Hacker News Reader

Comment by mehrdadn | Hacker News Reader

Yeah, I don't understand what's so cryptographically special about booting. If the system hadn't been turned off and on, it wouldn't have had a problem, right? Sounds like the real problem is some part of the system can't be bothered to persist its state. Either that, or the problem should also come up during normal usage, in which case there's a security hole somewhere.

View on HN

systemd isn't using the right interface for adding the entropy in the seed file, so the kernel uses it but counts it as zero bits of entropy

edit: but i guess that's on purpose to handle disk imaging

Yeah I was just reading that. Why was that not an obvious bug to be fixed immediately? I haven't read much of the GitHub issues but it seems so weird to recommend trusting the CPU RNG instead of just pushing out a quick fix...

because people are poorly educated and copy their system disks without changing the random seed. then, they generate similar long-term keys, e.g. RSA keys with common factors. digitalocean was well known to have identical SSH host keys for quite some time (one of many reasons I wouldn't trust them with any sort of remotely serious hosting job); I would be shocked if they knew to reset the random seed.

All this is literally just concern for the first ~minute or two of the very first boot after disk cloning? It should be irrelevant for subsequent boots once they reseed on the next boot... hardly seems like such a huge deal.

Except that SSH keys are often generated on first boot, so there's the risk that they're all the same--unless you put some code in your deployment recipes to regenerate them.

Isn't that the user's and the SSH devs' problem? SSH can just wait for new randomness if it sees that as a risk when generating keys. Key generation is already slow anyway. Why is systemd trying so hard to "fix" a problem that might by some chance come up in SSH during the first minute of the the first boot after an improper disk clone? Especially with the currently proposed solutions it seems like such a crazy way to attack somebody else's problem...

> Isn't that the user's and the SSH devs' problem?

Per some of the comments in the Debian bug(s): if you tell users and devs simply "you deal with it", you will get a multitude of ad hoc, half-baked solutions that may nor may not be secure. You will have dozens of people trying to re-invent the wheel.

The whole point of things like /dev/urandom and getentropy() is that the problem is solved once properly, and then you allow everyone to benefit. By having the above solutions not-work, we are basically going back to the days of where they did not exist--in which case what was the point of creating them?

The BSDs seem to not have a problem with this, so I have no idea why the Linux folks can't seem to get their act together.

> Key generation is already slow anyway.

Define "slow". Key generation was slow in the early 1990s when I first started using PGP (nee GPG) back in the day. It is generally not-slow nowadays IMHO--as in, it takes less than ten minutes to find p and q for RSA 1024 (nevermind 2048+).

example: https://factorable.net/weakkeys12.extended.pdf

> because people are poorly educated and copy their system disks without changing the random seed.

To a first approximation:

* SHA256( cat /var/lib/randomseed || ifconfig || date -u ) > /dev/urandom

will get any decent stream-cipher-based PRNG going.

Each machine will have a unique MAC address, so that's 48 bits right off the bat, even if randomseed is identical and every machine is booted at the exact same second.

MACs are very much not random. Especially on a series of machines for a given provider, you can likely predict one. Date of the bootup is also not random if it's got only seconds resolution and you can figure out when a machine comes online by constantly trying to connect. In practice you added much fewer than 48 bits.

I strongly suspect that this information is already included in the Linux kernel RNG, but it is not given much weight in the entropy calculation, on the basis that virtualized systems frequently have duplicated MAC addresses (as is allowed on a non-public system), and sometimes also have low initial timestamp resolution.

If a system will be communicating with other systems, it will need a unique Layer 2 address. This can be leveraged for some entropy (certainly you don't have to credit the full 48 bits, but are you tell saying that not even 1-2 bits of credit cannot be applied?).

If the system is not communicating with other systems... what attack tree are you actually worried about if it is inaccessible?

I would also be curious to know which virtualization environments generate duplicate MAC addresses over (say) dozens of systems?

MAC addresses are not used in a purely routed/virtualized network.

qemu uses the same MAC address by default.

date -u just returns 1-second resolution? It's trivial to go through all the possibilities...

$(date -u | wc --bytes) says there is 29 bytes worth of output, which is 232 bits.

Are you telling me there is not even, say, 8 bits' worth of entropy credit in there? (Especially when hashed with other data?)

Remember: the context of this discussion is VMs (and not embedded systems without a battery-backed RTC).

> Sounds like the real problem is some part of the system can't be bothered to persist its state.

Persistent state is hard. For example, some systems run with read-only filesystem.

I thought about that, but then, if you setup that kind of a system, isn't it your responsibility to worry about how to generate and persist cryptographic state? It's such a niche and advanced use-case by someone who should really know better considering the setup... why should all the other >95% of users pay such a price for it?

Edit: Replaced "host keys" with "cryptographic state".

The issue is not about host keys, which are generated once. It is about randomness that sshd needs for nonces and session keys. It needs fresh randomness during each boot.

Okay, I substituted "cryptographic state" for "host keys". It doesn't change my argument.

Reply on news.ycombinator.com