View Single Post
  #5   (View Single Post)  
Old 17th December 2013
thirdm thirdm is offline
Spam Deminer
 
Join Date: May 2009
Posts: 248
Default

Quote:
Originally Posted by jggimi View Post
You were mislead by yet another misleading article, Ninguem. De Raadt was not referring to crypto at all, but to programming practices for kernel services such as memory allocations. His complaint is that FreeBSD has developed but elected not to deploy several such security features, which began use in OpenBSD and have migrated to other platforms, including Linux and (gasp!) Windows. Look to his presentation, referenced in the article, and the last several slides.

Well wait now. True, that presentation mentioned at the end doesn't discuss the random number generator, but the article has quotes from Theo regarding this change: http://arstechnica.com/security/2013...evelopers-say/ and that OpenBSD and Linux didn't use (or provide an option to use, at least in kernel for populating /dev/random?) the hardware generated random numbers directly without combination with other sources of randomness.

If this ars technica quote isn't misleading by being out of context or something, it seems like FreeBSD has more pressure from users (particularly corporate users?) to do weird things:

"The meetings came shortly after a FreeBSD patch a third party submitted in July gave users the ability to use RDRAND and Padlock as direct sources of randomness. The update was committed without proper review and against established policy, and over the next few weeks all or parts of it were removed and recommitted FreeBSD Security Officer Dag-Erling Smørgrav told Ars. The change was finally removed altogether in October. Aside from the brief span earlier this year, no FreeBSD release has ever shipped with the option to feed the bit stream from a hardware pseudo-random number generator directly to /dev/random, Smørgrav said."

"Smørgrav went on to say that some corporate users have long expressed a wish that FreeBSD allow them to run raw feeds from RDRAND, because its underlying algorithm is listed as an approved deterministic RNG under the FIPS 140 standard defense contractors and other government workers are required to follow. Using a few lines of code, FreeBSD users can to use RDRAND directly in userland rather than use the kernel to feed it to them directly."

But one misleading thing maybe. These quotes make it sound like FreeBSD was doing the right thing previously but the July patch got by messing that up, not that only after 10 years have they mixed rdrand with other sources. Am I understanding that rightly?

I've been only sort of following the story. Is maybe the source of Theo's frustration that FreeBSD (maybe not purposely but because of the screwed up technical press) has gotten promoted for increasing security when in fact all they did was correct a slip up they made recently? Yeah, I'd be pissed off too if people came asking me why I hadn't slipped up in my security and then corrected it to public acclaim.

Thanks for pointing out the slides though. I'm really happy to have 5.4 working on my laptop (previously it wouldn't work properly) and to be a full time (well not including work but who counts that, eh) OpenBSD user again.
Reply With Quote