On Fri, 14 Dec 2018 11:04:40 +0100 Yves-Alexis Perez <cor...@debian.org> wrote: > On Fri, 2018-12-14 at 10:24 +0100, Yves-Alexis Perez wrote: > > Something puzzles me with all those issues: as far as I can tell, on most > > install, systemd-random-seed.service should save a seed at shutdown and > > restore it at startup, and this (I think) should be enough to properly init > > the RNG. > > > > Can you check if the service has been run in your case? > > Hi again, > > actually don't bother, I was pointed to [1] which has explanations. The random > seed load is done by just writing to /dev/urandom which doesn't credit > entropy [2]. Hi,
That service appears to be running normal on the machine with this bug. As you said, it cannot be the cause. > I don't have good solutions right now. With 4.19 and if your CPU has an RNG > you're willing to trust, you'll be able to pass random.trust_cpu=yes to the > kernel command line, which should help seeding the RNG. The CPU on the machine with the bug does have an hardware RNG. I will test this option once I have linux-image-amd64 4.19 installed.