]> git.ipfire.org Git - thirdparty/binutils-gdb.git/blob - gold/script.cc
* symtab.h (class Symbol_table): Add enum Defined.
[thirdparty/binutils-gdb.git] / gold / script.cc
1 // script.cc -- handle linker scripts for gold.
2
3 // Copyright 2006, 2007, 2008, 2009 Free Software Foundation, Inc.
4 // Written by Ian Lance Taylor <iant@google.com>.
5
6 // This file is part of gold.
7
8 // This program is free software; you can redistribute it and/or modify
9 // it under the terms of the GNU General Public License as published by
10 // the Free Software Foundation; either version 3 of the License, or
11 // (at your option) any later version.
12
13 // This program is distributed in the hope that it will be useful,
14 // but WITHOUT ANY WARRANTY; without even the implied warranty of
15 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 // GNU General Public License for more details.
17
18 // You should have received a copy of the GNU General Public License
19 // along with this program; if not, write to the Free Software
20 // Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston,
21 // MA 02110-1301, USA.
22
23 #include "gold.h"
24
25 #include <cstdio>
26 #include <cstdlib>
27 #include <cstring>
28 #include <fnmatch.h>
29 #include <string>
30 #include <vector>
31 #include "filenames.h"
32
33 #include "elfcpp.h"
34 #include "demangle.h"
35 #include "dirsearch.h"
36 #include "options.h"
37 #include "fileread.h"
38 #include "workqueue.h"
39 #include "readsyms.h"
40 #include "parameters.h"
41 #include "layout.h"
42 #include "symtab.h"
43 #include "target-select.h"
44 #include "script.h"
45 #include "script-c.h"
46 #include "incremental.h"
47
48 namespace gold
49 {
50
51 // A token read from a script file. We don't implement keywords here;
52 // all keywords are simply represented as a string.
53
54 class Token
55 {
56 public:
57 // Token classification.
58 enum Classification
59 {
60 // Token is invalid.
61 TOKEN_INVALID,
62 // Token indicates end of input.
63 TOKEN_EOF,
64 // Token is a string of characters.
65 TOKEN_STRING,
66 // Token is a quoted string of characters.
67 TOKEN_QUOTED_STRING,
68 // Token is an operator.
69 TOKEN_OPERATOR,
70 // Token is a number (an integer).
71 TOKEN_INTEGER
72 };
73
74 // We need an empty constructor so that we can put this STL objects.
75 Token()
76 : classification_(TOKEN_INVALID), value_(NULL), value_length_(0),
77 opcode_(0), lineno_(0), charpos_(0)
78 { }
79
80 // A general token with no value.
81 Token(Classification classification, int lineno, int charpos)
82 : classification_(classification), value_(NULL), value_length_(0),
83 opcode_(0), lineno_(lineno), charpos_(charpos)
84 {
85 gold_assert(classification == TOKEN_INVALID
86 || classification == TOKEN_EOF);
87 }
88
89 // A general token with a value.
90 Token(Classification classification, const char* value, size_t length,
91 int lineno, int charpos)
92 : classification_(classification), value_(value), value_length_(length),
93 opcode_(0), lineno_(lineno), charpos_(charpos)
94 {
95 gold_assert(classification != TOKEN_INVALID
96 && classification != TOKEN_EOF);
97 }
98
99 // A token representing an operator.
100 Token(int opcode, int lineno, int charpos)
101 : classification_(TOKEN_OPERATOR), value_(NULL), value_length_(0),
102 opcode_(opcode), lineno_(lineno), charpos_(charpos)
103 { }
104
105 // Return whether the token is invalid.
106 bool
107 is_invalid() const
108 { return this->classification_ == TOKEN_INVALID; }
109
110 // Return whether this is an EOF token.
111 bool
112 is_eof() const
113 { return this->classification_ == TOKEN_EOF; }
114
115 // Return the token classification.
116 Classification
117 classification() const
118 { return this->classification_; }
119
120 // Return the line number at which the token starts.
121 int
122 lineno() const
123 { return this->lineno_; }
124
125 // Return the character position at this the token starts.
126 int
127 charpos() const
128 { return this->charpos_; }
129
130 // Get the value of a token.
131
132 const char*
133 string_value(size_t* length) const
134 {
135 gold_assert(this->classification_ == TOKEN_STRING
136 || this->classification_ == TOKEN_QUOTED_STRING);
137 *length = this->value_length_;
138 return this->value_;
139 }
140
141 int
142 operator_value() const
143 {
144 gold_assert(this->classification_ == TOKEN_OPERATOR);
145 return this->opcode_;
146 }
147
148 uint64_t
149 integer_value() const
150 {
151 gold_assert(this->classification_ == TOKEN_INTEGER);
152 // Null terminate.
153 std::string s(this->value_, this->value_length_);
154 return strtoull(s.c_str(), NULL, 0);
155 }
156
157 private:
158 // The token classification.
159 Classification classification_;
160 // The token value, for TOKEN_STRING or TOKEN_QUOTED_STRING or
161 // TOKEN_INTEGER.
162 const char* value_;
163 // The length of the token value.
164 size_t value_length_;
165 // The token value, for TOKEN_OPERATOR.
166 int opcode_;
167 // The line number where this token started (one based).
168 int lineno_;
169 // The character position within the line where this token started
170 // (one based).
171 int charpos_;
172 };
173
174 // This class handles lexing a file into a sequence of tokens.
175
176 class Lex
177 {
178 public:
179 // We unfortunately have to support different lexing modes, because
180 // when reading different parts of a linker script we need to parse
181 // things differently.
182 enum Mode
183 {
184 // Reading an ordinary linker script.
185 LINKER_SCRIPT,
186 // Reading an expression in a linker script.
187 EXPRESSION,
188 // Reading a version script.
189 VERSION_SCRIPT,
190 // Reading a --dynamic-list file.
191 DYNAMIC_LIST
192 };
193
194 Lex(const char* input_string, size_t input_length, int parsing_token)
195 : input_string_(input_string), input_length_(input_length),
196 current_(input_string), mode_(LINKER_SCRIPT),
197 first_token_(parsing_token), token_(),
198 lineno_(1), linestart_(input_string)
199 { }
200
201 // Read a file into a string.
202 static void
203 read_file(Input_file*, std::string*);
204
205 // Return the next token.
206 const Token*
207 next_token();
208
209 // Return the current lexing mode.
210 Lex::Mode
211 mode() const
212 { return this->mode_; }
213
214 // Set the lexing mode.
215 void
216 set_mode(Mode mode)
217 { this->mode_ = mode; }
218
219 private:
220 Lex(const Lex&);
221 Lex& operator=(const Lex&);
222
223 // Make a general token with no value at the current location.
224 Token
225 make_token(Token::Classification c, const char* start) const
226 { return Token(c, this->lineno_, start - this->linestart_ + 1); }
227
228 // Make a general token with a value at the current location.
229 Token
230 make_token(Token::Classification c, const char* v, size_t len,
231 const char* start)
232 const
233 { return Token(c, v, len, this->lineno_, start - this->linestart_ + 1); }
234
235 // Make an operator token at the current location.
236 Token
237 make_token(int opcode, const char* start) const
238 { return Token(opcode, this->lineno_, start - this->linestart_ + 1); }
239
240 // Make an invalid token at the current location.
241 Token
242 make_invalid_token(const char* start)
243 { return this->make_token(Token::TOKEN_INVALID, start); }
244
245 // Make an EOF token at the current location.
246 Token
247 make_eof_token(const char* start)
248 { return this->make_token(Token::TOKEN_EOF, start); }
249
250 // Return whether C can be the first character in a name. C2 is the
251 // next character, since we sometimes need that.
252 inline bool
253 can_start_name(char c, char c2);
254
255 // If C can appear in a name which has already started, return a
256 // pointer to a character later in the token or just past
257 // it. Otherwise, return NULL.
258 inline const char*
259 can_continue_name(const char* c);
260
261 // Return whether C, C2, C3 can start a hex number.
262 inline bool
263 can_start_hex(char c, char c2, char c3);
264
265 // If C can appear in a hex number which has already started, return
266 // a pointer to a character later in the token or just past
267 // it. Otherwise, return NULL.
268 inline const char*
269 can_continue_hex(const char* c);
270
271 // Return whether C can start a non-hex number.
272 static inline bool
273 can_start_number(char c);
274
275 // If C can appear in a decimal number which has already started,
276 // return a pointer to a character later in the token or just past
277 // it. Otherwise, return NULL.
278 inline const char*
279 can_continue_number(const char* c)
280 { return Lex::can_start_number(*c) ? c + 1 : NULL; }
281
282 // If C1 C2 C3 form a valid three character operator, return the
283 // opcode. Otherwise return 0.
284 static inline int
285 three_char_operator(char c1, char c2, char c3);
286
287 // If C1 C2 form a valid two character operator, return the opcode.
288 // Otherwise return 0.
289 static inline int
290 two_char_operator(char c1, char c2);
291
292 // If C1 is a valid one character operator, return the opcode.
293 // Otherwise return 0.
294 static inline int
295 one_char_operator(char c1);
296
297 // Read the next token.
298 Token
299 get_token(const char**);
300
301 // Skip a C style /* */ comment. Return false if the comment did
302 // not end.
303 bool
304 skip_c_comment(const char**);
305
306 // Skip a line # comment. Return false if there was no newline.
307 bool
308 skip_line_comment(const char**);
309
310 // Build a token CLASSIFICATION from all characters that match
311 // CAN_CONTINUE_FN. The token starts at START. Start matching from
312 // MATCH. Set *PP to the character following the token.
313 inline Token
314 gather_token(Token::Classification,
315 const char* (Lex::*can_continue_fn)(const char*),
316 const char* start, const char* match, const char** pp);
317
318 // Build a token from a quoted string.
319 Token
320 gather_quoted_string(const char** pp);
321
322 // The string we are tokenizing.
323 const char* input_string_;
324 // The length of the string.
325 size_t input_length_;
326 // The current offset into the string.
327 const char* current_;
328 // The current lexing mode.
329 Mode mode_;
330 // The code to use for the first token. This is set to 0 after it
331 // is used.
332 int first_token_;
333 // The current token.
334 Token token_;
335 // The current line number.
336 int lineno_;
337 // The start of the current line in the string.
338 const char* linestart_;
339 };
340
341 // Read the whole file into memory. We don't expect linker scripts to
342 // be large, so we just use a std::string as a buffer. We ignore the
343 // data we've already read, so that we read aligned buffers.
344
345 void
346 Lex::read_file(Input_file* input_file, std::string* contents)
347 {
348 off_t filesize = input_file->file().filesize();
349 contents->clear();
350 contents->reserve(filesize);
351
352 off_t off = 0;
353 unsigned char buf[BUFSIZ];
354 while (off < filesize)
355 {
356 off_t get = BUFSIZ;
357 if (get > filesize - off)
358 get = filesize - off;
359 input_file->file().read(off, get, buf);
360 contents->append(reinterpret_cast<char*>(&buf[0]), get);
361 off += get;
362 }
363 }
364
365 // Return whether C can be the start of a name, if the next character
366 // is C2. A name can being with a letter, underscore, period, or
367 // dollar sign. Because a name can be a file name, we also permit
368 // forward slash, backslash, and tilde. Tilde is the tricky case
369 // here; GNU ld also uses it as a bitwise not operator. It is only
370 // recognized as the operator if it is not immediately followed by
371 // some character which can appear in a symbol. That is, when we
372 // don't know that we are looking at an expression, "~0" is a file
373 // name, and "~ 0" is an expression using bitwise not. We are
374 // compatible.
