diff options
| -rw-r--r-- | elfcpp/ChangeLog | 4 | ||||
| -rw-r--r-- | elfcpp/dwarf.h | 30 | ||||
| -rw-r--r-- | gold/ChangeLog | 25 | ||||
| -rw-r--r-- | gold/Makefile.am | 2 | ||||
| -rw-r--r-- | gold/Makefile.in | 10 | ||||
| -rw-r--r-- | gold/dwarf_reader.cc | 19 | ||||
| -rw-r--r-- | gold/dwarf_reader.h | 6 | ||||
| -rw-r--r-- | gold/layout.cc | 57 | ||||
| -rw-r--r-- | gold/layout.h | 6 | ||||
| -rw-r--r-- | gold/options.cc | 6 | ||||
| -rw-r--r-- | gold/options.h | 2 | ||||
| -rw-r--r-- | gold/reduced_debug_output.cc | 429 | ||||
| -rw-r--r-- | gold/reduced_debug_output.h | 140 |
13 files changed, 722 insertions, 14 deletions
diff --git a/elfcpp/ChangeLog b/elfcpp/ChangeLog index 63e0ac0267..1050b29b42 100644 --- a/elfcpp/ChangeLog +++ b/elfcpp/ChangeLog @@ -1,3 +1,7 @@ +2008-05-28 Caleb Howe <cshowe@google.com> + + * dwarf.h (enum DW_FORM): Define. + 2008-04-19 Ian Lance Taylor <iant@google.com> * elfcpp_file.h (class Elf_file): Add large_shndx_offset_ field. diff --git a/elfcpp/dwarf.h b/elfcpp/dwarf.h index b60bdbd601..c6ff402719 100644 --- a/elfcpp/dwarf.h +++ b/elfcpp/dwarf.h @@ -4,7 +4,7 @@ // Written by Ian Lance Taylor <iant@google.com>. // This file is part of elfcpp. - + // This program is free software; you can redistribute it and/or // modify it under the terms of the GNU Library General Public License // as published by the Free Software Foundation; either version 2, or @@ -124,6 +124,32 @@ enum DW_TAG DW_TAG_PGI_interface_block = 0xA020 }; +enum DW_FORM +{ + DW_FORM_null = 0x00, + DW_FORM_addr = 0x01, + DW_FORM_block2 = 0x03, + DW_FORM_block4 = 0x04, + DW_FORM_data2 = 0x05, + DW_FORM_data4 = 0x06, + DW_FORM_data8 = 0x07, + DW_FORM_string = 0x08, + DW_FORM_block = 0x09, + DW_FORM_block1 = 0x0a, + DW_FORM_data1 = 0x0b, + DW_FORM_flag = 0x0c, + DW_FORM_sdata = 0x0d, + DW_FORM_strp = 0x0e, + DW_FORM_udata = 0x0f, + DW_FORM_ref_addr = 0x10, + DW_FORM_ref1 = 0x11, + DW_FORM_ref2 = 0x12, + DW_FORM_ref4 = 0x13, + DW_FORM_ref8 = 0x14, + DW_FORM_ref_udata = 0x15, + DW_FORM_indirect = 0x16 +}; + // Frame unwind information. enum DW_EH_PE @@ -372,7 +398,7 @@ enum DW_OP DW_OP_call_frame_cfa =0x9c, DW_OP_bit_piece =0x9d, DW_OP_lo_user =0xe0, - DW_OP_hi_user =0xff, + DW_OP_hi_user =0xff, // GNU extensions DW_OP_GNU_push_tls_address =0xe0 }; diff --git a/gold/ChangeLog b/gold/ChangeLog index c8843cef8d..a20de3a116 100644 --- a/gold/ChangeLog +++ b/gold/ChangeLog @@ -1,3 +1,28 @@ +2008-05-28 Caleb Howe <cshowe@google.com> + + * reduced_debug_output.cc: New file. + * reduced_debug_output.h: New file. + * options.h (class General_optoins): Add --strip-debug-non-line. + * options.cc (General_options::finalize): Add strip_debug_non_line + to the strip heirarchy. + * layout.h (class Layout): Add debug_abbrev_ and debug_info_ + fields. + * layout.cc: Include "reduced_debug_output.h". + (Layout::Layout): Initialize new fields. + (line_only_debug_sections): New static array. + (is_lines_only_debug_sections): New static inline function. + (Layout::include_section): Handle --strip-debug-non-line. + (Layout::make_output_section): If --strip-debug-non-line, build + new output sections for .debug_abbrev and .debug_info. + * dwarf_reader.cc (read_unsigned_LEB_128): Move to namespace + gold. Warn about possible overflow. + (read_signed_LEB_128): Likewise. + * dwarf_reader.h: (read_unsigned_LEB_128): Declare. + (read_signed_LEB_128): Declare. + * Makefile.am (CCFILES): Add reduced_debug_output.cc. + (HFILES): Add reduced_debug_output.h. + * Makefile.in: Rebuild. + 2008-05-21 Ian Lance Taylor <iant@google.com> * mapfile.cc: New file. diff --git a/gold/Makefile.am b/gold/Makefile.am index 14c3c4bacb..ac9279613b 100644 --- a/gold/Makefile.am +++ b/gold/Makefile.am @@ -52,6 +52,7 @@ CCFILES = \ output.cc \ parameters.cc \ readsyms.cc \ + reduced_debug_output.cc \ reloc.cc \ resolve.cc \ script-sections.cc \ @@ -86,6 +87,7 @@ HFILES = \ output.h \ parameters.h \ readsyms.h \ + reduced_debug_output.h \ reloc.h \ reloc-types.h \ script-c.h \ diff --git a/gold/Makefile.in b/gold/Makefile.in index 5781ebcefb..992d229919 100644 --- a/gold/Makefile.in +++ b/gold/Makefile.in @@ -82,9 +82,10 @@ am__objects_1 = archive.