binutils-gdb.git
7 years agoPass $NOPIE_CFLAGS and $NOPIE_LDFLAGS to more ELF tests
H.J. Lu [Thu, 22 Jun 2017 16:53:33 +0000 (09:53 -0700)]
Pass $NOPIE_CFLAGS and $NOPIE_LDFLAGS to more ELF tests

PR ld/21090
* testsuite/ld-gc/gc.ex: Compile tmpdir/pr14265.o with
$NOPIE_CFLAGS.
* testsuite/ld-i386/i386.exp: Pass $NOPIE_CFLAGS and
$NOPIE_LDFLAGS if non-PIE is required.
* testsuite/ld-i386/no-plt.exp (NOPIE_CFLAGS): New.
(NOPIE_LDFLAGS): Likewise.
Pass $NOPIE_LDFLAGS if non-PIE is required.
* testsuite/ld-shared/shared.exp: Compile tmpdir/sh1np.o with
$NOPIE_CFLAGS.

7 years agoFix cached_frame allocation in py-unwind
Alan Hayward [Thu, 22 Jun 2017 15:30:15 +0000 (16:30 +0100)]
Fix cached_frame allocation in py-unwind

gdb/
* python/py-unwind.c (pyuw_sniffer): Allocate space for
registers.

7 years agoRemove an instance of MAX_REGISTER_SIZE from record-full.c
Alan Hayward [Thu, 22 Jun 2017 14:33:18 +0000 (15:33 +0100)]
Remove an instance of MAX_REGISTER_SIZE from record-full.c

gdb/
* record-full.c (record_full_exec_insn): Use byte_vector.

7 years agoRegenerate two regformats/i386/.dat files
Yao Qi [Thu, 22 Jun 2017 12:41:27 +0000 (13:41 +0100)]
Regenerate two regformats/i386/.dat files

The self tests which compare pre-generated target descriptions and
dynamically created target descriptions fail, and it turns out that two
pre-generated target descriptions are wrong, so regenerate them.

gdb:

2017-06-22  Yao Qi  <yao.qi@linaro.org>

* regformats/i386/amd64-avx-mpx-avx512-pku-linux.dat: Regenerated.
* regformats/i386/amd64-avx-mpx-avx512-pku.dat: Regenerated.

7 years agoRemove MAX_REGISTER_SIZE from py-unwind.c
Alan Hayward [Thu, 22 Jun 2017 13:09:52 +0000 (14:09 +0100)]
Remove MAX_REGISTER_SIZE from py-unwind.c

gdb/
* remote.c (cached_reg): Move from here...
* regcache.h (cached_reg): ...to here.
* python/py-unwind.c (struct reg_info): Remove.
(cached_frame_info): Use cached_reg_t.
(pyuw_prev_register): Likewise.
(pyuw_sniffer): Use cached_reg_t and allocate registers.
(pyuw_dealloc_cache): Free all registers.

7 years agox86: Support Intel Shadow Stack with SHSTK property
H.J. Lu [Thu, 22 Jun 2017 12:50:20 +0000 (05:50 -0700)]
x86: Support Intel Shadow Stack with SHSTK property

To support Intel Shadow Stack (SHSTK) in Intel Control-flow Enforcement
Technology (CET) instructions:

https://software.intel.com/sites/default/files/managed/4d/2a/control-flow-enforcement-technology-preview.pdf

 #define GNU_PROPERTY_X86_FEATURE_1_SHSTK (1U << 1)

is added to GNU program properties to indicate that all executable sections
are compatible with SHSTK where return address popped from shadow stack
always matches return address popped from normal stack.

GNU_PROPERTY_X86_FEATURE_1_SHSTK is set on output only if it is set on all
relocatable inputs.

bfd/

* elf32-i386.c (elf_i386_merge_gnu_properties): If info->shstk
is set, turn on GNU_PROPERTY_X86_FEATURE_1_SHSTK.
(elf_i386_link_setup_gnu_properties): If info->shstk is set,
turn on GNU_PROPERTY_X86_FEATURE_1_IBT.
* elf64-x86-64.c (elf_x86_64_merge_gnu_properties): If
info->shstk is set, turn on GNU_PROPERTY_X86_FEATURE_1_SHSTK.
(elf_x86_64_link_setup_gnu_properties): If info->shstk is set,
turn on GNU_PROPERTY_X86_FEATURE_1_IBT.

binutils/

* readelf.c (decode_x86_feature): Decode
GNU_PROPERTY_X86_FEATURE_1_SHSTK.
* testsuite/binutils-all/i386/shstk.d: New file.
* testsuite/binutils-all/i386/shstk.s: Likewise.
* testsuite/binutils-all/x86-64/shstk-x32.d: Likewise.
* testsuite/binutils-all/x86-64/shstk.d: Likewise.
* testsuite/binutils-all/x86-64/shstk.s: Likewise.

include/

* bfdlink.h (bfd_link_info): Add shstk.
* elf/common.h (GNU_PROPERTY_X86_FEATURE_1_SHSTK): New.

ld/

* NEWS: Mention -z shstk and GNU_PROPERTY_X86_FEATURE_1_SHSTK.
* emulparams/cet.sh (PARSE_AND_LIST_OPTIONS_CET): Add "-z shstk".
(PARSE_AND_LIST_ARGS_CASE_Z_CET): Support "-z shstk".
* ld.texinfo: Document -z shstk.
* testsuite/ld-i386/i386.exp: Run SHSTK tests.
* testsuite/ld-x86-64/x86-64.exp: Likewise.
* testsuite/ld-i386/property-x86-shstk.s: New file.
* testsuite/ld-i386/property-x86-shstk1a.d: Likewise.
* testsuite/ld-i386/property-x86-shstk1b.d: Likewise.
* testsuite/ld-i386/property-x86-shstk2.d: Likewise.
* testsuite/ld-i386/property-x86-shstk3a.d: Likewise.
* testsuite/ld-i386/property-x86-shstk3b.d: Likewise.
* testsuite/ld-i386/property-x86-shstk4.d: Likewise.
* testsuite/ld-i386/property-x86-shstk5.d: Likewise.
* testsuite/ld-x86-64/property-x86-shstk.s: Likewise.
* testsuite/ld-x86-64/property-x86-shstk1a-x32.d: Likewise.
* testsuite/ld-x86-64/property-x86-shstk1a.d: Likewise.
* testsuite/ld-x86-64/property-x86-shstk1b-x32.d: Likewise.
* testsuite/ld-x86-64/property-x86-shstk1b.d: Likewise.
* testsuite/ld-x86-64/property-x86-shstk2-x32.d: Likewise.
* testsuite/ld-x86-64/property-x86-shstk2.d: Likewise.
* testsuite/ld-x86-64/property-x86-shstk3a-x32.d: Likewise.
* testsuite/ld-x86-64/property-x86-shstk3a.d: Likewise.
* testsuite/ld-x86-64/property-x86-shstk3b-x32.d: Likewise.
* testsuite/ld-x86-64/property-x86-shstk3b.d: Likewise.
* testsuite/ld-x86-64/property-x86-shstk4-x32.d: Likewise.
* testsuite/ld-x86-64/property-x86-shstk4.d: Likewise.
* testsuite/ld-x86-64/property-x86-shstk5-x32.d: Likewise.
* testsuite/ld-x86-64/property-x86-shstk5.d: Likewise.

7 years agox86: Support Intel IBT with IBT property and IBT-enable PLT
H.J. Lu [Thu, 22 Jun 2017 12:44:37 +0000 (05:44 -0700)]
x86: Support Intel IBT with IBT property and IBT-enable PLT

To support IBT in Intel Control-flow Enforcement Technology (CET)
instructions:

https://software.intel.com/sites/default/files/managed/4d/2a/control-flow-enforcement-technology-preview.pdf

 #define GNU_PROPERTY_X86_FEATURE_1_AND 0xc0000002

 #define GNU_PROPERTY_X86_FEATURE_1_IBT (1U << 0)

are added to GNU program properties to indicate that all executable
sections are compatible with IBT when ENDBR instruction starts each
valid target where an indirect branch instruction can land.

GNU_PROPERTY_X86_FEATURE_1_IBT is set on output only if it is set on
all relocatable inputs.

The followings changes are made to the Procedure Linkage Table (PLT):

1. For 64-bit x86-64,  PLT is changed to

PLT0:  push       GOT[1]
       bnd jmp    *GOT[2]
       nop
...
PLTn:  endbr64
       push       namen_reloc_index
       bnd jmp    PLT0

together with the second PLT section:

PLTn:  endbr64
       bnd jmp   *GOT[namen_index]
       nop

BND prefix is also added so that IBT-enabled PLT is compatible with MPX.

2. For 32-bit x86-64 (x32) and i386,  PLT is changed to

PLT0:  push       GOT[1]
       jmp        *GOT[2]
       nop
...
PLTn:  endbr64                                 # endbr32 for i386.
       push       namen_reloc_index
       jmp        PLT0

together with the second PLT section:

PLTn:  endbr64                                 # endbr32 for i386.
       jmp       *GOT[namen_index]
       nop

BND prefix isn't used since MPX isn't supported on x32 and BND registers
aren't used in parameter passing on i386.

GOT is an array of addresses.  Initially, GOT[namen_index] is filled
with the address of the ENDBR instruction of the corresponding entry
in the first PLT section.  The function, namen, is called via the
ENDBR instruction in the second PLT entry.  GOT[namen_index] is updated
to the actual address of the function, namen, at run-time.

2 linker command line options are added:

1. -z ibtplt: Generate IBT-enabled PLT.
2. -z ibt: Generate GNU_PROPERTY_X86_FEATURE_1_IBT in GNU program
properties as well as IBT-enabled PLT.

bfd/

* elf32-i386.c (elf_i386_lazy_ibt_plt0_entry): New.
(elf_i386_lazy_ibt_plt_entry): Likewise.
(elf_i386_pic_lazy_ibt_plt0_entry): Likewise.
(elf_i386_non_lazy_ibt_plt_entry): Likewise.
(elf_i386_pic_non_lazy_ibt_plt_entry): Likewise.
(elf_i386_eh_frame_lazy_ibt_plt): Likewise.
(elf_i386_lazy_plt_layout): Likewise.
(elf_i386_non_lazy_plt_layout): Likewise.
(elf_i386_link_hash_entry): Add plt_second.
(elf_i386_link_hash_table): Add plt_second and
plt_second_eh_frame.
(elf_i386_allocate_dynrelocs): Use the second PLT if needed.
(elf_i386_size_dynamic_sections): Use .plt.got unwind info for
the second PLT.  Check the second PLT.
(elf_i386_relocate_section): Use the second PLT to resolve
PLT reference if needed.
(elf_i386_finish_dynamic_symbol): Fill and use the second PLT if
needed.
(elf_i386_finish_dynamic_sections): Set sh_entsize on the
second PLT.  Generate unwind info for the second PLT.
(elf_i386_plt_type): Add plt_second.
(elf_i386_get_synthetic_symtab): Support the second PLT.
(elf_i386_parse_gnu_properties): Support
GNU_PROPERTY_X86_FEATURE_1_AND.
(elf_i386_merge_gnu_properties): Support
GNU_PROPERTY_X86_FEATURE_1_AND.  If info->ibt is set, turn
on GNU_PROPERTY_X86_FEATURE_1_IBT
(elf_i386_link_setup_gnu_properties): If info->ibt is set,
turn on GNU_PROPERTY_X86_FEATURE_1_IBT.  Use IBT-enabled PLT
for info->ibtplt, info->ibt or GNU_PROPERTY_X86_FEATURE_1_IBT
is set on all relocatable inputs.
* elf64-x86-64.c (elf_x86_64_lazy_ibt_plt_entry): New.
(elf_x32_lazy_ibt_plt_entry): Likewise.
(elf_x86_64_non_lazy_ibt_plt_entry): Likewise.
(elf_x32_non_lazy_ibt_plt_entry): Likewise.
(elf_x86_64_eh_frame_lazy_ibt_plt): Likewise.
(elf_x32_eh_frame_lazy_ibt_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_get_synthetic_symtab): Support the second PLT.
(elf_x86_64_parse_gnu_properties): Support
GNU_PROPERTY_X86_FEATURE_1_AND.
(elf_x86_64_merge_gnu_properties): Support
GNU_PROPERTY_X86_FEATURE_1_AND.  If info->ibt is set, turn
on GNU_PROPERTY_X86_FEATURE_1_IBT
(elf_x86_64_link_setup_gnu_properties): If info->ibt is set,
turn on GNU_PROPERTY_X86_FEATURE_1_IBT.  Use IBT-enabled PLT
for info->ibtplt, info->ibt or GNU_PROPERTY_X86_FEATURE_1_IBT
is set on all relocatable inputs.

