#include <sys/mman.h>
#include <sys/stat.h>
#include <algorithm>
-#include "libiberty.h" // for unlink_if_ordinary()
+#include "libiberty.h"
#include "parameters.h"
#include "object.h"
# define MAP_ANONYMOUS MAP_ANON
#endif
+#ifndef HAVE_POSIX_FALLOCATE
+// A dummy, non general, version of posix_fallocate. Here we just set
+// the file size and hope that there is enough disk space. FIXME: We
+// could allocate disk space by walking block by block and writing a
+// zero byte into each block.
+static int
+posix_fallocate(int o, off_t offset, off_t len)
+{
+ return ftruncate(o, offset + len);
+}
+#endif // !defined(HAVE_POSIX_FALLOCATE)
+
namespace gold
{
unattached_section_list_(unattached_section_list),
secnamepool_(secnamepool),
shstrtab_section_(shstrtab_section)
+{
+}
+
+// Compute the current data size.
+
+off_t
+Output_section_headers::do_size() const
{
// Count all the sections. Start with 1 for the null section.
off_t count = 1;
if (!parameters->options().relocatable())
{
- for (Layout::Segment_list::const_iterator p = segment_list->begin();
- p != segment_list->end();
+ for (Layout::Segment_list::const_iterator p =
+ this->segment_list_->begin();
+ p != this->segment_list_->end();
++p)
if ((*p)->type() == elfcpp::PT_LOAD)
count += (*p)->output_section_count();
}
else
{
- for (Layout::Section_list::const_iterator p = section_list->begin();
- p != section_list->end();
+ for (Layout::Section_list::const_iterator p =
+ this->section_list_->begin();
+ p != this->section_list_->end();
++p)
if (((*p)->flags() & elfcpp::SHF_ALLOC) != 0)
++count;
}
- count += unattached_section_list->size();
+ count += this->unattached_section_list_->size();
const int size = parameters->target().get_size();
int shdr_size;
else
gold_unreachable();
- this->set_data_size(count * shdr_size);
+ return count * shdr_size;
}
// Write out the section headers.
const Layout::Segment_list& segment_list)
: segment_list_(segment_list)
{
- const int size = parameters->target().get_size();
- int phdr_size;
- if (size == 32)
- phdr_size = elfcpp::Elf_sizes<32>::phdr_size;
- else if (size == 64)
- phdr_size = elfcpp::Elf_sizes<64>::phdr_size;
- else
- gold_unreachable();
-
- this->set_data_size(segment_list.size() * phdr_size);
}
void
of->write_output_view(this->offset(), all_phdrs_size, view);
}
+off_t
+Output_segment_headers::do_size() const
+{
+ const int size = parameters->target().get_size();
+ int phdr_size;
+ if (size == 32)
+ phdr_size = elfcpp::Elf_sizes<32>::phdr_size;
+ else if (size == 64)
+ phdr_size = elfcpp::Elf_sizes<64>::phdr_size;
+ else
+ gold_unreachable();
+
+ return this->segment_list_.size() * phdr_size;
+}
+
// Output_file_header methods.
Output_file_header::Output_file_header(const Target* target,
shstrtab_(NULL),
entry_(entry)
{
- const int size = parameters->target().get_size();
- int ehdr_size;
- if (size == 32)
- ehdr_size = elfcpp::Elf_sizes<32>::ehdr_size;
- else if (size == 64)
- ehdr_size = elfcpp::Elf_sizes<64>::ehdr_size;
- else
- gold_unreachable();
-
- this->set_data_size(ehdr_size);
+ this->set_data_size(this->do_size());
}
// Set the section table information for a file header.
? elfcpp::ELFDATA2MSB
: elfcpp::ELFDATA2LSB);
e_ident[elfcpp::EI_VERSION] = elfcpp::EV_CURRENT;
- // FIXME: Some targets may need to set EI_OSABI and EI_ABIVERSION.
oehdr.put_e_ident(e_ident);
elfcpp::ET e_type;
if (parameters->options().relocatable())
e_type = elfcpp::ET_REL;
- else if (parameters->options().shared())
+ else if (parameters->options().output_is_position_independent())
e_type = elfcpp::ET_DYN;
else
e_type = elfcpp::ET_EXEC;
oehdr.put_e_phoff(this->segment_header_->offset());
oehdr.put_e_shoff(this->section_header_->offset());
-
- // FIXME: The target needs to set the flags.
- oehdr.put_e_flags(0);
-
+ oehdr.put_e_flags(this->target_->processor_specific_flags());
oehdr.put_e_ehsize(elfcpp::Elf_sizes<size>::ehdr_size);
if (this->segment_header_ == NULL)
else
oehdr.put_e_shstrndx(elfcpp::SHN_XINDEX);
+ // Let the target adjust the ELF header, e.g., to set EI_OSABI in
+ // the e_ident field.
+ parameters->target().adjust_elf_header(view, ehdr_size);
+
of->write_output_view(0, ehdr_size, view);
}
return v;
}
+// Compute the current data size.
+
+off_t
+Output_file_header::do_size() const
+{
+ const int size = parameters->target().get_size();
+ if (size == 32)
+ return elfcpp::Elf_sizes<32>::ehdr_size;
+ else if (size == 64)
+ return elfcpp::Elf_sizes<64>::ehdr_size;
+ else
+ gold_unreachable();
+}
+
// Output_data_const methods.
void
section_size_type entry_count,
elfcpp::Elf_Word flags,
std::vector<unsigned int>* input_shndxes)
- : Output_section_data(entry_count * 4, 4),
+ : Output_section_data(entry_count * 4, 4, false),
relobj_(relobj),
flags_(flags)
{
void
Output_data_dynamic::set_final_data_size()
{
- // Add the terminating entry.
