There are a few minor issues with a linear DC fan control design;
- 1) Linear dropout of BJT type regulators is high, FET type is low
- 2) Linear power dissipation is shared equally at half voltage, but @ 1.2W loss in the regulator is not too hard to cool yet too hot, if you make no attempt. ( hint; put heatsink near the fastest moving air )
- 3) When the fan stalls below say 3V how will know if any active parts inside need forced air cooling at the albeit reduced current consumption
- 4) Using a PWM approach for a Fan that has Hall sense commutation circuit may have aliasing acoustic noise in the windings with the Pulse modulated coils, can work but internal stress like cap surge current is unknown.
Nevertheless, I have used LM317 with a transistor and thermistor on a 48V supply with two 24V fans in series in production regulating at 45~55'C air temp on the power supply hotspot (XFMR).
BLDC Fans draw an average current like a resistor with voltage.
Here is my full range linear voltage regulator suitable for a 2.4W Fan.'
It works with any Op Amp as the inputs only drop down to the Vgs(th) of the NFET.

Adding >0.1uF decoupling cap is optional to the tap of pot. (10k~100k)
Note that since the FET inverts, negative feedback MUST use the +ve input.