORDER == _LITTLE_ENDIAN 456 p = ((char *)&yylval.l_int) + nbytes - 1; 457 for (q = s; nbytes != 0; nbytes--) 458 *p-- = *q++; 459 #else 460 bcopy(s, ((char *)&yylval.l_int) + 461 sizeof (yylval.l_int) - nbytes, nbytes); 462 #endif 463 return (DT_TOK_INT); 464 } 465 466 <S0>"/*" | 467 <S2>"/*" { 468 yypcb->pcb_cstate = (YYSTATE); 469 BEGIN(S1); 470 } 471 472 <S0>{RGX_INTERP} | 473 <S2>{RGX_INTERP} ; /* discard any #! lines */ 474 475 <S0>{RGX_CTL} | 476 <S2>{RGX_CTL} | 477 <S4>{RGX_CTL} { 478 assert(yypragma == NULL); 479 yypcb->pcb_cstate = (YYSTATE); 480 BEGIN(S3); 481 } 482 483 <S4>. ; /* discard */ 484 <S4>"\n" ; /* discard */ 485 486 <S0>"/" { 487 int c, tok; 488 489 /* 490 * The use of "/" as the predicate delimiter and as the 491 * integer division symbol requires special lookahead 492 * to avoid a shift/reduce conflict in the D grammar. 493 * We look ahead to the next non-whitespace character. 494 * If we encounter EOF, ";", "{", or "/", then this "/" 495 * closes the predicate and we return DT_TOK_EPRED. 496 * If we encounter anything else, it's DT_TOK_DIV. 497 */ 498 while ((c = input()) != 0) { 499 if (strchr("\f\n\r\t\v ", c) == NULL) 500 break; 501 } 502 503 if (c == 0 || c == ';' || c == '{' || c == '/') { 504 if (yypcb->pcb_parens != 0) { 505 yyerror("closing ) expected in " 506 "predicate before /\n"); 507 } 508 if (yypcb->pcb_brackets != 0) { 509 yyerror("closing ] expected in " 510 "predicate before /\n"); 511 } 512 tok = DT_TOK_EPRED; 513 } else 514 tok = DT_TOK_DIV; 515 516 unput(c); 517 return (tok); 518 } 519 520 <S0>"(" { 521 yypcb->pcb_parens++; 522 return (DT_TOK_LPAR); 523 } 524 525 <S0>")" { 526 if (--yypcb->pcb_parens < 0) 527 yyerror("extra ) in input stream\n"); 528 return (DT_TOK_RPAR); 529 } 530 531 <S0>"[" { 532 yypcb->pcb_brackets++; 533 return (DT_TOK_LBRAC); 534 } 535 536 <S0>"]" { 537 if (--yypcb->pcb_brackets < 0) 538 yyerror("extra ] in input stream\n"); 539 return (DT_TOK_RBRAC); 540 } 541 542 <S0>"{" | 543 <S2>"{" { 544 yypcb->pcb_braces++; 545 return ('{'); 546 } 547 548 <S0>"}" { 549 if (--yypcb->pcb_braces < 0) 550 yyerror("extra } in input stream\n"); 551 return ('}'); 552 } 553 554 <S0>"|" return (DT_TOK_BOR); 555 <S0>"^" return (DT_TOK_XOR); 556 <S0>"&" return (DT_TOK_BAND); 557 <S0>"&&" return (DT_TOK_LAND); 558 <S0>"^^" return (DT_TOK_LXOR); 559 <S0>"||" return (DT_TOK_LOR); 560 <S0>"==" return (DT_TOK_EQU); 561 <S0>"!=" return (DT_TOK_NEQ); 562 <S0>"<" return (DT_TOK_LT); 563 <S0>"<=" return (DT_TOK_LE); 564 <S0>">" return (DT_TOK_GT); 565 <S0>">=" return (DT_TOK_GE); 566 <S0>"<<" return (DT_TOK_LSH); 567 <S0>">>" return (DT_TOK_RSH); 568 <S0>"+" return (DT_TOK_ADD); 569 <S0>"-" return (DT_TOK_SUB); 570 <S0>"*" return (DT_TOK_MUL); 571 <S0>"%" return (DT_TOK_MOD); 572 <S0>"~" return (DT_TOK_BNEG); 573 <S0>"!" return (DT_TOK_LNEG); 574 <S0>"?" return (DT_TOK_QUESTION); 575 <S0>":" return (DT_TOK_COLON); 576 <S0>"." return (DT_TOK_DOT); 577 <S0>"->" return (DT_TOK_PTR); 