Use i386_target_description to get tdesc_i386

GDB can call function i386_target_description to get the right target
description rather than tdesc_i386

gdb:

2017-09-04  Yao Qi  <yao.qi@linaro.org>

* amd64-fbsd-nat.c (amd64fbsd_read_description): Call
i386_target_description.
* i386-fbsd-nat.c (i386fbsd_read_description): Call
i386_target_description.
* i386-tdep.c (i386_gdbarch_init): Likewise.

Use amd64_target_description to get tdesc_amd64

This patch changes amd64-*-tdep.c files to use function
amd64_target_description to get the right target description rather than
use the variable tdesd_amd64.

gdb:

2017-09-04  Yao Qi  <yao.qi@linaro.org>

* amd64-darwin-tdep.c: Include "x86-xstate.h".
(x86_darwin_init_abi_64): Call amd64_target_description.
* amd64-dicos-tdep.c: Likewise.
* amd64-fbsd-nat.c: Likewise.
* amd64-fbsd-tdep.c: Likewise.
* amd64-nbsd-tdep.c: Likewise.
* amd64-obsd-tdep.c: Likewise.
* amd64-sol2-tdep.c: Likewise.
* amd64-windows-tdep.c: Likewise.
* amd64-tdep.h (tdesc_amd64): Remove the declaration.

btrace: Store btrace_insn in an std::vector

Because it contains a non-POD type field (flags), the type btrace_insn
should be new'ed/delete'd.  Replace the VEC (btrace_insn_s) in
btrace_function with an std::vector.

gdb/ChangeLog:

* btrace.h (btrace_insn_s, DEF_VEC_O (btrace_insn_s)): Remove.
(btrace_function) <insn>: Change type to use std::vector.
* btrace.c (ftrace_debug, ftrace_call_num_insn,
ftrace_find_call, ftrace_new_gap, ftrace_update_function,
ftrace_update_insns, ftrace_compute_global_level_offset,
btrace_stitch_bts, btrace_clear, btrace_insn_get,
btrace_insn_end, btrace_insn_next, btrace_insn_prev): Adjust to
change to std::vector.
(ftrace_update_insns): Adjust to change to std::vector, change
type of INSN parameter.
(btrace_compute_ftrace_bts): Adjust call to ftrace_update_insns.
* record-btrace.c (btrace_call_history_insn_range,
btrace_compute_src_line_range,
record_btrace_frame_prev_register): Adjust to change to
std::vector.
* python/py-record-btrace.c (recpy_bt_func_instructions): Adjust
to change to std::vector.

Automatic date update in version.in

Use std::string in reopen_exec_file

This changes reopen_exec_file to use a std::string, removing a
cleanup.

ChangeLog
2017-09-03  Tom Tromey  <tom@tromey.com>

* corefile.c (reopen_exec_file): Use std::string.

Use std::string and unique_xmalloc_ptr in compile/ code

Change various things in the compile/ code to use std::string or
unique_xmalloc_ptr as appropriate.  This allows the removal of some
cleanups.

ChangeLog
2017-09-03  Tom Tromey  <tom@tromey.com>

* compile/compile.c (compile_register_name_mangled): Return
std::string.
* compile/compile-loc2c.c (pushf_register_address): Update.
(pushf_register): Update.
* compile/compile-c-types.c (convert_array): Update.
* compile/compile-c-symbols.c (generate_vla_size): Update.
(error_symbol_once): Use a gdb::unique_xmalloc_ptr.
(symbol_substitution_name): Return a gdb::unique_xmalloc_ptr.
(convert_one_symbol): Update.
(generate_c_for_for_one_variable): Update.
* compile/compile-c-support.c (c_get_range_decl_name): Return a
std::string.
(generate_register_struct): Update.
* compile/compile-internal.h (c_get_range_decl_name): Return a
std::string.
(compile_register_name_mangled): Return std::string.

Return std::string from perror_string

Change perror_string to return a std::string, removing a cleanup in
the process.

ChangeLog
2017-09-03  Tom Tromey  <tom@tromey.com>

* utils.c (perror_string): Return a std::string.
(throw_perror_with_name, perror_warning_with_name): Update.

Use std::string and unique_xmalloc_ptr in demangle_command

Change demangle_command to use std::string and unique_xmalloc_ptr,
removing some cleanups.

ChangeLog
2017-09-03  Tom Tromey  <tom@tromey.com>

* demangle.c (demangle_command): Use std::string,
unique_xmalloc_ptr.

Use std::string in do_set_command

Change do_set_command to use std::string, removing a cleanup and some
manual resizing code.

ChangeLog
2017-09-03  Tom Tromey  <tom@tromey.com>

* cli/cli-setshow.c (do_set_command): Use std::string.

Use unique_xmalloc_ptr in cd_command

Change cd_command to use unique_xmalloc_ptr, removing a cleanup.

ChangeLog
2017-09-03  Tom Tromey  <tom@tromey.com>

* cli/cli-cmds.c (cd_command): Use gdb::unique_xmalloc_ptr.

Use std::string in mi_cmd_interpreter_exec

Change mi_cmd_interpreter_exec to use std::string, removing a cleanup.

ChangeLog
2017-09-03  Tom Tromey  <tom@tromey.com>

* mi/mi-interp.c (mi_cmd_interpreter_exec): Use std::string.

Use unique_xmalloc_ptr in env_execute_cli_command

Change env_execute_cli_command to use unique_xmalloc_ptr, removing a
cleanup.

ChangeLog
2017-09-03  Tom Tromey  <tom@tromey.com>

* mi/mi-cmd-env.c (env_execute_cli_command): Use
gdb::unique_xmalloc_ptr.

Use std::string thread.c

This changes a few spots in thread.c to use std::string, removing some
cleanups.

ChangeLog
2017-09-03  Tom Tromey  <tom@tromey.com>

* thread.c (print_thread_info_1): Use string_printf.
(thread_apply_command, thread_apply_all_command): Use
std::string.

Return std::string from memory_error_message

This changes memory_error_message to return a std::string and fixes up
the callers.  This removes some cleanups.

ChangeLog
2017-09-03  Tom Tromey  <tom@tromey.com>

* valprint.c (val_print_string): Update.
* gdbcore.h (memory_error_message): Return std::string.
* corefile.c (memory_error_message): Return std::string.
(memory_error): Update.
* breakpoint.c (insert_bp_location): Update.

x86-64: Set tlsdesc_plt if GOT_TLS_GDESC_P is true

We need to set tlsdesc_plt for x86-64 if GOT_TLS_GDESC_P is true when
allocating dynamic relocations so that _bfd_x86_elf_size_dynamic_sections
will generate TLSDESC_PLT and TLSDESC_GOT in x86-64 output.

bfd/

PR ld/22071
* elfxx-x86.c (elf_x86_allocate_dynrelocs): Set tlsdesc_plt
for x86-64 if GOT_TLS_GDESC_P is true.

ld/

PR ld/22071
* testsuite/ld-x86-64/pr22071.d: New file.
* testsuite/ld-x86-64/pr22071.s: Likewise.
* testsuite/ld-x86-64/x86-64.exp: Run pr22071.

PR22067, x86 check_relocs invalid read

PR 22067
* elfxx-x86.h (elf_x86_hash_table): Check is_elf_hash_table first.

Make target_waitstatus_to_string return an std::string

A quite straightforward change.  It does "fix" leaks in record-btrace.c,
although since this is only used in debug printing code, it has no real
world impact.

gdb/ChangeLog:

* target/waitstatus.h (target_waitstatus_to_string): Change
return type to std::string.
* target/waitstatus.c (target_waitstatus_to_string): Return
std::string.
* target.h (target_waitstatus_to_string): Remove declaration.
* infrun.c (resume, clear_proceed_status_thread,
print_target_wait_results, do_target_wait, save_waitstatus,
stop_all_threads): Adjust.
* record-btrace.c (record_btrace_wait): Adjust.
* target-debug.h
(target_debug_print_struct_target_waitstatus_p): Adjust.

gdb/gdbserver/ChangeLog:

* linux-low.c (linux_wait_1): Adjust.
* server.c (queue_stop_reply_callback): Adjust.

Initialize tls_get_addr for x86-64 in one place

* elfxx-x86.c (_bfd_x86_elf_link_hash_table_create): Initialize
tls_get_addr for x86-64 in one place.

Add missing ChangeLog entries

Add elf64-x86-64.lo together with elfxx-x86.lo for 64-bit BFD

Since elfxx-x86.lo needs elf64-x86-64.lo with 64-bit BFD now, add
elf64-x86-64.lo together with elfxx-x86.lo to bfd_backends for 64-bit
BFD.

* configure.ac (bfd_backends): Add elf64-x86-64.lo together
with elfxx-x86.lo for 64-bit BFD.
* configure: Regenerated.

Automatic date update in version.in

x86: Add _bfd_x86_elf_size_dynamic_sections

elf_i386_size_dynamic_sections and elf_x86_64_size_dynamic_sections are
very similar, except for the followings:

1. elf_i386_size_dynamic_sections checks GOT_TLS_IE and GOT_TLS_IE_BOTH.
elf_x86_64_size_dynamic_sections checks only GOT_TLS_IE.  Since
GOT_TLS_IE_BOTH is never true for x86-64, it is OK to check GOT_TLS_IE
for both i386 and x86-64.
2, x86-64 sets tlsdesc_plt, but i386 doesn't.  We set tlsdesc_plt only
if target_id == X86_64_ELF_DATA.
3. x86-64 has

  if (s != htab->elf.srelplt)
    s->reloc_count = 0;

and i386 has

  s->reloc_count = 0;

i386 did have

  if (s != htab->srelplt)
    s->reloc_count = 0;

in the original commit:

commit 67a4f2b710581acc83afecff55424af285ecbc28
Author: Alexandre Oliva <aoliva@redhat.com>
Date:   Wed Jan 18 21:07:51 2006 +0000

But it was removed by

commit 5ae0bfb60a576344d7f701605346282c1144499e
Author: Richard Sandiford <rdsandiford@googlemail.com>
Date:   Tue Feb 28 07:16:12 2006 +0000

    bfd/
            * elf32-i386.c (elf_i386_link_hash_table): Add next_tls_desc_index.
            (elf_i386_link_hash_table_create): Initialize it.
            (elf_i386_compute_jump_table_size): Use it instead of
            srelplt->reloc_count.
            (allocate_dynrelocs): Likewise.
            (elf_i386_size_dynamic_sections): Likewise.
            (elf_i386_relocate_section): Likewise.