$(OBJEXT) binary.$(OBJEXT) common.$(OBJEXT) \ gold-threads.$(OBJEXT) layout.$(OBJEXT) mapfile.$(OBJEXT) \ merge.$(OBJEXT) object.$(OBJEXT) options.$(OBJEXT) \ output.$(OBJEXT) parameters.$(OBJEXT) readsyms.$(OBJEXT) \ - reloc.$(OBJEXT) resolve.$(OBJEXT) script-sections.$(OBJEXT) \ - script.$(OBJEXT) stringpool.$(OBJEXT) symtab.$(OBJEXT) \ - target-select.$(OBJEXT) version.$(OBJEXT) workqueue.$(OBJEXT) \ + reduced_debug_output.$(OBJEXT) reloc.$(OBJEXT) \ + resolve.$(OBJEXT) script-sections.$(OBJEXT) script.$(OBJEXT) \ + stringpool.$(OBJEXT) symtab.$(OBJEXT) target-select.$(OBJEXT) \ + version.$(OBJEXT) workqueue.$(OBJEXT) \ workqueue-threads.$(OBJEXT) am__objects_2 = am__objects_3 = yyscript.$(OBJEXT) @@ -332,6 +333,7 @@ CCFILES = \ output.cc \ parameters.cc \ readsyms.cc \ + reduced_debug_output.cc \ reloc.cc \ resolve.cc \ script-sections.cc \ @@ -366,6 +368,7 @@ HFILES = \ output.h \ parameters.h \ readsyms.h \ + reduced_debug_output.h \ reloc.h \ reloc-types.h \ script-c.h \ @@ -540,6 +543,7 @@ distclean-compile: @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/output.Po@am__quote@ @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/parameters.Po@am__quote@ @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/readsyms.Po@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/reduced_debug_output.Po@am__quote@ @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/reloc.Po@am__quote@ @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/resolve.Po@am__quote@ @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/script-sections.Po@am__quote@ diff --git a/gold/dwarf_reader.cc b/gold/dwarf_reader.cc index ec697e428e..3d0f65a4b5 100644 --- a/gold/dwarf_reader.cc +++ b/gold/dwarf_reader.cc @@ -32,7 +32,7 @@ #include "reloc.h" #include "dwarf_reader.h" -namespace { +namespace gold { // Read an unsigned LEB128 number. Each byte contains 7 bits of // information, plus one bit saying whether the number continues or @@ -48,6 +48,12 @@ read_unsigned_LEB_128(const unsigned char* buffer, size_t* len) do { + if (num_read >= 64 / 7) + { + gold_warning(_("Unusually large LEB128 decoded, " + "debug information may be corrupted")); + break; + } byte = *buffer++; num_read++; result |= (static_cast<uint64_t>(byte & 0x7f)) << shift; @@ -73,6 +79,12 @@ read_signed_LEB_128(const unsigned char* buffer, size_t* len) do { + if (num_read >= 64 / 7) + { + gold_warning(_("Unusually large LEB128 decoded, " + "debug information may be corrupted")); + break; + } byte = *buffer++; num_read++; result |= (static_cast<uint64_t>(byte & 0x7f) << shift); @@ -86,11 +98,6 @@ read_signed_LEB_128(const unsigned char* buffer, size_t* len) return result; } -} // End anonymous namespace. - - -namespace gold { - // This is the format of a DWARF2/3 line state machine that we process // opcodes using. There is no need for anything outside the lineinfo // processor to know how this works. diff --git a/gold/dwarf_reader.h b/gold/dwarf_reader.h index 173e8c2549..c9d760ce53 100644 --- a/gold/dwarf_reader.h +++ b/gold/dwarf_reader.h @@ -38,6 +38,12 @@ template<int size, bool big_endian> class Track_relocs; struct LineStateMachine; +uint64_t +read_unsigned_LEB_128(const unsigned char* buffer, size_t* len); + +int64_t +read_signed_LEB_128(const unsigned char* buffer, size_t* len); + // We can't do better than to keep the offsets in a sorted vector. // Here, offset is the