1 /* yyscript.y -- linker script grammar for gold. */
3 /* Copyright 2006, 2007, 2008 Free Software Foundation, Inc.
4 Written by Ian Lance Taylor <iant@google.com>.
6 This file is part of gold.
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.
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.
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. */
23 /* This is a bison grammar to parse a subset of the original GNU ld
24 linker script language. */
39 /* We need to use a pure parser because we might be multi-threaded.
40 We pass some arguments through the parser to the lexer. */
44 %parse-param {void* closure}
45 %lex-param {void* closure}
47 /* Since we require bison anyhow, we take advantage of it. */
51 /* The values associated with tokens. */
55 struct Parser_string string;
60 /* An output section header. */
61 struct Parser_output_section_header output_section_header;
62 /* An output section trailer. */
63 struct Parser_output_section_trailer output_section_trailer;
64 /* A section constraint. */
65 enum Section_constraint constraint;
66 /* A complete input section specification. */
67 struct Input_section_spec input_section_spec;
68 /* A list of wildcard specifications, with exclusions. */
69 struct Wildcard_sections wildcard_sections;
70 /* A single wildcard specification. */
71 struct Wildcard_section wildcard_section;
72 /* A list of strings. */
73 String_list_ptr string_list;
74 /* Information for a program header. */
75 struct Phdr_info phdr_info;
76 /* Used for version scripts and within VERSION {}. */
77 struct Version_dependency_list* deplist;
78 struct Version_expression_list* versyms;
79 struct Version_tree* versnode;
82 /* Operators, including a precedence table for expressions. */
84 %right PLUSEQ MINUSEQ MULTEQ DIVEQ '=' LSHIFTEQ RSHIFTEQ ANDEQ OREQ
97 /* A fake operator used to indicate unary operator precedence. */
102 %token <string> STRING
103 %token <string> QUOTED_STRING
104 %token <integer> INTEGER
106 /* Keywords. This list is taken from ldgram.y and ldlex.l in the old
107 GNU linker, with the keywords which only appear in MRI mode
108 removed. Not all these keywords are actually used in this grammar.
109 In most cases the keyword is recognized as the token name in upper
110 case. The comments indicate where this is not the case. */
114 %token ALIGN_K /* ALIGN */
116 %token ASSERT_K /* ASSERT */
124 %token CREATE_OBJECT_SYMBOLS
125 %token DATA_SEGMENT_ALIGN
126 %token DATA_SEGMENT_END
127 %token DATA_SEGMENT_RELRO_END
134 %token FORCE_COMMON_ALLOCATION
135 %token GLOBAL /* global */
139 %token INHIBIT_COMMON_ALLOCATION
142 %token LENGTH /* LENGTH, l, len */
144 %token LOCAL /* local */
147 %token MAX_K /* MAX */
149 %token MIN_K /* MIN */
155 %token ORIGIN /* ORIGIN, o, org */
162 %token PROVIDE_HIDDEN
169 %token SIZEOF_HEADERS /* SIZEOF_HEADERS, sizeof_headers */
170 %token SORT_BY_ALIGNMENT
177 %token TARGET_K /* TARGET */
179 %token VERSIONK /* VERSION */
181 /* Keywords, part 2. These are keywords that are unique to gold,
182 and not present in the old GNU linker. As before, unless the
183 comments say otherwise, the keyword is recognized as the token
184 name in upper case. */
188 /* Special tokens used to tell the grammar what type of tokens we are
189 parsing. The token stream always begins with one of these tokens.