A later commit:

commit e1f987424b7b3f5ac63a2a6ae044a202a44b8ff8
Author: H.J. Lu <hjl.tools@gmail.com>
Date:   Fri Oct 21 15:13:37 2011 +0000

    Put IRELATIVE relocations after JUMP_SLOT.

    bfd/

    2011-10-21  H.J. Lu  <hongjiu.lu@intel.com>

            PR ld/13302
            * elf32-i386.c (elf_i386_link_hash_table): Add next_jump_slot_index
            and next_irelative_index.
            (elf_i386_link_hash_table_create): Initialize next_jump_slot_index
            and next_irelative_index.
            (elf_i386_allocate_dynrelocs): Increment reloc_count instead of
            next_tls_desc_index.
            (elf_i386_size_dynamic_sections): Set next_tls_desc_index and
            next_irelative_index from reloc_count.
            (elf_i386_finish_dynamic_symbol): Put R_386_IRELATIVE after
            R_386_JUMP_SLOT.

changed it back to use reloc_count again. So it is correct to use

  if (s != htab->elf.srelplt)
    s->reloc_count = 0;

for both i386 and x86-64 now.
4. i386 and x86-64 use different DT_XXXs.  They are handled by adding
them to elf_x86_link_hash_table.

With these changes, we can share _bfd_x86_elf_size_dynamic_sections in
elf32-i386.c and elf64-x86-64.c.

* elf32-i386.c (elf_i386_convert_load): Renamed to ...
(_bfd_i386_elf_convert_load): This.  Remove static.
(elf_i386_size_dynamic_sections): Removed.
(elf_backend_size_dynamic_sections): Likewise.
* elf64-x86-64.c (elf_x86_64_convert_load): Renamed to ...
(_bfd_x86_64_elf_convert_load): This.  Remove static.
(elf_x86_64_size_dynamic_sections): Removed.
(elf_backend_size_dynamic_sections): Likewise.
* elfxx-x86.c (_bfd_x86_elf_allocate_dynrelocs): Renamed to ...
(elf_x86_allocate_dynrelocs): This.  Make it static.
(_bfd_x86_elf_allocate_local_dynrelocs): Renamed to ...
(elf_x86_allocate_local_dynreloc): This.  Make it static.
(elf_i386_is_reloc_section): New function.
(elf_x86_64_is_reloc_section): Likewise.
(_bfd_x86_elf_link_hash_table_create): Initialize convert_load,
is_reloc_section, dt_reloc, dt_reloc_sz and dt_reloc_ent.
Rearrange got_entry_size initialization.
(_bfd_x86_elf_size_dynamic_sections): New function.
* elfxx-x86.h (elf_x86_link_hash_table): Add convert_load,
is_reloc_section, dt_reloc, dt_reloc_sz and dt_reloc_ent.
(_bfd_i386_elf_convert_load): New.
(_bfd_x86_64_elf_convert_load): Likewise.
(_bfd_x86_elf_size_dynamic_sections): Likewise.
(elf_backend_size_dynamic_sections): Likewise.
(_bfd_x86_elf_allocate_dynrelocs): Removed.
(_bfd_x86_elf_allocate_local_dynrelocs): Likewise.

x86: Rearrange fields and update comments

* elfxx-x86.h (elf_x86_link_hash_table): Rearrange fields and
update comments.

i386: Update sgotplt_jump_table_size setting

elf_i386_size_dynamic_sections has

      htab->next_tls_desc_index = htab->elf.srelplt->reloc_count;
      htab->sgotplt_jump_table_size = htab->next_tls_desc_index * 4;

This patch changes it to

      htab->sgotplt_jump_table_size
= elf_x86_compute_jump_table_size (htab)

Since elf_x86_compute_jump_table_size is defined as

  ((htab)->elf.srelplt->reloc_count * (htab)->got_entry_size)

there is no change in output.  It makes elf_i386_size_dynamic_sections
the same as elf_x86_64_size_dynamic_sections.

* elf32-i386.c (elf_i386_size_dynamic_sections): Set
sgotplt_jump_table_size with elf_x86_compute_jump_table_size.

x86: Define PLT_CIE_LENGTH and PLT_FDE_* in elfxx-x86.h

Since PLT_CIE_LENGTH, PLT_FDE_LENGTH, PLT_FDE_START_OFFSET and
PLT_FDE_LEN_OFFSET are identical in elf32-i386.c and elf64-x86-64.c,
they can be defined in elfxx-x86.h.

* elf32-i386.c (PLT_CIE_LENGTH, PLT_FDE_LENGTH,
PLT_FDE_START_OFFSET, PLT_FDE_LEN_OFFSET): Moved to ...
* elfxx-x86.h (PLT_CIE_LENGTH, PLT_FDE_LENGTH,
PLT_FDE_START_OFFSET, PLT_FDE_LEN_OFFSET): Here.
* elf64-x86-64.c (PLT_CIE_LENGTH, PLT_FDE_LENGTH,
PLT_FDE_START_OFFSET, PLT_FDE_LEN_OFFSET): Removed.

x86: Add _bfd_x86_elf_allocate_dynrelocs

Share _bfd_x86_elf_allocate_dynrelocs in elf32-i386.c and elf64-x86-64.c.

* elf32-i386.c (elf_i386_allocate_dynrelocs): Removed.
(elf_i386_allocate_local_dynrelocs): Likewise.
(elf_i386_size_dynamic_sections): Replace
elf_i386_allocate_dynrelocs/elf_i386_allocate_local_dynrelocs
with _bfd_x86_elf_allocate_dynrelocs and
_bfd_x86_elf_allocate_local_dynrelocs.
* elf64-x86-64.c (elf_x86_64_allocate_dynrelocs): Removed.
(elf_x86_64_allocate_local_dynrelocs): Likewise.
(elf_x86_64_size_dynamic_sections): Replace
elf_x86_64_allocate_dynrelocs/elf_x86_64_allocate_local_dynrelocs
with _bfd_x86_elf_allocate_dynrelocs and
_bfd_x86_elf_allocate_local_dynrelocs.
* elfxx-x86.c (_bfd_x86_elf_allocate_dynrelocs): New function.
(_bfd_x86_elf_allocate_local_dynrelocs): Likewise.
* elfxx-x86.h (_bfd_x86_elf_allocate_dynrelocs): New prototype.
(_bfd_x86_elf_allocate_local_dynrelocs): Likewise.

x86: Add is_x86_elf

Share is_x86_elf in elf32-i386.c and elf64-x86-64.c.

* elf32-i386.c (is_i386_elf): Removed.
(elf_i386_check_relocs): Replace is_i386_elf with is_x86_elf.
(elf_i386_size_dynamic_sections): Likewise.
(elf_i386_relocate_section): Likewise.
* elf64-x86-64.c (is_x86_64_elf): Removed.
(elf_x86_64_check_relocs): Replace is_x86_64_elf with
is_x86_elf.
(elf_x86_64_size_dynamic_sections): Likewise.
(elf_x86_64_relocate_section): Likewise.
* elfxx-x86.c (_bfd_x86_elf_link_hash_table_create): Initialize
target_id.
* elfxx-x86.h (elf_x86_link_hash_table): Add target_id.
(is_x86_elf): New.

x86: Add elf_x86_compute_jump_table_size

Share elf_x86_compute_jump_table_size in elf32-i386.c and
elf64-x86-64.c.

* elf32-i386.c (elf_i386_compute_jump_table_size): Removed.
(elf_i386_allocate_dynrelocs): Replace
elf_i386_compute_jump_table_size with
elf_x86_compute_jump_table_size.
(elf_i386_size_dynamic_sections): Likewise.
* elf64-x86-64.c (elf_x86_64_compute_jump_table_size): Removed.
(elf_x86_64_allocate_dynrelocs): Replace
elf_x86_64_compute_jump_table_size with
elf_x86_compute_jump_table_size.
(elf_x86_64_size_dynamic_sections): Likewise.
* elfxx-x86.c (_bfd_x86_elf_link_hash_table_create): Initialize
got_entry_size.
* elfxx-x86.h (elf_x86_link_hash_table): Add got_entry_size.
(elf_x86_compute_jump_table_size): New.

x86: Add sizeof_reloc to elf_x86_link_hash_table

Initialize htab->sizeof_reloc once, instead of computing it every time.

* elfxx-x86.c (_bfd_x86_elf_link_hash_table_create): Initialize
sizeof_reloc.
(_bfd_x86_elf_adjust_dynamic_symbol): Use sizeof_reloc.
* elfxx-x86.h (elf_x86_link_hash_table): Add sizeof_reloc.

i386: Check VxWorks with htab->is_vxworks

* elf32-i386.c (elf_i386_allocate_dynrelocs): Check VxWorks
with htab->is_vxworks.
(elf_i386_size_dynamic_sections): Likewise.
(elf_i386_relocate_section): Likewise.
(elf_i386_finish_dynamic_symbol): Likewise.
(elf_i386_finish_dynamic_sections): Likewise.

x86: Move GOT_TLS_* in elf32-i386.c to elfxx-x86.h

elf64-x86-64.c can use GOT_TLS_* definitions in elf32-i386.c with
GOT_TLS_IE_POS, GOT_TLS_IE_NEG and GOT_TLS_IE_BOTH unused.

* elf32-i386.c (GOT_TLS_IE, GOT_TLS_IE_POS, GOT_TLS_IE_NEG,
GOT_TLS_IE_BOTH, GOT_TLS_GDESC, GOT_TLS_GD_BOTH_P,
GOT_TLS_GD_P, GOT_TLS_GDESC_P, GOT_TLS_GD_ANY_P): Moved to ...
* elfxx-x86.h (GOT_TLS_IE, GOT_TLS_IE_POS, GOT_TLS_IE_NEG,
GOT_TLS_IE_BOTH, GOT_TLS_GDESC, GOT_TLS_GD_BOTH_P,
GOT_TLS_GD_P, GOT_TLS_GDESC_P, GOT_TLS_GD_ANY_P): Here.
* elf64-x86-64.c (GOT_TLS_IE, GOT_TLS_GDESC, GOT_TLS_GD_BOTH_P,
GOT_TLS_GD_P, GOT_TLS_GDESC_P, GOT_TLS_GD_ANY_P): Removed.

LTO rescan archives

ld ought to be more clever about where it puts LTO recompiled objects.
Ideally the recompiled objects ought to be ordered to the same place
their IR objects were, and files extracted from archives on the second
pass ought to go in the same place as they would if extracted on the
first pass.  This patch addresses the archive problem.  Without this
fix, objects extracted from archives might be placed after the crt
files intended to go at the end of an executable or shared library,
possibly causing exception handling failures.

* ldlang.h (lang_input_statement_type): Expand comments.
(LANG_FOR_EACH_INPUT_STATEMENT): Rewrite without casts.
* ldlang.c (lang_for_each_input_file): Likewise.
(load_symbols): Set usrdata for archives.
(find_rescan_insertion): New function.
(lang_process): Trim off and reinsert entries added to file chain
when rescanning archives for LTO.
* ldmain.c (add_archive_element): Set my_archive input_statement
next pointer to last element added.

x86-64: Check ELF_COMMON_DEF_P for common symbols

bfd/

PR ld/22064
* elf64-x86-64.c (elf_x86_64_finish_dynamic_symbol): Check
ELF_COMMON_DEF_P for common symbols.

ld/

PR ld/22064
* testsuite/ld-x86-64/pr22064a.S: New file.
* testsuite/ld-x86-64/pr22064b.c: Likewise.
* testsuite/ld-x86-64/x86-64.exp: Run PR ld/22064 test.

