Hello!
If you are running fMRI group analysis in a standard template (i.e. MNI), then you need to first insure proper co-registration between your DWI and fMRI native space. I suggest you move the undistorted fMRI image to match the undistorted DWI’s and not the other way around. This way you reduce the risk of messing around the b-matrix (see https://www.ncbi.nlm.nih.gov/pubmed/20886566 for more detail about this issue). Once this is done, make sure that you save the inverse and forward transformation matrices during the segmentation/normalization step. This way you will be able to reverse normalize (invert transform) your clusters of activation back to the native space.
If you are only running single subject fMRI analysis, then I suggest to only insure proper co-registration between the modalities.
Enjoy