The need for them on the operand-less string insns has gone away with
the removal of maybe_adjust_templates() and associated logic. Since
i386_index_check() needs adjustment then anyway, take the opportunity
and also simplify it, possible again as a result of said removal (plus
the opcode template adjustments done here).
Having it in match_template() is unhelpful. Neither does looking for the
next template to possibly match make any sense in that case, nor is the
resulting diagnostic making clear what the problem is.
While moving the check, also generalize it to include all SIMD and VEX-
encoded insns. This way an existing conditional can be re-used in
md_assemble(). Note though that this still leaves a lof of insns which
are also wrong to use with these relocations.
Further fold the remaining check (BFD_RELOC_386_GOT32) with the XRELEASE
related one a few lines down. This again allows re-using an existing
conditional.
In commit 1212781b35 ("ix86: allow HLE store of accumulator to
absolute address") I was wrong to exclude 64-bit code. Dropping the
check also leads to better diagnostics in 64-bit code ("MOV", after
all, isn't invalid with "XRELEASE").
While there also limit the amount of further checks done: The operand
type checks that were there were effectively redundant with other ones
anyway, plus it's quite fine to also have "xrelease mov <disp>, %eax"
look for the next MOV template (in fact again also improving
diagnostics).
Have its use, except where actually legitimate, result in the same "only
supported in 64-bit mode" diagnostic as emitted for other 64-bit only
insns. Also suppress deriving of the suffix in Intel mode except in the
legitimate cases. This in exchange allows dropping the respective code
from match_template().
To maintain reasonable diagnostics (in particular to avoid "`mov' is
only supported in 64-bit mode" on the SIMD forms of MOVQ) we need to
defer parse_insn()'s emitting of errors unrelated to prefix parsing.
Utilize i.error just like match_template() does.
Oddly enough despite gcc's preference towards FILDQ and FIST{,T}Q we
had no testcase whatsoever for these. Therefore such tests are being
added. Note that the removed line in the x86-64-lfence-load testcase
was redundant with the exact same one a few lines up.
Having templates with a suffix explicitly present has always been
quirky. Introduce a 2nd matching pass in case the 1st one couldn't find
a suitable template _and_ didn't itself already need to trim off a
suffix to find a match at all. This requires error reporting adjustments
(albeit luckily fewer than I was afraid might be necessary), as errors
previously reported during matching now need deferring until after the
2nd pass (because, obviously, we must not emit any error if the 2nd pass
succeeds). While also related to PR gas/29524, it was requested that
move-with-sign-extend be left as broken as it always was.
PR gas/29525
Note that with the dropped CMPSD and MOVSD Intel Syntax string insn
templates taking operands, mixed IsString/non-IsString template groups
(with memory operands) cannot occur anymore. With that
maybe_adjust_templates() becomes unnecessary (and is hence being
removed).
PR gas/29526
Note further that while the additions to the intel16 testcase aren't
really proper Intel syntax, we've been permitting all of those except
for the MOVD variant. The test therefore is to avoid re-introducing such
an inconsistency.
The function doesn't alter its input buffer: Reflect this in its
prototype. To avoid using any kind of cast, simply calculate the update
of "line" from the function's input and output.
Speed up gas startup by avoiding runtime allocation of the instances of
type "templates". At the same time cut the memory requirement to just
very little over half (not even accounting for any overhead
notes_alloc() may incur) by reusing the "end" slot of a preceding entry
for the "start" slot of the subsequent one.
Now that the table is local to gas, ARRAY_SIZE() can be used to
determine the end of the table. Re-arrange the processing loop in
md_begin() accordingly, at the same time folding the two calls to
notes_alloc() into just one.
Unlike many other architectures, x86 does not share an opcode table
between assembly and disassembly. Any consumer of libopcodes would only
ever access one of the two. Since gas is the only consumer of the
assembly data, move it there. While doing so mark respective entities
"static" in i386-gen (we may want to do away with i386_regtab_size
altogether).
