>>> Lennart Poettering <lennart@xxxxxxxxxxxxxx> schrieb am 14.12.2022 um 18:34 in Nachricht <Y5oJICQrU0EuThkH@gardel-login>: > On Di, 13.12.22 22:34, Farblos (AKFKQU.9DF7RP@xxxxxxxxxxxxxxx) wrote: > >> I can imagine that the latter scenario is not supported or requires >> additional configuration (which?), but I have not found any hints on that, >> neither in systemd.resource‑control(5) nor in [1.] or [8.] from that man >> page. > > The relevant mechanisms are implemented via eBPF, which the kernel > restricts to privileged processes, which means ‑‑user systemd will > have a hard time. > > There were discussions and some work done to allow signed eBPF > programs which the kernel would then allow even from unpriv userspace, > but this didn't materialize so far. I think it would be great solution > for us. > > Most of our sandboxing settings degrade gracefully if the backing > kernel concept is not available in the kernel, or not accessible due > to privs. We generally value portability of service files more than > the sandboxing settings, currently. BUT: Shouldn't there be an error message for the --user case? > > Lennart > > ‑‑ > Lennart Poettering, Berlin