Status of stack right after call instruction is executed
- At this point, RSP ends in 8
- The function prologue will push 8 bytes to the stack so RSP ends in 0
- Then the function code can assume the stack is aligned

Note: stack grows downwards (towards lower addresses). Pusing something decreases the stack pointer)
RBP is used to refer to what's already on the stack (frame pointer). It is fixed througout a function
execution and allows to refer to local variables at fixed offsets from the register (whereas rsp
keeps moving as the code pushes and pops stuff).

Expected outcome:

^ HIGH ADDRESSES

------------------------- (stack misaligned here)
(hole for realigning the stack)
------------------------- (stack aligned on 16 bytes here)
USER_DATA_SELECTOR
user_rsp
flags
USER_CODE_SELECTOR
ip                  < ...
error_code          < RBP+8
vector              < RBP+4
iframe bottom       < RBP (ends in 0)
------------------------- (stack aligned on 16 bytes here)
return address      < RSP (ends in 8)

v LOW ADDRESSES



With current code:

^ HIGH ADDRESSES

------------------------- (stack misaligned here)
USER_DATA_SELECTOR
user_rsp
flags
USER_CODE_SELECTOR
ip                  < ...
error_code          < RBP+8
vector              < RBP+4
iframe bottom       < RBP (ends in 8)
(hole for realigning the stack)
------------------------- (stack aligned on 16 bytes here)
return address      < RSP (ends in 8)

v LOW ADDRESSES


By using andq/movq:

^ HIGH ADDRESSES

------------------------- (stack misaligned here)
USER_DATA_SELECTOR
user_rsp
flags
USER_CODE_SELECTOR
ip
error_code
vector
iframe bottom
(hole for realigning the stack) < RBP (now at the wrong place)
return address      < RSP (ends in 8)

v LOW ADDRESSES