Adjust PR ld/30791 tests:
1. Generic linker targets don't comply with all orhpan section merging
rules.
2. z80 fails since a, b, c, d are registers for z80.
3. hppa fails since .text sections aren't merged for relocatable link.
PR ld/30791
* testsuite/ld-elf/pr30791a.d: Xfail for generic and z80
targets.
* testsuite/ld-elf/pr30791b.d: Xfail for hppa and z80 targets.
Update ARC's elfarcv2 script template with:
- The .ivt section (Interrupt Vector Table) is mapped at the begining
of STARTUP_MEMORY when ivtbase_addr is not defined. Previously, it
was pointing to 0x00.
- MEMORY_FILE is a new emulation paramter and sets the name for the
linker script file which holds the MEMORY commands required by
arcv2elfx emulation.
- Four new linker variables are introduced available when arcv2elf emulation is used:
* __TEXT_REGION_ORIGIN__ Once defined it is setting the text region origin. By default it points to zero.
* __TEXT_REGION_LENGTH__ Once defined it is setting the text region length. By default it is set to 2M.
* __DATA_REGION_ORIGIN__ Once defined it is setting the data region origin. By default it is set to 0x80000000.
* __DATA_REGION_LENGTH__ Once defined it is setting the data region length. By default it is set to 2M.
ld/ChangeLog:
* scripttempl/elfarcv2.sc: Update script template.
Signed-off-by: Claudiu Zissulescu <claziss@gmail.com>
Following the arrangement in GCC select a 64-bit ABI by default, either
n32 or n64, rather than o32 for `mipsisa64*-*-linux*' targets, just as
with the corresponding `mips64*-*-linux*' targets.
This patch gets rid of the individual rules including the .Pc
dependency files made while generating e*.c files, replacing them with
a fancy GNU make pattern include. I've also moved creation of
ldscripts to the makefile since it is possible to run more than one
genscripts at once, resulting in "ldscripts: File exists" messages.
* Makefile.am: Replace individual include of *.Pc dependency
files with one pattern rule.
(.Pc): New dummy rule.
(ldscripts/stamp): New rule.
(GEN_DEPENDS): Add ldscripts/stamp.
(install-data-local): Exclude ldscripts/stamp from install.
* genscripts.sh: Don't make ldscripts dir.
* Makefile.in: Regenerate.
Use the pattern rule in a comment from commit 77ac17b8453f.
* Makefile.am (run-genscripts): Delete. Use pattern rule
e%.c instead.
* Makefile.in: Regenerate.
If a linker script does not place the PLT relocations and "normal"
relocations in separate ELF sections, `ld` will currently output incorrect
values for DT_JMPREL and DT_PLTRELSZ - they cover the entire ELF section,
rather than just the PLT relocations
Don't ignore the extent of the BFD section - use the size of the srelplt
BFD section and its offset from the output_secttion
bfd/
PR ld/30787
* elfxx-x86.c (_bfd_x86_elf_finish_dynamic_sections): Use input
section for DT_JMPREL and DT_PLTRELSZ.
ld/
PR ld/30787
* testsuite/ld-i386/i386.exp: Run pr30787.
* testsuite/ld-x86-64/x86-64.exp: Likewise.
* testsuite/ld-i386/pr30787.d: New file.
* testsuite/ld-i386/pr30787.s: Likewise.
* testsuite/ld-i386/pr30787.t: Likewise.
* testsuite/ld-x86-64/pr30787.d: Likewise.
* testsuite/ld-x86-64/pr30787.s: Likewise.
* testsuite/ld-x86-64/pr30787.t: Likewise.
bfd/
* Makefile.am: Move elf32-kvx.lo from BFD32_BACKENDS to
BFD64_BACKENDS. Remove elfxx-kvx.lo from BFD32_BACKENDS.
Remove elfxx-kvx.c from BFD32_BACKENDS_CFILES.
* Makefile.in: Regenerate.
* config.bfd: Adjust targ_defvec and targ_selvecs and gate them
behind BFD64.
* configure.ac: Add target_size=64 to kvx_elf64_*vec.