190 We do this because version scripts can appear embedded within
191 linker scripts, and because --defsym uses the expression
193 %token PARSING_LINKER_SCRIPT
194 %token PARSING_VERSION_SCRIPT
195 %token PARSING_DEFSYM
196 %token PARSING_DYNAMIC_LIST
198 /* Non-terminal types, where needed. */
200 %type <expr> parse_exp exp opt_address_and_section_type
201 %type <expr> opt_at opt_align opt_subalign opt_fill
202 %type <output_section_header> section_header
203 %type <output_section_trailer> section_trailer
204 %type <constraint> opt_constraint
205 %type <string_list> opt_phdr
206 %type <integer> data_length
207 %type <input_section_spec> input_section_no_keep
208 %type <wildcard_sections> wildcard_sections
209 %type <wildcard_section> wildcard_file wildcard_section
210 %type <string_list> exclude_names
211 %type <string> wildcard_name
212 %type <integer> phdr_type
213 %type <phdr_info> phdr_info
214 %type <versyms> vers_defns
215 %type <versnode> vers_tag
216 %type <deplist> verdep
217 %type <string> string
221 /* Read the special token to see what to read next. */
223 PARSING_LINKER_SCRIPT linker_script
224 | PARSING_VERSION_SCRIPT version_script
225 | PARSING_DEFSYM defsym_expr
226 | PARSING_DYNAMIC_LIST dynamic_list_expr
229 /* A file contains a list of commands. */
231 linker_script file_cmd
235 /* A command which may appear at top level of a linker script. */
237 FORCE_COMMON_ALLOCATION
238 { script_set_common_allocation(closure, 1); }
240 { script_start_group(closure); }
242 { script_end_group(closure); }
243 | INHIBIT_COMMON_ALLOCATION
244 { script_set_common_allocation(closure, 0); }
245 | INPUT '(' input_list ')'
246 | OPTION '(' string ')'
247 { script_parse_option(closure, $3.value, $3.length); }
248 | OUTPUT_FORMAT '(' string ')'
250 if (!script_check_output_format(closure, $3.value, $3.length,
254 | OUTPUT_FORMAT '(' string ',' string ',' string ')'
256 if (!script_check_output_format(closure, $3.value, $3.length,
258 $7.value, $7.length))
261 | PHDRS '{' phdrs_defs '}'
262 | SEARCH_DIR '(' string ')'
263 { script_add_search_dir(closure, $3.value, $3.length); }
265 { script_start_sections(closure); }
267 { script_finish_sections(closure); }
269 { script_push_lex_into_version_mode(closure); }
271 { script_pop_lex_mode(closure); }
272 | file_or_sections_cmd
277 /* Top level commands which we ignore. The GNU linker uses these to
278 select the output format, but we don't offer a choice. Ignoring
279 these is more-or-less OK since most scripts simply explicitly
280 choose the default. */
282 OUTPUT_ARCH '(' string ')'
285 /* A list of input file names. */
288 | input_list opt_comma input_list_element
291 /* An input file name. */
294 { script_add_file(closure, $1.value, $1.length); }
296 { script_start_as_needed(closure); }
298 { script_end_as_needed(closure); }
301 /* Commands in a SECTIONS block. */
303 sections_block section_block_cmd
307 /* A command which may appear within a SECTIONS block. */
310 | string section_header
311 { script_start_output_section(closure, $1.value, $1.length, &$2); }
312 '{' section_cmds '}' section_trailer
313 { script_finish_output_section(closure, &$7); }
316 /* The header of an output section in a SECTIONS block--everything
319 { script_push_lex_into_expression_mode(closure); }
320 opt_address_and_section_type opt_at opt_align opt_subalign
321 { script_pop_lex_mode(closure); }
325 $$.load_address = $3;
332 /* The optional address followed by the optional section type. This
333 is a separate nonterminal to avoid a shift/reduce conflict on
334 '(' in section_header. */
336 opt_address_and_section_type:
345 | exp '(' string ')' ':'