-Og warning fixes

Found when building with gcc 4.9.4 using -Og.

bfd/
* elf-eh-frame.c (offset_adjust): Avoid false positive gcc warning.
* elflink.c (bfd_elf_size_dynsym_hash_dynstr): Likewise.
* elfnn-aarch64.c (elfNN_aarch64_final_link_relocate): Likewise.
ld/
* emultempl/msp430.em (eval_upper_either_sections): Make base_sec_name
a const char*.
(eval_lower_either_sections): Likewise.
(msp430_elf_after_allocation): Likewise, and don't needlessly concat
and free.  Warning fix.

Automatic date update in version.in

x86: Add _bfd_x86_elf_gc_mark_hook

Since R_X86_64_GNU_VTINHERIT == R_386_GNU_VTINHERIT and
R_X86_64_GNU_VTENTRY == R_386_GNU_VTENTRY, we can share
_bfd_x86_elf_gc_mark_hook in elf32-i386.c and elf64-x86-64.c.

* elf32-i386.c (elf_i386_gc_mark_hook): Removed.
(elf_backend_gc_mark_hook): Likewise.
* elf64-x86-64.c (elf_x86_64_gc_mark_hook): Likewise.
(elf_backend_gc_mark_hook): Likewise.
* elfxx-x86.c (_bfd_x86_elf_gc_mark_hook): New function.
* elfxx-x86.h (_bfd_x86_elf_gc_mark_hook): New.
(elf_backend_gc_mark_hook): Likewise.

x86: Add _bfd_x86_elf_adjust_dynamic_symbol

Share _bfd_x86_elf_adjust_dynamic_symbol in elf32-i386.c and
elf64-x86-64.c.

* elf32-i386.c (elf_i386_adjust_dynamic_symbol): Removed.
(elf_backend_adjust_dynamic_symbol): Likewise.
* elf64-x86-64.c (elf_x86_64_adjust_dynamic_symbol): Likewise.
(elf_backend_adjust_dynamic_symbol): Likewise.
* elfxx-x86.c (_bfd_x86_elf_adjust_dynamic_symbol): New function.
(_bfd_x86_elf_link_setup_gnu_properties): Copy is_vxworks.
* elfxx-x86.h (elf_x86_link_hash_table): Add is_vxworks.
(_bfd_x86_elf_adjust_dynamic_symbol): New.
(elf_backend_adjust_dynamic_symbol): Likewise.

Correct ChangeLog entry

elfxx-x86.h: Fix a typo in comments

* elfxx-x86.h (elf_x86_plt_layout_table): Fix a typo in
comments.

x86: Add _bfd_x86_elf_mkobject

Share _bfd_x86_elf_mkobject in elf32-i386.c and elf64-x86-64.c.

* elf32-i386.c (elf_i386_mkobject): Removed.
(bfd_elf32_mkobject): Likewise.
* elf64-x86-64.c (elf_x86_64_mkobject): Likewise.
(bfd_elf64_mkobject): Likewise.
(bfd_elf32_mkobject): Likewise.
* elfxx-x86.c (_bfd_x86_elf_mkobject): New function.
(_bfd_x86_elf_mkobject): New.
(bfd_elf64_mkobject): Likewise.
(bfd_elf32_mkobject): Likewise.

x86: Add _bfd_x86_elf_link_setup_gnu_properties

Extract the common parts of elf_i386_link_setup_gnu_properties and
elf_x86_64_link_setup_gnu_properties into a new function.

For x86-64, since PIC PLT layouts are the same as non-PIC PLT layouts,
initialize pic_plt0_entry and pic_plt_entry fields in PLT layouts with
the non-PIC PLT entries.

* elf32-i386.c (elf_i386_link_setup_gnu_properties): Updated.
Call _bfd_x86_elf_link_setup_gnu_properties.
* elf64-x86-64.c (elf_x86_lazy_plt_layout): Initialize
pic_plt0_entry and pic_plt_entry fields with the non-PIC PLT
entries.
(elf_x86_64_non_lazy_plt): Likewise.
(elf_x86_64_lazy_bnd_plt): Likewise.
(elf_x86_64_non_lazy_bnd_plt): Likewise.
(elf_x86_64_lazy_ibt_plt): Likewise.
(elf_x32_lazy_ibt_plt): Likewise.
(elf_x86_64_non_lazy_ibt_plt): Likewise.
(elf_x32_non_lazy_ibt_plt): Likewise.
(elf_x86_64_nacl_plt): Likewise.
(elf_x86_64_link_setup_gnu_properties): Updated.  Call
_bfd_x86_elf_link_setup_gnu_properties.
* elfxx-x86.c: Include elf-vxworks.h".
(_bfd_x86_elf_link_setup_gnu_properties): New function.
* elfxx-x86.h (elf_x86_lazy_plt_layout): Remove "for i386 only"
comments for pic_plt0_entry and pic_plt_entry.
(elf_x86_non_lazy_plt_layout): Likewise.
(elf_x86_plt_layout_table): New.
(_bfd_x86_elf_link_setup_gnu_properties): Likewise.

Import latest fixes to libiberty from GCC

Fix warning for simple-object-elf.c.

2017-09-01  Martin Liska  <mliska@suse.cz>

* simple-object-elf.c (simple_object_elf_copy_lto_debug_sections):
Remove duplicite declaration.

x86: Correct unwind information for the second PLT

For i386, generate unwind information for the second PLT.  For x32,
correct alignment of .eh_frame section for the second PLT.

bfd/

PR ld/22061
* elf32-i386.c (elf_i386_link_setup_gnu_properties): Create
.eh_frame section for the second PLT.
* elf64-x86-64.c (elf_x86_64_link_setup_gnu_properties): Correct
alignment of .eh_frame section for the second PLT.

ld/

PR ld/22061
* testsuite/ld-i386/ibt-plt-1.d: Updated.
* testsuite/ld-i386/ibt-plt-2a.d: Likewise.
* testsuite/ld-i386/ibt-plt-2c.d: Likewise.
* testsuite/ld-i386/ibt-plt-3a.d: Likewise.
* testsuite/ld-i386/ibt-plt-3c.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-1-x32.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-2a-x32.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-2c-x32.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-3a-x32.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-3c-x32.d: Likewise.
* testsuite/ld-i386/ibt-plt-2b.d: Pass --hash-style=sysv to ld
and dump unwind information.
* testsuite/ld-i386/ibt-plt-2d.d: Likewise.
* testsuite/ld-i386/ibt-plt-3b.d: Likewise.
* testsuite/ld-i386/ibt-plt-3d.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-2b-x32.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-2b.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-2d-x32.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-2d.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-3b-x32.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-3b.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-3d-x32.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-3d.d: Likewise.

Enable support for the AArch64 dot-prod instruction in the Cortex A55 and A75 cpus.

* config/tc-aarch64.c (aarch64_cpus): Enable DOTPROD for
cortex-a55 and cortx-a75.

Prevent an address violation parsing corrupt DWARF information by fixing the test for an overlong debug line info structure.

PR 22059
* dwarf2.c (decode_line_info): Fix test for an overlong line info
structure.

Import latest fixes to libiberty from GCC.

PR lto/81968
* simple-object-elf.c (simple_object_elf_copy_lto_debug_section):
Keep names of removed global symbols.

* simple-object-xcoff.c (simple_object_xcoff_find_sections):
Improve .go_export csect handling.  Don't make assumptions
on containing section or number of auxiliary entries.

PR lto/81968
* simple-object-elf.c (simple_object_elf_copy_lto_debug_section):
Adjust field with for sh_type write, set SHF_EXCLUDE only for
removed sections.

PR lto/81925
* simple-object-elf.c (simple_object_elf_write_shdr): Adjust
type of sh_addralign and sh_entsize and properly write
sh_entsize as Elf_Addr.
(simple_object_elf_write_to_file): Read sh_entsize as Elf_Addr.

* simple-object-common.h (struct simple_object_functions): Add
copy_lto_debug_sections hook.
* simple-object.c: Include fcntl.h.
(handle_lto_debug_sections): New helper function.
(simple_object_copy_lto_debug_sections): New function copying
early LTO debug sections to regular debug sections in a new file.
(simple_object_start_write): Handle NULL segment_name.
* simple-object-coff.c (simple_object_coff_functions): Adjust
for not implemented copy_lto_debug_sections hook.
* simple-object-mach-o.c (simple_object_mach_o_functions): Likewise.
* simple-object-xcoff.c (simple_object_xcoff_functions): Likewise.
* simple-object-elf.c (SHT_NULL, SHT_SYMTAB, SHT_RELA, SHT_REL,
SHT_GROUP): Add various sectopn header types.
(SHF_EXCLUDE): Add flag.
(Elf32_External_Sym, Elf64_External_Sym): Add symbol struct.
(ELF_ST_BIND, ELF_ST_TYPE, ELF_ST_INFO): Add accessors.
(STT_OBJECT, STT_FUNC, STT_TLS, STT_GNU_IFUNC): Add Symbol types.
(STV_DEFAULT): Add symbol visibility.
(SHN_COMMON): Add special section index name.
(struct simple_object_elf_write): New.
(simple_object_elf_start_write): Adjust for new private data.
(simple_object_elf_write_shdr): Pass in values for all fields
we write.
(simple_object_elf_write_to_file): Adjust.  Copy from recorded
section headers if requested.
(simple_object_elf_release_write): Release private data.
(simple_object_elf_copy_lto_debug_sections): Copy and rename sections
as denoted by PFN and all their dependences, symbols and relocations
to the empty destination file.
(simple_object_elf_functions): Adjust for copy_lto_debug_sections hook.

* simple-object-xcoff.c (simple_object_xcoff_find_sections):
Search symbol table for .go_export symbol and apply pfn if found.

Fix buffer overrun when parsing an ELF attribute string that is not NUL terminated.

PR 22058
* elf-attrs.c (_bfd_elf_parse_attributes): Ensure that the
attribute buffer is NUL terminated.

PR gdb/22046: Fix T-stopped detach regression on old Linux kernels

On <=RHEL6 hosts Fedora/RHEL GDB started to 'kill -STOP' all processes it
detached.  Even those not originally T-stopped.  This is a Fedora-specific
patch which is based on upstream GDB's PROC_STATE_STOPPED state.

I believe (I did not verify) this patch did regress it:
commit d617208bb06bd461b52ce041d89f7127e3044762
Author: Pedro Alves <palves@redhat.com>
Date:   Mon Jul 25 12:42:17 2016 +0100
    linux-procfs: Introduce enum proc_state

As originally there was strstr() but now there is strcmp() and so the missing
trailing '\n' no longer matches.

The Bug was found by Michal Kolar.

Reproducibility:
$ gdb -p $PID
(gdb) quit
$ ...

