When generating notoc call and branch stubs without the benefit of
pc-relative insns, the stubs need to use LR to access the run time PC.
All LR changes must be described in .eh_frame if we're to support
unwinding through asynchronous exceptions. That's what this patch
does.
The patch has gone through way too many iterations. At first I
attempted to add multiple FDEs, one for each stub. That ran into
difficulties with do_plt_fde_location which is only capable of setting
the address of a single FDE per Output_data section, and with removing
any FDEs added on a previous do_relax pass. Removing FDEs (git commit
be897fb774) went overboard in matching the FDE contents. That means
either stashing the contents created for add_eh_frame_for_plt to use
when calling remove_eh_frame_for_plt, or recreating contents on the
fly (*) just to remove FDEs. In fact, FDE content matching is quite
unnecesary. FDEs added by a previous do_relax pass are those with
u_.from_linker.post_map set. So they can easily be recognised just by
looking at that flag. This patch keeps that part of the multiple FDE
changes.
In the end I went for just one FDE per stub group to describe the call
stubs. That's reasonably efficient for the common case of only
needing to describe the __tls_get_addr_opt call stub. We don't expect
to be making many calls using notoc stubs without pc-relative insns.
*) Which has it's own set of problems. The contents must be recreated
using the old stub layout, but .eh_frame size can affect stub
requirements so you need to temporarily keep the old .eh_frame size
when creating new stubs, then reset .eh_frame size before adding new
FDEs.
* ehframe.cc (Fde::operator==): Delete.
(Cie::remove_fde): Delete.
(Eh_frame::remove_ehframe_for_plt): Delete fde_data and fde_length
parameters. Remove all post-map plt FDEs.
* ehframe.h (Fde:post_map): Make const, add variant to compare plt.
(Fde::operator==): Delete.
(Cie::remove_fde): Implement here.
(Cie::last_fde): New accessor.
(Eh_frame::remove_ehframe_for_plt): Update prototype.
* layout.cc (Layout::remove_eh_frame_for_plt): Delete fde_data and
fde_length parameters.
* layout.h (Layout::remove_eh_frame_for_plt): Update prototype.
* powerpc.cc (Stub_table::tls_get_addr_opt_bctrl_): Delete.
(Stub_table::plt_fde_len_, plt_fde_, init_plt_fde): Delete.
(Stub_table::add_plt_call_entry): Don't set tls_get_addr_opt_bctrl_.
(eh_advance): New function.
(stub_sort): New function.
(Stub_table::add_eh_frame): Emit eh_frame for notoc plt calls and
branches as well as __tls_get_addr_opt plt call stub.
(Stub_table::remove_eh_frame): Update to suit.
There is a very small but non-zero probability that a stub group
contains stubs on one relax pass, but does not on the next. In that
case we would get an FDE covering a zero length address range.
(Actually, it's even worse. Alignment padding for stubs can mean the
address for the non-existent stubs is past the end of the original
section to which stubs are attached, and due to the way
do_plt_fde_location calculates the length we can get a negative
length.) Fixing this properly requires removing the FDE.
Also, I have been implementing the __tls_get_addr_opt support for
gold, and that stub needs something other than the default FDE. The
necessary FDE will depend on the offset to the __tls_get_addr_opt
stub, which of course can change during relaxation. That means at the
very least, rewriting the FDE on each pass, possibly changing the FDE
size. I think that is better done by completely recreating PLT
eh_frame FDEs.
* ehframe.cc (Fde::operator==): New.
(Cie::remove_fde, Eh_frame::remove_ehframe_for_plt): New.
* ehframe.h (Fde::operator==): Declare.
(Cie::remove_fde, Eh_frame::remove_ehframe_for_plt): Likewise.
* layout.cc (Layout::remove_eh_frame_for_plt): New.
* layout.h (Layout::remove_eh_frame_for_plt): Declare.
* powerpc.cc (Target_powerpc::do_relax): Remove old eh_frame FDEs.
(Stub_table::add_eh_frame): Delete eh_frame_added_ condition.
Don't add eh_frame for empty stub section.
(Stub_table::remove_eh_frame): New.
Because the __EH_FRAME_BEGIN__ symbol is provided in an empty .eh_frame
section in crtbeginT.o, if crt1.o has a non-empty .eh_frame section,
we place all optimized .eh_frame sections into the output section ahead
of the __EH_FRAME_BEGIN__ symbol, which breaks EH for statically-linked
binaries.