key, and file_num/line_num is the value. struct Offset_to_lineno_entry diff --git a/gold/layout.cc b/gold/layout.cc index a4003d5573..2c8c5241fb 100644 --- a/gold/layout.cc +++ b/gold/layout.cc @@ -43,6 +43,7 @@ #include "dynobj.h" #include "ehframe.h" #include "compressed_output.h" +#include "reduced_debug_output.h" #include "reloc.h" #include "layout.h" @@ -110,6 +111,8 @@ Layout::Layout(const General_options& options, Script_options* script_options) added_eh_frame_data_(false), eh_frame_hdr_section_(NULL), build_id_note_(NULL), + debug_abbrev_(NULL), + debug_info_(NULL), group_signatures_(), output_file_size_(-1), input_requires_executable_stack_(false), @@ -160,6 +163,19 @@ static const char* gdb_sections[] = ".debug_str", }; +static const char* lines_only_debug_sections[] = +{ ".debug_abbrev", + // ".debug_aranges", // not used by gdb as of 6.7.1 + // ".debug_frame", + ".debug_info", + ".debug_line", + // ".debug_loc", + // ".debug_macinfo", + // ".debug_pubnames", // not used by gdb as of 6.7.1 + // ".debug_ranges", + ".debug_str", +}; + static inline bool is_gdb_debug_section(const char* str) { @@ -170,6 +186,18 @@ is_gdb_debug_section(const char* str) return false; } +static inline bool +is_lines_only_debug_section(const char* str) +{ + // We can do this faster: binary search or a hashtable. But why bother? + for (size_t i = 0; + i < sizeof(lines_only_debug_sections)/sizeof(*lines_only_debug_sections); + ++i) + if (strcmp(str, lines_only_debug_sections[i]) == 0) + return true; + return false; +} + // Whether to include this section in the link. template<int size, bool big_endian> @@ -204,6 +232,14 @@ Layout::include_section(Sized_relobj<size, big_endian>*, const char* name, if (is_debug_info_section(name)) return false; } + if (parameters->options().strip_debug_non_line() + && (shdr.get_sh_flags() & elfcpp::SHF_ALLOC) == 0) + { + // Debugging sections can only be recognized by name. + if (is_prefix_of(".debug", name) + && !is_lines_only_debug_section(name)) + return false; + } if (parameters->options().strip_debug_gdb() && (shdr.get_sh_flags() & elfcpp::SHF_ALLOC) == 0) { @@ -702,7 +738,26 @@ Layout::make_output_section(const char* name, elfcpp::Elf_Word type, && strcmp(this->options_.compress_debug_sections(), "none") != 0 && is_compressible_debug_section(name)) os = new Output_compressed_section(&this->options_, name, type, flags); - else + + else if ((flags & elfcpp::SHF_ALLOC) == 0 + && this->options_.strip_debug_non_line() + && strcmp(".debug_abbrev", name) == 0) + { + os = this->debug_abbrev_ = new Output_reduced_debug_abbrev_section( + name, type, flags); + if (this->debug_info_) + this->debug_info_->set_abbreviations(this->debug_abbrev_); + } + else if ((flags & elfcpp::SHF_ALLOC) == 0 + && this->options_.strip_debug_non_line() + && strcmp(".debug_info", name) == 0) + { + os = this->debug_info_ = new Output_reduced_debug_info_section( + name, type, flags); + if (this->debug_abbrev_) + this->debug_info_->set_abbreviations(this->debug_abbrev_); + } + else os = new Output_section(name, type, flags); this->section_list_.push_back(os); diff --git a/gold/layout.h b/gold/layout.h index 1c71d47707..ca8b7c79cf 100644 --- a/gold/layout.h +++ b/gold/layout.h @@ -49,6 +49,8 @@ class Output_segment; class Output_data; class Output_data_dynamic; class Output_symtab_xindex; +class Output_reduced_debug_abbrev_section; +class Output_reduced_debug_info_section; class Eh_frame; class Target; @@ -667,6 +669,10 @@ class Layout Output_section* eh_frame_hdr_section_; // The space for the build ID checksum if there is one. Output_section_data* build_id_note_; + // The output section containing dwarf abbreviations + Output_reduced_debug_abbrev_section* debug_abbrev_; + // The output section containing the dwarf debug info tree + Output_reduced_debug_info_section* debug_info_; // A list of group sections and their signatures. Group_signatures group_signatures_; // The size of the output file. diff --git a/gold/options.cc b/gold/options.cc index da30a6da22..03686e03fd 100644 --- a/gold/options.cc +++ b/gold/options.cc @@ -640,11 +640,13 @@ void General_options::finalize() { // Normalize the strip modifiers. They have a total order: - // strip_all > strip_debug > strip_debug_gdb. If one is true, set - // all beneath it to true as well. + // strip_all > strip_debug > strip_non_line > strip_debug_gdb. + // If one is true, set all beneath it to true as well. if (this->strip_all()) this->set_strip_debug(true); if (this->strip_debug()) + this->set_strip_debug_non_line(true); + if (this->strip_debug_non_line()) this->set_strip_debug_gdb(true); // If the user specifies both -s and -r, convert the -s to -S. diff --git a/gold/options.h b/gold/options.h index e34e165ba7..653f5fc09d 100644 --- a/gold/options.h +++ b/gold/options.h @@ -708,6 +708,8 @@ class General_options N_("Strip all symbols"), NULL); DEFINE_bool(strip_debug, options::TWO_DASHES, 'S', false, N_("Strip debugging information"), NULL); + DEFINE_bool(strip_debug_non_line, options::TWO_DASHES, '\0', false, + N_("Emit only debug line number information"), NULL); DEFINE_bool(strip_debug_gdb, options::TWO_DASHES, '\0', false, N_("Strip debug symbols that are unused by gdb " "(at least versions <= 6.7)"), NULL); diff --git a/gold/reduced_debug_output.cc b/gold/reduced_debug_output.cc new file mode 100644 index 0000000000..d195cfdd40 --- /dev/null +++ b/gold/reduced_debug_output.cc @@ -0,0 +1,429 @@ +// reduced_debug_output.cc -- output reduced debugging information to save space + +// Copyright 2008 Free Software Foundation, Inc. +// Written by Caleb Howe <cshowe@google.com>. + +// This file is part of gold. + +// This program is free software; you can redistribute it and/or modify +// it under the terms of the GNU General Public License as published by +// the Free Software Foundation; either version 3 of the License, or +// (at your option) any later version. + +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +// GNU General Public License for more details. + +// You should have received a copy of the GNU General Public License +// along with this program; if not, write to the Free Software +// Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, +// MA 02110-1301, USA. + +#include "gold.h" + +#include "parameters.h" +#include "options.h" +#include "dwarf.h" +#include "dwarf_reader.h" +#include "reduced_debug_output.h" + +#include <vector> + +namespace gold +{ + +void +write_unsigned_LEB_128(std::vector<unsigned char>* buffer, uint64_t value) +{ + do + { + unsigned char current_byte = value & 0x7f; + value >>= 7; + if (value != 0) + { + current_byte |= 0x80; + } + buffer->push_back(current_byte); + } + while (value != 0); +} + +size_t +get_length_as_unsigned_LEB_128(uint64_t value) +{ + size_t length = 0; + do + { + unsigned char current_byte = value & 0x7f; + value >>= 7; + if (value != 0) + { + current_byte |= 0x80; + } + length++; + } + while (value != 0); + return length; +} + +template <int valsize> +void Insert_into_vector(std::vector<unsigned char>* destination, + typename elfcpp::Valtype_base<valsize>::Valtype value) +{ + union + { + unsigned char buffer[valsize / 8]; + long long align; + } u; + if (parameters->target().is_big_endian()) + elfcpp::Swap<valsize, true>::writeval(u.buffer, value); + else + elfcpp::Swap<valsize, false>::writeval(u.buffer, value); + destination->insert(destination->end(), u.buffer, u.buffer + valsize / 8); +} + +template <int valsize> +typename elfcpp::Valtype_base<valsize>::Valtype +read_from_pointer(unsigned char** source) +{ + typename