347 yyerror(closure, "section types are not supported");
352 /* The address at which an output section should be loaded. */
360 /* The alignment of an output section. */
364 | ALIGN_K '(' exp ')'
368 /* The input section alignment within an output section. */
372 | SUBALIGN '(' exp ')'
376 /* A section constraint. */
379 { $$ = CONSTRAINT_NONE; }
381 { $$ = CONSTRAINT_ONLY_IF_RO; }
383 { $$ = CONSTRAINT_ONLY_IF_RW; }
385 { $$ = CONSTRAINT_SPECIAL; }
388 /* The trailer of an output section in a SECTIONS block. */
390 opt_memspec opt_at_memspec opt_phdr opt_fill opt_comma
397 /* A memory specification for an output section. */
400 { yyerror(closure, "memory regions are not supported"); }
404 /* A memory specification for where to load an output section. */
407 { yyerror(closure, "memory regions are not supported"); }
411 /* The program segment an output section should go into. */
414 { $$ = script_string_list_push_back($1, $3.value, $3.length); }
419 /* The value to use to fill an output section. FIXME: This does not
420 handle a string of arbitrary length. */
428 /* Commands which may appear within the description of an output
429 section in a SECTIONS block. */
432 | section_cmds section_cmd
435 /* A command which may appear within the description of an output
436 section in a SECTIONS block. */
440 | data_length '(' parse_exp ')'
441 { script_add_data(closure, $1, $3); }
442 | ASSERT_K '(' parse_exp ',' string ')'
443 { script_add_assertion(closure, $3, $5.value, $5.length); }
444 | FILL '(' parse_exp ')'
445 { script_add_fill(closure, $3); }
448 /* The GNU linker uses CONSTRUCTORS for the a.out object
449 file format. It does nothing when using ELF. Since
450 some ELF linker scripts use it although it does
451 nothing, we accept it and ignore it. */
453 | SORT_BY_NAME '(' CONSTRUCTORS ')'
457 /* The length of data which may appear within the description of an
458 output section in a SECTIONS block. */
472 /* An input section specification. This may appear within the
473 description of an output section in a SECTIONS block. */
475 input_section_no_keep
476 { script_add_input_section(closure, &$1, 0); }
477 | KEEP '(' input_section_no_keep ')'
478 { script_add_input_section(closure, &$3, 1); }
481 /* An input section specification within a KEEP clause. */
482 input_section_no_keep:
486 $$.file.sort = SORT_WILDCARD_NONE;
487 $$.input_sections.sections = NULL;
488 $$.input_sections.exclude = NULL;
490 | wildcard_file '(' wildcard_sections ')'
493 $$.input_sections = $3;
497 /* A wildcard file specification. */
502 $$.sort = SORT_WILDCARD_NONE;
504 | SORT_BY_NAME '(' wildcard_name ')'
507 $$.sort = SORT_WILDCARD_BY_NAME;
511 /* A list of wild card section specifications. */
513 wildcard_sections opt_comma wildcard_section
515 $$.sections = script_string_sort_list_add($1.sections, &$3);
516 $$.exclude = $1.exclude;
520 $$.sections = script_new_string_sort_list(&$1);
523 | wildcard_sections opt_comma EXCLUDE_FILE '(' exclude_names ')'
525 $$.sections = $1.sections;
526 $$.exclude = script_string_list_append($1.exclude, $5);
528 | EXCLUDE_FILE '(' exclude_names ')'
535 /* A single wild card specification. */
540 $$.sort = SORT_WILDCARD_NONE;
542 | SORT_BY_NAME '(' wildcard_section ')'
547 case SORT_WILDCARD_NONE:
548 $$.sort = SORT_WILDCARD_BY_NAME;
550 case SORT_WILDCARD_BY_NAME:
551 case SORT_WILDCARD_BY_NAME_BY_ALIGNMENT:
553 case SORT_WILDCARD_BY_ALIGNMENT:
554 case SORT_WILDCARD_BY_ALIGNMENT_BY_NAME:
555 $$.sort = SORT_WILDCARD_BY_NAME_BY_ALIGNMENT;
561 | SORT_BY_ALIGNMENT '(' wildcard_section ')'
566 case SORT_WILDCARD_NONE:
567 $$.sort = SORT_WILDCARD_BY_ALIGNMENT;
569 case SORT_WILDCARD_BY_ALIGNMENT:
570 case SORT_WILDCARD_BY_ALIGNMENT_BY_NAME:
572 case SORT_WILDCARD_BY_NAME:
573 case SORT_WILDCARD_BY_NAME_BY_ALIGNMENT:
574 $$.sort = SORT_WILDCARD_BY_ALIGNMENT_BY_NAME;
582 /* A list of file names to exclude. */
584 exclude_names opt_comma wildcard_name
585 { $$ = script_string_list_push_back($1, $3.value, $3.length); }
587 { $$ = script_new_string_list($1.value, $1.length); }
590 /* A single wildcard name. We recognize '*' and '?' specially since
591 they are expression tokens. */
607 /* A command which may appear at the top level of a linker script, or
608 within a SECTIONS block. */
609 file_or_sections_cmd:
611 { script_set_entry(closure, $3.value, $3.length); }
613 | ASSERT_K '(' parse_exp ',' string ')'
614 { script_add_assertion(closure, $3, $5.value, $5.length); }
617 /* A list of program header definitions. */
623 /* A program header definition. */
625 string phdr_type phdr_info ';'
626 { script_add_phdr(closure, $1.value, $1.length, $2, &$3); }
629 /* A program header type. The GNU linker accepts a general expression
630 here, but that would be a pain because we would have to dig into
631 the expression structure. It's unlikely that anybody uses anything
632 other than a string or a number here, so that is all we expect. */
635 { $$ = script_phdr_string_to_type(closure, $1.value, $1.length); }
640 /* Additional information for a program header. */
643 { memset(&$$, 0, sizeof(struct Phdr_info)); }
647 if ($1.length == 7 && strncmp($1.value, "FILEHDR", 7) == 0)
648 $$.includes_filehdr = 1;
650 yyerror(closure, "PHDRS syntax error");
655 $$.includes_phdrs = 1;
657 | string '(' INTEGER ')' phdr_info
660 if ($1.length == 5 && strncmp($1.value, "FLAGS", 5) == 0)
662 $$.is_flags_valid = 1;
666 yyerror(closure, "PHDRS syntax error");
668 | AT '(' parse_exp ')' phdr_info
671 $$.load_address = $3;
675 /* Set a symbol to a value. */
678 { script_set_symbol(closure, $1.value, $1.length, $3, 0, 0); }
679 | string PLUSEQ parse_exp
681 Expression_ptr s = script_exp_string($1.value, $1.length);
682 Expression_ptr e = script_exp_binary_add(s, $3);
683 script_set_symbol(closure, $1.value, $1.length, e, 0, 0);
685 | string MINUSEQ parse_exp
687 Expression_ptr s = script_exp_string($1.value, $1.length);
688 Expression_ptr e = script_exp_binary_sub(s, $3);
689 script_set_symbol(closure, $1.value, $1.length, e, 0, 0);
691 | string MULTEQ parse_exp
693 Expression_ptr s = script_exp_string($1.value, $1.length);
694 Expression_ptr e = script_exp_binary_mult(s, $3);
695 script_set_symbol(closure, $1.value, $1.length, e, 0, 0);
697 | string DIVEQ parse_exp
699 Expression_ptr s = script_exp_string($1.value, $1.length);
700 Expression_ptr e = script_exp_binary_div(s, $3);
701 script_set_symbol(closure, $1.value, $1.length, e, 0, 0);
703 | string LSHIFTEQ parse_exp
705 Expression_ptr s = script_exp_string($1.value, $1.length);
706 Expression_ptr e = script_exp_binary_lshift(s, $3);
707 script_set_symbol(closure, $1.value, $1.length, e, 0, 0);
709 | string RSHIFTEQ parse_exp
711 Expression_ptr s = script_exp_string($1.value, $1.length);
712 Expression_ptr e = script_exp_binary_rshift(s, $3);
713 script_set_symbol(closure, $1.value, $1.length, e, 0, 0);
715 | string ANDEQ parse_exp
717 Expression_ptr s = script_exp_string($1.value, $1.length);
718 Expression_ptr e = script_exp_binary_bitwise_and(s, $3);
719 script_set_symbol(closure, $1.value, $1.length, e, 0, 0);
721 | string OREQ parse_exp
723 Expression_ptr s = script_exp_string($1.value, $1.length);
724 Expression_ptr e = script_exp_binary_bitwise_or(s, $3);
725 script_set_symbol(closure, $1.value, $1.length, e, 0, 0);
727 | PROVIDE '(' string '=' parse_exp ')'