375
376 inline bool
377 Lex::can_start_name(char c, char c2)
378 {
379 switch (c)
380 {
381 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
382 case 'G': case 'H': case 'I': case 'J': case 'K': case 'L':
383 case 'M': case 'N': case 'O': case 'Q': case 'P': case 'R':
384 case 'S': case 'T': case 'U': case 'V': case 'W': case 'X':
385 case 'Y': case 'Z':
386 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
387 case 'g': case 'h': case 'i': case 'j': case 'k': case 'l':
388 case 'm': case 'n': case 'o': case 'q': case 'p': case 'r':
389 case 's': case 't': case 'u': case 'v': case 'w': case 'x':
390 case 'y': case 'z':
391 case '_': case '.': case '$':
392 return true;
393
394 case '/': case '\\':
395 return this->mode_ == LINKER_SCRIPT;
396
397 case '~':
398 return this->mode_ == LINKER_SCRIPT && can_continue_name(&c2);
399
400 case '*': case '[':
401 return (this->mode_ == VERSION_SCRIPT
402 || this->mode_ == DYNAMIC_LIST
403 || (this->mode_ == LINKER_SCRIPT
404 && can_continue_name(&c2)));
405
406 default:
407 return false;
408 }
409 }
410
411 // Return whether C can continue a name which has already started.
412 // Subsequent characters in a name are the same as the leading
413 // characters, plus digits and "=+-:[],?*". So in general the linker
414 // script language requires spaces around operators, unless we know
415 // that we are parsing an expression.
416
417 inline const char*
418 Lex::can_continue_name(const char* c)
419 {
420 switch (*c)
421 {
422 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
423 case 'G': case 'H': case 'I': case 'J': case 'K': case 'L':
424 case 'M': case 'N': case 'O': case 'Q': case 'P': case 'R':
425 case 'S': case 'T': case 'U': case 'V': case 'W': case 'X':
426 case 'Y': case 'Z':
427 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
428 case 'g': case 'h': case 'i': case 'j': case 'k': case 'l':
429 case 'm': case 'n': case 'o': case 'q': case 'p': case 'r':
430 case 's': case 't': case 'u': case 'v': case 'w': case 'x':
431 case 'y': case 'z':
432 case '_': case '.': case '$':
433 case '0': case '1': case '2': case '3': case '4':
434 case '5': case '6': case '7': case '8': case '9':
435 return c + 1;
436
437 // TODO(csilvers): why not allow ~ in names for version-scripts?
438 case '/': case '\\': case '~':
439 case '=': case '+':
440 case ',':
441 if (this->mode_ == LINKER_SCRIPT)
442 return c + 1;
443 return NULL;
444
445 case '[': case ']': case '*': case '?': case '-':
446 if (this->mode_ == LINKER_SCRIPT || this->mode_ == VERSION_SCRIPT
447 || this->mode_ == DYNAMIC_LIST)
448 return c + 1;
449 return NULL;
450
451 // TODO(csilvers): why allow this? ^ is meaningless in version scripts.
452 case '^':
453 if (this->mode_ == VERSION_SCRIPT || this->mode_ == DYNAMIC_LIST)
454 return c + 1;
455 return NULL;
456
457 case ':':
458 if (this->mode_ == LINKER_SCRIPT)
459 return c + 1;
460 else if ((this->mode_ == VERSION_SCRIPT || this->mode_ == DYNAMIC_LIST)
461 && (c[1] == ':'))
462 {
463 // A name can have '::' in it, as that's a c++ namespace
464 // separator. But a single colon is not part of a name.
465 return c + 2;
466 }
467 return NULL;
468
469 default:
470 return NULL;
471 }
472 }
473
474 // For a number we accept 0x followed by hex digits, or any sequence
475 // of digits. The old linker accepts leading '$' for hex, and
476 // trailing HXBOD. Those are for MRI compatibility and we don't
477 // accept them. The old linker also accepts trailing MK for mega or
478 // kilo. FIXME: Those are mentioned in the documentation, and we
479 // should accept them.
480
481 // Return whether C1 C2 C3 can start a hex number.
482
483 inline bool
484 Lex::can_start_hex(char c1, char c2, char c3)
485 {
486 if (c1 == '0' && (c2 == 'x' || c2 == 'X'))
487 return this->can_continue_hex(&c3);
488 return false;
489 }
490
491 // Return whether C can appear in a hex number.
492
493 inline const char*
494 Lex::can_continue_hex(const char* c)
495 {
496 switch (*c)
497 {
498 case '0': case '1': case '2': case '3': case '4':
499 case '5': case '6': case '7': case '8': case '9':
500 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
501 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
502 return c + 1;
503
504 default:
505 return NULL;
506 }
507 }
508
509 // Return whether C can start a non-hex number.
510
511 inline bool
512 Lex::can_start_number(char c)
513 {
514 switch (c)
515 {
516 case '0': case '1': case '2': case '3': case '4':
517 case '5': case '6': case '7': case '8': case '9':
518 return true;
519
520 default:
521 return false;
522 }
523 }
524
525 // If C1 C2 C3 form a valid three character operator, return the
526 // opcode (defined in the yyscript.h file generated from yyscript.y).
527 // Otherwise return 0.
528
529 inline int
530 Lex::three_char_operator(char c1, char c2, char c3)
531 {
532 switch (c1)
533 {
534 case '<':
535 if (c2 == '<' && c3 == '=')
536 return LSHIFTEQ;
537 break;
538 case '>':
539 if (c2 == '>' && c3 == '=')
540 return RSHIFTEQ;
541 break;
542 default:
543 break;
544 }
545 return 0;
546 }
547
548 // If C1 C2 form a valid two character operator, return the opcode
549 // (defined in the yyscript.h file generated from yyscript.y).
550 // Otherwise return 0.
551
552 inline int
553 Lex::two_char_operator(char c1, char c2)
554 {
555 switch (c1)
556 {
557 case '=':
558 if (c2 == '=')
559 return EQ;
560 break;
561 case '!':
562 if (c2 == '=')
563 return NE;
564 break;
565 case '+':
566 if (c2 == '=')
567 return PLUSEQ;
568 break;
569 case '-':
570 if (c2 == '=')
571 return MINUSEQ;
572 break;
573 case '*':
574 if (c2 == '=')
575 return MULTEQ;
576 break;
577 case '/':
578 if (c2 == '=')
579 return DIVEQ;
580 break;
581 case '|':
582 if (c2 == '=')
583 return OREQ;
584 if (c2 == '|')
585 return OROR;
586 break;
587 case '&':
588 if (c2 == '=')
589 return ANDEQ;
590 if (c2 == '&')
591 return ANDAND;
592 break;
593 case '>':
594 if (c2 == '=')
595 return GE;
596 if (c2 == '>')
597 return RSHIFT;
598 break;
599 case '<':
600 if (c2 == '=')
601 return LE;
602 if (c2 == '<')
603 return LSHIFT;
604 break;
605 default:
606 break;
607 }
608 return 0;
609 }
610
611 // If C1 is a valid operator, return the opcode. Otherwise return 0.
612
613 inline int
614 Lex::one_char_operator(char c1)
615 {
616 switch (c1)
617 {
618 case '+':
619 case '-':
620 case '*':
621 case '/':
622 case '%':
623 case '!':
624 case '&':
625 case '|':
626 case '^':
627 case '~':
628 case '<':
629 case '>':
630 case '=':
631 case '?':
632 case ',':
633 case '(':
634 case ')':
635 case '{':
636 case '}':
637 case '[':
638 case ']':
639 case ':':
640 case ';':
641 return c1;
642 default:
643 return 0;
644 }
645 }
646
647 // Skip a C style comment. *PP points to just after the "/*". Return
648 // false if the comment did not end.
649
650 bool
651 Lex::skip_c_comment(const char** pp)
652 {
653 const char* p = *pp;
654 while (p[0] != '*' || p[1] != '/')
655 {
656 if (*p == '\0')
657 {
658 *pp = p;
659 return false;
660 }
661
662 if (*p == '\n')
663 {
664 ++this->lineno_;
665 this->linestart_ = p + 1;
666 }
667 ++p;
668 }
669
670 *pp = p + 2;
671 return true;
672 }
673
674 // Skip a line # comment. Return false if there was no newline.
675
676 bool
677 Lex::skip_line_comment(const char** pp)
678 {
679 const char* p = *pp;
680 size_t skip = strcspn(p, "\n");
681 if (p[skip] == '\0')
682 {
683 *pp = p + skip;
684 return false;
685 }
686
687 p += skip + 1;
688 ++this->lineno_;
689 this->linestart_ = p;
690 *pp = p;
691
692 return true;
693 }
694
695 // Build a token CLASSIFICATION from all characters that match
696 // CAN_CONTINUE_FN. Update *PP.
697
698 inline Token
699 Lex::gather_token(Token::Classification classification,
700 const char* (Lex::*can_continue_fn)(const char*),
701 const char* start,
702 const char* match,
703 const char **pp)
704 {
705 const char* new_match = NULL;
706 while ((new_match = (this->*can_continue_fn)(match)))
707 match = new_match;
708 *pp = match;
709 return this->make_token(classification, start, match - start, start);
710 }
711
712 // Build a token from a quoted string.
713
714 Token
715 Lex::gather_quoted_string(const char** pp)
716 {
717 const char* start = *pp;
718 const char* p = start;
719 ++p;
720 size_t skip = strcspn(p, "\"\n");
721 if (p[skip] != '"')
722 return this->make_invalid_token(start);
723 *pp = p + skip + 1;
724 return this->make_token(Token::TOKEN_QUOTED_STRING, p, skip, start);
725 }
726
727 // Return the next token at *PP. Update *PP. General guideline: we
728 // require linker scripts to be simple ASCII. No unicode linker
729 // scripts. In particular we can assume that any '\0' is the end of
730 // the input.
731
732 Token
733 Lex::get_token(const char** pp)
734 {
735 const char* p = *pp;
736
737 while (true)
738 {
739 if (*p == '\0')
740 {
741 *pp = p;
742 return this->make_eof_token(p);
743 }
744
745 // Skip whitespace quickly.
746 while (*p == ' ' || *p == '\t' || *p == '\r')
747 ++p;
748
749 if (*p == '\n')
750 {
751 ++p;
752 ++this->lineno_;
753 this->linestart_ = p;
754 continue;
755 }
756
757 // Skip C style comments.
758 if (p[0] == '/' && p[1] == '*')
759 {
760 int lineno = this->lineno_;
761 int charpos = p - this->linestart_ + 1;
762
763 *pp = p + 2;
764 if (!this->skip_c_comment(pp))
765 return Token(Token::TOKEN_INVALID, lineno, charpos);
766 p = *pp;
767
768 continue;
769 }
770
771 // Skip line comments.
772 if (*p == '#')
773 {
774 *pp = p + 1;
775 if (!this->skip_line_comment(pp))
776 return this->make_eof_token(p);
777 p = *pp;
778 continue;
779 }
780
781 // Check for a name.