This also shrinks the number of relocations to be processed for
libopcodes.so by about 30%.
With the general use of C99 there's no need anymore to have i386-gen
produce these. For more frequently used ones introduce local #define-s,
while others are simply spelled out directly. While doing this move
some static constants into more narrow scopes.
Note that as a "side effect" this corrects type_names[]'es imm8s entry.
For one, because of CheckRegSize, there's no need to check the size of
both (register) operands. And then in process_suffix() check opcode
space rather than the (potentially ambiguous) extension opcode.
Leverage the C (commutative) attribute to also reduce the number of XCHG
and TEST templates we have. This way the reg <-> r/m (and reg <-> reg for
XCHG) forms can also be folded into a single template each, utilizing D.
(Ab)using i386_opcode_modifier for this has been overkill, as the logic
doesn't really require the full structure. With the removal of
LONG_DOUBLE_MNEM_SUFFIX and No_ldSuf there's no good reason at all
anymore to pull out such a loop invariant: We're dealing a check of a
bit in the loop for a simple comparison. Do the original compares inside
the loop, thus also making it easier to understand what is actually
being checked.
With the removal of its use for FPU insns the suffix is now finally
properly misnamed. Drop its use altogether, replacing it by a separate
boolean instead.
As a comment near the top of match_template() already says: We really
only need this pseudo-suffix for far branch handling. Stop "deriving" it
for floating point insns. (Don't bother renaming the now properly
misnamed LONG_DOUBLE_MNEM_SUFFIX, to e.g. FAR_BRANCH_SUFFIX - it's going
to disappear anyway.)
In "x86/Intel: restrict suffix derivation" I think I screwed up
slightly, bringing a piece of code out of sync with its comment, and
resulting in a suffix potentially being derived when one isn't needed.
At the very least a comment in process_operands() is stale. Beyond that
there are effectively two options:
1) It is possible that FADDP and FMULP were mistakenly not marked as
being in need of dealing with the compiler anomaly, and hence the
respective templates weren't removed at the time when they should
have been.
2) It is also possible that there are indeed uses known beyond compiler
generated output for these two commutative opcodes, and hence the
templates need to stay.
To be on the safe side assume 2: Update the comment and fold the
templates into their "normal" ones (utilizing D), adjusting consuming
code accordingly.
For FMULP also add a comment paralleling a similar one FADDP has.
There are just 4 templates using it, which can be easily identified by
other means, as D is set only on a very limited number of FPU templates.
Also move the respective conditional out of the code path taken by all
"reverse match" insns (it probably should have been this way already
before, to avoid the one conditional in the common case).
With this the templates which had FloatR dropped no longer differ from
their AT&T syntax + mnemonic counterparts - the only difference is now
which of the two would be recognized. For this, however, we don't need
two templates - we can simply arrange the condition for setting
Opcode_FloatR accordingly.
Import include/xtensa-dynconfig.h that defines XCHAL_* macros as fields
of a structure returned from the xtensa_get_config_v<x> function call.
Define that structure and fill it with default parameter values
specified in the include/xtensa-config.h.
Define reusable function xtensa_load_config that tries to load
configuration and return an address of an exported object from it.
Define functions xtensa_get_config_v{1,2} that use xtensa_load_config
to get structures xtensa_config_v{1,2}, either dynamically configured
or the default.
bfd/
* Makefile.am (BFD32_BACKENDS, BFD32_BACKENDS_CFILES): Append
xtensa-dynconfig.c.
* Makefile.in: Regenerate.
* configure: Regenerate.
* configure.ac (xtensa_elf32_be_vec, xtensa_elf32_le_vec): Add
xtensa-dynconfig.lo to the tb.
* elf32-xtensa.c (xtensa-config.h): Replace #include with
xtensa-dynconfig.h.
