The rcscript addon udev-start.sh calls volume-managers to create its device nodes for already active volumes. This is only needed for the checkroot init-script, as / is already mounted, and after checkroot the normal addons for volume-managers are started. To not get more problems with volume-orders (not being configurable at that place) we should remove that part completely, see Bug #170027. Only remaining problem to be solved then, is: checkroot init-script, which calls "fsck /". fsck then tries to access device-node for root device (getting the info from fstab/its internal db). Uberlord suggested to just disable fs-checking in fstab for cases where /dev/mapper/.... (which is garanteed to exist) cannot be easily used in fstab.
i dont think i got back to you on this after we chatted about it in the base channel ... for baselayout-1.12.x, we'll happily move the volume crap into the checkfs script for 1.13.x and beyond, we'll move to proper init.d scripts
The problem is now reduced to baselayout-1. udev-start.sh no longer executes this code for baselayout-2.
@robbat, Cardoe: I found some code, that basically is/was the Phase I of volume-init we talked about yesterday. This code is from udev-start-104-r12.sh: # Create nodes that udev can't ebegin "Finalizing udev configuration" [ -x /sbin/dmsetup ] && /sbin/dmsetup mknodes &>/dev/null [ -x /sbin/lvm ] && \ /sbin/lvm vgscan -P --mknodes --ignorelockingfailure &>/dev/null # Running evms_activate on a LiveCD causes lots of headaches [ -z "${CDBOOT}" ] && [ -x /sbin/evms_activate ] && \ /sbin/evms_activate -q &>/dev/null eend 0 It got reduced for newer udev versions to (from udev-start-115-r6.sh): # Only do this for baselayout-1* if [ ! -e /lib/librc.so ]; then # Create nodes that udev can't ebegin "Finalizing udev configuration" [ -x /sbin/lvm ] && \ /sbin/lvm vgscan -P --mknodes --ignorelockingfailure &>/dev/null # Running evms_activate on a LiveCD causes lots of headaches [ -z "${CDBOOT}" -a -x /sbin/evms_activate ] && \ /sbin/evms_activate -q &>/dev/null eend 0 fi
zzam: yeah, make that code run always, and we should be good for the PhaseI.
I don't think is of any relevance with current stable, 197-r3 and the separate udev-init-scripts package. Did brief research using qlist, xargs and grep.