I am poking around some PS2 ELFs and I found this weird issue where Ghidra seems to be doubling the value of this byte for some reason.
Here's what Ghidra shows:
undefined v0_lo:1 <RETURN>
Money_get XREF: Entry Point(*)
001bcb90 ec 00 01 3c lui at,0xec
001bcb94 98 00 03 3c lui v1,0x98
001bcb98 e8 aa 25 8c lw a1,-0x5518(at)=>DAT_00ebaae8
And what IDA shows:
main:001BCB90 Money_get: # CODE XREF: plmove_find_02+178↑p
main:001BCB90 # Item_get+70↑p
main:001BCB90 lui $at, 0x76
main:001BCB94 lui $v1, 0x98
main:001BCB98 lw $a1, 0x7657A8
What IDA shows is the correct disassembly. The first line is supposed to be
lui $at, 0x76
But for some reason Ghidra has doubled the 0x76 to 0xEC. And because of that the address loaded in the 3rd line is incorrect. I've double checked that the input files are identical and I know IDA is the correct one because the gameshark code for this particular game modifies that address.
I even confirmed in a hex editor that the byte should be 0x76 but in Ghidra's own hex output it shows it as 0xEC. So during input it's doubling it for some reason.
So any reason why Ghidra is doing this? I tried reloading the file and analyzing as a generic MIPS (thinking it was the PS2 ELF loader I used) and it did the same thing.
So I did some more testing and when I strip off the ELF header and load the file as a raw binary then Ghidra loads it correctly. So I wonder what in the ELF header would cause Ghidra to modify that byte?