- this->add_constant(elfcpp::DT_NULL, 0);
+ // Add the terminating entry if it hasn't been added.
+ // Because of relaxation, we can run this multiple times.
+ if (this->entries_.empty()
+ || this->entries_.rbegin()->tag() != elfcpp::DT_NULL)
+ this->add_constant(elfcpp::DT_NULL, 0);
int dyn_size;
if (parameters->target().get_size() == 32)
for (Xindex_entries::const_iterator p = this->entries_.begin();
p != this->entries_.end();
++p)
- elfcpp::Swap<32, big_endian>::writeval(oview + p->first * 4, p->second);
+ {
+ unsigned int symndx = p->first;
+ gold_assert(symndx * 4 < this->data_size());
+ elfcpp::Swap<32, big_endian>::writeval(oview + symndx * 4, p->second);
+ }
}
// Output_section::Input_section methods.
this->u2_.posd->print_to_mapfile(mapfile);
break;
+ case RELAXED_INPUT_SECTION_CODE:
+ {
+ Output_relaxed_input_section* relaxed_section =
+ this->relaxed_input_section();
+ mapfile->print_input_section(relaxed_section->relobj(),
+ relaxed_section->shndx());
+ }
+ break;
default:
mapfile->print_input_section(this->u2_.object, this->shndx_);
break;
attached_input_sections_are_sorted_(false),
is_relro_(false),
is_relro_local_(false),
- tls_offset_(0)
+ is_small_section_(false),
+ is_large_section_(false),
+ is_interp_(false),
+ is_dynamic_linker_section_(false),
+ generate_code_fills_at_write_(false),
+ tls_offset_(0),
+ checkpoint_(NULL),
+ merge_section_map_(),
+ merge_section_by_properties_map_(),
+ relaxed_input_section_map_(),
+ is_relaxed_input_section_map_valid_(true)
{
// An unallocated section has no address. Forcing this means that
// we don't need special treatment for symbols defined in debug
Output_section::~Output_section()
{
+ delete this->checkpoint_;
}
// Set the entry size.
off_t aligned_offset_in_section = align_address(offset_in_section,
addralign);
+ // Determine if we want to delay code-fill generation until the output
+ // section is written. When the target is relaxing, we want to delay fill
+ // generating to avoid adjusting them during relaxation.
+ if (!this->generate_code_fills_at_write_
+ && !have_sections_script
+ && (sh_flags & elfcpp::SHF_EXECINSTR) != 0
+ && parameters->target().has_code_fill()
+ && parameters->target().may_relax())
+ {
+ gold_assert(this->fills_.empty());
+ this->generate_code_fills_at_write_ = true;
+ }
+
if (aligned_offset_in_section > offset_in_section
+ && !this->generate_code_fills_at_write_
&& !have_sections_script
&& (sh_flags & elfcpp::SHF_EXECINSTR) != 0
- && object->target()->has_code_fill())
+ && parameters->target().has_code_fill())
{
// We need to add some fill data. Using fill_list_ when
// possible is an optimization, since we will often have fill
this->fills_.push_back(Fill(offset_in_section, fill_len));
else
{
- // FIXME: When relaxing, the size needs to adjust to
- // maintain a constant alignment.
- std::string fill_data(object->target()->code_fill(fill_len));
+ std::string fill_data(parameters->target().code_fill(fill_len));
Output_data_const* odc = new Output_data_const(fill_data, 1);
this->input_sections_.push_back(Input_section(odc));
}
// We need to keep track of this section if we are already keeping
// track of sections, or if we are relaxing. Also, if this is a
// section which requires sorting, or which may require sorting in
- // the future, we keep track of the sections. FIXME: Add test for
- // relaxing.
+ // the future, we keep track of the sections.
if (have_sections_script
|| !this->input_sections_.empty()
|| this->may_sort_attached_input_sections()
|| this->must_sort_attached_input_sections()
- || parameters->options().user_set_Map())
+ || parameters->options().user_set_Map()
+ || parameters->target().may_relax())
this->input_sections_.push_back(Input_section(object, shndx,
shdr.get_sh_size(),
addralign));
}
}
+// Add a relaxed input section.
+
+void
+Output_section::add_relaxed_input_section(Output_relaxed_input_section* poris)
+{
+ Input_section inp(poris);
+ this->add_output_section_data(&inp);
+ if (this->is_relaxed_input_section_map_valid_)
+ {
+ Input_section_specifier iss(poris->relobj(), poris->shndx());
+ this->relaxed_input_section_map_[iss] = poris;
+ }
+
+ // For a relaxed section, we use the current data size. Linker scripts
+ // get all the input sections, including relaxed one from an output
+ // section and add them back to them same output section to compute the
+ // output section size. If we do not account for sizes of relaxed input
+ // sections, an output section would be incorrectly sized.
+ off_t offset_in_section = this->current_data_size_for_child();
+ off_t aligned_offset_in_section = align_address(offset_in_section,
+ poris->addralign());
+ this->set_current_data_size_for_child(aligned_offset_in_section
+ + poris->current_data_size());
+}
+
// Add arbitrary data to an output section by Input_section.
void
if (is_string && addralign > entsize)
return false;
- Input_section_list::iterator p;
- for (p = this->input_sections_.begin();
- p != this->input_sections_.end();
- ++p)
- if (p->is_merge_section(is_string, entsize, addralign))
- {
- p->add_input_section(object, shndx);
- return true;
- }
+ // We cannot restore merged input section states.
+ gold_assert(this->checkpoint_ == NULL);
+
+ // Look up merge sections by required properties.
