2009-10-16 Doug Kwan <dougkwan@google.com>
[deliverable/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 0, // value
878 0, // size
879 elfcpp::STT_NOTYPE,
880 elfcpp::STB_GLOBAL,
881 vis,
882 0, // nonvis
883 this->provide_,
884 true); // force_override
885 }
886
887 // Finalize a symbol value.
888
889 void
890 Symbol_assignment::finalize(Symbol_table* symtab, const Layout* layout)
891 {
892 this->finalize_maybe_dot(symtab, layout, false, 0, NULL);
893 }
894
895 // Finalize a symbol value which can refer to the dot symbol.
896
897 void
898 Symbol_assignment::finalize_with_dot(Symbol_table* symtab,
899 const Layout* layout,
900 uint64_t dot_value,
901 Output_section* dot_section)
902 {
903 this->finalize_maybe_dot(symtab, layout, true, dot_value, dot_section);
904 }
905
906 // Finalize a symbol value, internal version.
907
908 void
909 Symbol_assignment::finalize_maybe_dot(Symbol_table* symtab,
910 const Layout* layout,
911 bool is_dot_available,
912 uint64_t dot_value,
913 Output_section* dot_section)
914 {
915 // If we were only supposed to provide this symbol, the sym_ field
916 // will be NULL if the symbol was not referenced.
917 if (this->sym_ == NULL)
918 {
919 gold_assert(this->provide_);
920 return;
921 }
922
923 if (parameters->target().get_size() == 32)
924 {
925 #if defined(HAVE_TARGET_32_LITTLE) || defined(HAVE_TARGET_32_BIG)
926 this->sized_finalize<32>(symtab, layout, is_dot_available, dot_value,
927 dot_section);
928 #else
929 gold_unreachable();
930 #endif
931 }
932 else if (parameters->target().get_size() == 64)
933 {
934 #if defined(HAVE_TARGET_64_LITTLE) || defined(HAVE_TARGET_64_BIG)
935 this->sized_finalize<64>(symtab, layout, is_dot_available, dot_value,
936 dot_section);
937 #else
938 gold_unreachable();
939 #endif
940 }
941 else
942 gold_unreachable();
943 }
944
945 template<int size>
946 void
947 Symbol_assignment::sized_finalize(Symbol_table* symtab, const Layout* layout,
948 bool is_dot_available, uint64_t dot_value,
949 Output_section* dot_section)
950 {
951 Output_section* section;
952 uint64_t final_val = this->val_->eval_maybe_dot(symtab, layout, true,
953 is_dot_available,
954 dot_value, dot_section,
955 &section);
956 Sized_symbol<size>* ssym = symtab->get_sized_symbol<size>(this->sym_);
957 ssym->set_value(final_val);
958 if (section != NULL)
959 ssym->set_output_section(section);
960 }
961
962 // Set the symbol value if the expression yields an absolute value.
963
964 void
965 Symbol_assignment::set_if_absolute(Symbol_table* symtab, const Layout* layout,
966 bool is_dot_available, uint64_t dot_value)
967 {
968 if (this->sym_ == NULL)
969 return;
970
971 Output_section* val_section;
972 uint64_t val = this->val_->eval_maybe_dot(symtab, layout, false,
973 is_dot_available, dot_value,
974 NULL, &val_section);
975 if (val_section != NULL)
976 return;
977
978 if (parameters->target().get_size() == 32)
979 {
980 #if defined(HAVE_TARGET_32_LITTLE) || defined(HAVE_TARGET_32_BIG)
981 Sized_symbol<32>* ssym = symtab->get_sized_symbol<32>(this->sym_);
982 ssym->set_value(val);
983 #else
984 gold_unreachable();
985 #endif
986 }
987 else if (parameters->target().get_size() == 64)
988 {
989 #if defined(HAVE_TARGET_64_LITTLE) || defined(HAVE_TARGET_64_BIG)
990 Sized_symbol<64>* ssym = symtab->get_sized_symbol<64>(this->sym_);
991 ssym->set_value(val);
992 #else
993 gold_unreachable();
994 #endif
995 }
996 else
997 gold_unreachable();
998 }
999
1000 // Print for debugging.
1001
1002 void
1003 Symbol_assignment::print(FILE* f) const
1004 {
1005 if (this->provide_ && this->hidden_)
1006 fprintf(f, "PROVIDE_HIDDEN(");
1007 else if (this->provide_)
1008 fprintf(f, "PROVIDE(");
1009 else if (this->hidden_)
1010 gold_unreachable();
1011
1012 fprintf(f, "%s = ", this->name_.c_str());
1013 this->val_->print(f);
1014
1015 if (this->provide_ || this->hidden_)
1016 fprintf(f, ")");
1017
1018 fprintf(f, "\n");
1019 }
1020
1021 // Class Script_assertion.
1022
1023 // Check the assertion.
1024
1025 void
1026 Script_assertion::check(const Symbol_table* symtab, const Layout* layout)
1027 {
1028 if (!this->check_->eval(symtab, layout, true))
1029 gold_error("%s", this->message_.c_str());
1030 }
1031
1032 // Print for debugging.
1033
1034 void
1035 Script_assertion::print(FILE* f) const
1036 {
1037 fprintf(f, "ASSERT(");
1038 this->check_->print(f);
1039 fprintf(f, ", \"%s\")\n", this->message_.c_str());
1040 }
1041
1042 // Class Script_options.
1043
1044 Script_options::Script_options()
1045 : entry_(), symbol_assignments_(), version_script_info_(),
1046 script_sections_()
1047 {
1048 }
1049
1050 // Add a symbol to be defined.
1051
1052 void
1053 Script_options::add_symbol_assignment(const char* name, size_t length,
1054 Expression* value, bool provide,
1055 bool hidden)
1056 {
1057 if (length != 1 || name[0] != '.')
1058 {
1059 if (this->script_sections_.in_sections_clause())
1060 this->script_sections_.add_symbol_assignment(name, length, value,
1061 provide, hidden);
1062 else
1063 {
1064 Symbol_assignment* p = new Symbol_assignment(name, length, value,
1065 provide, hidden);
1066 this->symbol_assignments_.push_back(p);
1067 }
1068 }
1069 else
1070 {
1071 if (provide || hidden)
1072 gold_error(_("invalid use of PROVIDE for dot symbol"));
1073
1074 // The GNU linker permits assignments to dot outside of SECTIONS
1075 // clauses and treats them as occurring inside, so we don't
1076 // check in_sections_clause here.
1077 this->script_sections_.add_dot_assignment(value);
1078 }
1079 }
1080
1081 // Add an assertion.
1082
1083 void
1084 Script_options::add_assertion(Expression* check, const char* message,
1085 size_t messagelen)
1086 {
1087 if (this->script_sections_.in_sections_clause())
1088 this->script_sections_.add_assertion(check, message, messagelen);
1089 else
1090 {
1091 Script_assertion* p = new Script_assertion(check, message, messagelen);
1092 this->assertions_.push_back(p);
1093 }
1094 }
1095
1096 // Create sections required by any linker scripts.
1097
1098 void
1099 Script_options::create_script_sections(Layout* layout)
1100 {
1101 if (this->saw_sections_clause())
1102 this->script_sections_.create_sections(layout);
1103 }
1104
1105 // Add any symbols we are defining to the symbol table.
1106
1107 void
1108 Script_options::add_symbols_to_table(Symbol_table* symtab)
1109 {
1110 for (Symbol_assignments::iterator p = this->symbol_assignments_.begin();
1111 p != this->symbol_assignments_.end();
1112 ++p)
1113 (*p)->add_to_table(symtab);
1114 this->script_sections_.add_symbols_to_table(symtab);
1115 }
1116
1117 // Finalize symbol values. Also check assertions.
1118
1119 void
1120 Script_options::finalize_symbols(Symbol_table* symtab, const Layout* layout)
1121 {
1122 // We finalize the symbols defined in SECTIONS first, because they
1123 // are the ones which may have changed. This way if symbol outside
1124 // SECTIONS are defined in terms of symbols inside SECTIONS, they
1125 // will get the right value.