binutils/

* readelf.c (decode_x86_feature): New.
(print_gnu_property_note): Call decode_x86_feature on
GNU_PROPERTY_X86_FEATURE_1_AND.
* testsuite/binutils-all/i386/empty.d: New file.
* testsuite/binutils-all/i386/empty.s: Likewise.
* testsuite/binutils-all/i386/ibt.d: Likewise.
* testsuite/binutils-all/i386/ibt.s: Likewise.
* testsuite/binutils-all/x86-64/empty-x32.d: Likewise.
* testsuite/binutils-all/x86-64/empty.d: Likewise.
* testsuite/binutils-all/x86-64/empty.s: Likewise.
* testsuite/binutils-all/x86-64/ibt-x32.d: Likewise.
* testsuite/binutils-all/x86-64/ibt.d: Likewise.
* testsuite/binutils-all/x86-64/ibt.s: Likewise.

include/

* bfdlink.h (bfd_link_info): Add ibtplt and ibt.
* elf/common.h (GNU_PROPERTY_X86_FEATURE_1_AND): New.
(GNU_PROPERTY_X86_FEATURE_1_IBT): Likewise.

ld/

* Makefile.am (ELF_X86_DEPS): Add $(srcdir)/emulparams/cet.sh.
* Makefile.in: Regenerated.
* NEWS: Mention GNU_PROPERTY_X86_FEATURE_1_IBT, -z ibtplt
and -z ibt.
* emulparams/cet.sh: New file.
* testsuite/ld-i386/ibt-plt-1.d: Likewise.
* testsuite/ld-i386/ibt-plt-1.s: Likewise.
* testsuite/ld-i386/ibt-plt-2.s: Likewise.
* testsuite/ld-i386/ibt-plt-2a.d: Likewise.
* testsuite/ld-i386/ibt-plt-2b.d: Likewise.
* testsuite/ld-i386/ibt-plt-2c.d: Likewise.
* testsuite/ld-i386/ibt-plt-2d.d: Likewise.
* testsuite/ld-i386/ibt-plt-3.s: Likewise.
* testsuite/ld-i386/ibt-plt-3a.d: Likewise.
* testsuite/ld-i386/ibt-plt-3b.d: Likewise.
* testsuite/ld-i386/ibt-plt-3c.d: Likewise.
* testsuite/ld-i386/ibt-plt-3d.d: Likewise.
* testsuite/ld-i386/plt-main-ibt.dd: Likewise.
* testsuite/ld-i386/plt-pie-ibt.dd: Likewise.
* testsuite/ld-i386/property-x86-empty.s: Likewise.
* testsuite/ld-i386/property-x86-ibt.s: Likewise.
* testsuite/ld-i386/property-x86-ibt1a.d: Likewise.
* testsuite/ld-i386/property-x86-ibt1b.d: Likewise.
* testsuite/ld-i386/property-x86-ibt2.d: Likewise.
* testsuite/ld-i386/property-x86-ibt3a.d: Likewise.
* testsuite/ld-i386/property-x86-ibt3b.d: Likewise.
* testsuite/ld-i386/property-x86-ibt4.d: Likewise.
* testsuite/ld-i386/property-x86-ibt5.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-1-x32.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-1.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-1.s: Likewise.
* testsuite/ld-x86-64/ibt-plt-2.s: Likewise.
* testsuite/ld-x86-64/ibt-plt-2a-x32.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-2a.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-2c-x32.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-2c.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-3.s: Likewise.
* testsuite/ld-x86-64/ibt-plt-3a-x32.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-3a.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-3c-x32.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-3c.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-3d-x32.d: Likewise.
* testsuite/ld-x86-64/ibt-plt-3d.d: Likewise.
* testsuite/ld-x86-64/plt-main-ibt-now.rd: Likewise.
* testsuite/ld-x86-64/plt-main-ibt-x32.dd: Likewise.
* testsuite/ld-x86-64/plt-main-ibt.dd: Likewise.
* testsuite/ld-x86-64/property-x86-empty.s: Likewise.
* testsuite/ld-x86-64/property-x86-ibt.s: Likewise.
* testsuite/ld-x86-64/property-x86-ibt1a-x32.d: Likewise.
* testsuite/ld-x86-64/property-x86-ibt1a.d: Likewise.
* testsuite/ld-x86-64/property-x86-ibt1b-x32.d: Likewise.
* testsuite/ld-x86-64/property-x86-ibt1b.d: Likewise.
* testsuite/ld-x86-64/property-x86-ibt2-x32.d: Likewise.
* testsuite/ld-x86-64/property-x86-ibt2.d: Likewise.
* testsuite/ld-x86-64/property-x86-ibt3a-x32.d: Likewise.
* testsuite/ld-x86-64/property-x86-ibt3a.d: Likewise.
* testsuite/ld-x86-64/property-x86-ibt3b-x32.d: Likewise.
* testsuite/ld-x86-64/property-x86-ibt3b.d: Likewise.
* testsuite/ld-x86-64/property-x86-ibt4-x32.d: Likewise.
* testsuite/ld-x86-64/property-x86-ibt4.d: Likewise.
* testsuite/ld-x86-64/property-x86-ibt5-x32.d: Likewise.
* testsuite/ld-x86-64/property-x86-ibt5.d: Likewise.
* emulparams/elf32_x86_64.sh: Source emulparams/cet.sh.
(TINY_READONLY_SECTION): Add .plt.sec.
* emulparams/elf_i386.sh: Likewise.
* emulparams/elf_x86_64.sh: Source emulparams/cet.sh.
* ld.texinfo: Document -z ibtplt and -z ibt.
* testsuite/ld-i386/i386.exp: Run IBT and IBT PLT tests.
* testsuite/ld-x86-64/x86-64.exp: Likewise.
* testsuite/ld-x86-64/pr21481b.S (check): Updated for x32.

7 years agoenviron-selftests: Ignore -Wself-move warning
Pedro Alves [Thu, 22 Jun 2017 08:31:20 +0000 (10:31 +0200)]
environ-selftests: Ignore -Wself-move warning

clang gives this warning:

 ..../gdb/unittests/environ-selftests.c:139:7: error: explicitly moving variable of type 'gdb_environ' to itself [-Werror,-Wself-move]
   env = std::move (env);
   ~~~ ^            ~~~

Ignoring the warning locally is the right thing to do, since it warns
about behavior we want to unit test, while an explicit self-move in
real code would likely be a mistake that we'd want to catch.

To avoid cluttering the code with preprocessor conditionals, this
commit adds the file common/diagnostics.h, in which we can put macros
used to control compiler diagnostics.

GCC enhancement request here:
  https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81159

gdb/ChangeLog:
2017-06-22  Pedro Alves  <palves@redhat.com>
    Simon Marchi  <simon.marchi@ericsson.com>

* unittests/environ-selftests.c (run_tests): Ignore -Wself-move
warning.
* common/diagnostics.h: New file.

7 years agoAdd STRINGIFY to gdb/common/preprocessor.h
Pedro Alves [Thu, 22 Jun 2017 09:23:30 +0000 (10:23 +0100)]
Add STRINGIFY to gdb/common/preprocessor.h

We have several copies of this common idiom under gdb/ currently.
This commit moves them / factors them out to gdb/common/preprocessor.h.

gdb/ChangeLog:
2017-06-22  Pedro Alves  <palves@redhat.com>

* common/agent.h: Include "common/preprocessor.h".
(STRINGIZE_1, STRINGIZE): Delete.
(IPA_SYM): Use STRINGIFY instead.
* common/preprocessor.h (STRINGIFY_1, STRINGIFY): New.
* compile/compile-c-support.c: Include "common/preprocessor.h".
(STR, STRINGIFY): Delete.
* ia64-libunwind-tdep.c: Include "common/preprocessor.h".
(STRINGIFY2, STRINGIFY): Delete.

7 years agocommon/agent.h: Add missing include guards
Pedro Alves [Thu, 22 Jun 2017 09:57:13 +0000 (10:57 +0100)]
common/agent.h: Add missing include guards

gdb/ChangeLog:
2017-06-22  Pedro Alves  <palves@redhat.com>

* common/agent.h: Add include guards.

7 years agoFix address violation parsing a corrupt SOM binary.
Nick Clifton [Thu, 22 Jun 2017 09:33:56 +0000 (10:33 +0100)]
Fix address violation parsing a corrupt SOM binary.

PR binutils/21649
* som.c (setup_sections): NUL terminate the space_strings buffer.
Check that the space.name field does not index beyond the end of
the space_strings buffer.

7 years agoFix compile time warning about unused static variable.
Nick Clifton [Thu, 22 Jun 2017 08:34:12 +0000 (09:34 +0100)]
Fix compile time warning about unused static variable.

* config/tc-arm.c (arm_ext_v7m): Add ATTRIBUTE_UNUSED.

7 years agoAutomatic date update in version.in
GDB Administrator [Thu, 22 Jun 2017 00:01:05 +0000 (00:01 +0000)]
Automatic date update in version.in

7 years agoUse DWARF_VMA_FMT to report error
H.J. Lu [Wed, 21 Jun 2017 22:29:38 +0000 (15:29 -0700)]
Use DWARF_VMA_FMT to report error

Use DWARF_VMA_FMT to report error to work for both 32-bit and 64-bit
builds.

* dwarf.c (READ_ULEB): Use DWARF_VMA_FMT to report error.
(READ_SLEB): Likewise.

7 years agoPass $NOPIE_CFLAGS and $NOPIE_LDFLAGS to some ELF tests
H.J. Lu [Wed, 21 Jun 2017 22:22:05 +0000 (15:22 -0700)]
Pass $NOPIE_CFLAGS and $NOPIE_LDFLAGS to some ELF tests

Some ELF tests will fail when PIE is used.

PR ld/21090
* testsuite/ld-elf/shared.exp: Pass $NOPIE_CFLAGS and
$NOPIE_LDFLAGS if non-PIE is required.

7 years agoPass $NOPIE_CFLAGS to NOCROSSREFS tests
H.J. Lu [Wed, 21 Jun 2017 21:57:53 +0000 (14:57 -0700)]
Pass $NOPIE_CFLAGS to NOCROSSREFS tests

PR ld/21090
* testsuite/ld-scripts/crossref.exp: Also pass $NOPIE_CFLAGS
to CC.

7 years agoAdd missing ChangeLog entries
H.J. Lu [Wed, 21 Jun 2017 21:49:30 +0000 (14:49 -0700)]
Add missing ChangeLog entries

7 years agoPass $NOPIE_LDFLAGS size tests
H.J. Lu [Wed, 21 Jun 2017 21:45:16 +0000 (14:45 -0700)]
Pass $NOPIE_LDFLAGS size tests

PR ld/21090
* testsuite/ld-size/size.exp: Pass $NOPIE_LDFLAGS to size-4a,
size-4b, size-5a, size-5b, size-6 and size-8 tests.

7 years agoUse noncapturing subpattern/parens in gdb_test implementation
Kevin Buettner [Sat, 3 Jun 2017 00:58:12 +0000 (17:58 -0700)]
Use noncapturing subpattern/parens in gdb_test implementation

This is the portion of gdb_test which performs the match against
the RE (regular expression) passed to it:

    return [gdb_test_multiple $command $message {
        -re "\[\r\n\]*($pattern)\[\r\n\]+$gdb_prompt $" {
            if ![string match "" $message] then {
                pass "$message"
            }
        }

In a test that I've been working on recently, I wanted to use
a backreference - that's the \1 in the the RE below:

gdb_test "info threads"  \
{.*[\r\n]+\* +([0-9]+) +Thread[^\r\n]* do_something \(n=\1\) at.*}

Put into English, I wanted to make sure that the value of n passed to
do_something() is the same as the thread number shown in the "info
threads" Id column.  (I've structured the test case so that this
*should* be the case.)

It didn't work though.  It turned out that ($pattern) in the RE
noted above is capturing the attempted backreference.  So, in this
case, the backreference does not refer to ([0-9]+) as intended, but
instead refers to ($pattern).  This is wrong because it's not what I
intended, but is also wrong because, if allowed, it could only match a
string of infinite length.

This problem can be fixed by using parens for a "noncapturing
subpattern".  The way that this is done, syntactically, is to use
(?:$pattern) instead of ($pattern).

My research shows that this feature has been present since tcl8.1 which
was released in 1999.

The current tcl version is 8.6 - at least that's what I have on my
machine.  It appears to me that mingw uses some subversion of tcl8.4
which will also have this feature (since 8.4 > 8.1).

So it seems to me that any platform upon which we might wish to test
GDB will have a version of tcl which has this feature.  That being the
case, my hope is that there won't be any objections to its use.

When I looked at the implementation of gdb_test, I wondered whether
the parens were needed at all.  I've concluded that they are.  In the
event that $pattern is an RE which uses alternation at the top level,
e.g. a|b, we need to make $pattern a subpattern (via parens) to limit
the extend of the alternation.  I.e, we don't want the alternation to
extend to the other portions of the RE which gdb_test uses to match
potential blank lines at the beginning of the pattern or the gdb
prompt at the end.

gdb/testsuite/ChangeLog:

* gdb.exp (gdb_test): Using noncapturing parens for the $pattern
subpattern.

7 years agoFix address violation when reading corrupt DWARF data.
Nick Clifton [Wed, 21 Jun 2017 17:05:44 +0000 (18:05 +0100)]
Fix address violation when reading corrupt DWARF data.

PR binutils/21648
* dwarf.c (LEB): Rename to SKIP_ULEB and READ_ULEB.  Add check for
reading a value that is too big for the containing variable.
(SLEB): Rename to SKIP_SLEB and READ_SLEB.  Add similar check.
Replace uses of LEB and SLEB with appropriate new macro.
(display_debug_frames): Use an unsigned int for the 'reg'
variable.  Use a signed long for the 'l' variable.

7 years ago[ARM] Rework Tag_CPU_arch build attribute value selection
Thomas Preud'homme [Wed, 21 Jun 2017 15:32:40 +0000 (16:32 +0100)]
[ARM] Rework Tag_CPU_arch build attribute value selection

=== Context ===

This patch is part of a patch series to add support for ARMv8-R
architecture. Its purpose is to rework the Tag_CPU_arch build attribute
value selection to (i) match architecture or CPU if specified by user
without any need for hack and (ii) match an architecture with all the
features used if in autodetection mode or return an error.

=== Motivation ===

Currently, Tag_CPU_arch build attribute value selection assumes that an
architecture is always a superset of architectures released earlier. As
such, the logic is to browse architectures in chronological order of
release and selecting the Tag_CPU_arch value of the last one to
contribute a feature used[1]/requested[2] not contributed by earlier
architectures.

[1] in case of autodetection mode
[2] otherwise, ie. in case of -mcpu/-march or associated directives

This logic fails the two objectives mentionned in the Context section.
First, due to the assumption it does an architecture can be selected
while not having all the features used/requested which fails the second
objective. Second, not doing an exact match when an architecture or CPU
is selected by the user means the wrong value is chosen when a later
architecture provides a subset of the feature bits of an earlier
architecture. This is the case for the implementation of ARMv8-R (see
later patch).

An added benefit of this patch is that it is possible to easily generate
more consistent build attribute by setting the feature bits from the
architecture matched in aeabi_set_public_attributes in autodetection
mode. This is better done as a separate patch because lots of testcase'
expected results must then be updated accordingly.

=== Patch description ===

The patch changes the main logic for Tag_CPU_arch and
Tag_CPU_arch_profile
values selection to:
- look for an exact match in case an architecture or CPU was specified
  on the command line or in a directive
- select the first released architecture that provides a superset of the
  feature used in the autodetection case
- select the most featureful architecture in case of -march=all
The array cpu_arch_ver is updated to include all architectures in order
to make the first point work.

Note that when looking for an exact match, the architecture with
selected extension is tried first and then only the architecture. This
is because some architectures are exactly equivalent to an earlier
architecture with its extensions selected. ARMv6S-M (= ARMv6-M + OS
extension) and ARMv6KZ (ARMv6K + security extension) are two such
examples.

Other adjustments are also necessary in aeabi_set_public_attributes to
make this change work.

1) The logic to set Tag_ARM_ISA_use and Tag_THUMB_ISA_use must check the
absence of feature bit used/requested to decide whether to give the
default value for empty files (see EABI attribute defaults test). It was
previously checking that arch == 0 which would only happen if no feature
bit could be matched by any architecture, ie there is no feature bit to
match.

2) A fallback to a superset match must exist when no_cpu_selected ()
returns true. This is because aeabi_set_public_attributes is called
again after relaxation and at this point selected_cpu is set from the
previous execution of that function. There is therefore no way to check
whether the user specified an architecture or CPU.

3) Tag_CPU_arch lines are removed from expected output when the
detected architecture should be pre-ARMv4, since 0 is the Tag_CPU_arch
value for pre-ARMv4 architectures and default value for an absent entry
is 0.

2017-06-21  Thomas Preud'homme  <thomas.preudhomme@arm.com>

gas/
* config/tc-arm.c (fpu_any): Defined from FPU_ANY.
(cpu_arch_ver): Add all architectures and sort by release date.
(have_ext_for_needed_feat_p): New.
(get_aeabi_cpu_arch_from_fset): New.
(aeabi_set_public_attributes): Call above function to determine
Tag_CPU_arch and Tag_CPU_arch_profile values.  Adapt Tag_ARM_ISA_use
and Tag_THUMB_ISA_use selection logic to check absence of feature bit
accordingly.
* testsuite/gas/arm/attr-march-armv1.d: Fix expected Tag_CPU_arch build
attribute value.
* testsuite/gas/arm/attr-march-armv2.d: Likewise.
* testsuite/gas/arm/attr-march-armv2a.d: Likewise.
* testsuite/gas/arm/attr-march-armv2s.d: Likewise.
* testsuite/gas/arm/attr-march-armv3.d: Likewise.
* testsuite/gas/arm/attr-march-armv3m.d: Likewise.
* testsuite/gas/arm/pr12198-2.d: Likewise.

include/
* opcode/arm.h (FPU_ANY): New macro.

7 years agoFix addrss violation when processing a corrupt SH COFF binary.
Nick Clifton [Wed, 21 Jun 2017 15:36:44 +0000 (16:36 +0100)]
Fix addrss violation when processing a corrupt SH COFF binary.

PR binutils/21646
* coff-sh.c (sh_reloc): Check for an out of range reloc.

7 years agox86: CET v2.0: Update incssp and setssbsy
