CODEWITHSUNDEEP
assemblyx86-64
Alexander Zhak
Alexander Zhak

Reputation: 9282

address-size override with near calls in long mode

Intel manual (Vol.2 paragraph 2.2.1.6) says:

The use of the address-size prefix does not disable RIP-relative addressing. The effect of the address-size prefix is to truncate and zero-extend the computed effective address to 32 bits.

CALL rel32 description says:

Call near, relative, displacement relative to next instruction. 32-bit displacement sign extended to 64-bits in 64-bit mode

Thus, rel32 is immediate rather than address, as operand-size prefix overrides the default value.

But if use address-size override prefix with near call (or jmp), will it zero-extend effective address as described above or it won't have any effect?

Upvotes: 0

Views: 229

Answers (1)

user555045
user555045

Reputation: 64913

That isn't the type of RIP-relative addressing that a 67-prefix might target, a 67-prefix has no effect on call/jmp/jcc offsets (rather it is the 66-prefix that affects, for example, call rel16 vs call rel32).

The type of RIP-relative addressing par 2.2.1.6 talks about is the sort of thing encoded by a ModRM byte that looks like 00***101, for example in something like

mov rax, [eip]
; or,
addr32 mov rax, [eip]
; or if your assembler doesn't let you write either of those,
db 67
mov rax, [rip]

this is one of the cases they meant, where rip-relative addressing is still active but truncated. If addressing reverted completely like how it is in 32bit mode, a ModRM like 00***101 would encode [sword] with no eip-involvement at all.

Upvotes: 1

Related Questions