1126 this->script_sections_.finalize_symbols(symtab, layout);
1127
1128 for (Symbol_assignments::iterator p = this->symbol_assignments_.begin();
1129 p != this->symbol_assignments_.end();
1130 ++p)
1131 (*p)->finalize(symtab, layout);
1132
1133 for (Assertions::iterator p = this->assertions_.begin();
1134 p != this->assertions_.end();
1135 ++p)
1136 (*p)->check(symtab, layout);
1137 }
1138
1139 // Set section addresses. We set all the symbols which have absolute
1140 // values. Then we let the SECTIONS clause do its thing. This
1141 // returns the segment which holds the file header and segment
1142 // headers, if any.
1143
1144 Output_segment*
1145 Script_options::set_section_addresses(Symbol_table* symtab, Layout* layout)
1146 {
1147 for (Symbol_assignments::iterator p = this->symbol_assignments_.begin();
1148 p != this->symbol_assignments_.end();
1149 ++p)
1150 (*p)->set_if_absolute(symtab, layout, false, 0);
1151
1152 return this->script_sections_.set_section_addresses(symtab, layout);
1153 }
1154
1155 // This class holds data passed through the parser to the lexer and to
1156 // the parser support functions. This avoids global variables. We
1157 // can't use global variables because we need not be called by a
1158 // singleton thread.
1159
1160 class Parser_closure
1161 {
1162 public:
1163 Parser_closure(const char* filename,
1164 const Position_dependent_options& posdep_options,
1165 bool in_group, bool is_in_sysroot,
1166 Command_line* command_line,
1167 Script_options* script_options,
1168 Lex* lex,
1169 bool skip_on_incompatible_target)
1170 : filename_(filename), posdep_options_(posdep_options),
1171 in_group_(in_group), is_in_sysroot_(is_in_sysroot),
1172 skip_on_incompatible_target_(skip_on_incompatible_target),
1173 found_incompatible_target_(false),
1174 command_line_(command_line), script_options_(script_options),
1175 version_script_info_(script_options->version_script_info()),
1176 lex_(lex), lineno_(0), charpos_(0), lex_mode_stack_(), inputs_(NULL)
1177 {
1178 // We start out processing C symbols in the default lex mode.
1179 language_stack_.push_back("");
1180 lex_mode_stack_.push_back(lex->mode());
1181 }
1182
1183 // Return the file name.
1184 const char*
1185 filename() const
1186 { return this->filename_; }
1187
1188 // Return the position dependent options. The caller may modify
1189 // this.
1190 Position_dependent_options&
1191 position_dependent_options()
1192 { return this->posdep_options_; }
1193
1194 // Return whether this script is being run in a group.
1195 bool
1196 in_group() const
1197 { return this->in_group_; }
1198
1199 // Return whether this script was found using a directory in the
1200 // sysroot.
1201 bool
1202 is_in_sysroot() const
1203 { return this->is_in_sysroot_; }
1204
1205 // Whether to skip to the next file with the same name if we find an
1206 // incompatible target in an OUTPUT_FORMAT statement.
1207 bool
1208 skip_on_incompatible_target() const
1209 { return this->skip_on_incompatible_target_; }
1210
1211 // Stop skipping to the next file on an incompatible target. This
1212 // is called when we make some unrevocable change to the data
1213 // structures.
1214 void
1215 clear_skip_on_incompatible_target()
1216 { this->skip_on_incompatible_target_ = false; }
1217
1218 // Whether we found an incompatible target in an OUTPUT_FORMAT
1219 // statement.
1220 bool
1221 found_incompatible_target() const
1222 { return this->found_incompatible_target_; }
1223
1224 // Note that we found an incompatible target.
1225 void
1226 set_found_incompatible_target()
1227 { this->found_incompatible_target_ = true; }
1228
1229 // Returns the Command_line structure passed in at constructor time.
1230 // This value may be NULL. The caller may modify this, which modifies
1231 // the passed-in Command_line object (not a copy).
1232 Command_line*
1233 command_line()
1234 { return this->command_line_; }
1235
1236 // Return the options which may be set by a script.
1237 Script_options*
1238 script_options()
1239 { return this->script_options_; }
1240
1241 // Return the object in which version script information should be stored.
1242 Version_script_info*
1243 version_script()
1244 { return this->version_script_info_; }
1245
1246 // Return the next token, and advance.
1247 const Token*
1248 next_token()
1249 {
1250 const Token* token = this->lex_->next_token();
1251 this->lineno_ = token->lineno();
1252 this->charpos_ = token->charpos();
1253 return token;
1254 }
1255
1256 // Set a new lexer mode, pushing the current one.
1257 void
1258 push_lex_mode(Lex::Mode mode)
1259 {
1260 this->lex_mode_stack_.push_back(this->lex_->mode());
1261 this->lex_->set_mode(mode);
1262 }
1263
1264 // Pop the lexer mode.
1265 void
1266 pop_lex_mode()
1267 {
1268 gold_assert(!this->lex_mode_stack_.empty());
1269 this->lex_->set_mode(this->lex_mode_stack_.back());
1270 this->lex_mode_stack_.pop_back();
1271 }
1272
1273 // Return the current lexer mode.
1274 Lex::Mode
1275 lex_mode() const
1276 { return this->lex_mode_stack_.back(); }
1277
1278 // Return the line number of the last token.
1279 int
1280 lineno() const
1281 { return this->lineno_; }
1282
1283 // Return the character position in the line of the last token.
1284 int
1285 charpos() const
1286 { return this->charpos_; }
1287
1288 // Return the list of input files, creating it if necessary. This
1289 // is a space leak--we never free the INPUTS_ pointer.
1290 Input_arguments*
1291 inputs()
1292 {
1293 if (this->inputs_ == NULL)
1294 this->inputs_ = new Input_arguments();
1295 return this->inputs_;
1296 }
1297
1298 // Return whether we saw any input files.
1299 bool
1300 saw_inputs() const
1301 { return this->inputs_ != NULL && !this->inputs_->empty(); }
1302
1303 // Return the current language being processed in a version script
1304 // (eg, "C++"). The empty string represents unmangled C names.
1305 const std::string&
1306 get_current_language() const
1307 { return this->language_stack_.back(); }
1308
1309 // Push a language onto the stack when entering an extern block.
1310 void push_language(const std::string& lang)
1311 { this->language_stack_.push_back(lang); }
1312
1313 // Pop a language off of the stack when exiting an extern block.
1314 void pop_language()
1315 {
1316 gold_assert(!this->language_stack_.empty());
1317 this->language_stack_.pop_back();
1318 }
1319
1320 private:
1321 // The name of the file we are reading.
1322 const char* filename_;
1323 // The position dependent options.
1324 Position_dependent_options posdep_options_;
1325 // Whether we are currently in a --start-group/--end-group.
1326 bool in_group_;
1327 // Whether the script was found in a sysrooted directory.
1328 bool is_in_sysroot_;
1329 // If this is true, then if we find an OUTPUT_FORMAT with an
1330 // incompatible target, then we tell the parser to abort so that we
1331 // can search for the next file with the same name.
1332 bool skip_on_incompatible_target_;
1333 // True if we found an OUTPUT_FORMAT with an incompatible target.
1334 bool found_incompatible_target_;
1335 // May be NULL if the user chooses not to pass one in.
1336 Command_line* command_line_;
1337 // Options which may be set from any linker script.
1338 Script_options* script_options_;
1339 // Information parsed from a version script.
1340 Version_script_info* version_script_info_;
1341 // The lexer.
1342 Lex* lex_;
1343 // The line number of the last token returned by next_token.
1344 int lineno_;
1345 // The column number of the last token returned by next_token.
1346 int charpos_;
1347 // A stack of lexer modes.
1348 std::vector<Lex::Mode> lex_mode_stack_;
1349 // A stack of which extern/language block we're inside. Can be C++,
1350 // java, or empty for C.
1351 std::vector<std::string> language_stack_;
1352 // New input files found to add to the link.
1353 Input_arguments* inputs_;
1354 };
1355
1356 // FILE was found as an argument on the command line. Try to read it
1357 // as a script. Return true if the file was handled.