H.J. Lu [Wed, 21 Jun 2017 15:32:38 +0000 (08:32 -0700)]
x86: CET v2.0: Update incssp and setssbsy

Update x86 assembler and disassembler for CET v2.0:

https://software.intel.com/sites/default/files/managed/4d/2a/control-flow-enforcement-technology-preview.pdf

1. incsspd and incsspq are changed to take a register opeand with a
different opcode.
2. setssbsy is changed to take no opeand with a different opcode.

gas/

* testsuite/gas/i386/cet-intel.d: Updated.
* testsuite/gas/i386/cet.d: Likewise.
* testsuite/gas/i386/x86-64-cet-intel.d: Likewise.
* testsuite/gas/i386/x86-64-cet.d: Likewise.
* testsuite/gas/i386/cet.s: Update incsspd and setssbsy tests.
* testsuite/gas/i386/x86-64-cet.s: Likewise.

opcodes/

* i386-dis.c (RM_0FAE_REG_5): Removed.
(PREFIX_MOD_3_0F01_REG_5_RM_1): Likewise.
(PREFIX_MOD_3_0F01_REG_5_RM_0): New.
(PREFIX_MOD_3_0FAE_REG_5): Likewise.
(prefix_table): Remove PREFIX_MOD_3_0F01_REG_5_RM_1.  Add
PREFIX_MOD_3_0F01_REG_5_RM_0.
(prefix_table): Update PREFIX_MOD_0_0FAE_REG_5.  Add
PREFIX_MOD_3_0FAE_REG_5.
(mod_table): Update MOD_0FAE_REG_5.
(rm_table): Update RM_0F01_REG_5.  Remove RM_0FAE_REG_5.
* i386-opc.tbl: Update incsspd, incsspq and setssbsy.
* i386-tbl.h: Regenerated.

7 years agox86: CET v2.0: Rename savessp to saveprevssp
H.J. Lu [Wed, 21 Jun 2017 15:30:01 +0000 (08:30 -0700)]
x86: CET v2.0: Rename savessp to saveprevssp

Replace savessp with saveprevssp for CET v2.0:

https://software.intel.com/sites/default/files/managed/4d/2a/control-flow-enforcement-technology-preview.pdf

gas/

* testsuite/gas/i386/cet-intel.d: Updated.
* testsuite/gas/i386/cet.d: Likewise.
* testsuite/gas/i386/x86-64-cet-intel.d: Likewise.
* testsuite/gas/i386/x86-64-cet.d: Likewise.
* testsuite/gas/i386/cet.s: Replace savessp with saveprevssp.
* testsuite/gas/i386/x86-64-cet.s: Likewise.

opcodes/

* i386-dis.c (prefix_table): Replace savessp with saveprevssp.
* i386-opc.tbl: Likewise.
* i386-tbl.h: Regenerated.

7 years agox86: CET v2.0: Update NOTRACK prefix
H.J. Lu [Wed, 21 Jun 2017 15:28:30 +0000 (08:28 -0700)]
x86: CET v2.0: Update NOTRACK prefix

Update NOTRACK prefix handling to support memory indirect branch for
CET v2.0:

https://software.intel.com/sites/default/files/managed/4d/2a/control-flow-enforcement-technology-preview.pdf

gas/

* config/tc-i386.c (md_assemble): Update NOTRACK prefix check.
* testsuite/gas/i386/notrack-intel.d: Updated.
* testsuite/gas/i386/notrack.d: Likewise.
* testsuite/gas/i386/notrackbad.l: Likewise.
* testsuite/gas/i386/x86-64-notrack-intel.d: Likewise.
* testsuite/gas/i386/x86-64-notrack.d: Likewise.
* testsuite/gas/i386/x86-64-notrackbad.l: Likewise.
* testsuite/gas/i386/notrack.s: Add NOTRACK prefix tests with
memory indirect branch.
* testsuite/gas/i386/x86-64-notrack.s: Likewise.
* testsuite/gas/i386/notrackbad.s: Remove memory indirect branch
with NOTRACK prefix.
* testsuite/gas/i386/x86-64-notrackbad.s: Likewise.

opcodes/

* i386-dis.c (reg_table): Swap indirEv with NOTRACK on "call{&|}"
and "jmp{&|}".
(NOTRACK_Fixup): Support memory indirect branch with NOTRACK
prefix.

7 years agoFix address violation parsing a corrupt Alpha VMS binary file.
Nick Clifton [Wed, 21 Jun 2017 14:21:11 +0000 (15:21 +0100)]
Fix address violation parsing a corrupt Alpha VMS binary file.

PR binutils/21639
* vms-misc.c (_bfd_vms_save_sized_string): Use unsigned int as
type of the size parameter.
(_bfd_vms_save_counted_string): Add second parameter - the maximum
length of the counted string.
* vms.h (_bfd_vms_save_sized_string): Update prototype.
(_bfd_vms_save_counted_string): Likewise.
* vms-alpha.c (_bfd_vms_slurp_eisd): Update calls to
_bfd_vms_save_counted_string.
(_bfd_vms_slurp_ehdr): Likewise.
(_bfd_vms_slurp_egsd): Likewise.
(Parse_module): Likewise.

7 years ago[ARM] Allow Thumb division as an extension for ARMv7
Thomas Preud'homme [Wed, 21 Jun 2017 14:06:51 +0000 (15:06 +0100)]
[ARM] Allow Thumb division as an extension for ARMv7

=== Context ===

This patch is part of a patch series to add support for ARMv8-R
architecture. Its purpose is to allow ARMv7 to be selected in automatic
architecture selection in presence of Thumb division instructions.

=== Motivation ===

any-idiv.d and automatic-sdiv.d testcases in GAS testsuite expect
autodetection code to select ARMv7 in presence of Thumb integer
division. However, the definition of ARM_AEXT_V7 and thus ARM_ARCH_V7 do
not contain these instructions and the idiv extension is only available
for ARMv7-A and ARMv7-R. Therefore, under the stricter automatic
detection code proposed in the subsequent patch of the series ARMv7 is
refused if a Thumb division instruction is present.

=== Patch description ===

This patch adds a new "idiv" extension after the existing one that is
available to all ARMv7 targets. This new entry is ignored by all current
code parsing arm_extensions such that it would be unavailable on the
command-line and remain a purely internal hack, easily removed in favor
of a better solution later. This is considered though by the subsequent
patch reworking automatic detection of build attributes such that ARMv7
is allowed to match in present of Thumb division instructions. For good
measure, comments are added in all instances of code browsing
arm_extensions to mention the expected behavior in case of duplicate
entries as well as a new testcase.

2017-06-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>

gas/
* config/tc-arm.c (arm_extensions): New duplicate idiv entry to enable
Thumb division for ARMv7 architecture.
(arm_parse_extension): Document expected behavior for duplicate
entries.
(s_arm_arch_extension): Likewise.
* testsuite/gas/arm/forbid-armv7-idiv-ext.d: New test.
* testsuite/gas/arm/forbid-armv7-idiv-ext.l: New expected output for
above test.

7 years ago[ARM] Rework selection of feature bits to base build attributes on
Thomas Preud'homme [Wed, 21 Jun 2017 13:57:53 +0000 (14:57 +0100)]
[ARM] Rework selection of feature bits to base build attributes on

=== Context ===

This patch is part of a patch series to add support for ARMv8-R
architecture. Its purpose is to set the feature bits on which to decide
what the build attributes should be according to the mode
(autodetection, user specified architecture or CPU, or
-march/-mcpu=all).

=== Motivation ===

Currently, the flags variable which is used to determine the build
attribute is constructed from the instruction used (arm_arch_used and
thumb_arch_used) as well as the user specified architecture or CPU
(selected_cpu). This means when several .arch are specified the
resulting feature bits can be such that no architecture provide them
all and can thus result in incorrect Tag_CPU_arch. See for instance
what having both .arch armv8-a and .arch armv8-m.base would result in.
This is not caught by the testsuite because of further bugs in the
Tag_CPU_arch build attribute value selection logic (see next patch in
the series).

=== Patch description ===

As one would expect, this patch solves the problem by setting flags
from feature bits used if in autodetection mode [1] and from
selected_cpu otherwise. The logic to set arm_ext_v1, arm_ext_v4t and
arm_ext_os feature bits is also moved to only run in autodetection mode
since otherwise the architecture or CPU would have a consistent set of
feature bits already.

[1] No architecture or CPU was specified by the user

2017-06-21  Thomas Preud'homme  <thomas.preudhomme@arm.com>

gas/
* config/tc-arm.c (aeabi_set_public_attributes): Populate flags from
feature bits used or selected_cpu depending on whether a CPU was
selected by the user.

7 years ago[GOLD] PowerPC move plt indx_ out of unordered map key
Alan Modra [Wed, 21 Jun 2017 00:40:00 +0000 (10:10 +0930)]
[GOLD] PowerPC move plt indx_ out of unordered map key

I was lazy when adding indx_ to Plt_stub_ent.  The field isn't part of
the key, so ought to be part of the mapped type.  Make it so.

* powerpc.cc (Plt_stub_key): Rename from Plt_stub_ent.  Remove indx_.
(Plt_stub_key_hash): Rename from Plt_stub_ent_hash.
(struct Plt_stub_ent): New.
(Plt_stub_entries): Map from Plt_stub_key to Plt_stub_ent.  Adjust
use throughout file.

7 years agoPowerPC64 localentry:0 plt calls
Alan Modra [Wed, 21 Jun 2017 05:33:25 +0000 (15:03 +0930)]
PowerPC64 localentry:0 plt calls

These don't need a following nop.  Also, a localentry:0 plt call
marked with an R_PPC64_TOCSAVE reloc should ignore the tocsave.
There's no need to save r2 in the prologue for such calls.

* elf64-ppc.c (ppc64_elf_size_stubs): Test for localentry:0 plt
calls before tocsave calls.
(ppc64_elf_relocate_section): Allow localentry:0 plt calls without
following nop.

7 years agoPowerPC64 tocsave testcases
Alan Modra [Tue, 20 Jun 2017 00:00:23 +0000 (09:30 +0930)]
PowerPC64 tocsave testcases

* testsuite/ld-powerpc/powerpc.exp: Run TOCSAVE tests.
* testsuite/ld-powerpc/tocsave1.s,
* testsuite/ld-powerpc/tocsave1a.d,
* testsuite/ld-powerpc/tocsave1s.d,
* testsuite/ld-powerpc/tocsave2.s,
* testsuite/ld-powerpc/tocsave2a.d,
* testsuite/ld-powerpc/tocsave2s.d,
* testsuite/ld-powerpc/tocsavelib.s: New files.

7 years ago[ARM] Simplify Tag_DSP_extension selection logic
Thomas Preud'homme [Wed, 21 Jun 2017 13:09:38 +0000 (14:09 +0100)]
[ARM] Simplify Tag_DSP_extension selection logic

=== Context ===

This patch is part of a patch series to add support for ARMv8-R
architecture. Its purpose is to simplify the logic to decide whether to
set Tag_DSP_extension.

=== Motivation ===

To decide whether to set Tag_DSP_extension, the current code was
checking whether the flags had DSP instruction but the architecture
selected for Tag_CPU_arch did not have any. This was necessary because
extension feature bit were not available separately. This is no longer
necessary and can be simplified.

=== Patch description ===

The patch change the logic to set Tag_DSP_extension to check whether any
DSP feature bit is set in the extension feature bit, as per the
definition of that build attribute. The patch also removes all
definitions of arm_arch which is now unneeded.

2017-06-21  Thomas Preud'homme  <thomas.preudhomme@arm.com>

gas/
* config/tc-arm.c (aeabi_set_public_attributes): Test *mcpu_ext_opt to
decide whether to set Tag_DSP_extension build attribute value.  Remove
now useless arm_arch variable.

7 years ago[ARM] Keep separation between extensions and architecture bits throughout execution
Thomas Preud'homme [Wed, 21 Jun 2017 13:08:08 +0000 (14:08 +0100)]
[ARM] Keep separation between extensions and architecture bits throughout execution

=== Context ===

This patch is part of a patch series to add support for ARMv8-R
architecture. Its purpose is to keep the distinction between
architecture feature bits and extension ones after parsing has occured.

=== Motivation ===

This distinction is necessary to allow the Tag_CPU_arch build attribute
value to be exactly as per the architecture of the selected CPU. With
mixed architecture and extension feature bit, it is impossible to find
an architecture with an exact match of feature bit and the build
attribute value logic must then select the closest match which might not
be the right architecture. The previous patch in the patch series makes
the distinction possible when parsing -mcpu and .cpu directives but the
distinction gets lost after. Similarly feature bits contributed by
extensions in -march or .arch_extensions directive are mixed together
with architecture extensions.

=== Patch description ===

The patch adds new feature bit pointer for extension feature bits.
Information from the parsing regarding extensions can then be kept
separate in those. This requires adapting arm_parse_extension to deal
with two feature bits, allowing the architecture bits to be marked as
const. It also requires extra care when setting cpu_variant and
selected_cpu because the extension bits are optional since there might
not be any extension in use.

Note that contrary to cpu feature bits, the extension feature bits are