Actual results:
===
RHEL6.9 x86_64 # scl enable devtoolset-7 bash
RHEL6.9 x86_64 # which gdb
/opt/rh/devtoolset-7/root/usr/bin/gdb
RHEL6.9 x86_64 # ./testcase.sh
24737 pts/0    S+     0:00 /bin/sleep 4
24737 pts/0    T+     0:00 /bin/sleep 4
RHEL6.9 x86_64 #
===

Expected results:
===
RHEL6.9 x86_64 # which gdb
/usr/bin/gdb
RHEL6.9 x86_64 # ./testcase.sh
24708 pts/0    S+     0:00 /bin/sleep 4
24708 pts/0    S+     0:00 /bin/sleep 4
./testcase.sh: line 20: kill: (24708) - No such process
RHEL6.9 x86_64 #
===

gdb/ChangeLog
2017-09-01  Jan Kratochvil  <jan.kratochvil@redhat.com>

PR gdb/22046
* nat/linux-procfs.c (parse_proc_status_state): Fix PROC_STATE_STOPPED
detection.

Automatic date update in version.in

Implement the ability to set/unset environment variables to GDBserver when starting the inferior

This patch implements the ability to set/unset environment variables
on the remote target, mimicking what GDB already offers to the user.
There are two features present here: user-set and user-unset
environment variables.

User-set environment variables are only the variables that are
explicitly set by the user, using the 'set environment' command.  This
means that variables that were already present in the environment when
starting GDB/GDBserver are not transmitted/considered by this feature.

User-unset environment variables are variables that are explicitly
unset by the user, using the 'unset environment' command.

The idea behind this patch is to store user-set and user-unset
environment variables in two separate sets, both part of gdb_environ.
Then, when extended_remote_create_inferior is preparing to start the
inferior, it will iterate over the two sets and set/unset variables
accordingly.  Three new packets are introduced:

- QEnvironmentHexEncoded, which is used to set environment variables,
  and contains an hex-encoded string in the format "VAR=VALUE" (VALUE
  can be empty if the user set a variable with a null value, by doing
  'set environment VAR=').

- QEnvironmentUnset, which is used to unset environment variables, and
  contains an hex-encoded string in the format "VAR".

- QEnvironmentReset, which is always the first packet to be
  transmitted, and is used to reset the environment, i.e., discard any
  changes made by the user on previous runs.

The QEnvironmentHexEncoded packet is inspired on LLDB's extensions to
the RSP.  Details about it can be seen here:

  <https://raw.githubusercontent.com/llvm-mirror/lldb/master/docs/lldb-gdb-remote.txt>

I decided not to implement the QEnvironment packet because it is
considered deprecated by LLDB.  This packet, on LLDB, serves the same
purpose of QEnvironmentHexEncoded, but sends the information using a
plain text, non-hex-encoded string.

The other two packets are new.

This patch also includes updates to the documentation, testsuite, and
unit tests, without introducing regressions.

gdb/ChangeLog:
2017-08-31  Sergio Durigan Junior  <sergiodj@redhat.com>

* NEWS (Changes since GDB 8.0): Add entry mentioning new support
for setting/unsetting environment variables on the remote target.
(New remote packets): Add entries for QEnvironmentHexEncoded,
QEnvironmentUnset and QEnvironmentReset.
* common/environ.c (gdb_environ::operator=): Extend method to
handle m_user_set_env_list and m_user_unset_env_list.
(gdb_environ::clear): Likewise.
(match_var_in_string): Change type of first parameter from 'char
*' to 'const char *'.
(gdb_environ::set): Extend method to handle
m_user_set_env_list and m_user_unset_env_list.
(gdb_environ::unset): Likewise.
(gdb_environ::clear_user_set_env): New method.
(gdb_environ::user_set_envp): Likewise.
(gdb_environ::user_unset_envp): Likewise.
* common/environ.h (gdb_environ): Handle m_user_set_env_list and
m_user_unset_env_list on move constructor/assignment.
(unset): Add new default parameter 'update_unset_list = true'.
(clear_user_set_env): New method.
(user_set_envp): Likewise.
(user_unset_envp): Likewise.
(m_user_set_env_list): New std::set.
(m_user_unset_env_list): Likewise.
* common/rsp-low.c (hex2str): New function.
(bin2hex): New overload for bin2hex function.
* common/rsp-low.c (hex2str): New prototype.
(str2hex): New overload prototype.
* remote.c: Include "environ.h". Add QEnvironmentHexEncoded,
QEnvironmentUnset and QEnvironmentReset.
(remote_protocol_features): Add QEnvironmentHexEncoded,
QEnvironmentUnset and QEnvironmentReset packets.
(send_environment_packet): New function.
(extended_remote_environment_support): Likewise.
(extended_remote_create_inferior): Call
extended_remote_environment_support.
(_initialize_remote): Add QEnvironmentHexEncoded,
QEnvironmentUnset and QEnvironmentReset packet configs.
* unittests/environ-selftests.c (gdb_selftest_env_var):
New variable.
(test_vector_initialization): New function.
(test_init_from_host_environ): Likewise.
(test_reinit_from_host_environ): Likewise.
(test_set_A_unset_B_unset_A_cannot_find_A_can_find_B):
Likewise.
(test_unset_set_empty_vector): Likewise.
(test_vector_clear): Likewise.
(test_std_move): Likewise.
(test_move_constructor):
(test_self_move): Likewise.
(test_set_unset_reset): Likewise.
(run_tests): Rewrite in terms of the functions above.

gdb/gdbserver/ChangeLog:
2017-08-31  Sergio Durigan Junior  <sergiodj@redhat.com>

* server.c (handle_general_set): Handle QEnvironmentHexEncoded,
QEnvironmentUnset and QEnvironmentReset packets.
(handle_query): Inform remote that QEnvironmentHexEncoded,
QEnvironmentUnset and QEnvironmentReset are supported.

gdb/doc/ChangeLog:
2017-08-31  Sergio Durigan Junior  <sergiodj@redhat.com>

* gdb.texinfo (set environment): Add @anchor.  Explain that
environment variables set by the user are sent to GDBserver.
(unset environment): Likewise, but for unsetting variables.
(Connecting) <Remote Packet>: Add "environment-hex-encoded",
"QEnvironmentHexEncoded", "environment-unset", "QEnvironmentUnset",
"environment-reset" and "QEnvironmentReset" to the table.
(Remote Protocol) <QEnvironmentHexEncoded, QEnvironmentUnset,
QEnvironmentReset>: New item, explaining the packet.

gdb/testsuite/ChangeLog:
2017-08-31  Sergio Durigan Junior  <sergiodj@redhat.com>

* gdb.base/share-env-with-gdbserver.c: New file.
* gdb.base/share-env-with-gdbserver.exp: Likewise.

Fix buffer read overrun by ensuring that DWARF sections containing strings always end in a NUL byte.

PR 22047
* dwarf2.c (read_section): If necessary add a terminating NUL byte
to dwarf string sections.

Add updated French translations for opcodes and gprof

Add a test for PR ld/22048

PR ld/22048
* testsuite/ld-x86-64/pr22048.d: New file.
* testsuite/ld-x86-64/pr22048a.s: Likewise.
* testsuite/ld-x86-64/pr22048b.s: Likewise.
* testsuite/ld-x86-64/x86-64.exp: Run pr22048.

Regen SRC-POTFILES.in

* po/SRC-POTFILES.in: Regenerate.

Remove .eh_frame zero terminators

The machinery to do this was there, but not enabled if the terminator
was the only thing in the section.

bfd/
* elf-eh-frame.c (_bfd_elf_parse_eh_frame): Don't exit early
for a section containing just a terminator.  Allow multiple
terminators at end of section.
* elflink.c (bfd_elf_discard_info): Iterate over .eh_frame
sections when not adding alignment.  Assert on terminator in
the middle of FDEs.
ld/
* testsuite/ld-elf/eh3.d: Update.
* testsuite/ld-elf/eh4.d: Update.

Unbreak gdb build on 32-bit host with ADI support

The problem of failing to build with arm-linux-gnueabihf-g++-4.8 was
that type CORE_ADDR is of "unsigned long" on a 64-bit machine so it's
OK to use %lx but is of type "unsigned long long" on a 32 bit system.

Fixed the problem in three places - (1) use a temp variable of type
CORE_ADDR as argument 3 when calling target_auxv_search() then assign
its value to "blksize" and "nbits" in 2 calls; (2) redo
adi_normalize_address() using masks and xor operators to calculate
normalized address; (3) call paddress() to print CORE_ADDR in either
printf_filtered() or error(). Thank you, Pedro, for all your
suggestions.

gdb/ChangeLog:
2017-08-31  Weimin Pan  <weimin.pan@oracle.com>

* sparc64-tdep.c (adi_stat_t): Fix comment formatting.
(adi_available): Use a temp variable of type CORE_ADDR as argument
3 when calling target_auxv_search.
(adi_normalize_address): Use masks and xor operators to calculate
normalized address.
(adi_read_versions, adi_write_versions, adi_print_versions)
(do_examine, do_assign): Use paddress.

PR22048, Incorrect .eh_frame section in libc.so

PR 21441
PR 22048
* elflink.c (bfd_elf_discard_info): Don't pad embedded zero
terminators.

Add elf64.lo together with elfxx-x86.lo for 64-bit BFD

Since elfxx-x86.lo needs elf64.lo with 64-bit BFD, add elf64.lo together
with elfxx-x86.lo to bfd_backends for 64-bit BFD.

* configure.ac (bfd_backends): Add elf64.lo together with
elfxx-x86.lo for 64-bit BFD.
* configure: Regenerated.

FT32: improve disassembly readability

For opcode fields that are not addresses, display as
integers instead of using print_address_func.

opcodes/ChangeLog:

2017-08-31  James Bowman  <james.bowman@ftdichip.com>

* ft32-dis.c (print_insn_ft32): Correct display of non-address
fields.

FT32: Permit R_FT32_18 overflow

The howto for R_FT32_18 was using complain_overflow_signed. But some
valid address calculations exceed the range of this reloc.  Changing it
to complain_overflow_dont allows them.

bfd/ChangeLog:

* elf32-ft32.c (ft32_elf_howto_table): Use
complain_overflow_dont for R_FT32_18.

bfd_close_all_done calling _close_and_cleanup

elf64_vms_close_and_cleanup calls bfd_get_size, which calls
iovec->bstat.  cache_bstat ends up adding the bfd to the cache lru
list, negating the bfd_cache_close call in bfd_close_all_done.  So
there is a dangling pointer into the freed and then reused bfd.  Thus,
bfd_cache_close must be called after _close_and_cleanup, or better,
via iovec->bclose.

PR binutils/22032
* opncls.c (bfd_close_all_done): Don't call bfd_cache_close
before _close_and_cleanup.  Call iovec->bclose after.
(bfd_close): Remove code common to, and call, bfd_close_all_done.

Automatic date update in version.in

LD/testsuite: Also discard `.MIPS.options' in orphan tests

Complement commit 5b5f4e6f8cd2 ("ld: Early detection of orphans we know
will be discarded") and add `.MIPS.options' to the list of sections
discarded with orphan tests, removing failures like:

./ld-new: error: unplaced orphan section `.MIPS.options' from `tmpdir/orphan-11.o'.
FAIL: ld-elf/orphan-11

and:

./ld-new: error: unplaced orphan section `.MIPS.options' from `tmpdir/orphan-12.o'.
FAIL: ld-elf/orphan-12

from n64 MIPS testing.

ld/
* testsuite/ld-elf/orphan-11.ld: Also discard `.MIPS.options'
sections.

x86: Add _bfd_x86_elf_get_synthetic_symtab

Move the common codes in elf_i386_get_synthetic_symtab and
elf_x86_64_get_synthetic_symtab to _bfd_x86_elf_get_synthetic_symtab.