This patch fixes the problem by delaying the attachment of the optimized
.eh_frame sections to the output section until we see the end marker
section (or to the end of pass 1 if we never see an end marker).
gold/
PR gold/14675
* ehframe.cc (Eh_frame::add_ehframe_input_section): Change return type;
return enum indicating whether .eh_frame section is empty, optimizable,
unrecognized, or an end marker. Adjust explicit instantiations.
* ehframe.h (Eh_frame::Eh_frame_section_disposition): New enum type.
(Eh_frame::add_ehframe_input_section): Change return type.
* gold.cc (queue_middle_tasks): Call Layout::finalize_eh_frame_section.
* layout.cc (Layout::layout_eh_frame): Don't add optimized sections
to the .eh_frame output section until we see the end marker.
(Layout::finalize_eh_frame_section): New.
* layout.h: (Layout::finalize_eh_frame_section): New.
2015-03-02 Rafael Ávila de Espíndola <rafael.espindola@gmail.com>
* ehframe.cc (Cie::set_output_offset): Pass in and use a
Output_section_data instead of a Merge_map.
(Eh_frame::Eh_frame): Don't initialize merge_map_.
(Eh_frame::read_cie): Use add_merge_mapping instead of
Merge_map::add_mapping.
(Eh_frame::read_fde): Ditto.
(Eh_frame::set_final_data_size): Use this instead of this->merge_map_.
(Eh_frame::do_output_offset): Use merge_output_offset istead of
merge_map_->get_output_offset.
(Eh_frame::do_is_merge_section_for): Delete.
* ehframe.h (Fde::add_mapping): Pass in and use a Output_section_data
instead of a Merge_map.
(Cie::set_output_offset): Pass in a Output_section_data instead of a
Merge_map.
(Eh_frame::do_is_merge_section_for): Delete.
(Eh_frame::merge_map_): Delete.
* merge.cc (Object_merge_map::get_or_make_input_merge_map): Pass in
and use a Output_section_data instead of a Merge_map.
(Object_merge_map::add_mapping): Ditto.
(Object_merge_map::get_output_offset): Remove the merge_map argument.
(Object_merge_map::is_merge_section_for): Pass in and use a
Output_section_data instead of a Merge_map.
(Merge_map): Delete.
(Output_merge_base::do_output_offset): Use merge_output_offset instead
of merge_map_.get_output_offset.
(Output_merge_base::do_is_merge_section_for): Delete.
(Output_merge_data::do_add_input_section): Use
object->add_merge_mapping instead of add_mapping.
(Output_merge_string<Char_type>::finalize_merged_data): Ditto.
* merge.h (Merge_map): Delete forward declaration.
(Object_merge_map::add_mapping): Pass in and use a Output_section_data
instead of a Merge_map.
(Object_merge_map::get_output_offset): Remove the merge_map argument.
(Object_merge_map::is_merge_section_for): Pass in and use a
Output_section_data instead of a Merge_map.
(Input_merge_map::Object_merge_map::merge_map): Replace with
output_data.
(Object_merge_map::get_or_make_input_merge_map): Pass in and use a
Output_section_data instead of a Merge_map.
(Merge_map): Delete.
(Output_merge_base::Output_merge_base): Don't initialize merge_map_.
(Output_merge_base::do_is_merge_section_for): Delete.
(Output_merge_base::add_mapping): Delete.
(Output_merge_base::merge_map_): Delete.
* object.cc (Relobj::initialize_input_to_output_map): New.
(Relobj::initialize_input_to_output_map): New.
(Relobj::merge_output_offset): New.
(Relobj::is_merge_section_for): New.
(Relobj::initialize_input_to_output_map): Instantiate for 32 and 64
bits.
* object.h (Relobj::merge_map): Delete.
(initialize_input_to_output_map): New.
(set_merge_map): Delete.
(add_merge_mapping): New.
(merge_output_offset): New.
(is_merge_section_for): New.
* output.cc (Output_section::Input_section::is_merge_section_for):
Use object->is_merge_section_for.
* output.h (Output_section_data::is_merge_section_for): Delete.
(Output_section_data::do_is_merge_section_for): Delete.
* reloc.cc (Merged_symbol_value<size>::initialize_input_to_output_map):
Use object->initialize_input_to_output_map.
(Merged_symbol_value<size>::value_from_output_section): Use
object->merge_output_offset.
