When loading a core without an executable like so:
$ gdb --core core
for example often the gdbarch won't contain the
iterate_over_regset_sections method. For example on ARM.
This will generate a call to get_core_register_section with a NULL regset
like at corelow.c:628
get_core_register_section (regcache, NULL, ".reg", 0, 0, "general-purpose", 1);
However a check for REGSET_VARIABLE_SIZE in get_core_register_section
assumes that regset is != NULL thus leading to a crash with this backtrace:
(gdb) bt
#0 0x000000000065907b in get_core_register_section
(regcache=regcache@entry=0x2c26260, regset=regset@entry=0x0,
name=name@entry=0xdbf7b2 ".reg", min_size=min_size@entry=0,
which=which@entry=0, human_name=human_name@entry=0xdbac28
"general-purpose", required=1)
at ../../gdb/corelow.c:542
#1 0x0000000000659b70 in get_core_registers (ops=<optimized out>,
regcache=0x2c26260, regno=<optimized out>) at ../../gdb/corelow.c:628
#2 0x000000000076e5fb in target_fetch_registers
(regcache=regcache@entry=0x2c26260, regno=regno@entry=15)
at ../../gdb/target.c:3590
Note that commit: f962539ad2 ("Warn if core file register
section is larger than expected") introduced this issue.
Thus releases > 7.8.2 are affected.
Also, this would have been caught by gdb.base/corefile.exp but the
problem is that this triggers only if the core dump is missing some data
so that it's not recognized as a linux core dump, or it's not a linux core
dump and the core file register section is larger than expected.
So if you just create a core and read it on linux with ARM the osabi is
detected properly and iterate_over_regset_sections is present and so the
problem is not triggered.
Thus creating a linux test for this with a crafted core that meets the
problem requirements is non-trivial.
This patch fixes this crash by adding a check for regset existence before
running the condition.
gdb/ChangeLog:
* corelow.c (get_core_register_section): Check for regset
existence before checking for REGSET_VARIABLE_SIZE.
Hi,
GDB calls some APIs from opcodes to do disassembly and provide some
call backs. This model makes troubles on C++ exception unwinding,
because GDB is a C++ program, and opcodes is still compiled as C.
As we can see, frame #10 and #12 are C++, while #frame 11 is C,
#10 0x0000000000544228 in memory_error (err=TARGET_XFER_E_IO, memaddr=<optimized out>) at ../../binutils-gdb/gdb/corefile.c:237
#11 0x00000000006b0a54 in print_insn_aarch64 (pc=0, info=0xffffffffeeb0) at ../../binutils-gdb/opcodes/aarch64-dis.c:3185
#12 0x0000000000553590 in gdb_pretty_print_insn (gdbarch=gdbarch@entry=0xbbceb0, uiout=uiout@entry=0xbc73d0, di=di@entry=0xffffffffeeb0,
insn=0xffffffffed40, insn@entry=0xffffffffed90, flags=flags@entry=0,
C++ exception unwinder can't go across frame #11 unless it has
unwind table. However, C program on many architectures doesn't
have it in default. As a result, GDB aborts, which is described
in PR 20939.
This is not the first time we see this kind of problem. We've
had a commit 89525768cd
"Propagate GDB/C++ exceptions across readline using sj/lj-based TRY/CATCH".
We can fix the disassembly bug in a similar way, this is the option one.
Since opcodes is built with gdb, we fix this problem in a different
way as we did for the same issue with readline. Instead of throwing
exception in dis_asm_memory_error, we record the failed memory
address, and throw exception when GDB returns from opcodes disassemblers.
gdb:
2017-01-26 Yao Qi <yao.qi@linaro.org>
Pedro Alves <palves@redhat.com>
PR gdb/20939
* disasm.c (gdb_disassembler::dis_asm_memory_error): Don't
call memory_error, save memaddr instead.
(gdb_disassembler::print_insn): If gdbarch_print_insn returns
negative, cal memory_error.
* disasm.h (gdb_disassembler) <m_err_memaddr>: New field.
gdb/testsuite:
2017-01-26 Yao Qi <yao.qi@linaro.org>
* gdb.base/all-architectures.exp.in (do_arch_tests): Test
disassemble on address 0.
This patch adds a unit test about memory error occurs on reading
memory, and check MEMORY_ERROR exception is always thrown.
gdb:
2017-01-26 Yao Qi <yao.qi@linaro.org>
* disasm-selftests.c (memory_error_test): New function.