728 { script_set_symbol(closure, $3.value, $3.length, $5, 1, 0); }
729 | PROVIDE_HIDDEN '(' string '=' parse_exp ')'
730 { script_set_symbol(closure, $3.value, $3.length, $5, 1, 1); }
733 /* Parse an expression, putting the lexer into the right mode. */
735 { script_push_lex_into_expression_mode(closure); }
738 script_pop_lex_mode(closure);
747 | '-' exp %prec UNARY
748 { $$ = script_exp_unary_minus($2); }
749 | '!' exp %prec UNARY
750 { $$ = script_exp_unary_logical_not($2); }
751 | '~' exp %prec UNARY
752 { $$ = script_exp_unary_bitwise_not($2); }
753 | '+' exp %prec UNARY
756 { $$ = script_exp_binary_mult($1, $3); }
758 { $$ = script_exp_binary_div($1, $3); }
760 { $$ = script_exp_binary_mod($1, $3); }
762 { $$ = script_exp_binary_add($1, $3); }
764 { $$ = script_exp_binary_sub($1, $3); }
766 { $$ = script_exp_binary_lshift($1, $3); }
768 { $$ = script_exp_binary_rshift($1, $3); }
770 { $$ = script_exp_binary_eq($1, $3); }
772 { $$ = script_exp_binary_ne($1, $3); }
774 { $$ = script_exp_binary_le($1, $3); }
776 { $$ = script_exp_binary_ge($1, $3); }
778 { $$ = script_exp_binary_lt($1, $3); }
780 { $$ = script_exp_binary_gt($1, $3); }
782 { $$ = script_exp_binary_bitwise_and($1, $3); }
784 { $$ = script_exp_binary_bitwise_xor($1, $3); }
786 { $$ = script_exp_binary_bitwise_or($1, $3); }
788 { $$ = script_exp_binary_logical_and($1, $3); }
790 { $$ = script_exp_binary_logical_or($1, $3); }
791 | exp '?' exp ':' exp
792 { $$ = script_exp_trinary_cond($1, $3, $5); }
794 { $$ = script_exp_integer($1); }
796 { $$ = script_exp_string($1.value, $1.length); }
797 | MAX_K '(' exp ',' exp ')'
798 { $$ = script_exp_function_max($3, $5); }
799 | MIN_K '(' exp ',' exp ')'
800 { $$ = script_exp_function_min($3, $5); }
801 | DEFINED '(' string ')'
802 { $$ = script_exp_function_defined($3.value, $3.length); }
804 { $$ = script_exp_function_sizeof_headers(); }
805 | ALIGNOF '(' string ')'
806 { $$ = script_exp_function_alignof($3.value, $3.length); }
807 | SIZEOF '(' string ')'
808 { $$ = script_exp_function_sizeof($3.value, $3.length); }
809 | ADDR '(' string ')'
810 { $$ = script_exp_function_addr($3.value, $3.length); }
811 | LOADADDR '(' string ')'
812 { $$ = script_exp_function_loadaddr($3.value, $3.length); }
813 | ORIGIN '(' string ')'
814 { $$ = script_exp_function_origin($3.value, $3.length); }
815 | LENGTH '(' string ')'
816 { $$ = script_exp_function_length($3.value, $3.length); }
817 | CONSTANT '(' string ')'
818 { $$ = script_exp_function_constant($3.value, $3.length); }
819 | ABSOLUTE '(' exp ')'
820 { $$ = script_exp_function_absolute($3); }
821 | ALIGN_K '(' exp ')'
822 { $$ = script_exp_function_align(script_exp_string(".", 1), $3); }
823 | ALIGN_K '(' exp ',' exp ')'
824 { $$ = script_exp_function_align($3, $5); }
826 { $$ = script_exp_function_align(script_exp_string(".", 1), $3); }
827 | DATA_SEGMENT_ALIGN '(' exp ',' exp ')'
829 script_data_segment_align(closure);
830 $$ = script_exp_function_data_segment_align($3, $5);
832 | DATA_SEGMENT_RELRO_END '(' exp ',' exp ')'
834 script_data_segment_relro_end(closure);
835 $$ = script_exp_function_data_segment_relro_end($3, $5);
837 | DATA_SEGMENT_END '(' exp ')'
838 { $$ = script_exp_function_data_segment_end($3); }
839 | SEGMENT_START '(' string ',' exp ')'
841 $$ = script_exp_function_segment_start($3.value, $3.length, $5);
843 | ASSERT_K '(' exp ',' string ')'
844 { $$ = script_exp_function_assert($3, $5.value, $5.length); }
847 /* Handle the --defsym option. */
850 { script_set_symbol(closure, $1.value, $1.length, $3, 0, 0); }
853 /* Handle the --dynamic-list option. A dynamic list has the format
854 { sym1; sym2; extern "C++" { namespace::sym3 }; };