When --sort-section=name is used, gold will sort the linker-generated contents
of .eh_frame (after optimization) after the endcap provided by crtendS.o.
This causes two problems: the .eh_frame_hdr section is generated assuming that
the optimized .eh_frame contents will be placed at the very beginning of the
section, and the endcap no longer appears at the end of the section.
This patch fixes the first problem by adjusting FDE offsets to take into account
the actual starting offset within the output section, and fixes the second
problem by sorting linker-generated (Output_section_data) sections based on the
name of the output section.
gold/
PR gold/17005
* ehframe.cc (Fde::write): Add output_offset parameter.
(Cie::write): Likewise.
(Eh_frame::set_final_data_size): Account for offset within output
section.
(Eh_frame::do_sized_write): Likewise.
* ehframe.h (Fde::write): Add output_offset parameter.
(Cie::write): Likewise.
* output.cc (Output_section::Input_section_sort_entry): Remove
section_has_name_; add output_section_name parameter. Use
output section name for non-input sections.
(Output_section::Input_section_sort_entry::section_has_name): Remove.
(Output_section::Input_section_sort_entry::section_has_name_): Remove.
(Output_section::Input_section_sort_compare): Remove logic for
sections without names.
(Output_section::Input_section_sort_init_fini_compare): Likewise.
(Output_section::Input_section_sort_section_prefix_special_ordering_compare):
Likewise.
(Output_section::Input_section_sort_section_name_compare): Likewise.
gold/
PR gold/14860
* ehframe.cc (Eh_frame_hdr::Eh_frame_hdr): Initialize lock_.
(Ehframe_hdr::set_final_data_size): Allocate a Lock.
* ehframe.h (Eh_frame_hdr::record_fde): Hold the lock while
updating fde_offsets_.
(Eh_frame_hdr::lock_): New data member.
* target.cc (Target::do_plt_fde_location): New function.
* ehframe.h (class FDE): Add post_map field to u_.from_linker,
accessor function, and constructor param.
(struct Post_fde, Post_fdes): Declare.
(Cie::write): Add post_fdes param.
* ehframe.cc (Fde::write): Use plt_fde_location.
(struct Post_fde): Define.
(Cie::write): Stash FDEs added post merge mapping.
(Eh_frame::add_ehframe_for_plt): Assert no new CIEs after mapping.
Adjust Fde constructor call. Bump final_data_size_ for post map FDEs.
(Eh_frame::do_sized_write): Arrange to write post map FDES after
other FDEs.
* powerpc.cc (Target_powerpc::do_plt_fde_location): New function.
(Target_powerpc::has_glink): New function.
(Target_powerpc::do_relax): Add eh_frame info for stubs.
(struct Eh_cie, eh_frame_cie, glink_eh_frame_fde_64,
glink_eh_frame_fde_32, default_fde): New data.
(Stub_table::eh_frame_added_): New var.
(Stub_table::find_long_branch_entry, stub_address, stub_offset):
Make const.
(Stub_table::add_eh_frame): New function.
(Output_data_glink::add_eh_frame): New function.
(Target_powerpc::make_glink_section): Call add_eh_frame.
* options.h (class General_options): Add
--ld-generated-unwind-info.
* ehframe.cc (Fde::write): Add address parameter. Change all
callers. If associated with PLT, fill in address and size.
(Cie::set_output_offset): Only add merge mapping if there is an
object.
(Cie::write): Add address parameter. Change all callers.
(Eh_frame::add_ehframe_for_plt): New function.
* ehframe.h (class Fde): Update declarations. Move shndx_ and
input_offset_ fields into union u_, with new plt field.
(Fde::Fde): Adjust for new union field.
(Fde::Fde) [Output_data version]: New constructor.
(Fde::add_mapping): Only add merge mapping if there is an object.
(class Cie): Update declarations.
(class Eh_frame): Declare add_ehframe_for_plt.
* layout.cc (Layout::layout_eh_frame): Break out code into
make_eh_frame_section, and call it.
(Layout::make_eh_frame_section): New function.
(Layout::add_eh_frame_for_plt): New function.
* layout.h (class Layout): Update declarations.
* merge.cc (Merge_map::add_mapping): Add assertion.
* i386.cc: Include "dwarf.h".
(class Output_data_plt_i386): Make first_plt_entry,
dyn_first_plt_entry, exec_plt_entry, and dyn_plt_entry const. Add
plt_eh_frame_cie_size, plt_eh_frame_fde_size, plt_eh_frame_cie,
and plt_eh_frame_fde.
