On Mar 16, 2013 10:57 AM, "Zbigniew Jędrzejewski-Szmek" <<a href="mailto:zbyszek@in.waw.pl">zbyszek@in.waw.pl</a>> wrote: > > On Wed, Mar 13, 2013 at 05:06:01PM +0900, Tom Gundersen wrote: > > On Wed, Mar 13, 2013 at 3:41 AM, <<a href="mailto:harald@redhat.com">harald@redhat.com</a>> wrote: > > > From: Harald Hoyer <<a href="mailto:harald@redhat.com">harald@redhat.com</a>> > > > > > > Instead of using local-fs*.target in the initrd, use root-fs.target for > > > sysroot.mount and initrd-fs.target for /sysroot/usr and friends. > > > > > > Using local-fs.target would mean to carry over the activated > > > local-fs.target to the isolated initrd-switch-root.target and thus in > > > the real root. Having local-fs.target already active after > > > deserialization causes ordering problems with the real root services and > > > targets. > > > > > > We better isolate to targets for initrd-switch-root.target, which are > > > only available in the initrd. > > > > Looks good. > > > > This means that we should probably stop reusing units at all in the > > initramfs. In some cases I guess it works (udev/journal), but perhaps > > we should avoid it also there for consistency? > > > > In particular any storage daemons will need initrd-specific versions > > which are ordered against these new targets rather than > > local-fs.target. I don't see a problem with this, just something we > > need to be aware of. > Sorry fot the belated comment, but isn't it possible to exploit the > fact that dependencies on non-existent unints are ignored and use > something like: > > WantedBy=local-fs.target initrd-fs.target > Before=local-fs.target initrd-fs.target > > which should work both in the initrd and on the real system? > > Zbyszek Yes, that would work. As long as we don't mind the same unit being possibly activated twice: once in the real root and once in the init rd. I guess in general it would cause similar problems as what this was meant to solve? T