I just had a closer look at your schematic.
TL072 is a bad opamp to use with a grounded power supply. For this opamp there should be at least 3 volt headroom between the inputs and either power supply voltage. It's not a bad opamp, but it is definately not a rail-to-rail opamp.
An opamp like the LM358 can work with input voltages very near to the GND level and is a better choice.
Had an extra peek at its datasheet, and it looks like the LM358 still works OK with an input voltage of 300mV below the GND pin.
You have also connected "everything" to the same 12V rails.
Consider this scenario:
The MOSfets draws a lot of current from the 12V rails and it sags a bit. which results in a lower voltage on the non-inverting input on the opamp via the potentiometer. A lower voltage on the potentiometer results in the opamp reduicing the gate voltage to the MOSfet, which may result in a rise on the 12V rails and the cycle repeats, which is a cause for oscillations.
As MarkF already wrote, it's much better to use a stable reference for the potentiometer, which decouples the current setpoint from the power supply. An easy choise would be 1.25V from an LM317L (in TO92) or even a simple led with resitor and a stabilizing cap. (LED voltage depends mostly on the color of the LED.
After a reference voltage has been chosen, then calculate the right current shunt resistor for the wanted maximum current at that reference voltage.