This was the following in fmtconst_val, x is unsigned int.
x = SIGNEXTEND (x, constant_formats[cf].nbits);
Problem is, the SIGNEXTEND macro assumed its arg was a long and sign
extended by shifting left then shifting right, and didn't cast the
arg. So don't do the silly shift thing. It's not guaranteed to work
anyway according to the C standard. ">>" might do a logical shift
even if its args are signed.
* bfin-dis.c (HOST_LONG_WORD_SIZE, XFIELD): Delete.
(SIGNBIT): New.
(MASKBITS, SIGNEXTEND): Rewrite.
(fmtconst): Don't use ? expression now that SIGNEXTEND uses
unsigned arithmetic, instead assign result of SIGNEXTEND back
to x.
(fmtconst_val): Use 1u in shift expression.
include * dis-asm.h (struct disassemble_info): Add stop_vma field.
binuti * objdump.c (disassemble_bytes): Set the stop_vma field in the
disassemble_info structure when disassembling code sections with
-d.
* doc/binutils.texi (objdump): Document the discrepancy between -d
and -D.
opcodes * dis-buf.c (buffer_read_memory): Fail is stop_vma is set and the
requested region lies beyond it.
* bfin-dis.c (print_insn_bfin): Ignore sysop instructions when
looking for 32-bit insns.
* mcore-dis.c (print_insn_mcore): Disable stop_vma when reading
data.
* sh-dis.c (print_insn_sh): Likewise.
* tic6x-dis.c (print_insn_tic6x): Disable stop_vma when reading
blocks of instructions.
* vax-dis.c (print_insn_vax): Check that the requested address
does not clash with the stop_vma.
tests * gas/arm/backslash-at.s: Add extra .byte directives so that the
foo symbol does not appear to point half way through an
instruction.
* gas/arm/backslash-at.d: Update expected disassembly.
* gas/i386/ilp32/x86-64-opcode-inval-intel.d: Likewise.
* gas/i386/ilp32/x86-64-opcode-inval.d: Likewise.
* gas/i386/x86-64-opcode-inval-intel.d: Likewise.
* gas/i386/x86-64-opcode-inval.d: Likewise.
These multiple if statements can be condensed down into a single if
statement and an array of strings.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
The variables used to track insn state should be pushed down into the
private_data structure to avoid pollution across calls.
This also happens to fix the output when hitting comments/invalid insns
which needs to tweak a gas test.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
If the user gives us an unaligned PC, then dump an error as such.
Otherwise if you try to disassemble at an odd address, the output
will look weird (it'll read one byte earlier).
This can be seen in one of the gas tests where data is in the middle
of .text, so move the data to .data like it should be in the first place.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
The current code ignores memory read errors which isn't a great idea.
So add a helper function which takes care of error checking and update
the code to use that.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
This fix the build time warning:
warning: format not a string literal, argument types not checked [-Wformat-nonliteral]
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
If the MAC1 part of the insn is disabled, then the (M) flag is ignored.
Rather than include it in the decode, move the MM clearing to the MAC0
portion of the code.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Parts of the disassembler rely on the disasm info never being NULL (such
as being able to read memory to disassemble in the first place). So drop
useless null checks in the OUTS helper.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
We have an OUTS helper to handle outf fprintf_func logic, so conver the
few places not using it over.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Mark the state static, punt unused members, unify indexable register
lookups, and abort when there is a register lookup failure. Otherwise
we return NULL and the calling code assumes a valid pointer is returned.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
The LoopSetup insn is only valid when the reg field is 0-7, so
don't go decoding it incorrectly when reg is 8-15.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
The single cycle dual mac ABS insn was incorrectly decoding the mac1
part of the insn.
Once we fix the decode, update the gas tests to have the correct output.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
When assigning to a register half, the mac0 part of the mult insn
was not decoding properly. It would always show a full dreg instead
of the dreg low half.
