I usually just compress the section of feedback so it isn't ear-splitting on playback. We've all heard feedback at performances before, and I find I prefer the familiar feedback itself - albeit compressed so it's not earsplitting - to the unexpected and more unnatural artifacts resulting from attempted removal (see / listen below for the latter).
Alternatively, you might try using a spectral view editor and/or notch filters to manually remove the feedback's fundamental frequency (in this case, ~1500 Hz) and harmonic frequencies (integer multiples of the fundamental...e.g. approximately 3000, 4500, 6000, etc.). You could even go crazy and tediously attempt (and probably fail) to manually remove the inharmonic frequencies (non-integer multiples of the fundamental), as well. (Personally, I don't think it's worth the effort.) In either of the above cases, IME you'll still have plenty of artifact leftover. But you may find it less offensive to your ears. For a quick and dirty example, in the attachment I've very swiftly and not very carefully removed the fundamental frequency and harmonic frequencies using Adobe Audition's spectral view. As you'll hear, plenty of artifact remains, but to some it may prove less offensive to the ears.
I suspect in short order someone else will chime in about the wonders of iZotope RX, with which I have no direct experience.