elfcpp::Valtype_base<valsize>::Valtype return_value; + if (parameters->target().is_big_endian()) + return_value = elfcpp::Swap_unaligned<valsize, true>::readval(*source); + else + return_value = elfcpp::Swap_unaligned<valsize, false>::readval(*source); + *source += valsize / 8; + return return_value; +} + +// Given a pointer to the beginning of a die and the beginning of the associated +// abbreviation fills in die_end with the end of the information entry. If +// successful returns true. Get_die_end also takes a pointer to the end of the +// buffer containing the die. If die_end would be beyond the end of the +// buffer, or if an unsupported dwarf form is encountered returns false. +bool +Output_reduced_debug_info_section::get_die_end( + unsigned char* die, unsigned char* abbrev, unsigned char** die_end, + unsigned char* buffer_end, int address_size, bool is64) +{ + size_t LEB_size; + uint64_t LEB_decoded; + for(;;) + { + uint64_t attribute = read_unsigned_LEB_128(abbrev, &LEB_size); + abbrev += LEB_size; + elfcpp::DW_FORM form = + static_cast<elfcpp::DW_FORM>(read_unsigned_LEB_128(abbrev, + &LEB_size)); + abbrev += LEB_size; + if (!(attribute || form)) + break; + if (die >= buffer_end) + return false; + switch(form) + { + case elfcpp::DW_FORM_null: + break; + case elfcpp::DW_FORM_strp: + die += is64 ? 8 : 4; + break; + case elfcpp::DW_FORM_addr: + case elfcpp::DW_FORM_ref_addr: + die += address_size; + break; + case elfcpp::DW_FORM_block1: + die += *die; + die += 1; + break; + case elfcpp::DW_FORM_block2: + { + uint16_t block_size; + block_size = read_from_pointer<16>(&die); + die += block_size; + break; + } + case elfcpp::DW_FORM_block4: + { + uint32_t block_size; + block_size = read_from_pointer<32>(&die); + die += block_size; + break; + } + case elfcpp::DW_FORM_block: + LEB_decoded = read_unsigned_LEB_128(die, &LEB_size); + die += (LEB_decoded + LEB_size); + break; + case elfcpp::DW_FORM_data1: + case elfcpp::DW_FORM_ref1: + case elfcpp::DW_FORM_flag: + die += 1; + break; + case elfcpp::DW_FORM_data2: + case elfcpp::DW_FORM_ref2: + die += 2; + break; + case elfcpp::DW_FORM_data4: + case elfcpp::DW_FORM_ref4: + die += 4; + break; + case elfcpp::DW_FORM_data8: + case elfcpp::DW_FORM_ref8: + die += 8; + break; + case elfcpp::DW_FORM_ref_udata: + case elfcpp::DW_FORM_udata: + read_unsigned_LEB_128(die, &LEB_size); + die += LEB_size; + break; + case elfcpp::DW_FORM_string: + { + size_t length = strlen(reinterpret_cast<char*>(die)); + die += length + 1; + break; + } + case elfcpp::DW_FORM_sdata: + case elfcpp::DW_FORM_indirect: + return false; + } + } + *die_end = die; + return true; +} + +void +Output_reduced_debug_abbrev_section::set_final_data_size() +{ + if (this->sized_ || this->failed_) + return; + + uint64_t abbrev_number; + size_t LEB_size; + unsigned char* abbrev_data = this->postprocessing_buffer(); + unsigned char* abbrev_end = this->postprocessing_buffer() + + this->postprocessing_buffer_size(); + this->write_to_postprocessing_buffer(); + while(abbrev_data < abbrev_end) + { + uint64_t abbrev_offset = abbrev_data - this->postprocessing_buffer(); + while((abbrev_number = read_unsigned_LEB_128(abbrev_data, &LEB_size))) + { + if (abbrev_data >= abbrev_end) + { + failed("Debug abbreviations extend beyond .debug_abbrev " + "section; failed to reduce debug abbreviations"); + return; + } + abbrev_data += LEB_size; + + // Together with the abbreviation number these fields make up + // the header for each abbreviation + uint64_t abbrev_type = read_unsigned_LEB_128(abbrev_data, &LEB_size); + abbrev_data += LEB_size; + + // This would ordinarily be the has_children field of the + // abbreviation. But it's going to be false after reducting the + // information, so there's no point in storing it + abbrev_data++; + + // Read