(XSHAL_ABI, XTHAL_ABI_WINDOWED, XTHAL_ABI_CALL0): Remove
definitions.
* xtensa-dynconfig.c: New file.
* xtensa-isa.c (xtensa-dynconfig.h): New #include.
(xtensa_get_modules): New function.
(xtensa_isa_init): Call xtensa_get_modules instead of taking
address of global xtensa_modules.
gas/
* config/tc-xtensa.c (xtensa-config.h): Replace #include with
xtensa-dynconfig.h.
(XTHAL_ABI_WINDOWED, XTHAL_ABI_CALL0, XTENSA_MARCH_EARLIEST):
Remove definitions.
* config/tc-xtensa.h (xtensa-config.h): Replace #include with
xtensa-dynconfig.h.
* config/xtensa-relax.c (xtensa-config.h): Replace #include with
xtensa-dynconfig.h.
(XCHAL_HAVE_WIDE_BRANCHES): Remove definition.
include/
* xtensa-dynconfig.h: New file.
ld/
* emultempl/xtensaelf.em (xtensa-config.h): Replace #include
with xtensa-dynconfig.h.
(XTHAL_ABI_WINDOWED, XTHAL_ABI_CALL0): Remove definitions.
Commit bb996692bd ("RISC-V/gas: allow generating up to 176-bit
instructions with .insn") tried to start supporting long instructions but
it was insufficient.
1. It heavily depended on the bignum internals (radix of 2^16),
2. It generates "value conflicts with instruction length" even if a big
number instruction encoding does not exceed its expected length and
3. Because long opcode was handled separately (from struct riscv_cl_insn),
some information like DWARF line number correspondence was missing.
To resolve these problems, this commit:
1. Handles bignum (and its encodings) precisely and
2. Incorporates long opcode handling into regular instruction handling.
This commit will be tested on the separate commit.
gas/ChangeLog:
* config/tc-riscv.c (struct riscv_cl_insn): Add long opcode field.
(create_insn) Clear long opcode marker.
(install_insn) Install longer opcode as well.
(s_riscv_insn) Likewise.
(riscv_ip_hardcode): Make big number handling stricter. Length and
the value conflicts only if the bignum size exceeds the expected
maximum length.
First of all make operand_type_register_match() apply to all sized
operands, i.e. in Intel Syntax also to respective memory ones. This
addresses gas wrongly accepting certain SIMD insns where register and
memory operand sizes should match but don't. This apparently has
affected all templates with one memory-only operand and one or more
register ones, both permitting at least two sizes, due to CheckRegSize
not taking effect.
Then also add CheckRegSize to a couple of non-SIMD templates matching
that same pattern of memory-only vs register operands. This replaces
bogus (for Intel Syntax) diagnostics referring to a wrong suffix (when
none was used at all) by "type mismatch" ones, just like already emitted
for insns where the template allows a register operand alongside a
memory one at any particular position.
This also is a prereq to limiting (ideally eliminating in the long run)
suffix "derivation" in Intel Syntax mode.
While making the code adjustment also flip order of checks to do the
cheaper one first in both cases.
Both uniformly only ever take 16-bit memory operands while at the same
time requiring matching (in size) register operands, which then also
should disassemble that way. This in particular requires splitting each
of the templates for the assembler and separating decode of the
register and memory forms in the disassembler.
This commit adds 'Ssstateen' extension, which is a supervisor-visible view
of the 'Smstateen' extension. It means, this extension implements sstateen*
and hstateen* CSRs of the 'Smstateen' extension.
Note that 'Smstateen' extension itself is unchanged but due to
implementation simplicity, it is implemented so that 'Smstateen' implies
'Ssstateen' (just like 'M' implies 'Zmmul').
This is based on the latest version of RISC-V Profiles
(version 0.9-draft, Frozen):
<226b7f6430>
bfd/ChangeLog:
* elfxx-riscv.c (riscv_implicit_subsets): Update implication rules.
