Newbie FFT Convolution question

Hello,

I would like to apply an FIR filter (at least 512 points) to a 1D array of
audio data (512 points).  This convolution would take place within a audio
callback function so the input and output data arrays both need to be the
same length.

My problem is (if I understand correctly) convolving the filter kernel and
input data would result in an output signal N + M -1 long, i.e. different
input/output lengths.

Is there an algorithm I could use for avoiding this problem?  Does the
overlap-save or overlap-add avoid this?

Many thanks for your help,

David