random and urandom are more or less the same... /dev/random is still a PRNG, using the same algorithms as urandom, it just maintains a record of some desirable degree of estimated entropy and just blocks until it's satisfied.
They also use separate entropy buffers, called the blocking and non-blocking pools, respectively. The latter doesn't directly seed itself from the former like you might expect.
Both use the SHA1 hash function on their buffers, presumably to prevent any practical leakage of raw data from the pool to the outside world, and both then mix this hash back in to their pool before outputting just half of it to the user.
They also use separate entropy buffers, called the blocking and non-blocking pools, respectively. The latter doesn't directly seed itself from the former like you might expect.
Both use the SHA1 hash function on their buffers, presumably to prevent any practical leakage of raw data from the pool to the outside world, and both then mix this hash back in to their pool before outputting just half of it to the user.