578 <S0>"=" return (DT_TOK_ASGN); 579 <S0>"+=" return (DT_TOK_ADD_EQ); 580 <S0>"-=" return (DT_TOK_SUB_EQ); 581 <S0>"*=" return (DT_TOK_MUL_EQ); 582 <S0>"/=" return (DT_TOK_DIV_EQ); 583 <S0>"%=" return (DT_TOK_MOD_EQ); 584 <S0>"&=" return (DT_TOK_AND_EQ); 585 <S0>"^=" return (DT_TOK_XOR_EQ); 586 <S0>"|=" return (DT_TOK_OR_EQ); 587 <S0>"<<=" return (DT_TOK_LSH_EQ); 588 <S0>">>=" return (DT_TOK_RSH_EQ); 589 <S0>"++" return (DT_TOK_ADDADD); 590 <S0>"--" return (DT_TOK_SUBSUB); 591 <S0>"..." return (DT_TOK_ELLIPSIS); 592 <S0>"," return (DT_TOK_COMMA); 593 <S0>";" return (';'); 594 <S0>{RGX_WS} ; /* discard */ 595 <S0>"\\"\n ; /* discard */ 596 <S0>. yyerror("syntax error near \"%c\"\n", yytext[0]); 597 598 <S1>"/*" yyerror("/* encountered inside a comment\n"); 599 <S1>"*/" BEGIN(yypcb->pcb_cstate); 600 <S1>.|\n ; /* discard */ 601 602 <S2>{RGX_PSPEC} { 603 /* 604 * S2 has an ambiguity because RGX_PSPEC includes '*' 605 * as a glob character and '*' also can be DT_TOK_STAR. 606 * Since lex always matches the longest token, this 607 * rule can be matched by an input string like "int*", 608 * which could begin a global variable declaration such 609 * as "int*x;" or could begin a RGX_PSPEC with globbing 610 * such as "int* { trace(timestamp); }". If C_PSPEC is 611 * not set, we must resolve the ambiguity in favor of 612 * the type and perform lexer pushback if the fragment 613 * before '*' or entire fragment matches a type name. 614 * If C_PSPEC is set, we always return a PSPEC token. 615 * If C_PSPEC is off, the user can avoid ambiguity by 616 * including a ':' delimiter in the specifier, which 617 * they should be doing anyway to specify the provider. 618 */ 619 if (!(yypcb->pcb_cflags & DTRACE_C_PSPEC) && 620 strchr(yytext, ':') == NULL) { 621 622 char *p = strchr(yytext, '*'); 623 char *q = yytext + yyleng - 1; 624 625 if (p != NULL && p > yytext) 626 *p = '\0'; /* prune yytext */ 627 628 if (dt_type_lookup(yytext, NULL) == 0) { 629 yylval.l_str = strdup(yytext); 630 631 if (yylval.l_str == NULL) { 632 longjmp(yypcb->pcb_jmpbuf, 633 EDT_NOMEM); 634 } 635 636 if (p != NULL && p > yytext) { 637 for (*p = '*'; q >= p; q--) 638 unput(*q); 639 } 640 641 yybegin(YYS_EXPR); 642 return (DT_TOK_TNAME); 643 } 644 645 if (p != NULL && p > yytext) 646 *p = '*'; /* restore yytext */ 647 } 648 649 if ((yylval.l_str = strdup(yytext)) == NULL) 650 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 651 652 return (DT_TOK_PSPEC); 653 } 654 655 <S2>"/" return (DT_TOK_DIV); 656 <S2>"," return (DT_TOK_COMMA); 657 658 <S2>{RGX_WS} ; /* discard */ 659 <S2>. yyerror("syntax error near \"%c\"\n", yytext[0]); 660 661 <S3>\n { 662 dt_pragma(yypragma); 663 yypragma = NULL; 664 BEGIN(yypcb->pcb_cstate); 665 } 666 667 <S3>[\f\t\v ]+ ; /* discard */ 668 669 <S3>[^\f\n\t\v "]+ { 670 dt_node_t *dnp; 671 672 if ((yylval.l_str = strdup(yytext)) == NULL) 673 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 674 675 /* 676 * We want to call dt_node_ident() here, but we can't 677 * because it will expand inlined identifiers, which we 678 * don't want to do from #pragma context in order to 679 * support pragmas that apply to the ident itself. We 680 * call dt_node_string() and then reset dn_op instead. 