If you are going to use P2 for data AND control you have to ensure that when you move the data to the port that you do not destroy the RS & RW settings.
In "lcd_data:" you set up RW & RS at the port, isolate the upper and lower nibbles to U & L and then move them to A as a byte. In "loop:" you simply dump A to Port 2, destroying RS & RW.

When you move U or L to A, set bits 5 & 6 in A so that RS & RW won't be modified when you dump A to the port. In fact, as long as you are going to dump A to the port, defining RS & RW at the port is meaningless, because A is going to override whatever you set P2.4 and P2.5 to.

Hal