* configure: Regenerate.
* elfnn-kvx.c (elfNN_kvx_stub_name): Cast rel->r_addend to
uint64_t to match format string.
(elfNN_kvx_relocate_section): Similarly for r_offset, and
use PRIx64 in format string.
* targets.c (_bfd_target_vector <kvx_elf32_vec>): Move inside
#ifdef BFD64.
ld/
* Makefile.am: Move eelf32kvx.c from ALL_EMULATION_SOURCES to
ALL_64_EMULATION_SOURCES.
* Makefile.in: Regenerate.
The elf32-kvx.c and elf64-kvx.c rules in the bfd makefile are
different to the other similar generated files, and that reminded me
that we need to have $srcdir in the generated #line reference back to
the source for debugging, but don't want it for comments in bfd.pot
(because then bfd.pot will likely reference Nick's source tree).
This patch fixes that by making all the #line use $srcdir by virtue of
using $<, and edits bfd.pot.
I also uniq list of files to remove duplicated elfxx-x86.c, sort lists
of files and regen with our standard automake/autoconf.
* configure: Regenerate.
bfd/
* Makefile.am: Sort various lists of files. Use $< in #line
directive of generated C files.
(po/SRC-POTFILES.in): uniq SRC_POTFILES.
(po/BLD-POTFILES.in): uniq BFD_POTFILES.
* Makefile.in: Regenerate.
* po/Make-in (bfd.pot): Edit out source dir from comments.
* po/SRC-POTFILES.in: Regenerate.
gas/
* Makefile.in: Regenerate.
* configure: Regenerate.
* po/POTFILES.in: Regenerate.
ld/
* Makefile.am (ALL_64_EMULATION_SOURCES): Sort.
* Makefile.in: Regenerate.
If we're using dump_prog in a test which fails, log the dump.out contents
to ld.log to aid debugging.
This avoids needing to ask reporters to manually run e.g. `objdump` commands
when making bug reports.
PR30722
* ld/testsuite/lib/ld-lib.exp: Log failed dump.out contents to aid
debugging.
Approved-by: Nick Clifton <nickc@redhat.com>
Signed-off-by: Sam James <sam@gentoo.org>
pr23169b test only works with lazy binding. To work with linker which
disables lazy binding by default, build pr23169b binaries with -z lazy.
PR ld/30698
* ld-ifunc/ifunc.exp: Build pr23169b binaries with -z lazy.
Fix issue reported by Dave and Alan.
Put back the old pattern for hppa-*-linux* and add hppa[12]*-*-linux* to cover
Gentoo's hppa1.1 and hppa2.0 without including hppa64 inadvertently like I did
before.
ld/
PR 30733
PR 30734
* ld/testsuite/ld-elf/relocatable.d: Use better pattern to exclude hppa64
but include hppa1.1, hppa2.0.
* ld/testsuite/ld-elf/retain7a.d: Ditto.
Fixes: 0e339f6b4f2df25ed351cb94dc7fe16868626f49
Fixes: e3b66187192ce6840df283c00f6395bb0ff15cf5
Signed-off-by: Sam James <sam@gentoo.org>
This regenerates config files changed by the previous 44 commits.
Note that subject lines in these commits mostly match the gcc git
originating commit.
These were renamed from bfd_read and bfd_write back in 2001 when they
lost an unnecessary parameter. Rename them back, and get rid of a few
casts that are only needed without prototyped functions (K&R C).
PR 30733
* ld/testsuite/ld-elf/retain7a.d: Fix XFAIL entry for hppa to match
hppa{1.1,2.0}*, like hppa2.0-unknown-linux-gnu which Gentoo uses.
Signed-off-by: Sam James <sam@gentoo.org>
PR 30734
* ld/testsuite/ld-elf/relocatable.d: Fix notarget entry for hppa to match
hppa{1.1,2.0}*, like hppa2.0-unknown-linux-gnu which Gentoo uses.