+ Merge_section_properties msp(is_string, entsize, addralign);
+ Merge_section_by_properties_map::const_iterator p =
+ this->merge_section_by_properties_map_.find(msp);
+ if (p != this->merge_section_by_properties_map_.end())
+ {
+ Output_merge_base* merge_section = p->second;
+ merge_section->add_input_section(object, shndx);
+ gold_assert(merge_section->is_string() == is_string
+ && merge_section->entsize() == entsize
+ && merge_section->addralign() == addralign);
+
+ // Link input section to found merge section.
+ Input_section_specifier iss(object, shndx);
+ this->merge_section_map_[iss] = merge_section;
+ return true;
+ }
// We handle the actual constant merging in Output_merge_data or
// Output_merge_string_data.
- Output_section_data* posd;
+ Output_merge_base* pomb;
if (!is_string)
- posd = new Output_merge_data(entsize, addralign);
+ pomb = new Output_merge_data(entsize, addralign);
else
{
switch (entsize)
{
case 1:
- posd = new Output_merge_string<char>(addralign);
+ pomb = new Output_merge_string<char>(addralign);
break;
case 2:
- posd = new Output_merge_string<uint16_t>(addralign);
+ pomb = new Output_merge_string<uint16_t>(addralign);
break;
case 4:
- posd = new Output_merge_string<uint32_t>(addralign);
+ pomb = new Output_merge_string<uint32_t>(addralign);
break;
default:
return false;
}
}
- this->add_output_merge_section(posd, is_string, entsize);
- posd->add_input_section(object, shndx);
+ // Add new merge section to this output section and link merge section
+ // properties to new merge section in map.
+ this->add_output_merge_section(pomb, is_string, entsize);
+ this->merge_section_by_properties_map_[msp] = pomb;
+
+ // Add input section to new merge section and link input section to new
+ // merge section in map.
+ pomb->add_input_section(object, shndx);
+ Input_section_specifier iss(object, shndx);
+ this->merge_section_map_[iss] = pomb;
return true;
}
+// Build a relaxation map to speed up relaxation of existing input sections.
+// Look up to the first LIMIT elements in INPUT_SECTIONS.
+
+void
+Output_section::build_relaxation_map(
+ const Input_section_list& input_sections,
+ size_t limit,
+ Relaxation_map* relaxation_map) const
+{
+ for (size_t i = 0; i < limit; ++i)
+ {
+ const Input_section& is(input_sections[i]);
+ if (is.is_input_section() || is.is_relaxed_input_section())
+ {
+ Input_section_specifier iss(is.relobj(), is.shndx());
+ (*relaxation_map)[iss] = i;
+ }
+ }
+}
+
+// Convert regular input sections in INPUT_SECTIONS into relaxed input
+// sections in RELAXED_SECTIONS. MAP is a prebuilt map from input section
+// specifier to indices of INPUT_SECTIONS.
+
+void
+Output_section::convert_input_sections_in_list_to_relaxed_sections(
+ const std::vector<Output_relaxed_input_section*>& relaxed_sections,
+ const Relaxation_map& map,
+ Input_section_list* input_sections)
+{
+ for (size_t i = 0; i < relaxed_sections.size(); ++i)
+ {
+ Output_relaxed_input_section* poris = relaxed_sections[i];
+ Input_section_specifier iss(poris->relobj(), poris->shndx());
+ Relaxation_map::const_iterator p = map.find(iss);
+ gold_assert(p != map.end());
+ gold_assert((*input_sections)[p->second].is_input_section());
+ (*input_sections)[p->second] = Input_section(poris);
+ }
+}
+
+// Convert regular input sections into relaxed input sections. RELAXED_SECTIONS
+// is a vector of pointers to Output_relaxed_input_section or its derived
+// classes. The relaxed sections must correspond to existing input sections.
+
+void
+Output_section::convert_input_sections_to_relaxed_sections(
+ const std::vector<Output_relaxed_input_section*>& relaxed_sections)
+{
+ gold_assert(parameters->target().may_relax());
+
+ // We want to make sure that restore_states does not undo the effect of
+ // this. If there is no checkpoint active, just search the current
+ // input section list and replace the sections there. If there is
+ // a checkpoint, also replace the sections there.
+
+ // By default, we look at the whole list.
+ size_t limit = this->input_sections_.size();
+
+ if (this->checkpoint_ != NULL)
+ {
+ // Replace input sections with relaxed input section in the saved
+ // copy of the input section list.
+ if (this->checkpoint_->input_sections_saved())
+ {
+ Relaxation_map map;
+ this->build_relaxation_map(
+ *(this->checkpoint_->input_sections()),
+ this->checkpoint_->input_sections()->size(),
+ &map);
+ this->convert_input_sections_in_list_to_relaxed_sections(
+ relaxed_sections,
+ map,
+ this->checkpoint_->input_sections());
+ }
+ else
+ {
+ // We have not copied the input section list yet. Instead, just
+ // look at the portion that would be saved.
+ limit = this->checkpoint_->input_sections_size();
+ }
+ }
+
+ // Convert input sections in input_section_list.
+ Relaxation_map map;
+ this->build_relaxation_map(this->input_sections_, limit, &map);
+ this->convert_input_sections_in_list_to_relaxed_sections(
+ relaxed_sections,
+ map,
+ &this->input_sections_);
+}
+
+// Update the output section flags based on input section flags.
+
+void
+Output_section::update_flags_for_input_section(elfcpp::Elf_Xword flags)
+{
+ // If we created the section with SHF_ALLOC clear, we set the
+ // address. If we are now setting the SHF_ALLOC flag, we need to
+ // undo that.