(Output_data_plt_i386::Output_data_plt_i386): Align to 16-byte
boundary. Call add_eh_frame_for_plt if appropriate.
* x86_64.cc: Include "dwarf.h".
(class Output_data_plt_x86_64): Align to 16-byte boundary. Make
first_plt_entry, plt_entry and tlsdesc_plt_entry const. Add
plt_eh_frame_cie_size, plt_eh_frame_fde_size, plt_eh_frame_cie,
and plt_eh_frame_fde.
(Output_data_plt_x86_64::init): Call add_eh_frame_for_plt if
appropriate.
info; adjust display of GOT entries.
* incremental.cc (Sized_incremental_binary::setup_readers): Allocate
vector of input objects; remove file_status_.
(Sized_incremental_binary::do_reserve_layout): Remove file_status_.
(Sized_incremental_binary::do_process_got_plt): Adjust calls to
got_plt reader; call target hooks to reserve GOT entries.
(Output_section_incremental_inputs::set_final_data_size): Adjust size
of input file info header and GOT info entry.
(Output_section_incremental_inputs::write_info_blocks): Write dynamic
relocation info.
(Got_plt_view_info::got_descriptor): Remove.
(Got_plt_view_info::sym_index): New data member.
(Got_plt_view_info::input_index): New data member.
(Local_got_offset_visitor::visit): Write input file index.
(Global_got_offset_visitor::visit): Write 0 for input file index.
(Global_symbol_visitor_got_plt::operator()): Replace got_descriptor
with sym_index and input_index.
(Output_section_incremental_inputs::write_got_plt): Adjust size of
incremental info GOT entry; replace got_descriptor with input_index.
(Sized_relobj_incr::Sized_relobj_incr): Adjust initializers; record
map from input file index to object.
(Sized_relobj_incr::do_layout): Replace direct data member reference
with accessor function.
(Sized_relobj_incr::do_for_all_local_got_entries): Move to base class.
* incremental.h (Incremental_input_entry_reader::get_symbol_offset):
Adjust size of input file info header.
(Incremental_input_entry_reader::get_first_dyn_reloc): New function.
(Incremental_input_entry_reader::get_dyn_reloc_count): New function.
(Incremental_input_entry_reader::get_input_section): Adjust size of
input file info header.
(Incremental_got_plt_reader::Incremental_got_plt_reader): Adjust size
of incremental info GOT entry.
(Incremental_got_plt_reader::get_got_desc): Remove.
(Incremental_got_plt_reader::get_got_symndx): New function.
(Incremental_got_plt_reader::get_got_input_index): New function.
(Sized_incremental_binary::Sized_incremental_binary): Remove
file_status_; add input_objects_.
(Sized_incremental_binary::~Sized_incremental_binary): Remove.
(Sized_incremental_binary::set_file_is_unchanged): Remove.
(Sized_incremental_binary::file_is_unchanged): Remove.
(Sized_incremental_binary::set_input_object): New function.
(Sized_incremental_binary::input_object): New function.
(Sized_incremental_binary::file_status_): Remove.
(Sized_incremental_binary::input_objects_): New data member.
(Sized_relobj_incr): Rename Sized_incr_relobj to this; adjust all
references.
(Sized_relobj_incr::invalid_address): Move to base class.
(Sized_relobj_incr::is_output_section_offset_invalid): Move to base
class.
(Sized_relobj_incr::do_output_section_offset): Likewise.
(Sized_relobj_incr::do_for_all_local_got_entries): Likewise.
(Sized_relobj_incr::section_offsets_): Likewise.
* object.cc (Sized_relobj::do_for_all_local_got_entries): New
function.
(Sized_relobj_file::Sized_relobj_file): Remove local_got_offsets_.
(Sized_relobj_file::layout_section): Replace refs to section_offsets_
with accessor function.
(Sized_relobj_file::do_layout): Likewise.
(Sized_relobj_file::do_layout_deferred_sections): Likewise.
(Sized_relobj_file::do_for_all_local_got_entries): Move to base class.
(Sized_relobj_file::compute_final_local_value): Replace refs to
section_offsets_ with accessor function.
(Sized_relobj_file::do_finalize_local_symbols): Likewise.
* object.h (Relobj::Relobj): Initialize new data members.
(Relobj::add_dyn_reloc): New function.
