X-Git-Url: http://review.tizen.org/git/?a=blobdiff_plain;f=gold%2Fscript.cc;h=eff9a81a6cefa611debeea73c30f4bedebc050c3;hb=14ea52eecb11790c7a03eda1b13568b587238b66;hp=973c05ce6df9d5c7c351e5da4a24622802689b80;hpb=3802b2dd6b937e2904b6e2de087e224437eab493;p=platform%2Fupstream%2Fbinutils.git diff --git a/gold/script.cc b/gold/script.cc index 973c05c..eff9a81 100644 --- a/gold/script.cc +++ b/gold/script.cc @@ -1,6 +1,6 @@ // script.cc -- handle linker scripts for gold. -// Copyright 2006, 2007, 2008 Free Software Foundation, Inc. +// Copyright (C) 2006-2014 Free Software Foundation, Inc. // Written by Ian Lance Taylor . // This file is part of gold. @@ -22,11 +22,12 @@ #include "gold.h" +#include +#include +#include #include #include #include -#include -#include #include "filenames.h" #include "elfcpp.h" @@ -39,8 +40,10 @@ #include "parameters.h" #include "layout.h" #include "symtab.h" +#include "target-select.h" #include "script.h" #include "script-c.h" +#include "incremental.h" namespace gold { @@ -143,13 +146,7 @@ class Token } uint64_t - integer_value() const - { - gold_assert(this->classification_ == TOKEN_INTEGER); - // Null terminate. - std::string s(this->value_, this->value_length_); - return strtoull(s.c_str(), NULL, 0); - } + integer_value() const; private: // The token classification. @@ -168,6 +165,35 @@ class Token int charpos_; }; +// Return the value of a TOKEN_INTEGER. + +uint64_t +Token::integer_value() const +{ + gold_assert(this->classification_ == TOKEN_INTEGER); + + size_t len = this->value_length_; + + uint64_t multiplier = 1; + char last = this->value_[len - 1]; + if (last == 'm' || last == 'M') + { + multiplier = 1024 * 1024; + --len; + } + else if (last == 'k' || last == 'K') + { + multiplier = 1024; + --len; + } + + char *end; + uint64_t ret = strtoull(this->value_, &end, 0); + gold_assert(static_cast(end - this->value_) == len); + + return ret * multiplier; +} + // This class handles lexing a file into a sequence of tokens. class Lex @@ -183,7 +209,9 @@ class Lex // Reading an expression in a linker script. EXPRESSION, // Reading a version script. - VERSION_SCRIPT + VERSION_SCRIPT, + // Reading a --dynamic-list file. + DYNAMIC_LIST }; Lex(const char* input_string, size_t input_length, int parsing_token) @@ -392,8 +420,9 @@ Lex::can_start_name(char c, char c2) case '~': return this->mode_ == LINKER_SCRIPT && can_continue_name(&c2); - case '*': case '[': + case '*': case '[': return (this->mode_ == VERSION_SCRIPT + || this->mode_ == DYNAMIC_LIST || (this->mode_ == LINKER_SCRIPT && can_continue_name(&c2))); @@ -428,27 +457,31 @@ Lex::can_continue_name(const char* c) case '5': case '6': case '7': case '8': case '9': return c + 1; + // TODO(csilvers): why not allow ~ in names for version-scripts? case '/': case '\\': case '~': case '=': case '+': - case ',': case '?': + case ',': if (this->mode_ == LINKER_SCRIPT) return c + 1; return NULL; - case '[': case ']': case '*': case '-': - if (this->mode_ == LINKER_SCRIPT || this->mode_ == VERSION_SCRIPT) + case '[': case ']': case '*': case '?': case '-': + if (this->mode_ == LINKER_SCRIPT || this->mode_ == VERSION_SCRIPT + || this->mode_ == DYNAMIC_LIST) return c + 1; return NULL; + // TODO(csilvers): why allow this? ^ is meaningless in version scripts. case '^': - if (this->mode_ == VERSION_SCRIPT) + if (this->mode_ == VERSION_SCRIPT || this->mode_ == DYNAMIC_LIST) return c + 1; return NULL; case ':': if (this->mode_ == LINKER_SCRIPT) return c + 1; - else if (this->mode_ == VERSION_SCRIPT && (c[1] == ':')) + else if ((this->mode_ == VERSION_SCRIPT || this->mode_ == DYNAMIC_LIST) + && (c[1] == ':')) { // A name can have '::' in it, as that's a c++ namespace // separator. But a single colon is not part of a name. @@ -464,9 +497,7 @@ Lex::can_continue_name(const char* c) // For a number we accept 0x followed by hex digits, or any sequence // of digits. The old linker accepts leading '$' for hex, and // trailing HXBOD. Those are for MRI compatibility and we don't -// accept them. The old linker also accepts trailing MK for mega or -// kilo. FIXME: Those are mentioned in the documentation, and we -// should accept them. +// accept them. // Return whether C1 C2 C3 can start a hex number. @@ -690,11 +721,18 @@ Lex::gather_token(Token::Classification classification, const char* (Lex::*can_continue_fn)(const char*), const char* start, const char* match, - const char **pp) + const char** pp) { const char* new_match = NULL; - while ((new_match = (this->*can_continue_fn)(match))) + while ((new_match = (this->*can_continue_fn)(match)) != NULL) match = new_match; + + // A special case: integers may be followed by a single M or K, + // case-insensitive. + if (classification == Token::TOKEN_INTEGER + && (*match == 'm' || *match == 'M' || *match == 'k' || *match == 'K')) + ++match; + *pp = match; return this->make_token(classification, start, match - start, start); } @@ -733,7 +771,7 @@ Lex::get_token(const char** pp) } // Skip whitespace quickly. - while (*p == ' ' || *p == '\t') + while (*p == ' ' || *p == '\t' || *p == '\r') ++p; if (*p == '\n') @@ -846,47 +884,6 @@ Lex::next_token() return &this->token_; } -// A trivial task which waits for THIS_BLOCKER to be clear and then -// clears NEXT_BLOCKER. THIS_BLOCKER may be NULL. - -class Script_unblock : public Task -{ - public: - Script_unblock(Task_token* this_blocker, Task_token* next_blocker) - : this_blocker_(this_blocker), next_blocker_(next_blocker) - { } - - ~Script_unblock() - { - if (this->this_blocker_ != NULL) - delete this->this_blocker_; - } - - Task_token* - is_runnable() - { - if (this->this_blocker_ != NULL && this->this_blocker_->is_blocked()) - return this->this_blocker_; - return NULL; - } - - void - locks(Task_locker* tl) - { tl->add(this, this->next_blocker_); } - - void - run(Workqueue*) - { } - - std::string - get_name() const - { return "Script_unblock"; } - - private: - Task_token* this_blocker_; - Task_token* next_blocker_; -}; - // class Symbol_assignment. // Add the symbol to the symbol table. This makes sure the symbol is @@ -895,19 +892,27 @@ class Script_unblock : public Task // necessarily evaluate the expression until all ordinary symbols have // been finalized. +// The GNU linker lets symbol assignments in the linker script +// silently override defined symbols in object files. We are +// compatible. FIXME: Should we issue a warning? + void Symbol_assignment::add_to_table(Symbol_table* symtab) { elfcpp::STV vis = this->hidden_ ? elfcpp::STV_HIDDEN : elfcpp::STV_DEFAULT; this->sym_ = symtab->define_as_constant(this->name_.c_str(), NULL, // version + (this->is_defsym_ + ? Symbol_table::DEFSYM + : Symbol_table::SCRIPT), 0, // value 0, // size elfcpp::STT_NOTYPE, elfcpp::STB_GLOBAL, vis, 0, // nonvis - this->provide_); + this->provide_, + true); // force_override } // Finalize a symbol value. @@ -915,7 +920,7 @@ Symbol_assignment::add_to_table(Symbol_table* symtab) void Symbol_assignment::finalize(Symbol_table* symtab, const Layout* layout) { - this->finalize_maybe_dot(symtab, layout, false, false, 0); + this->finalize_maybe_dot(symtab, layout, false, 0, NULL); } // Finalize a symbol value which can refer to the dot symbol. @@ -923,10 +928,10 @@ Symbol_assignment::finalize(Symbol_table* symtab, const Layout* layout) void Symbol_assignment::finalize_with_dot(Symbol_table* symtab, const Layout* layout, - bool dot_has_value, - uint64_t dot_value) + uint64_t dot_value, + Output_section* dot_section) { - this->finalize_maybe_dot(symtab, layout, true, dot_has_value, dot_value); + this->finalize_maybe_dot(symtab, layout, true, dot_value, dot_section); } // Finalize a symbol value, internal version. @@ -935,8 +940,8 @@ void Symbol_assignment::finalize_maybe_dot(Symbol_table* symtab, const Layout* layout, bool is_dot_available, - bool dot_has_value, - uint64_t dot_value) + uint64_t dot_value, + Output_section* dot_section) { // If we were only supposed to provide this symbol, the sym_ field // will be NULL if the symbol was not referenced. @@ -946,20 +951,20 @@ Symbol_assignment::finalize_maybe_dot(Symbol_table* symtab, return; } - if (parameters->get_size() == 32) + if (parameters->target().get_size() == 32) { #if defined(HAVE_TARGET_32_LITTLE) || defined(HAVE_TARGET_32_BIG) - this->sized_finalize<32>(symtab, layout, is_dot_available, dot_has_value, - dot_value); + this->sized_finalize<32>(symtab, layout, is_dot_available, dot_value, + dot_section); #else gold_unreachable(); #endif } - else if (parameters->get_size() == 64) + else if (parameters->target().get_size() == 64) { #if