782 if (this->can_start_name(p[0], p[1]))
783 return this->gather_token(Token::TOKEN_STRING,
784 &Lex::can_continue_name,
785 p, p + 1, pp);
786
787 // We accept any arbitrary name in double quotes, as long as it
788 // does not cross a line boundary.
789 if (*p == '"')
790 {
791 *pp = p;
792 return this->gather_quoted_string(pp);
793 }
794
795 // Check for a number.
796
797 if (this->can_start_hex(p[0], p[1], p[2]))
798 return this->gather_token(Token::TOKEN_INTEGER,
799 &Lex::can_continue_hex,
800 p, p + 3, pp);
801
802 if (Lex::can_start_number(p[0]))
803 return this->gather_token(Token::TOKEN_INTEGER,
804 &Lex::can_continue_number,
805 p, p + 1, pp);
806
807 // Check for operators.
808
809 int opcode = Lex::three_char_operator(p[0], p[1], p[2]);
810 if (opcode != 0)
811 {
812 *pp = p + 3;
813 return this->make_token(opcode, p);
814 }
815
816 opcode = Lex::two_char_operator(p[0], p[1]);
817 if (opcode != 0)
818 {
819 *pp = p + 2;
820 return this->make_token(opcode, p);
821 }
822
823 opcode = Lex::one_char_operator(p[0]);
824 if (opcode != 0)
825 {
826 *pp = p + 1;
827 return this->make_token(opcode, p);
828 }
829
830 return this->make_token(Token::TOKEN_INVALID, p);
831 }
832 }
833
834 // Return the next token.
835
836 const Token*
837 Lex::next_token()
838 {
839 // The first token is special.
840 if (this->first_token_ != 0)
841 {
842 this->token_ = Token(this->first_token_, 0, 0);
843 this->first_token_ = 0;
844 return &this->token_;
845 }
846
847 this->token_ = this->get_token(&this->current_);
848
849 // Don't let an early null byte fool us into thinking that we've
850 // reached the end of the file.
851 if (this->token_.is_eof()
852 && (static_cast<size_t>(this->current_ - this->input_string_)
853 < this->input_length_))
854 this->token_ = this->make_invalid_token(this->current_);
855
856 return &this->token_;
857 }
858
859 // class Symbol_assignment.
860
861 // Add the symbol to the symbol table. This makes sure the symbol is
862 // there and defined. The actual value is stored later. We can't
863 // determine the actual value at this point, because we can't
864 // necessarily evaluate the expression until all ordinary symbols have
865 // been finalized.
866
867 // The GNU linker lets symbol assignments in the linker script
868 // silently override defined symbols in object files. We are
869 // compatible. FIXME: Should we issue a warning?
870
871 void
872 Symbol_assignment::add_to_table(Symbol_table* symtab)
873 {
874 elfcpp::STV vis = this->hidden_ ? elfcpp::STV_HIDDEN : elfcpp::STV_DEFAULT;
875 this->sym_ = symtab->define_as_constant(this->name_.c_str(),
876 NULL, // version
877 (this->is_defsym_
878 ? Symbol_table::DEFSYM
879 : Symbol_table::SCRIPT),
880 0, // value
881 0, // size
882 elfcpp::STT_NOTYPE,
883 elfcpp::STB_GLOBAL,
884 vis,
885 0, // nonvis
886 this->provide_,
887 true); // force_override
888 }
889
890 // Finalize a symbol value.
891
892 void
893 Symbol_assignment::finalize(Symbol_table* symtab, const Layout* layout)
894 {
895 this->finalize_maybe_dot(symtab, layout, false, 0, NULL);
896 }
897
898 // Finalize a symbol value which can refer to the dot symbol.
899
900 void
901 Symbol_assignment::finalize_with_dot(Symbol_table* symtab,
902 const Layout* layout,
903 uint64_t dot_value,
904 Output_section* dot_section)
905 {
906 this->finalize_maybe_dot(symtab, layout, true, dot_value, dot_section);
907 }
908
909 // Finalize a symbol value, internal version.
910
911 void
912 Symbol_assignment::finalize_maybe_dot(Symbol_table* symtab,
913 const Layout* layout,
914 bool is_dot_available,
915 uint64_t dot_value,
916 Output_section* dot_section)
917 {
918 // If we were only supposed to provide this symbol, the sym_ field
919 // will be NULL if the symbol was not referenced.
920 if (this->sym_ == NULL)
921 {
922 gold_assert(this->provide_);
923 return;
924 }
925
926 if (parameters->target().get_size() == 32)
927 {
928 #if defined(HAVE_TARGET_32_LITTLE) || defined(HAVE_TARGET_32_BIG)
929 this->sized_finalize<32>(symtab, layout, is_dot_available, dot_value,
930 dot_section);
931 #else
932 gold_unreachable();
933 #endif
934 }
935 else if (parameters->target().get_size() == 64)
936 {
937 #if defined(HAVE_TARGET_64_LITTLE) || defined(HAVE_TARGET_64_BIG)
938 this->sized_finalize<64>(symtab, layout, is_dot_available, dot_value,
939 dot_section);
940 #else
941 gold_unreachable();
942 #endif
943 }
944 else
945 gold_unreachable();
946 }
947
948 template<int size>
949 void
950 Symbol_assignment::sized_finalize(Symbol_table* symtab, const Layout* layout,
951 bool is_dot_available, uint64_t dot_value,
952 Output_section* dot_section)
953 {
954 Output_section* section;
955 uint64_t final_val = this->val_->eval_maybe_dot(symtab, layout, true,
956 is_dot_available,
957 dot_value, dot_section,
958 &section);
959 Sized_symbol<size>* ssym = symtab->get_sized_symbol<size>(this->sym_);
960 ssym->set_value(final_val);
961 if (section != NULL)
962 ssym->set_output_section(section);
963 }
964
965 // Set the symbol value if the expression yields an absolute value.
966
967 void
968 Symbol_assignment::set_if_absolute(Symbol_table* symtab, const Layout* layout,
969 bool is_dot_available, uint64_t dot_value)
970 {
971 if (this->sym_ == NULL)
972 return;
973
974 Output_section* val_section;
975 uint64_t val = this->val_->eval_maybe_dot(symtab, layout, false,
976 is_dot_available, dot_value,
977 NULL, &val_section);
978 if (val_section != NULL)
979 return;
980
981 if (parameters->target().get_size() == 32)
982 {
983 #if defined(HAVE_TARGET_32_LITTLE) || defined(HAVE_TARGET_32_BIG)
984 Sized_symbol<32>* ssym = symtab->get_sized_symbol<32>(this->sym_);
985 ssym->set_value(val);
986 #else
987 gold_unreachable();
988 #endif
989 }
990 else if (parameters->target().get_size() == 64)
991 {
992 #if defined(HAVE_TARGET_64_LITTLE) || defined(HAVE_TARGET_64_BIG)
993 Sized_symbol<64>* ssym = symtab->get_sized_symbol<64>(this->sym_);
994 ssym->set_value(val);
995 #else
996 gold_unreachable();
997 #endif
998 }
999 else
1000 gold_unreachable();
1001 }
1002
1003 // Print for debugging.
1004
1005 void
1006 Symbol_assignment::print(FILE* f) const
1007 {
1008 if (this->provide_ && this->hidden_)
1009 fprintf(f, "PROVIDE_HIDDEN(");
1010 else if (this->provide_)
1011 fprintf(f, "PROVIDE(");
1012 else if (this->hidden_)
1013 gold_unreachable();
1014
1015 fprintf(f, "%s = ", this->name_.c_str());
1016 this->val_->print(f);
1017
1018 if (this->provide_ || this->hidden_)
1019 fprintf(f, ")");
1020
1021 fprintf(f, "\n");
1022 }
1023
1024 // Class Script_assertion.
1025
1026 // Check the assertion.
1027
1028 void
1029 Script_assertion::check(const Symbol_table* symtab, const Layout* layout)
1030 {
1031 if (!this->check_->eval(symtab, layout, true))
1032 gold_error("%s", this->message_.c_str());
1033 }
1034
1035 // Print for debugging.
1036
1037 void
1038 Script_assertion::print(FILE* f) const
1039 {
1040 fprintf(f, "ASSERT(");
1041 this->check_->print(f);
1042 fprintf(f, ", \"%s\")\n", this->message_.c_str());
1043 }
1044
1045 // Class Script_options.
1046
1047 Script_options::Script_options()
1048 : entry_(), symbol_assignments_(), version_script_info_(),
1049 script_sections_()
1050 {
1051 }
1052
1053 // Add a symbol to be defined.
1054
1055 void
1056 Script_options::add_symbol_assignment(const char* name, size_t length,
1057 bool is_defsym, Expression* value,
1058 bool provide, bool hidden)
1059 {
1060 if (length != 1 || name[0] != '.')
1061 {
1062 if (this->script_sections_.in_sections_clause())
1063 {
1064 gold_assert(!is_defsym);
1065 this->script_sections_.add_symbol_assignment(name, length, value,
1066 provide, hidden);
1067 }
1068 else
1069 {
1070 Symbol_assignment* p = new Symbol_assignment(name, length, is_defsym,
1071 value, provide, hidden);
1072 this->symbol_assignments_.push_back(p);
1073 }
1074 }
1075 else
1076 {
1077 if (provide || hidden)
1078 gold_error(_("invalid use of PROVIDE for dot symbol"));
1079
1080 // The GNU linker permits assignments to dot outside of SECTIONS
1081 // clauses and treats them as occurring inside, so we don't
1082 // check in_sections_clause here.
1083 this->script_sections_.add_dot_assignment(value);
1084 }
1085 }
1086
1087 // Add an assertion.
1088
1089 void
1090 Script_options::add_assertion(Expression* check, const char* message,
1091 size_t messagelen)
1092 {
1093 if (this->script_sections_.in_sections_clause())
1094 this->script_sections_.add_assertion(check, message, messagelen);
1095 else
1096 {
1097 Script_assertion* p = new Script_assertion(check, message, messagelen);
1098 this->assertions_.push_back(p);
1099 }
1100 }
1101
1102 // Create sections required by any linker scripts.
1103
1104 void
1105 Script_options::create_script_sections(Layout* layout)
1106 {
1107 if (this->saw_sections_clause())
1108 this->script_sections_.create_sections(layout);
1109 }
1110
1111 // Add any symbols we are defining to the symbol table.
1112
1113 void
1114 Script_options::add_symbols_to_table(Symbol_table* symtab)
1115 {
1116 for (Symbol_assignments::iterator p = this->symbol_assignments_.begin();
1117 p != this->symbol_assignments_.end();
1118 ++p)
1119 (*p)->add_to_table(symtab);
1120 this->script_sections_.add_symbols_to_table(symtab);
1121 }
1122
1123 // Finalize symbol values. Also check assertions.
1124
1125 void
1126 Script_options::finalize_symbols(Symbol_table* symtab, const Layout* layout)
1127 {
1128 // We finalize the symbols defined in SECTIONS first, because they
1129 // are the ones which may have changed. This way if symbol outside
1130 // SECTIONS are defined in terms of symbols inside SECTIONS, they
1131 // will get the right value.