681 */ 682 dnp = dt_node_string(yylval.l_str); 683 dnp->dn_kind = DT_NODE_IDENT; 684 dnp->dn_op = DT_TOK_IDENT; 685 yypragma = dt_node_link(yypragma, dnp); 686 } 687 688 <S3>. yyerror("syntax error near \"%c\"\n", yytext[0]); 689 690 %% 691 692 /* 693 * yybegin provides a wrapper for use from C code around the lex BEGIN() macro. 694 * We use two main states for lexing because probe descriptions use a syntax 695 * that is incompatible with the normal D tokens (e.g. names can contain "-"). 696 * yybegin also handles the job of switching between two lists of dt_nodes 697 * as we allocate persistent definitions, like inlines, and transient nodes 698 * that will be freed once we are done parsing the current program file. 699 */ 700 void 701 yybegin(yystate_t state) 702 { 703 #ifdef YYDEBUG 704 yydebug = _dtrace_debug; 705 #endif 706 if (yypcb->pcb_yystate == state) 707 return; /* nothing to do if we're in the state already */ 708 709 if (yypcb->pcb_yystate == YYS_DEFINE) { 710 yypcb->pcb_list = yypcb->pcb_hold; 711 yypcb->pcb_hold = NULL; 712 } 713 714 switch (state) { 715 case YYS_CLAUSE: 716 BEGIN(S2); 717 break; 718 case YYS_DEFINE: 719 assert(yypcb->pcb_hold == NULL); 720 yypcb->pcb_hold = yypcb->pcb_list; 721 yypcb->pcb_list = NULL; 722 /*FALLTHRU*/ 723 case YYS_EXPR: 724 BEGIN(S0); 725 break; 726 case YYS_DONE: 727 break; 728 case YYS_CONTROL: 729 BEGIN(S4); 730 break; 731 default: 732 xyerror(D_UNKNOWN, "internal error -- bad yystate %d\n", state); 733 } 734 735 yypcb->pcb_yystate = state; 736 } 737 738 void 739 yyinit(dt_pcb_t *pcb) 740 { 741 yypcb = pcb; 742 yylineno = 1; 743 yypragma = NULL; 744 #ifdef illumos 745 yysptr = yysbuf; 746 #endif 747 YY_FLUSH_BUFFER; 748 } 749 750 /* 751 * Given a lexeme 's' (typically yytext), set yylval and return an appropriate 752 * token to the parser indicating either an identifier or a typedef name. 753 * User-defined global variables always take precedence over types, but we do 754 * use some heuristics because D programs can look at an ever-changing set of 755 * kernel types and also can implicitly instantiate variables by assignment, 756 * unlike in C. The code here is ordered carefully as lookups are not cheap. 757 */ 758 static int 759 id_or_type(const char *s) 760 { 761 dtrace_hdl_t *dtp = yypcb->pcb_hdl; 762 dt_decl_t *ddp = yypcb->pcb_dstack.ds_decl; 763 int c0, c1, ttok = DT_TOK_TNAME; 764 dt_ident_t *idp; 765 766 if ((s = yylval.l_str = strdup(s)) == NULL) 767 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 768 769 /* 770 * If the lexeme is a global variable or likely identifier or *not* a 771 * type_name, then it is an identifier token. 