A [cryptographic] hash function is a one-way operation - that is, it is not a bijective function and is subject to the Pigeonhole principle.
That being said, given a restricted domain, a collision (should one exist) can be discovered by generating a hash for each value in the domain and comparing it with the expected hash. This is known as brute-force attack.
In this case the input domain looks to be limited to "a.b.c" (3 of the 4 octets of an IPv4 address), where a, b, and c are values in the range 0..255: thus there is an upper-bound of 256^3 ~ 2^24 ~ 16 million inputs to check1.
Again, this is not reversing the hash, but rather finding some input - which might not be the original value! - that results in a particular hash value.
1 Modern CPUs - not to mention GPUs and specialized hash hardware - don't even blink at 16 million hash generations. This is why password hashing must use a distinct salt (to avoid rainbow tables) and a slow hash function such as bcrypt or scrypt.