1132 this->script_sections_.finalize_symbols(symtab, layout);
1133
1134 for (Symbol_assignments::iterator p = this->symbol_assignments_.begin();
1135 p != this->symbol_assignments_.end();
1136 ++p)
1137 (*p)->finalize(symtab, layout);
1138
1139 for (Assertions::iterator p = this->assertions_.begin();
1140 p != this->assertions_.end();
1141 ++p)
1142 (*p)->check(symtab, layout);
1143 }
1144
1145 // Set section addresses. We set all the symbols which have absolute
1146 // values. Then we let the SECTIONS clause do its thing. This
1147 // returns the segment which holds the file header and segment
1148 // headers, if any.
1149
1150 Output_segment*
1151 Script_options::set_section_addresses(Symbol_table* symtab, Layout* layout)
1152 {
1153 for (Symbol_assignments::iterator p = this->symbol_assignments_.begin();
1154 p != this->symbol_assignments_.end();
1155 ++p)
1156 (*p)->set_if_absolute(symtab, layout, false, 0);
1157
1158 return this->script_sections_.set_section_addresses(symtab, layout);
1159 }
1160
1161 // This class holds data passed through the parser to the lexer and to
1162 // the parser support functions. This avoids global variables. We
1163 // can't use global variables because we need not be called by a
1164 // singleton thread.
1165
1166 class Parser_closure
1167 {
1168 public:
1169 Parser_closure(const char* filename,
1170 const Position_dependent_options& posdep_options,
1171 bool parsing_defsym, bool in_group, bool is_in_sysroot,
1172 Command_line* command_line,
1173 Script_options* script_options,
1174 Lex* lex,
1175 bool skip_on_incompatible_target)
1176 : filename_(filename), posdep_options_(posdep_options),
1177 parsing_defsym_(parsing_defsym), in_group_(in_group),
1178 is_in_sysroot_(is_in_sysroot),
1179 skip_on_incompatible_target_(skip_on_incompatible_target),
1180 found_incompatible_target_(false),
1181 command_line_(command_line), script_options_(script_options),
1182 version_script_info_(script_options->version_script_info()),
1183 lex_(lex), lineno_(0), charpos_(0), lex_mode_stack_(), inputs_(NULL)
1184 {
1185 // We start out processing C symbols in the default lex mode.
1186 language_stack_.push_back("");
1187 lex_mode_stack_.push_back(lex->mode());
1188 }
1189
1190 // Return the file name.
1191 const char*
1192 filename() const
1193 { return this->filename_; }
1194
1195 // Return the position dependent options. The caller may modify
1196 // this.
1197 Position_dependent_options&
1198 position_dependent_options()
1199 { return this->posdep_options_; }
1200
1201 // Whether we are parsing a --defsym.
1202 bool
1203 parsing_defsym() const
1204 { return this->parsing_defsym_; }
1205
1206 // Return whether this script is being run in a group.
1207 bool
1208 in_group() const
1209 { return this->in_group_; }
1210
1211 // Return whether this script was found using a directory in the
1212 // sysroot.
1213 bool
1214 is_in_sysroot() const
1215 { return this->is_in_sysroot_; }
1216
1217 // Whether to skip to the next file with the same name if we find an
1218 // incompatible target in an OUTPUT_FORMAT statement.
1219 bool
1220 skip_on_incompatible_target() const
1221 { return this->skip_on_incompatible_target_; }
1222
1223 // Stop skipping to the next file on an incompatible target. This
1224 // is called when we make some unrevocable change to the data
1225 // structures.
1226 void
1227 clear_skip_on_incompatible_target()
1228 { this->skip_on_incompatible_target_ = false; }
1229
1230 // Whether we found an incompatible target in an OUTPUT_FORMAT
1231 // statement.
1232 bool
1233 found_incompatible_target() const
1234 { return this->found_incompatible_target_; }
1235
1236 // Note that we found an incompatible target.
1237 void
1238 set_found_incompatible_target()
1239 { this->found_incompatible_target_ = true; }
1240
1241 // Returns the Command_line structure passed in at constructor time.
1242 // This value may be NULL. The caller may modify this, which modifies
1243 // the passed-in Command_line object (not a copy).
1244 Command_line*
1245 command_line()
1246 { return this->command_line_; }
1247
1248 // Return the options which may be set by a script.
1249 Script_options*
1250 script_options()
1251 { return this->script_options_; }
1252
1253 // Return the object in which version script information should be stored.
1254 Version_script_info*
1255 version_script()
1256 { return this->version_script_info_; }
1257
1258 // Return the next token, and advance.
1259 const Token*
1260 next_token()
1261 {
1262 const Token* token = this->lex_->next_token();
1263 this->lineno_ = token->lineno();
1264 this->charpos_ = token->charpos();
1265 return token;
1266 }
1267
1268 // Set a new lexer mode, pushing the current one.
1269 void
1270 push_lex_mode(Lex::Mode mode)
1271 {
1272 this->lex_mode_stack_.push_back(this->lex_->mode());
1273 this->lex_->set_mode(mode);
1274 }
1275
1276 // Pop the lexer mode.
1277 void
1278 pop_lex_mode()
1279 {
1280 gold_assert(!this->lex_mode_stack_.empty());
1281 this->lex_->set_mode(this->lex_mode_stack_.back());
1282 this->lex_mode_stack_.pop_back();
1283 }
1284
1285 // Return the current lexer mode.
1286 Lex::Mode
1287 lex_mode() const
1288 { return this->lex_mode_stack_.back(); }
1289
1290 // Return the line number of the last token.
1291 int
1292 lineno() const
1293 { return this->lineno_; }
1294
1295 // Return the character position in the line of the last token.
1296 int
1297 charpos() const
1298 { return this->charpos_; }
1299
1300 // Return the list of input files, creating it if necessary. This
1301 // is a space leak--we never free the INPUTS_ pointer.
1302 Input_arguments*
1303 inputs()
1304 {
1305 if (this->inputs_ == NULL)
1306 this->inputs_ = new Input_arguments();
1307 return this->inputs_;
1308 }
1309
1310 // Return whether we saw any input files.
1311 bool
1312 saw_inputs() const
1313 { return this->inputs_ != NULL && !this->inputs_->empty(); }
1314
1315 // Return the current language being processed in a version script
1316 // (eg, "C++"). The empty string represents unmangled C names.
1317 const std::string&
1318 get_current_language() const
1319 { return this->language_stack_.back(); }
1320
1321 // Push a language onto the stack when entering an extern block.
1322 void push_language(const std::string& lang)
1323 { this->language_stack_.push_back(lang); }
1324
1325 // Pop a language off of the stack when exiting an extern block.
1326 void pop_language()
1327 {
1328 gold_assert(!this->language_stack_.empty());
1329 this->language_stack_.pop_back();
1330 }
1331
1332 private:
1333 // The name of the file we are reading.
1334 const char* filename_;
1335 // The position dependent options.
1336 Position_dependent_options posdep_options_;
1337 // True if we are parsing a --defsym.
1338 bool parsing_defsym_;
1339 // Whether we are currently in a --start-group/--end-group.
1340 bool in_group_;
1341 // Whether the script was found in a sysrooted directory.
1342 bool is_in_sysroot_;
1343 // If this is true, then if we find an OUTPUT_FORMAT with an
1344 // incompatible target, then we tell the parser to abort so that we
1345 // can search for the next file with the same name.
1346 bool skip_on_incompatible_target_;
1347 // True if we found an OUTPUT_FORMAT with an incompatible target.
1348 bool found_incompatible_target_;
1349 // May be NULL if the user chooses not to pass one in.
1350 Command_line* command_line_;
1351 // Options which may be set from any linker script.
1352 Script_options* script_options_;
1353 // Information parsed from a version script.
1354 Version_script_info* version_script_info_;
1355 // The lexer.
1356 Lex* lex_;
1357 // The line number of the last token returned by next_token.
1358 int lineno_;
1359 // The column number of the last token returned by next_token.
1360 int charpos_;
1361 // A stack of lexer modes.
1362 std::vector<Lex::Mode> lex_mode_stack_;
1363 // A stack of which extern/language block we're inside. Can be C++,
1364 // java, or empty for C.
1365 std::vector<std::string> language_stack_;
1366 // New input files found to add to the link.
1367 Input_arguments* inputs_;
1368 };
1369
1370 // FILE was found as an argument on the command line. Try to read it
1371 // as a script. Return true if the file was handled.
1372
1373 bool
1374 read_input_script(Workqueue* workqueue, Symbol_table* symtab, Layout* layout,
1375 Dirsearch* dirsearch, int dirindex,
1376 Input_objects* input_objects, Mapfile* mapfile,
1377 Input_group* input_group,
1378 const Input_argument* input_argument,
1379 Input_file* input_file, Task_token* next_blocker,
1380 bool* used_next_blocker)
1381 {
1382 *used_next_blocker = false;
1383
1384 std::string input_string;
1385 Lex::read_file(input_file, &input_string);
1386
1387 Lex lex(input_string.c_str(), input_string.length(), PARSING_LINKER_SCRIPT);
1388
1389 Parser_closure closure(input_file->filename().c_str(),
1390 input_argument->file().options(),
1391 false,
1392 input_group != NULL,
1393 input_file->is_in_sysroot(),
1394 NULL,
1395 layout->script_options(),
1396 &lex,
1397 input_file->will_search_for());
1398
1399 if (yyparse(&closure) != 0)
1400 {
1401 if (closure.found_incompatible_target())
1402 {
1403 Read_symbols::incompatible_warning(input_argument, input_file);
1404 Read_symbols::requeue(workqueue, input_objects, symtab, layout,
1405 dirsearch, dirindex, mapfile, input_argument,
1406 input_group, next_blocker);
1407 return true;
1408 }
1409 return false;
1410 }
1411
1412 if (!closure.saw_inputs())
1413 return true;
1414
1415 Task_token* this_blocker = NULL;
1416 for (Input_arguments::const_iterator p = closure.inputs()->begin();
1417 p != closure.inputs()->end();
1418 ++p)
1419 {
1420 Task_token* nb;
1421 if (p + 1 == closure.inputs()->end())
1422 nb = next_blocker;
1423 else
1424 {
1425 nb = new Task_token(true);
1426 nb->add_blocker();
1427 }
1428 workqueue->queue_soon(new Read_symbols(input_objects, symtab,
1429 layout, dirsearch, 0, mapfile, &*p,
1430 input_group, this_blocker, nb));
1431 this_blocker = nb;
1432 }
1433
1434 if (layout->incremental_inputs())
1435 {
1436 // Like new Read_symbols(...) above, we rely on close.inputs()
1437 // getting leaked by closure.
1438 Script_info* info = new Script_info(closure.inputs());
1439 layout->incremental_inputs()->report_script(
1440 input_argument,
1441 input_file->file().get_mtime(),
1442 info);
1443 }
1444 *used_next_blocker = true;
1445
1446 return true;
1447 }
1448
1449 // Helper function for read_version_script() and
1450 // read_commandline_script(). Processes the given file in the mode
1451 // indicated by first_token and lex_mode.
1452
1453 static bool
1454 read_script_file(const char* filename, Command_line* cmdline,
1455 Script_options* script_options,
1456 int first_token, Lex::Mode lex_mode)
1457 {
1458 // TODO: if filename is a relative filename, search for it manually
1459 // using "." + cmdline->options()->search_path() -- not dirsearch.
