Matthew Wahab [Fri, 27 Nov 2015 15:02:26 +0000 (15:02 +0000)]
[AArch64] Let aliased instructions be their preferred form.
Although the AArch64 backend supports aliased instructions, the aliasing
forms are always preferred over the real instruction. This makes it
awkward to handle instructions which have aliases but which are their
own preferred form.
This patch includes the instruction being aliased in the list of
alternatives which is searched when considering which form to use.
opcodes/
2015-11-27 Matthew Wahab <matthew.wahab@arm.com>
* aarch64-asm-2.c: Regenerate.
* aarch64-dis-2.c: Regenerate.
* aarch64-dis.c: Weaken assert.
* aarch64-gen.c: Include the instruction in the list of its
possible aliases.
Matthew Wahab [Fri, 27 Nov 2015 13:44:10 +0000 (13:44 +0000)]
[Aarch64] Support an ARMv8.2 system register.
ARMv8.2 adds a new system register id_aa64mmfr2_el1. This patch adds
support for the register to binutils, making it available when
-march=armv8.2-a is selected.
opcodes/
2015-11-27 Matthew Wahab <matthew.wahab@arm.com>
Alan Modra [Mon, 7 Dec 2015 03:22:01 +0000 (13:52 +1030)]
PowerPC ifunc with local symbols
This fixes some cases where the linker would incorrectly error on plt
relocs to local ifunc symbols. I've also tidied plt and ifunc
handling for ppc64, where check_relocs was allowing for the
possibility of plt calls via addr14/addr24 relocs but relocate_section
was not.
* elf32-ppc.c (ppc_elf_check_relocs): Don't error on local ifunc
plt call. Wrap long lines.
(ppc_elf_relocate_section): Wrap long lines.
* elf64-ppc.c (ppc64_elf_check_relocs): Don't error on local ifunc
plt calls. Move __tls_get_addr checks later. Don't create plt
for addr14/addr24 relocs.
(ppc64_elf_gc_sweep_hook): Adjust to suit check_relocs changes.
(ppc64_elf_relocate_section): Correct local ifunc handling for
PLT64, PLT32 and PLT16 relocs.
Alan Modra [Mon, 7 Dec 2015 03:11:36 +0000 (13:41 +1030)]
PR19323 memory allocation greater than 4G
On 32-bit targets, memory requested for program/section headers on a
fuzzed binary can wrap to 0. A bfd_alloc of zero bytes actually
returns a one byte allocation rather than a NULL pointer. This then
leads to buffer overflows.
Making this check unconditional triggers an extremely annoying gcc-5
warning.
PR 19323
* elfcode.h (elf_object_p): Check for ridiculous e_shnum and
e_phnum values.
Alan Modra [Wed, 2 Dec 2015 05:16:48 +0000 (15:46 +1030)]
addr2line vs. inlined C functions called from C++
In this case the inlined function doesn't have DW_AT_linkage_name in
.debug_info, but the language is C++ so find_nearest_line goes looking
in the symbol table. Since the function is inlined the enclosing
non-inline function symbol is returned from _bfd_elf_find_function,
which is wrong. This patch only uses a symbol if its address matches.
PR binutils/19315
* dwarf2.c (_bfd_elf_find_function): Return symbol matched.
(_bfd_dwarf2_find_nearest_line): Check symbol returned above
against dwarf range.
* elf-bfd.h (_bfd_elf_find_function): Update prototype.
Alan Modra [Wed, 2 Dec 2015 08:19:53 +0000 (18:49 +1030)]
Fix powerpc64 segfault caused by zero r_symndx relocs.
Fixes a segfault in ppc64_elf_tls_optimize found when testing
R_PPC64_ENTRY, and potential for trouble in other places found by
code inspection.
* elf64-ppc.c (ppc64_elf_tls_optimize): Don't segfault on NULL
symbol section or output section.
(ppc64_elf_edit_toc): Similarly for ld -R objects.
(ppc64_elf_size_stubs): Likewise.
Andrew Burgess [Fri, 16 Oct 2015 09:58:02 +0000 (11:58 +0200)]
objdump: Handle 32-bit base address in debug_ranges / debug_loc.