defined(HAVE_TARGET_64_LITTLE) || defined(HAVE_TARGET_64_BIG) - this->sized_finalize<64>(symtab, layout, is_dot_available, dot_has_value, - dot_value); + this->sized_finalize<64>(symtab, layout, is_dot_available, dot_value, + dot_section); #else gold_unreachable(); #endif @@ -971,36 +976,47 @@ Symbol_assignment::finalize_maybe_dot(Symbol_table* symtab, template void Symbol_assignment::sized_finalize(Symbol_table* symtab, const Layout* layout, - bool is_dot_available, bool dot_has_value, - uint64_t dot_value) -{ - bool dummy; - uint64_t final_val = this->val_->eval_maybe_dot(symtab, layout, + bool is_dot_available, uint64_t dot_value, + Output_section* dot_section) +{ + Output_section* section; + elfcpp::STT type = elfcpp::STT_NOTYPE; + elfcpp::STV vis = elfcpp::STV_DEFAULT; + unsigned char nonvis = 0; + uint64_t final_val = this->val_->eval_maybe_dot(symtab, layout, true, is_dot_available, - dot_has_value, dot_value, - &dummy); + dot_value, dot_section, + §ion, NULL, &type, + &vis, &nonvis, false); Sized_symbol* ssym = symtab->get_sized_symbol(this->sym_); ssym->set_value(final_val); + ssym->set_type(type); + ssym->set_visibility(vis); + ssym->set_nonvis(nonvis); + if (section != NULL) + ssym->set_output_section(section); } -// Set the symbol value if the expression yields an absolute value. +// Set the symbol value if the expression yields an absolute value or +// a value relative to DOT_SECTION. void Symbol_assignment::set_if_absolute(Symbol_table* symtab, const Layout* layout, - bool is_dot_available, bool dot_has_value, - uint64_t dot_value) + bool is_dot_available, uint64_t dot_value, + Output_section* dot_section) { if (this->sym_ == NULL) return; - bool is_absolute; - uint64_t val = this->val_->eval_maybe_dot(symtab, layout, is_dot_available, - dot_has_value, dot_value, - &is_absolute); - if (!is_absolute) + Output_section* val_section; + uint64_t val = this->val_->eval_maybe_dot(symtab, layout, false, + is_dot_available, dot_value, + dot_section, &val_section, NULL, + NULL, NULL, NULL, false); + if (val_section != NULL && val_section != dot_section) return; - if (parameters->get_size() == 32) + if (parameters->target().get_size() == 32) { #if defined(HAVE_TARGET_32_LITTLE) || defined(HAVE_TARGET_32_BIG) Sized_symbol<32>* ssym = symtab->get_sized_symbol<32>(this->sym_); @@ -1009,7 +1025,7 @@ Symbol_assignment::set_if_absolute(Symbol_table* symtab, const Layout* layout, gold_unreachable(); #endif } - else if (parameters->get_size() == 64) + else if (parameters->target().get_size() == 64) { #if defined(HAVE_TARGET_64_LITTLE) || defined(HAVE_TARGET_64_BIG) Sized_symbol<64>* ssym = symtab->get_sized_symbol<64>(this->sym_); @@ -1020,6 +1036,8 @@ Symbol_assignment::set_if_absolute(Symbol_table* symtab, const Layout* layout, } else gold_unreachable(); + if (val_section != NULL) + this->sym_->set_output_section(val_section); } // Print for debugging. @@ -1050,7 +1068,7 @@ Symbol_assignment::print(FILE* f) const void Script_assertion::check(const Symbol_table* symtab, const Layout* layout) { - if (!this->check_->eval(symtab, layout)) + if (!this->check_->eval(symtab, layout, true)) gold_error("%s", this->message_.c_str()); } @@ -1067,38 +1085,76 @@ Script_assertion::print(FILE* f) const // Class Script_options. Script_options::Script_options() - : entry_(), symbol_assignments_(), version_script_info_(), - script_sections_() + : entry_(), symbol_assignments_(), symbol_definitions_(), + symbol_references_(), version_script_info_(), script_sections_() +{ +} + +// Returns true if NAME is on the list of symbol assignments waiting +// to be processed. + +bool +Script_options::is_pending_assignment(const char* name) { + for (Symbol_assignments::iterator p = this->symbol_assignments_.begin(); + p != this->symbol_assignments_.end(); + ++p) + if ((*p)->name() == name) + return true; + return false; } // Add a symbol to be defined. void Script_options::add_symbol_assignment(const char* name, size_t length, - Expression* value, bool provide, - bool hidden) + bool is_defsym, Expression* value, + bool provide, bool hidden) { if (length != 1 || name[0] != '.') { if (this->script_sections_.in_sections_clause()) - this->script_sections_.add_symbol_assignment(name, length, value, - provide, hidden); - else { - Symbol_assignment* p = new Symbol_assignment(name, length, value, + gold_assert(!is_defsym); + this->script_sections_.add_symbol_assignment(name, length, value, provide, hidden); + } + else + { + Symbol_assignment* p = new Symbol_assignment(name, length, is_defsym, + value, provide, hidden); this->symbol_assignments_.push_back(p); } + + if (!provide) + { + std::string n(name, length); + this->symbol_definitions_.insert(n); + this->symbol_references_.erase(n); + } } else { if (provide || hidden) gold_error(_("invalid use of PROVIDE for dot symbol")); - if (!this->script_sections_.in_sections_clause()) - gold_error(_("invalid assignment to dot outside of SECTIONS")); - else - this->script_sections_.add_dot_assignment(value); + + // The GNU linker permits assignments to dot outside of SECTIONS + // clauses and treats them as occurring inside, so we don't + // check in_sections_clause here. + this->script_sections_.add_dot_assignment(value); + } +} + +// Add a reference to a symbol. + +void +Script_options::add_symbol_reference(const char* name, size_t length) +{ + if (length != 1 || name[0] != '.') + { + std::string n(name, length); + if (this->symbol_definitions_.find(n) == this->symbol_definitions_.end()) + this->symbol_references_.insert(n); } } @@ -1117,6 +1173,15 @@ Script_options::add_assertion(Expression* check, const char* message, } } +// Create sections required by any linker scripts. + +void +Script_options::create_script_sections(Layout* layout) +{ + if (this->saw_sections_clause()) + this->script_sections_.create_sections(layout); +} + // Add any symbols we are defining to the symbol table. void @@ -1134,6 +1199,12 @@ Script_options::add_symbols_to_table(Symbol_table* symtab) void Script_options::finalize_symbols(Symbol_table* symtab, const Layout* layout) { + // We finalize the symbols defined in SECTIONS first, because they + // are the ones which may have changed. This way if symbol outside + // SECTIONS are defined in terms of symbols inside SECTIONS, they + // will get the right value. + this->script_sections_.finalize_symbols(symtab, layout); + for (Symbol_assignments::iterator p = this->symbol_assignments_.begin(); p != this->symbol_assignments_.end(); ++p) @@ -1143,8 +1214,6 @@ Script_options::finalize_symbols(Symbol_table* symtab, const Layout* layout) p != this->assertions_.end(); ++p) (*p)->check(symtab, layout); - - this->script_sections_.finalize_symbols(symtab, layout); } // Set section addresses. We set all the symbols which have absolute @@ -1158,7 +1227,7 @@ Script_options::set_section_addresses(Symbol_table* symtab, Layout* layout) for (Symbol_assignments::iterator p = this->symbol_assignments_.begin(); p != this->symbol_assignments_.end(); ++p) - (*p)->set_if_absolute(symtab, layout, false, false, 0); + (*p)->set_if_absolute(symtab, layout, false, 0, NULL); return this->script_sections_.set_section_addresses(symtab, layout); } @@ -1173,19 +1242,25 @@ class Parser_closure public: Parser_closure(const char* filename, const Position_dependent_options& posdep_options, - bool in_group, bool is_in_sysroot, + bool parsing_defsym, bool in_group, bool is_in_sysroot, Command_line* command_line, Script_options* script_options, - Lex* lex) + Lex* lex, + bool skip_on_incompatible_target, + Script_info* script_info) : filename_(filename), posdep_options_(posdep_options), - in_group_(in_group), is_in_sysroot_(is_in_sysroot), + parsing_defsym_(parsing_defsym), in_group_(in_group), + is_in_sysroot_(is_in_sysroot), + skip_on_incompatible_target_(skip_on_incompatible_target), + found_incompatible_target_(false), command_line_(command_line), script_options_(script_options), version_script_info_(script_options->version_script_info()), - lex_(lex), lineno_(0), charpos_(0), lex_mode_stack_(), inputs_(NULL) - { + lex_(lex), lineno_(0), charpos_(0), lex_mode_stack_(), inputs_(NULL), + script_info_(script_info) + { // We start out processing C symbols in the default lex mode. - language_stack_.push_back(""); - lex_mode_stack_.push_back(lex->mode()); + this->language_stack_.push_back(Version_script_info::LANGUAGE_C); + this->lex_mode_stack_.push_back(lex->mode()); } // Return the file name. @@ -1199,6 +1274,11 @@ class Parser_closure position_dependent_options() { return this->posdep_options_; } + // Whether we are parsing a --defsym. + bool + parsing_defsym() const + { return this->parsing_defsym_; } + // Return whether this script is being run in a group. bool in_group() const @@ -1210,6 +1290,30 @@ class Parser_closure is_in_sysroot() const { return this->is_in_sysroot_; } + // Whether to skip to the next file with the same name if we find an + // incompatible target in an OUTPUT_FORMAT statement. + bool + skip_on_incompatible_target() const + { return this->skip_on_incompatible_target_; } + + // Stop skipping to the next file on an incompatible target. This + // is called when we make some unrevocable change to the data + // structures. + void + clear_skip_on_incompatible_target() + { this->skip_on_incompatible_target_ = false; } + + // Whether we found an incompatible target in an OUTPUT_FORMAT + // statement. + bool + found_incompatible_target() const + { return this->found_incompatible_target_; } + + // Note that we found an incompatible target. + void + set_found_incompatible_target() + { this->found_incompatible_target_ = true; } + // Returns the Command_line structure passed in at constructor time. // This value may be NULL. The caller may modify this, which modifies // the passed-in Command_line object (not a copy). @@ -1286,30 +1390,45 @@ class Parser_closure // Return the current language being processed in a version script // (eg, "C++"). The empty string represents unmangled C names. - const std::string& + Version_script_info::Language get_current_language() const { return this->language_stack_.back(); } // Push a language onto the stack when entering an extern block. - void push_language(const std::string& lang) + void + push_language(Version_script_info::Language lang) { this->language_stack_.push_back(lang); } // Pop a language off of the stack when exiting an extern block. - void pop_language() + void + pop_language() { gold_assert(!this->language_stack_.empty()); this->language_stack_.pop_back(); } + // Return a pointer to the incremental info. + Script_info* + script_info() + { return this->script_info_; } + private: // The name of the file we are reading. const char* filename_; // The position dependent options. Position_dependent_options posdep_options_; + // True if we are parsing a --defsym. + bool parsing_defsym_; // Whether we are currently in a --start-group/--end-group. bool in_group_; // Whether the script was found in a sysrooted directory. bool is_in_sysroot_; + // If this is true, then if we find an OUTPUT_FORMAT with an + // incompatible target, then we tell the parser to abort so that we + // can search for the next file with the same name. + bool skip_on_incompatible_target_; + // True if we found an OUTPUT_FORMAT with an incompatible target. + bool found_incompatible_target_; // May be NULL if the user chooses not to pass one in. Command_line* command_line_; // Options which may be set from any linker script. @@ -1326,54 +1445,80 @@ class Parser_closure std::vector lex_mode_stack_; // A stack of which extern/language block we're inside. Can be C++, // java, or empty for C. - std::vector language_stack_; + std::vector language_stack_; // New input files found to add to the link. Input_arguments* inputs_; + // Pointer to incremental linking info. + Script_info* script_info_; }; // FILE was found as an argument on the command line. Try to read it -// as a script. We've already read BYTES of data into P, but we -// ignore that. Return true if the file was handled. +// as a script. Return true if the file was handled. bool -read_input_script(Workqueue* workqueue, const General_options& options, - Symbol_table* symtab, Layout* layout, - Dirsearch* dirsearch, Input_objects* input_objects, +read_input_script(Workqueue* workqueue, Symbol_table* symtab, Layout* layout, + Dirsearch* dirsearch, int dirindex, + Input_objects* input_objects, Mapfile* mapfile, Input_group* input_group, const Input_argument* input_argument, - Input_file* input_file, const unsigned char*, off_t, - Task_token* this_blocker, Task_token* next_blocker) + Input_file* input_file, Task_token* next_blocker, + bool* used_next_blocker) { + *used_next_blocker = false; + std::string input_string; Lex::read_file(input_file, &input_string); Lex lex(input_string.c_str(), input_string.length(), PARSING_LINKER_SCRIPT); + Script_info* script_info = NULL; + if (layout->incremental_inputs() != NULL) + { + const std::string& filename = input_file->filename(); + Timespec mtime = input_file->file().get_mtime(); + unsigned int arg_serial = input_argument->file().arg_serial(); + script_info = new Script_info(filename); + layout->incremental_inputs()->report_script(script_info, arg_serial, + mtime); + } + Parser_closure closure(input_file->filename().c_str(), input_argument->file().options(), + false, input_group != NULL, input_file->is_in_sysroot(), NULL, layout->script_options(), - &lex); + &lex, + input_file->will_search_for(), + script_info); + + bool old_saw_sections_clause = + layout->script_options()->saw_sections_clause(); if (yyparse(&closure) != 0) - return false; + { + if (closure.found_incompatible_target()) + { + Read_symbols::incompatible_warning(input_argument, input_file); + Read_symbols::requeue(workqueue, input_objects, symtab, layout, + dirsearch, dirindex, mapfile, input_argument, + input_group, next_blocker); + return true; + } + return false; + } - // THIS_BLOCKER must be clear before we may add anything to the - // symbol table. We are responsible for unblocking NEXT_BLOCKER - // when we are done. We are responsible for deleting THIS_BLOCKER - // when it is unblocked. + if (!old_saw_sections_clause + && layout->script_options()->saw_sections_clause() + && layout->have_added_input_section()) + gold_error(_("%s: SECTIONS seen after other input files; try -T/--script"), + input_file->filename().c_str()); if (!closure.saw_inputs()) - { - // The script did not add any files to read. Note that we are - // not permitted to call NEXT_BLOCKER->unblock() here even if - // THIS_BLOCKER is NULL, as we do not hold the workqueue lock. - workqueue->queue(new Script_unblock(this_blocker, next_blocker)); - return true; - } + return true; + Task_token* this_blocker = NULL; for (Input_arguments::const_iterator p = closure.inputs()->begin(); p != closure.inputs()->end(); ++p) @@ -1386,36 +1531,53 @@ read_input_script(Workqueue* workqueue, const General_options& options, nb = new Task_token(true); nb->add_blocker(); } - workqueue->queue(new Read_symbols(options, input_objects, symtab, - layout, dirsearch, &*p, - input_group, this_blocker, nb)); + workqueue->queue_soon(new Read_symbols(input_objects, symtab, + layout, dirsearch, 0, mapfile, &*p, + input_group, NULL, this_blocker, nb)); this_blocker = nb; } + *used_next_blocker = true; + return true; } -// Helper function for read_version_script() and -// read_commandline_script(). Processes the given file in the mode -// indicated by first_token and lex_mode. +// Helper function for read_version_script(), read_commandline_script() and +// script_include_directive(). Processes the given file in the mode indicated +// by first_token and lex_mode. static bool read_script_file(const char* filename, Command_line* cmdline, + Script_options* script_options, int first_token, Lex::Mode lex_mode) { - // TODO: if filename is a relative filename, search for it manually - // using "." + cmdline->options()->search_path() -- not dirsearch. Dirsearch dirsearch; + std::string name = filename; + + // If filename is a relative filename, search for it manually using "." + + // cmdline->options()->library_path() -- not dirsearch. + if (!IS_ABSOLUTE_PATH(filename)) + { + const General_options::Dir_list& search_path = + cmdline->options().library_path(); + name = Dirsearch::find_file_in_dir_list(name, search_path, "."); + } // The file locking code wants to record a Task, but we haven't // started the workqueue yet. This is only for debugging purposes, // so we invent a fake value. const Task* task = reinterpret_cast(-1); - Input_file_argument input_argument(filename, false, "", - cmdline->position_dependent_options()); + // We don't want this file to be opened in binary mode. + Position_dependent_options posdep = cmdline->position_dependent_options(); + if (posdep.format_enum() == General_options::OBJECT_FORMAT_BINARY) + posdep.set_format_enum(General_options::OBJECT_FORMAT_ELF); + Input_file_argument input_argument(name.c_str(), + Input_file_argument::INPUT_FILE_TYPE_FILE, + "", false, posdep); Input_file input_file(&input_argument); - if (!input_file.open(cmdline->options(), dirsearch, task)) + int dummy = 0; + if (!input_file.open(dirsearch, task, &dummy)) return