1460 Dirsearch dirsearch;
1461
1462 // The file locking code wants to record a Task, but we haven't
1463 // started the workqueue yet. This is only for debugging purposes,
1464 // so we invent a fake value.
1465 const Task* task = reinterpret_cast<const Task*>(-1);
1466
1467 // We don't want this file to be opened in binary mode.
1468 Position_dependent_options posdep = cmdline->position_dependent_options();
1469 if (posdep.format_enum() == General_options::OBJECT_FORMAT_BINARY)
1470 posdep.set_format_enum(General_options::OBJECT_FORMAT_ELF);
1471 Input_file_argument input_argument(filename,
1472 Input_file_argument::INPUT_FILE_TYPE_FILE,
1473 "", false, posdep);
1474 Input_file input_file(&input_argument);
1475 int dummy = 0;
1476 if (!input_file.open(dirsearch, task, &dummy))
1477 return false;
1478
1479 std::string input_string;
1480 Lex::read_file(&input_file, &input_string);
1481
1482 Lex lex(input_string.c_str(), input_string.length(), first_token);
1483 lex.set_mode(lex_mode);
1484
1485 Parser_closure closure(filename,
1486 cmdline->position_dependent_options(),
1487 first_token == Lex::DYNAMIC_LIST,
1488 false,
1489 input_file.is_in_sysroot(),
1490 cmdline,
1491 script_options,
1492 &lex,
1493 false);
1494 if (yyparse(&closure) != 0)
1495 {
1496 input_file.file().unlock(task);
1497 return false;
1498 }
1499
1500 input_file.file().unlock(task);
1501
1502 gold_assert(!closure.saw_inputs());
1503
1504 return true;
1505 }
1506
1507 // FILENAME was found as an argument to --script (-T).
1508 // Read it as a script, and execute its contents immediately.
1509
1510 bool
1511 read_commandline_script(const char* filename, Command_line* cmdline)
1512 {
1513 return read_script_file(filename, cmdline, &cmdline->script_options(),
1514 PARSING_LINKER_SCRIPT, Lex::LINKER_SCRIPT);
1515 }
1516
1517 // FILENAME was found as an argument to --version-script. Read it as
1518 // a version script, and store its contents in
1519 // cmdline->script_options()->version_script_info().
1520
1521 bool
1522 read_version_script(const char* filename, Command_line* cmdline)
1523 {
1524 return read_script_file(filename, cmdline, &cmdline->script_options(),
1525 PARSING_VERSION_SCRIPT, Lex::VERSION_SCRIPT);
1526 }
1527
1528 // FILENAME was found as an argument to --dynamic-list. Read it as a
1529 // list of symbols, and store its contents in DYNAMIC_LIST.
1530
1531 bool
1532 read_dynamic_list(const char* filename, Command_line* cmdline,
1533 Script_options* dynamic_list)
1534 {
1535 return read_script_file(filename, cmdline, dynamic_list,
1536 PARSING_DYNAMIC_LIST, Lex::DYNAMIC_LIST);
1537 }
1538
1539 // Implement the --defsym option on the command line. Return true if
1540 // all is well.
1541
1542 bool
1543 Script_options::define_symbol(const char* definition)
1544 {
1545 Lex lex(definition, strlen(definition), PARSING_DEFSYM);
1546 lex.set_mode(Lex::EXPRESSION);
1547
1548 // Dummy value.
1549 Position_dependent_options posdep_options;
1550
1551 Parser_closure closure("command line", posdep_options, true,
1552 false, false, NULL, this, &lex, false);
1553
1554 if (yyparse(&closure) != 0)
1555 return false;
1556
1557 gold_assert(!closure.saw_inputs());
1558
1559 return true;
1560 }
1561
1562 // Print the script to F for debugging.
1563
1564 void
1565 Script_options::print(FILE* f) const
1566 {
1567 fprintf(f, "%s: Dumping linker script\n", program_name);
1568
1569 if (!this->entry_.empty())
1570 fprintf(f, "ENTRY(%s)\n", this->entry_.c_str());
1571
1572 for (Symbol_assignments::const_iterator p =
1573 this->symbol_assignments_.begin();
1574 p != this->symbol_assignments_.end();
1575 ++p)
1576 (*p)->print(f);
1577
1578 for (Assertions::const_iterator p = this->assertions_.begin();
1579 p != this->assertions_.end();
1580 ++p)
1581 (*p)->print(f);
1582
1583 this->script_sections_.print(f);
1584
1585 this->version_script_info_.print(f);
1586 }
1587
1588 // Manage mapping from keywords to the codes expected by the bison
1589 // parser. We construct one global object for each lex mode with
1590 // keywords.
1591
1592 class Keyword_to_parsecode
1593 {
1594 public:
1595 // The structure which maps keywords to parsecodes.
1596 struct Keyword_parsecode
1597 {
1598 // Keyword.
1599 const char* keyword;
1600 // Corresponding parsecode.
1601 int parsecode;
1602 };
1603
1604 Keyword_to_parsecode(const Keyword_parsecode* keywords,
1605 int keyword_count)
1606 : keyword_parsecodes_(keywords), keyword_count_(keyword_count)
1607 { }
1608
1609 // Return the parsecode corresponding KEYWORD, or 0 if it is not a
1610 // keyword.
1611 int
1612 keyword_to_parsecode(const char* keyword, size_t len) const;
1613
1614 private:
1615 const Keyword_parsecode* keyword_parsecodes_;
1616 const int keyword_count_;
1617 };
1618
1619 // Mapping from keyword string to keyword parsecode. This array must
1620 // be kept in sorted order. Parsecodes are looked up using bsearch.
1621 // This array must correspond to the list of parsecodes in yyscript.y.
1622
1623 static const Keyword_to_parsecode::Keyword_parsecode
1624 script_keyword_parsecodes[] =
1625 {
1626 { "ABSOLUTE", ABSOLUTE },
1627 { "ADDR", ADDR },
1628 { "ALIGN", ALIGN_K },
1629 { "ALIGNOF", ALIGNOF },
1630 { "ASSERT", ASSERT_K },
1631 { "AS_NEEDED", AS_NEEDED },
1632 { "AT", AT },
1633 { "BIND", BIND },
1634 { "BLOCK", BLOCK },
1635 { "BYTE", BYTE },
1636 { "CONSTANT", CONSTANT },
1637 { "CONSTRUCTORS", CONSTRUCTORS },
1638 { "CREATE_OBJECT_SYMBOLS", CREATE_OBJECT_SYMBOLS },
1639 { "DATA_SEGMENT_ALIGN", DATA_SEGMENT_ALIGN },
1640 { "DATA_SEGMENT_END", DATA_SEGMENT_END },
1641 { "DATA_SEGMENT_RELRO_END", DATA_SEGMENT_RELRO_END },
1642 { "DEFINED", DEFINED },
1643 { "ENTRY", ENTRY },
1644 { "EXCLUDE_FILE", EXCLUDE_FILE },
1645 { "EXTERN", EXTERN },
1646 { "FILL", FILL },
1647 { "FLOAT", FLOAT },
1648 { "FORCE_COMMON_ALLOCATION", FORCE_COMMON_ALLOCATION },
1649 { "GROUP", GROUP },
1650 { "HLL", HLL },
1651 { "INCLUDE", INCLUDE },
1652 { "INHIBIT_COMMON_ALLOCATION", INHIBIT_COMMON_ALLOCATION },
1653 { "INPUT", INPUT },
1654 { "KEEP", KEEP },
1655 { "LENGTH", LENGTH },
1656 { "LOADADDR", LOADADDR },
1657 { "LONG", LONG },
1658 { "MAP", MAP },
1659 { "MAX", MAX_K },
1660 { "MEMORY", MEMORY },
1661 { "MIN", MIN_K },
1662 { "NEXT", NEXT },
1663 { "NOCROSSREFS", NOCROSSREFS },
1664 { "NOFLOAT", NOFLOAT },
1665 { "ONLY_IF_RO", ONLY_IF_RO },
1666 { "ONLY_IF_RW", ONLY_IF_RW },
1667 { "OPTION", OPTION },
1668 { "ORIGIN", ORIGIN },
1669 { "OUTPUT", OUTPUT },
1670 { "OUTPUT_ARCH", OUTPUT_ARCH },
1671 { "OUTPUT_FORMAT", OUTPUT_FORMAT },
1672 { "OVERLAY", OVERLAY },
1673 { "PHDRS", PHDRS },
1674 { "PROVIDE", PROVIDE },
1675 { "PROVIDE_HIDDEN", PROVIDE_HIDDEN },
1676 { "QUAD", QUAD },
1677 { "SEARCH_DIR", SEARCH_DIR },
1678 { "SECTIONS", SECTIONS },
1679 { "SEGMENT_START", SEGMENT_START },
1680 { "SHORT", SHORT },
1681 { "SIZEOF", SIZEOF },
1682 { "SIZEOF_HEADERS", SIZEOF_HEADERS },
1683 { "SORT", SORT_BY_NAME },
1684 { "SORT_BY_ALIGNMENT", SORT_BY_ALIGNMENT },
1685 { "SORT_BY_NAME", SORT_BY_NAME },
1686 { "SPECIAL", SPECIAL },
1687 { "SQUAD", SQUAD },
1688 { "STARTUP", STARTUP },
1689 { "SUBALIGN", SUBALIGN },
1690 { "SYSLIB", SYSLIB },
1691 { "TARGET", TARGET_K },
1692 { "TRUNCATE", TRUNCATE },
1693 { "VERSION", VERSIONK },
1694 { "global", GLOBAL },
1695 { "l", LENGTH },
1696 { "len", LENGTH },
1697 { "local", LOCAL },
1698 { "o", ORIGIN },
1699 { "org", ORIGIN },
1700 { "sizeof_headers", SIZEOF_HEADERS },
1701 };
1702
1703 static const Keyword_to_parsecode
1704 script_keywords(&script_keyword_parsecodes[0],
1705 (sizeof(script_keyword_parsecodes)
1706 / sizeof(script_keyword_parsecodes[0])));
1707
1708 static const Keyword_to_parsecode::Keyword_parsecode
1709 version_script_keyword_parsecodes[] =
1710 {
1711 { "extern", EXTERN },
1712 { "global", GLOBAL },
1713 { "local", LOCAL },
1714 };
1715
1716 static const Keyword_to_parsecode
1717 version_script_keywords(&version_script_keyword_parsecodes[0],
1718 (sizeof(version_script_keyword_parsecodes)
1719 / sizeof(version_script_keyword_parsecodes[0])));
1720
1721 static const Keyword_to_parsecode::Keyword_parsecode
1722 dynamic_list_keyword_parsecodes[] =
1723 {
1724 { "extern", EXTERN },
1725 };
1726
1727 static const Keyword_to_parsecode
1728 dynamic_list_keywords(&dynamic_list_keyword_parsecodes[0],
1729 (sizeof(dynamic_list_keyword_parsecodes)
1730 / sizeof(dynamic_list_keyword_parsecodes[0])));
1731
1732
1733
1734 // Comparison function passed to bsearch.