When the DWARF address size is 32-bit, but the host machine is 64-bit,
objdump fails to spot base addresses specified in the .debug_ranges and
.debug_loc lists.
As an example, here is the output when dumping an example .debug_ranges
section with the pre-patched objdump:
Alan Modra [Thu, 19 Nov 2015 04:30:13 +0000 (15:00 +1030)]
Account for .tbss alignment when adjusting start of relro
Another option might be to not bump "dot" for .tbss alignment in the
main section sizing loop, but that could leak some of the following
section into the TLS segment. Leakage shouldn't matter since it will
be to bytes past the end of .tdata, but for now this is a safer
option.
PR ld/19264
* ldlang.c (lang_size_sections): Don't ignore .tbss when
adjusting start of relro region.
Matthew Wahab [Wed, 9 Dec 2015 13:10:59 +0000 (13:10 +0000)]
[AArch64] Add ARMv8.2 command line option and feature flag.
ARMv8.2 is an architectural extension of ARMv8. This patch adds an
architecture feature macro for ARMv8.2 to the binutils AArch64 target
with GAS command line option -march=armv8.2-a.
gas/
2015-12-09 Matthew Wahab <matthew.wahab@arm.com>
Matthew Wahab [Tue, 8 Dec 2015 09:51:10 +0000 (09:51 +0000)]
[AArch64] Add support for ARMv8.1 Virtualization Host Extensions.
The ARMv8.1 architecture includes the Virtualization Host Extensions
which add a number of system registers. This patch adds support for
these system registers, making them available when -march=armv8.1-a is
selected.
include/opcode/
2015-12-08 Matthew Wahab <matthew.wahab@arm.com>
H.J. Lu [Tue, 1 Dec 2015 22:45:51 +0000 (14:45 -0800)]
Properly check symbol defined by assignment in linker script
Symbol defined by a linker assignment may have type bfd_link_hash_new
or bfd_link_hash_undefined. And h->def_regular is always set.
elf_i386_convert_load and elf_x86_64_convert_load should check
h->def_regular as well as bfd_link_hash_undefined and bfd_link_hash_new
to see if a symbol is defined by a linker script.
bfd/
PR ld/19319
* elf32-i386.c (elf_i386_convert_load): Check h->def_regular
instead of bfd_link_hash_new.
* elf64-x86-64.c (elf_x86_64_convert_load): Likewise. Skip
relocation overflow for bfd_link_hash_undefined and
bfd_link_hash_new if h->def_regular is set.
James Greenhalgh [Thu, 12 Nov 2015 12:04:22 +0000 (12:04 +0000)]
[AArch64] Add support for Cortex-A35
This patch adds support to the AArch64 back-end for the Cortex-A35
processor, as recently announced by ARM. The ARM Cortex-A35 provides
full support for the ARMv8-A architecture, including the CRC extension,
with optional Advanced-SIMD and Floating-Point support. We therefore set
feature flags for this CPU to AARCH64_ARCH_V8 and AARCH64_FEATURE_CRC, in
the same fashion as Cortex-A53 and Cortex-A57.
Tested in a cross environment for AArch64 with no issues.
The assembly code for emitting the proper tracepointable instruction
was duplicated in many places. Keep it in one place, to reduce work
needed for new targets.
gdb/testsuite/ChangeLog:
* gdb.trace/change-loc.h: include "trace-common.h", remove SYMBOL
macro.
(func5): Removed.
(func4): Use FAST_TRACEPOINT_LABEL.
* gdb.trace/ftrace-lock.c: include "trace-common.h", remove SYMBOL
macro.
(func): Removed.
(thread_function): Use FAST_TRACEPOINT_LABEL.
* gdb.trace/ftrace.c: include "trace-common.h", remove SYMBOL macro.
(func): Remove.
(marker): Use FAST_TRACEPOINT_LABEL.
* gdb.trace/pendshr1.c: include "trace-common.h", remove SYMBOL macro.
(pendfunc1): Remove.
(pendfunc): Use FAST_TRACEPOINT_LABEL.