772 */ 773 if (dt_idstack_lookup(&yypcb->pcb_globals, s) != NULL || 774 dt_idhash_lookup(yypcb->pcb_idents, s) != NULL || 775 dt_type_lookup(s, NULL) != 0) 776 return (DT_TOK_IDENT); 777 778 /* 779 * If we're in the midst of parsing a declaration and a type_specifier 780 * has already been shifted, then return DT_TOK_IDENT instead of TNAME. 781 * This semantic is necessary to permit valid ISO C code such as: 782 * 783 * typedef int foo; 784 * struct s { foo foo; }; 785 * 786 * without causing shift/reduce conflicts in the direct_declarator part 787 * of the grammar. The result is that we must check for conflicting 788 * redeclarations of the same identifier as part of dt_node_decl(). 789 */ 790 if (ddp != NULL && ddp->dd_name != NULL) 791 return (DT_TOK_IDENT); 792 793 /* 794 * If the lexeme is a type name and we are not in a program clause, 795 * then always interpret it as a type and return DT_TOK_TNAME. 796 */ 797 if ((YYSTATE) != S0) 798 return (DT_TOK_TNAME); 799 800 /* 801 * If the lexeme matches a type name but is in a program clause, then 802 * it could be a type or it could be an undefined variable. Peek at 803 * the next token to decide. If we see ++, --, [, or =, we know there 804 * might be an assignment that is trying to create a global variable, 805 * so we optimistically return DT_TOK_IDENT. There is no harm in being 806 * wrong: a type_name followed by ++, --, [, or = is a syntax error. 807 */ 808 while ((c0 = input()) != 0) { 809 if (strchr("\f\n\r\t\v ", c0) == NULL) 810 break; 811 } 812 813 switch (c0) { 814 case '+': 815 case '-': 816 if ((c1 = input()) == c0) 817 ttok = DT_TOK_IDENT; 818 unput(c1); 819 break; 820 821 case '=': 822 if ((c1 = input()) != c0) 823 ttok = DT_TOK_IDENT; 824 unput(c1); 825 break; 826 case '[': 827 ttok = DT_TOK_IDENT; 828 break; 829 } 830 831 if (ttok == DT_TOK_IDENT) { 832 idp = dt_idhash_insert(yypcb->pcb_idents, s, DT_IDENT_SCALAR, 0, 833 0, _dtrace_defattr, 0, &dt_idops_thaw, NULL, dtp->dt_gen); 834 835 if (idp == NULL) 836 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 837 } 838 839 unput(c0); 840 return (ttok); 841 } 842 843 #ifdef illumos 844 static int 845 input(void) 846 { 847 int c; 848 849 if (yysptr > yysbuf) 850 c = *--yysptr; 851 else if (yypcb->pcb_fileptr != NULL) 852 c = fgetc(yypcb->pcb_fileptr); 853 else if (yypcb->pcb_strptr < yypcb->pcb_string + yypcb->pcb_strlen) 854 c = *(unsigned char *)(yypcb->pcb_strptr++); 855 else 856 c = EOF; 857 858 if (c == '\n') 859 yylineno++; 860 861 if (c != EOF) 862 return (c); 863 864 if ((YYSTATE) == S1) 865 yyerror("end-of-file encountered before matching */\n"); 866 867 if ((YYSTATE) == S3) 868 yyerror("end-of-file encountered before end of control line\n"); 869 870 if (yypcb->pcb_fileptr != NULL && ferror(yypcb->pcb_fileptr)) 871 longjmp(yypcb->pcb_jmpbuf, EDT_FIO); 872 873 return (0); /* EOF */ 874 } 875 876 static void 877 unput(int c) 878 { 879 if (c == '\n') 880 yylineno--; 881 882 *yysptr++ = c; 883 yytchar = c; 884 } 885 #endif /* illumos */ 886