+ if ((this->flags_ & elfcpp::SHF_ALLOC) == 0
+ && (flags & elfcpp::SHF_ALLOC) != 0)
+ this->mark_address_invalid();
+
+ this->flags_ |= (flags
+ & (elfcpp::SHF_WRITE
+ | elfcpp::SHF_ALLOC
+ | elfcpp::SHF_EXECINSTR));
+}
+
+// Find the merge section into which an input section with index SHNDX in
+// OBJECT has been added. Return NULL if none found.
+
+Output_section_data*
+Output_section::find_merge_section(const Relobj* object,
+ unsigned int shndx) const
+{
+ Input_section_specifier iss(object, shndx);
+ Output_section_data_by_input_section_map::const_iterator p =
+ this->merge_section_map_.find(iss);
+ if (p != this->merge_section_map_.end())
+ {
+ Output_section_data* posd = p->second;
+ gold_assert(posd->is_merge_section_for(object, shndx));
+ return posd;
+ }
+ else
+ return NULL;
+}
+
+// Find an relaxed input section corresponding to an input section
+// in OBJECT with index SHNDX.
+
+const Output_section_data*
+Output_section::find_relaxed_input_section(const Relobj* object,
+ unsigned int shndx) const
+{
+ // Be careful that the map may not be valid due to input section export
+ // to scripts or a check-point restore.
+ if (!this->is_relaxed_input_section_map_valid_)
+ {
+ // Rebuild the map as needed.
+ this->relaxed_input_section_map_.clear();
+ for (Input_section_list::const_iterator p = this->input_sections_.begin();
+ p != this->input_sections_.end();
+ ++p)
+ if (p->is_relaxed_input_section())
+ {
+ Input_section_specifier iss(p->relobj(), p->shndx());
+ this->relaxed_input_section_map_[iss] =
+ p->relaxed_input_section();
+ }
+ this->is_relaxed_input_section_map_valid_ = true;
+ }
+
+ Input_section_specifier iss(object, shndx);
+ Output_section_data_by_input_section_map::const_iterator p =
+ this->relaxed_input_section_map_.find(iss);
+ if (p != this->relaxed_input_section_map_.end())
+ return p->second;
+ else
+ return NULL;
+}
+
// Given an address OFFSET relative to the start of input section
// SHNDX in OBJECT, return whether this address is being included in
// the final link. This should only be called if SHNDX in OBJECT has
unsigned int shndx,
off_t offset) const
{
+ // Look at the Output_section_data_maps first.
+ const Output_section_data* posd = this->find_merge_section(object, shndx);
+ if (posd == NULL)
+ posd = this->find_relaxed_input_section(object, shndx);
+
+ if (posd != NULL)
+ {
+ section_offset_type output_offset;
+ bool found = posd->output_offset(object, shndx, offset, &output_offset);
+ gold_assert(found);
+ return output_offset != -1;
+ }
+
+ // Fall back to the slow look-up.
for (Input_section_list::const_iterator p = this->input_sections_.begin();
p != this->input_sections_.end();
++p)
Output_section::output_offset(const Relobj* object, unsigned int shndx,
section_offset_type offset) const
{
- // This can only be called meaningfully when layout is complete.
- gold_assert(Output_data::is_layout_complete());
+ // This can only be called meaningfully when we know the data size
+ // of this.
+ gold_assert(this->is_data_size_valid());
+
+ // Look at the Output_section_data_maps first.
+ const Output_section_data* posd = this->find_merge_section(object, shndx);
+ if (posd == NULL)
+ posd = this->find_relaxed_input_section(object, shndx);
+ if (posd != NULL)
+ {
+ section_offset_type output_offset;
+ bool found = posd->output_offset(object, shndx, offset, &output_offset);
+ gold_assert(found);
+ return output_offset;
+ }
+ // Fall back to the slow look-up.
for (Input_section_list::const_iterator p = this->input_sections_.begin();
p != this->input_sections_.end();
++p)
off_t offset) const
{
uint64_t addr = this->address() + this->first_input_offset_;
+
+ // Look at the Output_section_data_maps first.
+ const Output_section_data* posd = this->find_merge_section(object, shndx);
+ if (posd == NULL)
+ posd = this->find_relaxed_input_section(object, shndx);
+ if (posd != NULL && posd->is_address_valid())
+ {
+ section_offset_type output_offset;
+ bool found = posd->output_offset(object, shndx, offset, &output_offset);
+ gold_assert(found);
+ return posd->address() + output_offset;
+ }
+
+ // Fall back to the slow look-up.
for (Input_section_list::const_iterator p = this->input_sections_.begin();
p != this->input_sections_.end();
++p)
unsigned int shndx,
uint64_t* paddr) const
{
+ // FIXME: This becomes a bottle-neck if we have many relaxed sections.
+ // Looking up the merge section map does not always work as we sometimes
+ // find a merge section without its address set.
uint64_t addr = this->address() + this->first_input_offset_;
for (Input_section_list::const_iterator p = this->input_sections_.begin();
p != this->input_sections_.end();
void
Output_section::do_reset_address_and_file_offset()
{
+ // An unallocated section has no address. Forcing this means that
+ // we don't need special treatment for symbols defined in debug
+ // sections. We do the same in the constructor.
+ if ((this->flags_ & elfcpp::SHF_ALLOC) == 0)
+ this->set_address(0);
+
for (Input_section_list::iterator p = this->input_sections_.begin();
p != this->input_sections_.end();
++p)
p->reset_address_and_file_offset();
}
+
+// Return true if address and file offset have the values after reset.
+
+bool
+Output_section::do_address_and_file_offset_have_reset_values() const
+{
+ if (this->is_offset_valid())
+ return false;
+
+ // An unallocated section has address 0 after its construction or a reset.
