binutils-gdb/ld/emulparams/elf64mmix.sh
Hans-Peter Nilsson 4609ada49f ld: optionally emit _etext last, before .data
So, here's my suggestion for making _init .. __etext cover .text +
.rodata (including things like the read-only exception tables) for
elf64mmix.  A quick web search gives that __etext (and friends) isn't
well defined, so each target can interpret the "end of text segment"
to their own liking.  It seems likely this change is also a better fit
than the default for other ports, at least those with .rodata after
.text in the same segment.

The presence of a separate rodata-segment is optional (and not true
for elf64mmix).  This is reflected in the name as SEPARATE_TEXT /
SEPARATE_CODE isn't considered, to keep it simple; each target has to
make sure their settings of variables make sense.

ld:
	* scripttempl/elf.sc (ETEXT_LAST_IN_RODATA_SEGMENT): New variable.
	* emulparams/elf64mmix.sh (ETEXT_LAST_IN_RODATA_SEGMENT): Define.
	* testsuite/ld-mmix/sec-1.d: Adjust.
2020-07-19 06:08:07 +02:00

66 lines
2.3 KiB
Bash

TEMPLATE_NAME=elf
GENERATE_SHLIB_SCRIPT=yes
ELFSIZE=64
SCRIPT_NAME=elf
OUTPUT_FORMAT="elf64-mmix"
NO_REL_RELOCS=yes
ENTRY=_start.
# Default to 0 as mmixal does.
TEXT_START_ADDR='DEFINED (__.MMIX.start..text) ? __.MMIX.start..text : 0'
# Don't add SIZEOF_HEADERS.
# Don't set EMBEDDED, that would be misleading; it's not that kind of system.
TEXT_BASE_ADDRESS=$TEXT_START_ADDR
DATA_ADDR='DEFINED (__.MMIX.start..data) ? __.MMIX.start..data : 0x2000000000000000'
MAXPAGESIZE="CONSTANT (MAXPAGESIZE)"
ARCH=mmix
MACHINE=
COMPILE_IN=yes
EXTRA_EM_FILE=mmixelf
# The existence of a symbol __start (or _start) should overrule Main, so
# it can be a user symbol without the associated mmixal magic. We
# also want to provide Main as a synonym for _start, if Main wasn't
# defined but is referred to, and _start was defined.
#
# The reason we use a symbol "_start." as a mediator is to avoid
# causing ld to force the type of _start to object rather than no
# type, which confuses disassembly; we also make it alphanumerically
# a successor of _start for similar reasons. Perhaps it's a linker
# bug that linker-defined symbols set the symbol-type.
#
# Note that we smuggle this into OTHER_TEXT_SECTIONS (at the end
# of .text) rather than TEXT_START_SYMBOLS. This is necessary, as
# DEFINED wouldn't find the symbol if it was at the top; presumably
# before the definition, if the definition is not in the first file.
# FIXME: Arguably a linker bug.
# Only do this for a final link, or else we'll mess up e.g. error
# messages.
OTHER_TEXT_SECTIONS="
${RELOCATING+
_start. = (DEFINED (_start) ? _start
: (DEFINED (Main) ? Main : (DEFINED (.text) ? .text : 0)));
PROVIDE (Main = DEFINED (Main) ? Main : (DEFINED (_start) ? _start : _start.));
}"
# We need a symbol at the end of the read-only data, which is
# colocated with the code. We make __etext fit.
ETEXT_LAST_IN_RODATA_SEGMENT=" "
OTHER_SECTIONS='
.MMIX.reg_contents :
{
/* Note that this section always has a fixed VMA - that of its
first register * 8. */
*(.MMIX.reg_contents.linker_allocated);
*(.MMIX.reg_contents);
}
'
# FIXME: Also bit by the PROVIDE bug? If not, this could be
# EXECUTABLE_SYMBOLS.
# By default, put the high end of the stack where the register stack
# begins. They grow in opposite directions. */
OTHER_SYMBOLS="PROVIDE (__Stack_start = 0x6000000000000000);"