* gdb.trace/pendshr2.c: include "trace-common.h", remove SYMBOL macro.
(foo): Remove.
(pendfunc2): Use FAST_TRACEPOINT_LABEL.
* gdb.trace/trace-break.c: include "trace-common.h", remove SYMBOL
macro.
(func): Remove.
(marker): Use FAST_TRACEPOINT_LABEL.
* gdb.trace/trace-common.h: New header.
* gdb.trace/trace-condition.c: include "trace-common.h", remove SYMBOL
macro.
(func): Remove.
(marker): Use FAST_TRACEPOINT_LABEL.
* gdb.trace/trace-mt.c: include "trace-common.h", remove SYMBOL macro.
(func): Remove.
(thread_function): Use FAST_TRACEPOINT_LABEL.
These variables were used in many gdb.trace tests. Keep them in one place,
to reduce work needed for new targets.
gdb/testsuite/ChangeLog:
* gdb.trace/backtrace.exp: Use global fpreg/spreg definition, add $
in front.
* gdb.trace/change-loc.exp: Use global pcreg definition.
* gdb.trace/collection.exp: Use global pcreg/fpreg/spreg definition.
* gdb.trace/entry-values.exp: Use global spreg definition, add $
in front.
* gdb.trace/mi-trace-frame-collected.exp: Use global pcreg definition.
* gdb.trace/pending.exp: Likewise.
* gdb.trace/report.exp: Use global pcreg/fpreg/spreg definition.
* gdb.trace/trace-break.exp: Likewise.
* gdb.trace/trace-condition.exp: Use global pcreg definition, add $
in front.
* gdb.trace/unavailable.exp: Use global pcreg/fpreg/spreg definition.
* gdb.trace/while-dyn.exp: Use global fpreg definition, add $
in front.
* lib/trace-support.exp: Define fpreg, spreg, pcreg variables.
* elf32-i386.c (elf_i386_relocate_section): Use read and write
pointers to reloc array, rather than memmove when deleting a
reloc. Don't use RELOC_AGAINST_DISCARDED_SECTION. Adjust
reloc counts at end of loop.
* elf64-x86-64.c (elf_x86_64_relocate_section): Likewise.
Alan Modra [Mon, 9 Nov 2015 04:33:29 +0000 (15:03 +1030)]
Fix performance regression due to ld -r memmove
The idea here is that instead of using memmove to shuffle the relocs
array every time one is deleted, to add a "wrel" pointer and copy from
rel[0] to wrel[0] as we go.
* elf64-ppc.c (ppc64_elf_relocate_section): Use read and write
pointers to reloc array, rather than memmove when deleting a
reloc. Don't use RELOC_AGAINST_DISCARDED_SECTION. Adjust
reloc counts at end of loop.
* elf32-ppc.c (ppc_elf_relocate_section): Likewise.
Joel Brobecker [Mon, 9 Nov 2015 17:58:16 +0000 (09:58 -0800)]
[Ada] GDB crash during "finish" of function with out parameters
Consider a function with the following signature...
function F (R : out Rec_Type) return Enum_Type;
... where Rec_Type is a simple record:
type Rec_Type is record
Cur : Integer;
end record;
Trying to "finish" from that function causes GDB to SEGV:
(gdb) fin
Run till exit from #0 bar.f (r=...) at bar.adb:5
0x00000000004022fe in foo () at foo.adb:5
5 I : Enum_Type := F (R);
[1] 18949 segmentation fault (core dumped) /[..]/gdb
This is related to the fact that funtion F has a parameter (R)
which is an "out" parameter being passed by copy. For those,
GNAT transforms the return value to be a record with multiple
fields: The first one is called "RETVAL" and contains the return
value shown in the source, and the remaining fields have the same
name as the "out" or "in out" parameters which are passed by copy.
So, in the example above, function F returns a struct that has
one field who name is "r".
