From: Björn Töpel <bjorn@rivosinc.com>

RISC-V has some GNU disassembly quirks, e.g. it requires '-D' to
properly disassemble .2byte directives similar to Arm [1]. Further,
GNU objdump groups RISC-V instruction by 2 or 4 byte chunks, instead
doing byte-for-byte.

Add the required switches, and translate from short/word to bytes when
ARCH is "riscv".

An example how to invoke decodecode for RISC-V:
  $ echo 'Code: bf45 f793 1007 f7d9 50ef 37af d541 b7d9 7097 00c8 (80e7)
  6140' | AFLAGS="-march=rv64imac_zicbom_zihintpause"  \
  ARCH=riscv CROSS_COMPILE=riscv64-linux-gnu- ./scripts/decodecode
  Code: bf45 f793 1007 f7d9 50ef 37af d541 b7d9 7097 00c8 (80e7) 6140
  All code
  ========
     0:   bf45                    c.j     0xffffffffffffffb0
     2:   1007f793                andi    a5,a5,256
     6:   f7d9                    c.bnez  a5,0xffffffffffffff94
     8:   37af50ef                jal     ra,0xf5382
     c:   d541                    c.beqz  a0,0xffffffffffffff94
     e:   b7d9                    c.j     0xffffffffffffffd4
    10:   00c87097                auipc   ra,0xc87
    14:*  614080e7                jalr    ra,1556(ra) # 0xc87624          <-- trapping instruction

  Code starting with the faulting instruction
  ===========================================
     0:   614080e7                jalr    ra,1556(ra)

[1] https://sourceware.org/bugzilla/show_bug.cgi?id=10263

Signed-off-by: Björn Töpel <bjorn@rivosinc.com>
---
 scripts/decodecode | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/scripts/decodecode b/scripts/decodecode
index b28fd2686561..8fe71c292381 100755
--- a/scripts/decodecode
+++ b/scripts/decodecode
@@ -93,6 +93,11 @@ disas() {
 		${CROSS_COMPILE}strip $t.o
 	fi
 
+	if [ "$ARCH" = "riscv" ]; then
+		OBJDUMPFLAGS="-M no-aliases --section=.text -D"
+		${CROSS_COMPILE}strip $t.o
+	fi
+
 	if [ $pc_sub -ne 0 ]; then
 		if [ $PC ]; then
 			adj_vma=$(( $PC - $pc_sub ))
@@ -126,8 +131,13 @@ get_substr_opcode_bytes_num()
 	do
 		substr+="$opc"
 
+		opcode="$substr"
+		if [ "$ARCH" = "riscv" ]; then
+			opcode=$(echo $opcode | tr ' ' '\n' | tac | tr -d '\n')
+		fi
+
 		# return if opcode bytes do not match @opline anymore
-		if ! echo $opline | grep -q "$substr";
+		if ! echo $opline | grep -q "$opcode";
 		then
 			break
 		fi
-- 
2.37.2

Hi Björn,

On 1/19/23 08:47, Björn Töpel wrote:
> From: Björn Töpel <bjorn@rivosinc.com>
>
> RISC-V has some GNU disassembly quirks, e.g. it requires '-D' to
> properly disassemble .2byte directives similar to Arm [1]. Further,
> GNU objdump groups RISC-V instruction by 2 or 4 byte chunks, instead
> doing byte-for-byte.
>
> Add the required switches, and translate from short/word to bytes when
> ARCH is "riscv".
>
> An example how to invoke decodecode for RISC-V:
>    $ echo 'Code: bf45 f793 1007 f7d9 50ef 37af d541 b7d9 7097 00c8 (80e7)
>    6140' | AFLAGS="-march=rv64imac_zicbom_zihintpause"  \
>    ARCH=riscv CROSS_COMPILE=riscv64-linux-gnu- ./scripts/decodecode
>    Code: bf45 f793 1007 f7d9 50ef 37af d541 b7d9 7097 00c8 (80e7) 6140
>    All code
>    ========
>       0:   bf45                    c.j     0xffffffffffffffb0
>       2:   1007f793                andi    a5,a5,256
>       6:   f7d9                    c.bnez  a5,0xffffffffffffff94
>       8:   37af50ef                jal     ra,0xf5382
>       c:   d541                    c.beqz  a0,0xffffffffffffff94
>       e:   b7d9                    c.j     0xffffffffffffffd4
>      10:   00c87097                auipc   ra,0xc87
>      14:*  614080e7                jalr    ra,1556(ra) # 0xc87624          <-- trapping instruction
>
>    Code starting with the faulting instruction
>    ===========================================
>       0:   614080e7                jalr    ra,1556(ra)
>
> [1] https://sourceware.org/bugzilla/show_bug.cgi?id=10263
>
> Signed-off-by: Björn Töpel <bjorn@rivosinc.com>
> ---
>   scripts/decodecode | 12 +++++++++++-
>   1 file changed, 11 insertions(+), 1 deletion(-)
>
> diff --git a/scripts/decodecode b/scripts/decodecode
> index b28fd2686561..8fe71c292381 100755
> --- a/scripts/decodecode
> +++ b/scripts/decodecode
> @@ -93,6 +93,11 @@ disas() {
>   		${CROSS_COMPILE}strip $t.o
>   	fi
>   
> +	if [ "$ARCH" = "riscv" ]; then
> +		OBJDUMPFLAGS="-M no-aliases --section=.text -D"
> +		${CROSS_COMPILE}strip $t.o
> +	fi
> +
>   	if [ $pc_sub -ne 0 ]; then
>   		if [ $PC ]; then
>   			adj_vma=$(( $PC - $pc_sub ))
> @@ -126,8 +131,13 @@ get_substr_opcode_bytes_num()
>   	do
>   		substr+="$opc"
>   
> +		opcode="$substr"
> +		if [ "$ARCH" = "riscv" ]; then
> +			opcode=$(echo $opcode | tr ' ' '\n' | tac | tr -d '\n')
> +		fi
> +
>   		# return if opcode bytes do not match @opline anymore
> -		if ! echo $opline | grep -q "$substr";
> +		if ! echo $opline | grep -q "$opcode";
>   		then
>   			break
>   		fi


Tested-by: Alexandre Ghiti <alexghiti@rivosinc.com>

Thanks, that's very useful!

Alex