to the end of the current abbreviation + // This is indicated by two zero unsigned LEBs in a row. We don't + // need to parse the data yet, so we just scan through the data + // looking for two consecutive 0 bytes indicating the end of the + // abbreviation + unsigned char* current_abbrev; + for (current_abbrev = abbrev_data; + current_abbrev[0] || current_abbrev[1]; + current_abbrev++) + { + if (current_abbrev >= abbrev_end) + { + this->failed(_("Debug abbreviations extend beyond " + ".debug_abbrev section; failed to reduce " + "debug abbreviations")); + return; + } + } + // Account for the two nulls and advance to the start of the + // next abbreviation. + current_abbrev += 2; + + // We're eliminating every entry except for compile units, so we + // only need to store abbreviations that describe them + if (abbrev_type == elfcpp::DW_TAG_compile_unit) + { + write_unsigned_LEB_128(&this->data_, ++this->abbrev_count_); + write_unsigned_LEB_128(&this->data_, abbrev_type); + // has_children is false for all entries + this->data_.push_back(0); + this->abbrev_mapping_[std::make_pair(abbrev_offset, + abbrev_number)] = + std::make_pair(abbrev_count_, this->data_.size()); + this->data_.insert(this->data_.end(), abbrev_data, + current_abbrev); + } + abbrev_data = current_abbrev; + } + gold_assert(LEB_size == 1); + abbrev_data += LEB_size; + } + // Null terminate the list of abbreviations + this->data_.push_back(0); + this->set_data_size(data_.size()); + this->sized_ = true; +} + +void +Output_reduced_debug_abbrev_section::do_write(Output_file* of) +{ + off_t offset = this->offset(); + off_t data_size = this->data_size(); + unsigned char* view = of->get_output_view(offset, data_size); + if (this->failed_) + memcpy(view, this->postprocessing_buffer(), + this->postprocessing_buffer_size()); + else + memcpy(view, &this->data_.front(), data_size); + of->write_output_view(offset, data_size, view); +} + +// Locates the abbreviation with abbreviation_number abbrev_number in the +// abbreviation table at offset abbrev_offset. abbrev_number is updated with +// its new abbreviation number and a pointer to the beginning of the +// abbreviation is returned. +unsigned char* +Output_reduced_debug_abbrev_section::get_new_abbrev( + uint64_t* abbrev_number, uint64_t abbrev_offset) +{ + set_final_data_size(); + std::pair<uint64_t, uint64_t> abbrev_info = + this->abbrev_mapping_[std::make_pair(abbrev_offset, *abbrev_number)]; + *abbrev_number = abbrev_info.first; + return &this->data_[abbrev_info.second]; +} + +void Output_reduced_debug_info_section::set_final_data_size() +{ + if (this->failed_) + return; + unsigned char* debug_info = this->postprocessing_buffer(); + unsigned char* debug_info_end = (this->postprocessing_buffer() + + this->postprocessing_buffer_size()); + unsigned char* next_compile_unit; + this->write_to_postprocessing_buffer(); + + while (debug_info < debug_info_end) + { + uint32_t compile_unit_start = read_from_pointer<32>(&debug_info); + // The first 4 bytes of each compile unit determine whether or + // not we're using dwarf32 or dwarf64. This is not necessarily + // related to whether the binary is 32 or 64 bits. + if (compile_unit_start == 0xFFFFFFFF) + { + // Technically the size can be up to 96 bits. Rather than handle + // 96/128 bit integers we just truncate the size at 64 bits. + if (0 != read_from_pointer<32>(&debug_info)) + { + this->failed(_("Extremely large compile unit in debug info; " + "failed to reduce debug info")); + return; + } + const int dwarf64_header_size = sizeof(uint64_t) + sizeof(uint16_t) + + sizeof(uint64_t) + sizeof(uint8_t); + if (debug_info + dwarf64_header_size >= debug_info_end) + { + this->failed(_("Debug info extends beyond .debug_info section;" + "failed to reduce debug