Because "RETVAL" starts with "R", GDB thinks it's a wrapper field,
because it looks like the encoding used for variant records:
-- member_name ::= {choice} | others_choice
-- choice ::= simple_choice | range_choice
-- simple_choice ::= S number
-- range_choice ::= R number T number <<<<<----- here
-- number ::= {decimal_digit} [m]
-- others_choice ::= O (upper case letter O)
... which is a problem since print_field_values assumes that
the type it is given ("TYPE_FIELD_TYPE (type, i)" here), is also
a record type. However, that's not the case, since RETVAL is
an enum. That eventually leads GDB to a NULL type when trying to
extract fields out of the enum, which then leads to a SEGV when
trying to dereference it.
Ideally, we'd want to be a little more careful in identifying
wrapper fields, by enhancing ada_is_wrapper_field to be a little
more complete in its analysis of the field name before declaring
it a variant record wrapper. However, it's not super easy to do
so, considering that the choices can be combined together when
complex choices are used. Eg:
-- [...] the choice 1 .. 4 | 7 | -10 would be represented by
-- R1T4S7S10m
Given that we are working towards getting rid of GNAT encodings,
which means that the above will eventually disappear, we took
the more pragmatic approach is just treating RETVAL as a special
case.
gdb/ChangeLog:
* ada-lang.c (ada_is_wrapper_field): Add special handling
for fields called "RETVAL".
Alan Modra [Mon, 9 Nov 2015 04:47:05 +0000 (15:17 +1030)]
Configury changes for obstack optimization
Provides defines used to determine whether glibc obstacks are
compatible. Generally speaking, 32-bit targets won't need to use
obstack.o from libiberty if glibc is used, while 64-bit targets will,
until glibc gets the new obstack code.
libiberty/
* configure.ac: Get size of size_t.
* config.in: Regenerate.
* configure: Regenerate.
Alan Modra [Mon, 9 Nov 2015 04:45:51 +0000 (15:15 +1030)]
Silence obstack.c -Wc++compat warning
Fixes
warning: request for implicit conversion from ‘void *’ to ‘struct _obstack_chunk *’ not permitted in C++ [-Wc++-compat]
I moved the assignment to h->chunk to fix an overlong line, then
decided it would be better after the alloc failure check just to do
things the same way as in _obstack_newchunk.
* obstack.c (_obstack_newchunk): Silence -Wc++compat warning.
(_obstack_begin_worker): Likewise. Move assignment to h->chunk
after alloc failure check.
Kevin Buettner [Fri, 30 Oct 2015 04:53:51 +0000 (21:53 -0700)]
gdb.dwarf2: Don't hardcode certain constants in Dwarf::assemble constructs
Two tests in gdb.dwarf2, data-loc.exp and dynarr-ptr.exp assume that
sizeof(int) is 4. This patch looks up the integer size and uses this
constant for DW_AT_byte_size, DW_AT_lower_bound, and DW_AT_upper_bound.
I discovered this problem while looking at test results for this
msp430 multilib:
As I recall, there are still (other) problems with msp430 multilibs
which don't use -mlarge.
gdb/testsuite/ChangeLog:
* gdb.dwarf2/data-loc.exp (Dwarf::assemble): Don't hardcode
value associated with DW_AT_byte_size.
* gdb.dwarf2/dynarr-ptr.exp (Dwarf::assemble): Don't hardcode
constants for DW_AT_byte_size, DW_AT_lower_bound, and
DW_AT_upper_bound.
Kevin Buettner [Fri, 6 Nov 2015 04:40:53 +0000 (21:40 -0700)]
testsuite: Define and use gdb_target_symbol_prefix_flags_asm.
Some of the source code for the test cases in the GDB testsuite
reside in .S files containing assembly code. These files typically
define a symbol - such as main - which may, depending on the target,
require a prefix such as underscore.
For example, gdb.dwarf2/dw-compdir-oldgcc.S defines the symbol main:
main: .globl main
Some targets, such as rx-elf, require main to have an underscore
prefix. (If it doesn't, a linker error results due to not being able
to find _main required by crt0.o.) So, instead, the above should look
like this for rx-elf and other targets with this same requirement:
_main: .globl _main
This patch defines a new tcl proc in lib/gdb named
gdb_target_symbol_prefix_flags_asm. This proc returns a string
which will - assuming everything else is wired up correctly - cause
-DSYMBOL_PREFIX=_ to be passed on the command line to the compiler.
