On Tue, 12.03.13 10:05, Harald Hoyer ([email protected]) wrote: > We have a big problem now with local-fs.target in the > initrd-switch-root.target > now. Because local-fs.target is active now after > serialization/deserialization, > the whole boot process after the switch-root is screwed. All units depending > on > local-fs.target are started now immediately!! > > Solutions I can think of is: > - not to serialize the targets as we do with the jobs now > - set all targets to TARGET_DEAD on switch_root
No, please don't. What's the story behind wanting to use local-fs.target for so many different things? I mean, you now want to reach it three times. Twice from the initrd, and once from the host, and everytime it shall have a different meaning and be gone until the next time again, in some magical way. THis sounds really wrong to me. Instead it appears natural to me to simply have three different targets. i.e. "root-fs.target" + "initrd-fs.target" in the initrd, and then "local-fs.target" is left for the host? That way, after boot you can easily follow what has been achieved when. What's the rationale here? I don't follow? I'd really prefer not having to add special semantics for the switch root transition serialization. To underline this: I really think every unit should be activated once during boot, and not forgotten and then activated again. Lennart -- Lennart Poettering - Red Hat, Inc. _______________________________________________ systemd-devel mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/systemd-devel