1358
1359 bool
1360 read_input_script(Workqueue* workqueue, Symbol_table* symtab, Layout* layout,
1361 Dirsearch* dirsearch, int dirindex,
1362 Input_objects* input_objects, Mapfile* mapfile,
1363 Input_group* input_group,
1364 const Input_argument* input_argument,
1365 Input_file* input_file, Task_token* next_blocker,
1366 bool* used_next_blocker)
1367 {
1368 *used_next_blocker = false;
1369
1370 std::string input_string;
1371 Lex::read_file(input_file, &input_string);
1372
1373 Lex lex(input_string.c_str(), input_string.length(), PARSING_LINKER_SCRIPT);
1374
1375 Parser_closure closure(input_file->filename().c_str(),
1376 input_argument->file().options(),
1377 input_group != NULL,
1378 input_file->is_in_sysroot(),
1379 NULL,
1380 layout->script_options(),
1381 &lex,
1382 input_file->will_search_for());
1383
1384 if (yyparse(&closure) != 0)
1385 {
1386 if (closure.found_incompatible_target())
1387 {
1388 Read_symbols::incompatible_warning(input_argument, input_file);
1389 Read_symbols::requeue(workqueue, input_objects, symtab, layout,
1390 dirsearch, dirindex, mapfile, input_argument,
1391 input_group, next_blocker);
1392 return true;
1393 }
1394 return false;
1395 }
1396
1397 if (!closure.saw_inputs())
1398 return true;
1399
1400 Task_token* this_blocker = NULL;
1401 for (Input_arguments::const_iterator p = closure.inputs()->begin();
1402 p != closure.inputs()->end();
1403 ++p)
1404 {
1405 Task_token* nb;
1406 if (p + 1 == closure.inputs()->end())
1407 nb = next_blocker;
1408 else
1409 {
1410 nb = new Task_token(true);
1411 nb->add_blocker();
1412 }
1413 workqueue->queue_soon(new Read_symbols(input_objects, symtab,
1414 layout, dirsearch, 0, mapfile, &*p,
1415 input_group, this_blocker, nb));
1416 this_blocker = nb;
1417 }
1418
1419 if (layout->incremental_inputs())
1420 {
1421 // Like new Read_symbols(...) above, we rely on close.inputs()
1422 // getting leaked by closure.
1423 Script_info* info = new Script_info(closure.inputs());
1424 layout->incremental_inputs()->report_script(
1425 input_argument,
1426 input_file->file().get_mtime(),
1427 info);
1428 }
1429 *used_next_blocker = true;
1430
1431 return true;
1432 }
1433
1434 // Helper function for read_version_script() and
1435 // read_commandline_script(). Processes the given file in the mode
1436 // indicated by first_token and lex_mode.
1437
1438 static bool
1439 read_script_file(const char* filename, Command_line* cmdline,
1440 Script_options* script_options,
1441 int first_token, Lex::Mode lex_mode)
1442 {
1443 // TODO: if filename is a relative filename, search for it manually
1444 // using "." + cmdline->options()->search_path() -- not dirsearch.
1445 Dirsearch dirsearch;
1446
1447 // The file locking code wants to record a Task, but we haven't
1448 // started the workqueue yet. This is only for debugging purposes,
1449 // so we invent a fake value.
1450 const Task* task = reinterpret_cast<const Task*>(-1);
1451
1452 // We don't want this file to be opened in binary mode.
1453 Position_dependent_options posdep = cmdline->position_dependent_options();
1454 if (posdep.format_enum() == General_options::OBJECT_FORMAT_BINARY)
1455 posdep.set_format_enum(General_options::OBJECT_FORMAT_ELF);
1456 Input_file_argument input_argument(filename,
1457 Input_file_argument::INPUT_FILE_TYPE_FILE,
1458 "", false, posdep);
1459 Input_file input_file(&input_argument);
1460 int dummy = 0;
1461 if (!input_file.open(dirsearch, task, &dummy))
1462 return false;
1463
1464 std::string input_string;
1465 Lex::read_file(&input_file, &input_string);
1466
1467 Lex lex(input_string.c_str(), input_string.length(), first_token);
1468 lex.set_mode(lex_mode);
1469
1470 Parser_closure closure(filename,
1471 cmdline->position_dependent_options(),
1472 false,
1473 input_file.is_in_sysroot(),
1474 cmdline,
1475 script_options,
1476 &lex,
1477 false);
1478 if (yyparse(&closure) != 0)
1479 {
1480 input_file.file().unlock(task);
1481 return false;
1482 }
1483
1484 input_file.file().unlock(task);
1485
1486 gold_assert(!closure.saw_inputs());
1487
1488 return true;
1489 }
1490
1491 // FILENAME was found as an argument to --script (-T).
1492 // Read it as a script, and execute its contents immediately.
1493
1494 bool
1495 read_commandline_script(const char* filename, Command_line* cmdline)
1496 {
1497 return read_script_file(filename, cmdline, &cmdline->script_options(),
1498 PARSING_LINKER_SCRIPT, Lex::LINKER_SCRIPT);
1499 }
1500
1501 // FILENAME was found as an argument to --version-script. Read it as
1502 // a version script, and store its contents in
1503 // cmdline->script_options()->version_script_info().
1504
1505 bool
1506 read_version_script(const char* filename, Command_line* cmdline)
1507 {
1508 return read_script_file(filename, cmdline, &cmdline->script_options(),
1509 PARSING_VERSION_SCRIPT, Lex::VERSION_SCRIPT);
1510 }
1511
1512 // FILENAME was found as an argument to --dynamic-list. Read it as a
1513 // list of symbols, and store its contents in DYNAMIC_LIST.
1514
1515 bool
1516 read_dynamic_list(const char* filename, Command_line* cmdline,
1517 Script_options* dynamic_list)
1518 {
1519 return read_script_file(filename, cmdline, dynamic_list,
1520 PARSING_DYNAMIC_LIST, Lex::DYNAMIC_LIST);
1521 }
1522
1523 // Implement the --defsym option on the command line. Return true if
1524 // all is well.
1525
1526 bool
1527 Script_options::define_symbol(const char* definition)
1528 {
1529 Lex lex(definition, strlen(definition), PARSING_DEFSYM);
1530 lex.set_mode(Lex::EXPRESSION);
1531
1532 // Dummy value.
1533 Position_dependent_options posdep_options;
1534
1535 Parser_closure closure("command line", posdep_options, false, false, NULL,
1536 this, &lex, false);
1537
1538 if (yyparse(&closure) != 0)
1539 return false;
1540
1541 gold_assert(!closure.saw_inputs());
1542
1543 return true;
1544 }
1545
1546 // Print the script to F for debugging.
1547
1548 void
1549 Script_options::print(FILE* f) const
1550 {
1551 fprintf(f, "%s: Dumping linker script\n", program_name);
1552
1553 if (!this->entry_.empty())
1554 fprintf(f, "ENTRY(%s)\n", this->entry_.c_str());
1555
1556 for (Symbol_assignments::const_iterator p =
1557 this->symbol_assignments_.begin();
1558 p != this->symbol_assignments_.end();
1559 ++p)
1560 (*p)->print(f);
1561
1562 for (Assertions::const_iterator p = this->assertions_.begin();
1563 p != this->assertions_.end();
1564 ++p)
1565 (*p)->print(f);
1566
1567 this->script_sections_.print(f);
1568
1569 this->version_script_info_.print(f);
1570 }
1571
1572 // Manage mapping from keywords to the codes expected by the bison
1573 // parser. We construct one global object for each lex mode with
1574 // keywords.
1575
1576 class Keyword_to_parsecode
1577 {
1578 public:
1579 // The structure which maps keywords to parsecodes.
1580 struct Keyword_parsecode
1581 {
1582 // Keyword.
1583 const char* keyword;
1584 // Corresponding parsecode.
1585 int parsecode;
1586 };
1587
1588 Keyword_to_parsecode(const Keyword_parsecode* keywords,
1589 int keyword_count)
1590 : keyword_parsecodes_(keywords), keyword_count_(keyword_count)
1591 { }
1592
1593 // Return the parsecode corresponding KEYWORD, or 0 if it is not a
1594 // keyword.