1735
1736 extern "C"
1737 {
1738
1739 struct Ktt_key
1740 {
1741 const char* str;
1742 size_t len;
1743 };
1744
1745 static int
1746 ktt_compare(const void* keyv, const void* kttv)
1747 {
1748 const Ktt_key* key = static_cast<const Ktt_key*>(keyv);
1749 const Keyword_to_parsecode::Keyword_parsecode* ktt =
1750 static_cast<const Keyword_to_parsecode::Keyword_parsecode*>(kttv);
1751 int i = strncmp(key->str, ktt->keyword, key->len);
1752 if (i != 0)
1753 return i;
1754 if (ktt->keyword[key->len] != '\0')
1755 return -1;
1756 return 0;
1757 }
1758
1759 } // End extern "C".
1760
1761 int
1762 Keyword_to_parsecode::keyword_to_parsecode(const char* keyword,
1763 size_t len) const
1764 {
1765 Ktt_key key;
1766 key.str = keyword;
1767 key.len = len;
1768 void* kttv = bsearch(&key,
1769 this->keyword_parsecodes_,
1770 this->keyword_count_,
1771 sizeof(this->keyword_parsecodes_[0]),
1772 ktt_compare);
1773 if (kttv == NULL)
1774 return 0;
1775 Keyword_parsecode* ktt = static_cast<Keyword_parsecode*>(kttv);
1776 return ktt->parsecode;
1777 }
1778
1779 // Helper class that calls cplus_demangle when needed and takes care of freeing
1780 // the result.
1781
1782 class Lazy_demangler
1783 {
1784 public:
1785 Lazy_demangler(const char* symbol, int options)
1786 : symbol_(symbol), options_(options), demangled_(NULL), did_demangle_(false)
1787 { }
1788
1789 ~Lazy_demangler()
1790 { free(this->demangled_); }
1791
1792 // Return the demangled name. The actual demangling happens on the first call,
1793 // and the result is later cached.
1794
1795 inline char*
1796 get();
1797
1798 private:
1799 // The symbol to demangle.
1800 const char *symbol_;
1801 // Option flags to pass to cplus_demagle.
1802 const int options_;
1803 // The cached demangled value, or NULL if demangling didn't happen yet or
1804 // failed.
1805 char *demangled_;
1806 // Whether we already called cplus_demangle
1807 bool did_demangle_;
1808 };
1809
1810 // Return the demangled name. The actual demangling happens on the first call,
1811 // and the result is later cached. Returns NULL if the symbol cannot be
1812 // demangled.
1813
1814 inline char*
1815 Lazy_demangler::get()
1816 {
1817 if (!this->did_demangle_)
1818 {
1819 this->demangled_ = cplus_demangle(this->symbol_, this->options_);
1820 this->did_demangle_ = true;
1821 }
1822 return this->demangled_;
1823 }
1824
1825 // The following structs are used within the VersionInfo class as well
1826 // as in the bison helper functions. They store the information
1827 // parsed from the version script.
1828
1829 // A single version expression.
1830 // For example, pattern="std::map*" and language="C++".
1831 // pattern and language should be from the stringpool
1832 struct Version_expression {
1833 Version_expression(const std::string& pattern,
1834 const std::string& language,
1835 bool exact_match)
1836 : pattern(pattern), language(language), exact_match(exact_match) {}
1837
1838 std::string pattern;
1839 std::string language;
1840 // If false, we use glob() to match pattern. If true, we use strcmp().
1841 bool exact_match;
1842 };
1843
1844
1845 // A list of expressions.
1846 struct Version_expression_list {
1847 std::vector<struct Version_expression> expressions;
1848 };
1849
1850
1851 // A list of which versions upon which another version depends.
1852 // Strings should be from the Stringpool.
1853 struct Version_dependency_list {
1854 std::vector<std::string> dependencies;
1855 };
1856
1857
1858 // The total definition of a version. It includes the tag for the
1859 // version, its global and local expressions, and any dependencies.
1860 struct Version_tree {
1861 Version_tree()
1862 : tag(), global(NULL), local(NULL), dependencies(NULL) {}
1863
1864 std::string tag;
1865 const struct Version_expression_list* global;
1866 const struct Version_expression_list* local;
1867 const struct Version_dependency_list* dependencies;
1868 };
1869
1870 Version_script_info::~Version_script_info()
1871 {
1872 this->clear();
1873 }
1874
1875 void
1876 Version_script_info::clear()
1877 {
1878 for (size_t k = 0; k < dependency_lists_.size(); ++k)
1879 delete dependency_lists_[k];
1880 this->dependency_lists_.clear();
1881 for (size_t k = 0; k < version_trees_.size(); ++k)
1882 delete version_trees_[k];
1883 this->version_trees_.clear();
1884 for (size_t k = 0; k < expression_lists_.size(); ++k)
1885 delete expression_lists_[k];
1886 this->expression_lists_.clear();
1887 }
1888
1889 std::vector<std::string>
1890 Version_script_info::get_versions() const
1891 {
1892 std::vector<std::string> ret;
1893 for (size_t j = 0; j < version_trees_.size(); ++j)
1894 if (!this->version_trees_[j]->tag.empty())
1895 ret.push_back(this->version_trees_[j]->tag);
1896 return ret;
1897 }
1898
1899 std::vector<std::string>
1900 Version_script_info::get_dependencies(const char* version) const
1901 {
1902 std::vector<std::string> ret;
1903 for (size_t j = 0; j < version_trees_.size(); ++j)
1904 if (version_trees_[j]->tag == version)
1905 {
1906 const struct Version_dependency_list* deps =
1907 version_trees_[j]->dependencies;
1908 if (deps != NULL)
1909 for (size_t k = 0; k < deps->dependencies.size(); ++k)
1910 ret.push_back(deps->dependencies[k]);
1911 return ret;
1912 }
1913 return ret;
1914 }
1915
1916 // Look up SYMBOL_NAME in the list of versions. If CHECK_GLOBAL is
1917 // true look at the globally visible symbols, otherwise look at the
1918 // symbols listed as "local:". Return true if the symbol is found,
1919 // false otherwise. If the symbol is found, then if PVERSION is not
1920 // NULL, set *PVERSION to the version.
1921
1922 bool
1923 Version_script_info::get_symbol_version_helper(const char* symbol_name,
1924 bool check_global,
1925 std::string* pversion) const
1926 {
1927 Lazy_demangler cpp_demangled_name(symbol_name, DMGL_ANSI | DMGL_PARAMS);
1928 Lazy_demangler java_demangled_name(symbol_name,
1929 DMGL_ANSI | DMGL_PARAMS | DMGL_JAVA);
1930 for (size_t j = 0; j < version_trees_.size(); ++j)
1931 {
1932 // Is it a global symbol for this version?
1933 const Version_expression_list* explist =
1934 check_global ? version_trees_[j]->global : version_trees_[j]->local;
1935 if (explist != NULL)
1936 for (size_t k = 0; k < explist->expressions.size(); ++k)
1937 {
1938 const char* name_to_match = symbol_name;
1939 const struct Version_expression& exp = explist->expressions[k];
1940 if (exp.language == "C++")
1941 {
1942 name_to_match = cpp_demangled_name.get();
1943 // This isn't a C++ symbol.
1944 if (name_to_match == NULL)
1945 continue;
1946 }
1947 else if (exp.language == "Java")
1948 {
1949 name_to_match = java_demangled_name.get();
1950 // This isn't a Java symbol.
1951 if (name_to_match == NULL)
1952 continue;
1953 }
1954 bool matched;
1955 if (exp.exact_match)
1956 matched = strcmp(exp.pattern.c_str(), name_to_match) == 0;
1957 else
1958 matched = fnmatch(exp.pattern.c_str(), name_to_match,
1959 FNM_NOESCAPE) == 0;
1960 if (matched)
1961 {
1962 if (pversion != NULL)
1963 *pversion = this->version_trees_[j]->tag;
1964 return true;
1965 }
1966 }
1967 }
1968 return false;
1969 }
1970
1971 struct Version_dependency_list*
1972 Version_script_info::allocate_dependency_list()
1973 {
1974 dependency_lists_.push_back(new Version_dependency_list);
1975 return dependency_lists_.back();
1976 }
1977
1978 struct Version_expression_list*
1979 Version_script_info::allocate_expression_list()
1980 {
1981 expression_lists_.push_back(new Version_expression_list);
1982 return expression_lists_.back();
1983 }
1984
1985 struct Version_tree*
1986 Version_script_info::allocate_version_tree()
1987 {
1988 version_trees_.push_back(new Version_tree);
1989 return version_trees_.back();
1990 }
1991
1992 // Print for debugging.
1993
1994 void
1995 Version_script_info::print(FILE* f) const
1996 {
1997 if (this->empty())
1998 return;
1999
2000 fprintf(f, "VERSION {");
2001
2002 for (size_t i = 0; i < this->version_trees_.size(); ++i)
2003 {
2004 const Version_tree* vt = this->version_trees_[i];
2005
2006 if (vt->tag.empty())
2007 fprintf(f, " {\n");
2008 else
2009 fprintf(f, " %s {\n", vt->tag.c_str());
2010
2011 if (vt->global != NULL)
2012 {
2013 fprintf(f, " global :\n");
2014 this->print_expression_list(f, vt->global);
2015 }
2016
2017 if (vt->local != NULL)
2018 {
2019 fprintf(f, " local :\n");
2020 this->print_expression_list(f, vt->local);
2021 }
2022
2023 fprintf(f, " }");
2024 if (vt->dependencies != NULL)
2025 {
2026 const Version_dependency_list* deps = vt->dependencies;
2027 for (size_t j = 0; j < deps->dependencies.size(); ++j)
2028 {
2029 if (j < deps->dependencies.size() - 1)
2030 fprintf(f, "\n");
2031 fprintf(f, " %s", deps->dependencies[j].c_str());
2032 }
2033 }
2034 fprintf(f, ";\n");
2035 }
2036
2037 fprintf(f, "}\n");
2038 }
2039
2040 void
2041 Version_script_info::print_expression_list(
2042 FILE* f,
2043 const Version_expression_list* vel) const
2044 {
2045 std::string current_language;
2046 for (size_t i = 0; i < vel->expressions.size(); ++i)
2047 {
2048 const Version_expression& ve(vel->expressions[i]);
2049
2050 if (ve.language != current_language)
2051 {
2052 if (!current_language.empty())
2053 fprintf(f, " }\n");
2054 fprintf(f, " extern \"%s\" {\n", ve.language.c_str());
2055 current_language = ve.language;
2056 }
2057
2058 fprintf(f, " ");
2059 if (!current_language.empty())
2060 fprintf(f, " ");
2061
2062 if (ve.exact_match)
2063 fprintf(f, "\"");
2064 fprintf(f, "%s", ve.pattern.c_str());
2065 if (ve.exact_match)
2066 fprintf(f, "\"");
2067
2068 fprintf(f, "\n");
2069 }
2070
2071 if (!current_language.empty())
2072 fprintf(f, " }\n");
2073 }
2074
2075 } // End namespace gold.
2076
2077 // The remaining functions are extern "C", so it's clearer to not put
2078 // them in namespace gold.
2079
2080 using namespace gold;
2081
2082 // This function is called by the bison parser to return the next
2083 // token.