(riscv_supported_std_s_ext) Add 'Ssstateen' extension.
gas/ChangeLog:
* config/tc-riscv.c (enum riscv_csr_class): Rename
CSR_CLASS_SMSTATEEN_AND_H{,_32} to CSR_CLASS_SSSTATEEN_...
Add CSR_CLASS_SSSTATEEN.
(riscv_csr_address): Support new/renamed CSR classes.
* testsuite/gas/riscv/csr.s: Add 'Ssstateen' extension to comment.
* testsuite/gas/riscv/csr-version-1p9p1.l: Reflect changes to
error messages.
* testsuite/gas/riscv/csr-version-1p10.l: Likewise.
* testsuite/gas/riscv/csr-version-1p11.l: Likewise.
* testsuite/gas/riscv/csr-version-1p12.l: Likewise.
* testsuite/gas/riscv/ssstateen-csr.s: Test for 'Ssstateen' CSRs.
* testsuite/gas/riscv/ssstateen-csr.d: Likewise.
* testsuite/gas/riscv/smstateen-csr-s.d: Test to make sure that
supervisor/hypervisor part of 'Smstateen' CSRs are accessible from
'RV32IH_Smstateen', not just from 'RV32IH_Ssstateen' that is tested
in ssstateen-csr.d.
include/ChangeLog:
* opcode/riscv-opc.h: Update DECLARE_CSR declarations with
new CSR classes.
SFrame is supported for ELF only. Keep the definitions and declarations
guarded with OBJ_ELF consistently.
ChangeLog:
* gas/config/tc-aarch64.h: Guard SFrame related definitions
with OBJ_ELF.
Currently supported for x86_64 and aarch64 only.
[PS: Currently, the compiler has not been adapted to generate
".cfi_sections" with ".sframe" in it. The newly added command line
option of --gsframe provides an easy way to try out .sframe support
in the toolchain.]
gas interprets the CFI directives to generate DWARF-based .eh_frame
info. These internal DWARF structures are now consumed by
gen-sframe.[ch] sub-system to, in turn, create the SFrame unwind
information. These internal DWARF structures are read-only for the
purpose of SFrame unwind info generation.
SFrame unwind info generation does not impact .eh_frame unwind info
generation. Both .eh_frame and .sframe can co-exist in an ELF file,
if so desired by the user.
Recall that SFrame unwind information only contains the minimal
necessary information to generate backtraces and does not provide
information to recover all callee-saved registers. The reason being
that callee-saved registers other than FP are not needed for stack
unwinding, and hence are not included in the .sframe section.
Consequently, gen-sframe.[ch] only needs to interpret a subset of
DWARF opcodes in gas. More details follow.
[Set 1, Interpreted] The following opcodes are interpreted:
- DW_CFA_advance_loc
- DW_CFA_def_cfa
- DW_CFA_def_cfa_register
- DW_CFA_def_cfa_offset
- DW_CFA_offset
- DW_CFA_remember_state
- DW_CFA_restore_state
- DW_CFA_restore
[Set 2, Bypassed] The following opcodes are acknowledged but are not
necessary for generating SFrame unwind info:
- DW_CFA_undefined
- DW_CFA_same_value
Anything else apart from the two above-mentioned sets is skipped
altogether. This means that any function containing a CFI directive not
in Set 1 or Set 2 above, will not have any SFrame unwind information
generated for them. Holes in instructions covered by FREs of a single
FDE are not representable in the SFrame unwind format.
As few examples, following opcodes are not processed for .sframe
generation, and are skipped:
- .cfi_personality*
- .cfi_*lsda
- .cfi_escape
- .cfi_negate_ra_state
- ...
Not processing .cfi_escape, .cfi_negate_ra_state will cause SFrame
unwind information to be absent for SFrame FDEs that contain these CFI
directives, hence affecting the asynchronicity.
x86-64 and aarch64 backends need to have a few new definitions and
functions for .sframe generation. These provide gas with architecture
specific information like the SP/FP/RA register numbers and an
SFrame-specific ABI marker.
