udis86 cannot into SSE

[ghost]

https://www.dropbox.com/s/aolwqzmqo7ffw7p/test
https://www.dropbox.com/s/yxi4lhjlhwfai2n/test.c

[vmt]

Hi, sorry, I couldn't make sense of you report. Can you describe the problem?

[radare]

This bug was reported in radare2 before. The problem is that udis86 doesn't seems to properly disassemble SSE instructions. If you go to 0x00000824 you'll see something like this:

      | |   0x00000824      8b85fcfeffff     mov eax, [rbp-0x104]
      | |   0x0000082a      8b9d00ffffff     mov ebx, [rbp-0x100]
      | |   0x00000830      899dd8feffff     mov [rbp-0x128], ebx
            0x00000836      c5               invalid
            0x00000837      f9               stc
            0x00000838      6e               outsb
            0x00000839      95               xchg ebp, eax
            0x0000083a      d8fe             fdivr st0, st6
            0x0000083c      ff               invalid
            0x0000083d      ffc4             inc esp
            0x0000083f      e369             jrcxz 0x8aa
            0x00000841      22c0             and al, al
            0x00000843      018b8504ffff     add [rbx-0xfb7b], ecx
            0x00000849      ff8b9d08ffff     dec dword [rbx-0xf763]
            0x0000084f      ff899dd8feff     dec dword [rcx-0x12763]
            0x00000855      ffc5             inc ebp
            0x00000857      f9               stc

According to objdump, this should be disassembled as:

 830:   89 9d d8 fe ff ff       mov    %ebx,-0x128(%rbp)
 836:   c5 f9 6e 95 d8 fe ff    vmovd  -0x128(%rbp),%xmm2
 83d:   ff
 83e:   c4 e3 69 22 c0 01       vpinsrd $0x1,%eax,%xmm2,%xmm0
 844:   8b 85 04 ff ff ff       mov    -0xfc(%rbp),%eax
 84a:   8b 9d 08 ff ff ff       mov    -0xf8(%rbp),%ebx
 850:   89 9d d8 fe ff ff       mov    %ebx,-0x128(%rbp)
 856:   c5 f9 6e 9d d8 fe ff    vmovd  -0x128(%rbp),%xmm3
 85d:   ff
 85e:   c4 e3 61 22 c8 01       vpinsrd $0x1,%eax,%xmm3,%xmm1
 864:   c5 f1 6c c0             vpunpcklqdq %xmm0,%xmm1,%xmm0
 868:   c5 f9 7f 85 70 ff ff    vmovdqa %xmm0,-0x90(%rbp)
 86f:   ff
 870:   c5 f9 6f 85 70 ff ff    vmovdqa -0x90(%rbp),%xmm0
 877:   ff
 878:   c5 f9 7f 85 40 ff ff    vmovdqa %xmm0,-0xc0(%rbp)
 87f:   ff
 880:   8b 45 bc                mov    -0x44(%rbp),%eax
 883:   8b 55 b4                mov    -0x4c(%rbp),%edx

[XVilka]

Any updates/decisions on this?

[vmt]

Can you try top of tree?

$ echo c5 f9 7f 85 70 ff ff ff | udcli/udcli -x -32
0000000000000000 c5f97f8570ffffff vmovdqa [ebp-0x90], xmm0

$ echo c5 f1 6c c0 | udcli/udcli -x -32
0000000000000000 c5f16cc0         vpunpcklqdq xmm0, xmm1, xmm0

[ghost]

echo c5 f9 7f 85 70 ff ff ff | udcli -x -32
0000000000000000 c5f97f8570ffffff vmovdqa [ebp-0x90], xmm0

echo c5 f1 6c c0 | udcli -x -32
0000000000000000 c5f16cc0 vpunpcklqdq xmm0, xmm1, xmm0

Looks good

[radare]

Looks good in r2 too after updating to latest udis86 code. Thanks! IMHO you can close the bug

$ rasm2 -d "c4 e3 69 22 c0 01"
vpinsrd xmm0, xmm2, eax, 0x1
$ rasm2 -d c5f97f8570ffffff
vmovdqa [ebp-0x90], xmm0
$ rasm2 -d c5f16cc0
vpunpcklqdq xmm0, xmm1, xmm0

[vmt]

Thanks!