false; std::string input_string; @@ -1426,11 +1588,14 @@ read_script_file(const char* filename, Command_line* cmdline, Parser_closure closure(filename, cmdline->position_dependent_options(), + first_token == Lex::DYNAMIC_LIST, false, input_file.is_in_sysroot(), cmdline, - cmdline->script_options(), - &lex); + script_options, + &lex, + false, + NULL); if (yyparse(&closure) != 0) { input_file.file().unlock(task); @@ -1450,21 +1615,32 @@ read_script_file(const char* filename, Command_line* cmdline, bool read_commandline_script(const char* filename, Command_line* cmdline) { - return read_script_file(filename, cmdline, + return read_script_file(filename, cmdline, &cmdline->script_options(), PARSING_LINKER_SCRIPT, Lex::LINKER_SCRIPT); } -// FILE was found as an argument to --version-script. Read it as a -// version script, and store its contents in +// FILENAME was found as an argument to --version-script. Read it as +// a version script, and store its contents in // cmdline->script_options()->version_script_info(). bool read_version_script(const char* filename, Command_line* cmdline) { - return read_script_file(filename, cmdline, + return read_script_file(filename, cmdline, &cmdline->script_options(), PARSING_VERSION_SCRIPT, Lex::VERSION_SCRIPT); } +// FILENAME was found as an argument to --dynamic-list. Read it as a +// list of symbols, and store its contents in DYNAMIC_LIST. + +bool +read_dynamic_list(const char* filename, Command_line* cmdline, + Script_options* dynamic_list) +{ + return read_script_file(filename, cmdline, dynamic_list, + PARSING_DYNAMIC_LIST, Lex::DYNAMIC_LIST); +} + // Implement the --defsym option on the command line. Return true if // all is well. @@ -1477,8 +1653,8 @@ Script_options::define_symbol(const char* definition) // Dummy value. Position_dependent_options posdep_options; - Parser_closure closure("command line", posdep_options, false, false, NULL, - this, &lex); + Parser_closure closure("command line", posdep_options, true, + false, false, NULL, this, &lex, false, NULL); if (yyparse(&closure) != 0) return false; @@ -1564,11 +1740,13 @@ script_keyword_parsecodes[] = { "BYTE", BYTE }, { "CONSTANT", CONSTANT }, { "CONSTRUCTORS", CONSTRUCTORS }, + { "COPY", COPY }, { "CREATE_OBJECT_SYMBOLS", CREATE_OBJECT_SYMBOLS }, { "DATA_SEGMENT_ALIGN", DATA_SEGMENT_ALIGN }, { "DATA_SEGMENT_END", DATA_SEGMENT_END }, { "DATA_SEGMENT_RELRO_END", DATA_SEGMENT_RELRO_END }, { "DEFINED", DEFINED }, + { "DSECT", DSECT }, { "ENTRY", ENTRY }, { "EXCLUDE_FILE", EXCLUDE_FILE }, { "EXTERN", EXTERN }, @@ -1578,6 +1756,7 @@ script_keyword_parsecodes[] = { "GROUP", GROUP }, { "HLL", HLL }, { "INCLUDE", INCLUDE }, + { "INFO", INFO }, { "INHIBIT_COMMON_ALLOCATION", INHIBIT_COMMON_ALLOCATION }, { "INPUT", INPUT }, { "KEEP", KEEP }, @@ -1591,6 +1770,7 @@ script_keyword_parsecodes[] = { "NEXT", NEXT }, { "NOCROSSREFS", NOCROSSREFS }, { "NOFLOAT", NOFLOAT }, + { "NOLOAD", NOLOAD }, { "ONLY_IF_RO", ONLY_IF_RO }, { "ONLY_IF_RW", ONLY_IF_RW }, { "OPTION", OPTION }, @@ -1647,6 +1827,19 @@ version_script_keywords(&version_script_keyword_parsecodes[0], (sizeof(version_script_keyword_parsecodes) / sizeof(version_script_keyword_parsecodes[0]))); +static const Keyword_to_parsecode::Keyword_parsecode +dynamic_list_keyword_parsecodes[] = +{ + { "extern", EXTERN }, +}; + +static const Keyword_to_parsecode +dynamic_list_keywords(&dynamic_list_keyword_parsecodes[0], + (sizeof(dynamic_list_keyword_parsecodes) + / sizeof(dynamic_list_keyword_parsecodes[0]))); + + + // Comparison function passed to bsearch. extern "C" @@ -1698,38 +1891,45 @@ Keyword_to_parsecode::keyword_to_parsecode(const char* keyword, // A single version expression. // For example, pattern="std::map*" and language="C++". -// pattern and language should be from the stringpool -struct Version_expression { - Version_expression(const std::string& pattern, - const std::string& language, - bool exact_match) - : pattern(pattern), language(language), exact_match(exact_match) {} +struct Version_expression +{ + Version_expression(const std::string& a_pattern, + Version_script_info::Language a_language, + bool a_exact_match) + : pattern(a_pattern), language(a_language), exact_match(a_exact_match), + was_matched_by_symbol(false) + { } std::string pattern; - std::string language; + Version_script_info::Language language; // If false, we use glob() to match pattern. If true, we use strcmp(). bool exact_match; + // True if --no-undefined-version is in effect and we found this + // version in get_symbol_version. We use mutable because this + // struct is generally not modifiable after it has been created. + mutable bool was_matched_by_symbol; }; - // A list of expressions. -struct Version_expression_list { +struct Version_expression_list +{ std::vector expressions; }; - // A list of which versions upon which another version depends. // Strings should be from the Stringpool. -struct Version_dependency_list { +struct Version_dependency_list +{ std::vector dependencies; }; - // The total definition of a version. It includes the tag for the // version, its global and local expressions, and any dependencies. -struct Version_tree { +struct Version_tree +{ Version_tree() - : tag(), global(NULL), local(NULL), dependencies(NULL) {} + : tag(), global(NULL), local(NULL), dependencies(NULL) + { } std::string tag; const struct Version_expression_list* global; @@ -1737,34 +1937,116 @@ struct Version_tree { const struct Version_dependency_list* dependencies; }; +// Helper class that calls cplus_demangle when needed and takes care of freeing +// the result. + +class Lazy_demangler +{ + public: + Lazy_demangler(const char* symbol, int options) + : symbol_(symbol), options_(options), demangled_(NULL), did_demangle_(false) + { } + + ~Lazy_demangler() + { free(this->demangled_); } + + // Return the demangled name. The actual demangling happens on the first call, + // and the result is later cached. + inline char* + get(); + + private: + // The symbol to demangle. + const char* symbol_; + // Option flags to pass to cplus_demagle. + const int options_; + // The cached demangled value, or NULL if demangling didn't happen yet or + // failed. + char* demangled_; + // Whether we already called cplus_demangle + bool did_demangle_; +}; + +// Return the demangled name. The actual demangling happens on the first call, +// and the result is later cached. Returns NULL if the symbol cannot be +// demangled. + +inline char* +Lazy_demangler::get() +{ + if (!this->did_demangle_) + { + this->demangled_ = cplus_demangle(this->symbol_, this->options_); + this->did_demangle_ = true; + } + return this->demangled_; +} + +// Class Version_script_info. + +Version_script_info::Version_script_info() + : dependency_lists_(), expression_lists_(), version_trees_(), globs_(), + default_version_(NULL), default_is_global_(false), is_finalized_(false) +{ + for (int i = 0; i < LANGUAGE_COUNT; ++i) + this->exact_[i] = NULL; +} + Version_script_info::~Version_script_info() { - for (size_t k = 0; k < dependency_lists_.size(); ++k) - delete dependency_lists_[k]; - for (size_t k = 0; k < version_trees_.size(); ++k) - delete version_trees_[k]; - for (size_t k = 0; k < expression_lists_.size(); ++k) - delete expression_lists_[k]; } +// Forget all the known version script information. + +void +Version_script_info::clear() +{ + for (size_t k = 0; k < this->dependency_lists_.size(); ++k) + delete this->dependency_lists_[k]; + this->dependency_lists_.clear(); + for (size_t k = 0; k < this->version_trees_.size(); ++k) + delete this->version_trees_[k]; + this->version_trees_.clear(); + for (size_t k = 0; k < this->expression_lists_.size(); ++k) + delete this->expression_lists_[k]; + this->expression_lists_.clear(); +} + +// Finalize the version script information. + +void +Version_script_info::finalize() +{ + if (!this->is_finalized_) + { + this->build_lookup_tables(); + this->is_finalized_ = true; + } +} + +// Return all the versions. + std::vector Version_script_info::get_versions() const { std::vector ret; - for (size_t j = 0; j < version_trees_.size(); ++j) - ret.push_back(version_trees_[j]->tag); + for (size_t j = 0; j < this->version_trees_.size(); ++j) + if (!this->version_trees_[j]->tag.empty()) + ret.push_back(this->version_trees_[j]->tag); return ret; } +// Return the dependencies of VERSION. + std::vector Version_script_info::get_dependencies(const char* version) const { std::vector ret; - for (size_t j = 0; j < version_trees_.size(); ++j) - if (version_trees_[j]->tag == version) + for (size_t j = 0; j < this->version_trees_.size(); ++j) + if (this->version_trees_[j]->tag == version) { const struct Version_dependency_list* deps = - version_trees_[j]->dependencies; + this->version_trees_[j]->dependencies; if (deps != NULL) for (size_t k = 0; k < deps->dependencies.size(); ++k) ret.push_back(deps->dependencies[k]); @@ -1773,54 +2055,387 @@ Version_script_info::get_dependencies(const char* version) const return ret; } -const std::string& -Version_script_info::get_symbol_version_helper(const char* symbol_name, - bool check_global) const +// A version script essentially maps a symbol name to a version tag +// and an indication of whether symbol is global or local within that +// version tag. Each symbol maps to at most one version tag. +// Unfortunately, in practice, version scripts are ambiguous, and list +// symbols multiple times. Thus, we have to document the matching +// process. + +// This is a description of what the GNU linker does as of 2010-01-11. +// It walks through the version tags in the order in which they appear +// in the version script. For each tag, it first walks through the +// global patterns for that tag, then the local patterns. When +// looking at a single pattern, it first applies any language specific +// demangling as specified for the pattern, and then matches the +// resulting symbol name to the pattern. If it finds an exact match +// for a literal pattern (a pattern enclosed in quotes or with no +// wildcard characters), then that is the match that it uses. If +// finds a match with a wildcard pattern, then it saves it and +// continues searching. Wildcard patterns that are exactly "*" are +// saved separately. + +// If no exact match with a literal pattern is ever found, then if a +// wildcard match with a global pattern was found it is used, +// otherwise if a wildcard match with a local pattern was found it is +// used. + +// This is the result: +// * If there is an exact match, then we use the first tag in the +// version script where it matches. +// + If the exact match in that tag is global, it is used. +// + Otherwise the exact match in that tag is local, and is used. +// * Otherwise, if there is any match with a global wildcard pattern: +// + If there is any match with a wildcard pattern which is not +// "*", then we use the tag in which the *last* such pattern +// appears. +// + Otherwise, we matched "*". If there is no match with a local +// wildcard pattern which is not "*", then we use the *last* +// match with a global "*". Otherwise, continue. +// * Otherwise, if there is any match with a local wildcard pattern: +// + If there is any match with a wildcard pattern which is not +// "*", then we use the tag in which the *last* such pattern +// appears. +// + Otherwise, we matched "*", and we use the tag in which the +// *last* such match occurred. + +// There is an additional wrinkle. When the GNU linker finds a symbol +// with a version defined in an object file due to a .symver +// directive, it looks up that symbol name in that version tag. If it +// finds it, it matches the symbol name against the patterns for that +// version. If there is no match with a global pattern, but there is +// a match with a local pattern, then the GNU linker marks the symbol +// as local. + +// We want gold to be generally compatible, but we also want gold to +// be fast. These are the rules that gold implements: +// * If there is an exact match for the mangled name, we use it. +// + If there is more than one exact match, we give a warning, and +// we use the first tag in the script which matches. +// + If a symbol has an exact match as both global and local for +// the same version tag, we give an error. +// * Otherwise, we look for an extern C++ or an extern Java exact +// match. If we find an exact match, we use it. +// + If there is more than one exact match, we give a warning, and +// we use the first tag in the script which matches. +// + If a symbol has an exact match as both global and local for +// the same version tag, we give an error. +// * Otherwise, we look through the wildcard patterns, ignoring "*" +// patterns. We look through the version tags in reverse order. +// For each version tag, we look through the global patterns and +// then the local patterns. We use the first match we find (i.e., +// the last matching version tag in the file). +// * Otherwise, we use the "*" pattern if there is one. We give an +// error if there are multiple "*" patterns. + +// At least for now, gold does not look up the version tag for a +// symbol version found in an object file to see if it should be +// forced local. There are other ways to force a symbol to be local, +// and I don't understand why this one is useful. + +// Build a set of fast lookup tables for a version script. + +void +Version_script_info::build_lookup_tables() { - for (size_t j = 0; j < version_trees_.size(); ++j) + size_t size = this->version_trees_.size(); + for (size_t j = 0; j < size; ++j) { - // Is it a global symbol for this version? - const Version_expression_list* explist = - check_global ? version_trees_[j]->global : version_trees_[j]->local; - if (explist != NULL) - for (size_t k = 0; k < explist->expressions.size(); ++k) - { - const char* name_to_match = symbol_name; - const struct Version_expression& exp = explist->expressions[k]; - char* demangled_name = NULL; - if (exp.language == "C++") - { - demangled_name = cplus_demangle(symbol_name, - DMGL_ANSI | DMGL_PARAMS); - // This isn't a C++ symbol. - if (demangled_name == NULL) - continue; - name_to_match = demangled_name; - } - else if (exp.language == "Java") - { - demangled_name = cplus_demangle(symbol_name, - (DMGL_ANSI | DMGL_PARAMS - | DMGL_JAVA)); - // This isn't a Java symbol. - if (demangled_name == NULL) - continue; - name_to_match = demangled_name; - } - bool matched; - if (exp.exact_match) - matched = strcmp(exp.pattern.c_str(), name_to_match) == 0; - else - matched = fnmatch(exp.pattern.c_str(), name_to_match, - FNM_NOESCAPE) == 0; - if (demangled_name != NULL) - free(demangled_name); - if (matched) - return version_trees_[j]->tag; - } + const Version_tree* v = this->version_trees_[j]; + this->build_expression_list_lookup(v->local, v, false); + this->build_expression_list_lookup(v->global, v, true); + } +} + +// If a pattern has backlashes but no unquoted wildcard characters, +// then we apply backslash unquoting and look for an exact match. +// Otherwise we treat it as a wildcard pattern. This function returns +// true for a wildcard pattern. Otherwise, it does backslash +// unquoting on *PATTERN and returns false. If this returns true, +// *PATTERN may have been partially unquoted. + +bool +Version_script_info::unquote(std::string* pattern) const +{ + bool saw_backslash = false; + size_t len = pattern->length(); + size_t j = 0; + for (size_t i = 0; i < len; ++i) + { + if (saw_backslash) + saw_backslash = false; + else + { + switch ((*pattern)[i]) + { + case '?': case '[': case '*': + return true; + case '\\': + saw_backslash = true; + continue; + default: + break; + } + } + + if (i != j) + (*pattern)[j] = (*pattern)[i]; + ++j; + } + return false; +} + +// Add an exact match for MATCH to *PE. The result of the match is +// V/IS_GLOBAL. + +void +Version_script_info::add_exact_match(const std::string& match, + const Version_tree* v, bool is_global, + const Version_expression* ve, + Exact* pe) +{ + std::pair ins = + pe->insert(std::make_pair(match, Version_tree_match(v, is_global, ve))); + if (ins.second) + { + // This is the first time we have seen this match. + return; + } + + Version_tree_match& vtm(ins.first->second); + if (vtm.real->tag != v->tag) + { + // This is an ambiguous match. We still return the + // first version that we found in the script, but we + // record the new version to issue a warning if we + // wind up looking up this symbol. + if (vtm.ambiguous == NULL) + vtm.ambiguous = v; + } + else if (is_global != vtm.is_global) + { + // We have a match for both the global and local entries for a + // version tag. That's got to be wrong. + gold_error(_("'%s' appears as both a global and a local symbol " + "for version '%s' in script"), + match.c_str(), v->tag.c_str()); + } +} + +// Build fast lookup information for EXPLIST and store it in LOOKUP. +// All matches go to V, and IS_GLOBAL is true if they are global +// matches. + +void +Version_script_info::build_expression_list_lookup( + const Version_expression_list* explist, + const Version_tree* v, + bool is_global) +{ + if (explist == NULL) + return; + size_t size = explist->expressions.size(); + for (size_t i = 0; i < size; ++i) + { + const