(_initialize_disasm_selftests): Register memory_error_test.
This patch adds one unit test, which disassemble one instruction for
every gdbarch if available. The test needs one valid instruction of
each gdbarch, and most of them are got from breakpoint instruction.
For the rest gdbarch whose breakpoint instruction isn't a valid
instruction, I copy one instruction from the gas/testsuite/gas/
directory.
I get the valid instruction of most gdbarch except ia64, mep, mips,
tic6x, and xtensa. People familiar with these arch should be easy
to extend the test.
In order to achieve "do the unit test for every gdbarch", I add
selftest-arch.[c,h], so that we can register a function pointer,
which has one argument gdbarch. selftest.c will iterate over all
gdbarches to call the registered function pointer.
gdb:
2017-01-26 Yao Qi <yao.qi@linaro.org>
* Makefile.in (SFILES): Add disasm-selftests.c and
selftest-arch.c.
(COMMON_OBS): Add disasm-selftests.o and selftest-arch.o.
* disasm-selftests.c: New file.
* selftest-arch.c: New file.
* selftest-arch.h: New file.
opcodes/mep-dis.c:mep_print_insn has already had the code to
handle the case when info->section is NULL,
/* Picking the right ISA bitmask for the current context is tricky. */
if (info->section)
{
}
else /* sid or gdb */
{
}
so that we can still cal print_insn_mep even section can't be found.
On the other hand, user can disassemble an arbitrary address which
doesn't map to any section at all.
gdb:
2017-01-26 Yao Qi <yao.qi@linaro.org>
* mep-tdep.c (mep_gdb_print_insn): Set info->arch
to bfd_arch_mep. Don't return 0 if section is not
found. Call print_insn_mep.
This patch addes class gdb_disassembler, and refactor
code to use it. The gdb_disassembler object is saved
in disassember_info.application_data. However,
disassember_info.application_data is already used by
gdb for arm, mips spu, and scm-disasm. In arm and mips,
.application_data is gdbarch, but we can still get gdbarch
from gdb_disassember.
The use of application_data in spu is a little bit
complicated. It creates its own disassemble_info, and
save spu_dis_asm_data in .application_data. This will
overwrite the pointer to gdb_disassembler, so we need
to find another place to save spu_dis_asm_data. I
extend disassemble_info, and put "id" there.
gdb:
2017-01-26 Pedro Alves <palves@redhat.com>
Yao Qi <yao.qi@linaro.org>
* arm-tdep.c: Include "disasm.h".
(gdb_print_insn_arm): Update code to get gdbarch.
* disasm.c (dis_asm_read_memory): Change it to
gdb_disassembler::dis_asm_read_memory.
(dis_asm_memory_error): Likewise.
(dis_asm_print_address): Likewise.
(gdb_pretty_print_insn): Change it to
gdb_disassembler::pretty_print_insn.
(dump_insns): Add one argument gdb_disassemlber. All
callers updated.
(do_mixed_source_and_assembly_deprecated): Likewise.
(do_mixed_source_and_assembly): Likewise.
(do_assembly_only): Likewise.
(gdb_disassembler::gdb_disassembler): New.
(gdb_disassembler::print_insn): New.
* disasm.h (class gdb_disassembler): New.
(gdb_pretty_print_insn): Remove declaration.
(gdb_disassemble_info): Likewise.
* guile/scm-disasm.c (class gdbscm_disassembler): New.
(gdbscm_disasm_read_memory_worker): Update.
(gdbscm_disasm_read_memory): Update.
(gdbscm_disasm_memory_error): Remove.
(gdbscm_disasm_print_address): Remove.
(gdbscm_disassembler::gdbscm_disassembler): New.
(gdbscm_print_insn_from_port): Update.
* mips-tdep.c: Include disasm.h.
(gdb_print_insn_mips): Update code to get gdbarch.
* record-btrace.c (btrace_insn_history): Update.
* spu-tdep.c: Include disasm.h.
(struct spu_dis_asm_data): Remove.
(struct spu_dis_asm_info): New.
(spu_dis_asm_print_address): Use spu_dis_asm_info to get
SPU id.
(gdb_print_insn_spu): Cast disassemble_info to
spu_dis_asm_info.