855 We store the symbol we see in the "local" list; that is where
856 Command_line::in_dynamic_list() will look to do its check.
857 TODO(csilvers): More than one of these brace-lists can appear, and
858 should just be merged and treated as a single list. */
859 dynamic_list_expr: dynamic_list_nodes ;
863 | dynamic_list_nodes dynamic_list_node
867 '{' vers_defns ';' '}' ';'
868 { script_new_vers_node (closure, NULL, $2); }
871 /* A version script. */
878 | vers_nodes vers_node
884 script_register_vers_node (closure, NULL, 0, $2, NULL);
886 | string '{' vers_tag '}' ';'
888 script_register_vers_node (closure, $1.value, $1.length, $3,
891 | string '{' vers_tag '}' verdep ';'
893 script_register_vers_node (closure, $1.value, $1.length, $3, $5);
900 $$ = script_add_vers_depend (closure, NULL, $1.value, $1.length);
904 $$ = script_add_vers_depend (closure, $1, $2.value, $2.length);
910 { $$ = script_new_vers_node (closure, NULL, NULL); }
912 { $$ = script_new_vers_node (closure, $1, NULL); }
913 | GLOBAL ':' vers_defns ';'
914 { $$ = script_new_vers_node (closure, $3, NULL); }
915 | LOCAL ':' vers_defns ';'
916 { $$ = script_new_vers_node (closure, NULL, $3); }
917 | GLOBAL ':' vers_defns ';' LOCAL ':' vers_defns ';'
918 { $$ = script_new_vers_node (closure, $3, $7); }
921 /* Here is one of the rare places we care about the distinction
922 between STRING and QUOTED_STRING. For QUOTED_STRING, we do exact
923 matching on the pattern, so we pass in true for the exact_match
924 parameter. For STRING, we do glob matching and pass in false. */
928 $$ = script_new_vers_pattern (closure, NULL, $1.value,
933 $$ = script_new_vers_pattern (closure, NULL, $1.value,
936 | vers_defns ';' STRING
938 $$ = script_new_vers_pattern (closure, $1, $3.value,
941 | vers_defns ';' QUOTED_STRING
943 $$ = script_new_vers_pattern (closure, $1, $3.value,
946 | /* Push string on the language stack. */
948 { version_script_push_lang (closure, $2.value, $2.length); }
949 vers_defns opt_semicolon '}'
952 version_script_pop_lang(closure);
954 | /* Push string on the language stack. This is more complicated
955 than the other cases because we need to merge the linked-list
956 state from the pre-EXTERN defns and the post-EXTERN defns. */
957 vers_defns ';' EXTERN string '{'
958 { version_script_push_lang (closure, $4.value, $4.length); }
959 vers_defns opt_semicolon '}'
961 $$ = script_merge_expressions ($1, $7);
962 version_script_pop_lang(closure);
964 | EXTERN // "extern" as a symbol name
966 $$ = script_new_vers_pattern (closure, NULL, "extern",
967 sizeof("extern") - 1, 1);
969 | vers_defns ';' EXTERN
971 $$ = script_new_vers_pattern (closure, $1, "extern",
972 sizeof("extern") - 1, 1);
976 /* A string can be either a STRING or a QUOTED_STRING. Almost all the
977 time we don't care, and we use this rule. */
985 /* Some statements require a terminator, which may be a semicolon or a
992 /* An optional semicolon. */
998 /* An optional comma. */