Signed-off-by: Sam James <sam@gentoo.org>
This fixes miscounting of dynamic relocations on GOT entries when
a) there are both local-dynamic and global-dynamic tls accesss for a
given symbol, and
b) the symbol is global with non-default visibility, and
c) the __tls_get_addr calls aren't optimised away.
PR 30697
bfd/
* elf32-ppc.c (allocate_dynrelocs): Correct local-dynamic
reloc count.
ld/
* testsuite/ld-powerpc/tls32ldgd.d,
* testsuite/ld-powerpc/tls32ldgd.s: New test.
* testsuite/ld-powerpc/powerpc.exp: Run it.
* configure.ac (stpcpy): AC_CHECK_DECLS.
* sysdep.h (stpcpy): Add fallback declaraion.
* config.in: Regenerate.
* configure: Regenerate.
* emultempl/pe.em (open_dynamic_archive): Use
stpcpy rather than sprintf plus strlen.
* emultempl/pep.em (open_dynamic_archive): Likewise.
* emultempl/xtensaelf.em (elf_xtensa_before_allocation): Use
auto rather than malloc'd buffer. Use sprintf count.
* ldelf.c (ldelf_search_needed): Use memcpy in place of sprintf.
* pe-dll.c (pe_process_import_defs): Use string already formed
for alias match rather than recreating.
Make the n64 ABI the default for 64-bit Linux targets specified with
`-gnuabi64' suffix included in the target triplet, for configurations
such as the Debian mips64el and mips64r6el ports. Adjust testsuite
configuration accordingly.
There are the following regressions with the new target triplet:
mips64-linux-gnuabi64 +FAIL: readelf -S bintest
mips64-linux-gnuabi64 +FAIL: MIPS reloc estimation 1
mips64el-linux-gnuabi64 +FAIL: readelf -S bintest
mips64el-linux-gnuabi64 +FAIL: MIPS reloc estimation 1
The `readelf' issue comes from a difference in section headers produced
that the `binutils/testsuite/binutils-all/readelf.s-64' pattern template
does not match. While there has been a precedent it does not appear to
me that there is a clear advantage from adding more and more variations
to the template rather than forking the existing template into multiple
ones for a more exact match. So this is best deferred to a separate
discussion.
The MIPS reloc estimation issue is an actual bug in `objdump', which
discards a number of trailing entries from output here for n64 composed
relocations:
DYNAMIC RELOCATION RECORDS
OFFSET TYPE VALUE
0000000000000000 R_MIPS_NONE *ABS*
0000000000000000 R_MIPS_NONE *ABS*
and consequently `ld/testsuite/ld-mips-elf/reloc-estimate-1.d' does not
match even though ELF output produced is correct according to `readelf':
Relocation section '.rel.dyn' at offset 0x10400 contains 2 entries:
Offset Info Type Sym. Value Sym. Name
000000000000 000000000000 R_MIPS_NONE
Type2: R_MIPS_NONE
Type3: R_MIPS_NONE
000000010000 000300001203 R_MIPS_REL32 0000000000010010 foo@@V2
Type2: R_MIPS_64
Type3: R_MIPS_NONE
As a genuine bug this has to be handled separately.
Co-Authored by: Maciej W. Rozycki <macro@orcam.me.uk>
bfd/
* config.bfd: Add `mips64*el-*-linux*-gnuabi64' and
`mips64*-*-linux*-gnuabi64' targets.
binutils/
* testsuite/binutils-all/mips/mips.exp: Handle `*-*-*-gnuabi64'
targets.
* testsuite/binutils-all/objcopy.exp: Handle
`mips64*-*-*-gnuabi64' targets.
* testsuite/binutils-all/remove-relocs-01.d: Likewise.
* testsuite/binutils-all/remove-relocs-04.d: Likewise.
* testsuite/binutils-all/remove-relocs-05.d: Likewise.
* testsuite/binutils-all/remove-relocs-06.d: Likewise.
gas/
* configure.ac: Handle `mips64*-linux-gnuabi64' targets.
* configure: Regenerate.
* testsuite/gas/mips/compact-eh-eb-7.d: Handle
`mips64*-*-*-gnuabi64' targets.