This patch adds a new function null_stream, which returns a null
stream. The null stream can be used in multiple places. It is
used in gdb_insn_length, and the following patches will use it too.
gdb:
2017-01-26 Yao Qi <yao.qi@linaro.org>
* disasm.c (do_ui_file_delete): Delete.
(gdb_insn_length): Move code creating stream to ...
* utils.c (null_stream): ... here. New function.
* utils.h (null_stream): Declare.
This patch adds a DW_OP_implicit_value in dwarf assembler, and uses
dwarf assembler in implptr-64bit.exp. Using dwarf assembler in
implptr-64bit.exp exposes some limitations in dwarf assembler,
- some variables are not evaluated in the caller's context, so we
can not pass variable to assembler, like this
Dwarf::assemble $asm_file {
cu {
version $dwarf_version
addr_size $addr_size
is_64 $is_64
} {
}
and
{DW_AT_type :$struct_label "DW_FORM_ref$ref_addr_size"}
this limitation is fixed by adding "uplevel" and "subst".
- dwarf assembler doesn't emit DW_FORM_ref_addr for label referencing.
this limitation is fixed by adding a new character "%",
{ type %$int_label }
this means we want to emit DW_FORM_ref_addr for label referencing.
- we can't set the form of label referencing offset in dwarf assembler.
Nowadays, dwarf assembler guesses the form of labels, which is
DW_FORM_ref4. However, in implptr-64bit.exp, both DW_FORM_ref4
and DW_FORM_ref8 is used (see REF_ADDR in implptr-64bit.S). This
patch adds the flexibility of setting the form of label reference.
Both of them below are valid,
{DW_AT_type :$struct_label}
{DW_AT_type :$struct_label DW_FORM_ref8}
the former form is the default DW_FORM_ref4.
I compared the .debug_info of objects without and with this patch
applied. There is no changes except abbrev numbers.
gdb/testsuite:
2017-01-25 Andreas Arnez <arnez@linux.vnet.ibm.com>
Yao Qi <yao.qi@linaro.org>
* gdb.dwarf2/implptr-64bit.exp: Use dwarf assembler.
* gdb.dwarf2/implptr-64bit.S: Remove.
* lib/dwarf.exp (Dwarf): Handle character "%". Evaluate some
variables in caller's context. Add DW_OP_implicit_value.
DW_OP_GNU_implicit_pointer refers to a DIE with an offset of different
sizes in different dwarf versions. In v2, the size is the pointer size,
while in v3 and above, it is the ref_addr size. This patch fixes
dwarf assembler to emit the correct size of offset. We've already fixed
this size issue in gdb,
https://sourceware.org/ml/gdb-patches/2011-09/msg00451.html
gdb/testsuite:
2017-01-25 Yao Qi <yao.qi@linaro.org>
* lib/dwarf.exp (Dwarf::_location): Handle
DW_OP_GNU_implicit_pointer with proper size.
bfd/
* config.bfd (powerpcle-*-rtems*): Do not mark as removed.
(arm-*-rtems*): Move to (arm*-*-eabi*).
(i[3-7]86-*-rtems*): Move to (i[3-7]86-*-elf*).
(m68-*-rtems*): Move to (m68*-*-elf*).
ld/
* configure.tgt (arm-*-rtems*): Move to (arm*-*-eabi*).
(bfin-*-rtems*): Move to (bfin*-*-elf*).
(i[3-7]86-*-rtems*): Move to (i[3-7]86*-*-elf*).
(m68*-*-rtems*): Move to (m68*-*-elf*).
(mips*-*-rtems*): Move to (mips*-*-elf*).
(or1k*-*-rtems*): Move to (or1k*-*-elf*).
(powerpc*-*-rtems*): Move to (powerpc*-*-elf*).
(sparc*-*-rtems*): Move to (sparc*-*-elf*).
(sparc64*-*-rtems*): Move to (sparc64*-*-elf*).
The use of te-linux.h is unnecessary since the TE_LINUX define is unused
and LOCAL_LABELS_FB is defined to 1 in tc-riscv.h as well.
gas/
* configure.tgt (riscv*-*-*): Remove em=linux.