made read/write and are always dynamically allocated. This allows to
unconditionally free them in arm_md_post_relax added for this occasion,
thereby fixing a longstanding memory leak when arm_parse_extension was
invoked (XNEW of ext_fset without corresponding XDELETE).
Introduction of arm_md_post_relax is necessary to only free the
extension feature bits after aeabi_set_attribute_string has been called
for the last time.

2017-06-21  Thomas Preud'homme  <thomas.preudhomme@arm.com>

gas/
* config/tc-arm.c (dyn_mcpu_ext_opt): New static variable.
(dyn_march_ext_opt): Likewise.
(md_begin): Copy extension feature bits alongside architecture ones.
Merge extensions feature bits in selected_cpu and cpu_variant if there
is some.
(arm_parse_extension): Pass architecture and extension feature bits in
separate parameters, with architecture bits being read only.  Update
**opt_p directly rather than *ext_set and initialize it if needed.
(arm_parse_cpu): Stop merging architecture and extension feature bits
and instead use mcpu_cpu_opt and dyn_mcpu_ext_opt to memorize them
respectively.  Adapt to change in parameters of arm_parse_extension.
(arm_parse_arch): Adapt to change in parameters of arm_parse_extension.
(aeabi_set_attribute_string): Make function static.
(arm_md_post_relax): New function.
(s_arm_cpu): Stop merging architecture and extension feature bits and
instead use mcpu_cpu_opt and dyn_mcpu_ext_opt to memorize them
respectively.  Merge extension feature bits in cpu_variant
if there is any.
(s_arm_arch): Reset extension feature bit.  Set selected_cpu from
*mcpu_cpu_opt and cpu_variant from selected_cpu and *mfpu_opt for
consistency with s_arm_cpu.
(s_arm_arch_extension): Update *dyn_mcpu_ext_opt rather than
selected_cpu, allocating it before hand if needed.  Set selected_cpu
from it and then cpu_variant.
(s_arm_fpu): Merge *mcpu_ext_opt feature bits if any in cpu_variant.
* config/tc-arm.h (md_post_relax_hook): Set to arm_md_post_relax.
(aeabi_set_public_attributes): Delete external declaration.
(arm_md_post_relax): Declare externally.

7 years ago[ARM] Separate extensions from architectures in arm_cpus
Thomas Preud'homme [Wed, 21 Jun 2017 12:16:56 +0000 (13:16 +0100)]
[ARM] Separate extensions from architectures in arm_cpus

=== Context ===

This patch is part of a patch series to add support for ARMv8-R
architecture. Its purpose is to distinguish for a CPU the feature bits
coming from its architecture from the feature bits coming from
extension(s) available in this CPU.

=== Motivation ===

This distinction is necessary to allow the Tag_CPU_arch build attribute
value to be exactly as per the architecture of the selected CPU. With
mixed architecture and extension feature bit, it is impossible to find
an architecture with an exact match of feature bit and the build
attribute value logic must then select the closest match which might not
be the right architecture.

=== Patch description ===

The patch creates a new field in the arm_cpus table to hold the feature
set for the extensions available in each CPU. The existing architecture
feature set is then updated to remove those feature bit. The patch also
takes advantage of all the lines being changed to reindent the whole
table.

Note: This patch *adds* a memory leak due to mcpu_cpu_opt sometimes
pointing to dynamically allocated feature bits which is never freeed.
The subsequent patch in the series solves this issue as well as a
preexisting identical issue in arm_parse_extension. The patches are kept
separate for ease of review since they are both big enough already.

2017-06-21  Thomas Preud'homme  <thomas.preudhomme@arm.com>

gas/
* config/tc-arm.c (struct arm_cpu_option_table): New ext field.
(ARM_CPU_OPT): Add parameter to set new ext field and reorder canonical
name field just after the name field.
(arm_cpus): Move extension feature bit from value field to ext field,
reorder parameter according to changes in ARM_CPU_OPT and reindent.
(arm_parse_cpu): Point mcpu_cpu_opt to a bitfield merging the value and
ext field from the selected arm_cpus entry.
(s_arm_cpu): Likewise.

7 years agoFix seg-fault in the BFD parsing a corrupt input binary.
Nick Clifton [Wed, 21 Jun 2017 12:18:46 +0000 (13:18 +0100)]
Fix seg-fault in the BFD parsing a corrupt input binary.

PR binutils/21645
* reloc.c (bfd_generic_get_relocated_section_contents): Fail if
bfd_get_full_section_contents returns no contents.

7 years agoChange to_xfer_partial doc to use addressable memory units
Simon Marchi [Wed, 21 Jun 2017 10:28:03 +0000 (12:28 +0200)]
Change to_xfer_partial doc to use addressable memory units

The commit

  d309493  target: consider addressable unit size when reading/writing memory

introduced the possibility of reading memory of targets with
non-8-bit-bytes (e.g. memories that store 16 bits at each address).
The documentation of target_read and target_write was updated
accordingly, but to_xfer_partial, which is very related, wasn't updated.
This commit fixes that.

gdb/ChangeLog:

* target.h (struct target_ops) <to_xfer_partial>: Update doc to
talk about addressable units instead of bytes.

7 years agoFix potential address violation parsing a corrupt Alpha VMS file.
Nick Clifton [Wed, 21 Jun 2017 11:04:07 +0000 (12:04 +0100)]
Fix potential address violation parsing a corrupt Alpha VMS file.

PR binutils/21638
* vms-alpha.c (_bfd_vms_slurp_egsd): Check for an undersized
record.

7 years agoFix address violation parsing a corrupt IEEE Alpha binary.
Nick Clifton [Wed, 21 Jun 2017 10:13:49 +0000 (11:13 +0100)]
Fix address violation parsing a corrupt IEEE Alpha binary.

PR binutils/21637
* vms-alpha.c (_bfd_vms_slurp_egsd): Check for an empty section
list.
(image_set_ptr): Likewise.
(alpha_vms_fix_sec_rel): Likewise.
(alpha_vms_slurp_relocs): Likewise.

7 years agoFix address violation when parsing a corrupt IEEE binary.
Nick Clifton [Wed, 21 Jun 2017 09:54:04 +0000 (10:54 +0100)]
Fix address violation when parsing a corrupt IEEE binary.

PR binutils/21633
* ieee.c (ieee_slurp_sections): Check for a NULL return from
read_id.
(ieee_archive_p): Likewise.
(ieee_object_p): Likewise.

7 years agoFix seg-fault reading a corrupt ELF binary.
Nick Clifton [Wed, 21 Jun 2017 09:36:58 +0000 (10:36 +0100)]
Fix seg-fault reading a corrupt ELF binary.

PR binutils/21640
* elf.c (setup_group): Zero the group section pointer list after
allocation so that loops can be caught.  Check for NULL pointers
when processing a group list.

7 years agoAdd support for the Cortex-A55 and Cortex-A75 versions of the AArch64 architecture.
James Greenhalgh [Wed, 21 Jun 2017 08:13:25 +0000 (09:13 +0100)]
Add support for the Cortex-A55 and Cortex-A75 versions of the AArch64 architecture.

* config/tc-aarch64.c (aarch64_cpus): Add cortex-a55 and cortex-a75.
* doc/c-aarch64.texi (-mcpu): Document cortex-a55 and cortex-a75.

7 years agoFix PR gdb/21606: SYMBOL_FUNCTIONS_DOMAIN misspelled in documentation
Sergio Durigan Junior [Wed, 21 Jun 2017 01:30:11 +0000 (21:30 -0400)]
Fix PR gdb/21606: SYMBOL_FUNCTIONS_DOMAIN misspelled in documentation

Both Python and Guile documentations misspelled
SYMBOL_FUNCTIONS_DOMAIN, writing SYMBOL_FUNCTION_DOMAIN instead.  This
obvious commit fixes it.

gdb/doc/ChangeLog:
2017-06-20  Sergio Durigan Junior  <sergiodj@redhat.com>

PR gdb/21606
* python.texi (Python representation of Symbols.): Replace
SYMBOL_FUNCTION_DOMAIN by SYMBOL_FUNCTIONS_DOMAIN, fixing typo.
* guile.texi (Guile representation of Symbols.): Likewise.

7 years agoAutomatic date update in version.in
GDB Administrator [Wed, 21 Jun 2017 00:00:36 +0000 (00:00 +0000)]
Automatic date update in version.in

7 years ago2017-06-20 Eric Christopher <echristo@gmail.com>
Eric Christopher [Tue, 20 Jun 2017 23:18:58 +0000 (16:18 -0700)]
2017-06-20  Eric Christopher  <echristo@gmail.com>

        * aarch64.cc (scan_reloc_for_stub): Use plt_address_for_global to
        calculate the symbol value.
        (scan_reloc_section_for_stubs): Allow stubs to be created for
        section symbols.
        (maybe_apply_stub): Handle creating stubs for weak symbols to
        match the code in scan_reloc_for_stub.

7 years agogdbserver/Makefile.in: Sort IPA_OBJS
Simon Marchi [Tue, 20 Jun 2017 14:59:03 +0000 (16:59 +0200)]
gdbserver/Makefile.in: Sort IPA_OBJS

gdb/gdbserver/ChangeLog:

* Makefile.in (IPA_OBJS): Sort and format one item per line.

7 years agoUse '::iterator' instead of '::const_iterator' on environ.c (and fix breakage on...
Sergio Durigan Junior [Tue, 20 Jun 2017 13:33:53 +0000 (09:33 -0400)]
Use '::iterator' instead of '::const_iterator' on environ.c (and fix breakage on early versions of libstdc++)

Even though C++11 supports modifying containers using a const_iterator
(e.g., calling the 'erase' method of a std::vector), early versions of
libstdc++ did not implement that.  Some of our buildslaves are using
these versions (e.g., the AArch64 buildslave uses gcc 4.8.8), and my
previous commit causes a breakage on them.  The solution is simple:
just use a normal iterator, without const.

gdb/ChangeLog:
2017-06-20  Sergio Durigan Junior  <sergiodj@redhat.com>

* common/environ.c (gdb_environ::unset): Use '::iterator' instead
of '::const_iterator'.

7 years agoC++ify gdb/common/environ.c
Sergio Durigan Junior [Sat, 11 Feb 2017 02:19:44 +0000 (21:19 -0500)]
C++ify gdb/common/environ.c

As part of the preparation necessary for my upcoming task, I'd like to
propose that we turn gdb_environ into a class.  The approach taken
here is simple: the class gdb_environ contains everything that is
needed to manipulate the environment variables.  These variables are
stored in an std::vector<char *>, which can be converted to a 'char
**' and passed as argument to functions that need it.

The usage has not changed much.  As per Pedro's suggestion, this class
uses a static factory method initialization.  This means that when an
instance is created, it is initially empty.  When needed, it has to be
initialized using the static method 'from_host_environ'.

As mentioned before, this is a preparation for an upcoming work that I
will be posting in the next few weeks or so.  For that work, I'll
probably create another data structure that will contain all the
environment variables that were set by the user using the 'set
environment' command, because I'll need access to them.  This will be
much easier with the class-ification of gdb_environ.

As noted, this has been regression-tested with the new version of
environ.exp and no regressions were found.

gdb/ChangeLog:
2017-06-20  Sergio Durigan Junior  <sergiodj@redhat.com>

* Makefile.in (SUBDIR_UNITTESTS_SRCS): Add
'unittests/environ-selftests.c'.
(SUBDIR_UNITTESTS_OBS): Add 'environ-selftests.o'.
* charset.c (find_charset_names): Declare object 'iconv_env'.
Update code to use 'iconv_env' object.  Remove call to
'free_environ'.
* common/environ.c: Include <utility>.
(make_environ): Delete function.
(free_environ): Delete function.
(gdb_environ::clear): New function.
(gdb_environ::operator=): New function.
(gdb_environ::get): Likewise.
(environ_vector): Delete function.
(set_in_environ): Delete function.
(gdb_environ::set): New function.
(unset_in_environ): Delete function.
(gdb_environ::unset): New function.
(gdb_environ::envp): Likewise.
* common/environ.h: Include <vector>.
(struct gdb_environ): Delete; transform into...
(class gdb_environ): ... this class.
(free_environ): Delete prototype.
(init_environ, get_in_environ, set_in_environ, unset_in_environ,
environ_vector): Likewise.
* infcmd.c (run_command_1): Update code to call
'envp' from 'gdb_environ' class.
(environment_info): Update code to call methods from 'gdb_environ'
class.
(unset_environment_command): Likewise.
(path_info): Likewise.
(path_command): Likewise.
* inferior.c (inferior::~inferior): Delete call to 'free_environ'.
(inferior::inferior): Initialize 'environment' using the host's
information.
* inferior.h: Remove forward declaration of 'struct gdb_environ'.
Include "environ.h".
