Is it possible to keep redundant prefixes for an assembly instruction in NASM?

Question

I need to test if a windows emulator correctly identifies and handles an instruction that is too long (more than 15 bytes long). I compile the code with NASM. The only way I know of specifying an instruction that is too long is by adding prefixes to the instruction. Let's say I do the following:

rep O32 rep O32 rep O32 rep O32 rep O32 mov eax, [fs:0]

In this case, I should get a 16-byte-long instruction, since mov eax, [fs:0] is a 6-byte-long instruction, plus the 10 O32 and rep prefixes equal 16 bytes.

At least that's the theory ... but when I compile this with NASM and disassemble the resulting binary, what I find is that the unnecessary prefixes got plucked from there: only 1 rep prefix stays and obviously the fs segment prefix. Thus, the total length of the instruction is reduced to just 7 bytes. I guess this is how it is supposed to work, but is there a way to make it so the instruction length actually grows beyond bounds?

Answer 1

You state that mov eax, [fs:0] is a 6-byte-long instruction. This implies that you're using 16-bit code. (32-bit code would have taken 7 bytes)

Thus mov ax,1 will use 3 bytes. You've prepended 12 operand size prefixes. This totals up to 15 which is still perfectly legal. Therefore you've not yet tested your 'windows emulator'.

If the solution presented is 32-bit code then the total will be 16 bytes.

Just curious : did the emulator complain?