Complement commit 81ff47b3a5 ("PR ld/20828: Fix linker script symbols
wrongly forced local with section GC") and move the symbol sweep stage
of section GC from `elf_gc_sweep' to `bfd_elf_size_dynamic_sections',
avoiding the need to clear the `forced_local' marker, problematic for
targets that have special processing in their `elf_backend_hide_symbol'
handler. Set `mark' instead in `bfd_elf_record_link_assignment' and,
matching changes from commit 3bd43ebcb6 ("ld --gc-sections fail with
__tls_get_addr_opt"), also in PowerPC `__tls_get_addr_opt' handling
code, removing a:
FAIL: PR ld/20828 dynamic symbols with section GC (version script)
test suite failure with the `score-elf' target.
The rationale is it is enough if symbols are swept at the beginning of
`bfd_elf_size_dynamic_sections' as it is only in this function that the
size of the GOT, the dynamic symbol table and other dynamic sections is
determined, which will depend on the number of symbols making it to the
dynamic symbol table. It is also appropriate to do the sweep at this
point as it is already after any changes have been made to symbols with
`bfd_elf_record_link_assignment', and not possible any earlier as calls
to that function are only made just beforehand -- barring audit entry
processing -- via `gld${EMULATION_NAME}_find_statement_assignment'
invoked from `gld${EMULATION_NAME}_before_allocation' which is the ELF
handler for `ldemul_before_allocation'.
bfd/
PR ld/20828
* elflink.c (bfd_elf_record_link_assignment): Revert last
change and don't ever clear `forced_local'. Set `mark'
unconditionally.
(elf_gc_sweep_symbol_info, elf_gc_sweep_symbol): Reorder within
file.
(elf_gc_sweep): Move the call to `elf_gc_sweep_symbol'...
(bfd_elf_size_dynamic_sections): ... here.
* elf32-ppc.c (ppc_elf_tls_setup): Don't clear `forced_local'
and set `mark' instead in `__tls_get_addr_opt' processing.
* elf64-ppc.c (ppc64_elf_tls_setup): Likewise.
Correct a bug in Solaris 2 linker emulation code triggered by a test
introduced with commit 81ff47b3a5 ("PR ld/20828: Fix linker script
symbols wrongly forced local with section GC") and only create implicit
version nodes if versioning is actually introduced with a version script
(or VERSION command) rather than only global vs local symbol visibility
selected, fixing an:
ld: anonymous version tag cannot be combined with other version tags
linker error produced whenever a version script (or VERSION command) is
used that does not assign symbol versions, such as:
{ global: foo; bar; local: *; };
and consequently removing a:
FAIL: PR ld/20828 dynamic symbols with section GC (version script)
test suite failure with the `x86_64-solaris2' target.
ld/
* emultempl/solaris2.em (elf_solaris2_before_allocation): Do not
add implicit version nodes if an anonymous version tag is being
used.
This patch fixes a number of issues with powerpc dynamic relocations.
1) Both ppc and ppc64 were emitting more dynamic symbols and
relocations than necessary, due to not supporting static linker
resolution of tls_index entries for __tls_get_addr_opt. This meant
that any @got@tlsgd or @got@tlsld reloc needed to make their symbols
dynamic and generate dptmod and dtprel relocs for the dynamic linker.
That would have been passable, but what happened was that practically
all @got relocations resulted in their symbols being made dynamic and
dynamic relocations emitted against the GOT entries. (Mostly visible
on ppc32 executables since ppc64 gcc really only uses @got style
relocs for TLS.)
2) The PowerOpen syntax was not supported with __tls_get_addr_opt.
DTPMOD/DTPREL relocs on tls_index TOC entries did not use the trick of
forcing dynamic symbols and relocations so those entries always
resulted in the full __tls_get_addr processing. gcc doesn't use the
PowerOpen syntax for TLS, and normally such code would be optimized to
TLS IE or LE so the impact of missing this support was minimal.
3) In an executable, relocations against GNU indirect functions always
used the value of their PLT stub. While this is correct, it is
better in some cases to use a dynamic relocation. An extra dynamic
relocation can mean that calls via function pointers need not bounce
through the PLT stub at runtime.
The patch also tidies the PLT handling code in ppc32
allocate_dynrelocs. Allocating PLT entries after other dynamic relocs
allows the PLT loop to omit special handling for undefined weak
symbols, and that in turn allows the loop to be simplified.
bfd/
* elf32-ppc.c (ppc_elf_adjust_dynamic_symbol): Merge two cases
where dynamic relocs are preferable. Allow ifunc too.
