The length of output is source length multipled with "upFactor". I checked ippsFIR_32f and ippsMRFIRDirect_32f, the bugs is same. Is it realy a bug?
Which downFactor do you use? If downFactor=1 output will be the same, but for downFactor=2 output will be two times shorter and so on. Look please ippsSampleDown function documentation in Intel Integrated Performance Primitives, Reference Manual, Volume 1, Signal Processing.