(class inferior) <environment>: Change type from 'struct
gdb_environ' to 'gdb_environ'.
* mi/mi-cmd-env.c (mi_cmd_env_path): Update code to call
methods from 'gdb_environ' class.
* solib.c (solib_find_1): Likewise
* unittests/environ-selftests.c: New file.

gdb/gdbserver/ChangeLog:
2017-06-20  Sergio Durigan Junior  <sergiodj@redhat.com>

* linux-low.c (linux_create_inferior): Adjust code to access the
environment information via 'gdb_environ' class.
* lynx-low.c (lynx_create_inferior): Likewise.
* server.c (our_environ): Make it an instance of 'gdb_environ'.
(get_environ): Return a pointer to 'our_environ'.
(captured_main): Initialize 'our_environ'.
* server.h (get_environ): Adjust prototype.
* spu-low.c (spu_create_inferior): Adjust code to access the
environment information via 'gdb_environ' class.

7 years agoAdjust the order of 32bit-linux.xml and 32bit-sse.xml in i386/i386-linux.xml
Yao Qi [Tue, 20 Jun 2017 11:08:33 +0000 (12:08 +0100)]
Adjust the order of 32bit-linux.xml and 32bit-sse.xml in i386/i386-linux.xml

Exchange the order of 32bit-linux.xml and 32bit-sse.xml in
i386/i386-linux.xml, to align with other i386 linux .xml files.

gdb:

2017-06-20  Yao Qi  <yao.qi@linaro.org>

* features/i386/i386-linux.xml: Exchange the order of including
32bit-linux.xml and 32bit-sse.xml.
* features/i386/i386-linux.c: Regenerated.

7 years agoClass-fy tdesc_reg tdesc_type and tdesc_feature
Yao Qi [Tue, 20 Jun 2017 10:29:17 +0000 (11:29 +0100)]
Class-fy tdesc_reg tdesc_type and tdesc_feature

This patch class-fies them, adding ctor, dtor, and deleting
copy ctor and assignment operator.

gdb:

2017-06-20  Yao Qi  <yao.qi@linaro.org>

* target-descriptions.c (tdesc_reg): Add ctor, dtor.
Delete copy ctor and assignment operator.
(tdesc_type): Likewise.
(tdesc_feature): Likewise.
(tdesc_free_reg): Remove.
(tdesc_create_reg): Use new.
(tdesc_free_type): Remove.
(tdesc_create_vector): Use new.
(tdesc_create_union): Likewise.
(tdesc_create_flags): Likewise.
(tdesc_create_enum): Likewise.
(tdesc_free_feature): Delete.
(free_target_description): Use delete.

7 years ago[GOLD] Avoid duplicate PLT stub symbols on ppc32
James Clarke [Tue, 20 Jun 2017 08:31:52 +0000 (18:01 +0930)]
[GOLD] Avoid duplicate PLT stub symbols on ppc32

If two objects are compiled with -fPIC or -fPIE and call the same
function, two different PLT entries are created, one for each object,
but the same stub symbol name is used for both.

* powerpc.cc (Stub_table::define_stub_syms): Always include object's
uniq_ value.

7 years agoCheck the DYNAMIC bit for input shared objects
H.J. Lu [Tue, 20 Jun 2017 03:30:20 +0000 (20:30 -0700)]
Check the DYNAMIC bit for input shared objects

Since the BFD section count may not be cleared for shared objects during
linking, we should check the DYNAMIC bit for input shared objects.

bfd/

PR ld/21626
* elf-properties.c (_bfd_elf_link_setup_gnu_properties): Check
the DYNAMIC bit instead of bfd_count_sections.

ld/

PR ld/21626
* testsuite/ld-i386/i386.exp: Run ld/21626 tests.
* testsuite/ld-x86-64/x86-64.exp: Likewise.

7 years agoAutomatic date update in version.in
GDB Administrator [Tue, 20 Jun 2017 00:00:38 +0000 (00:00 +0000)]
Automatic date update in version.in

7 years agoDon't throw an error in 'info registers' for unavailable MIPS registers.
John Baldwin [Mon, 19 Jun 2017 21:30:24 +0000 (14:30 -0700)]
Don't throw an error in 'info registers' for unavailable MIPS registers.

'info registers' for MIPS throws an error and when it first encounters
an unavailable register.  This does not match other architectures
which annotate unavailable registers and continue to print out the
values of subsequent registers.  Replace the error by displaying an
aligned "<unavailable>".  This string is truncated to "<unavl>" when
displaying a 32-bit register.

gdb/ChangeLog:

* mips-tdep.c (print_gp_register_row): Don't error for unavailable
registers.

7 years agoUpdate GDB test case for new lnia extended mnemonic.
Peter Bergner [Mon, 19 Jun 2017 18:04:13 +0000 (13:04 -0500)]
Update GDB test case for new lnia extended mnemonic.

When I added the new lnia extended mnemonic for addpcis, I updated the
assembler/disassembler test cases, but overlooked the GDB test cases.
This patch fixes that oversight and associated test case failure.

* gdb.arch/powerpc-power9.exp: Update test case for new lnia
extended mnemonic.
* gdb.arch/powerpc-power9.s: Likewise.

7 years agoFix address violation when attempting to display disassembled data.
Nick Clifton [Mon, 19 Jun 2017 14:57:19 +0000 (15:57 +0100)]
Fix address violation when attempting to display disassembled data.

PR binutils/21619
* objdump.c (disassemble_bytes): Check that there is sufficient
data available before attempting to display it.

7 years agoFix address violations when reading corrupt VMS records.
Nick Clifton [Mon, 19 Jun 2017 13:52:36 +0000 (14:52 +0100)]
Fix address violations when reading corrupt VMS records.

PR binutils/21618
* vms-alpha.c (evax_bfd_print_emh): Check for insufficient record
length.
(evax_bfd_print_eeom): Likewise.
(evax_bfd_print_egsd): Check for an overlarge record length.
(evax_bfd_print_etir): Likewise.

7 years agoPrevent address violation when attempting to disassemble a corrupt score binary.
Nick Clifton [Mon, 19 Jun 2017 13:15:57 +0000 (14:15 +0100)]
Prevent address violation when attempting to disassemble a corrupt score binary.

PR binutils/21614
* score-dis.c (score_opcodes): Add sentinel.

7 years agoFix access violation when parsing a corrupt IEEE binary.
Nick Clifton [Mon, 19 Jun 2017 13:06:53 +0000 (14:06 +0100)]
Fix access violation when parsing a corrupt IEEE binary.

PR binutils/21612
* libieee.h (struct common_header_type): Add end_p field.
* ieee.c (this_byte_and_next): Do not advance input_p beyond
end_p.
(read_id): Check for a length that exceeds the remaining bytes in
the input buffer.
(ieee_seek): Initialise end_p.
(ieee_archive_p): Likewise.
(ieee_object_p): Likewise.

7 years agoFix access violation disassembling a corrupt VMS binary.
Nick Clifton [Mon, 19 Jun 2017 12:01:57 +0000 (13:01 +0100)]
Fix access violation disassembling a corrupt VMS binary.

PR binutils/21611
* vms-alpha.c (_bfd_vms_slurp_eihs): Check for invalid offset
before reading the EIHS structure entries.

7 years ago.gdb_index writer: close the file before unlinking it
Pedro Alves [Mon, 19 Jun 2017 11:46:47 +0000 (12:46 +0100)]
.gdb_index writer: close the file before unlinking it

We should close the file before unlinking because on MS-Windows one
cannot delete a file that is still open.

I considered making 'gdb::unlinker::unlinker(const char *)'
'noexcept(true)' and then adding
  static_assert (noexcept (gdb::unlinker (filename.c_str ())), "");

but that doesn't really work because gdb::unlinker has a gdb_assert,
which can throw a QUIT if/when the assertion fails.  'noexcept(true)'
would cause GDB to abruptly terminate if/when the assertion fails.

gdb/ChangeLog:
2017-06-19  Pedro Alves  <palves@redhat.com>

* dwarf2read.c (write_psymtabs_to_index): Construct file_closer
after gdb::unlinker.

7 years agoFix access violation when disassembling a corrupt VMS binary.
Nick Clifton [Mon, 19 Jun 2017 11:31:07 +0000 (12:31 +0100)]
Fix access violation when disassembling a corrupt VMS binary.

PR 21615
* vms-alpha.c (_bfd_vms_slurp_egsd): Use unsigned int for
gsd_size.  Check that there are enough bytes remaining to read the
type and size of the next egsd.  Check that the size of the egsd
does not exceed the size of the record.

7 years agoCorrect target_underscore for cris
Alan Modra [Mon, 19 Jun 2017 06:34:16 +0000 (16:04 +0930)]
Correct target_underscore for cris

* config.bfd: Correct targ_underscore for cris.

7 years agoUse getenv instead of gdb_environ on mi-cmd-env.c
Sergio Durigan Junior [Fri, 16 Jun 2017 22:10:27 +0000 (18:10 -0400)]
Use getenv instead of gdb_environ on mi-cmd-env.c

This is a spinoff of
<https://sourceware.org/ml/gdb-patches/2017-06/msg00437.html>.
mi-cmd-env.c is using the whole gdb_environ machinery in order to
access just one variable, which can be easily replaced by a simple
call to getenv.  This patch does that, and doesn't cause regressions.

gdb/ChangeLog:
2017-06-18  Sergio Durigan Junior  <sergiodj@redhat.com>

* mi/mi-cm-env.c (_initialize_mi_cmd_env): Use getenv instead of
gdb_environ to access an environment variable.

7 years agoAutomatic date update in version.in
GDB Administrator [Mon, 19 Jun 2017 00:00:23 +0000 (00:00 +0000)]
Automatic date update in version.in

7 years agonat/linux-ptrace.c: add missing gdb_byte* cast
Thomas Petazzoni [Sun, 18 Jun 2017 21:28:56 +0000 (23:28 +0200)]
nat/linux-ptrace.c: add missing gdb_byte* cast

On noMMU platforms, the following code gets compiled:

  child_stack = xmalloc (STACK_SIZE * 4);

Where child_stack is a gdb_byte*, and xmalloc() returns a void*. While
the lack of cast is valid in C, it is not in C++, causing the
following build failure:

../nat/linux-ptrace.c: In function 'int linux_fork_to_function(gdb_byte*, int (*)(void*))':
../nat/linux-ptrace.c:273:29: error: invalid conversion from 'void*' to 'gdb_byte* {aka unsigned char*}' [-fpermissive]
       child_stack = xmalloc (STACK_SIZE * 4);

Therefore, this commit adds the appropriate cast.

gdb/ChangeLog:

* nat/linux-ptrace.c (linux_fork_to_function): Add cast to
gdb_byte*.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
7 years agois_underscore_target for ld-elf tests
Alan Modra [Sat, 17 Jun 2017 23:49:34 +0000 (09:19 +0930)]
is_underscore_target for ld-elf tests

and correct targ_underscore in config.bfd

bfd/
* config.bfd: Correct targ_underscore for epiphany, ip2k,
m32c, mn10200, pru, rl78, rx, crisv32 and v850.
ld/
* testsuite/lib/ld-lib.exp (is_underscore_target): New.
* testsuite/ld-elf/elf.exp (ASFLAGS): Define UNDERSCORE.
* testsuite/ld-elf/pr21562a.s: If UNDERSCORE defined,
reference sym with prefix.
* testsuite/ld-elf/pr21562b.s: Likewise.
* testsuite/ld-elf/sizeof.s: Likewise.
* testsuite/ld-elf/startof.s: Likewise.
* testsuite/ld-elf/pr14156a.d: Adjust for extra symbols.
* testsuite/ld-elf/pr21562a.d: Remove underscore target from
xfails, and match prefixed symbol.
* testsuite/ld-elf/pr21562b.d: Likewise.
* testsuite/ld-elf/pr21562c.d: Likewise.
* testsuite/ld-elf/pr21562d.d: Likewise.
* testsuite/ld-elf/pr21562e.d: Likewise.
* testsuite/ld-elf/pr21562f.d: Likewise.
* testsuite/ld-elf/pr21562g.d: Likewise.
* testsuite/ld-elf/pr21562h.d: Likewise.
* testsuite/ld-elf/pr21562i.d: Likewise.
* testsuite/ld-elf/pr21562j.d: Likewise.
* testsuite/ld-elf/pr21562k.d: Likewise.
* testsuite/ld-elf/pr21562l.d: Likewise.
* testsuite/ld-elf/pr21562m.d: Likewise.
* testsuite/ld-elf/pr21562n.d: Likewise.
* testsuite/ld-elf/sizeofa.d: Likewise.
* testsuite/ld-elf/sizeofb.d: Likewise.
* testsuite/ld-elf/sizeofc.d: Likewise.
* testsuite/ld-elf/startofa.d: Likewise.
* testsuite/ld-elf/startofb.d: Likewise.
* testsuite/ld-elf/startofc.d: Likewise.

7 years agoAutomatic date update in version.in
GDB Administrator [Sun, 18 Jun 2017 00:00:37 +0000 (00:00 +0000)]
Automatic date update in version.in

7 years agoAdd ATTRIBUTE_PRINTF to trace_start_error
Simon Marchi [Sat, 17 Jun 2017 21:19:25 +0000 (23:19 +0200)]
Add ATTRIBUTE_PRINTF to trace_start_error

clang complains that the fmt passed to vwarning in trace_start_error is
not a literal.  This looks like a fair warning, which can be removed by
adding ATTRIBUTE_PRINTF to the declaration of trace_start_error.

gdb/ChangeLog:

* nat/fork-inferior.h (trace_start_error): Add ATTRIBUTE_PRINTF.

7 years agolinux-low: Remove usage of "register" keyword
Simon Marchi [Sat, 17 Jun 2017 21:19:08 +0000 (23:19 +0200)]
linux-low: Remove usage of "register" keyword

AFAIK, the register keyword is not relevant today, and clang complains
about it:

/home/emaisin/src/binutils-gdb/gdb/gdbserver/linux-low.c:5873:3: error: 'register' storage class specifier is deprecated and incompatible with C++1z
      [-Werror,-Wdeprecated-register]