+ if ((this->flags_ & elfcpp::SHF_ALLOC) == 0)
+ return this->is_address_valid() && this->address() == 0;
+ else
+ return !this->is_address_valid();
+}
// Set the TLS offset. Called only for SHT_TLS sections.
Input_section_sort_entry(const Input_section& input_section,
unsigned int index)
: input_section_(input_section), index_(index),
- section_has_name_(input_section.is_input_section())
+ section_has_name_(input_section.is_input_section()
+ || input_section.is_relaxed_input_section())
{
if (this->section_has_name_)
{
// so it is OK to lock. Unfortunately we have no way to pass
// in a Task token.
const Task* dummy_task = reinterpret_cast<const Task*>(-1);
- Object* obj = input_section.relobj();
+ Object* obj = (input_section.is_input_section()
+ ? input_section.relobj()
+ : input_section.relaxed_input_section()->relobj());
Task_lock_obj<Object> tl(dummy_task, obj);
// This is a slow operation, which should be cached in
if (this->attached_input_sections_are_sorted_)
return;
+ if (this->checkpoint_ != NULL
+ && !this->checkpoint_->input_sections_saved())
+ this->checkpoint_->save_input_sections();
+
// The only thing we know about an input section is the object and
// the section index. We need the section name. Recomputing this
// is slow but this is an unusual case. If this becomes a speed
{
gold_assert(!this->requires_postprocessing());
+ // If the target performs relaxation, we delay filler generation until now.
+ gold_assert(!this->generate_code_fills_at_write_ || this->fills_.empty());
+
off_t output_section_file_offset = this->offset();
for (Fill_list::iterator p = this->fills_.begin();
p != this->fills_.end();
fill_data.data(), fill_data.size());
}
+ off_t off = this->offset() + this->first_input_offset_;
for (Input_section_list::iterator p = this->input_sections_.begin();
p != this->input_sections_.end();
++p)
- p->write(of);
+ {
+ off_t aligned_off = align_address(off, p->addralign());
+ if (this->generate_code_fills_at_write_ && (off != aligned_off))
+ {
+ size_t fill_len = aligned_off - off;
+ std::string fill_data(parameters->target().code_fill(fill_len));
+ of->write(off, fill_data.data(), fill_data.size());
+ }
+
+ p->write(of);
+ off = aligned_off + p->data_size();
+ }
}
// If a section requires postprocessing, create the buffer to use.
{
gold_assert(this->requires_postprocessing());
+ // If the target performs relaxation, we delay filler generation until now.
+ gold_assert(!this->generate_code_fills_at_write_ || this->fills_.empty());
+
unsigned char* buffer = this->postprocessing_buffer();
for (Fill_list::iterator p = this->fills_.begin();
p != this->fills_.end();
p != this->input_sections_.end();
++p)
{
- off = align_address(off, p->addralign());
- p->write_to_buffer(buffer + off);
- off += p->data_size();
+ off_t aligned_off = align_address(off, p->addralign());
+ if (this->generate_code_fills_at_write_ && (off != aligned_off))
+ {
+ size_t fill_len = aligned_off - off;
+ std::string fill_data(parameters->target().code_fill(fill_len));
+ memcpy(buffer + off, fill_data.data(), fill_data.size());
+ }
+
+ p->write_to_buffer(buffer + aligned_off);
+ off = aligned_off + p->data_size();
}
}
Output_section::get_input_sections(
uint64_t address,
const std::string& fill,
- std::list<std::pair<Relobj*, unsigned int> >* input_sections)
+ std::list<Simple_input_section>* input_sections)
{
+ if (this->checkpoint_ != NULL
+ && !this->checkpoint_->input_sections_saved())
+ this->checkpoint_->save_input_sections();
+
+ // Invalidate the relaxed input section map.
+ this->is_relaxed_input_section_map_valid_ = false;
+
uint64_t orig_address = address;
address = align_address(address, this->addralign());
++p)
{
if (p->is_input_section())
- input_sections->push_back(std::make_pair(p->relobj(), p->shndx()));
+ input_sections->push_back(Simple_input_section(p->relobj(),
+ p->shndx()));
+ else if (p->is_relaxed_input_section())
+ input_sections->push_back(
+ Simple_input_section(p->relaxed_input_section()));
else
{
uint64_t aligned_address = align_address(address, p->addralign());
// Add an input section from a script.
void
-Output_section::add_input_section_for_script(Relobj* object,
- unsigned int shndx,
+Output_section::add_input_section_for_script(const Simple_input_section& sis,
off_t data_size,
uint64_t addralign)
{
this->set_current_data_size_for_child(aligned_offset_in_section
+ data_size);
- this->input_sections_.push_back(Input_section(object, shndx,
- data_size, addralign));
+ Input_section is =
+ (sis.is_relaxed_input_section()
+ ? Input_section(sis.relaxed_input_section())
+ : Input_section(sis.relobj(), sis.shndx(), data_size, addralign));
+ this->input_sections_.push_back(is);
+}
+
+//
+
+void
+Output_section::save_states()
+{
+ gold_assert(this->checkpoint_ == NULL);
+ Checkpoint_output_section* checkpoint =
+ new Checkpoint_output_section(this->addralign_, this->flags_,
+ this->input_sections_,
+ this->first_input_offset_,
+ this->attached_input_sections_are_sorted_);
+ this->checkpoint_ = checkpoint;
+ gold_assert(this->fills_.empty());
+}
+
+void
+Output_section::restore_states()
+{
+ gold_assert(this->checkpoint_ != NULL);
+ Checkpoint_output_section* checkpoint = this->checkpoint_;
+
+ this->addralign_ = checkpoint->addralign();
+ this->flags_ = checkpoint->flags();
+ this->first_input_offset_ = checkpoint->first_input_offset();
+
+ if (!checkpoint->input_sections_saved())
+ {
+ // If we have not copied the input sections, just resize it.