* elf32-i386.c (elf_i386_plt_type): Removed.
(elf_i386_plt): Likewise.
(elf_i386_get_synthetic_symtab): Updated.   Call
_bfd_x86_elf_get_synthetic_symtab.
* elf64-x86-64.c (elf_x86_64_plt_type): Removed.
(elf_x86_64_plt): Likewise.
(elf_x86_64_get_synthetic_symtab): Updated.  Call
_bfd_x86_elf_get_synthetic_symtab.
* elfxx-x86.c (elf_i386_get_plt_got_vma): New function.
(elf_x86_64_get_plt_got_vma): Likewise.
(elf_i386_valid_plt_reloc_p): Likewise.
(elf_x86_64_valid_plt_reloc_p): Likewise.
(_bfd_x86_elf_get_synthetic_symtab): Likewise.
* elfxx-x86.h (elf_x86_plt_type): New.
(elf_x86_plt): Likewise.
(_bfd_x86_elf_get_synthetic_symtab): Likewise.

x86: Check target_id instead of elf_machine_code

Since both elf32-i386.c and elf64-x86-64.c support targets with
different ELF_MACHINE_CODEs, _bfd_x86_elf_link_hash_table_create
should check target_id instead of elf_machine_code.

* elfxx-x86.c (_bfd_x86_elf_link_hash_table_create): Check
target_id instead of elf_machine_code.

x86: Add _bfd_x86_elf_link_hash_table_create

Share _bfd_x86_elf_link_hash_table_create in elf32-i386.c and
elf64-x86-64.c by:

1. Replace elf_i386_lazy_plt_layout, elf_i386_non_lazy_plt_layout,
elf_i386_plt_layout, elf_x86_64_lazy_plt_layout,
elf_x86_64_non_lazy_plt_layout and elf_x86_64_plt_layout with
elf_x86_lazy_plt_layout, elf_x86_non_lazy_plt_layout and
elf_x86_plt_layout.
2. Move plt, lazy_plt, non_lazy_plt, srelplt2 and next_tls_desc_index
from elf_i386_link_hash_table to elf_x86_link_hash_table.
3. Remove elf_i386_link_hash_table and elf_x86_64_link_hash_table.

* elf32-i386.c (ELF_DYNAMIC_INTERPRETER): Removed.
(elf_i386_lazy_plt_layout): Likewise.
(elf_i386_non_lazy_plt_layout): Likewise.
(elf_i386_plt_layout): Likewise.
(elf_i386_link_hash_table): Likewise.
(elf_i386_next_tls_desc_index): Likewise.
(elf_i386_srelplt2): Likewise.
(elf_i386_plt): Likewise.
(elf_i386_lazy_plt): Likewise.
(elf_i386_non_lazy_plt): Likewise.
(elf_i386_link_hash_table_create): Likewise.
(bfd_elf32_bfd_link_hash_table_create): Likewise.
(elf_i386_lazy_plt): Updated.
(elf_i386_non_lazy_plt): Likewise.
(elf_i386_lazy_ibt_plt): Likewise.
(elf_i386_non_lazy_ibt_plt): Likewise.
(elf_i386_allocate_dynrelocs): Likewise.
(elf_i386_size_dynamic_sections): Likewise.
(elf_i386_relocate_section): Likewise.
(elf_i386_finish_dynamic_symbol): Likewise.
(elf_i386_finish_dynamic_sections): Likewise.
(elf_i386_get_synthetic_symtab): Likewise.
(elf_i386_link_setup_gnu_properties): Likewise.
(elf_i386_nacl_plt): Likewise.
* elf64-x86-64.c (ABI_64_P): Removed.
(ELF64_DYNAMIC_INTERPRETER): Likewise.
(ELF32_DYNAMIC_INTERPRETER): Likewise.
(elf_x86_64_lazy_plt_layout): Likewise.
(elf_x86_64_non_lazy_plt_layout): Likewise.
(elf_x86_64_plt_layout): Likewise.
(elf_x86_64_link_hash_table): Likewise.
(elf_x86_64_plt): Likewise.
(elf_x86_64_lazy_plt): Likewise.
(elf_x86_64_non_lazy_plt): Likewise.
(elf_x86_64_link_hash_table_create): Likewise.
(bfd_elf64_bfd_link_hash_table_create): Likewise.
(bfd_elf32_bfd_link_hash_table_create): Likewise.
(elf_x86_64_lazy_plt): Updated.
(elf_x86_64_non_lazy_plt): Likewise.
(elf_x86_64_lazy_bnd_plt): Likewise.
(elf_x86_64_non_lazy_bnd_plt): Likewise.
(elf_x86_64_lazy_ibt_plt): Likewise.
(elf_x32_lazy_ibt_plt): Likewise.
(elf_x86_64_non_lazy_ibt_plt): Likewise.
(elf_x32_non_lazy_ibt_plt): Likewise.
(elf_x86_64_allocate_dynrelocs): Likewise.
(elf_x86_64_size_dynamic_sections): Likewise.
(elf_x86_64_relocate_section): Likewise.
(elf_x86_64_finish_dynamic_symbol): Likewise.
(elf_x86_64_finish_dynamic_sections): Likewise.
(elf_x86_64_get_synthetic_symtab): Likewise.
(elf_x86_64_link_setup_gnu_properties): Likewise.
(elf_x86_64_nacl_plt): Likewise.
* elfxx-x86.c: Include "objalloc.h", "elf/i386.h" and
"elf/x86-64.h".
(ELF32_DYNAMIC_INTERPRETER): New.
(ELF64_DYNAMIC_INTERPRETER): Likewise.
(ELFX32_DYNAMIC_INTERPRETER): Likewise.
(_bfd_x86_elf_link_hash_table_create): Likewise.
(_bfd_x86_elf_link_hash_table_free): Renamed to ...
(elf_x86_link_hash_table_free): This.  Make it static.
* elfxx-x86.h: Don't include "objalloc.h".
(ABI_64_P): New.
(elf_x86_lazy_plt_layout): Likewise.
(elf_x86_non_lazy_plt_layout): Likewise.
(elf_x86_plt_layout): Likewise.
(_bfd_x86_elf_link_hash_table_create): Likewise.
(bfd_elf64_bfd_link_hash_table_create): Likewise.
(bfd_elf32_bfd_link_hash_table_create): Likewise.
(elf_x86_link_hash_table): Add plt, lazy_plt, non_lazy_plt,
srelplt2 and next_tls_desc_index.
(_bfd_x86_elf_link_hash_table_free): Removed.

MIPS/BFD: Correct microMIPS cross-mode BAL to JALX relaxation

Fix a bug in commit a6ebf6169a1b ("MIPS: Convert cross-mode BAL to
JALX") and in BFD linker relaxation correct the microMIPS interpretation
of the branch offset, which is supposed to be shifted by 1 bit, rather
than 2 as in the regular MIPS case.

bfd/
* elfxx-mips.c (mips_elf_perform_relocation): Correct microMIPS
branch offset interpretation.

gas/
* testsuite/gas/mips/branch-addend-micromips.d: New test.
* testsuite/gas/mips/branch-addend-micromips-n32.d: New test.
* testsuite/gas/mips/branch-addend-micromips-n64.d: New test.
* testsuite/gas/mips/branch-addend-micromips.s: New test source.
* testsuite/gas/mips/mips.exp: Run the new tests.

ld/
* testsuite/ld-mips-elf/bal-jalx-addend-micromips.d: New test.
* testsuite/ld-mips-elf/bal-jalx-addend-micromips-n32.d: New
test.
* testsuite/ld-mips-elf/bal-jalx-addend-micromips-n64.d: New
test.
* testsuite/ld-mips-elf/bal-jalx-local-micromips.d: New test.
* testsuite/ld-mips-elf/bal-jalx-local-micromips-n32.d: New
test.
* testsuite/ld-mips-elf/bal-jalx-local-micromips-n64.d: New
test.
* testsuite/ld-mips-elf/bal-jalx-pic-micromips.d: New test.
* testsuite/ld-mips-elf/bal-jalx-pic-micromips-n32.d: New test.
* testsuite/ld-mips-elf/bal-jalx-pic-micromips-n64.d: New test.
* testsuite/ld-mips-elf/bal-jalx-pic-ignore-micromips.d: New
test.
* testsuite/ld-mips-elf/bal-jalx-pic-ignore-micromips-n32.d: New
test.
* testsuite/ld-mips-elf/bal-jalx-pic-ignore-micromips-n64.d: New
test.
* testsuite/ld-mips-elf/mips-elf.exp: Run the new tests.

MIPS/GAS: Also respect `-mignore-branch-isa' with MIPS16 code

Fix a bug in commit 8b10b0b3e100 ("MIPS: Add options to control branch
ISA checks") and with the `-mignore-branch-isa' command-line option also
lift a GAS check for invalid MIPS16 branches between ISA modes, which is
made separately from regular MIPS and microMIPS checks.

gas/
* config/tc-mips.c (md_convert_frag): Respect
`mips_ignore_branch_isa'.
* testsuite/gas/mips/branch-local-5.d: New test.
* testsuite/gas/mips/branch-local-n32-5.d: New test.
* testsuite/gas/mips/branch-local-n64-5.d: New test.
* testsuite/gas/mips/branch-local-6.d: New test.
* testsuite/gas/mips/branch-local-n32-6.d: New test.
* testsuite/gas/mips/branch-local-n64-6.d: New test.
* testsuite/gas/mips/branch-local-7.d: New test.
* testsuite/gas/mips/branch-local-n32-7.d: New test.
* testsuite/gas/mips/branch-local-n64-7.d: New test.
* testsuite/gas/mips/branch-local-ignore-5.d: New test.
* testsuite/gas/mips/branch-local-ignore-n32-5.d: New test.
* testsuite/gas/mips/branch-local-ignore-n64-5.d: New test.
* testsuite/gas/mips/branch-local-ignore-6.d: New test.
* testsuite/gas/mips/branch-local-ignore-n32-6.d: New test.
* testsuite/gas/mips/branch-local-ignore-n64-6.d: New test.
* testsuite/gas/mips/branch-local-5.l: New stderr output.
* testsuite/gas/mips/branch-local-6.l: New stderr output.
* testsuite/gas/mips/branch-local-5.s: New test source.
* testsuite/gas/mips/branch-local-6.s: New test source.
* testsuite/gas/mips/branch-local-7.s: New test source.
* testsuite/gas/mips/mips.exp: Run the new tests.

Call _close_and_cleanup in bfd_close_all_done

PR binutils/22032
* opncls.c (bfd_close_all_done): Call _close_and_cleanup.

Remove elf_x86_64_next_tls_desc_index/elf_x86_64_srelplt2

They are unused.

* elf64-x86-64.c (elf_x86_64_next_tls_desc_index): Removed.
(elf_x86_64_srelplt2): Likewise.