  register PTRACE_XFER_TYPE *buffer;
  ^~~~~~~~~

I think we can safely remove it.

gdb/gdbserver/ChangeLog:

* linux-low.c (linux_read_memory, linux_write_memory): Remove
usage of "register" keyword.

7 years agogdb: Add -Wno-mismatched-tags
Simon Marchi [Sat, 17 Jun 2017 21:18:49 +0000 (23:18 +0200)]
gdb: Add -Wno-mismatched-tags

clang complains that for some types, we use both the class and struct
keywords in different places.  It's not really a problem, so I think we
can safely turn this warning off.

gdb/ChangeLog:

* configure: Re-generate.
* warning.m4 (build_warnings): Add -Wno-mismatched-tags.

gdb/gdbserver/ChangeLog:

* configure: Re-generate.

7 years agogdb: Use -Werror when checking for (un)supported warning flags
Simon Marchi [Sat, 17 Jun 2017 21:18:20 +0000 (23:18 +0200)]
gdb: Use -Werror when checking for (un)supported warning flags

In warning.m4, we pass all the warning flags one by one to the compiler
to test if they are supported by this particular compiler.  If the
compiler exits with an error, we conclude that this warning flag is not
supported and exclude it.  This allows us to use warning flags without
having to worry about which versions of which compilers support each
flag.

clang, by default, only emits a warning if an unknown flag is passed:

  warning: unknown warning option '-Wfoo' [-Wunknown-warning-option]

The result is that we think that all the warning flags we use are
supported by clang (they are not), and the compilation fails later when
building with -Werror, since the aforementioned warning becomes an
error.  The fix is to also pass -Werror when probing for supported
flags, then we'll correctly get an error when using an unknown warning,
and we'll exclude it:

  error: unknown warning option '-Wfoo' [-Werror,-Wunknown-warning-option]

I am not sure why there is a change in a random comment in
gdbserver/configure, but I suppose it's a leftfover from a previous
patch, so I included it.

gdb/ChangeLog:

* configure: Re-generate.
* warning.m4: Pass -Werror to compiler when checking for
supported warning flags.

gdb/gdbserver/ChangeLog:

* configure: Re-generate.

7 years agogdb: Pass -x c++ to the compiler
Simon Marchi [Sat, 17 Jun 2017 21:17:00 +0000 (23:17 +0200)]
gdb: Pass -x c++ to the compiler

Because we are compiling .c files containing C++ code, clang++ complains
with:

  clang: error: treating 'c' input as 'c++' when in C++ mode, this behavior is deprecated

If renaming all the source files to .cpp is out of the question, an
alternative is to pass "-x c++" to convince the compiler that we are
really compiling C++.  It works fine with GCC too.

gdb/ChangeLog:

* Makefile.in (COMPILE.pre): Add "-x c++".

gdb/gdbserver/ChangeLog:

* Makefile.in (COMPILE.pre): Add "-x c++".

7 years agoAutomatic date update in version.in
GDB Administrator [Sat, 17 Jun 2017 00:00:41 +0000 (00:00 +0000)]
Automatic date update in version.in

7 years agoextract/store integer function template
Yao Qi [Fri, 16 Jun 2017 14:38:42 +0000 (15:38 +0100)]
extract/store integer function template

This patch converts functions extract_{unsigned,signed}_integer
to a function template extract_integer, which has two instantiations.  It
also does the similar changes to store__{unsigned,signed}_integer,
regcache::raw_read_{unsigned,signed}, regcache::raw_write_{unsigned,signed},
regcache::cooked_read_{unsigned,signed},
regcache::cooked_write_{unsigned,signed}.

This patch was posted here
https://sourceware.org/ml/gdb-patches/2017-05/msg00492.html but the
problem was fixed in a different way.  However, I think the patch is still
useful to shorten the code.

gdb:

2017-06-16  Alan Hayward  <alan.hayward@arm.com>
    Pedro Alves  <palves@redhat.com>
    Yao Qi  <yao.qi@linaro.org>

* defs.h (RequireLongest): New.
(extract_integer): Declare function template.
(extract_signed_integer): Remove the declaration, but define it
static inline.
(extract_unsigned_integer): Likewise.
(store_integer): Declare function template.
(store_signed_integer): Remove the declaration, but define it
static inline.
(store_unsigned_integer): Likewise.
* findvar.c (extract_integer): New function template.
(extract_signed_integer): Remove.
(extract_unsigned_integer): Remove.
(extract_integer<LONGEST>, extract_integer<ULONGEST>): Explicit
instantiations.
(store_integer): New function template.
(store_signed_integer): Remove.
(store_unsigned_integer): Remove.
(store_integer): Explicit instantiations.
* regcache.c (regcache_raw_read_signed): Update.
(regcache::raw_read): New function.
(regcache::raw_read_signed): Remove.
(regcache::raw_read_unsigned): Remove.
(regcache_raw_read_unsigned): Update.
(regcache_raw_write_unsigned): Update.
(regcache::raw_write_signed): Remove.
(regcache::raw_write): New function.
(regcache_cooked_read_signed): Update.
(regcache::raw_write_unsigned): Remove.
(regcache::cooked_read_signed): Remove.
(regcache_cooked_read_unsigned): Update.
(regcache::cooked_read_unsigned): Remove.
(regcache_cooked_write_signed): Update.
(regcache_cooked_write_unsigned): Update.
* regcache.h (regcache) <raw_read_signed>: Remove.
<raw_write_signed, raw_read_unsigned, raw_write_unsigned>: Remove.
<raw_read, raw_write>: New.
<cooked_read_signed, cooked_write_signed>: Remove.
<cooked_write_unsigned, cooked_read_unsigned>: Remove.
<cooked_read, cooked_write>: New.
* sh64-tdep.c (sh64_pseudo_register_read): Update.
(sh64_pseudo_register_write): Update.

7 years agoRegen rx-decode.c
Alan Modra [Fri, 16 Jun 2017 12:20:43 +0000 (21:50 +0930)]
Regen rx-decode.c

opcodes/
* rx-decode.c: Regenerate.

7 years agoFixing linking configure generated tests of ifunc support.
Nick Clifton [Fri, 16 Jun 2017 14:16:19 +0000 (15:16 +0100)]
Fixing linking configure generated tests of ifunc support.

* elflink.c (bfd_elf_size_dynsym_hash_dynstr): Do not fail if the
bucketlist is empty because there are no symbols to add to the
list.

7 years agoRewrite __start and __stop symbol handling
Alan Modra [Fri, 16 Jun 2017 10:11:41 +0000 (19:41 +0930)]
Rewrite __start and __stop symbol handling

This arranges for __start and __stop symbols to be defined before
garbage collection, for all target formats.  That should allow the
COFF and PE --gc-sections to keep a singleton orphan input section,
a feature lost by 2017-06-13 commit cbd0eecf26.  The fancier ELF
treatment of keeping all input sections associated with a __start or
__stop symbol, from 2015-10-23 commit 1cce69b9dc, is retained.

.startof. and .sizeof. symbols are deliberately not defined before
garbage collection, so these won't affect garbage collection of
sections.

The patch also ensures __start, __stop, .startof. and .sizeof. symbols
are defined before target size_dynamic_sections is called, albeit
with a preliminary value, so that target code doesn't need to cope
with a symbol changing from undefined at size_dynamic_sections to
defined at relocate_section.

Also, a number of problems with the testcases have been fixed.

PR ld/20022
PR ld/21557
PR ld/21562
PR ld/21571
include/
* bfdlink.h (struct bfd_link_hash_entry): Delete undef.section.
bfd/
* targets.c (struct bfd_target): Add _bfd_define_start_stop.
(BFD_JUMP_TABLE_LINK): Likewise.
* elf-bfd.h (bfd_elf_define_start_stop): Declare.
* elflink.c (_bfd_elf_gc_mark_rsec): Update comment.
(bfd_elf_define_start_stop): New function.
* linker.c (bfd_generic_define_start_stop): New function.
* coff64-rs6000.c (rs6000_xcoff64_vec, rs6000_xcoff64_aix_vec): Init
new field.
* aout-adobe.c (aout_32_bfd_define_start_stop): Define.
* aout-target.h (MY_bfd_define_start_stop): Define.
* aout-tic30.c (MY_bfd_define_start_stop): Define.
* binary.c (binary_bfd_define_start_stop): Define.
* bout.c (b_out_bfd_define_start_stop): Define.
* coff-alpha.c (_bfd_ecoff_bfd_define_start_stop): Define.
* coff-mips.c (_bfd_ecoff_bfd_define_start_stop): Define.
* coff-rs6000.c (_bfd_xcoff_bfd_define_start_stop): Define.
* coffcode.h (coff_bfd_define_start_stop): Define.
* elfxx-target.h (bfd_elfNN_bfd_define_start_stop): Define.
* i386msdos.c (msdos_bfd_define_start_stop): Define.
* i386os9k.c (os9k_bfd_define_start_stop): Define.
* ieee.c (ieee_bfd_define_start_stop): Define.
* ihex.c (ihex_bfd_define_start_stop): Define.
* libbfd-in.h (_bfd_nolink_bfd_define_start_stop): Define.
* mach-o-target.c (bfd_mach_o_bfd_define_start_stop): Define.
* mmo.c (mmo_bfd_define_start_stop): Define.
* nlm-target.h (nlm_bfd_define_start_stop): Define.
* oasys.c (oasys_bfd_define_start_stop): Define.
* pef.c (bfd_pef_bfd_define_start_stop): Define.
* plugin.c (bfd_plugin_bfd_define_start_stop): Define.
* ppcboot.c (ppcboot_bfd_define_start_stop): Define.
* som.c (som_bfd_define_start_stop): Define.
* srec.c (srec_bfd_define_start_stop): Define.
* tekhex.c (tekhex_bfd_define_start_stop): Define.
* versados.c (versados_bfd_define_start_stop): Define.
* vms-alpha.c (vms_bfd_define_start_stop): Define.
(alpha_vms_bfd_define_start_stop): Define.
* xsym.c (bfd_sym_bfd_define_start_stop): Define.
* bfd-in2.h: Regenerate.
* libbfd.h: Regenerate.
ld/
* emultempl/elf32.em (gld${EMULATION_NAME}_after_open): Don't set
__start/__stop syms here.
* ldlang.c (lang_set_startof): Delete.
(start_stop_syms, start_stop_count, start_stop_alloc): New vars.
(lang_define_start_stop, lang_init_start_stop, foreach_start_stop,
undef_start_stop, lang_undef_start_stop, lang_init_startof_sizeof,
set_start_stop, lang_finalize_start_stop): New functions.
(lang_process): Call _start_stop functions.
* testsuite/ld-elf/pr21562a.d: Use xfail rather than notarget.
Correct typos and list of xfail targets.
* testsuite/ld-elf/pr21562b.d: Likewise.
* testsuite/ld-elf/pr21562c.d: Likewise.
* testsuite/ld-elf/pr21562d.d: Likewise.
* testsuite/ld-elf/pr21562e.d: Likewise.
* testsuite/ld-elf/pr21562f.d: Likewise.
* testsuite/ld-elf/pr21562g.d: Likewise.
* testsuite/ld-elf/pr21562h.d: Likewise.
* testsuite/ld-elf/pr21562i.d: Likewise.
* testsuite/ld-elf/pr21562j.d: Likewise.
* testsuite/ld-elf/pr21562k.d: Likewise.
* testsuite/ld-elf/pr21562l.d: Likewise.
* testsuite/ld-elf/pr21562m.d: Likewise.
* testsuite/ld-elf/pr21562n.d: Likewise.
* testsuite/ld-elf/sizeofa.d: Likewise.  Adjust to pass for generic ELF.
* testsuite/ld-elf/sizeofb.d: Likewise.
* testsuite/ld-elf/startofa.d: Likewise.
* testsuite/ld-elf/startofb.d: Likewise.

7 years ago[AArch64] Use SYMBOL_REFERENCES_LOCAL in one symbol check
Jiong Wang [Fri, 16 Jun 2017 13:20:38 +0000 (14:20 +0100)]
[AArch64] Use SYMBOL_REFERENCES_LOCAL in one symbol check

For some pc-relative relocations we want to allow them under PIC mode while
a normal global symbol defined in the same dynamic object can still bind
externally through copy relocation.  So, we should not allow pc-relative
relocation against such symbol.

SYMBOL_REFERENCES_LOCAL should be used and is more accurate than the original
individual checks.

bfd/
* elfnn-aarch64.c (elfNN_aarch64_final_link_relocate): Use
SYMBOL_REFERENCES_LOCAL.
ld/
* testsuite/ld-aarch64/aarch64-elf.exp: Update test name
* testsuite/ld-aarch64/pcrel.s: Add new testcases.
* testsuite/ld-aarch64/pcrel_pic_undefined.d: Update the expected
warnings.
* testsuite/ld-aarch64/pcrel_pic_defined_local.d: Rename ...
* testsuite/ld-aarch64/pcrel_pic_defined.d: ... to this.
Update expected warnings.

7 years agoarc: Select CPU model properly before disassembling
Anton Kolesov [Fri, 10 Feb 2017 11:11:55 +0000 (14:11 +0300)]
arc: Select CPU model properly before disassembling

Enforce CPU model for disassembler via its options, if it was specified in XML
target description, otherwise use default method of determining CPU implemented
in disassembler - scanning ELF private header.  The latter requires
disassemble_info->section to be properly initialized.  To make sure that
info->section is set in all cases this patch partially reverts [1] for ARC: it
reinstates arc_delayed_print_insn as a "print_insn" function for ARC, but
now this function only sets disassemble_info->section and then calls