1595 int
1596 keyword_to_parsecode(const char* keyword, size_t len) const;
1597
1598 private:
1599 const Keyword_parsecode* keyword_parsecodes_;
1600 const int keyword_count_;
1601 };
1602
1603 // Mapping from keyword string to keyword parsecode. This array must
1604 // be kept in sorted order. Parsecodes are looked up using bsearch.
1605 // This array must correspond to the list of parsecodes in yyscript.y.
1606
1607 static const Keyword_to_parsecode::Keyword_parsecode
1608 script_keyword_parsecodes[] =
1609 {
1610 { "ABSOLUTE", ABSOLUTE },
1611 { "ADDR", ADDR },
1612 { "ALIGN", ALIGN_K },
1613 { "ALIGNOF", ALIGNOF },
1614 { "ASSERT", ASSERT_K },
1615 { "AS_NEEDED", AS_NEEDED },
1616 { "AT", AT },
1617 { "BIND", BIND },
1618 { "BLOCK", BLOCK },
1619 { "BYTE", BYTE },
1620 { "CONSTANT", CONSTANT },
1621 { "CONSTRUCTORS", CONSTRUCTORS },
1622 { "CREATE_OBJECT_SYMBOLS", CREATE_OBJECT_SYMBOLS },
1623 { "DATA_SEGMENT_ALIGN", DATA_SEGMENT_ALIGN },
1624 { "DATA_SEGMENT_END", DATA_SEGMENT_END },
1625 { "DATA_SEGMENT_RELRO_END", DATA_SEGMENT_RELRO_END },
1626 { "DEFINED", DEFINED },
1627 { "ENTRY", ENTRY },
1628 { "EXCLUDE_FILE", EXCLUDE_FILE },
1629 { "EXTERN", EXTERN },
1630 { "FILL", FILL },
1631 { "FLOAT", FLOAT },
1632 { "FORCE_COMMON_ALLOCATION", FORCE_COMMON_ALLOCATION },
1633 { "GROUP", GROUP },
1634 { "HLL", HLL },
1635 { "INCLUDE", INCLUDE },
1636 { "INHIBIT_COMMON_ALLOCATION", INHIBIT_COMMON_ALLOCATION },
1637 { "INPUT", INPUT },
1638 { "KEEP", KEEP },
1639 { "LENGTH", LENGTH },
1640 { "LOADADDR", LOADADDR },
1641 { "LONG", LONG },
1642 { "MAP", MAP },
1643 { "MAX", MAX_K },
1644 { "MEMORY", MEMORY },
1645 { "MIN", MIN_K },
1646 { "NEXT", NEXT },
1647 { "NOCROSSREFS", NOCROSSREFS },
1648 { "NOFLOAT", NOFLOAT },
1649 { "ONLY_IF_RO", ONLY_IF_RO },
1650 { "ONLY_IF_RW", ONLY_IF_RW },
1651 { "OPTION", OPTION },
1652 { "ORIGIN", ORIGIN },
1653 { "OUTPUT", OUTPUT },
1654 { "OUTPUT_ARCH", OUTPUT_ARCH },
1655 { "OUTPUT_FORMAT", OUTPUT_FORMAT },
1656 { "OVERLAY", OVERLAY },
1657 { "PHDRS", PHDRS },
1658 { "PROVIDE", PROVIDE },
1659 { "PROVIDE_HIDDEN", PROVIDE_HIDDEN },
1660 { "QUAD", QUAD },
1661 { "SEARCH_DIR", SEARCH_DIR },
1662 { "SECTIONS", SECTIONS },
1663 { "SEGMENT_START", SEGMENT_START },
1664 { "SHORT", SHORT },
1665 { "SIZEOF", SIZEOF },
1666 { "SIZEOF_HEADERS", SIZEOF_HEADERS },
1667 { "SORT", SORT_BY_NAME },
1668 { "SORT_BY_ALIGNMENT", SORT_BY_ALIGNMENT },
1669 { "SORT_BY_NAME", SORT_BY_NAME },
1670 { "SPECIAL", SPECIAL },
1671 { "SQUAD", SQUAD },
1672 { "STARTUP", STARTUP },
1673 { "SUBALIGN", SUBALIGN },
1674 { "SYSLIB", SYSLIB },
1675 { "TARGET", TARGET_K },
1676 { "TRUNCATE", TRUNCATE },
1677 { "VERSION", VERSIONK },
1678 { "global", GLOBAL },
1679 { "l", LENGTH },
1680 { "len", LENGTH },
1681 { "local", LOCAL },
1682 { "o", ORIGIN },
1683 { "org", ORIGIN },
1684 { "sizeof_headers", SIZEOF_HEADERS },
1685 };
1686
1687 static const Keyword_to_parsecode
1688 script_keywords(&script_keyword_parsecodes[0],
1689 (sizeof(script_keyword_parsecodes)
1690 / sizeof(script_keyword_parsecodes[0])));
1691
1692 static const Keyword_to_parsecode::Keyword_parsecode
1693 version_script_keyword_parsecodes[] =
1694 {
1695 { "extern", EXTERN },
1696 { "global", GLOBAL },
1697 { "local", LOCAL },
1698 };
1699
1700 static const Keyword_to_parsecode
1701 version_script_keywords(&version_script_keyword_parsecodes[0],
1702 (sizeof(version_script_keyword_parsecodes)
1703 / sizeof(version_script_keyword_parsecodes[0])));
1704
1705 static const Keyword_to_parsecode::Keyword_parsecode
1706 dynamic_list_keyword_parsecodes[] =
1707 {
1708 { "extern", EXTERN },
1709 };
1710
1711 static const Keyword_to_parsecode
1712 dynamic_list_keywords(&dynamic_list_keyword_parsecodes[0],
1713 (sizeof(dynamic_list_keyword_parsecodes)
1714 / sizeof(dynamic_list_keyword_parsecodes[0])));
1715
1716
1717
1718 // Comparison function passed to bsearch.
1719
1720 extern "C"
1721 {
1722
1723 struct Ktt_key
1724 {
1725 const char* str;
1726 size_t len;
1727 };
1728
1729 static int
1730 ktt_compare(const void* keyv, const void* kttv)
1731 {
1732 const Ktt_key* key = static_cast<const Ktt_key*>(keyv);
1733 const Keyword_to_parsecode::Keyword_parsecode* ktt =
1734 static_cast<const Keyword_to_parsecode::Keyword_parsecode*>(kttv);
1735 int i = strncmp(key->str, ktt->keyword, key->len);
1736 if (i != 0)
1737 return i;
1738 if (ktt->keyword[key->len] != '\0')
1739 return -1;
1740 return 0;
1741 }
1742
1743 } // End extern "C".
1744
1745 int
1746 Keyword_to_parsecode::keyword_to_parsecode(const char* keyword,
1747 size_t len) const
1748 {
1749 Ktt_key key;
1750 key.str = keyword;
1751 key.len = len;
1752 void* kttv = bsearch(&key,
1753 this->keyword_parsecodes_,
1754 this->keyword_count_,
1755 sizeof(this->keyword_parsecodes_[0]),
1756 ktt_compare);
1757 if (kttv == NULL)
1758 return 0;
1759 Keyword_parsecode* ktt = static_cast<Keyword_parsecode*>(kttv);
1760 return ktt->parsecode;
1761 }
1762
1763 // Helper class that calls cplus_demangle when needed and takes care of freeing
1764 // the result.
1765
1766 class Lazy_demangler
1767 {
1768 public:
1769 Lazy_demangler(const char* symbol, int options)
1770 : symbol_(symbol), options_(options), demangled_(NULL), did_demangle_(false)
1771 { }
1772
1773 ~Lazy_demangler()
1774 { free(this->demangled_); }
1775
1776 // Return the demangled name. The actual demangling happens on the first call,
1777 // and the result is later cached.
1778
1779 inline char*
1780 get();
1781
1782 private:
1783 // The symbol to demangle.