(ensure_undefweak_dynamic): New function.
(allocate_dynrelocs): Use it here. Move plt handling last and
don't make symbols dynamic, simplifying loop. Only make undef
weak symbols with GOT entries dynamic. Correct condition
for GOT relocs. Handle dynamic relocs on ifuncs. Correct
comments. Remove goto.
(ppc_elf_relocate_section): Correct test for using dynamic
symbol on GOT relocs. Rearrange test for emitting GOT relocs
to suit. Set up explicit tls_index entries and implicit GOT
tls_index entries resolvable at link time for
__tls_get_addr_opt. Simplify test to clear mem for prelink.
* elf64-ppc.c (allocate_got): Correct condition for GOT relocs.
(ensure_undefweak_dynamic): New function.
(allocate_dynrelocs): Use it here. Only make undef weak symbols
with GOT entries dynamic. Remove unnecessary test of
WILL_CALL_FINISH_DYNAMIC_SYMBOL in PLT handling.
(ppc64_elf_relocate_section): Correct test for using dynamic
symbol on GOT relocs. Rearrange test for emitting GOT relocs
to suit. Set up explicit tls_index entries and implicit GOT
tls_index entries resolvable at link time for __tls_get_addr_opt.
Simplify expression to clear mem for prelink.
ld/
* testsuite/ld-powerpc/tlsexe.r: Update for fewer dynamic relocs
and symbols.
* testsuite/ld-powerpc/tlsexe.d: Likewise.
* testsuite/ld-powerpc/tlsexe.g: Likewise.
gold/
* mips.cc (Mips_output_data_plt::rel_plt): Remove const from return
type.
(Target_mips::make_plt_entry): Make the sh_info field of .rel.plt
point to .plt.
gold/
PR gold/21054
* mips.cc (Mips_got_info::record_global_got_symbol): Don't add symbol
to the dynamic symbol table if it is forced to local visibility.
(Target_mips::do_finalize_sections): Don't add __RLD_MAP symbol to the
dynamic symbol table if it is forced to local visibility.
Since the reference to the Inferior Python object is managed by
gdbpy_ref (RAII), we can return directly from the loop. It's just a
leftover from the cleanups era.
gdb/ChangeLog:
* python/py-inferior.c (find_thread_object): Return directly
from the loop. Remove "found" variable.
bfd * elfnn-aarch64.c: Fix relaxations for ILP32 mode.
ld * testsuite/ld-aarch64/aarch64-elf.exp: Run new tests.
* testsuite/ld-aarch64/tls-desc-ie-ilp32.d: New test.
* testsuite/ld-aarch64/tls-relax-all-ilp32.d: New test.
* testsuite/ld-aarch64/tls-relax-gd-le-ilp32.d: New test.
* testsuite/ld-aarch64/tls-relax-gdesc-le-2-ilp32.d: New test.
* testsuite/ld-aarch64/tls-relax-gdesc-le-ilp32.d: New test.
* testsuite/ld-aarch64/tls-relax-ie-le-2-ilp32.d: New test.
* testsuite/ld-aarch64/tls-relax-ie-le-3-ilp32.d: New test.
* testsuite/ld-aarch64/tls-relax-ie-le-ilp32.d: New test.
* testsuite/ld-aarch64/tls-tiny-desc-ie-ilp32.d: New test.
* testsuite/ld-aarch64/tls-tiny-desc-le-ilp32.d: New test.
* testsuite/ld-aarch64/tls-tiny-gd-ie-ilp32.d: New test.
* testsuite/ld-aarch64/tls-tiny-gd-le-ilp32.d: New test.
* configure.ac: Restore old behaviour of only enabling multilibs
when a target subdirectory is defined. This allows building with
srcdir == builddir.
* configure: Regenerate.
* configure.ac: Update year in copyright notice.
Sync from FSF GCC mainline, bringing in the following patches.
* Makefile.def: Likewise.
* Makefile.tpl: Likewise.
* configure: Regenerate.
* Makefile.in: Regenerate.
2016-12-21 Jakub Jelinek <jakub@redhat.com>
* configure.ac: Don't bootstrap libmpx unless --with-build-config
includes bootstrap-mpx.
2016-12-01 Matthias Klose <doko@ubuntu.com>
* configure.ac: Don't use pkg-config to check for bdw-gc.