Version_expression& exp(explist->expressions[i]); + + if (exp.pattern.length() == 1 && exp.pattern[0] == '*') + { + if (this->default_version_ != NULL + && this->default_version_->tag != v->tag) + gold_warning(_("wildcard match appears in both version '%s' " + "and '%s' in script"), + this->default_version_->tag.c_str(), v->tag.c_str()); + else if (this->default_version_ != NULL + && this->default_is_global_ != is_global) + gold_error(_("wildcard match appears as both global and local " + "in version '%s' in script"), + v->tag.c_str()); + this->default_version_ = v; + this->default_is_global_ = is_global; + continue; + } + + std::string pattern = exp.pattern; + if (!exp.exact_match) + { + if (this->unquote(&pattern)) + { + this->globs_.push_back(Glob(&exp, v, is_global)); + continue; + } + } + + if (this->exact_[exp.language] == NULL) + this->exact_[exp.language] = new Exact(); + this->add_exact_match(pattern, v, is_global, &exp, + this->exact_[exp.language]); + } +} + +// Return the name to match given a name, a language code, and two +// lazy demanglers. + +const char* +Version_script_info::get_name_to_match(const char* name, + int language, + Lazy_demangler* cpp_demangler, + Lazy_demangler* java_demangler) const +{ + switch (language) + { + case LANGUAGE_C: + return name; + case LANGUAGE_CXX: + return cpp_demangler->get(); + case LANGUAGE_JAVA: + return java_demangler->get(); + default: + gold_unreachable(); + } +} + +// Look up SYMBOL_NAME in the list of versions. Return true if the +// symbol is found, false if not. If the symbol is found, then if +// PVERSION is not NULL, set *PVERSION to the version tag, and if +// P_IS_GLOBAL is not NULL, set *P_IS_GLOBAL according to whether the +// symbol is global or not. + +bool +Version_script_info::get_symbol_version(const char* symbol_name, + std::string* pversion, + bool* p_is_global) const +{ + Lazy_demangler cpp_demangled_name(symbol_name, DMGL_ANSI | DMGL_PARAMS); + Lazy_demangler java_demangled_name(symbol_name, + DMGL_ANSI | DMGL_PARAMS | DMGL_JAVA); + + gold_assert(this->is_finalized_); + for (int i = 0; i < LANGUAGE_COUNT; ++i) + { + Exact* exact = this->exact_[i]; + if (exact == NULL) + continue; + + const char* name_to_match = this->get_name_to_match(symbol_name, i, + &cpp_demangled_name, + &java_demangled_name); + if (name_to_match == NULL) + { + // If the name can not be demangled, the GNU linker goes + // ahead and tries to match it anyhow. That does not + // make sense to me and I have not implemented it. + continue; + } + + Exact::const_iterator pe = exact->find(name_to_match); + if (pe != exact->end()) + { + const Version_tree_match& vtm(pe->second); + if (vtm.ambiguous != NULL) + gold_warning(_("using '%s' as version for '%s' which is also " + "named in version '%s' in script"), + vtm.real->tag.c_str(), name_to_match, + vtm.ambiguous->tag.c_str()); + + if (pversion != NULL) + *pversion = vtm.real->tag; + if (p_is_global != NULL) + *p_is_global = vtm.is_global; + + // If we are using --no-undefined-version, and this is a + // global symbol, we have to record that we have found this + // symbol, so that we don't warn about it. We have to do + // this now, because otherwise we have no way to get from a + // non-C language back to the demangled name that we + // matched. + if (p_is_global != NULL && vtm.is_global) + vtm.expression->was_matched_by_symbol = true; + + return true; + } + } + + // Look through the glob patterns in reverse order. + + for (Globs::const_reverse_iterator p = this->globs_.rbegin(); + p != this->globs_.rend(); + ++p) + { + int language = p->expression->language; + const char* name_to_match = this->get_name_to_match(symbol_name, + language, + &cpp_demangled_name, + &java_demangled_name); + if (name_to_match == NULL) + continue; + + if (fnmatch(p->expression->pattern.c_str(), name_to_match, + FNM_NOESCAPE) == 0) + { + if (pversion != NULL) + *pversion = p->version->tag; + if (p_is_global != NULL) + *p_is_global = p->is_global; + return true; + } + } + + // Finally, there may be a wildcard. + if (this->default_version_ != NULL) + { + if (pversion != NULL) + *pversion = this->default_version_->tag; + if (p_is_global != NULL) + *p_is_global = this->default_is_global_; + return true; + } + + return false; +} + +// Give an error if any exact symbol names (not wildcards) appear in a +// version script, but there is no such symbol. + +void +Version_script_info::check_unmatched_names(const Symbol_table* symtab) const +{ + for (size_t i = 0; i < this->version_trees_.size(); ++i) + { + const Version_tree* vt = this->version_trees_[i]; + if (vt->global == NULL) + continue; + for (size_t j = 0; j < vt->global->expressions.size(); ++j) + { + const Version_expression& expression(vt->global->expressions[j]); + + // Ignore cases where we used the version because we saw a + // symbol that we looked up. Note that + // WAS_MATCHED_BY_SYMBOL will be true even if the symbol was + // not a definition. That's OK as in that case we most + // likely gave an undefined symbol error anyhow. + if (expression.was_matched_by_symbol) + continue; + + // Just ignore names which are in languages other than C. + // We have no way to look them up in the symbol table. + if (expression.language != LANGUAGE_C) + continue; + + // Remove backslash quoting, and ignore wildcard patterns. + std::string pattern = expression.pattern; + if (!expression.exact_match) + { + if (this->unquote(&pattern)) + continue; + } + + if (symtab->lookup(pattern.c_str(), vt->tag.c_str()) == NULL) + gold_error(_("version script assignment of %s to symbol %s " + "failed: symbol not defined"), + vt->tag.c_str(), pattern.c_str()); + } } - static const std::string empty = ""; - return empty; } struct Version_dependency_list* @@ -1897,21 +2512,33 @@ Version_script_info::print_expression_list( FILE* f, const Version_expression_list* vel) const { - std::string current_language; + Version_script_info::Language current_language = LANGUAGE_C; for (size_t i = 0; i < vel->expressions.size(); ++i) { const Version_expression& ve(vel->expressions[i]); if (ve.language != current_language) { - if (!current_language.empty()) + if (current_language != LANGUAGE_C) fprintf(f, " }\n"); - fprintf(f, " extern \"%s\" {\n", ve.language.c_str()); + switch (ve.language) + { + case LANGUAGE_C: + break; + case LANGUAGE_CXX: + fprintf(f, " extern \"C++\" {\n"); + break; + case LANGUAGE_JAVA: + fprintf(f, " extern \"Java\" {\n"); + break; + default: + gold_unreachable(); + } current_language = ve.language; } fprintf(f, " "); - if (!current_language.empty()) + if (current_language != LANGUAGE_C) fprintf(f, " "); if (ve.exact_match) @@ -1923,7 +2550,7 @@ Version_script_info::print_expression_list( fprintf(f, "\n"); } - if (!current_language.empty()) + if (current_language != LANGUAGE_C) fprintf(f, " }\n"); } @@ -1968,6 +2595,9 @@ yylex(YYSTYPE* lvalp, void* closurev) case Lex::VERSION_SCRIPT: parsecode = version_script_keywords.keyword_to_parsecode(str, len); break; + case Lex::DYNAMIC_LIST: + parsecode = dynamic_list_keywords.keyword_to_parsecode(str, len); + break; default: break; } @@ -2001,6 +2631,15 @@ yyerror(void* closurev, const char* message) closure->charpos(), message); } +// Called by the bison parser to add an external symbol to the link. + +extern "C" void +script_add_extern(void* closurev, const char* name, size_t length) +{ + Parser_closure* closure = static_cast(closurev); + closure->script_options()->add_symbol_reference(name, length); +} + // Called by the bison parser to add a file to the link. extern "C" void @@ -2019,7 +2658,7 @@ script_add_file(void* closurev, const char* name, size_t length) { if (closure->is_in_sysroot()) { - const std::string& sysroot(parameters->sysroot()); + const std::string& sysroot(parameters->options().sysroot()); gold_assert(!sysroot.empty()); name_string = sysroot + name_string; } @@ -2028,7 +2667,7 @@ script_add_file(void* closurev, const char* name, size_t length) { // In addition to checking the normal library search path, we // also want to check in the script-directory. - const char *slash = strrchr(closure->filename(), '/'); + const char* slash = strrchr(closure->filename(), '/'); if (slash != NULL) { script_directory.assign(closure->filename(), @@ -2037,9 +2676,31 @@ script_add_file(void* closurev, const char* name, size_t length) } } - Input_file_argument file(name_string.c_str(), false, extra_search_path, + Input_file_argument file(name_string.c_str(), + Input_file_argument::INPUT_FILE_TYPE_FILE, + extra_search_path, false, + closure->position_dependent_options()); + Input_argument& arg = closure->inputs()->add_file(file); + arg.set_script_info(closure->script_info()); +} + +// Called by the bison parser to add a library to the link. + +extern "C" void +script_add_library(void* closurev, const char* name, size_t length) +{ + Parser_closure* closure = static_cast(closurev); + std::string name_string(name, length); + + if (name_string[0] != 'l') + gold_error(_("library name must be prefixed with -l")); + + Input_file_argument file(name_string.c_str() + 1, + Input_file_argument::INPUT_FILE_TYPE_LIBRARY, + "", false, closure->position_dependent_options()); - closure->inputs()->add_file(file); + Input_argument& arg = closure->inputs()->add_file(file); + arg.set_script_info(closure->script_info()); } // Called by the bison parser to start a group. If we are already in @@ -2077,7 +2738,7 @@ extern "C" void script_start_as_needed(void* closurev) { Parser_closure* closure = static_cast(closurev); - closure->position_dependent_options().set_as_needed(); + closure->position_dependent_options().set_as_needed(true); } // Called by the bison parser at the end of an AS_NEEDED list. @@ -2086,7 +2747,7 @@ extern "C" void script_end_as_needed(void* closurev) { Parser_closure* closure = static_cast(closurev); - closure->position_dependent_options().clear_as_needed(); + closure->position_dependent_options().set_as_needed(false); } // Called by the bison parser to set the entry symbol. @@ -2094,8 +2755,31 @@ script_end_as_needed(void* closurev) extern "C" void script_set_entry(void* closurev, const char* entry, size_t length) { + // We'll parse this exactly the same as --entry=ENTRY on the commandline + // TODO(csilvers): FIXME -- call set_entry directly. + std::string arg("--entry="); + arg.append(entry, length); + script_parse_option(closurev, arg.c_str(), arg.size()); +} + +// Called by the bison parser to set whether to define common symbols. + +extern "C" void +script_set_common_allocation(void* closurev, int set) +{ + const char* arg = set != 0 ? "--define-common" : "--no-define-common"; + script_parse_option(closurev, arg, strlen(arg)); +} + +// Called by the bison parser to refer to a symbol. + +extern "C" Expression* +script_symbol(void* closurev, const char* name, size_t length) +{ Parser_closure* closure = static_cast(closurev); - closure->script_options()->set_entry(entry, length); + if (length != 1 || name[0] != '.') + closure->script_options()->add_symbol_reference(name, length); + return script_exp_string(name, length); } // Called by the bison parser to define a symbol. @@ -2107,8 +2791,10 @@ script_set_symbol(void* closurev, const char* name, size_t length, Parser_closure* closure = static_cast(closurev); const bool provide = providei != 0; const bool hidden = hiddeni != 0; - closure->script_options()->add_symbol_assignment(name, length, value, - provide, hidden); + closure->script_options()->add_symbol_assignment(name, length, + closure->parsing_defsym(), + value, provide, hidden); + closure->clear_skip_on_incompatible_target(); } // Called by the bison parser to add an assertion. @@ -2119,6 +2805,7 @@ script_add_assertion(void* closurev, Expression* check, const char* message, { Parser_closure* closure = static_cast(closurev); closure->script_options()->add_assertion(check, message, messagelen); + closure->clear_skip_on_incompatible_target(); } // Called by the bison parser to parse an OPTION. @@ -2140,12 +2827,54 @@ script_parse_option(void* closurev, const char* option, size_t length) else { bool past_a_double_dash_option = false; - char* mutable_option = strndup(option, length); + const char* mutable_option = strndup(option, length); gold_assert(mutable_option != NULL); closure->command_line()->process_one_option(1, &mutable_option, 0, &past_a_double_dash_option); - free(mutable_option); + // The General_options class will quite possibly store a pointer + // into mutable_option, so we can't free it. In cases the class + // does not store such a pointer, this is a memory leak. Alas. :( + } + closure->clear_skip_on_incompatible_target(); +} + +// Called by the bison parser to handle OUTPUT_FORMAT. OUTPUT_FORMAT +// takes either one or three arguments. In the three argument case, +// the format depends on the endianness option, which we don't +// currently support (FIXME). If we see an OUTPUT_FORMAT for the +// wrong format, then we want to search for a new file. Returning 0 +// here will cause the parser to immediately abort. + +extern "C" int +script_check_output_format(void* closurev, + const char* default_name, size_t default_length, + const char*, size_t, const char*, size_t) +{ + Parser_closure* closure = static_cast(closurev); + std::string name(default_name, default_length); + Target* target = select_target_by_bfd_name(name.c_str()); + if (target == NULL || !parameters->is_compatible_target(target)) + { + if (closure->skip_on_incompatible_target()) + { + closure->set_found_incompatible_target(); + return 0; + } + // FIXME: Should we warn about the unknown target? } + return 1; +} + +// Called by the bison parser to handle TARGET. + +extern "C" void +script_set_target(void* closurev, const char* target, size_t len) +{ + Parser_closure* closure = static_cast(closurev); + std::string s(target, len); + General_options::Object_format format_enum; + format_enum = General_options::string_to_object_format(s.c_str()); + closure->position_dependent_options().set_format_enum(format_enum); } // Called by the bison parser to handle SEARCH_DIR. This is handled @@ -2159,7 +2888,7 @@ script_add_search_dir(void* closurev, const char* option, size_t length) gold_warning(_("%s:%d:%d: ignoring SEARCH_DIR; SEARCH_DIR is only valid" " for scripts specified via -T/--script"), closure->filename(), closure->lineno(), closure->charpos()); - else + else if (!closure->command_line()->options().nostdlib()) { std::string s = "-L" + std::string(option, length); script_parse_option(closurev, s.c_str(), s.size()); @@ -2183,6 +2912,9 @@ extern "C" void script_push_lex_into_version_mode(void* closurev) { Parser_closure* closure = static_cast(closurev); + if (closure->version_script()->is_finalized()) + gold_error(_("%s:%d:%d: invalid use of VERSION in input file"), + closure->filename(), closure->lineno(), closure->charpos()); closure->push_lex_mode(Lex::VERSION_SCRIPT); } @@ -2209,22 +2941,22 @@ extern "C" void script_register_vers_node(void*, const char* tag, int taglen, - struct Version_tree *tree, - struct Version_dependency_list *deps) + struct Version_tree* tree, + struct Version_dependency_list* deps) { gold_assert(tree != NULL); - gold_assert(tag != NULL); tree->dependencies = deps; - tree->tag = std::string(tag, taglen); + if (tag != NULL) + tree->tag = std::string(tag, taglen); } // Add a dependencies to the list of existing dependencies, if any, // and return the expanded list. -extern "C" struct Version_dependency_list * +extern "C" struct Version_dependency_list* script_add_vers_depend(void* closurev, - struct Version_dependency_list *all_deps, - const char *depend_to_add, int deplen) + struct Version_dependency_list* all_deps, + const char* depend_to_add, int deplen) { Parser_closure* closure = static_cast(closurev); if (all_deps == NULL) @@ -2234,13 +2966,11 @@ script_add_vers_depend(void* closurev, } // Add a pattern expression to an existing list of expressions, if any. -// TODO: In the old linker, the last argument used to be a bool, but I -// don't know what it meant. -extern "C" struct Version_expression_list * +extern "C" struct Version_expression_list* script_new_vers_pattern(void* closurev, - struct Version_expression_list *expressions, - const char *pattern, int patlen, int exact_match) + struct Version_expression_list* expressions, + const char* pattern, int patlen, int exact_match) { Parser_closure* closure = static_cast(closurev); if (expressions == NULL) @@ -2255,8 +2985,8 @@ script_new_vers_pattern(void* closurev, // Attaches b to the end of a, and clears b. So a = a + b and b = {}. extern "C" struct Version_expression_list* -script_merge_expressions(struct Version_expression_list *a, - struct Version_expression_list *b) +script_merge_expressions(struct Version_expression_list* a, + struct Version_expression_list* b) { a->expressions.insert(a->expressions.end(), b->expressions.begin(), b->expressions.end()); @@ -2268,10 +2998,10 @@ script_merge_expressions(struct Version_expression_list *a, // Combine the global and local expressions into a a Version_tree. -extern "C" struct Version_tree * +extern "C" struct Version_tree* script_new_vers_node(void* closurev, - struct Version_expression_list *global, - struct