The test cases are augmented with a macro definition for SYMBOL
as follows:
#define CONCAT1(a, b) CONCAT2(a, b)
#define CONCAT2(a, b) a ## b
Symbols, such as main shown in the example earlier are then wrapped
with SYMBOL like this:
SYMBOL(main): .globl SYMBOL(main)
The net effect will be to add a prefix for those targets which need
it and add no prefix for those targets which do not.
It should be noted that there was already a proc in lib/gdb.exp
called gdb_target_symbol_prefix_flags. It still exists, but has
been significantly rewritten. (There is only one small difference
between the two versions.)
That proc used to explicitly list targets which were known to
require an underscore prefix. This is no longer done; the recently
added proc, gdb_target_symbol_prefix, is now invoked to dynamically
discover whether or not a prefix is required for that particular
target.
The difference between gdb_target_symbol_prefix_flags_asm
and gdb_target_symbol_prefix_flags is that the former returns
a bare prefix while the latter returns the prefix enclosed in
double quotes. I.e. assuming that the discovered prefix is
underscore, gdb_target_symbol_prefix_flags_asm returns:
additional_flags=-DSYMBOL_PREFIX=_
while gdb_target_symbol_prefix_flags returns:
additional_flags=-DSYMBOL_PREFIX="_"
The double-quoted version is not suitable for using with .S files
containing assembly code; there is no way to strip the double quotes
using C preprocessor constructs.
It would be possible to use the bare (non double quoted) version in
C source code. However, the supporting macros become more complicated
and therefore more difficult to maintain.
gdb/testsuite/ChangeLog:
* lib/gdb (gdb_target_symbol_prefix_flags_asm): New proc.
(gdb_target_symbol_prefix_flags): Define in terms of _asm
version.
* gdb.arch/i386-float.exp, gdb.arch/i386-permbkpt.exp,
gdb.dwarf2/dw2-canonicalize-type.exp,
gdb.dwarf2/dw2-compdir-oldgcc.exp, gdb.dwarf2/dw2-minsym-in-cu.exp,
gdb.dwarf2/dw2-op-stack-value.exp, gdb.dwarf2/dw2-unresolved.exp,
gdb.dwarf2/fission-reread.exp, gdb.dwarf2/pr13961.exp: Use flags
provided by gdb_target_symbol_prefix_flags_asm.
* gdb.dwarf2/dw2-canonicalize-type.S, gdb.dwarf2/dw2-compdir-oldgcc.S,
testsuite/gdb.dwarf2/dw2-minsym-in-cu.S,
testsuite/gdb.dwarf2/dw2-unresolved-main.c,
testsuite/gdb.dwarf2/dw2-unresolved.S, gdb.dwarf2/fission-reread.S,
gdb.dwarf2/pr13961.S: Define and use SYMBOL macro (and supporting
macros where needed). Use this macro for symbols which require
the prefix provided by SYMBOL_PREFIX.
Joel Brobecker [Fri, 6 Nov 2015 21:39:19 +0000 (13:39 -0800)]
Do not use libiberty's getpagesize on Android
Building libiberty on Android currently fails with the error message
shown below. This was discovered by trying to build GDBserver
for Android, which stopped building after libiberty became
a GDBserver dependency.
Here is the error message:
[...]/getpagesize.c:64:1: error: redefinition of 'getpagesize'
In file included from /[...]/getpagesize.c:34:0:
/[...]/usr/include/unistd.h:171:23: note: previous definition of 'getpagesize' was here
And looking at the definition, one can see that it defined as
a static inline function...
static __inline__ int getpagesize(void) {
extern unsigned int __page_size;
return __page_size;
}
... which explains why the AC_CHECK_FUNCS test failed to detect
the function, since there is no associated symbol to be linked in.
This patch prevents getpagesize.c to be compiled in by hard-coding
the fact that getpagesize is available on android hosts.
libiberty/ChangeLog:
* configure.ac: Set AC_CV_FUNC_GETPAGESIZE to "yes" on
Android hosts.
