On 2019-07-10, Andy Smith <a...@strugglers.net> wrote: > Secondly, the reason I asked you what you would like done is that in > the message I replied to you said that the release notes were > something that users don't read. But your proposed solution is to > put more things in the release notes.
I said users don't read the release notes? I don't remember saying that. I remember saying we can't assume or expect the "regular user" (for any arbitrary definition of that) is following the technical discussions of the development team. I do think, though, all users are responsible for reading the release notes. That's life in the big city, as Mom used to say. >> Further, I would like to know whether the patch will be "baked into the >> kernel" or whether it can be toggled on and/or off at the *user's* >> discretion. I don't remember being clear on this point after reading the >> notes (maybe it's there and I missed it). >> >> It wasn't clear to me, either, in the release-notes, the recommended way >> forward for those with amd64 cpus lacking the RDRAND instruction (and who >> therefore cannot "benefit" from the patch). > > In the release notes in the relevant section (5.1.4) the last > paragraph is: > > See the wiki (https://wiki.debian.org/BoottimeEntropyStarvation) > and DLange's overview of the issue > (https://daniel-lange.com/archives/152-hello-buster.html) for > other options. > Both those pages have a list of different solutions and both mention > that this RDRAND thing is enabled. Neither of them say how to > disable it but they do say: > > for amd64: use a recent kernel with CONFIG_RANDOM_TRUST_CPU set > (less recent kernels may need random.trust_cpu=on added to the > commandline) > > which kind of makes it obvious to me that to disable it you would do > "random.trust_cpu=off". If you don't find this obvious maybe the > wiki page could do with editing to improve that. So for those with a recent kernel, who have a cpu that supports the RDRAND instruction, to unset the 'CONFIG_RANDOM_TRUST_CPU' default, they should put 'random.trust_cpu=off' on the kernel command line? And for those using a less recent kernel with a cpu that supports the RDRAND instruction, it might be enough not to add "random.trust_cpu=on" to the kernel command line? Sorry, but I'm still unclear about this. > So what's lacking? Is it just that the release notes and linked pages > mention that the user trusts the CPU but do not mention why > some feel that is a bad idea? > Maybe you could draft an extra paragraph for the release notes that > contains a suitable warning? Although I do think if you are going to > use the "even the author of the patch thinks it's bad!" argument > then you should probably check with Ted Ts'o that that's an accurate > representation of his views on using RDRAND for boot-time entropy. His views shouldn't be "represented" at all. I thought it would be honest to have something brief, up front, and simple, in the notes themselves, like: For amd64 systems supporting the RDRAND instruction this issue is avoided by the Debian kernel using this instruction by default (CONFIG_RANDOM_TRUST_CPU). There are security implications in regard to applying this default in the view of some. See the comments of Theodore Ts'o (author of the patch) about this config option here: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=39a8883a2b989d1d21bd8dd99f5557f0c5e89694 Sufficiently wishy-washy, I think (*on ne se mouille pas trop*), and those who travel to the link can interpret Ts'o's remarks via their own brains and arrive at their own conclusions inside of same. > As for the recommended way forward, I'm not sure that there is an > easy answer if RDRAND isn't an option. There are complex trade-offs > and I think it's probably right that users in this position read the > wiki page and work out what's best for them. > > I do note that for a person in your situation (real hardware [not a > virtual machine] with no RDRAND and no TPM), every listed solution > has at least one expert that says it is a very bad idea! I don't > think there is consensus here yet. > > In your position I think I'd probably hold my nose (as it says) and > use haveged. What about jiggling my mouse for a while? > Cheers, > Andy > -- "These findings demonstrate that under appropriate conditions the isolated, intact large mammalian brain possesses an underappreciated capacity for restoration of microcirculation and molecular and cellular activity after a prolonged post-mortem interval." From a recent article in *Nature*. Holy shit.