+ size_t old_size = checkpoint->input_sections_size();
+ gold_assert(this->input_sections_.size() >= old_size);
+ this->input_sections_.resize(old_size);
+ }
+ else
+ {
+ // We need to copy the whole list. This is not efficient for
+ // extremely large output with hundreads of thousands of input
+ // objects. We may need to re-think how we should pass sections
+ // to scripts.
+ this->input_sections_ = *checkpoint->input_sections();
+ }
+
+ this->attached_input_sections_are_sorted_ =
+ checkpoint->attached_input_sections_are_sorted();
+
+ // Simply invalidate the relaxed input section map since we do not keep
+ // track of it.
+ this->is_relaxed_input_section_map_valid_ = false;
}
// Print to the map file.
type_(type),
flags_(flags),
is_max_align_known_(false),
- are_addresses_set_(false)
+ are_addresses_set_(false),
+ is_large_data_segment_(false)
{
}
void
Output_segment::add_output_section(Output_section* os,
- elfcpp::Elf_Word seg_flags)
+ elfcpp::Elf_Word seg_flags,
+ bool do_sort)
{
gold_assert((os->flags() & elfcpp::SHF_ALLOC) != 0);
gold_assert(!this->is_max_align_known_);
+ gold_assert(os->is_large_data_section() == this->is_large_data_segment());
+ gold_assert(this->type() == elfcpp::PT_LOAD || !do_sort);
// Update the segment flags.
this->flags_ |= seg_flags;
else
pdl = &this->output_data_;
- // So that PT_NOTE segments will work correctly, we need to ensure
- // that all SHT_NOTE sections are adjacent. This will normally
- // happen automatically, because all the SHT_NOTE input sections
- // will wind up in the same output section. However, it is possible
- // for multiple SHT_NOTE input sections to have different section
- // flags, and thus be in different output sections, but for the
- // different section flags to map into the same segment flags and
- // thus the same output segment.
-
// Note that while there may be many input sections in an output
// section, there are normally only a few output sections in an
- // output segment. This loop is expected to be fast.
+ // output segment. The loops below are expected to be fast.
+ // So that PT_NOTE segments will work correctly, we need to ensure
+ // that all SHT_NOTE sections are adjacent.
if (os->type() == elfcpp::SHT_NOTE && !pdl->empty())
{
Output_segment::Output_data_list::iterator p = pdl->end();
// case: we group the SHF_TLS/SHT_NOBITS sections right after the
// SHF_TLS/SHT_PROGBITS sections. This lets us set up PT_TLS
// correctly. SHF_TLS sections get added to both a PT_LOAD segment
- // and the PT_TLS segment -- we do this grouping only for the
- // PT_LOAD segment.
+ // and the PT_TLS segment; we do this grouping only for the PT_LOAD
+ // segment.
if (this->type_ != elfcpp::PT_TLS
&& (os->flags() & elfcpp::SHF_TLS) != 0)
{
pdl = &this->output_data_;
- bool nobits = os->type() == elfcpp::SHT_NOBITS;
- bool sawtls = false;
- Output_segment::Output_data_list::iterator p = pdl->end();
- do
+ if (!pdl->empty())
{
- --p;
- bool insert;
- if ((*p)->is_section_flag_set(elfcpp::SHF_TLS))
- {
- sawtls = true;
- // Put a NOBITS section after the first TLS section.
- // Put a PROGBITS section after the first TLS/PROGBITS
- // section.
- insert = nobits || !(*p)->is_section_type(elfcpp::SHT_NOBITS);
- }
- else
- {
- // If we've gone past the TLS sections, but we've seen a
- // TLS section, then we need to insert this section now.
- insert = sawtls;
- }
-
- if (insert)
+ bool nobits = os->type() == elfcpp::SHT_NOBITS;
+ bool sawtls = false;
+ Output_segment::Output_data_list::iterator p = pdl->end();
+ gold_assert(p != pdl->begin());
+ do
{
- ++p;
- pdl->insert(p, os);
- return;
+ --p;
+ bool insert;
+ if ((*p)->is_section_flag_set(elfcpp::SHF_TLS))
+ {
+ sawtls = true;
+ // Put a NOBITS section after the first TLS section.
+ // Put a PROGBITS section after the first
+ // TLS/PROGBITS section.
+ insert = nobits || !(*p)->is_section_type(elfcpp::SHT_NOBITS);
+ }
+ else
+ {
+ // If we've gone past the TLS sections, but we've
+ // seen a TLS section, then we need to insert this
+ // section now.
+ insert = sawtls;
+ }
+
+ if (insert)
+ {
+ ++p;
+ pdl->insert(p, os);
+ return;
+ }
}
+ while (p != pdl->begin());
}
- while (p != pdl->begin());
// There are no TLS sections yet; put this one at the requested
// location in the section list.
return;
}
+ // Small data sections go at the end of the list of data sections.
+ // If OS is not small, and there are small sections, we have to
+ // insert it before the first small section.
+ if (os->type() != elfcpp::SHT_NOBITS
+ && !os->is_small_section()
+ && !pdl->empty()
+ && pdl->back()->is_section()
+ && pdl->back()->output_section()->is_small_section())
+ {
+ for (Output_segment::Output_data_list::iterator p = pdl->begin();
+ p != pdl->end();
+ ++p)
+ {
+ if ((*p)->is_section()
+ && (*p)->output_section()->is_small_section())
+ {
+ pdl->insert(p, os);
+ return;
+ }
+ }
+ gold_unreachable();
+ }
+
+ // A small BSS section goes at the start of the BSS sections, after
+ // other small BSS sections.
