Re: use of add_interrupt_randomness in drivers missing in many drivers

From: tytso@mit.edu
Date: Thu Oct 19 2000 - 15:50:37 EST


   From: daw@mozart.cs.berkeley.edu (David Wagner)
   Date: 18 Oct 2000 20:29:33 GMT

   Adding more bits to the pool should never hurt; the cryptographic
   mixing ensures this. What _can_ hurt is adding predictable bits but
   (erroneously) bumping up the entropy counter.

Yes; and writing to /dev/random only mixes the contents into the pool;
it does *not* bump the entropy counter. Hence, writing to /dev/random
is always safe; it can't hurt, and can help. For this reason, it's safe
to have /dev/random to be world writeable; some folks have been overly
paranoid and making /dev/random be mode 444, or some such.

If you want to add random data to the pool and bump the estimate of the
entropy, you need to be root, and use a special ioctl which does this
atomically. The intent is that a user-mode daemon would read data from
/dev/microphone, post-processes it a lot (filter out 60 Hz hum, compress
it, whatever), get an estimate of the entropy in the sample (which may
not be the same as its size), and then call the ioctl to push that data
into the entropy pool.

                                                - Ted
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Mon Oct 23 2000 - 21:00:16 EST