at&t assembly code ebp register undefined behavior

Question

Since starting to learn at&t assembly about a month ago I was taught that the %ebp register starting at %ebp-4 and going down (ex. -8/-12) were the local variables. Now that I am getting deeper into assembly though I am noticing calls to %ebp-10 and other non multiples of 4. I am wondering how this works and what the significance of this is.

0x08048e2b <+6>:    lea    -0x10(%ebp),%eax

0x08048e2b <+6>:    lea    -0x10(%ebp),%eax

Above are two examples from disassembling a program I am working on now using gdb. How is this data used correctly if, as I am deducing, it is only half way through the variable or supposed data. Wouldn't %ebp-8 or %ebp-12 be the correct version instead of %ebp-10?

I feel like there is probably an intuitive answer to this question, but I'm just not seeing it and haven't been able to find any resources online about it.

Answer 1

0x10 is a multiple of four. It's a hexadecimal number equal to 16 in decimal which, last I looked into the matter, was equal to 4x4. I'm old but I'm pretty sure maths hasn't changed that much since I went to school :-)

In any case, things on the stack aren't required to be aligned on a four-byte boundary. Alignment may speed up certain operations, but:

• it isn't a mandatory thing on the x86 platforms (on some platforms, accessing misaligned data causes a fault to be raised); and
• alignment tends to match the argument size (such as a two-byte value "needing" two-byte alignment).