x86 Assembly (Advanced Indexing Mode): How do I access memory at a variable amount ahead of a start address?

Question

I am trying to read a section of memory, byte by byte, starting from the memory address stored in the EAX register. I push the current byte to the EBX register. However, when I execute the following line of code (advanced indexing mode):

movb byteCount(%eax), %ebx

I get the following error:

Program received signal SIGSEGV, Segmentation fault.

It seems that Assembly doesn't supporting using a label in advanced indexing. How would I go about emulating this same action without using a label or register before (%eax)? Below is the data section:

.data
    str:
        .string "abcdefg"
        
    byteCount:
        .int 1

Answer 1

In that context, byteCount is the label address, not the dword that happens to be in memory there. You need to load all runtime-variable things into registers to use them in addressing modes; x86 doesn't do memory-indirect addressing.

   mov    byteCount, %edx           # dword load
   movzbl (%eax, %edx), %ebx        # zero-extending byte load

Or of course you could use add byteCount, %eax and deref (%eax). Or better, you could keep byteCount in a register in the first place, where you need it anyway. That's what registers are for. In x86 assembly, when should I use global variables instead of local variables? (usually you shouldn't).

---

movb into EBX is an error (operand-size mismatch), so certainly that's not what you actually ran. But that would access a byte at bytecount[ (uintptr_t)eax ]. If you'd used just mov, it would be a dword load.

But regardless, the sum of two addresses is rarely a valid address, so it segfaults. Your debugger should have told you the faulting address so you could see it was far from str.

---

Related:

• How i can access a variable data using a variable value in adress like [ var_+[second_byte] ]? (Intel-syntax near duplicate).
• Referencing the contents of a memory location. (x86 addressing modes)
• A couple of questions about [base + index*scale + disp] (AT&T addressing mode syntax)