* testsuite/gas/mips/compact-eh-el-7.d: Likewise.
ld/
* configure.tgt: Add `mips64*el-*-linux-gnuabi64' and
`mips64*-*-linux-gnuabi64' targets.
* testsuite/ld-undefined/undefined.exp: Handle
`mips64*-*-*-gnuabi64' targets.
* testsuite/ld-mips-elf/attr-gnu-4-10.d: Likewise.
* testsuite/ld-mips-elf/compact-eh6.d: Likewise.
* testsuite/ld-mips-elf/mips-elf.exp: Handle `*-*-*-gnuabi64'
targets.
IRIX 6 does not have MIPS16 stub section support in its n32 linker
scripts, causing such input sections to be propagated to the respective
output sections rather than `.text', causing dump pattern mismatches.
Expect IRIX 6 to fail with n32 testing then, removing this regression:
mips-sgi-irix6 -FAIL: MIPS16 interlinking for local functions 1 (n32)
We may choose to update IRIX 6 n32 linker scripts sometime, as it seems
a harmless change.
ld/
* testsuite/ld-mips-elf/mips-elf.exp: Expect IRIX 6 to fail with
n32 `mips16-local-stubs-1' testing.
The MIPS16 interlinking test for local functions expects to be assembled
with 32-bit addressing, otherwise causing assembly warnings:
.../ld/testsuite/ld-mips-elf/mips16-local-stubs-1.s: Assembler messages:
.../ld/testsuite/ld-mips-elf/mips16-local-stubs-1.s:16: Warning: la used to load 64-bit address; recommend using dla instead
Use the per-ABI framework then to run the test explicitly for o32 and
n32 ABIs only, replacing the `-mips4' option from the `as' tag with
`.module mips4' pseudo-op within the source itself so as to avoid
assembly errors:
Assembler messages:
Error: -mips4 conflicts with the other architecture options, which imply -mips3
with n32 testing for some targets, and ultimately removing these
regressions:
mips64-openbsd -FAIL: MIPS16 interlinking for local functions 1
mips64el-openbsd -FAIL: MIPS16 interlinking for local functions 1
Co-Authored-By: Maciej W. Rozycki <macro@orcam.me.uk>
ld/
* testsuite/ld-mips-elf/mips16-local-stubs-1.d: Remove `-mips4'
from the `as' tag.
* testsuite/ld-mips-elf/mips16-local-stubs-1.s: Add `.module
mips4'.
* testsuite/ld-mips-elf/mips-elf.exp: Run `mips16-local-stubs-1'
for o32 and n32 ABIs only.
The `got-dump-1' test case uses 32-bit addressing, so it makes no sense
to run it with the n64 ABI. And there is a corresponding `got-dump-2'
test already for the n64 ABI.
Use the per-ABI framework then to run the `got-dump-1' test explicitly
for o32 and n32 ABIs only.
ld/
* testsuite/ld-mips-elf/mips-elf.exp: Run `got-dump-1' for o32
and n32 ABIs only.
OpenBSD targets produce ELF64 files while the pattern dump expects ELF32
output and specific header sizes. Disable it for `mips64*-*-openbsd*'
for these targets then, removing these failures:
mips64-openbsd -FAIL: ld-mips-elf/attr-gnu-4-10
mips64el-openbsd -FAIL: ld-mips-elf/attr-gnu-4-10
ld/
* testsuite/ld-mips-elf/attr-gnu-4-10.d: Add `notarget' tag with
`mips64*-*-openbsd*'.
IRIX targets do not enable the production of a `.pdr' section in GAS by
default, which causes a failure with the `attr-gnu-4-10' test case due
to a difference resulting in the number and indices of sections produced
in linker output.
As the presence or absence of this section is not relevant to this test
case, just enable it unconditionally, fixing these regressions:
mips-sgi-irix5 -FAIL: ld-mips-elf/attr-gnu-4-10
mips-sgi-irix6 -FAIL: ld-mips-elf/attr-gnu-4-10
ld/
* testsuite/ld-mips-elf/attr-gnu-4-10.d: Add `as' tag with
`-mpdr'.