2016-11-30 Matthias Klose <doko@ubuntu.com>
* Makefile.def: Remove reference to boehm-gc target module.
* configure.ac: Include pkg.m4, check for --with-target-bdw-gc
options and for the bdw-gc pkg-config module.
2016-11-15 Matthias Klose <doko@ubuntu.com>
* config-ml.in: Remove references to GCJ.
* configure.ac: Likewise.
2016-09-30 Jakub Jelinek <jakub@redhat.com>
* configure.ac: Add target-libffi to target_libraries.
Readd libgcj target disablings, modified to only target-libffi.
Readd target addition of go to unsupported languages.
2016-09-30 Andrew Haley <aph@redhat.com>
* Makefile.def: Remove libjava.
* Makefile.tpl: Likewise.
* configure.ac: Likewise.
2016-09-26 Anton Kolesov <Anton.Kolesov@synopsys.com>
* configure.ac: Disable "sim" directory for arc*-*-*.
2016-09-12 Maciej W. Rozycki <macro@imgtec.com>
* configure.ac: Check for the minimum in-tree MPFR version
handled.
Complement commit 81ff47b3a5 ("PR ld/20828: Fix linker script symbols
wrongly forced local with section GC") and add `.plt' to the list of
output sections created, fixing:
FAIL: PR ld/20828 dynamic symbols with section GC (auxiliary shared library)
FAIL: PR ld/20828 dynamic symbols with section GC (plain)
FAIL: PR ld/20828 dynamic symbols with section GC (version script)
failures with `riscv32-elf', `riscv32-linux', `riscv64-elf' and
`riscv64-linux' targets caused by LD crashing in the absence of such a
section.
ld/
PR ld/20828
* testsuite/ld-elf/pr20828.ld: Add `.plt'.
Complement commit 81ff47b3a5 ("PR ld/20828: Fix linker script symbols
wrongly forced local with section GC") and remove the leading underscore
from `_fdata' and `_edata' symbols used in tests, fixing a:
FAIL: PR ld/20828 dynamic symbols with section GC (version script)
failure with targets such as: `bfin-elf', `bfin-uclinux', `metag-elf',
`metag-linux' `mn10300-elf', `sh-elf', `sh64-elf', and possibly other
ones, that have `_' set (with `elf_symbol_leading_char') as the leading
character for symbols. As from commit 93252b1cf4 ("bfd/ld: handle ABI
prefixes in version scripts") these targets strip the leading underscore
before applying version script rules, because the (default) syntax for
symbol names is that of the C language rather than their low-level
symbol table encoding.
ld/
PR ld/20828
* testsuite/ld-elf/pr20828.ld: Rename `_fdata' and `_edata' to
`fdata' and `edata' respectively.
* testsuite/ld-elf/pr20828.ver: Adjust accordingly.
* testsuite/ld-elf/pr20828-a.sd: Likewise.
* testsuite/ld-elf/pr20828-b.sd: Likewise.
* testsuite/ld-elf/pr20828-c.sd: Likewise.
Complement commit 81ff47b3a5 ("PR ld/20828: Fix linker script symbols
wrongly forced local with section GC") and make tests check for the
presence of global `_fdata' and `_edata' symbols separately, removing
any dependency on symbol table ordering for tests to succeed and
removing:
FAIL: PR ld/20828 dynamic symbols with section GC (auxiliary shared library)
FAIL: PR ld/20828 dynamic symbols with section GC (plain)
failures with the `x86_64-solaris2' target, which has additional
intervening entries:
Symbol table '.dynsym' contains 6 entries:
Num: Value Size Type Bind Vis Ndx Name
0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND
1: 0000000000000000 0 NOTYPE GLOBAL DEFAULT 1 _fdata
2: 0000000000000000 0 OBJECT GLOBAL DEFAULT 1 _DYNAMIC
3: 0000000000000000 0 OBJECT GLOBAL DEFAULT ABS _PROCEDURE_LINKAGE_TABLE_
4: 0000000000000000 0 NOTYPE GLOBAL DEFAULT 1 _edata
5: 00000000000001b8 0 OBJECT GLOBAL DEFAULT 4 _GLOBAL_OFFSET_TABLE_
Rename dump pattern files accordingly for consistency.
ld/
PR ld/20828
* testsuite/ld-elf/pr20828-1.sd: Remove test.