2084
2085 extern "C" int
2086 yylex(YYSTYPE* lvalp, void* closurev)
2087 {
2088 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2089 const Token* token = closure->next_token();
2090 switch (token->classification())
2091 {
2092 default:
2093 gold_unreachable();
2094
2095 case Token::TOKEN_INVALID:
2096 yyerror(closurev, "invalid character");
2097 return 0;
2098
2099 case Token::TOKEN_EOF:
2100 return 0;
2101
2102 case Token::TOKEN_STRING:
2103 {
2104 // This is either a keyword or a STRING.
2105 size_t len;
2106 const char* str = token->string_value(&len);
2107 int parsecode = 0;
2108 switch (closure->lex_mode())
2109 {
2110 case Lex::LINKER_SCRIPT:
2111 parsecode = script_keywords.keyword_to_parsecode(str, len);
2112 break;
2113 case Lex::VERSION_SCRIPT:
2114 parsecode = version_script_keywords.keyword_to_parsecode(str, len);
2115 break;
2116 case Lex::DYNAMIC_LIST:
2117 parsecode = dynamic_list_keywords.keyword_to_parsecode(str, len);
2118 break;
2119 default:
2120 break;
2121 }
2122 if (parsecode != 0)
2123 return parsecode;
2124 lvalp->string.value = str;
2125 lvalp->string.length = len;
2126 return STRING;
2127 }
2128
2129 case Token::TOKEN_QUOTED_STRING:
2130 lvalp->string.value = token->string_value(&lvalp->string.length);
2131 return QUOTED_STRING;
2132
2133 case Token::TOKEN_OPERATOR:
2134 return token->operator_value();
2135
2136 case Token::TOKEN_INTEGER:
2137 lvalp->integer = token->integer_value();
2138 return INTEGER;
2139 }
2140 }
2141
2142 // This function is called by the bison parser to report an error.
2143
2144 extern "C" void
2145 yyerror(void* closurev, const char* message)
2146 {
2147 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2148 gold_error(_("%s:%d:%d: %s"), closure->filename(), closure->lineno(),
2149 closure->charpos(), message);
2150 }
2151
2152 // Called by the bison parser to add an external symbol to the link.
2153
2154 extern "C" void
2155 script_add_extern(void* closurev, const char* name, size_t length)
2156 {
2157 // We treat exactly like -u NAME. FIXME: If it seems useful, we
2158 // could handle this after the command line has been read, by adding
2159 // entries to the symbol table directly.
2160 std::string arg("--undefined=");
2161 arg.append(name, length);
2162 script_parse_option(closurev, arg.c_str(), arg.size());
2163 }
2164
2165 // Called by the bison parser to add a file to the link.
2166
2167 extern "C" void
2168 script_add_file(void* closurev, const char* name, size_t length)
2169 {
2170 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2171
2172 // If this is an absolute path, and we found the script in the
2173 // sysroot, then we want to prepend the sysroot to the file name.
2174 // For example, this is how we handle a cross link to the x86_64
2175 // libc.so, which refers to /lib/libc.so.6.
2176 std::string name_string(name, length);
2177 const char* extra_search_path = ".";
2178 std::string script_directory;
2179 if (IS_ABSOLUTE_PATH(name_string.c_str()))
2180 {
2181 if (closure->is_in_sysroot())
2182 {
2183 const std::string& sysroot(parameters->options().sysroot());
2184 gold_assert(!sysroot.empty());
2185 name_string = sysroot + name_string;
2186 }
2187 }
2188 else
2189 {
2190 // In addition to checking the normal library search path, we
2191 // also want to check in the script-directory.
2192 const char *slash = strrchr(closure->filename(), '/');
2193 if (slash != NULL)
2194 {
2195 script_directory.assign(closure->filename(),
2196 slash - closure->filename() + 1);
2197 extra_search_path = script_directory.c_str();
2198 }
2199 }
2200
2201 Input_file_argument file(name_string.c_str(),
2202 Input_file_argument::INPUT_FILE_TYPE_FILE,
2203 extra_search_path, false,
2204 closure->position_dependent_options());
2205 closure->inputs()->add_file(file);
2206 }
2207
2208 // Called by the bison parser to start a group. If we are already in
2209 // a group, that means that this script was invoked within a
2210 // --start-group --end-group sequence on the command line, or that
2211 // this script was found in a GROUP of another script. In that case,
2212 // we simply continue the existing group, rather than starting a new
2213 // one. It is possible to construct a case in which this will do
2214 // something other than what would happen if we did a recursive group,
2215 // but it's hard to imagine why the different behaviour would be
2216 // useful for a real program. Avoiding recursive groups is simpler
2217 // and more efficient.
2218
2219 extern "C" void
2220 script_start_group(void* closurev)
2221 {
2222 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2223 if (!closure->in_group())
2224 closure->inputs()->start_group();
2225 }
2226
2227 // Called by the bison parser at the end of a group.
2228
2229 extern "C" void
2230 script_end_group(void* closurev)
2231 {
2232 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2233 if (!closure->in_group())
2234 closure->inputs()->end_group();
2235 }
2236
2237 // Called by the bison parser to start an AS_NEEDED list.
2238
2239 extern "C" void
2240 script_start_as_needed(void* closurev)
2241 {
2242 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2243 closure->position_dependent_options().set_as_needed(true);
2244 }
2245
2246 // Called by the bison parser at the end of an AS_NEEDED list.
2247
2248 extern "C" void
2249 script_end_as_needed(void* closurev)
2250 {
2251 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2252 closure->position_dependent_options().set_as_needed(false);
2253 }
2254
2255 // Called by the bison parser to set the entry symbol.
2256
2257 extern "C" void
2258 script_set_entry(void* closurev, const char* entry, size_t length)
2259 {
2260 // We'll parse this exactly the same as --entry=ENTRY on the commandline
2261 // TODO(csilvers): FIXME -- call set_entry directly.
2262 std::string arg("--entry=");
2263 arg.append(entry, length);
2264 script_parse_option(closurev, arg.c_str(), arg.size());
2265 }
2266
2267 // Called by the bison parser to set whether to define common symbols.
2268
2269 extern "C" void
2270 script_set_common_allocation(void* closurev, int set)
2271 {
2272 const char* arg = set != 0 ? "--define-common" : "--no-define-common";
2273 script_parse_option(closurev, arg, strlen(arg));
2274 }
2275
2276 // Called by the bison parser to define a symbol.
2277
2278 extern "C" void
2279 script_set_symbol(void* closurev, const char* name, size_t length,
2280 Expression* value, int providei, int hiddeni)
2281 {
2282 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2283 const bool provide = providei != 0;
2284 const bool hidden = hiddeni != 0;
2285 closure->script_options()->add_symbol_assignment(name, length,
2286 closure->parsing_defsym(),
2287 value, provide, hidden);
2288 closure->clear_skip_on_incompatible_target();
2289 }
2290
2291 // Called by the bison parser to add an assertion.
2292
2293 extern "C" void
2294 script_add_assertion(void* closurev, Expression* check, const char* message,
2295 size_t messagelen)
2296 {
2297 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2298 closure->script_options()->add_assertion(check, message, messagelen);
2299 closure->clear_skip_on_incompatible_target();
2300 }
2301
2302 // Called by the bison parser to parse an OPTION.
2303
2304 extern "C" void
2305 script_parse_option(void* closurev, const char* option, size_t length)
2306 {
2307 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2308 // We treat the option as a single command-line option, even if
2309 // it has internal whitespace.
2310 if (closure->command_line() == NULL)
2311 {
2312 // There are some options that we could handle here--e.g.,
2313 // -lLIBRARY. Should we bother?
2314 gold_warning(_("%s:%d:%d: ignoring command OPTION; OPTION is only valid"
2315 " for scripts specified via -T/--script"),
2316 closure->filename(), closure->lineno(), closure->charpos());
2317 }
2318 else
2319 {
2320 bool past_a_double_dash_option = false;
2321 const char* mutable_option = strndup(option, length);
2322 gold_assert(mutable_option != NULL);
2323 closure->command_line()->process_one_option(1, &mutable_option, 0,
2324 &past_a_double_dash_option);
2325 // The General_options class will quite possibly store a pointer
2326 // into mutable_option, so we can't free it. In cases the class
2327 // does not store such a pointer, this is a memory leak. Alas. :(
2328 }
2329 closure->clear_skip_on_incompatible_target();
2330 }
2331
2332 // Called by the bison parser to handle OUTPUT_FORMAT. OUTPUT_FORMAT
2333 // takes either one or three arguments. In the three argument case,
2334 // the format depends on the endianness option, which we don't
2335 // currently support (FIXME). If we see an OUTPUT_FORMAT for the
2336 // wrong format, then we want to search for a new file. Returning 0
2337 // here will cause the parser to immediately abort.
2338
2339 extern "C" int
2340 script_check_output_format(void* closurev,
2341 const char* default_name, size_t default_length,
2342 const char*, size_t, const char*, size_t)
2343 {
2344 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2345 std::string name(default_name, default_length);
2346 Target* target = select_target_by_name(name.c_str());
2347 if (target == NULL || !parameters->is_compatible_target(target))
2348 {
2349 if (closure->skip_on_incompatible_target())
2350 {
2351 closure->set_found_incompatible_target();
2352 return 0;
2353 }
2354 // FIXME: Should we warn about the unknown target?
2355 }
2356 return 1;
2357 }
2358
2359 // Called by the bison parser to handle TARGET.
2360
2361 extern "C" void
2362 script_set_target(void* closurev, const char* target, size_t len)
2363 {
2364 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2365 std::string s(target, len);
2366 General_options::Object_format format_enum;
2367 format_enum = General_options::string_to_object_format(s.c_str());
2368 closure->position_dependent_options().set_format_enum(format_enum);
2369 }
2370
2371 // Called by the bison parser to handle SEARCH_DIR. This is handled
2372 // exactly like a -L option.
2373
2374 extern "C" void
2375 script_add_search_dir(void* closurev, const char* option, size_t length)
2376 {
2377 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2378 if (closure->command_line() == NULL)
2379 gold_warning(_("%s:%d:%d: ignoring SEARCH_DIR; SEARCH_DIR is only valid"
2380 " for scripts specified via -T/--script"),
2381 closure->filename(), closure->lineno(), closure->charpos());
2382 else
2383 {
2384 std::string s = "-L" + std::string(option, length);
2385 script_parse_option(closurev, s.c_str(), s.size());
2386 }
2387 }
2388
2389 /* Called by the bison parser to push the lexer into expression
2390 mode. */
2391
2392 extern "C" void
2393 script_push_lex_into_expression_mode(void* closurev)
2394 {
2395 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2396 closure->push_lex_mode(Lex::EXPRESSION);
2397 }
2398
2399 /* Called by the bison parser to push the lexer into version
2400 mode. */
2401
2402 extern "C" void
2403 script_push_lex_into_version_mode(void* closurev)
2404 {
2405 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2406 closure->push_lex_mode(Lex::VERSION_SCRIPT);
2407 }
2408
2409 /* Called by the bison parser to pop the lexer mode. */
2410
2411 extern "C" void
2412 script_pop_lex_mode(void* closurev)
2413 {
2414 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2415 closure->pop_lex_mode();
2416 }
2417
2418 // Register an entire version node. For example:
2419 //
2420 // GLIBC_2.1 {
2421 // global: foo;
2422 // } GLIBC_2.0;
2423 //
2424 // - tag is "GLIBC_2.1"
2425 // - tree contains the information "global: foo"
2426 // - deps contains "GLIBC_2.0"
2427
2428 extern "C" void
2429 script_register_vers_node(void*,
2430 const char* tag,
2431 int taglen,
2432 struct Version_tree *tree,
2433 struct Version_dependency_list *deps)
2434 {
2435 gold_assert(tree != NULL);
2436 tree->dependencies = deps;
2437 if (tag != NULL)
2438 tree->tag = std::string(tag, taglen);
2439 }
2440
2441 // Add a dependencies to the list of existing dependencies, if any,
2442 // and return the expanded list.