info")); + return; + } + + uint64_t compile_unit_size = read_from_pointer<64>(&debug_info); + next_compile_unit = debug_info + compile_unit_size; + uint16_t version = read_from_pointer<16>(&debug_info); + uint64_t abbrev_offset = read_from_pointer<64>(&debug_info); + uint8_t address_size = read_from_pointer<8>(&debug_info); + size_t LEB_size; + uint64_t abbreviation_number = read_unsigned_LEB_128(debug_info, + &LEB_size); + debug_info += LEB_size; + unsigned char* die_abbrev = this->associated_abbrev_->get_new_abbrev( + &abbreviation_number, abbrev_offset); + unsigned char* die_end; + if (!this->get_die_end(debug_info, die_abbrev, &die_end, + debug_info_end, address_size, true)) + { + this->failed(_("Invalid DIE in debug info; " + "failed to reduce debug info")); + return; + } + + Insert_into_vector<32>(&this->data_, 0xFFFFFFFF); + Insert_into_vector<32>(&this->data_, 0); + Insert_into_vector<64>( + &this->data_, + (11 + get_length_as_unsigned_LEB_128(abbreviation_number) + + die_end - debug_info)); + Insert_into_vector<16>(&this->data_, version); + Insert_into_vector<64>(&this->data_, 0); + Insert_into_vector<8>(&this->data_, address_size); + write_unsigned_LEB_128(&this->data_, abbreviation_number); + this->data_.insert(this->data_.end(), debug_info, die_end); + } + else + { + const int dwarf32_header_size = + sizeof(uint16_t) + sizeof(uint32_t) + sizeof(uint8_t); + if (debug_info + dwarf32_header_size >= debug_info_end) + { + this->failed(_("Debug info extends beyond .debug_info section; " + "failed to reduce debug info")); + return; + } + uint32_t compile_unit_size = compile_unit_start; + next_compile_unit = debug_info + compile_unit_size; + uint16_t version = read_from_pointer<16>(&debug_info); + uint32_t abbrev_offset = read_from_pointer<32>(&debug_info); + uint8_t address_size = read_from_pointer<8>(&debug_info); + size_t LEB_size; + uint64_t abbreviation_number = read_unsigned_LEB_128(debug_info, + &LEB_size); + debug_info += LEB_size; + unsigned char* die_abbrev = this->associated_abbrev_->get_new_abbrev( + &abbreviation_number, abbrev_offset); + unsigned char* die_end; + if (!this->get_die_end(debug_info, die_abbrev, &die_end, + debug_info_end, address_size, false)) + { + this->failed(_("Invalid DIE in debug info; " + "failed to reduce debug info")); + return; + } + + Insert_into_vector<32>( + &this->data_, + (7 + get_length_as_unsigned_LEB_128(abbreviation_number) + + die_end - debug_info)); + Insert_into_vector<16>(&this->data_, version); + Insert_into_vector<32>(&this->data_, 0); + Insert_into_vector<8>(&this->data_, address_size); + write_unsigned_LEB_128(&this->data_, abbreviation_number); + this->data_.insert(this->data_.end(), debug_info, die_end); + } + debug_info = next_compile_unit; + } + this->set_data_size(data_.size()); +} + +void Output_reduced_debug_info_section::do_write(Output_file* of) +{ + off_t offset = this->offset(); + off_t data_size = this->data_size(); + unsigned char* view = of->get_output_view(offset, data_size); + if (this->failed_) + memcpy(view, this->postprocessing_buffer(), + this->postprocessing_buffer_size()); + else + memcpy(view, &this->data_.front(), data_size); + of->write_output_view(offset, data_size, view); +} + +} // End namespace gold. diff --git a/gold/reduced_debug_output.h b/gold/reduced_debug_output.h new file mode 100644 index 0000000000..bd8da2237c --- /dev/null +++ b/gold/reduced_debug_output.h @@ -0,0 +1,140 @@ +// reduced_debug_output.h -- reduce debugging information -*- C++ -*- + +// Copyright 2008 Free Software Foundation, Inc. +// Written by Caleb Howe <cshowe@google.com>. + +// This file is part of gold. + +// This program is free software; you can redistribute it and/or modify +// it under the terms of the GNU General Public License as published by +// the Free