We would generally recommend registering your nodiff (b=0) diffusion image to your structural image and then applying that transformation to your FA image. This is likely to work the best for your within-subject registration problem.
If you want to transfer your hippocampus segmentation, then do the above steps, check that they work visually, and then invert the transformation and apply the inverted transformation to your hippocampal segmentation, to get it into the diffusion space
(and aligned with the FA).
Thank you in advance for your time.
Jim C.