2443
2444 extern "C" struct Version_dependency_list *
2445 script_add_vers_depend(void* closurev,
2446 struct Version_dependency_list *all_deps,
2447 const char *depend_to_add, int deplen)
2448 {
2449 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2450 if (all_deps == NULL)
2451 all_deps = closure->version_script()->allocate_dependency_list();
2452 all_deps->dependencies.push_back(std::string(depend_to_add, deplen));
2453 return all_deps;
2454 }
2455
2456 // Add a pattern expression to an existing list of expressions, if any.
2457 // TODO: In the old linker, the last argument used to be a bool, but I
2458 // don't know what it meant.
2459
2460 extern "C" struct Version_expression_list *
2461 script_new_vers_pattern(void* closurev,
2462 struct Version_expression_list *expressions,
2463 const char *pattern, int patlen, int exact_match)
2464 {
2465 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2466 if (expressions == NULL)
2467 expressions = closure->version_script()->allocate_expression_list();
2468 expressions->expressions.push_back(
2469 Version_expression(std::string(pattern, patlen),
2470 closure->get_current_language(),
2471 static_cast<bool>(exact_match)));
2472 return expressions;
2473 }
2474
2475 // Attaches b to the end of a, and clears b. So a = a + b and b = {}.
2476
2477 extern "C" struct Version_expression_list*
2478 script_merge_expressions(struct Version_expression_list *a,
2479 struct Version_expression_list *b)
2480 {
2481 a->expressions.insert(a->expressions.end(),
2482 b->expressions.begin(), b->expressions.end());
2483 // We could delete b and remove it from expressions_lists_, but
2484 // that's a lot of work. This works just as well.
2485 b->expressions.clear();
2486 return a;
2487 }
2488
2489 // Combine the global and local expressions into a a Version_tree.
2490
2491 extern "C" struct Version_tree *
2492 script_new_vers_node(void* closurev,
2493 struct Version_expression_list *global,
2494 struct Version_expression_list *local)
2495 {
2496 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2497 Version_tree* tree = closure->version_script()->allocate_version_tree();
2498 tree->global = global;
2499 tree->local = local;
2500 return tree;
2501 }
2502
2503 // Handle a transition in language, such as at the
2504 // start or end of 'extern "C++"'
2505
2506 extern "C" void
2507 version_script_push_lang(void* closurev, const char* lang, int langlen)
2508 {
2509 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2510 closure->push_language(std::string(lang, langlen));
2511 }
2512
2513 extern "C" void
2514 version_script_pop_lang(void* closurev)
2515 {
2516 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2517 closure->pop_language();
2518 }
2519
2520 // Called by the bison parser to start a SECTIONS clause.
2521
2522 extern "C" void
2523 script_start_sections(void* closurev)
2524 {
2525 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2526 closure->script_options()->script_sections()->start_sections();
2527 closure->clear_skip_on_incompatible_target();
2528 }
2529
2530 // Called by the bison parser to finish a SECTIONS clause.
2531
2532 extern "C" void
2533 script_finish_sections(void* closurev)
2534 {
2535 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2536 closure->script_options()->script_sections()->finish_sections();
2537 }
2538
2539 // Start processing entries for an output section.
2540
2541 extern "C" void
2542 script_start_output_section(void* closurev, const char* name, size_t namelen,
2543 const struct Parser_output_section_header* header)
2544 {
2545 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2546 closure->script_options()->script_sections()->start_output_section(name,
2547 namelen,
2548 header);
2549 }
2550
2551 // Finish processing entries for an output section.
2552
2553 extern "C" void
2554 script_finish_output_section(void* closurev,
2555 const struct Parser_output_section_trailer* trail)
2556 {
2557 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2558 closure->script_options()->script_sections()->finish_output_section(trail);
2559 }
2560
2561 // Add a data item (e.g., "WORD (0)") to the current output section.
2562
2563 extern "C" void
2564 script_add_data(void* closurev, int data_token, Expression* val)
2565 {
2566 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2567 int size;
2568 bool is_signed = true;
2569 switch (data_token)
2570 {
2571 case QUAD:
2572 size = 8;
2573 is_signed = false;
2574 break;
2575 case SQUAD:
2576 size = 8;
2577 break;
2578 case LONG:
2579 size = 4;
2580 break;
2581 case SHORT:
2582 size = 2;
2583 break;
2584 case BYTE:
2585 size = 1;
2586 break;
2587 default:
2588 gold_unreachable();
2589 }
2590 closure->script_options()->script_sections()->add_data(size, is_signed, val);
2591 }
2592
2593 // Add a clause setting the fill value to the current output section.
2594
2595 extern "C" void
2596 script_add_fill(void* closurev, Expression* val)
2597 {
2598 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2599 closure->script_options()->script_sections()->add_fill(val);
2600 }
2601
2602 // Add a new input section specification to the current output
2603 // section.
2604
2605 extern "C" void
2606 script_add_input_section(void* closurev,
2607 const struct Input_section_spec* spec,
2608 int keepi)
2609 {
2610 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2611 bool keep = keepi != 0;
2612 closure->script_options()->script_sections()->add_input_section(spec, keep);
2613 }
2614
2615 // When we see DATA_SEGMENT_ALIGN we record that following output
2616 // sections may be relro.
2617
2618 extern "C" void
2619 script_data_segment_align(void* closurev)
2620 {
2621 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2622 if (!closure->script_options()->saw_sections_clause())
2623 gold_error(_("%s:%d:%d: DATA_SEGMENT_ALIGN not in SECTIONS clause"),
2624 closure->filename(), closure->lineno(), closure->charpos());
2625 else
2626 closure->script_options()->script_sections()->data_segment_align();
2627 }
2628
2629 // When we see DATA_SEGMENT_RELRO_END we know that all output sections
2630 // since DATA_SEGMENT_ALIGN should be relro.
2631
2632 extern "C" void
2633 script_data_segment_relro_end(void* closurev)
2634 {
2635 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2636 if (!closure->script_options()->saw_sections_clause())
2637 gold_error(_("%s:%d:%d: DATA_SEGMENT_ALIGN not in SECTIONS clause"),
2638 closure->filename(), closure->lineno(), closure->charpos());
2639 else
2640 closure->script_options()->script_sections()->data_segment_relro_end();
2641 }
2642
2643 // Create a new list of string/sort pairs.
2644
2645 extern "C" String_sort_list_ptr
2646 script_new_string_sort_list(const struct Wildcard_section* string_sort)
2647 {
2648 return new String_sort_list(1, *string_sort);
2649 }
2650
2651 // Add an entry to a list of string/sort pairs. The way the parser
2652 // works permits us to simply modify the first parameter, rather than
2653 // copy the vector.
2654
2655 extern "C" String_sort_list_ptr
2656 script_string_sort_list_add(String_sort_list_ptr pv,
2657 const struct Wildcard_section* string_sort)
2658 {
2659 if (pv == NULL)
2660 return script_new_string_sort_list(string_sort);
2661 else
2662 {
2663 pv->push_back(*string_sort);
2664 return pv;
2665 }
2666 }
2667
2668 // Create a new list of strings.
2669
2670 extern "C" String_list_ptr
2671 script_new_string_list(const char* str, size_t len)
2672 {
2673 return new String_list(1, std::string(str, len));
2674 }
2675
2676 // Add an element to a list of strings. The way the parser works
2677 // permits us to simply modify the first parameter, rather than copy
2678 // the vector.
2679
2680 extern "C" String_list_ptr
2681 script_string_list_push_back(String_list_ptr pv, const char* str, size_t len)
2682 {
2683 if (pv == NULL)
2684 return script_new_string_list(str, len);
2685 else
2686 {
2687 pv->push_back(std::string(str, len));
2688 return pv;
2689 }
2690 }
2691
2692 // Concatenate two string lists. Either or both may be NULL. The way
2693 // the parser works permits us to modify the parameters, rather than
2694 // copy the vector.
2695
2696 extern "C" String_list_ptr
2697 script_string_list_append(String_list_ptr pv1, String_list_ptr pv2)
2698 {
2699 if (pv1 == NULL)
2700 return pv2;
2701 if (pv2 == NULL)
2702 return pv1;
2703 pv1->insert(pv1->end(), pv2->begin(), pv2->end());
2704 return pv1;
2705 }
2706
2707 // Add a new program header.
2708
2709 extern "C" void
2710 script_add_phdr(void* closurev, const char* name, size_t namelen,
2711 unsigned int type, const Phdr_info* info)
2712 {
2713 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2714 bool includes_filehdr = info->includes_filehdr != 0;
2715 bool includes_phdrs = info->includes_phdrs != 0;
2716 bool is_flags_valid = info->is_flags_valid != 0;
2717 Script_sections* ss = closure->script_options()->script_sections();
2718 ss->add_phdr(name, namelen, type, includes_filehdr, includes_phdrs,
2719 is_flags_valid, info->flags, info->load_address);
2720 closure->clear_skip_on_incompatible_target();
2721 }
2722
2723 // Convert a program header string to a type.
2724
2725 #define PHDR_TYPE(NAME) { #NAME, sizeof(#NAME) - 1, elfcpp::NAME }
2726
2727 static struct
2728 {
2729 const char* name;
2730 size_t namelen;
2731 unsigned int val;
2732 } phdr_type_names[] =
2733 {
2734 PHDR_TYPE(PT_NULL),
2735 PHDR_TYPE(PT_LOAD),
2736 PHDR_TYPE(PT_DYNAMIC),
2737 PHDR_TYPE(PT_INTERP),
2738 PHDR_TYPE(PT_NOTE),
2739 PHDR_TYPE(PT_SHLIB),
2740 PHDR_TYPE(PT_PHDR),
2741 PHDR_TYPE(PT_TLS),
2742 PHDR_TYPE(PT_GNU_EH_FRAME),
2743 PHDR_TYPE(PT_GNU_STACK),
2744 PHDR_TYPE(PT_GNU_RELRO)
2745 };
2746
2747 extern "C" unsigned int
2748 script_phdr_string_to_type(void* closurev, const char* name, size_t namelen)
2749 {
2750 for (unsigned int i = 0;
2751 i < sizeof(phdr_type_names) / sizeof(phdr_type_names[0]);
2752 ++i)
2753 if (namelen == phdr_type_names[i].namelen
2754 && strncmp(name, phdr_type_names[i].name, namelen) == 0)
2755 return phdr_type_names[i].val;
2756 yyerror(closurev, _("unknown PHDR type (try integer)"));
2757 return elfcpp::PT_NULL;
2758 }
2759
2760 extern "C" void
2761 script_saw_segment_start_expression(void* closurev)
2762 {
2763 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2764 Script_sections* ss = closure->script_options()->script_sections();
2765 ss->set_saw_segment_start_expression(true);
2766 }