1784 const char *symbol_;
1785 // Option flags to pass to cplus_demagle.
1786 const int options_;
1787 // The cached demangled value, or NULL if demangling didn't happen yet or
1788 // failed.
1789 char *demangled_;
1790 // Whether we already called cplus_demangle
1791 bool did_demangle_;
1792 };
1793
1794 // Return the demangled name. The actual demangling happens on the first call,
1795 // and the result is later cached. Returns NULL if the symbol cannot be
1796 // demangled.
1797
1798 inline char*
1799 Lazy_demangler::get()
1800 {
1801 if (!this->did_demangle_)
1802 {
1803 this->demangled_ = cplus_demangle(this->symbol_, this->options_);
1804 this->did_demangle_ = true;
1805 }
1806 return this->demangled_;
1807 }
1808
1809 // The following structs are used within the VersionInfo class as well
1810 // as in the bison helper functions. They store the information
1811 // parsed from the version script.
1812
1813 // A single version expression.
1814 // For example, pattern="std::map*" and language="C++".
1815 // pattern and language should be from the stringpool
1816 struct Version_expression {
1817 Version_expression(const std::string& pattern,
1818 const std::string& language,
1819 bool exact_match)
1820 : pattern(pattern), language(language), exact_match(exact_match) {}
1821
1822 std::string pattern;
1823 std::string language;
1824 // If false, we use glob() to match pattern. If true, we use strcmp().
1825 bool exact_match;
1826 };
1827
1828
1829 // A list of expressions.
1830 struct Version_expression_list {
1831 std::vector<struct Version_expression> expressions;
1832 };
1833
1834
1835 // A list of which versions upon which another version depends.
1836 // Strings should be from the Stringpool.
1837 struct Version_dependency_list {
1838 std::vector<std::string> dependencies;
1839 };
1840
1841
1842 // The total definition of a version. It includes the tag for the
1843 // version, its global and local expressions, and any dependencies.
1844 struct Version_tree {
1845 Version_tree()
1846 : tag(), global(NULL), local(NULL), dependencies(NULL) {}
1847
1848 std::string tag;
1849 const struct Version_expression_list* global;
1850 const struct Version_expression_list* local;
1851 const struct Version_dependency_list* dependencies;
1852 };
1853
1854 Version_script_info::~Version_script_info()
1855 {
1856 this->clear();
1857 }
1858
1859 void
1860 Version_script_info::clear()
1861 {
1862 for (size_t k = 0; k < dependency_lists_.size(); ++k)
1863 delete dependency_lists_[k];
1864 this->dependency_lists_.clear();
1865 for (size_t k = 0; k < version_trees_.size(); ++k)
1866 delete version_trees_[k];
1867 this->version_trees_.clear();
1868 for (size_t k = 0; k < expression_lists_.size(); ++k)
1869 delete expression_lists_[k];
1870 this->expression_lists_.clear();
1871 }
1872
1873 std::vector<std::string>
1874 Version_script_info::get_versions() const
1875 {
1876 std::vector<std::string> ret;
1877 for (size_t j = 0; j < version_trees_.size(); ++j)
1878 if (!this->version_trees_[j]->tag.empty())
1879 ret.push_back(this->version_trees_[j]->tag);
1880 return ret;
1881 }
1882
1883 std::vector<std::string>
1884 Version_script_info::get_dependencies(const char* version) const
1885 {
1886 std::vector<std::string> ret;
1887 for (size_t j = 0; j < version_trees_.size(); ++j)
1888 if (version_trees_[j]->tag == version)
1889 {
1890 const struct Version_dependency_list* deps =
1891 version_trees_[j]->dependencies;
1892 if (deps != NULL)
1893 for (size_t k = 0; k < deps->dependencies.size(); ++k)
1894 ret.push_back(deps->dependencies[k]);
1895 return ret;
1896 }
1897 return ret;
1898 }
1899
1900 // Look up SYMBOL_NAME in the list of versions. If CHECK_GLOBAL is
1901 // true look at the globally visible symbols, otherwise look at the
1902 // symbols listed as "local:". Return true if the symbol is found,
1903 // false otherwise. If the symbol is found, then if PVERSION is not
1904 // NULL, set *PVERSION to the version.
1905
1906 bool
1907 Version_script_info::get_symbol_version_helper(const char* symbol_name,
1908 bool check_global,
1909 std::string* pversion) const
1910 {
1911 Lazy_demangler cpp_demangled_name(symbol_name, DMGL_ANSI | DMGL_PARAMS);
1912 Lazy_demangler java_demangled_name(symbol_name,
1913 DMGL_ANSI | DMGL_PARAMS | DMGL_JAVA);
1914 for (size_t j = 0; j < version_trees_.size(); ++j)
1915 {
1916 // Is it a global symbol for this version?
1917 const Version_expression_list* explist =
1918 check_global ? version_trees_[j]->global : version_trees_[j]->local;
1919 if (explist != NULL)
1920 for (size_t k = 0; k < explist->expressions.size(); ++k)
1921 {
1922 const char* name_to_match = symbol_name;
1923 const struct Version_expression& exp = explist->expressions[k];
1924 if (exp.language == "C++")
1925 {
1926 name_to_match = cpp_demangled_name.get();
1927 // This isn't a C++ symbol.
1928 if (name_to_match == NULL)
1929 continue;
1930 }
1931 else if (exp.language == "Java")
1932 {
1933 name_to_match = java_demangled_name.get();
1934 // This isn't a Java symbol.
1935 if (name_to_match == NULL)
1936 continue;
1937 }
1938 bool matched;
1939 if (exp.exact_match)
1940 matched = strcmp(exp.pattern.c_str(), name_to_match) == 0;
1941 else
1942 matched = fnmatch(exp.pattern.c_str(), name_to_match,
1943 FNM_NOESCAPE) == 0;
1944 if (matched)
1945 {
1946 if (pversion != NULL)
1947 *pversion = this->version_trees_[j]->tag;
1948 return true;
1949 }
1950 }
1951 }
1952 return false;
1953 }
1954
1955 struct Version_dependency_list*
1956 Version_script_info::allocate_dependency_list()
1957 {
1958 dependency_lists_.push_back(new Version_dependency_list);
1959 return dependency_lists_.back();
1960 }
1961
1962 struct Version_expression_list*
1963 Version_script_info::allocate_expression_list()
1964 {
1965 expression_lists_.push_back(new Version_expression_list);
1966 return expression_lists_.back();
1967 }
1968
1969 struct Version_tree*
1970 Version_script_info::allocate_version_tree()
1971 {
1972 version_trees_.push_back(new Version_tree);
1973 return version_trees_.back();
1974 }
1975
1976 // Print for debugging.
1977
1978 void
1979 Version_script_info::print(FILE* f) const
1980 {
1981 if (this->empty())
1982 return;
1983
1984 fprintf(f, "VERSION {");
1985
1986 for (size_t i = 0; i < this->version_trees_.size(); ++i)
1987 {
1988 const Version_tree* vt = this->version_trees_[i];
1989
1990 if (vt->tag.empty())
1991 fprintf(f, " {\n");
1992 else
1993 fprintf(f, " %s {\n", vt->tag.c_str());
1994
1995 if (vt->global != NULL)
1996 {
1997 fprintf(f, " global :\n");
1998 this->print_expression_list(f, vt->global);
1999 }
2000
2001 if (vt->local != NULL)
2002 {
2003 fprintf(f, " local :\n");
2004 this->print_expression_list(f, vt->local);
2005 }
2006
2007 fprintf(f, " }");
2008 if (vt->dependencies != NULL)
2009 {
2010 const Version_dependency_list* deps = vt->dependencies;
2011 for (size_t j = 0; j < deps->dependencies.size(); ++j)
2012 {
2013 if (j < deps->dependencies.size() - 1)
2014 fprintf(f, "\n");
2015 fprintf(f, " %s", deps->dependencies[j].c_str());
2016 }
2017 }
2018 fprintf(f, ";\n");
2019 }
2020
2021 fprintf(f, "}\n");
2022 }
2023
2024 void
2025 Version_script_info::print_expression_list(
2026 FILE* f,
2027 const Version_expression_list* vel) const
2028 {
2029 std::string current_language;
2030 for (size_t i = 0; i < vel->expressions.size(); ++i)
2031 {
2032 const Version_expression& ve(vel->expressions[i]);
2033
2034 if (ve.language != current_language)
2035 {
2036 if (!current_language.empty())
2037 fprintf(f, " }\n");
2038 fprintf(f, " extern \"%s\" {\n", ve.language.c_str());
2039 current_language = ve.language;
2040 }
2041
2042 fprintf(f, " ");
2043 if (!current_language.empty())
2044 fprintf(f, " ");
2045
2046 if (ve.exact_match)
2047 fprintf(f, "\"");
2048 fprintf(f, "%s", ve.pattern.c_str());
2049 if (ve.exact_match)
2050 fprintf(f, "\"");
2051
2052 fprintf(f, "\n");
2053 }
2054
2055 if (!current_language.empty())
2056 fprintf(f, " }\n");
2057 }
2058
2059 } // End namespace gold.