The `mips-sgi-irix6' target only supports IRIX linker emulations, but
most JALR relaxation tests request the relevant traditional emulation
instead, causing a link failure:
./ld-new: unrecognised emulation mode: elf32btsmipn32
Supported emulations: elf32bmipn32 elf32bsmip elf64bmip
This is clearly an omission from the conversion to use the per-ABI
framework made with commit 78da84f99405 ("MIPS/LD/testsuite: Correct
mips-elf.exp test ABI/emul/endian arrangement"). These tests are also
endianness agnostic, which was missed in the conversion as well.
Remove the unnecessary explicit ABI and endianness options then and rely
on the per-ABI framework to get things right, removing this regression:
mips-sgi-irix6 -FAIL: MIPS relax-jalr-shared n32
ld/
* testsuite/ld-mips-elf/relax-jalr-n32-shared.d: Remove flags
related to ABI and endianness selection from the `as' and `ld'
tags.
* testsuite/ld-mips-elf/relax-jalr-n64.d: Likewise.
* testsuite/ld-mips-elf/relax-jalr-n64-shared.d: Likewise.
* testsuite/ld-mips-elf/mips-elf.exp: Remove `as' and `ld' tag
additions from the invocation of JALR relaxation tests.
There are only n64 linker emulations included with `mips64*-*-openbsd*'
targets, however the unaligned JALX tests insist on running across all
targets and force the n32 ABI, causing link errors with the targets
concerned, e.g.:
./ld-new: tmpdir/unaligned-jalx-0.o: ABI is incompatible with that of the selected emulation
./ld-new: failed to merge target specific data of file tmpdir/unaligned-jalx-0.o
./ld-new: tmpdir/unaligned-insn.o: ABI is incompatible with that of the selected emulation
./ld-new: failed to merge target specific data of file tmpdir/unaligned-insn.o
Convert the tests then to use the per-ABI framework and run them for the
o32 and n32 ABIs, removing these regressions:
mips64-openbsd -FAIL: MIPS JALX to unaligned symbol 0
mips64-openbsd -FAIL: MIPS JALX to unaligned symbol 1
mips64-openbsd -FAIL: MIPS JALX to unaligned symbol 2
mips64-openbsd -FAIL: MIPS JALX to unaligned symbol 3
mips64-openbsd -FAIL: MIPS16 JALX to unaligned symbol 0
mips64-openbsd -FAIL: MIPS16 JALX to unaligned symbol 1
mips64-openbsd -FAIL: microMIPS JALX to unaligned symbol 0
mips64-openbsd -FAIL: microMIPS JALX to unaligned symbol 1
mips64el-openbsd -FAIL: MIPS JALX to unaligned symbol 0
mips64el-openbsd -FAIL: MIPS JALX to unaligned symbol 1
mips64el-openbsd -FAIL: MIPS JALX to unaligned symbol 2
mips64el-openbsd -FAIL: MIPS JALX to unaligned symbol 3
mips64el-openbsd -FAIL: MIPS16 JALX to unaligned symbol 0
mips64el-openbsd -FAIL: MIPS16 JALX to unaligned symbol 1
mips64el-openbsd -FAIL: microMIPS JALX to unaligned symbol 0
mips64el-openbsd -FAIL: microMIPS JALX to unaligned symbol 1
Similar tests for the n64 ABI can be added separately, using suitable
dump patterns.
ld/
* testsuite/ld-mips-elf/unaligned-jalx-0.d: Remove `-32' from
the `as' tag.
* testsuite/ld-mips-elf/unaligned-jalx-1.d: Likewise.
* testsuite/ld-mips-elf/unaligned-jalx-2.d: Likewise.
* testsuite/ld-mips-elf/unaligned-jalx-3.d: Likewise.
* testsuite/ld-mips-elf/unaligned-jalx-mips16-0.d: Likewise.
* testsuite/ld-mips-elf/unaligned-jalx-mips16-1.d: Likewise.
* testsuite/ld-mips-elf/unaligned-jalx-micromips-0.d: Likewise.
* testsuite/ld-mips-elf/unaligned-jalx-micromips-1.d: Likewise.