Version_expression_list *local) + struct Version_expression_list* global, + struct Version_expression_list* local) { Parser_closure* closure = static_cast(closurev); Version_tree* tree = closure->version_script()->allocate_version_tree(); @@ -2287,7 +3017,25 @@ extern "C" void version_script_push_lang(void* closurev, const char* lang, int langlen) { Parser_closure* closure = static_cast(closurev); - closure->push_language(std::string(lang, langlen)); + std::string language(lang, langlen); + Version_script_info::Language code; + if (language.empty() || language == "C") + code = Version_script_info::LANGUAGE_C; + else if (language == "C++") + code = Version_script_info::LANGUAGE_CXX; + else if (language == "Java") + code = Version_script_info::LANGUAGE_JAVA; + else + { + char* buf = new char[langlen + 100]; + snprintf(buf, langlen + 100, + _("unrecognized version script language '%s'"), + language.c_str()); + yyerror(closurev, buf); + delete[] buf; + code = Version_script_info::LANGUAGE_C; + } + closure->push_language(code); } extern "C" void @@ -2304,6 +3052,7 @@ script_start_sections(void* closurev) { Parser_closure* closure = static_cast(closurev); closure->script_options()->script_sections()->start_sections(); + closure->clear_skip_on_incompatible_target(); } // Called by the bison parser to finish a SECTIONS clause. @@ -2330,7 +3079,7 @@ script_start_output_section(void* closurev, const char* name, size_t namelen, // Finish processing entries for an output section. extern "C" void -script_finish_output_section(void* closurev, +script_finish_output_section(void* closurev, const struct Parser_output_section_trailer* trail) { Parser_closure* closure = static_cast(closurev); @@ -2391,6 +3140,34 @@ script_add_input_section(void* closurev, closure->script_options()->script_sections()->add_input_section(spec, keep); } +// When we see DATA_SEGMENT_ALIGN we record that following output +// sections may be relro. + +extern "C" void +script_data_segment_align(void* closurev) +{ + Parser_closure* closure = static_cast(closurev); + if (!closure->script_options()->saw_sections_clause()) + gold_error(_("%s:%d:%d: DATA_SEGMENT_ALIGN not in SECTIONS clause"), + closure->filename(), closure->lineno(), closure->charpos()); + else + closure->script_options()->script_sections()->data_segment_align(); +} + +// When we see DATA_SEGMENT_RELRO_END we know that all output sections +// since DATA_SEGMENT_ALIGN should be relro. + +extern "C" void +script_data_segment_relro_end(void* closurev) +{ + Parser_closure* closure = static_cast(closurev); + if (!closure->script_options()->saw_sections_clause()) + gold_error(_("%s:%d:%d: DATA_SEGMENT_ALIGN not in SECTIONS clause"), + closure->filename(), closure->lineno(), closure->charpos()); + else + closure->script_options()->script_sections()->data_segment_relro_end(); +} + // Create a new list of string/sort pairs. extern "C" String_sort_list_ptr @@ -2431,8 +3208,13 @@ script_new_string_list(const char* str, size_t len) extern "C" String_list_ptr script_string_list_push_back(String_list_ptr pv, const char* str, size_t len) { - pv->push_back(std::string(str, len)); - return pv; + if (pv == NULL) + return script_new_string_list(str, len); + else + { + pv->push_back(std::string(str, len)); + return pv; + } } // Concatenate two string lists. Either or both may be NULL. The way @@ -2449,3 +3231,186 @@ script_string_list_append(String_list_ptr pv1, String_list_ptr pv2) pv1->insert(pv1->end(), pv2->begin(), pv2->end()); return pv1; } + +// Add a new program header. + +extern "C" void +script_add_phdr(void* closurev, const char* name, size_t namelen, + unsigned int type, const Phdr_info* info) +{ + Parser_closure* closure = static_cast(closurev); + bool includes_filehdr = info->includes_filehdr != 0; + bool includes_phdrs = info->includes_phdrs != 0; + bool is_flags_valid = info->is_flags_valid != 0; + Script_sections* ss = closure->script_options()->script_sections(); + ss->add_phdr(name, namelen, type, includes_filehdr, includes_phdrs, + is_flags_valid, info->flags, info->load_address); + closure->clear_skip_on_incompatible_target(); +} + +// Convert a program header string to a type. + +#define PHDR_TYPE(NAME) { #NAME, sizeof(#NAME) - 1, elfcpp::NAME } + +static struct +{ + const char* name; + size_t namelen; + unsigned int val; +} phdr_type_names[] = +{ + PHDR_TYPE(PT_NULL), + PHDR_TYPE(PT_LOAD), + PHDR_TYPE(PT_DYNAMIC), + PHDR_TYPE(PT_INTERP), + PHDR_TYPE(PT_NOTE), + PHDR_TYPE(PT_SHLIB), + PHDR_TYPE(PT_PHDR), + PHDR_TYPE(PT_TLS), + PHDR_TYPE(PT_GNU_EH_FRAME), + PHDR_TYPE(PT_GNU_STACK), + PHDR_TYPE(PT_GNU_RELRO) +}; + +extern "C" unsigned int +script_phdr_string_to_type(void* closurev, const char* name, size_t namelen) +{ + for (unsigned int i = 0; + i < sizeof(phdr_type_names) / sizeof(phdr_type_names[0]); + ++i) + if (namelen == phdr_type_names[i].namelen + && strncmp(name, phdr_type_names[i].name, namelen) == 0) + return phdr_type_names[i].val; + yyerror(closurev, _("unknown PHDR type (try integer)")); + return elfcpp::PT_NULL; +} + +extern "C" void +script_saw_segment_start_expression(void* closurev) +{ + Parser_closure* closure = static_cast(closurev); + Script_sections* ss = closure->script_options()->script_sections(); + ss->set_saw_segment_start_expression(true); +} + +extern "C" void +script_set_section_region(void* closurev, const char* name, size_t namelen, + int set_vma) +{ + Parser_closure* closure = static_cast(closurev); + if (!closure->script_options()->saw_sections_clause()) + { + gold_error(_("%s:%d:%d: MEMORY region '%.*s' referred to outside of " + "SECTIONS clause"), + closure->filename(), closure->lineno(), closure->charpos(), + static_cast(namelen), name); + return; + } + + Script_sections* ss = closure->script_options()->script_sections(); + Memory_region* mr = ss->find_memory_region(name, namelen); + if (mr == NULL) + { + gold_error(_("%s:%d:%d: MEMORY region '%.*s' not declared"), + closure->filename(), closure->lineno(), closure->charpos(), + static_cast(namelen), name); + return; + } + + ss->set_memory_region(mr, set_vma); +} + +extern "C" void +script_add_memory(void* closurev, const char* name, size_t namelen, + unsigned int attrs, Expression* origin, Expression* length) +{ + Parser_closure* closure = static_cast(closurev); + Script_sections* ss = closure->script_options()->script_sections(); + ss->add_memory_region(name, namelen, attrs, origin, length); +} + +extern "C" unsigned int +script_parse_memory_attr(void* closurev, const char* attrs, size_t attrlen, + int invert) +{ + int attributes = 0; + + while (attrlen--) + switch (*attrs++) + { + case 'R': + case 'r': + attributes |= MEM_READABLE; break; + case 'W': + case 'w': + attributes |= MEM_READABLE | MEM_WRITEABLE; break; + case 'X': + case 'x': + attributes |= MEM_EXECUTABLE; break; + case 'A': + case 'a': + attributes |= MEM_ALLOCATABLE; break; + case 'I': + case 'i': + case 'L': + case 'l': + attributes |= MEM_INITIALIZED; break; + default: + yyerror(closurev, _("unknown MEMORY attribute")); + } + + if (invert) + attributes = (~ attributes) & MEM_ATTR_MASK; + + return attributes; +} + +extern "C" void +script_include_directive(void* closurev, const char* filename, size_t length) +{ + Parser_closure* closure = static_cast(closurev); + std::string name(filename, length); + Command_line* cmdline = closure->command_line(); + read_script_file(name.c_str(), cmdline, &cmdline->script_options(), + PARSING_LINKER_SCRIPT, Lex::LINKER_SCRIPT); +} + +// Functions for memory regions. + +extern "C" Expression* +script_exp_function_origin(void* closurev, const char* name, size_t namelen) +{ + Parser_closure* closure = static_cast(closurev); + Script_sections* ss = closure->script_options()->script_sections(); + Expression* origin = ss->find_memory_region_origin(name, namelen); + + if (origin == NULL) + { + gold_error(_("undefined memory region '%s' referenced " + "in ORIGIN expression"), + name); + // Create a dummy expression to prevent crashes later on. + origin = script_exp_integer(0); + } + + return origin; +} + +extern "C" Expression* +script_exp_function_length(void* closurev, const char* name, size_t namelen) +{ + Parser_closure* closure = static_cast(closurev); + Script_sections* ss = closure->script_options()->script_sections(); + Expression* length = ss->find_memory_region_length(name, namelen); + + if (length == NULL) + { + gold_error(_("undefined memory region '%s' referenced " + "in LENGTH expression"), + name); + // Create a dummy expression to prevent crashes later on. + length = script_exp_integer(0); + } + + return length; +}