(Relobj::first_dyn_reloc): New function.
(Relobj::dyn_reloc_count): New function.
(Relobj::first_dyn_reloc_): New data member.
(Relobj::dyn_reloc_count_): New data member.
(Sized_relobj): Rename Sized_relobj_base to this; adjust all
references.
(Sized_relobj::Address): New typedef.
(Sized_relobj::invalid_address): Move here from child class.
(Sized_relobj::Sized_relobj): Initialize new data members.
(Sized_relobj::sized_relobj): New function.
(Sized_relobj::is_output_section_offset_invalid): Move here from
child class.
(Sized_relobj::get_output_section_offset): Likewise.
(Sized_relobj::local_has_got_offset): Likewise.
(Sized_relobj::local_got_offset): Likewise.
(Sized_relobj::set_local_got_offset): Likewise.
(Sized_relobj::do_for_all_local_got_entries): Likewise.
(Sized_relobj::clear_got_offsets): New function.
(Sized_relobj::section_offsets): Move here from child class.
(Sized_relobj::do_output_section_offset): Likewise.
(Sized_relobj::do_set_section_offset): Likewise.
(Sized_relobj::Local_got_offsets): Likewise.
(Sized_relobj::local_got_offsets_): Likewise.
(Sized_relobj::section_offsets_): Likewise.
(Sized_relobj_file): Rename Sized_relobj to this; adjust all
references.
(Sized_relobj_file::is_output_section_offset_invalid): Move to base
class.
(Sized_relobj_file::sized_relobj): New function
(Sized_relobj_file::local_has_got_offset): Move to base class.
(Sized_relobj_file::local_got_offset): Likewise.
(Sized_relobj_file::set_local_got_offset): Likewise.
(Sized_relobj_file::get_output_section_offset): Likewise.
(Sized_relobj_file::do_for_all_local_got_entries): Likewise.
(Sized_relobj_file::do_output_section_offset): Likewise.
(Sized_relobj_file::do_set_section_offset): Likewise.
(Sized_relobj_file::Local_got_offsets): Likewise.
(Sized_relobj_file::local_got_offsets_): Likewise.
(Sized_relobj_file::section_offsets_): Likewise.
* output.cc (Output_reloc::Output_reloc): Adjust type of relobj
(all constructors).
(set_needs_dynsym_index): Convert relobj to derived class pointer.
(Output_reloc::get_symbol_index): Likewise.
(Output_reloc::local_section_offset): Likewise.
(Output_reloc::get_address): Likewise.
(Output_reloc::symbol_value): Likewise.
(Output_data_got::reserve_slot): Move to class definition.
(Output_data_got::reserve_local): New function.
(Output_data_got::reserve_slot_for_global): Remove.
(Output_data_got::reserve_global): New function.
* output.h (Output_reloc::Output_reloc): Adjust type of relobj
(all constructors, two instantiations).
(Output_reloc::get_relobj): New function (two instantiations).
(Output_reloc::u1_.relobj, Output_reloc::u2_.relobj): Adjust type.
(Output_data_reloc_base::add): Convert relobj to derived class pointer.
(Output_data_reloc::add_global): Adjust type of relobj.
(Output_data_reloc::add_global_relative): Likewise.
(Output_data_reloc::add_symbolless_global_addend): Likewise.
(Output_data_reloc::add_local): Likewise.
(Output_data_reloc::add_local_relative): Likewise.
(Output_data_reloc::add_symbolless_local_addend): Likewise.
(Output_data_reloc::add_local_section): Likewise.
(Output_data_reloc::add_output_section): Likewise.
(Output_data_reloc::add_absolute): Likewise.
(Output_data_reloc::add_target_specific): Likewise.
(Output_data_got::reserve_slot): Move definition here.
(Output_data_got::reserve_local): New function.
(Output_data_got::reserve_global): New function.
* reloc.cc (Sized_relobj_file::do_read_relocs): Replace refs to
section_offsets_ with accessor function.
(Sized_relobj_file::write_sections): Likewise.
(Sized_relobj_file::do_relocate_sections): Likewise.
* target.h (Sized_target::reserve_local_got_entry): New function.
(Sized_target::reserve_global_got_entry): New function.
* x86_64.cc (Target_x86_64::reserve_local_got_entry): New function.
(Target_x86_64::reserve_global_got_entry): New function.
(Target_x86_64::init_got_plt_for_update): Create rela_dyn section.