Once we fix the disassembler, we have to update a few of the gas
tests as their previous expected output was incorrect.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
The BYTEOP2M insn was part of the initial Blackfin designs, but never made
it into any actual silicon. So punt support for it.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
There never was a "GP" register, so punt it from the decode map. It's
a hold over from a very old processor definition and never made it into
actual silicon.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
The mmod field is decoded in a few places (gas/opcodes/sim), so move it to
a common place to avoid duplication.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
The BYTEOP2M parser incorrectly calls BYTEOP2P to generate the opcode.
Once we've fixed that, it's easy to see that the disassembler also likes
to decode this insn incorrectly. So fix that and then add some tests.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
The Blackfin ISA is very exact with regards to address truncation when
under/over flowing its 32bit range. On a 32bit system, things work the
same and so addresses are decoded properly. On a 64bit system though,
the decoded addresses may include the bits that are supposed to have
been truncated. So force a 32bit truncation after the address has been
calculated.
Signed-off-by: Robin Getz <robin.getz@analog.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Many register move insns were not being decoded properly, so rewrite
the whole function to be a bit more manageable in terms of valid
combinations.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
The Blackfin disassembler was originally based on the premise of parsing
valid opcodes all the time, so some of the opcode checking can be a bit
fuzzy. This is exemplified in decoding of parallel insns where many
times things are decoded as invalid when in reality, they may not be
used in parallel combinations. So add parallel checking to most insn
decoding routines so we see ILLEGAL and not just whatever insn happens
to be close to a valid mnemonic, as well as some additional sub-opcode
checks.
Signed-off-by: Robin Getz <robin.getz@analog.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
The push/pop multiple insn has a 3 bit field for the P register range,
but only values of 0...5 are valid (P0 - P5). There is no such P6 or
P7 register, so mark these insns as illegal.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
The saturation bit was missed when decoding a vector shift insn
leading to the output looking the same as the non-saturating insn.
Signed-off-by: Robin Getz <robin.getz@analog.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
All ASTAT bits work in the hardware even though they aren't part of the
official Blackfin ISA. So decode every ASTAT field to make the output
a bit nicer when working with hand generated opcodes.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Sometimes the encoding in the opcode is a 4 bit field which defines a
register number. However, register numbers are only 0-7, so make sure
we call illegal for when the opcode register number is greater than 8.
Signed-off-by: Robin Getz <robin.getz@analog.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Some extended registers when given to the DBG/DBGCMPLX pseudo insns are
not encoded properly. So fix them, fix the display of them when being
disassembled, and add testcases.
Signed-off-by: Robin Getz <robin.getz@analog.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Support a few more ASTAT bits with the standard insns that operate on
ASTAT bits directly.
Signed-off-by: Robin Getz <robin.getz@analog.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
The disassembler has partial (but incomplete/broken) support already for
the pseudo debug insn OUTC, so let's fix it up and finish it. And now
that the disassembler can handle it, make sure our assembler can output
it too.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
The Blackfin ISA does not have a "SHIFT" insn, it has either LSHIFT,
ASHIFT, or BXORSHIFT. So be specific when disassembling.
As fall out of this change, we need to update some assembler tests.
Signed-off-by: Robin Getz <robin.getz@analog.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
No need for these local structures related to formatting of output to
be writable, so constify the whole shebang.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
* config/bfin-parse.y (asm_1): Implement HLT instruction.
Fix comments for DBGA, DBGAH and DBGAL.
* config/tc-bfin.c (bfin_gen_pseudodbg_assert): Change according
to the new encoding of DBGA, DBGAH, and DBGAL.
include/
* opcode/bfin.h (PseudoDbg_Assert): Add bits_grp and mask_grp.
(PseudoDbg_Assert_grp_bits, PseudoDbg_Assert_grp_mask): Define.
(PseudoDbg_Assert_dbgop_bits, PseudoDbg_Assert_dbgop_mask,
PseudoDbg_Assert_dontcare_bits, PseudoDbg_Assert_dontcare_mask):
Adjust accordingly.
(init_PseudoDbg_Assert): Add PseudoDbg_Assert_grp_bits and
PseudoDbg_Assert_grp_mask.
opcodes/
* bfin-dis.c (decode_pseudodbg_assert_0): Change according
to the new encoding of DBGA, DBGAH, and DBGAL.
(_print_insn_bfin): Likewise.