* testsuite/ld-elf/pr20828-a.sd: New test.
* testsuite/ld-elf/pr20828-2a.sd: Rename test to...
* testsuite/ld-elf/pr20828-b.sd: ... this.
* testsuite/ld-elf/pr20828-2b.sd: Rename test to...
* testsuite/ld-elf/pr20828-c.sd: ... this.
* testsuite/ld-elf/shared.exp: Adjust accordingly.
ld/
* testsuite/lib/ld-lib.exp (run_ld_link_exec_tests): Append
board_cflags as gcc is used as linker driver.
* testsuite/ld-unique/unique.exp: Likewise
When the gdbpy_ref objects get destroyed, they call Py_DECREF to
decrement the reference counter of the python object they hold a
reference to. Any time we call into the Python API, we should be
holding the GIL. The gdbpy_enter object does that for us in an
RAII-fashion.
However, if gdbpy_enter is declared after a gdbpy_ref object in a
function, gdbpy_enter's destructor will be called (and the GIL will be
released) before gdbpy_ref's destructor is called. Therefore, we will
end up calling Py_DECREF without holding the GIL.
This became obvious with Python 3.6, where memory management functions
have asserts to make sure that the GIL is held. This was exposed by
tests py-as-string.exp, py-function.exp and py-xmethods. For example:
(gdb) p $_as_string(enum_valid)
Fatal Python error: Python memory allocator called without holding the GIL
Current thread 0x00007f7f7b21c780 (most recent call first):
[1] 18678 abort (core dumped) ./gdb -nx testsuite/outputs/gdb.python/py-as-string/py-as-string
#0 0x00007ffff618bc37 in raise () from /lib/x86_64-linux-gnu/libc.so.6
#1 0x00007ffff618f028 in abort () from /lib/x86_64-linux-gnu/libc.so.6
#2 0x00007ffff6b104d6 in Py_FatalError (msg=msg@entry=0x7ffff6ba15b8 "Python memory allocator called without holding the GIL") at Python/pylifecycle.c:1457
#3 0x00007ffff6a37a68 in _PyMem_DebugCheckGIL () at Objects/obmalloc.c:1972
#4 0x00007ffff6a3804e in _PyMem_DebugFree (ctx=0x7ffff6e65290 <_PyMem_Debug+48>, ptr=0x24f8830) at Objects/obmalloc.c:1994
#5 0x00007ffff6a38e1d in PyMem_Free (ptr=<optimized out>) at Objects/obmalloc.c:442
#6 0x00007ffff6b866c6 in _PyFaulthandler_Fini () at ./Modules/faulthandler.c:1369
#7 0x00007ffff6b104bd in Py_FatalError (msg=msg@entry=0x7ffff6ba15b8 "Python memory allocator called without holding the GIL") at Python/pylifecycle.c:1431
#8 0x00007ffff6a37a68 in _PyMem_DebugCheckGIL () at Objects/obmalloc.c:1972
#9 0x00007ffff6a3804e in _PyMem_DebugFree (ctx=0x7ffff6e652c0 <_PyMem_Debug+96>, ptr=0x7ffff46b6040) at Objects/obmalloc.c:1994
#10 0x00007ffff6a38f55 in PyObject_Free (ptr=<optimized out>) at Objects/obmalloc.c:503
#11 0x00007ffff6a5f27e in unicode_dealloc (unicode=unicode@entry=0x7ffff46b6040) at Objects/unicodeobject.c:1794
#12 0x00007ffff6a352a9 in _Py_Dealloc (op=0x7ffff46b6040) at Objects/object.c:1786
#13 0x000000000063f28b in gdb_Py_DECREF (op=0x7ffff46b6040) at /home/emaisin/src/binutils-gdb/gdb/python/python-internal.h:192
#14 0x000000000063fa33 in gdbpy_ref_policy::decref (ptr=0x7ffff46b6040) at /home/emaisin/src/binutils-gdb/gdb/python/py-ref.h:35
#15 0x000000000063fa77 in gdb::ref_ptr<_object, gdbpy_ref_policy>::~ref_ptr (this=0x7fffffffcdf0, __in_chrg=<optimized out>) at /home/emaisin/src/binutils-gdb/gdb/common/gdb_ref_ptr.h:91
#16 0x000000000064d8b8 in fnpy_call (gdbarch=0x2b50010, language=0x115d2c0 <c_language_defn>, cookie=0x7ffff46b7468, argc=1, argv=0x7fffffffcf48)
at /home/emaisin/src/binutils-gdb/gdb/python/py-function.c:145
The fix is to place the gdbpy_enter first in the function. I also
cleaned up the comments a bit and removed the unnecessary initialization
of the value variable.
gdb/ChangeLog:
* python/py-function.c (fnpy_call): Reorder declarations to have
the gdbpy_enter object declared first.