* testsuite/ld-mips-elf/mips-elf.exp: Run unaligned JALX tests
with `run_dump_test_o32' and `run_dump_test_n32' rather than
`run_dump_test'.
In ld/emulparams/elf64bmip-defs.sh there is no explicit handling of the
`.interp' section, which causes it to be positioned in output at an odd
place.
Let's include it with INITIAL_READONLY_SECTIONS, just like o32/n32 do,
fixing a regression from commit 5a8e7be242f3 ("INITIAL_READONLY_SECTIONS
in elf.sc"), where the handling of n64 was missed due to an unfortunate
sequence of events where ld/emulparams/elf64bmip-defs.sh was only added
with commit 94bb04b3c611 ("Use .reginfo rather than .MIPS.options in n32
linker scripts") the day before.
Add test cases covering section ordering across the three ABIs. This
change also fixes ld/pr23658-2:
FAIL: Build pr23658-2
Co-Authored-By: Maciej W. Rozycki <macro@orcam.me.uk>
ld/ChangeLog:
* emulparams/elf64bmip-defs.sh: Include `.interp' with
INITIAL_READONLY_SECTIONS.
* testsuite/ld-mips-elf/pie-n64.d: Adjust addresses.
* testsuite/ld-mips-elf/sections-1-o32.rd: New test.
* testsuite/ld-mips-elf/sections-1-o32t.rd: New test.
* testsuite/ld-mips-elf/sections-1-n32.rd: New test.
* testsuite/ld-mips-elf/sections-1-n32t.rd: New test.
* testsuite/ld-mips-elf/sections-1-n32p.rd: New test.
* testsuite/ld-mips-elf/sections-1-n64.rd: New test.
* testsuite/ld-mips-elf/sections-1-n64t.rd: New test.
* testsuite/ld-mips-elf/sections-2-o32.rd: New test.
* testsuite/ld-mips-elf/sections-2-o32t.rd: New test.
* testsuite/ld-mips-elf/sections-2-n32.rd: New test.
* testsuite/ld-mips-elf/sections-2-n32t.rd: New test.
* testsuite/ld-mips-elf/sections-2-n32p.rd: New test.
* testsuite/ld-mips-elf/sections-2-n64.rd: New test.
* testsuite/ld-mips-elf/sections-2-n64t.rd: New test.
* testsuite/ld-mips-elf/sections.s: New test source.
* testsuite/ld-mips-elf/mips-elf.exp: Run the new tests.
This reverts commit 32f1c80375ebe8ad25d9805ee5889f0006c51e59. It had
two unrelated changes lumped together, one of which changed the meaning
of the `mipsisa64*-*-linux*' target triplets, which was not properly
evaluated.
The empty-address tests check that if a section is removed by ld due
to being empty then properties of that section don't affect following
addresses. The xcoff backend doesn't remove the empty .data section
created by empty-address-2* and empty-address-3* for some reason, and
therefore fails the test.
* testsuite/ld-scripts/empty-address-1.d: Accept more symbols.
* testsuite/ld-scripts/empty-address-2a.d: xfail for xcoff.
* testsuite/ld-scripts/empty-address-2b.d: Likewise.
* testsuite/ld-scripts/empty-address-3a.d: Likewise.
* testsuite/ld-scripts/empty-address-3b.d: Likewise.
The canonical form to discard all sections not mentioned earlier in
the script is
/DISCARD/ : { *(*) }
not
/DISCARD/ : { *(.*) }
".*" happens to work with the usual section names starting with a dot,
but let's not promote something not quite right.
bfd
* peXXigen.c (_bfd_XXi_only_swap_filehdr_out): If inserting a timestamp, use the value held in the SOURCE_DATE_EPOCH environment variable, if it is defined.
binutils
* doc/binutils.texi (objcopy): Document change in behaviour of objcopy's --preserve-dates command line option.
ld
* pe-dll.c (fill_edata): If inserting a timestamp, use the value held in the SOURCE_DATE_EPOCH environment variable, if it is defined.