Add missing ChangeLog entries

MIPS/GAS/testsuite: Deduplicate error lists of branch local tests

Complement commit 7795a8f8bdde ("MIPS/GAS/testsuite: Convert branch
local list tests to dump tests") and share identical error lists among
branch local tests, removing duplicate copies.

gas/
* testsuite/gas/mips/branch-local-n32-2.d: Use `branch-local-2.l'
for `error-output'.
* testsuite/gas/mips/branch-local-n64-2.d: Likewise.
* testsuite/gas/mips/branch-local-n32-3.d: Use `branch-local-3.l'
for `error-output'.
* testsuite/gas/mips/branch-local-n64-3.d: Likewise.
* testsuite/gas/mips/branch-local-n32-2.l: Remove file.
* testsuite/gas/mips/branch-local-n64-2.l: Remove file.
* testsuite/gas/mips/branch-local-n32-3.l: Remove file.
* testsuite/gas/mips/branch-local-n64-3.l: Remove file.

PowerPC TPREL16_HA/LO reloc optimization

In the TLS GD/LD to LE optimization, ld replaces a sequence like

 addi 3,2,x@got@tlsgd R_PPC64_GOT_TLSGD16 x
 bl __tls_get_addr(x@tlsgd) R_PPC64_TLSGD x
R_PPC64_REL24 __tls_get_addr
 nop

with

 addis 3,13,x@tprel@ha R_PPC64_TPREL16_HA x
 addi 3,3,x@tprel@l R_PPC64_TPREL16_LO x
 nop

When the tprel offset is small, this can be further optimized to

 nop
 addi 3,13,x@tprel
 nop

bfd/
* elf64-ppc.c (struct ppc_link_hash_table): Add do_tls_opt.
(ppc64_elf_tls_optimize): Set it.
(ppc64_elf_relocate_section): Nop addis on TPREL16_HA, and convert
insn on TPREL16_LO and TPREL16_LO_DS relocs to use r13 when
addis would add zero.
* elf32-ppc.c (struct ppc_elf_link_hash_table): Add do_tls_opt.
(ppc_elf_tls_optimize): Set it.
(ppc_elf_relocate_section): Nop addis on TPREL16_HA, and convert
insn on TPREL16_LO relocs to use r2 when addis would add zero.
gold/
* powerpc.cc (Target_powerpc::Relocate::relocate): Nop addis on
TPREL16_HA, and convert insn on TPREL16_LO and TPREL16_LO_DS
relocs to use r2/r13 when addis would add zero.
ld/
* testsuite/ld-powerpc/tls.s: Add calls with tls markers.
* testsuite/ld-powerpc/tls32.s: Likewise.
* testsuite/ld-powerpc/powerpc.exp: Run tls marker tests.
* testsuite/ld-powerpc/tls.d: Adjust for TPREL16_HA/LO optimization.
* testsuite/ld-powerpc/tlsexe.d: Likewise.
* testsuite/ld-powerpc/tlsexetoc.d: Likewise.
* testsuite/ld-powerpc/tlsld.d: Likewise.
* testsuite/ld-powerpc/tlsmark.d: Likewise.
* testsuite/ld-powerpc/tlsopt4.d: Likewise.
* testsuite/ld-powerpc/tlstoc.d: Likewise.

PowerPC64 __tls_get_addr sequence optimization

There isn't a good reason for ld.bfd to behave differently from gold
in the code generated by TLS GD/LD to LE optimization.

bfd/
* elf64-ppc.c (ppc64_elf_relocate_section): When optimizing
__tls_get_addr call sequences to LE, don't move the addi down
to the nop.  Replace the bl with addi and leave the nop alone.
ld/
* testsuite/ld-powerpc/tls.d: Update.
* testsuite/ld-powerpc/tlsexe.d: Update.
* testsuite/ld-powerpc/tlsexetoc.d: Update.
* testsuite/ld-powerpc/tlsld.d: Update.
* testsuite/ld-powerpc/tlsmark.d: Update.
* testsuite/ld-powerpc/tlsopt4.d: Update.
* testsuite/ld-powerpc/tlstoc.d: Update.

testsuite/ld-cris various files: Run ld with --hash-style=sysv.

* testsuite/ld-cris/dso-pltdis1.d: Run ld with --hash-style=sysv.
* testsuite/ld-cris/dso-pltdis2.d,
testsuite/ld-cris/dso12-pltdis.d, testsuite/ld-cris/expdyn1.d,
testsuite/ld-cris/expdyn5.d, testsuite/ld-cris/expdyn6.d,
testsuite/ld-cris/expdyn7.d, testsuite/ld-cris/gotplt1.d,
testsuite/ld-cris/gotplt2.d, testsuite/ld-cris/gotplt3.d,
testsuite/ld-cris/hiddef1.d, testsuite/ld-cris/libdso-11.d,
testsuite/ld-cris/libdso-12.d, testsuite/ld-cris/libdso-12b.d,
testsuite/ld-cris/libdso-12c.d, testsuite/ld-cris/libdso-13.d,
testsuite/ld-cris/libdso-13b.d, testsuite/ld-cris/libdso-14.d,
testsuite/ld-cris/libdso-15.d, testsuite/ld-cris/libdso-15b.d,
testsuite/ld-cris/libdso-1b.d, testsuite/ld-cris/libdso-1c.d,
testsuite/ld-cris/libdso-1d.d, testsuite/ld-cris/libdso-4.d,
testsuite/ld-cris/pr16044.d, testsuite/ld-cris/pv32-1.d,
testsuite/ld-cris/tls-dso-dtpoffd2.d,
testsuite/ld-cris/tls-dso-dtpoffd4.d,
testsuite/ld-cris/tls-dso-tpoffgotcomm1.d,
testsuite/ld-cris/tls-dso-x1x2-1.d, testsuite/ld-cris/tls-gc-71.d,
testsuite/ld-cris/tls-ie-78.d, testsuite/ld-cris/tls-js1.d,
testsuite/ld-cris/tls-ldgdex-14.d,
testsuite/ld-cris/tls-ldgdex-15.d,
testsuite/ld-cris/tls-legdx-16.d,
testsuite/ld-cris/tls-legdx-17.d,
testsuite/ld-cris/tls-local-63.d,
testsuite/ld-cris/tls-local-64.d, testsuite/ld-cris/tls-ok-30.d,
testsuite/ld-cris/tls-ok-32.d, testsuite/ld-cris/tls-ok-34.d,
testsuite/ld-cris/tls-und-38.d, testsuite/ld-cris/tls-und-42.d,
testsuite/ld-cris/tls-und-46.d, testsuite/ld-cris/tls-und-50.d,
testsuite/ld-cris/weakref2.d, testsuite/ld-cris/weakref3.d,
testsuite/ld-cris/weakref4.d: Likewise.

Automatic date update in version.in

Look for FIR in the last FreeBSD/mips floating-point register.

FreeBSD/mips kernels were recently changed to include the floating
point implementation revision register in the floating point register
set exported in process cores and via ptrace() (r318067).  This change
will first ship in FreeBSD 12.0 when it is eventually released.  The
space used to hold FIR was previously reserved in 'struct fpreg' as a
zero-filled dummy for padding, so 'struct fpreg' has not changed in
size.  Since FIR should be non-zero on all MIPS processors supported
by FreeBSD, ignore a value of 0 from 'struct fpreg' and only report
non-zero values as a valid FIR register.

gdb/ChangeLog:

* mips-fbsd-nat.c (getfpregs_supplies): Return true for FIR.
* mips-fbsd-tdep.c (mips_fbsd_supply_fpregs): Split supply of FSR
out of loop and add supply of FIR.
(mips_fbsd_collect_fpregs): Split collect of FSR out of loop and
add collect of FIR.

x86: Re-indent elf32-i386.c/elf64-x86-64.c

* elf32-i386.c (elf_i386_pie_finish_undefweak_symbol):
Re-indent.
* elf64-x86-64.c (elf_x86_64_finish_local_dynamic_symbol):
Likewise.
(elf_x86_64_pie_finish_undefweak_symbol): Likewise.

Add elfxx-x86.h and elfxx-x86.c

Move common codes in elf32-i386.c and elf64-x86-64.c to elfxx-x86.c.

