Possible bugs in datafiles and xed code

[a0rland0]

Hi,
In file "datafiles\xed-addressing-modes-new.txt" (line 483) shouldn't these lines be:

REXB4=1 REXB=0 SIBBASE=0b101 MOD=0b01 | BASE0=Ar21() SEG0=FINAL_DSEG() DISP_WIDTH=8
REXB4=1 REXB=0 SIBBASE=0b101 MOD=0b10 | BASE0=Ar21() SEG0=FINAL_DSEG() DISP_WIDTH=32

instead of:

REXB4=1 REXB=0 SIBBASE=0b101 MOD=0b01 | BASE0=Ar21() SEG0=FINAL_SSEG() DISP_WIDTH=8
REXB4=1 REXB=0 SIBBASE=0b101 MOD=0b10 | BASE0=Ar21() SEG0=FINAL_SSEG() DISP_WIDTH=32

The same logic is applied to Ar13() and Ar29(), so I wonder if this is an error.
Thanks.

[a0rland0]

Hi,
A few other things that might be wrong too (imo).

1. In xed-decode.c the function "check_all_regs_match" should be ifdef'd with XED_ATTRIBUTE_NO_REG_MATCH_DEFINED instead of XED_ATTRIBUTE_NO_SRC_DEST_MATCH_DEFINED. This will make the code consistent with function xed_decode_finalize_operand_storage_fields.

2. NOP5 (and NOP6, etc.) will not work properly in 16-bit mode. A valid NOP using 5 bytes would be 66 0F 1F 44 00 i.e.

{
ICLASS : NOP5
CPL : 3
CATEGORY : WIDENOP
EXTENSION : BASE
ISA_SET : FAT_NOP
PATTERN : 0x0F 0x1F 0x44 0x00 66_prefix mode16
OPERANDS :
PATTERN : 0x0F 0x1F 0x44 0x00 0x00 mode32
OPERANDS :
PATTERN : 0x0F 0x1F 0x44 0x00 0x00 mode64
OPERANDS :
}

Thanks.

[kkhalail]

Hi, A few other things that might be wrong too (imo).

1. In xed-decode.c the function "check_all_regs_match" should be ifdef'd with XED_ATTRIBUTE_NO_REG_MATCH_DEFINED instead of XED_ATTRIBUTE_NO_SRC_DEST_MATCH_DEFINED. This will make the code consistent with function xed_decode_finalize_operand_storage_fields.
2. NOP5 (and NOP6, etc.) will not work properly in 16-bit mode. A valid NOP using 5 bytes would be 66 0F 1F 44 00 i.e.

{ ICLASS : NOP5 CPL : 3 CATEGORY : WIDENOP EXTENSION : BASE ISA_SET : FAT_NOP PATTERN : 0x0F 0x1F 0x44 0x00 66_prefix mode16 OPERANDS : PATTERN : 0x0F 0x1F 0x44 0x00 0x00 mode32 OPERANDS : PATTERN : 0x0F 0x1F 0x44 0x00 0x00 mode64 OPERANDS : }

Thanks.

Hi,
your input is greatly appreciated! The first issue has been fixed and will be updated in an upcoming release! Other issues will be under investigation soon.

[kkhalail]

Hi, In file "datafiles\xed-addressing-modes-new.txt" (line 483) shouldn't these lines be:

REXB4=1 REXB=0 SIBBASE=0b101 MOD=0b01 | BASE0=Ar21() SEG0=FINAL_DSEG() DISP_WIDTH=8 REXB4=1 REXB=0 SIBBASE=0b101 MOD=0b10 | BASE0=Ar21() SEG0=FINAL_DSEG() DISP_WIDTH=32

instead of:

REXB4=1 REXB=0 SIBBASE=0b101 MOD=0b01 | BASE0=Ar21() SEG0=FINAL_SSEG() DISP_WIDTH=8 REXB4=1 REXB=0 SIBBASE=0b101 MOD=0b10 | BASE0=Ar21() SEG0=FINAL_SSEG() DISP_WIDTH=32

The same logic is applied to Ar13() and Ar29(), so I wonder if this is an error. Thanks.

Thanks for your contribution, time and effort! A fix will be available in an upcoming release.

[kkhalail]

Hi, A few other things that might be wrong too (imo).

1. In xed-decode.c the function "check_all_regs_match" should be ifdef'd with XED_ATTRIBUTE_NO_REG_MATCH_DEFINED instead of XED_ATTRIBUTE_NO_SRC_DEST_MATCH_DEFINED. This will make the code consistent with function xed_decode_finalize_operand_storage_fields.
2. NOP5 (and NOP6, etc.) will not work properly in 16-bit mode. A valid NOP using 5 bytes would be 66 0F 1F 44 00 i.e.

{ ICLASS : NOP5 CPL : 3 CATEGORY : WIDENOP EXTENSION : BASE ISA_SET : FAT_NOP PATTERN : 0x0F 0x1F 0x44 0x00 66_prefix mode16 OPERANDS : PATTERN : 0x0F 0x1F 0x44 0x00 0x00 mode32 OPERANDS : PATTERN : 0x0F 0x1F 0x44 0x00 0x00 mode64 OPERANDS : }

Thanks.

Regarding the FAT NOPS issue:
These instructions are known only to the encoder, basically to provide the ability to encode the SDM recommended multi-byte NOPs.
The SDM lists 2-byte NOPs and up to 9-byte NOPs recommended for modes 32 and 64 only. See reference below:

So we've decided that encoding of these instructions won't be supported in 16b mode. Nonetheless, I suggest the following unofficial workaround: you can always add a sequence of multiple ignored 66prefixes to the NOP in order to generate multi-byte NOPs in 16bit mode.

[a0rland0]

Thank you for the fixes and explanations.