* ld.texi (--insert-timestamp): Document change in behaviour.
In LoongArch, the R_LARCH_IRELATIVE relocations for local ifunc symbols are
in .rela.dyn. Before, this is done by loongarch_elf_finish_dynamic_sections.
But this function is called after elf_link_sort_relocs, it need to find a
null slot to insert IRELATIVE relocation.
Now, it is processed by elf_loongarch_output_arch_local_syms before
elf_link_sort_relocs, just need to call loongarch_elf_append_rela to
insert IRELATIVE relocation.
bfd/ChangeLog:
* elfnn-loongarch.c (elfNN_allocate_local_ifunc_dynrelocs): Return
type change to int.
(loongarch_elf_size_dynamic_sections): Delete (void *).
(loongarch_elf_finish_dynamic_symbol): Use loongarch_elf_append_rela
insert IRELATIVE relocation to .rela.dyn.
(elfNN_loongarch_finish_local_dynamic_symbol): Return type change to
int.
(loongarch_elf_finish_dynamic_sections): Delete process of local
ifunc symbols.
(elf_backend_output_arch_local_syms): New.
ld/ChangeLog:
* testsuite/ld-loongarch-elf/local-ifunc-reloc.d: Regenerated.
CGEN is cool, but the BPF architecture is simply too bizarre for it.
The weird way of BPF to handle endianness in instruction encoding, the
weird C-like alternative assembly syntax, the weird abuse of
multi-byte (or infra-byte) instruction fields as opcodes, the unusual
presence of opcodes beyond the first 32-bits of some instructions, are
all examples of what makes it a PITA to continue using CGEN for this
port. The bpf.cpu file is becoming so complex and so nested with
p-macros that it is very difficult to read, and quite challenging to
update. Also, every time we are forced to change something in CGEN to
accommodate BPF requirements (which is often) we have to do extensive
testing to make sure we do not break any other target using CGEN.
This is getting un-maintenable.
So I have decided to bite the bullet and revamp/rewrite the port so it
no longer uses CGEN. Overall, this involved:
* To remove the cpu/bpf.{cpu,opc} descriptions.
* To remove the CGEN generated files.
* To replace the CGEN generated opcodes table with a new hand-written
opcodes table for BPF.
* To replace the CGEN generated disassembler wih a new disassembler
that uses the new opcodes.
* To replace the CGEN generated assembler with a new assembler that uses the
new opcodes.
* To replace the CGEN generated simulator with a new simulator that uses the
new opcodes. [This is pushed in GDB in another patch.]
* To adapt the build systems to the new situation.
Additionally, this patch introduces some extensions and improvements:
* A new BPF relocation BPF_RELOC_BPF_DISP16 plus corresponding ELF
relocation R_BPF_GNU_64_16 are added to the BPF BFD port. These
relocations are used for section-relative 16-bit offsets used in
load/store instructions.
* The disassembler now has support for the "pseudo-c" assembly syntax of
BPF. What dialect to use when disassembling is controlled by a command
line option.
* The disassembler now has support for dumping instruction immediates in
either octal, hexadecimal or decimal. The used output base is controlled
by a new command-line option.
* The GAS BPF test suite has been re-structured and expanded in order to
test the disassembler pseudoc syntax support. Minor bugs have been also
fixed there. The assembler generic tests that were disabled for bpf-*-*
targets due to the previous implementation of pseudoc syntax are now
re-enabled. Additional tests have been added to test the new features of
the assembler. .dump files are no longer used.
* The linker BPF test suite has been adapted to the command line options
used by the new disassembler.
The result is very satisfactory. This patchs adds 3448 lines of code
and removes 10542 lines of code.
Tested in:
* Target bpf-unknown-none with 64-bit little-endian host and 32-bit
little-endian host.
* Target x86-64-linux-gnu with --enable-targets=all
Note that I have not tested in a big-endian host yet. I will do so
once this lands upstream so I can use the GCC compiler farm.
I have not included ChangeLog entries in this patch: these would be
massive and not very useful, considering this is pretty much a rewrite
of the port. I beg the indulgence of the global maintainers.
