fasmg examples: error in x64.inc about RIP-addressing

Index > Macroinstructions > fasmg examples: error in x64.inc about RIP-addressing

Author

Thread

zhak

Joined: 12 Apr 2005
Posts: 501
Location: Belarus

zhak 07 May 2017, 18:03

Instructions like the following don't compile:

Code:

  include 'x64.inc'

  use64
  org 0x140000000

  mov eax, [0]
  add eax, [0]

The first one may be encoded as

Code:

A1 00 00 00 00 00 00 00 00

The second one as

Code:

40 03 04 25 00 00 00 00

using SIB-form (see Intel 64 manual, vol 2 paragraph 2.2.1.6 RIP-Relative Addressing)

07 May 2017, 18:03

Tomasz Grysztar

Joined: 16 Jun 2003
Posts: 8466
Location: Kraków, Poland

Tomasz Grysztar 07 May 2017, 18:56

Have you tried it with fasm 1? It signals the same error here and the macros aim at fasm compatibility. You can only choose the absolute (not RIP-relative) addressing with address size specifier:

Code:

  mov eax, [qword 0]

07 May 2017, 18:56

zhak

Joined: 12 Apr 2005
Posts: 501
Location: Belarus

zhak 07 May 2017, 20:14

Thanks, Tomasz. However, I started getting another error

Code:

        add eax, [qword 0]
macro add [21] macro store_instruction [24]:
        if (reg or rm_operand.rm) and x86.REX_REQUIRED
Processed: if (@dest.rm or @src.rm) and x86.REX_REQUIRED
Error: symbol 'rm' is undefined or out of scope.

07 May 2017, 20:14

Tomasz Grysztar

Joined: 16 Jun 2003
Posts: 8466
Location: Kraków, Poland

Tomasz Grysztar 07 May 2017, 21:30

In this case it should work with:

Code:

add eax,[dword 0]

But you're right that there is a bug there when "qword" is used.

07 May 2017, 21:30

Tomasz Grysztar

Joined: 16 Jun 2003
Posts: 8466
Location: Kraków, Poland

Tomasz Grysztar 07 May 2017, 21:49

I have corrected it, now this works the same as in fasm 1:

Code:

use64
add eax, [dword -1]  ; 67 03 04 25 FF FF FF FF
add eax, [qword -1]  ; 03 04 25 FF FF FF FF

07 May 2017, 21:49

zhak

Joined: 12 Apr 2005
Posts: 501
Location: Belarus

zhak 08 May 2017, 14:32

I tried the above instructions with fasm 1, and there's some inconsistency in the output:

Code:

use64
add eax, [dword -1]    ; 67 03 04 25 FF FF FF FF
add eax, [qword -1]    ; 03 04 25 FF FF FF FF
add [dword 0], eax     ; 01 04 25 00 00 00 00
add dword [dword 0], 1 ; 83 04 25 00 00 00 00 01
add [dword 1], eax     ; 01 04 25 01 00 00 00
add dword [dword 1], 1 ; 83 04 25 01 00 00 00 01

address-size prefix is not added for positive moffs

08 May 2017, 14:32

Tomasz Grysztar

Joined: 16 Jun 2003
Posts: 8466
Location: Kraków, Poland

Tomasz Grysztar 08 May 2017, 14:37

zhak wrote:

address-size prefix is not added for positive moffs

The 67h prefix here has an effect equivalent to a choice between sign-extending (when absent) and zero-extending (when present) the 32-bit displacement value into 64-bit address. Therefore the prefix is only needed when 32-bit displacement has the sign bit set but the upper 32 bit need to be zero, like in the case of "[dword -1]" or "[dword 0FFFFFFFFh].

08 May 2017, 14:37

< Last Thread | Next Thread >

Forum Rules:

You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You cannot attach files in this forum
You can download files in this forum