Software Foundation; either version 3 of the License, or +// (at your option) any later version. + +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +// GNU General Public License for more details. + +// You should have received a copy of the GNU General Public License +// along with this program; if not, write to the Free Software +// Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, +// MA 02110-1301, USA. + +// Reduce the size of the debug sections by emitting only debug line number +// information. We still need to emit skeleton debug_info and debug_abbrev +// sections for standard tools to parse the debug information correctly. These +// classes remove all debug information entries from the .debug_info section +// except for those describing compilation units as these DIEs contain +// references to the debug line information needed by most parsers. + +#ifndef GOLD_REDUCED_DEBUG_OUTPUT_H +#define GOLD_REDUCED_DEBUG_OUTPUT_H + +#include <map> +#include <utility> +#include <vector> + +#include "output.h" + +namespace gold +{ + +class Output_reduced_debug_abbrev_section : public Output_section +{ + public: + Output_reduced_debug_abbrev_section(const char* name, elfcpp::Elf_Word flags, + elfcpp::Elf_Xword type) + : Output_section(name, flags, type), sized_(false), + abbrev_count_(0), failed_(false) + { this->set_requires_postprocessing(); } + + unsigned char* get_new_abbrev(uint64_t* abbrev_number, + uint64_t abbrev_offset); + + protected: + // Set the final data size. + void + set_final_data_size(); + + // Write out the new debug abbreviations + void + do_write(Output_file*); + + private: + void + failed(std::string reason) + { + gold_warning(reason.c_str()); + failed_ = true; + } + + // The reduced debug abbreviations + std::vector<unsigned char> data_; + + // We map the abbreviation table offset and abbreviation number of the + // old abbreviation to the number and size of the new abbreviation. + std::map<std::pair<uint64_t, uint64_t>, + std::pair<uint64_t, uint64_t> > abbrev_mapping_; + + bool sized_; + + // The count of abbreviations in the output data + int abbrev_count_; + + // Whether or not the debug reduction has failed for any reason + bool failed_; +}; + +class Output_reduced_debug_info_section : public Output_section +{ + public: + Output_reduced_debug_info_section(const char* name, elfcpp::Elf_Word flags, + elfcpp::Elf_Xword type) + : Output_section(name, flags, type), failed_(false) + { this->set_requires_postprocessing(); } + + void + set_abbreviations(Output_reduced_debug_abbrev_section* abbrevs) + { associated_abbrev_ = abbrevs; } + + protected: + // Set the final data size. + void + set_final_data_size(); + + // Write out the new debug info + void + do_write(Output_file*); + + private: + void + failed(std::string reason) + { + gold_warning(reason.c_str()); + this->failed_ = true; + } + + // Given a pointer to the beginning of a die and the beginning of the + // associated abbreviation fills in die_end with the end of the information + // entry. If successful returns true. Get_die_end also takes a pointer to + // the end of the buffer containing the die. If die_end would be beyond the + // end of the buffer, or if an unsupported dwarf form is encountered returns + // false. + bool + get_die_end(unsigned char* die, unsigned char* abbrev, + unsigned char** die_end, unsigned char* buffer_end, + int address_size, bool is64); + + // The reduced debug info + std::vector<unsigned char> data_; + + // Each debug info section needs to be associated with a debug abbrev section + Output_reduced_debug_abbrev_section* associated_abbrev_; + + // Whether or not the debug reduction has failed for any reason + bool failed_; +}; + +} // End namespace gold. + +#endif // !defined(GOLD_REDUCED_DEBUG_OUTPUT_H) |