Post by @netbsd@mastodon.sdf.org

@netbsd@mastodon.sdf.org

2026-08-10 07:38 UTC

@marshray@infosec.exchange @jackemled@furry.engineer It does do that. It just doesn't count it as increasing the entropy estimate because there isn't a scientific paper detailing how much entropy a hard disk produces. Writing 256 bits to /dev/random should be enough to shut the estimator up permanently, providing safe shutdowns (with shutdown -p now).

Replies (1)

@marshray@infosec.exchange 2026-08-10 13:26

@netbsd@mastodon.sdf.org @jackemled@furry.engineer We are unlikely to see a scientific paper on that because it’s applied engineering specific to a given scenario. Entropy gathering is a basic job of a kernel because it’s the code in position to observe high resolution interrupt timings. That CPU has multiple independent GHz oscillators, temperature and other sensors that vary according to unique (physical) thermal profiles. There are likely thousands of interrupts during initial boot. Each is a potential opportunity to sample a high resolution timer (RDTSC). How many samples do you need to accumulate 256 bits of entropy during initial boot? Answering this is a basic job of an OS implementer and many others have managed it successfully in the past. Simply pushing a complex security decision onto the user who has even less information is just setting them up to fail.

Open ##5349617