Ouch.
Security requires a threat model. Something can only be "the worst possible vulnerability" if it causes a large amount of damage consistent via the threats which are in the model. The threats that you're thinking of are not usually contained within the models of cutting-edge-dev-systems, which will be protected by firewalls etc. on intranets.
There's really a need in both cases for "an active attacker to intersect with the vulnerable host while it's still there." A broken RNG in such circumstances is much, much weaker than kernel RCE.
This is indeed a serious issue, and I'm disappointed it was in the source tree for four months. But -CURRENT is the development branch, and is not supported by the FreeBSD security team, or supported for use in production.
From https://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/c...:
FreeBSD-CURRENT is made available for three primary interest groups:
Members of the FreeBSD community who are actively working on some part of the source tree.
Members of the FreeBSD community who are active testers. They are willing to spend time solving problems, making topical suggestions on changes and the general direction of FreeBSD, and submitting patches.
Users who wish to keep an eye on things, use the current source for reference purposes, or make the occasional comment or code contribution.
FreeBSD-CURRENT should not be considered a fast-track to getting new features before the next release as pre-release features are not yet fully tested and most likely contain bugs. It is not a quick way of getting bug fixes as any given commit is just as likely to introduce new bugs as to fix existing ones. FreeBSD-CURRENT is not in any way “officially supported”.
I can understand the sentiment that you should be prepared for even horrible bugs like this if you're running -CURRENT, but "if you can deal with crashing, you can deal with this" is completely wrong. Instability is a completely different beast from potentially generating easily crackable crypto keys.
I'm not sure what you're trying to argue for/against unless you are looking at things from a pure number theory/computer science research point of view? What would people do then? Create perfect flawless code that never malfunctions even as they write it? Code that never has bugs does not exist.
Perfect algorithms don't exist either.
-CURRENT is exactly the place for this because RNG bugs exist in reality.
If you read emaste's post above, this bug "potentially generates easily crackable crypto keys" for the safe to a monopoly game that holds monopoly money.
It's a development kernel.
Normally you could use a system like this on a trial basis and keep the results. For example, you might run some image processing on it. You'd want to make sure the output was good, but having done so you could then use those images for real-world tasks without worry.
But replace "image processing" with "generate a private key" and now you're deeply screwed.
Well, now we know. Don't generate crypto keys on pre-release OSes that you're going to use for anything important. That's reasonable. But I don't think many people would have thought that way before.
And of course pre-release OSes aren't the only place the danger lies. Debian had a similar problem that made it to releases and stayed undetected for a couple of years. The point is just that "silently break all your crypto" is a much worse vulnerability than mere crashes, or even corruption.
My point is that -CURRENT may be rather unstable at any given time, and is explicitly not supported. One should not be using it a production deployment where sensitive key material may exist.
In that regard the 'crashability' of the branch is indeed more significant than a random-numbers exploit: the crashability of the branch is a mnemonic for its general permissiveness towards unaudited (or not-sufficiently-audited) code. When you remember that your adversaries can inject code into the OS, that changes your threat modeling.
Of course, as you say, not-crashing back-doors can make it into releases, too. Open source's generic promise that more eyes will make bugs shallow is not a guarantee.
What we know about every system that installed FreeBSD-CURRENT is that the systems administrators at the time fully accepted an operating system:
1. That is not in any way "officially supported". (FreeBSD's words, not mine.)
2. that may for short periods of time "not be buildable."
3. that "is not a quick way of getting bug fixes as any given commit is just as likely to introduce new bugs as to fix existing ones".
4. that is much weaker in guarantees than the FreeBSD-STABLE branch, which expressly disclaims, "one should not blindly track FreeBSD-STABLE. It is particularly important not to update any production servers to FreeBSD-STABLE without thoroughly testing the code in a development or testing environment."
If someone has signed off on these topics, then there is no such thing as "game over". The server isn't important enough for "game over". If it is, then the security vulnerability was not the broken RNG but tracking FreeBSD-CURRENT in the first place.
I say it's a game-ON plan because thank goodness it got caught in -CURRENT now - that's the way the development process is supposed to work.
I myself run -CURRENT 2 ways - one is a sandbox development physical box that has no access to any of my other servers (I use it to try ports to see if they will still work on -CURRENT), and the 2nd way is in a VM on my laptop to fire it up to see if it boots from time to time.
How do you reach this conclusion? If nobody runs -current, then the damage associated with a -current RCE must also be scaled down by a similar factor, no?
On the other hand, a -current test system which later gets upgraded to the next release for production will still have busted keys on it, but would have had the RCE fixed.
I guess if my house is burglered because I left the door unlocked, you'd say I'm blameless as long as my "threat model" assumed that any burglers always come in through the window.
What you're trying to point out is that sometimes the real threat model (i.e. the security concerns list of the stakeholders) is not the same as the explicit modeling they do of threats. And that's actually very common: HTTPS, for example, is based on a model of eavesdropping threats which only applies to a small case of real-world security problems (public WiFi), models a lot of threats which haven't been so important (ISP malfeasance, DNS hijacking) and the resulting certificate system may be worse for some bigger matters of network security (protecting from government eavesdropping, for example) when compared to something like SSH.
But HTTPS is a good security standard for what it does, and we don't claim that HTTPS sucks just because a lot of people have malware on their PCs which can trivially transmit their traffic to untrusted third parties. Rather, it's your responsibility to revise your explicit threat models to accord with your own implicit expectations. Your clients' students are suddenly upset that your client was storing Social Security Numbers on your servers, which could be read by the public via your API? You have hereby discovered that your "explicit" threat model, which left the front door unlocked, was failing to live up to an "implicit" threat model which you had. You now need to either punt the problem off to the client ("we're not storing your sensitive info, our service is not appropriate for that!") or update your explicit model ("we're re-securing our systems and assuming that everything is potentially sensitive, so new permissions systems will guard access").
Running and continuing to run -CURRENT when you know that occasionally the front door's locks can fall off for any or no reason, means that you've accepted this reality.
BTW, links to the referenced revisions:
http://fxr.watson.org/fxr/source/dev/random/randomdev.c?im=3...
For some, it's a gleeful traipse into the 5m deep concrete pour, briefly regretted. I salute those brave souls! Why they do what they do may never be known, even to those who partake, but they do it with conviction! And, it's a damn fine thing, if you ask me!
In short, I think it's a bit worse if you're actually vulnerable, but vulnerable systems will be much more rare.