<clever>
kaotika: let me see what i can glean from the drivers.....
<clever>
./drivers/char/hw_random/bcm2835-rng.c
<clever>
kaotika: i see a warp-up count, where it says it will discard the first 0x40000 values, because they are less random, it just gets written to a register at offset 4
<clever>
then it sets an enable flag
<clever>
within the status register, is an 8bit word count, the driver will then read from a 32bit data register, up to that many times
<clever>
so basically, the hw is filling a FIFO with random 32bit values, and reporting how much data is within the FIFO
<clever>
linux will then read data from there, possibly sleeping if the fifo has no data, but it can read less then you requested, as all file IO can
<kaotika>
clever: i am thinking ring oscillators because of the warm-up discard
<kaotika>
it fails many of the diehard tests but the min-entropy if very good.
<kaotika>
ah, got to rest. bbl!
<clever>
there is also an interruptlisted in the device-tree, but the driver never uses it
<clever>
i assume that is to interrupt when the fifo is full and more rng is available
<kaotika>
what is it for do you think?
<kaotika>
oh
<clever>
so you can be fed rng on an event basis, rather then polling for more rng