On Sun, 10 Jul 2011 16:38:43 +0400 Pan Tsu <iny...@gmail.com> wrote: > "Christopher J. Ruwe" <c...@cruwe.de> writes: > > > Nearly a week ago I posted this question to freebsd-fs, but > > probalby my question is a) worded too complicatedly, b) not really a > > filesystem-issue or c) both. > > > > To rephrase: In setups requiring one or more ZFS-dataset to be > > mounted before another service is activated (GELI in my case) and > > the rest of the ZFS-datasets after that service is activated > > (because they require GELI), it seems to be necessary to add a `zfs > > mount -a` to mountcritlocal. Is this considered correct behaviour > > and wouldn't it make sense to add such a line to mountcritlocal in > > the standard setup? > [...] > > Have you tried to set zfs_enable=YES in rc.conf? Based on rcorder(8) > output rc.d/zfs should come just after rc.d/mountcritlocal.
zfs_enable="YES" is set. rcorder gives curious output, so maybe my etcs are wrong? $> rcorder /etc/rc.d/zfs rcorder: file `/etc/rc.d/zfs' is before unknown provision `mountlate' /etc/rc.d/zfs $> rcorder /etc/rc.d/mountcritlocal rcorder: requirement `root' in file `/etc/rc.d/mountcritlocal' has no providers. /etc/rc.d/mountcritlocal However, I fear I have not made my intent clear. My boot-sequence should be as follows (intermittent steps left out): 1) mount zfs root-fs, which is on plain standard zpool A 2) unlock another, GELI-encrypted zpool B 3) mount all other fs (/usr,...), which reside on zpool B What my system does is first to mount the fs on zpool A, then GELI-unlock and then halt because the contents of /usr are not accessible (yet) What I want my system to do is to first mount root, then unlock GELI and then mount all other remaining fs on zpool B. I could either mount all remaining zfs'es in mountcritlocal, which requires another line there, which I have added locally as put in my patch. I cannot shift the order so that GELI-unlock comes first, because my keys for GELI reside on /boot, which resides on zpool A. So, is my setup anything from unfortunate to plain stupid or is mountcritlocal missing a statement catering for such cases as I described? Thank you for your help, cheers, -- Christopher J. Ruwe TZ GMT + 2
signature.asc
Description: PGP signature