* Makefile.am (BFD32_BACKENDS): Add elfxx-x86.lo.
(BFD64_BACKENDS): Likewise.
(BFD32_BACKENDS_CFILES): Add elfxx-x86.c.
(BFD64_BACKENDS_CFILES): Likewise.
* Makefile.in: Regenerated.
* configure.ac (bfd_backends): Add elfxx-x86.lo together with
elf32-i386.lo and elf64-x86-64.lo.
* configure: Regenerated.
* elf32-i386.c: Include "elfxx-x86.h" instead of "sysdep.h",
"bfd.h", "bfdlink.h", "libbfd.h", "elf-bfd.h", "bfd_stdint.h",
"objalloc.h" and "hashtab.h".
(ELIMINATE_COPY_RELOCS): Removed.
(UNDEFINED_WEAK_RESOLVED_TO_ZERO):Likewise.
(SYMBOL_NO_COPYRELOC): Likewise.
(elf_i386_link_hash_entry): Likewise.
(GOT_UNKNOWN): Likewise.
(GOT_NORMAL): Likewise.
(GOT_TLS_GD): Likewise.
(elf_i386_hash_entry): Likewise.
(elf_i386_obj_tdata): Likewise.
(elf_i386_tdata): Likewise.
(elf_i386_local_got_tls_type): Likewise.
(elf_i386_local_tlsdesc_gotent): Likewise.
(elf_i386_hash_table): Likewise.
(elf_i386_link_hash_newfunc): Likewise.
(elf_i386_local_htab_hash): Likewise.
(elf_i386_local_htab_eq): Likewise.
(elf_i386_get_local_sym_hash): Likewise.
(elf_i386_link_hash_table_free): Likewise.
(elf_i386_copy_indirect_symbol): Likewise.
(elf_i386_fixup_symbol): Likewise.
(elf_i386_readonly_dynrelocs): Likewise.
(elf_i386_always_size_sections): Likewise.
(elf_i386_set_tls_module_base): Likewise.
(elf_i386_dtpoff_base): Likewise.
(compare_relocs): Likewise.
(elf_i386_hash_symbol): Likewise.
(elf_i386_parse_gnu_properties): Likewise.
(elf_i386_merge_gnu_properties): Likewise.
(elf_i386_link_check_relocs): Likewise.
(elf_i386_merge_symbol_attribute): Likewise.
(bfd_elf32_bfd_link_check_relocs): Likewise.
(elf_backend_copy_indirect_symbol): Likewise.
(elf_backend_always_size_sections): Likewise.
(elf_backend_omit_section_dynsym): Likewise.
(elf_backend_hash_symbol): Likewise.
(elf_backend_fixup_symbol): Likewise.
(elf_backend_parse_gnu_properties): Likewise.
(elf_backend_merge_gnu_properties): Likewise.
(elf_backend_merge_symbol_attribute): Likewise.
(elf_i386_mkobject): Updated.
(elf_i386_link_hash_table_create): Likewise.
(elf_i386_check_tls_transition): Likewise.
(elf_i386_tls_transition): Likewise.
(elf_i386_convert_load_reloc): Likewise.
(elf_i386_check_relocs): Likewise.
(elf_i386_adjust_dynamic_symbol): Likewise.
(elf_i386_allocate_dynrelocs): Likewise.
(elf_i386_convert_load): Likewise.
(elf_i386_size_dynamic_sections): Likewise.
(elf_i386_relocate_section): Likewise.
(elf_i386_finish_dynamic_symbol ): Likewise.
(elf_i386_finish_dynamic_sections): Likewise.
(elf_i386_output_arch_local_syms): Likewise.
(elf_i386_get_synthetic_symtab): Likewise.
(elf_i386_link_setup_gnu_properties): Likewise.
(elf_i386_link_hash_table): Use elf_x86_link_hash_table, instead
of elf_link_hash_table, as base.
(elf_i386_next_tls_desc_index): New.
(elf_i386_srelplt2): Likewise.
(elf_i386_plt): Likewise.
(elf_i386_lazy_plt): Likewise.
(elf_i386_non_lazy_plt): Likewise.
* elf32-x86-64.c: Include "elfxx-x86.h" instead of "sysdep.h",
"bfd.h", "bfdlink.h", "libbfd.h", "elf-bfd.h", "bfd_stdint.h",
"objalloc.h" and "hashtab.h".
(ELIMINATE_COPY_RELOCS): Removed.
(UNDEFINED_WEAK_RESOLVED_TO_ZERO):Likewise.
(SYMBOL_NO_COPYRELOC): Likewise.
(elf_x86_64_link_hash_entry): Likewise.
(GOT_UNKNOWN): Likewise.
(GOT_NORMAL): Likewise.
(GOT_TLS_GD): Likewise.
(elf_x86_64_hash_entry): Likewise.
(elf_x86_64_obj_tdata): Likewise.
(elf_x86_64_tdata): Likewise.
(elf_x86_64_local_got_tls_type): Likewise.
(elf_x86_64_local_tlsdesc_gotent): Likewise.
(elf_x86_64_hash_table): Likewise.
(elf_x86_64_link_hash_newfunc): Likewise.
(elf_x86_64_local_htab_hash): Likewise.
(elf_x86_64_local_htab_eq): Likewise.
(elf_x86_64_get_local_sym_hash): Likewise.
(elf_x86_64_link_hash_table_free): Likewise.
(elf_x86_64_copy_indirect_symbol): Likewise.
(elf_x86_64_fixup_symbol): Likewise.
(elf_x86_64_readonly_dynrelocs): Likewise.
(elf_x86_64_always_size_sections): Likewise.
(elf_x86_64_set_tls_module_base): Likewise.
(elf_x86_64_dtpoff_base): Likewise.
(compare_relocs): Likewise.
(elf_x86_64_merge_symbol_attribute): Likewise.
(elf_x86_64_hash_symbol): Likewise.
(elf_x86_64_parse_gnu_properties): Likewise.
(elf_x86_64_merge_gnu_properties): Likewise.
(elf_x86_64_link_check_relocs): Likewise.
(elf_backend_copy_indirect_symbol): Likewise.
(elf_backend_always_size_sections): Likewise.
(bfd_elf64_bfd_link_check_relocs): Likewise.
(elf_backend_merge_symbol_attribute): Likewise.
(elf_backend_hash_symbol): Likewise.
(elf_backend_omit_section_dynsym): Likewise.
(elf_backend_fixup_symbol): Likewise.
(elf_backend_parse_gnu_properties): Likewise.
(elf_backend_merge_gnu_properties): Likewise.
(bfd_elf32_bfd_link_check_relocs): Likewise.
(elf_x86_64_mkobject): Updated.
(elf_x86_64_link_hash_table_create): Likewise.
(elf_x86_64_check_tls_transition): Likewise.
(elf_x86_64_tls_transition): Likewise.
(elf_x86_64_convert_load_reloc): Likewise.
(elf_x86_64_check_relocs): Likewise.
(elf_x86_64_adjust_dynamic_symbol): Likewise.
(elf_x86_64_allocate_dynrelocs): Likewise.
(elf_x86_64_convert_load): Likewise.
(elf_x86_64_size_dynamic_sections): Likewise.
(elf_x86_64_relocate_section): Likewise.
(elf_x86_64_finish_dynamic_symbol ): Likewise.
(elf_x86_64_finish_dynamic_sections): Likewise.
(elf_x86_64_output_arch_local_syms): Likewise.
(elf_x86_64_get_synthetic_symtab): Likewise.
(elf_x86_64_link_setup_gnu_properties): Likewise.
(elf_x86_64_link_hash_table): Use elf_x86_link_hash_table,
instead of elf_link_hash_table, as base.
(elf_x86_64_next_tls_desc_index): New.
(elf_x86_64_srelplt2): Likewise.
(elf_x86_64_plt): Likewise.
(elf_x86_64_lazy_plt): Likewise.
(elf_x86_64_non_lazy_plt): Likewise.
* elfxx-x86.c: New file.
* elfxx-x86.h: Likewise.

x86: Simplify bad return in get_synthetic_symtab

* elf32-i386.c (elf_i386_get_synthetic_symtab): Simplify bad
return.
* elf64-x86-64.c (elf_x86_64_get_synthetic_symtab): Likewise.

x86: Check valid PLT sections before checking dynamic relocations

Update x86 get_synthetic_symtab to check valid PLT sections before
checking dynamic relocations and free invalid PLT section contents.

* elf32-i386.c (elf_i386_get_synthetic_symtab): Check valid PLT
sections before checking dynamic relocations and free invalid
PLT section contents.
* elf64-x86-64.c (elf_x86_64_get_synthetic_symtab): Likewise.

Improve MSP430 section placement.

ld * emultempl/msp430.em (change_output_section): New function.
(move_prefixed_section): New function.
(add_region_prefix): New function.
(msp430_elf_after_open): New function.
(gld${EMULATION_NAME}_add_options): Implement.
(gld${EMULATION_NAME}_list_options): Implement.
(gld${EMULATION_NAME}_handle_option): Implement.
* ld.texinfo: Document new options.
* testsuite/ld-msp430-elf/main-bss-lower.d: New.
* testsuite/ld-msp430-elf/main-bss-upper.d: New.
* testsuite/ld-msp430-elf/main-const-lower.d: New.
* testsuite/ld-msp430-elf/main-const-upper.d: New.
* testsuite/ld-msp430-elf/main-text-lower.d: New.
* testsuite/ld-msp430-elf/main-text-upper.d: New.
* testsuite/ld-msp430-elf/main-var-lower.d: New.
* testsuite/ld-msp430-elf/main-var-upper.d: New.
* testsuite/ld-msp430-elf/main-with-data-bss-unique-sec.s: New.
* testsuite/ld-msp430-elf/main-with-data-bss.s: New.
* testsuite/ld-msp430-elf/main-with-text-rodata-unique-sec.s: New.
* testsuite/ld-msp430-elf/main-with-text-rodata.s: New.
* testsuite/ld-msp430-elf/msp430-elf.exp: New.
* testsuite/ld-msp430-elf/msp430-no-lower.ld: New.
* testsuite/ld-msp430-elf/msp430.ld: New.
* emultempl/msp430.em (data_statement_size): New.
(eval_upper_either_sections): New.
(eval_lower_either_sections): New.
(intermediate_relax_sections): New.
(msp430_elf_after_allocation): New.
* emultempl/msp430.em (gld${EMULATION_NAME}_place_orphan): Always
place sections in the lower region.

gas * config/tc-msp430.c (md_parse_option): Define high data and high
bss symbols if -mdata-region is passed.
Define -mdata-region open.
* doc/c-msp430.texi: Document -mdata-region.
* testsuite/gas/msp430/high-data-bss-sym.d: New test.
* testsuite/gas/msp430/high-data-bss-sym.s: New.
* testsuite/gas/msp430/msp430.exp: Add -mdata-region tests.

Fix simulation of MSP430's open system call.

* sim/msp430/msp430-sim.c (maybe_perform_syscall): Fix passing of
arguments for variadic syscall "open".

[GOLD] PowerPC tls_get_addr_optimize

This implements the special __tls_get_addr_opt call stub for powerpc
gold that returns __thread variable addresses without actually making
a call to __tls_get_addr in most cases.  Shared libraries that are
loaded at program load time (ie. dlopen is not used) have a known
layout for their __thread variables, and thus DTPMOD64/DPTREL64 pairs
describing those variables can be set up by ld.so for the
__tls_get_addr_opt call stub fast exit.
Ref https://sourceware.org/ml/libc-alpha/2015-03/msg00626.html

I really, really wish I'd used a differently versioned __tls_get_addr
symbol than the base symbol to indicate glibc support for the
optimized call, rather than having glibc export __tls_get_addr_opt.  A
lot of the messing around here, flipping symbols from __tls_get_addr
to __tls_get_addr_opt, is caused by that decision.  About the only
benefit is that a user can see at a glance that their disassembled
code is calling __tls_get_addr via the fancy call stub..  Anyway, we
need references to __tls_get_addr to seem like they were to
__tls_get_addr_opt, and in cases like the tsan interceptor, a
definition of __tls_get_addr to seem like one of __tls_get_addr_opt
as well.  That's the reason for Symbol::clear_in_reg and
Symbol_table::clone, and why symbols are substituted in Scan::global
and other places dealing with dynamic linking.

elfcpp/
* elfcpp.h (DT_PPC_OPT): Define.
* powerpc.h (PPC_OPT_TLS): Define.
gold/
* options.h (tls_get_addr_optimize): New option.
* symtab.h (Symbol::clear_in_reg, clone): New functions.
(Sized_symbol::clone): New function.
(Symbol_table::clone): New function.
* resolve.cc (Symbol::clone, Sized_symbol::clone): New functions.
* powerpc.cc (Target_powerpc::has_tls_get_addr_opt_,
tls_get_addr_, tls_get_addr_opt_): New vars.
(Target_powerpc::tls_get_addr_opt, tls_get_addr,
is_tls_get_addr_opt, replace_tls_get_addr,
set_has_tls_get_addr_opt, stk_linker): New functions.
(Target_powerpc::Track_tls::maybe_skip_tls_get_addr_call): Add
target param.  Update callers.  Compare symbols rather than names.
(Target_powerpc::do_define_standard_symbols): Init tls_get_addr_
and tls_get_addr_opt_.
(Target_powerpc::Branch_info::mark_pltcall): Translate tls_get_addr
sym to tls_get_addr_opt.
(Target_powerpc::Branch_info::make_stub): Likewise.
(Stub_table::define_stub_syms): Likewise.
(Target_powerpc::Scan::global): Likewise.
(Target_powerpc::Relocate::relocate): Likewise.
(add_3_12_2, add_3_12_13, bctrl, beqlr, cmpdi_11_0, cmpwi_11_0,
ld_11_1, ld_11_3, ld_12_3, lwz_11_3, lwz_12_3, mr_0_3, mr_3_0,
mtlr_11, std_11_1): New constants.
(Stub_table::eh_frame_added_): Delete.
(Stub_table::tls_get_addr_opt_bctrl_, plt_fde_len_, plt_fde_): New vars.
(Stub_table::init_plt_fde): New functions.
(Stub_table::add_eh_frame, replace_eh_frame): Move definition out
of line.  Init and use plt_fde_.
(Stub_table::plt_call_size): Return size for tls_get_addr stub.
Extract alignment code to..
(Stub_table::plt_call_align): ..this new function.  Adjust all callers.
(Stub_table::add_plt_call_entry): Set has_tls_get_addr_opt and
tls_get_addr_opt_bctrl, and align after that.
(Stub_table::do_write): Write out tls_get_addr stub.
(Target_powerpc::do_finalize_sections): Emit DT_PPC_OPT
PPC_OPT_TLS/PPC64_OPT_TLS bit.
(Target_powerpc::Relocate::relocate): Don't check for or modify
nop following bl for tls_get_addr stub.