Lastly, the patch also implements an optimization for size, where
specific fragments containing SFrame FRE start address and SFrame FDE
function are fixed up. This is similar to other similar optimizations
in gas, where fragments are sized and fixed up when the associated
symbols can be resolved. This optimization is controlled by a #define
SFRAME_FRE_TYPE_SELECTION_OPT and should be easy to turn off if needed.
The optimization is on by default for both x86_64 and aarch64.
ChangeLog:
* gas/Makefile.am: Include gen-sframe.c and sframe-opt.c.
* gas/Makefile.in: Regenerated.
* gas/as.h (enum _relax_state): Add new state rs_sframe.
(sframe_estimate_size_before_relax): New function.
(sframe_relax_frag): Likewise.
(sframe_convert_frag): Likewise.
* gas/config/tc-aarch64.c (aarch64_support_sframe_p): New
definition.
(aarch64_sframe_ra_tracking_p): Likewise.
(aarch64_sframe_cfa_ra_offset): Likewise.
(aarch64_sframe_get_abi_arch): Likewise.
(md_begin): Set values of sp/fp/ra registers.
* gas/config/tc-aarch64.h (aarch64_support_sframe_p): New
declaration.
(support_sframe_p): Likewise.
(SFRAME_CFA_SP_REG): Likewise.
(SFRAME_CFA_FP_REG): Likewise.
(SFRAME_CFA_RA_REG): Likewise.
(aarch64_sframe_ra_tracking_p): Likewise.
(sframe_ra_tracking_p): Likewise.
(aarch64_sframe_cfa_ra_offset): Likewise.
(sframe_cfa_ra_offset): Likewise.
(aarch64_sframe_get_abi_arch): Likewise.
(sframe_get_abi_arch): Likewise.
* gas/config/tc-i386.c (x86_support_sframe_p): New definition.
(x86_sframe_ra_tracking_p): Likewise.
(x86_sframe_cfa_ra_offset): Likewise.
(x86_sframe_get_abi_arch): Likewise.
* gas/config/tc-i386.h (x86_support_sframe_p): New declaration.
(support_sframe_p): Likewise.
(SFRAME_CFA_SP_REG): Likewise.
(SFRAME_CFA_FP_REG): Likewise.
(x86_sframe_ra_tracking_p): Likewise.
(sframe_ra_tracking_p): Likewise.
(x86_sframe_cfa_ra_offset): Likewise.
(sframe_cfa_ra_offset): Likewise.
(x86_sframe_get_abi_arch): Likewise.
(sframe_get_abi_arch): Likewise.
* gas/config/tc-xtensa.c (unrelaxed_frag_max_size): Add case for
rs_sframe.
* gas/doc/as.texi: Add .sframe to the documentation for
.cfi_sections.
* gas/dw2gencfi.c (cfi_finish): Create a .sframe section.
* gas/dw2gencfi.h (CFI_EMIT_sframe): New definition.
* gas/write.c (cvt_frag_to_fill): Handle rs_sframe.
(relax_segment): Likewise.
* gas/gen-sframe.c: New file.
* gas/gen-sframe.h: New file.
* gas/sframe-opt.c: New file.
Without commit b66e671854, this:
.p2align 4
nop
.p2align 3
nop
results in an error when coff_frob_section attempts to pad out the
section to a 16-byte boundary. Due to miscalculating the pad pattern
repeat count, write.c:write_contents attempts to shove 16 bytes of
padding into the remaining 4 bytes of the .text section.
* config/obj-coff.c (coff_frob_section): Correct fill count.
Don't pad after errors.
Attributes which aren't used together in any single insn template can be
converted from individual booleans to a single enum, as was done for a few
other attributes before. This is more space efficient. Collect together
all attributes which express special operand constraints (and which fit
the criteria for folding).
