Currently, all fib hash tables use the same hashing function. This leads
to a situation where feeding routes through a pipe from one table to
another causes significant number of collisions, as routes are fed in the
order of increasing hash values, but dst tables are sized based on the
number of stored routes.
The patch makes fib hashing function parametric and chooses random
parameter for each table. Also generally improves quality of hashing
functions.
Unfortunately, while this patch fixes the issue with initial collisions,
having different hashing functions leads to 2x slowdown of pipe feeding,
presumably due to worse cache behavior in dst tables. Also, the original
issue significantly affects just the initial part of feed, when the dst
table is small, so even ideal fix would not improve that much.
Therefore, no merge for this patch.
Add a wrapper function in sysdep to get random bytes, and required checks
in configure.ac to select how to do it. The configure script tries, in
order, getrandom(), getentropy() and reading from /dev/urandom.