* python/py-xmethods.c (gdbpy_get_xmethod_arg_types): Likewise.
New in v2:
- Define PyMem_RawMalloc as PyMem_Malloc for Python < 3.4 and use
PyMem_RawMalloc in the code.
Since Python 3.4, the callback installed in PyOS_ReadlineFunctionPointer
should return a value allocated with PyMem_RawMalloc instead of
PyMem_Malloc. The reason is that PyMem_Malloc must be called with the
Python Global Interpreter Lock (GIL) held, which is not the case in the
context where this function is called. PyMem_RawMalloc was introduced
for cases like this.
In Python 3.6, it looks like they added an assert to verify that
PyMem_Malloc was not called without the GIL. The consequence is that
typing anything in the python-interactive mode of gdb crashes the
process. The same behavior was observed with the official package on
Arch Linux as well as with a manual Python build on Ubuntu 14.04.
This is what is shown with a debug build of Python 3.6 (the error with a
non-debug build is far less clear):
(gdb) pi
>>> print(1)
Fatal Python error: Python memory allocator called without holding the GIL
Current thread 0x00007f1459af8780 (most recent call first):
[1] 21326 abort ./gdb
and the backtrace:
#0 0x00007ffff618bc37 in raise () from /lib/x86_64-linux-gnu/libc.so.6
#1 0x00007ffff618f028 in abort () from /lib/x86_64-linux-gnu/libc.so.6
#2 0x00007ffff6b104d6 in Py_FatalError (msg=msg@entry=0x7ffff6ba15b8 "Python memory allocator called without holding the GIL") at Python/pylifecycle.c:1457
#3 0x00007ffff6a37a68 in _PyMem_DebugCheckGIL () at Objects/obmalloc.c:1972
#4 0x00007ffff6a3804e in _PyMem_DebugFree (ctx=0x7ffff6e65290 <_PyMem_Debug+48>, ptr=0x24f8830) at Objects/obmalloc.c:1994
#5 0x00007ffff6a38e1d in PyMem_Free (ptr=<optimized out>) at Objects/obmalloc.c:442
#6 0x00007ffff6b866c6 in _PyFaulthandler_Fini () at ./Modules/faulthandler.c:1369
#7 0x00007ffff6b104bd in Py_FatalError (msg=msg@entry=0x7ffff6ba15b8 "Python memory allocator called without holding the GIL") at Python/pylifecycle.c:1431
#8 0x00007ffff6a37a68 in _PyMem_DebugCheckGIL () at Objects/obmalloc.c:1972
#9 0x00007ffff6a37aa3 in _PyMem_DebugMalloc (ctx=0x7ffff6e65290 <_PyMem_Debug+48>, nbytes=5) at Objects/obmalloc.c:1980
#10 0x00007ffff6a38d91 in PyMem_Malloc (size=<optimized out>) at Objects/obmalloc.c:418
#11 0x000000000064dbe2 in gdbpy_readline_wrapper (sys_stdin=0x7ffff6514640 <_IO_2_1_stdin_>, sys_stdout=0x7ffff6514400 <_IO_2_1_stdout_>, prompt=0x7ffff4d4f7d0 ">>> ")
at /home/emaisin/src/binutils-gdb/gdb/python/py-gdb-readline.c:75
The documentation is very clear about it [1] and it was also mentioned
in the "What's New In Python 3.4" page [2].
[1] https://docs.python.org/3/c-api/veryhigh.html#c.PyOS_ReadlineFunctionPointer
[2] https://docs.python.org/3/whatsnew/3.4.html#changes-in-the-c-api
gdb/ChangeLog:
* python/python-internal.h (PyMem_RawMalloc): Define for
Python < 3.4.
* python/py-gdb-readline.c (gdbpy_readline_wrapper): Use
PyMem_RawMalloc instead of PyMem_Malloc.
