I've noticed that since updating my server to Debian Squeeze the amount of entropy as reported by /proc/sys/kernel/random/entropy_avail is much lower than it was before the upgrade. I would like to know if this lower pool size is big enough to function with or if I need to look into getting more entropy sources. I think having a way to log blocking reads of /dev/random would show whether I have enough entropy or not.
You can write a library with your own implementation of
read(...) or whatever function you use to read from
/dev/random to wrap the actual call in profiling and logging methods. Then you should use a
LD_PRELOAD env variable to insert your custom implementation of
read before the system one.