2060
2061 // The remaining functions are extern "C", so it's clearer to not put
2062 // them in namespace gold.
2063
2064 using namespace gold;
2065
2066 // This function is called by the bison parser to return the next
2067 // token.
2068
2069 extern "C" int
2070 yylex(YYSTYPE* lvalp, void* closurev)
2071 {
2072 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2073 const Token* token = closure->next_token();
2074 switch (token->classification())
2075 {
2076 default:
2077 gold_unreachable();
2078
2079 case Token::TOKEN_INVALID:
2080 yyerror(closurev, "invalid character");
2081 return 0;
2082
2083 case Token::TOKEN_EOF:
2084 return 0;
2085
2086 case Token::TOKEN_STRING:
2087 {
2088 // This is either a keyword or a STRING.
2089 size_t len;
2090 const char* str = token->string_value(&len);
2091 int parsecode = 0;
2092 switch (closure->lex_mode())
2093 {
2094 case Lex::LINKER_SCRIPT:
2095 parsecode = script_keywords.keyword_to_parsecode(str, len);
2096 break;
2097 case Lex::VERSION_SCRIPT:
2098 parsecode = version_script_keywords.keyword_to_parsecode(str, len);
2099 break;
2100 case Lex::DYNAMIC_LIST:
2101 parsecode = dynamic_list_keywords.keyword_to_parsecode(str, len);
2102 break;
2103 default:
2104 break;
2105 }
2106 if (parsecode != 0)
2107 return parsecode;
2108 lvalp->string.value = str;
2109 lvalp->string.length = len;
2110 return STRING;
2111 }
2112
2113 case Token::TOKEN_QUOTED_STRING:
2114 lvalp->string.value = token->string_value(&lvalp->string.length);
2115 return QUOTED_STRING;
2116
2117 case Token::TOKEN_OPERATOR:
2118 return token->operator_value();
2119
2120 case Token::TOKEN_INTEGER:
2121 lvalp->integer = token->integer_value();
2122 return INTEGER;
2123 }
2124 }
2125
2126 // This function is called by the bison parser to report an error.
2127
2128 extern "C" void
2129 yyerror(void* closurev, const char* message)
2130 {
2131 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2132 gold_error(_("%s:%d:%d: %s"), closure->filename(), closure->lineno(),
2133 closure->charpos(), message);
2134 }
2135
2136 // Called by the bison parser to add an external symbol to the link.
2137
2138 extern "C" void
2139 script_add_extern(void* closurev, const char* name, size_t length)
2140 {
2141 // We treat exactly like -u NAME. FIXME: If it seems useful, we
2142 // could handle this after the command line has been read, by adding
2143 // entries to the symbol table directly.
2144 std::string arg("--undefined=");
2145 arg.append(name, length);
2146 script_parse_option(closurev, arg.c_str(), arg.size());
2147 }
2148
2149 // Called by the bison parser to add a file to the link.
2150
2151 extern "C" void
2152 script_add_file(void* closurev, const char* name, size_t length)
2153 {
2154 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2155
2156 // If this is an absolute path, and we found the script in the
2157 // sysroot, then we want to prepend the sysroot to the file name.
2158 // For example, this is how we handle a cross link to the x86_64
2159 // libc.so, which refers to /lib/libc.so.6.
2160 std::string name_string(name, length);
2161 const char* extra_search_path = ".";
2162 std::string script_directory;
2163 if (IS_ABSOLUTE_PATH(name_string.c_str()))
2164 {
2165 if (closure->is_in_sysroot())
2166 {
2167 const std::string& sysroot(parameters->options().sysroot());
2168 gold_assert(!sysroot.empty());
2169 name_string = sysroot + name_string;
2170 }
2171 }
2172 else
2173 {
2174 // In addition to checking the normal library search path, we
2175 // also want to check in the script-directory.
2176 const char *slash = strrchr(closure->filename(), '/');
2177 if (slash != NULL)
2178 {
2179 script_directory.assign(closure->filename(),
2180 slash - closure->filename() + 1);
2181 extra_search_path = script_directory.c_str();
2182 }
2183 }
2184
2185 Input_file_argument file(name_string.c_str(),
2186 Input_file_argument::INPUT_FILE_TYPE_FILE,
2187 extra_search_path, false,
2188 closure->position_dependent_options());
2189 closure->inputs()->add_file(file);
2190 }
2191
2192 // Called by the bison parser to start a group. If we are already in
2193 // a group, that means that this script was invoked within a
2194 // --start-group --end-group sequence on the command line, or that
2195 // this script was found in a GROUP of another script. In that case,
2196 // we simply continue the existing group, rather than starting a new
2197 // one. It is possible to construct a case in which this will do
2198 // something other than what would happen if we did a recursive group,
2199 // but it's hard to imagine why the different behaviour would be
2200 // useful for a real program. Avoiding recursive groups is simpler
2201 // and more efficient.
2202
2203 extern "C" void
2204 script_start_group(void* closurev)
2205 {
2206 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2207 if (!closure->in_group())
2208 closure->inputs()->start_group();
2209 }
2210
2211 // Called by the bison parser at the end of a group.
2212
2213 extern "C" void
2214 script_end_group(void* closurev)
2215 {
2216 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2217 if (!closure->in_group())
2218 closure->inputs()->end_group();
2219 }
2220
2221 // Called by the bison parser to start an AS_NEEDED list.
2222
2223 extern "C" void
2224 script_start_as_needed(void* closurev)
2225 {
2226 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2227 closure->position_dependent_options().set_as_needed(true);
2228 }
2229
2230 // Called by the bison parser at the end of an AS_NEEDED list.
2231
2232 extern "C" void
2233 script_end_as_needed(void* closurev)
2234 {
2235 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2236 closure->position_dependent_options().set_as_needed(false);
2237 }
2238
2239 // Called by the bison parser to set the entry symbol.
2240
2241 extern "C" void
2242 script_set_entry(void* closurev, const char* entry, size_t length)
2243 {
2244 // We'll parse this exactly the same as --entry=ENTRY on the commandline
2245 // TODO(csilvers): FIXME -- call set_entry directly.
2246 std::string arg("--entry=");
2247 arg.append(entry, length);
2248 script_parse_option(closurev, arg.c_str(), arg.size());
2249 }
2250
2251 // Called by the bison parser to set whether to define common symbols.
2252
2253 extern "C" void
2254 script_set_common_allocation(void* closurev, int set)
2255 {
2256 const char* arg = set != 0 ? "--define-common" : "--no-define-common";
2257 script_parse_option(closurev, arg, strlen(arg));
2258 }
2259
2260 // Called by the bison parser to define a symbol.
2261
2262 extern "C" void
2263 script_set_symbol(void* closurev, const char* name, size_t length,
2264 Expression* value, int providei, int hiddeni)
2265 {
2266 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2267 const bool provide = providei != 0;
2268 const bool hidden = hiddeni != 0;
2269 closure->script_options()->add_symbol_assignment(name, length, value,
2270 provide, hidden);
2271 closure->clear_skip_on_incompatible_target();
2272 }
2273
2274 // Called by the bison parser to add an assertion.