+ if (os->type() == elfcpp::SHT_NOBITS && os->is_small_section())
+ {
+ for (Output_segment::Output_data_list::iterator p = pdl->begin();
+ p != pdl->end();
+ ++p)
+ {
+ if (!(*p)->is_section()
+ || !(*p)->output_section()->is_small_section())
+ {
+ pdl->insert(p, os);
+ return;
+ }
+ }
+ }
+
+ // A large BSS section goes at the end of the BSS sections, which
+ // means that one that is not large must come before the first large
+ // one.
+ if (os->type() == elfcpp::SHT_NOBITS
+ && !os->is_large_section()
+ && !pdl->empty()
+ && pdl->back()->is_section()
+ && pdl->back()->output_section()->is_large_section())
+ {
+ for (Output_segment::Output_data_list::iterator p = pdl->begin();
+ p != pdl->end();
+ ++p)
+ {
+ if ((*p)->is_section()
+ && (*p)->output_section()->is_large_section())
+ {
+ pdl->insert(p, os);
+ return;
+ }
+ }
+ gold_unreachable();
+ }
+
+ // We do some further output section sorting in order to make the
+ // generated program run more efficiently. We should only do this
+ // when not using a linker script, so it is controled by the DO_SORT
+ // parameter.
+ if (do_sort)
+ {
+ // FreeBSD requires the .interp section to be in the first page
+ // of the executable. That is a more efficient location anyhow
+ // for any OS, since it means that the kernel will have the data
+ // handy after it reads the program headers.
+ if (os->is_interp() && !pdl->empty())
+ {
+ pdl->insert(pdl->begin(), os);
+ return;
+ }
+
+ // Put loadable non-writable notes immediately after the .interp
+ // sections, so that the PT_NOTE segment is on the first page of
+ // the executable.
+ if (os->type() == elfcpp::SHT_NOTE
+ && (os->flags() & elfcpp::SHF_WRITE) == 0
+ && !pdl->empty())
+ {
+ Output_segment::Output_data_list::iterator p = pdl->begin();
+ if ((*p)->is_section() && (*p)->output_section()->is_interp())
+ ++p;
+ pdl->insert(p, os);
+ return;
+ }
+
+ // If this section is used by the dynamic linker, and it is not
+ // writable, then put it first, after the .interp section and
+ // any loadable notes. This makes it more likely that the
+ // dynamic linker will have to read less data from the disk.
+ if (os->is_dynamic_linker_section()
+ && !pdl->empty()
+ && (os->flags() & elfcpp::SHF_WRITE) == 0)
+ {
+ bool is_reloc = (os->type() == elfcpp::SHT_REL
+ || os->type() == elfcpp::SHT_RELA);
+ Output_segment::Output_data_list::iterator p = pdl->begin();
+ while (p != pdl->end()
+ && (*p)->is_section()
+ && ((*p)->output_section()->is_dynamic_linker_section()
+ || (*p)->output_section()->type() == elfcpp::SHT_NOTE))
+ {
+ // Put reloc sections after the other ones. Putting the
+ // dynamic reloc sections first confuses BFD, notably
+ // objcopy and strip.
+ if (!is_reloc
+ && ((*p)->output_section()->type() == elfcpp::SHT_REL
+ || (*p)->output_section()->type() == elfcpp::SHT_RELA))
+ break;
+ ++p;
+ }
+ pdl->insert(p, os);
+ return;
+ }
+ }
+
+ // If there were no constraints on the output section, just add it
+ // to the end of the list.
pdl->push_back(os);
}
{
// The script may have inserted a skip forward, but it
// better not have moved backward.
- gold_assert((*p)->address() >= addr + (off - startoff));
- off += (*p)->address() - (addr + (off - startoff));
+ if ((*p)->address() >= addr + (off - startoff))
+ off += (*p)->address() - (addr + (off - startoff));
+ else
+ {
+#if 0
+ if (!layout->script_options()->saw_sections_clause())
+ gold_unreachable();
+ else
+#endif
+ {
+ Output_section* os = (*p)->output_section();
+
+ // Cast to unsigned long long to avoid format warnings.
+ unsigned long long previous_dot =
+ static_cast<unsigned long long>(addr + (off - startoff));
+ unsigned long long dot =
+ static_cast<unsigned long long>((*p)->address());
+
+ if (os == NULL)
+ gold_error(_("dot moves backward in linker script "
+ "from 0x%llx to 0x%llx"), previous_dot, dot);
+ else
+ gold_error(_("address of section '%s' moves backward "
+ "from 0x%llx to 0x%llx"),
+ os->name(), previous_dot, dot);
+ }
+ }
(*p)->set_file_offset(off);
(*p)->finalize_data_size();
}
{
}
+// Try to open an existing file. Returns false if the file doesn't
+// exist, has a size of 0 or can't be mmapped.
+
+bool
+Output_file::open_for_modification()
+{
+ // The name "-" means "stdout".
+ if (strcmp(this->name_, "-") == 0)
+ return false;
+
+ // Don't bother opening files with a size of zero.
+ struct stat s;
+ if (::stat(this->name_, &s) != 0 || s.st_size == 0)
+ return false;
+
+ int o = open_descriptor(-1, this->name_, O_RDWR, 0);
+ if (o < 0)
+ gold_fatal(_("%s: open: %s"), this->name_, strerror(errno));
+ this->o_ = o;
+ this->file_size_ = s.st_size;
+
+ // If the file can't be mmapped, copying the content to an anonymous
+ // map will probably negate the performance benefits of incremental
+ // linking. This could be helped by using views and loading only
+ // the necessary parts, but this is not supported as of now.