This patch adds the assembler support for the new unwinder
directive ".pacspval" and encodes this directives with opcode
"0xb5". This opcode indicates the unwinder to use effective
vsp as modifier for PAC validation.
gas/ChangeLog:
2022-11-07 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
* doc/c-arm.texi: Document directive.
* config/tc-arm.c (s_arm_unwind_pacspval): Define function.
(md_pseudo_table): Add entry for pacspval directive.
* testsuite/gas/arm/ehabi-pacbti-m.d: New test.
* testsuite/gas/arm/ehabi-pacbti-m.s: Likewise.
(%dx) isn't a valid memory address in any modes. It is used as a special
memory operand for input/output port address in AT&T syntax and should
only be used with input/output instructions. Update i386_att_operand to
set i.input_output_operand to true for (%dx) and issue an error if (%dx)
is used with non-input/output instructions.
PR gas/29751
* config/tc-i386.c (_i386_insn): Add input_output_operand.
(md_assemble): Issue an error if input/output memory operand is
used with non-input/output instructions.
(i386_att_operand): Set i.input_output_operand to true for
(%dx).
* testsuite/gas/i386/inval.l: Updated.
* testsuite/gas/i386/x86-64-inval.l: Likewise.
* testsuite/gas/i386/inval.s: Add tests for invalid (%dx) usage.
* testsuite/gas/i386/x86-64-inval.s: Likewise.
Operand swapping was mistakenly suppressed when the first two operands
were immediate ones, not taking into account overall operand count. This
way EXTRQ / INSERTQ would have been accepted also with kind-of-AT&T
operand order.
For the testcase being extended, in order to not move around "GAS
LISTING" expectations, suppress pagination.
Comparing the sum of the relevant .imm<N> fields against a constant imo
makes more obvious what is actually meant. It allows dropping of two
static variables, with a 3rd drop requiring two more minor adjustments
elsewhere, utilizing that "i" is zeroed first thing in md_assemble().
This also increases the chances of the compiler doing the calculations
all in registers.
Consider the case,
.option arch, rv32i
.option norelax
.option arch, +c
.byte 1
.align 2
addi a0, zero, 1
Assembler adds $d for the odd .byte, and then adds $x+arch for the
alignment. Since norelax, riscv_add_odd_padding_symbol will add the
$d and $x for the odd alignment, but accidently remove the $x+arch because
it has the same address as $d. Therefore, we will get the unexpected result
before applying this patch,
.byte 1 # $d
.align 2 # odd alignment, $xrv32ic replaced by $d + $x
After this patch, the expected result should be,
.byte 1 # $d
.align 2 # odd alignment, $xrv32ic replaced by $d + $xrv32ic
gas/
* config/tc-riscv.c (make_mapping_symbol): If we are adding mapping symbol
for odd alignment, then we probably will remove the $x+arch by accidently
when it has the same address of $d. Try to add the removed $x+arch back
after the $d rather than just $x.
(riscv_mapping_state): Updated since parameters of make_mapping_symbol are
changed.
(riscv_add_odd_padding_symbol): Likewise.
(riscv_remove_mapping_symbol): Removed and moved the code into the
riscv_check_mapping_symbols.
(riscv_check_mapping_symbols): Updated.
* testsuite/gas/riscv/mapping-dis.d: Updated and added new testcase.
* testsuite/gas/riscv/mapping-symbols.d: Likewise.
* testsuite/gas/riscv/mapping.s: Likewise.
Earlier tidying still missed an opportunity: There's no need for the
"anyimm" static variable. Instead of using it in the loop to mask
"allowed" (which is necessary to satisfy operand_type_or()'s assertions)
simply use "mask", requiring it to be calculated first. That way the
post-loop masking by "mask" ahead of the operand_type_all_zero() can be
dropped.
