News for hackers. Stuff that matters.
Re:/dev/random – a new approach by matt_wulfeck on Sunday April 24, 2016 @06:52PM (#11561731)
Can someone help me understand why intel has both the rdrand and rdseed instruction set? If your device outputs high quality random data, why would it need two outlets?
Re:/dev/random – a new approach by JoshTriplett on Sunday April 24, 2016 @07:02PM (#11561757)
https://software.intel.com/en-us/blogs/2012/11/17/the-differ...

rdseed directly gives you the output of a hardware RNG, limited by the rate that hardware RNG produces random bits. rdrand gives you CSPRNG output seeded by rdseed.

Re:/dev/random – a new approach by ryuuchin on Sunday April 24, 2016 @11:49PM (#11562442)
Because the DRNG (Digital RNG) on Intel processors filter the data through a DRNG (Deterministic RNG) (i.e. NIST 800-90)[1] which is constantly reseeded. The rdseed instruction was added later to allow for accessing the hardware generator directly (or the bits from the AES CBC-MAC conditioner directly).

I don't know why rdseed was added a generation after rdrand. Whatever the case is you can guarantee that rdrand's internal generator gets reseeded by calling it a certain number of times[2]. You can also wait some amount of time between calls but it's likely faster just to generate more data to force the reseed.

[1] Like Windows it uses AES-256 (AFAIK) in CTR mode as specified in NIST 800-90 or whatever the latest version of that standard is.

[2] the DRNG reseeds automatically after generating 511 128-bit samples (from Intel's drng_samples source package)