+ if (!this->map_no_anonymous())
+ {
+ release_descriptor(o, true);
+ this->o_ = -1;
+ this->file_size_ = 0;
+ return false;
+ }
+
+ return true;
+}
+
// Open the output file.
void
else
{
struct stat s;
- if (::stat(this->name_, &s) == 0 && s.st_size != 0)
- unlink_if_ordinary(this->name_);
+ if (::stat(this->name_, &s) == 0
+ && (S_ISREG (s.st_mode) || S_ISLNK (s.st_mode)))
+ {
+ if (s.st_size != 0)
+ ::unlink(this->name_);
+ else if (!parameters->options().relocatable())
+ {
+ // If we don't unlink the existing file, add execute
+ // permission where read permissions already exist
+ // and where the umask permits.
+ int mask = ::umask(0);
+ ::umask(mask);
+ s.st_mode |= (s.st_mode & 0444) >> 2;
+ ::chmod(this->name_, s.st_mode & ~mask);
+ }
+ }
int mode = parameters->options().relocatable() ? 0666 : 0777;
int o = open_descriptor(-1, this->name_, O_RDWR | O_CREAT | O_TRUNC,
{
this->unmap();
this->file_size_ = file_size;
- this->map();
+ if (!this->map_no_anonymous())
+ gold_fatal(_("%s: mmap: %s"), this->name_, strerror(errno));
}
}
-// Map the file into memory.
+// Map an anonymous block of memory which will later be written to the
+// file. Return whether the map succeeded.
-void
-Output_file::map()
+bool
+Output_file::map_anonymous()
+{
+ void* base = ::mmap(NULL, this->file_size_, PROT_READ | PROT_WRITE,
+ MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+ if (base != MAP_FAILED)
+ {
+ this->map_is_anonymous_ = true;
+ this->base_ = static_cast<unsigned char*>(base);
+ return true;
+ }
+ return false;
+}
+
+// Map the file into memory. Return whether the mapping succeeded.
+
+bool
+Output_file::map_no_anonymous()
{
const int o = this->o_;
|| ::fstat(o, &statbuf) != 0
|| !S_ISREG(statbuf.st_mode)
|| this->is_temporary_)
- {
- this->map_is_anonymous_ = true;
- base = ::mmap(NULL, this->file_size_, PROT_READ | PROT_WRITE,
- MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
- }
- else
- {
- // Write out one byte to make the file the right size.
- if (::lseek(o, this->file_size_ - 1, SEEK_SET) < 0)
- gold_fatal(_("%s: lseek: %s"), this->name_, strerror(errno));
- char b = 0;
- if (::write(o, &b, 1) != 1)
- gold_fatal(_("%s: write: %s"), this->name_, strerror(errno));
+ return false;
- // Map the file into memory.
- this->map_is_anonymous_ = false;
- base = ::mmap(NULL, this->file_size_, PROT_READ | PROT_WRITE,
- MAP_SHARED, o, 0);
- }
+ // Ensure that we have disk space available for the file. If we
+ // don't do this, it is possible that we will call munmap, close,
+ // and exit with dirty buffers still in the cache with no assigned
+ // disk blocks. If the disk is out of space at that point, the
+ // output file will wind up incomplete, but we will have already
+ // exited. The alternative to fallocate would be to use fdatasync,
+ // but that would be a more significant performance hit.
+ if (::posix_fallocate(o, 0, this->file_size_) < 0)
+ gold_fatal(_("%s: %s"), this->name_, strerror(errno));
+
+ // Map the file into memory.
+ base = ::mmap(NULL, this->file_size_, PROT_READ | PROT_WRITE,
+ MAP_SHARED, o, 0);
+
+ // The mmap call might fail because of file system issues: the file
+ // system might not support mmap at all, or it might not support
+ // mmap with PROT_WRITE.
if (base == MAP_FAILED)
- gold_fatal(_("%s: mmap: %s"), this->name_, strerror(errno));
+ return false;
+
+ this->map_is_anonymous_ = false;
this->base_ = static_cast<unsigned char*>(base);
+ return true;
+}
+
+// Map the file into memory.
+
+void
+Output_file::map()
+{
+ if (this->map_no_anonymous())
+ return;
+
+ // The mmap call might fail because of file system issues: the file
+ // system might not support mmap at all, or it might not support
+ // mmap with PROT_WRITE. I'm not sure which errno values we will
+ // see in all cases, so if the mmap fails for any reason and we
+ // don't care about file contents, try for an anonymous map.
+ if (this->map_anonymous())
+ return;
+
+ gold_fatal(_("%s: mmap: failed to allocate %lu bytes for output file: %s"),
+ this->name_, static_cast<unsigned long>(this->file_size_),
+ strerror(errno));
}
// Unmap the file from memory.
if (this->map_is_anonymous_ && !this->is_temporary_)
{
size_t bytes_to_write = this->file_size_;
+ size_t offset = 0;
while (bytes_to_write > 0)
{
- ssize_t bytes_written = ::write(this->o_, this->base_, bytes_to_write);
+ ssize_t bytes_written = ::write(this->o_, this->base_ + offset,
+ bytes_to_write);
if (bytes_written == 0)
gold_error(_("%s: write: unexpected 0 return-value"), this->name_);
else if (bytes_written < 0)
gold_error(_("%s: write: %s"), this->name_, strerror(errno));
else
- bytes_to_write -= bytes_written;
+ {
+ bytes_to_write -= bytes_written;
+ offset += bytes_written;
+ }
}
}
this->unmap();