[GOLD] Reduce size of class Symbol

On 64-bit targets there is a 32-bit hole in symbol->u_, and another
due to symbol flags exceeding 32 bits.  By splitting the union,
the total size of the class reduces by one 64-bit word.

* symtab.h (Symbol): Split u_ into u1_ and u2_.  Adjust accessors
to suit.  Move plt_offset_ before got_offsets_.
* symtab.cc (Symbol::init_fields): Adjust for union change.
(Symbol::init_base_output_data): Likewise.
(Symbol::init_base_output_segment): Likewise.
(Symbol::allocate_base_common): Likewise.
(Symbol::output_section): Likewise.
(Symbol::set_output_section): Likewise.
(Symbol::set_output_segment): Likewise.
* resolve.cc (Symbol::override_base): Likewise.
(Symbol::override_base_with_special): Likewise.

Automatic date update in version.in

2017-08-28  James Clarke <jrtc27@jrtc27.com>

gold/ChangeLog:
PR gold/21868
* aarch64.cc (AArch64_relobj::try_fix_erratum_843419_optimized):
Add extra view offset argument to function.
(AArch64_relobj::fix_errata_and_relocate_erratum_stubs): Add
extra view offset set to the output offset when the view has
is_input_output_view set, since it has not already been
included. Pass this to try_fix_erratum_843419_optimized.

Remove unused _bfd_elf_ifunc_get_synthetic_symtab

Since _bfd_elf_ifunc_get_synthetic_symtab is no longer called by x86
get_synthetic_symtab, it is now unused.  This patch removes it.

* elf-bfd.h (_bfd_elf_ifunc_get_synthetic_symtab): Removed.
* elf-ifunc.c (_bfd_elf_ifunc_get_synthetic_symtab): Likewise.

gdb.base/commands.exp: Remove unused global references

There are a few unused references to the gdb_prompt global.

gdb/testsuite/ChangeLog:

* gdb.base/commands.exp (gdbvar_simple_if_test,
gdbvar_simple_if_test, gdbvar_complex_if_while_test,
progvar_simple_if_test, progvar_simple_while_test,
progvar_complex_if_while_test, user_defined_command_test,
user_defined_command_args_eval,
user_defined_command_args_stack_test,
user_defined_command_manyargs_test, bp_deleted_in_command_test,
temporary_breakpoint_commands,
gdb_test_no_prompt, redefine_hook_test,
redefine_backtrace_test): Remove "global gdb_prompt".

Add missing PR number in ChangeLog

This should have been included in the previous commit.

define_command: Don't convert command name to lower case

Commit

  Command names: make them case sensitive
  3d7b173c29900879c9a5958dd6029fd36666e57c

made command name lookup case sensitive.  However, define_command, used
when creating a user-defined command, converts the command name to
lowercase, assuming that the command name lookup works in a case
insensitive way.  This causes user-defined commands with capital letters
in their name to only be callable with a lowercase version:

  (gdb) define Foo
  Type commands for definition of "Foo".
  End with a line saying just "end".
  >print 1
  >end
  (gdb) Foo
  Undefined command: "Foo".  Try "help".
  (gdb) foo
  $1 = 1

This patch removes that conversion to lowercase, so that the user can
call the command with the same name they provided.

gdb/ChangeLog:

* cli/cli-script.c (define_command): Don't convert command name
to lower case.

gdb/testsuite/ChangeLog:

* gdb.base/commands.exp (user_defined_command_case_sensitivity):
New proc, call it from toplevel.

x86: Check for valid PLT section size

Update x86 get_synthetic_symtab to check for valid PLT section size
before examining PLT section contents.

PR binutils/22018
* elf32-i386.c (elf_i386_get_synthetic_symtab): Check for valid
PLT section size.
* elf64-x86-64.c (elf_x86_64_get_synthetic_symtab): Likewise.

Fix bad offset calculation for R_AARCH64_TLSDESC_* relocs.

If a custom linker script with an unexpected relative layout of .got
and .got.plt sections was used, gold might produce a wrong offset
when applying R_AARCH64_TLSDESC_* relocations.
This patch fixes the issue by calculating "got_tlsdesc_offset"
in a more direct way.

gold/
* aarch64.cc (Target_aarch64::Relocate::relocate_tls):
Make got_tlsdesc_offset signed and fix its calculation.
* testsuite/Makefile.am (aarch64_tlsdesc): New test.
* testsuite/Makefile.in: Regenerate.
* testsuite/aarch64_tlsdesc.s: New test source file.
* testsuite/aarch64_tlsdesc.sh: New test script.
* testsuite/aarch64_tlsdesc.t: New test linker script.

[GOLD] Symbol flag for PowerPC64 localentry:0 tracking

This patch provides a flag for PowerPC64 ELFv2 use in class Symbol,
and modifies Sized_target::resolve to return whether the symbol has
been resolved.  If not, normal processing continues.  I use this for
PowerPC64 ELFv2 to keep track of whether a symbol has any definition
with non-zero localentry, in order to disable --plt-localentry for
that symbol.

PR 21847
* powerpc.cc (Target_powerpc::is_elfv2_localentry0): Test
non_zero_localentry.
(Target_powerpc::resolve): New function.
(powerpc_info): Set has_resolve for 64-bit.
* target.h (Sized_target::resolve): Return bool.
* resolve.cc (Symbol_table::resolve): Continue with normal
processing when target resolve returns false.
* symtab.h (Symbol::non_zero_localentry, set_non_zero_localentry):
New accessors.
(Symbol::non_zero_localentry_): New flag bit.
* symtab.cc (Symbol::init_fields): Init non_zero_localentry_.

Automatic date update in version.in

x86: Initialize def_protected

* elf32-i386.c (elf_i386_link_hash_newfunc): Initialize
def_protected.
* elf64-x86-64.c (elf_x86_64_link_hash_newfunc): Likewise.

Disallow copy relocation against protected data symbol

We shpouldn't generate copy relocation to resolve reference to protected
data symbol defined in shared object with the NO_COPY_ON_PROTECTED
property.  This patch adds a bit to elf_obj_tdata as well as
elf_i386_link_hash_entry and elf_x86_64_link_hash_entry to track the bfd
with the NO_COPY_ON_PROTECTED property as well as protected symbol
defined in shared object.  extern_protected_data is set to FALSE if any
input relocatable file contains the NO_COPY_ON_PROTECTED property.

bfd/

PR ld/21997
* elf-bfd.h (elf_obj_tdata): Use ENUM_BITFIELD on object_id,
dyn_lib_class and has_gnu_symbols.  Change bad_symtab to bitfield.
Add a has_no_copy_on_protected bitfield.
(elf_has_no_copy_on_protected): New.
* elf-properties.c (_bfd_elf_parse_gnu_properties): Set
elf_has_no_copy_on_protected for GNU_PROPERTY_NO_COPY_ON_PROTECTED.
(elf_merge_gnu_property_list): Likewise.
(_bfd_elf_link_setup_gnu_properties): Set extern_protected_data
to FALSE for elf_has_no_copy_on_protected.
* elf32-i386.c (SYMBOL_NO_COPYRELOC): New.
(elf_i386_link_hash_entry): Add def_protected.
(elf_i386_adjust_dynamic_symbol): Also check SYMBOL_NO_COPYRELOC
when checking info->nocopyreloc.
(elf_i386_link_setup_gnu_properties): Don't set
extern_protected_data here.
(elf_i386_merge_symbol_attribute): New function.
(elf_backend_merge_symbol_attribute): New.
* elf64-x86-64.c (SYMBOL_NO_COPYRELOC): New.
(elf_x86_64_link_hash_entry): Add def_protected.
(elf_x86_64_need_pic): Report protected symbol for def_protected.
(elf_x86_64_adjust_dynamic_symbol): Also check SYMBOL_NO_COPYRELOC
when checking info->nocopyreloc.
(elf_x86_64_relocate_section): Also check for R_X86_64_PC32
relocation run-time overflow and unresolvable R_X86_64_32S
relocation against protected data symbol defined in shared object
with GNU_PROPERTY_NO_COPY_ON_PROTECTED.
(elf_x86_64_link_setup_gnu_properties): Don't set
extern_protected_data here.
(elf_x86_64_merge_symbol_attribute): New function.
(elf_backend_merge_symbol_attribute): New.

ld/

PR ld/21997
* testsuite/ld-i386/i386.exp: Run PR ld/21997 tests.
* testsuite/ld-x86-64/x86-64.exp: Likewise.
* testsuite/ld-i386/pr21997-1a.S: New file.
* testsuite/ld-i386/pr21997-1b.c: Likewise.
* testsuite/ld-i386/pr21997-1c.S: Likewise.
* testsuite/ld-x86-64/pr21997-1a.S: Likewise.
* testsuite/ld-x86-64/pr21997-1a.err: Likewise.
* testsuite/ld-x86-64/pr21997-1b.c: Likewise.
* testsuite/ld-x86-64/pr21997-1b.err: Likewise.
* testsuite/ld-x86-64/pr21997-1c.c: Likewise.

Automatic date update in version.in

PowerPC TPREL reloc handling

Tidy how these are handled in PIEs.

* elf32-ppc.c (must_be_dyn_reloc): Use bfd_link_dll.  Comment.
(ppc_elf_check_relocs): Only set DF_STATIC_TLS in shared libs.
(ppc_elf_relocate_section): Comment fix.
* elf64-ppc.c (must_be_dyn_reloc): Use bfd_link_dll.  Comment.
(ppc64_elf_check_relocs): Only set DF_STATIC_TLS in shared libs.
Support dynamic relocs for TPREL16 when non-pic too.
(dec_dynrel_count): Adjust TPREL16 handling as per check_relocs.
(ppc64_elf_relocate_section): Support dynamic relocs for TPREL16
when non-pic too.

remove param "dispp" from ada-lang.c::ada_lookup_struct_elt_type

The function is always called with DISPP set to NULL, so there is
no need for this parameter anymore. This patch removes it, and
eliminates some dead code associated to that.

gdb/ChangeLog:

        * ada-lang.c (ada_lookup_struct_elt_type): Remove parameter "dispp".
        Update all callers accordingly. Remove all code blocks handling
        the case where DISPP is not NULL.

Tested on x86_64-linux, no regression.