2275
2276 extern "C" void
2277 script_add_assertion(void* closurev, Expression* check, const char* message,
2278 size_t messagelen)
2279 {
2280 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2281 closure->script_options()->add_assertion(check, message, messagelen);
2282 closure->clear_skip_on_incompatible_target();
2283 }
2284
2285 // Called by the bison parser to parse an OPTION.
2286
2287 extern "C" void
2288 script_parse_option(void* closurev, const char* option, size_t length)
2289 {
2290 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2291 // We treat the option as a single command-line option, even if
2292 // it has internal whitespace.
2293 if (closure->command_line() == NULL)
2294 {
2295 // There are some options that we could handle here--e.g.,
2296 // -lLIBRARY. Should we bother?
2297 gold_warning(_("%s:%d:%d: ignoring command OPTION; OPTION is only valid"
2298 " for scripts specified via -T/--script"),
2299 closure->filename(), closure->lineno(), closure->charpos());
2300 }
2301 else
2302 {
2303 bool past_a_double_dash_option = false;
2304 const char* mutable_option = strndup(option, length);
2305 gold_assert(mutable_option != NULL);
2306 closure->command_line()->process_one_option(1, &mutable_option, 0,
2307 &past_a_double_dash_option);
2308 // The General_options class will quite possibly store a pointer
2309 // into mutable_option, so we can't free it. In cases the class
2310 // does not store such a pointer, this is a memory leak. Alas. :(
2311 }
2312 closure->clear_skip_on_incompatible_target();
2313 }
2314
2315 // Called by the bison parser to handle OUTPUT_FORMAT. OUTPUT_FORMAT
2316 // takes either one or three arguments. In the three argument case,
2317 // the format depends on the endianness option, which we don't
2318 // currently support (FIXME). If we see an OUTPUT_FORMAT for the
2319 // wrong format, then we want to search for a new file. Returning 0
2320 // here will cause the parser to immediately abort.
2321
2322 extern "C" int
2323 script_check_output_format(void* closurev,
2324 const char* default_name, size_t default_length,
2325 const char*, size_t, const char*, size_t)
2326 {
2327 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2328 std::string name(default_name, default_length);
2329 Target* target = select_target_by_name(name.c_str());
2330 if (target == NULL || !parameters->is_compatible_target(target))
2331 {
2332 if (closure->skip_on_incompatible_target())
2333 {
2334 closure->set_found_incompatible_target();
2335 return 0;
2336 }
2337 // FIXME: Should we warn about the unknown target?
2338 }
2339 return 1;
2340 }
2341
2342 // Called by the bison parser to handle TARGET.
2343
2344 extern "C" void
2345 script_set_target(void* closurev, const char* target, size_t len)
2346 {
2347 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2348 std::string s(target, len);
2349 General_options::Object_format format_enum;
2350 format_enum = General_options::string_to_object_format(s.c_str());
2351 closure->position_dependent_options().set_format_enum(format_enum);
2352 }
2353
2354 // Called by the bison parser to handle SEARCH_DIR. This is handled
2355 // exactly like a -L option.
2356
2357 extern "C" void
2358 script_add_search_dir(void* closurev, const char* option, size_t length)
2359 {
2360 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2361 if (closure->command_line() == NULL)
2362 gold_warning(_("%s:%d:%d: ignoring SEARCH_DIR; SEARCH_DIR is only valid"
2363 " for scripts specified via -T/--script"),
2364 closure->filename(), closure->lineno(), closure->charpos());
2365 else
2366 {
2367 std::string s = "-L" + std::string(option, length);
2368 script_parse_option(closurev, s.c_str(), s.size());
2369 }
2370 }
2371
2372 /* Called by the bison parser to push the lexer into expression
2373 mode. */
2374
2375 extern "C" void
2376 script_push_lex_into_expression_mode(void* closurev)
2377 {
2378 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2379 closure->push_lex_mode(Lex::EXPRESSION);
2380 }
2381
2382 /* Called by the bison parser to push the lexer into version
2383 mode. */
2384
2385 extern "C" void
2386 script_push_lex_into_version_mode(void* closurev)
2387 {
2388 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2389 closure->push_lex_mode(Lex::VERSION_SCRIPT);
2390 }
2391
2392 /* Called by the bison parser to pop the lexer mode. */
2393
2394 extern "C" void
2395 script_pop_lex_mode(void* closurev)
2396 {
2397 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2398 closure->pop_lex_mode();
2399 }
2400
2401 // Register an entire version node. For example:
2402 //
2403 // GLIBC_2.1 {
2404 // global: foo;
2405 // } GLIBC_2.0;
2406 //
2407 // - tag is "GLIBC_2.1"
2408 // - tree contains the information "global: foo"
2409 // - deps contains "GLIBC_2.0"
2410
2411 extern "C" void
2412 script_register_vers_node(void*,
2413 const char* tag,
2414 int taglen,
2415 struct Version_tree *tree,
2416 struct Version_dependency_list *deps)
2417 {
2418 gold_assert(tree != NULL);
2419 tree->dependencies = deps;
2420 if (tag != NULL)
2421 tree->tag = std::string(tag, taglen);
2422 }
2423
2424 // Add a dependencies to the list of existing dependencies, if any,
2425 // and return the expanded list.
2426
2427 extern "C" struct Version_dependency_list *
2428 script_add_vers_depend(void* closurev,
2429 struct Version_dependency_list *all_deps,
2430 const char *depend_to_add, int deplen)
2431 {
2432 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2433 if (all_deps == NULL)
2434 all_deps = closure->version_script()->allocate_dependency_list();
2435 all_deps->dependencies.push_back(std::string(depend_to_add, deplen));
2436 return all_deps;
2437 }
2438
2439 // Add a pattern expression to an existing list of expressions, if any.
2440 // TODO: In the old linker, the last argument used to be a bool, but I
2441 // don't know what it meant.
2442
2443 extern "C" struct Version_expression_list *
2444 script_new_vers_pattern(void* closurev,
2445 struct Version_expression_list *expressions,
2446 const char *pattern, int patlen, int exact_match)
2447 {
2448 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2449 if (expressions == NULL)
2450 expressions = closure->version_script()->allocate_expression_list();
2451 expressions->expressions.push_back(
2452 Version_expression(std::string(pattern, patlen),
2453 closure->get_current_language(),
2454 static_cast<bool>(exact_match)));
2455 return expressions;
2456 }
2457
2458 // Attaches b to the end of a, and clears b. So a = a + b and b = {}.
2459
2460 extern "C" struct Version_expression_list*
2461 script_merge_expressions(struct Version_expression_list *a,
2462 struct Version_expression_list *b)
2463 {
2464 a->expressions.insert(a->expressions.end(),
2465 b->expressions.begin(), b->expressions.end());
2466 // We could delete b and remove it from expressions_lists_, but
2467 // that's a lot of work. This works just as well.
2468 b->expressions.clear();
2469 return a;
2470 }
2471
2472 // Combine the global and local expressions into a a Version_tree.
2473
2474 extern "C" struct Version_tree *
2475 script_new_vers_node(void* closurev,
2476 struct Version_expression_list *global,
2477 struct Version_expression_list *local)
2478 {
2479 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2480 Version_tree* tree = closure->version_script()->allocate_version_tree();
2481 tree->global = global;
2482 tree->local = local;
2483 return tree;
2484 }
2485
2486 // Handle a transition in language, such as at the
2487 // start or end of 'extern "C++"'
2488
2489 extern "C" void
2490 version_script_push_lang(void* closurev, const char* lang, int langlen)
2491 {
2492 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2493 closure->push_language(std::string(lang, langlen));
2494 }
2495
2496 extern "C" void
2497 version_script_pop_lang(void* closurev)
2498 {
2499 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2500 closure->pop_language();
2501 }
2502
2503 // Called by the bison parser to start a SECTIONS clause.
2504
2505 extern "C" void
2506 script_start_sections(void* closurev)
2507 {
2508 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2509 closure->script_options()->script_sections()->start_sections();
2510 closure->clear_skip_on_incompatible_target();
2511 }
2512
2513 // Called by the bison parser to finish a SECTIONS clause.