default_print_insn to do the rest of the job.

Support for CPU in disassembler options for ARC has been added in [2].

[1] https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=39503f82427e22ed8e04d986ccdc8562091ec62e
[2] https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=10045478d984f9924cb945423388ba25b7dd3ffe

gdb/ChangeLog:

yyyy-mm-dd  Anton Kolesov  <anton.kolesov@synopsys.com>

* arc-tdep.c (arc_disassembler_options): New variable.
(arc_gdbarch_init): Set and use it. Use arc_delayed_print_insn instead
of default_print_insn.
(arc_delayed_print_insn): Set info->section when needed,
use default_print_insn to retrieve a disassembler.

7 years ago2017-06-15 Eric Christopher <echristo@gmail.com>
Eric Christopher [Fri, 16 Jun 2017 01:46:17 +0000 (18:46 -0700)]
2017-06-15  Eric Christopher  <echristo@gmail.com>

        * aarch64.cc: Fix a few typos and grammar-os.

7 years agoAutomatic date update in version.in
GDB Administrator [Fri, 16 Jun 2017 00:00:27 +0000 (00:00 +0000)]
Automatic date update in version.in

7 years agoCorrect "ld --help" to display "-z stack-size=SIZE"
H.J. Lu [Thu, 15 Jun 2017 19:30:53 +0000 (12:30 -0700)]
Correct "ld --help" to display "-z stack-size=SIZE"

ELF Linker command line option to set stack size is "-z stack-size=SIZE",
not "-z stacksize=SIZE".

* exsup.c (elf_shlib_list_options): Display "-z stack-size=SIZE"
instead of "-z stacksize=SIZE".

7 years ago[AArch64] Allow COPY relocation elimination
Jiong Wang [Thu, 15 Jun 2017 15:51:01 +0000 (16:51 +0100)]
[AArch64] Allow COPY relocation elimination

  As discussed at the PR, this patch tries to avoid COPY relocation generation
and propagate the original relocation into runtime if it was relocating on
writable section.  The ELIMINATE_COPY_RELOCS has been set to true and it's
underlying infrastructure has been improved so that the COPY reloc elimination
at least working on absoluate relocations (ABS64) on AArch64.

  BFD linker copy relocation elimination framwork requires the backend to always
allocate dynrelocs for all those relocation types that are possible to introduce
copy relocations.  This is for adjust_dynamic_symbol hook to be able to get all
symbol reference information.  Should one symbol is referenced by more than one
relocations, if there is any of them needs copy relocation then linker should
generate it.

bfd/
PR ld/21532
* elfnn-aarch64.c (ELIMINATE_COPY_RELOCS): Set to 1.
(elfNN_aarch64_final_link_relocate): Also propagate relocations to
runtime for if there needs copy relocation elimination.
(need_copy_relocation_p): New function.  Return true for symbol with
pc-relative references and if it's against read-only sections.
(elfNN_aarch64_adjust_dynamic_symbol): Use need_copy_relocation_p.
(elfNN_aarch64_check_relocs): Allocate dynrelocs for relocation types
that are related with accessing external objects.
(elfNN_aarch64_gc_sweep_hook): Sync the relocation types with the change
in elfNN_aarch64_check_relocs.

ld/
* testsuite/ld-aarch64/copy-reloc-exe-2.s: New test source file.
* testsuite/ld-aarch64/copy-reloc-2.d: New test.
* testsuite/ld-aarch64/copy-reloc-exe-eliminate.s: New test source file.
* testsuite/ld-aarch64/copy-reloc-eliminate.d: New test.
* testsuite/ld-aarch64/copy-reloc-so.s: Define new global objects.
* testsuite/ld-aarch64/aarch64-elf.exp: Run new tests.

7 years agoi386-dis: Add 2 tests with invalid bnd register
H.J. Lu [Thu, 15 Jun 2017 15:21:48 +0000 (08:21 -0700)]
i386-dis: Add 2 tests with invalid bnd register

PR binutils/21594
* testsuite/gas/i386/mpx.s: Add 2 tests with invalid bnd
register.
* testsuite/gas/i386/x86-64-mpx.s: Likewise.
* testsuite/gas/i386/mpx.d: Updated.
* testsuite/gas/i386/x86-64-mpx.d: Likewise.

7 years agoi386-dis: Check valid bnd register
H.J. Lu [Thu, 15 Jun 2017 13:40:17 +0000 (06:40 -0700)]
i386-dis: Check valid bnd register

Since there are only 4 bnd registers, return "(bad)" for register
number > 3.

PR binutils/21594
* i386-dis.c (OP_E_register): Check valid bnd register.
(OP_G): Likewise.

7 years agoPrevent address violation problem when disassembling corrupt aarch64 binary.
Nick Clifton [Thu, 15 Jun 2017 12:26:54 +0000 (13:26 +0100)]
Prevent address violation problem when disassembling corrupt aarch64 binary.

PR binutils/21595
* aarch64-dis.c (aarch64_ext_ldst_reglist): Check for an out of
range value.

7 years agoFix address violation parsing a corrupt ieee binary.
Nick Clifton [Thu, 15 Jun 2017 12:08:47 +0000 (13:08 +0100)]
Fix address violation parsing a corrupt ieee binary.

PR binutils/21581
(ieee_archive_p): Use a static buffer to avoid compiler bugs.

7 years agoAvoid a possible compiler bug by using a static buffer instead of a stack local buffer.
Nick Clifton [Thu, 15 Jun 2017 11:44:23 +0000 (12:44 +0100)]
Avoid a possible compiler bug by using a static buffer instead of a stack local buffer.

PR binutils/21582
* ieee.c (ieee_object_p): Use a static buffer to avoid compiler
bugs.

7 years agoFix address violation when disassembling a corrupt RL78 binary.
Nick Clifton [Thu, 15 Jun 2017 11:37:01 +0000 (12:37 +0100)]
Fix address violation when disassembling a corrupt RL78 binary.

PR binutils/21588
* rl78-decode.opc (OP_BUF_LEN): Define.
(GETBYTE): Check for the index exceeding OP_BUF_LEN.
(rl78_decode_opcode): Use OP_BUF_LEN as the length of the op_buf
array.
* rl78-decode.c: Regenerate.

7 years agoHandle EITR records in VMS Alpha binaries with overlarge command length parameters.
Nick Clifton [Thu, 15 Jun 2017 11:08:57 +0000 (12:08 +0100)]
Handle EITR records in VMS Alpha binaries with overlarge command length parameters.

PR binutils/21579
* vms-alpha.c (_bfd_vms_slurp_etir): Extend check of cmd_length.

7 years agoPrevent invalid array accesses when disassembling a corrupt bfin binary.
Nick Clifton [Thu, 15 Jun 2017 10:52:02 +0000 (11:52 +0100)]
Prevent invalid array accesses when disassembling a corrupt bfin binary.

PR binutils/21586
* bfin-dis.c (gregs): Clip index to prevent overflow.
(regs): Likewise.
(regs_lo): Likewise.
(regs_hi): Likewise.

7 years ago[GOLD] Don't install branch-to-stub for TLS relaxed ERRATUM 843419 sequences on AArch64
Jiong Wang [Thu, 15 Jun 2017 09:35:28 +0000 (10:35 +0100)]
[GOLD] Don't install branch-to-stub for TLS relaxed ERRATUM 843419 sequences on AArch64

TLS relaxation may change erratum 843419 sequences that those offending ADRP
instructions actually transformed into other instructions in which case there
is erratum 843419 risk anymore that we should avoid installing unnecessary
branch-to-stub.

gold/
        * aarch64.cc (Insn_utilities::is_mrs_tpidr_el0): New method.
        (AArch64_relobj<size, big_endian>::try_fix_erratum_843419_optimized):
        Return ture for some TLS relaxed sequences.

7 years agoPR gdb/21574: Mention $SHELL and startup-with-shell on "help run"
Sergio Durigan Junior [Wed, 14 Jun 2017 20:45:20 +0000 (16:45 -0400)]
PR gdb/21574: Mention $SHELL and startup-with-shell on "help run"

This simple patch updates the documentation of "help run" in order to
mention that the shell used to start the inferior comes from the
$SHELL environment variable.  It also mentions that this behaviour can
be disabled by using the "set startup-with-shell off" command.

gdb/ChangeLog:
2017-06-14  Sergio Durigan Junior  <sergiodj@redhat.com>

PR gdb/21574
* infcmd.c (_initialize_infcmd): Expand "help run" documentation
to mention $SHELL and startup-with-shell.

7 years agoAutomatic date update in version.in
GDB Administrator [Thu, 15 Jun 2017 00:00:48 +0000 (00:00 +0000)]
Automatic date update in version.in

7 years agoxtensa: don't expect XCHAL_* macros to be constant
Max Filippov [Wed, 17 May 2017 09:09:14 +0000 (02:09 -0700)]
xtensa: don't expect XCHAL_* macros to be constant

Get rid of the assumption that XCHAL_* macros are preprocessor
constants: don't use them in preprocessor conditionals or in static
variable initializers.

2017-06-14  Max Filippov  <jcmvbkbc@gmail.com>
bfd/
* elf32-xtensa.c (elf_xtensa_be_plt_entry,
elf_xtensa_le_plt_entry): Add dimension for the ABI to arrays,
keep both windowed and call0 ABI PLT definitions.
(elf_xtensa_create_plt_entry): Use selected ABI to choose upper
elf_xtensa_*_plt_entry endex.
(ELF_MAXPAGESIZE): Fix at minimal supported MMU page size.

gas/
* config/tc-xtensa.c (density_supported, xtensa_fetch_width,
absolute_literals_supported): Leave definitions uninitialized.
(directive_state): Leave entries for directive_density and
directive_absolute_literals initialized to false.
(xg_init_global_config, xtensa_init): New functions.
* config/tc-xtensa.h (TARGET_BYTES_BIG_ENDIAN): Define as 0.
(HOST_SPECIAL_INIT): New definition.
(xtensa_init): New declaration.

7 years agoMaxim Grigoriev stepping down as Xtensa maintainer
Max Filippov [Tue, 13 Jun 2017 17:26:55 +0000 (10:26 -0700)]
Maxim Grigoriev stepping down as Xtensa maintainer

gdb/
2017-06-14  Max Filippov  <jcmvbkbc@gmail.com>

* MAINTAINERS: Move Maxim Grigoriev to the Past Maintainers.

7 years agoFix seg-fault when trying to disassemble a corrupt score binary.
Nick Clifton [Wed, 14 Jun 2017 16:10:28 +0000 (17:10 +0100)]
Fix seg-fault when trying to disassemble a corrupt score binary.

PR binutils/21576
* score7-dis.c (score_opcodes): Add sentinel.

7 years agoFix address violation bug when disassembling a corrupt SH binary.
Nick Clifton [Wed, 14 Jun 2017 16:01:54 +0000 (17:01 +0100)]
Fix address violation bug when disassembling a corrupt SH binary.

PR binutils/21578
* elf32-sh.c (sh_elf_set_mach_from_flags): Fix check for invalid
flag value.

7 years agoFix address violation when disassembling a corrupt binary.
Nick Clifton [Wed, 14 Jun 2017 15:50:03 +0000 (16:50 +0100)]
Fix address violation when disassembling a corrupt binary.

PR binutils/21580
binutils * objdump.c (disassemble_bytes): Check for buffer overrun when
printing out rae insns.

ld * testsuite/ld-nds32/diff.d: Adjust expected output.

7 years agoDon't use print_insn_XXX in GDB
Yao Qi [Wed, 14 Jun 2017 15:28:30 +0000 (16:28 +0100)]
Don't use print_insn_XXX in GDB

This is a follow-up to

  [PATCH 0/6] Unify the disassembler selection in gdb and objdump
  https://sourceware.org/ml/binutils/2017-05/msg00192.html

that is, opcodes is able to select the right disassembler, so gdb
doesn't have to select them.  Instead, gdb can just use
default_print_insn.  As a result, these print_insn_XXX are not used
out of opcodes, so this patch also moves their declarations from
include/dis-asm.h to opcodes/disassemble.h.  With this change,
GDB doesn't use any print_insn_XXX directly any more.

gdb:

2017-06-14  Yao Qi  <yao.qi@linaro.org>

* aarch64-tdep.c (aarch64_gdb_print_insn): Call
default_print_insn instead of print_insn_aarch64.
* arm-tdep.c (gdb_print_insn_arm): Call
default_print_insn instead of print_insn_big_arm
and print_insn_little_arm.
* i386-tdep.c (i386_print_insn): Call default_print_insn
instead of print_insn_i386.
* ia64-tdep.c (ia64_print_insn): Call
default_print_insn instead of print_insn_ia64.
* mips-tdep.c (gdb_print_insn_mips): Call
default_print_insn instead of print_insn_big_mips
and print_insn_little_mips.
* spu-tdep.c (gdb_print_insn_spu): Call default_print_insn
instead of print_insn_spu.

include:

2017-06-14  Yao Qi  <yao.qi@linaro.org>

* dis-asm.h (print_insn_aarch64): Move it to opcodes/disassemble.h.
(print_insn_big_arm, print_insn_big_mips): Likewise.
(print_insn_i386, print_insn_ia64): Likewise.
(print_insn_little_arm, print_insn_little_mips): Likewise.
(print_insn_spu): Likewise.

opcodes:

2017-06-14  Yao Qi  <yao.qi@linaro.org>

* aarch64-dis.c: Include disassemble.h instead of dis-asm.h.
* arm-dis.c: Likewise.
* ia64-dis.c: Likewise.
* mips-dis.c: Likewise.
* spu-dis.c: Likewise.
* disassemble.h (print_insn_aarch64): New declaration, moved from
include/dis-asm.h.
(print_insn_big_arm, print_insn_big_mips): Likewise.
(print_insn_i386, print_insn_ia64): Likewise.
(print_insn_little_arm, print_insn_little_mips): Likewise.