* configure: Regenerate.
Kevin Buettner [Wed, 28 Oct 2015 18:36:06 +0000 (11:36 -0700)]
gdb.dwarf2: Define and use gdb_target_symbol for symbol prefixes
Some of the tests in gdb.dwarf2 which use Dwarf::assemble refer to
(minimal/linker) symbols created in the course of building a small
test program. Some targets use a prefix such as underscore ("_") on
these symbols. Many of the tests in gdb.dwarf2 do not take this into
account. As a consequence, these tests fail to build, resulting
either in failures or untested testcases.
Here is an example from gdb.dwarf2/dw2-regno-invalid.exp:
Dwarf::assemble $asm_file {
cu {} {
compile_unit {
{low_pc main DW_FORM_addr}
{high_pc main+0x10000 DW_FORM_addr}
} {
...
}
For targets which require an underscore prefix on linker symbols,
the two occurrences of "main" would have to have a prepended underscore,
i.e. _main instead of main.
For the above case, a call to the new proc gdb_target_symbol is used
prepend the correct prefix to the symbol. I.e. the above code is
rewritten (as shown in the patch) as follows:
I also found it necessary to make an adjustment to lib/dwarf.exp so that
expressions of more than just one list element can be used in DW_TAG_...
constructs. Both atomic-type.exp and dw2-bad-mips-linkage-name.exp require
this new functionality.
gdb/testsuite/ChangeLog:
* lib/gdb.exp (gdb_target_symbol_prefix, gdb_target_symbol):
New procs.
* lib/dwarf.exp (_handle_DW_TAG): Handle attribute values,
representing expressions, of more than one list element.
* gdb.dwarf2/atomic-type.exp (Dwarf::assemble): Use gdb_target_symbol
to prepend linker symbol prefix to f.
* gdb.dwarf2/data-loc.exp (Dwarf::assemble): Likewise, for
table_1 and table_2.
* gdb.dwarf2/dw2-bad-mips-linkage-name.exp (Dwarf::assemble):
Likewise, for f and g.
* gdb.dwarf2/dw2-ifort-parameter.exp (Dwarf::assemble): Likewise,
for ptr.
* gdb.dwarf2/dw2-regno-invalid.exp (Dwarf::assemble): Likewise,
for main.
* gdb.dwarf2/dynarr-ptr.exp (Dwarf::assemble): Likewise, for
table_1_ptr and table_2_ptr.
Cary Coutant [Thu, 5 Nov 2015 20:59:02 +0000 (12:59 -0800)]
Revert patch for PR 19119, which led to PR 19172 and 19197.
Gold does not support all the emulations that Gnu ld does, and supports
only one spelling per target. The -m option is used only in the rare case
where there are no ELF input files, and we produce an empty output file.
In those cases, users are expected to supply a -m option naming one of
the supported emulations. In the many cases where a build script provides
an unnecessary -m option naming an emulation that gold does not support,
we will simply ignore the option, as we did before the reverted patch.
* options.h (General_options): Remove "obsolete" from -m.
* parameters.cc (set_parameters_target): Check if input target
is compatible with output emulation set by "-m emulation".
With --no-apply-dynamic-relocs on aarch64 targets, gold will not apply
link-time values for absolute relocations that become dynamic relocations.
This provides a workaround for broken Android dynamic linkers that use
the link-time value as an extra addend to the relocation.
gold/
PR gold/19163
* aarch64.cc (Target_aarch64::Relocate::relocate): Don't apply
certain relocations if --no-apply-dynamic-relocs is set.
* options.h (--apply-dynamic-relocs): New aarch64-specific option.
Yao Qi [Thu, 5 Nov 2015 09:44:32 +0000 (09:44 +0000)]
Use aarch64_decode_insn in aarch64_analyze_prologue
This patch convert aarch64_analyze_prologue to using aarch64_decode_insn
to decode instructions. After this change, aarch64_analyze_prologue
looks much simple, and some aarch64_decode_* functions are removed
accordingly.
projects / thirdparty / binutils-gdb.git / log