2514
2515 extern "C" void
2516 script_finish_sections(void* closurev)
2517 {
2518 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2519 closure->script_options()->script_sections()->finish_sections();
2520 }
2521
2522 // Start processing entries for an output section.
2523
2524 extern "C" void
2525 script_start_output_section(void* closurev, const char* name, size_t namelen,
2526 const struct Parser_output_section_header* header)
2527 {
2528 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2529 closure->script_options()->script_sections()->start_output_section(name,
2530 namelen,
2531 header);
2532 }
2533
2534 // Finish processing entries for an output section.
2535
2536 extern "C" void
2537 script_finish_output_section(void* closurev,
2538 const struct Parser_output_section_trailer* trail)
2539 {
2540 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2541 closure->script_options()->script_sections()->finish_output_section(trail);
2542 }
2543
2544 // Add a data item (e.g., "WORD (0)") to the current output section.
2545
2546 extern "C" void
2547 script_add_data(void* closurev, int data_token, Expression* val)
2548 {
2549 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2550 int size;
2551 bool is_signed = true;
2552 switch (data_token)
2553 {
2554 case QUAD:
2555 size = 8;
2556 is_signed = false;
2557 break;
2558 case SQUAD:
2559 size = 8;
2560 break;
2561 case LONG:
2562 size = 4;
2563 break;
2564 case SHORT:
2565 size = 2;
2566 break;
2567 case BYTE:
2568 size = 1;
2569 break;
2570 default:
2571 gold_unreachable();
2572 }
2573 closure->script_options()->script_sections()->add_data(size, is_signed, val);
2574 }
2575
2576 // Add a clause setting the fill value to the current output section.
2577
2578 extern "C" void
2579 script_add_fill(void* closurev, Expression* val)
2580 {
2581 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2582 closure->script_options()->script_sections()->add_fill(val);
2583 }
2584
2585 // Add a new input section specification to the current output
2586 // section.
2587
2588 extern "C" void
2589 script_add_input_section(void* closurev,
2590 const struct Input_section_spec* spec,
2591 int keepi)
2592 {
2593 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2594 bool keep = keepi != 0;
2595 closure->script_options()->script_sections()->add_input_section(spec, keep);
2596 }
2597
2598 // When we see DATA_SEGMENT_ALIGN we record that following output
2599 // sections may be relro.
2600
2601 extern "C" void
2602 script_data_segment_align(void* closurev)
2603 {
2604 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2605 if (!closure->script_options()->saw_sections_clause())
2606 gold_error(_("%s:%d:%d: DATA_SEGMENT_ALIGN not in SECTIONS clause"),
2607 closure->filename(), closure->lineno(), closure->charpos());
2608 else
2609 closure->script_options()->script_sections()->data_segment_align();
2610 }
2611
2612 // When we see DATA_SEGMENT_RELRO_END we know that all output sections
2613 // since DATA_SEGMENT_ALIGN should be relro.
2614
2615 extern "C" void
2616 script_data_segment_relro_end(void* closurev)
2617 {
2618 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2619 if (!closure->script_options()->saw_sections_clause())
2620 gold_error(_("%s:%d:%d: DATA_SEGMENT_ALIGN not in SECTIONS clause"),
2621 closure->filename(), closure->lineno(), closure->charpos());
2622 else
2623 closure->script_options()->script_sections()->data_segment_relro_end();
2624 }
2625
2626 // Create a new list of string/sort pairs.
2627
2628 extern "C" String_sort_list_ptr
2629 script_new_string_sort_list(const struct Wildcard_section* string_sort)
2630 {
2631 return new String_sort_list(1, *string_sort);
2632 }
2633
2634 // Add an entry to a list of string/sort pairs. The way the parser
2635 // works permits us to simply modify the first parameter, rather than
2636 // copy the vector.
2637
2638 extern "C" String_sort_list_ptr
2639 script_string_sort_list_add(String_sort_list_ptr pv,
2640 const struct Wildcard_section* string_sort)
2641 {
2642 if (pv == NULL)
2643 return script_new_string_sort_list(string_sort);
2644 else
2645 {
2646 pv->push_back(*string_sort);
2647 return pv;
2648 }
2649 }
2650
2651 // Create a new list of strings.
2652
2653 extern "C" String_list_ptr
2654 script_new_string_list(const char* str, size_t len)
2655 {
2656 return new String_list(1, std::string(str, len));
2657 }
2658
2659 // Add an element to a list of strings. The way the parser works
2660 // permits us to simply modify the first parameter, rather than copy
2661 // the vector.
2662
2663 extern "C" String_list_ptr
2664 script_string_list_push_back(String_list_ptr pv, const char* str, size_t len)
2665 {
2666 if (pv == NULL)
2667 return script_new_string_list(str, len);
2668 else
2669 {
2670 pv->push_back(std::string(str, len));
2671 return pv;
2672 }
2673 }
2674
2675 // Concatenate two string lists. Either or both may be NULL. The way
2676 // the parser works permits us to modify the parameters, rather than
2677 // copy the vector.
2678
2679 extern "C" String_list_ptr
2680 script_string_list_append(String_list_ptr pv1, String_list_ptr pv2)
2681 {
2682 if (pv1 == NULL)
2683 return pv2;
2684 if (pv2 == NULL)
2685 return pv1;
2686 pv1->insert(pv1->end(), pv2->begin(), pv2->end());
2687 return pv1;
2688 }
2689
2690 // Add a new program header.
2691
2692 extern "C" void
2693 script_add_phdr(void* closurev, const char* name, size_t namelen,
2694 unsigned int type, const Phdr_info* info)
2695 {
2696 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2697 bool includes_filehdr = info->includes_filehdr != 0;
2698 bool includes_phdrs = info->includes_phdrs != 0;
2699 bool is_flags_valid = info->is_flags_valid != 0;
2700 Script_sections* ss = closure->script_options()->script_sections();
2701 ss->add_phdr(name, namelen, type, includes_filehdr, includes_phdrs,
2702 is_flags_valid, info->flags, info->load_address);
2703 closure->clear_skip_on_incompatible_target();
2704 }
2705
2706 // Convert a program header string to a type.
2707
2708 #define PHDR_TYPE(NAME) { #NAME, sizeof(#NAME) - 1, elfcpp::NAME }
2709
2710 static struct
2711 {
2712 const char* name;
2713 size_t namelen;
2714 unsigned int val;
2715 } phdr_type_names[] =
2716 {
2717 PHDR_TYPE(PT_NULL),
2718 PHDR_TYPE(PT_LOAD),
2719 PHDR_TYPE(PT_DYNAMIC),
2720 PHDR_TYPE(PT_INTERP),
2721 PHDR_TYPE(PT_NOTE),
2722 PHDR_TYPE(PT_SHLIB),
2723 PHDR_TYPE(PT_PHDR),
2724 PHDR_TYPE(PT_TLS),
2725 PHDR_TYPE(PT_GNU_EH_FRAME),
2726 PHDR_TYPE(PT_GNU_STACK),
2727 PHDR_TYPE(PT_GNU_RELRO)
2728 };
2729
2730 extern "C" unsigned int
2731 script_phdr_string_to_type(void* closurev, const char* name, size_t namelen)
2732 {
2733 for (unsigned int i = 0;
2734 i < sizeof(phdr_type_names) / sizeof(phdr_type_names[0]);
2735 ++i)
2736 if (namelen == phdr_type_names[i].namelen
2737 && strncmp(name, phdr_type_names[i].name, namelen) == 0)
2738 return phdr_type_names[i].val;
2739 yyerror(closurev, _("unknown PHDR type (try integer)"));
2740 return elfcpp::PT_NULL;
2741 }
2742
2743 extern "C" void
2744 script_saw_segment_start_expression(void* closurev)
2745 {
2746 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2747 Script_sections* ss = closure->script_options()->script_sections();
2748 ss->set_saw_segment_start_expression(true);
2749 }
This page took 0.148171 seconds and 5 git commands to generate.